[Ubuntu 16.10] Need to have an optimised atlas version

Bug #1592066 reported by bugproxy
6
This bug affects 1 person
Affects Status Importance Assigned to Milestone
The Ubuntu-power-systems project
Won't Fix
Medium
Unassigned
atlas (Ubuntu)
Invalid
Medium
Ubuntu on IBM Power Systems Bug Triage

Bug Description

Current ATLAS version is not optimised for ppc64el, disabling VSX use.
New version 3.11.38 under development is providing this feature.

This is a feature for Ubuntu 16.10 to have a Atlas optimized in 16.10/ppc64el

bugproxy (bugproxy)
tags: added: architecture-ppc64le bugnameltc-133878 severity-medium targetmilestone-inin1610
Changed in ubuntu:
assignee: nobody → Taco Screen team (taco-screen-team)
affects: ubuntu → atlas (Ubuntu)
Revision history for this message
bugproxy (bugproxy) wrote : Comment bridged from LTC Bugzilla

------- Comment From <email address hidden> 2016-07-27 07:04 EDT-------
Beta version 3.10.3BETA is currently being tested - On Ubuntu all tests are currently successful. Waiting for release to be out.

Revision history for this message
bugproxy (bugproxy) wrote :
Download full text (3.7 KiB)

------- Comment From <email address hidden> 2016-07-29 03:05 EDT-------
I am happy to announce the release of ATLAS 3.10.3. Please forward this
e-mail to any relevant mailing list, or to any parties that are
repackaging the stable release that you know about.

ATLAS 3.10.3 should be noticeably faster than 3.10.2 on modern hardware,
but the 3.11 series is almost always much faster on such systems. While
I was able to backport support for modern architectures, and even
provide some reasonable kernels for modern ISA extensions, the 3.11
series allows for much larger block factors and improved storage formats
that are required to get decent performance many modern machines
(including all AVX-enabled Intel chips). So, if you can use it, 3.11 is
still the best for modern machines by a long way.

I had hoped to have ATLAS 4.0 out by now, but various setbacks have
delayed the release, necessitating 3.10.3, since 3.10.2 was not
installing well on modern machines.

3.10.3 fixes these three bugs:
http://math-atlas.sourceforge.net/errata3.10.2.html#herkNaN
http://math-atlas.sourceforge.net/errata3.10.2.html#syr2kNaN
http://math-atlas.sourceforge.net/errata3.10.2.html#rotmg

I have tested 3.10.3 to work on the following OSes:
1. Linux
2. Windows64 (cygwin64 builds now work!)
3. AIX
4. OS X

For OSes 2-4, see special sections in the install guide for additional help:
http://math-atlas.sourceforge.net/atlas_install/node53.html
Hopefully other OSes (eg., Windows32, Solaris) still work from 3.10.2
testing.

Also note that clang can now be used to build ATLAS by adding:
--force-clang=/path/to/clang
to your configure line. For the open version of clang, performance
still tends to lag gcc, but is strongly improved from last release.
Apple's clang appears to be substantially faster, but I may be mistaken.

New architecture support available in 3.10.3 includes:
1. ARM32: a7, a9, a15 (auto-detect of SOFT/HARD ABI)
2. ARM64: xgene1, a53, a57
3. Intel: Corei3 & Corei4 (skylake)
4. IBM: Z series, POWER8 (including little/big endian)

Support for modern vector extensions in atlas_simd.h:
1. Intel AVX2
2. IBM VSX & Z-series VX
3. ARM64 Advanced SIMD
4. ARM32 NEON (only if -Si ieee 0 flag is thrown)

Regards,
Clint

ATLAS 3.10.3 released 07/28/16, highlights of changes from 3.10.2
* Updated F77 L1BLAS testers to those used LAPACK3.6.1
* Fixed bug in rotmg revealed by LAPACK3.6.1 testers
* Fixed bug in hprk/sprk that could cause NaN propogation in
HERK/SYRK due
to reading uninitialized memory in BETA=0 case
* Fixed bug in threaded SYR2K/HER2K that could cause NaN propogation due
to reading uninitialized memory
* Extended matrix/vector norm functions to detect NaNs
* Extended configure:
+ --force-clang=/path/to/clang : will use clang for all C compilers,
even goodgcc (assumes gcc flag & inline-assembly compatibility)
+ --cripple-atlas-performance: install despite failing throttle check
+ Can now use arch string rather than enum # for -A arg
+ --force-tids now affects ATLrun.sh as well as threaded build
+ ARM32 autodetects SOFTFP/HARDFP ABI
* backport of config & archdefs for:
+ POWER[7,8]le, IBMz[10,13,19], Corei[3,4], ARM[7,9,15,17],
ARM64[xgene,a53,a57]
+ archdefs ...

Read more...

Revision history for this message
bugproxy (bugproxy) wrote :

------- Comment From <email address hidden> 2016-08-01 07:18 EDT-------
Bug#833139 on Debian has been opened for same request of upgrade

Revision history for this message
bugproxy (bugproxy) wrote :

------- Comment From <email address hidden> 2016-08-26 03:52 EDT-------
Can we ensure compilation is using VSX

Revision history for this message
bugproxy (bugproxy) wrote :

------- Comment From <email address hidden> 2016-08-30 10:19 EDT-------
$ apt-cache madison atlas
atlas | 3.10.3-1ubuntu1 | http://fr.ports.ubuntu.com/ubuntu-ports yakkety/universe Sources

but from source
$ cat debian/patches/series
02_rename_lapack_atlas.diff
static_full_blas_lapack.diff
shared_libraries.diff
10_s390.diff
16_warning-removed.diff
17_hppa.diff
18_alpha.diff
20_armel.diff
21_mips2.diff
22_sh.diff
kfreebsd.diff
armel-is-v4t.diff
generic.diff
ppc64el-abiv2.patch
ppc64el-ifdef-files-with-lvx.patch
powerpc-dcbt.patch

so ppc64el-ifdef-files-with-lvx.patch is still applied and compilation doesn't use VSX for ppc64le.

Would you check the previously added patch

Revision history for this message
bugproxy (bugproxy) wrote : Patch to reuse VSX instruction on ppc64le

------- Comment (attachment only) From <email address hidden> 2016-08-30 10:18 EDT-------

Changed in atlas (Ubuntu):
assignee: Taco Screen team (taco-screen-team) → nobody
Steve Langasek (vorlon)
no longer affects: atlas (Debian)
Revision history for this message
Steve Langasek (vorlon) wrote :

The VSX patch is reasonably small and self-contained, and we could consider including that as a distro patch in 17.04. If we do so, does that resolve this bug, or are there further optimizations present in 3.11.38 that you believe are needed still?

Changed in atlas (Ubuntu):
assignee: nobody → Steve Langasek (vorlon)
status: New → Incomplete
Revision history for this message
Steve Langasek (vorlon) wrote :

This bug is blocked waiting for more information from IBM; reassigning back to triage team.

Changed in atlas (Ubuntu):
assignee: Steve Langasek (vorlon) → Ubuntu on IBM Power Systems Bug Triage (ubuntu-power-triage)
Frank Heimes (fheimes)
Changed in ubuntu-power-systems:
status: New → Incomplete
Manoj Iyer (manjo)
Changed in atlas (Ubuntu):
importance: Undecided → Medium
Changed in ubuntu-power-systems:
importance: Undecided → Medium
Manoj Iyer (manjo)
tags: added: triage-g
Revision history for this message
bugproxy (bugproxy) wrote : Comment bridged from LTC Bugzilla

------- Comment From <email address hidden> 2017-08-08 11:48 EDT-------
Canonical,

please close this bug, let's follow upstream for the upcoming release.

Revision history for this message
Andrew Cloke (andrew-cloke) wrote :

Ack. Thanks. Closing.

Changed in ubuntu-power-systems:
status: Incomplete → Won't Fix
Changed in atlas (Ubuntu):
status: Incomplete → Invalid
To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.