SPEC's CPU2000 fma3d had internal compiler error with Aarch64 2014.05 toolchain

[frank smith]

Using the linaro build 2014.05 of the Aarch64 toolchain the floating point benchmark 191.fma3d failed to compile.

/opt/gcc-linaro-aarch64-linux-gnu-4.9-2014.05_linux/bin/aarch64-linux-gnu-gfortran -c -o strain.o -static -g -O3 -march=armv8-a+crypto -ffast-math -ftree-vectorize -fno-use-linker-plugin -floop-interchange -static -g strain.f90
material_00.f90:348.17:

     & (TRUSS(N)%RES%Stress,MatID,SecID,Isv,Nsv)
                 1
Warning: Rank mismatch in argument 'stress' at (1) (rank-1 and scalar)
output.f90:54.53:

      CALL PRINT_COUNTERS (IO_UNIT%LSDO,'Number of ',NUMIF,PAGE_NUMBER)
                                                     1
Warning: Rank mismatch in argument 'icount' at (1) (rank-1 and scalar)
output.f90:55.53:

      CALL PRINT_COUNTERS (IO_UNIT%LSDO,'Mesh with ',MSHIF,PAGE_NUMBER)
                                                     1
Warning: Rank mismatch in argument 'icount' at (1) (rank-1 and scalar)
output.f90:56.53:

      CALL PRINT_COUNTERS (IO_UNIT%LSDO,'Restart w/',NRSIF,PAGE_NUMBER)
                                                     1
Warning: Rank mismatch in argument 'icount' at (1) (rank-1 and scalar)
output.f90:2302.36:

     & (IO_UNIT%LELO,'Number of ', NUMIF, PAGE_NUMBER)
                                    1
Warning: Rank mismatch in argument 'icount' at (1) (rank-1 and scalar)
output.f90:2304.36:

     & (IO_UNIT%LELO,'Mesh with ', MSHIF, PAGE_NUMBER)
                                    1
Warning: Rank mismatch in argument 'icount' at (1) (rank-1 and scalar)
output.f90:2306.36:

     & (IO_UNIT%LELO,'Restart w/', NRSIF, PAGE_NUMBER)
                                    1
Warning: Rank mismatch in argument 'icount' at (1) (rank-1 and scalar)
strain.f90: In function 'gauge2d_strain_integration':
strain.f90:900:0: internal compiler error: in simplify_const_unary_operation, at simplify-rtx.c:1539
       END
 ^
no stack trace because unwind library not available
Please submit a full bug report,
with preprocessed source if appropriate.
See <https://bugs.launchpad.net/gcc-linaro> for instructions.
specmake: *** [strain.o] Error 1
specmake options 2> options.err | tee options.out
COMP: /opt/gcc-linaro-aarch64-linux-gnu-4.9-2014.05_linux/bin/aarch64-linux-gnu-gfortran -c -o options.o -static -g -O3 -march=armv8-a+crypto -ffast-math -ftree-vectorize -fno-use-linker-plugin -floop-interchange -static -g
LINK: /opt/gcc-linaro-aarch64-linux-gnu-4.9-2014.05_linux/bin/aarch64-linux-gnu-gfortran -O3 -march=armv8-a+crypto -ffast-math -ftree-vectorize -fno-use-linker-plugin -floop-interchange -static -g -static -g -o options
  Some files did not appear to be built: fma3d
*** Error building 191.fma3d

[frank smith (smithfra)]

REMOVED -floop-interchange to see if this had any effect. No.

Warning: Rank mismatch in argument 'icount' at (1) (rank-1 and scalar)
strain.f90: In function 'gauge2d_strain_integration':
strain.f90:900:0: internal compiler error: in simplify_const_unary_operation, at simplify-rtx.c:1539
       END
 ^
no stack trace because unwind library not available
Please submit a full bug report,
with preprocessed source if appropriate.
See <https://bugs.launchpad.net/gcc-linaro> for instructions.
specmake: *** [strain.o] Error 1
specmake options 2> options.err | tee options.out
COMP: /opt/gcc-linaro-aarch64-linux-gnu-4.9-2014.05_linux/bin/aarch64-linux-gnu-gfortran -c -o options.o -static -g -O3 -march=armv8-a+crypto -ffast-math -ftree-vectorize -fno-use-linker-plugin -static -g
LINK: /opt/gcc-linaro-aarch64-linux-gnu-4.9-2014.05_linux/bin/aarch64-linux-gnu-gfortran -O3 -march=armv8-a+crypto -ffast-math -ftree-vectorize -fno-use-linker-plugin -static -g -static -g -o options
  Some files did not appear to be built: fma3d
*** Error building 191.fma3d

[frank smith (smithfra)]

Forgot to note that this is a regression with respect to the 2104.04 release of the AArch64 toolchain.

[Kugan Vivekanandarajah (kugan-vivekanandarajah)]

Is this reproducible with 4.9-2014.05 also? what version of 2104.04 did you use? Can you please provide the gcc -v and the exact compiler option that can be used to reproduce this.

[frank smith (smithfra)]

The bug is reproducible, I eliminated one of the flags (see below) and got the same error.

This bug is against the 4.9-2014.05 aarch64 toolchain. This did NOT happen in the 4.8-21014.04 aarch64 toolchain.

fma3d code is Fortran.

FC = /opt/gcc-linaro-aarch64-linux-gnu-4.9-2014.05_linux/bin/aarch64-linux-gnu-gfortran

smithfra:~/Desktop/SPEC2000/cpu2000$ /opt/gcc-linaro-aarch64-linux-gnu-4.9-2014.05_linux/bin/aarch64-linux-gnu-gfortran --version
GNU Fortran (crosstool-NG linaro-1.13.1-4.9-2014.05 - Linaro GCC 4.9-2014.05) 4.9.1 20140505 (prerelease)
Copyright (C) 2014 Free Software Foundation, Inc.

Following is extracted from the SPEC config file that I used to build the cpu2000 floating-point binaries
191.fma3d=default=default=default:
FOPTIMIZE = -O3 -march=armv8-a+crypto -ffast-math -ftree-vectorize -fno-use-linker-plugin -floop-interchange -static -g

and had the same problem after removing -floop-interchange
FOPTIMIZE = -O3 -march=armv8-a+crypto -ffast-math -ftree-vectorize -fno-use-linker-plugin -static -g

[Maxim Kuvyrkov (maxim-kuvyrkov)]

The same-looking bug reported against AOSP: https://bugs.linaro.org/show_bug.cgi?id=373

Since bugzilla is now the default place for bugs in Linaro toolchains, bugs.linaro.org shall be the place to track this problem.

Thank you.

[Kugan Vivekanandarajah (kugan-vivekanandarajah)]

I cant reproduce this with Linaro GCC 2014.09. with:

CC = /home/tcwg/kugan/cbuild/builds/destdir/aarch64-unknown-linux-gnu/bin/aarch64-none-linux-gnu-gcc -Wl,--dynamic-linker=/home/tcwg/kugan/cbuild/builds/sysroot-aarch64-none-linux-gnu/lib/ld-linux-aarch64.so.1 -Wl,-rpath=/home/tcwg/kugan/cbuild/builds/sysroot-aarch64-none-linux-gnu/libc/lib/aarch64-linux-gnu:/home/tcwg/kugan/cbuild/builds/sysroot-aarch64-none-linux-gnu/lib

COPTIMIZE = -O3 -march=armv8-a+crypto -ffast-math -ftree-vectorize -fno-use-linker-plugin -flto -fno-common -mcpu=cortex-a57

Please let me know with the exact command line you used in the spec config to reproduce.

[Kugan Vivekanandarajah (kugan-vivekanandarajah)]

Please also note that we are now using https://bugs.linaro.org/ for tracking bugs. If you still see this regression we should take it there.