[polly] r230624 - Allow non-affine control flow by default

[Johannes Doerfert]

I found the problem,.. dominance + SCEVExpander + RegionMap can make a
bad combination. I will look into it and fix it at the latest tomorrow.

On 02/26, Johannes Doerfert wrote:
> On 02/26, Tobias Grosser wrote:
> > On 26.02.2015 16:12, Johannes Doerfert wrote:
> > >On 02/26, Tobias Grosser wrote:
> > >>On 26.02.2015 12:09, Johannes Doerfert wrote:
> > >>>Author: jdoerfert
> > >>>Date: Thu Feb 26 05:09:24 2015
> > >>>New Revision: 230624
> > >>>
> > >>>URL: http://llvm.org/viewvc/llvm-project?rev=230624&view=rev
> > >>>Log:
> > >>>Allow non-affine control flow by default
> > >>>
> > >>>   With the patches r230325, r230329 and r230340 we can handle non-affine
> > >>>   control flow in (loop-free) subregions. As all LLVM test-suite tests pass and
> > >>>   we get ~20% more non-trivial SCoPs, we activate it now by default.
> > >>
> > >>Our buildbots report three compile-time failures:
> > >>
> > >>http://lab.llvm.org:8011/builders/perf-x86_64-penryn-O3-polly-fast/builds/3987
> > >>
> > >>Did you verify the correctness on LNT with a non-asserts build?
> > >My HEAD is at:
> > >   commit 63329c0fb9dc14633887b121b20bc7a17dc492b7
> > >   Author: Johannes Doerfert <doerfert at cs.uni-saarland.de>
> > >   Date:   Thu Feb 26 11:09:24 2015 +0000
> > >
> > >       Allow non-affine control flow by default
> > >
> > >I build a Debug and a Release version with cmake/ninja and run lnt on
> > >the 3 failing tests:
> > >   FAIL: MultiSource/Benchmarks/mediabench/jpeg/jpeg-6a/cjpeg.compile_time
> > >   FAIL: MultiSource/Benchmarks/MiBench/consumer-jpeg/consumer-jpeg.compile_time
> > >   FAIL: MultiSource/Benchmarks/MiBench/consumer-lame/consumer-lame.compile_time
> > >
> > >However, I pass them with both, the Release as well as the Debug build.
> > >
> > >My lnt command line:
> > >lnt  runtest --verbose  nt --sandbox nt  --cc `which clang` --cxx `which
> > >clang++` --without-llvm --test-suite ~/repos/llvm-test-suite/
> > >--no-machdep-info pollyperf-O3-polly --multisample=1 --cflag=-O3
> > >--mllvm=-polly  -j 6  --only-test=...
> > >
> > >Do you have any idea why I do not get the same failures as the bot?
> > 
> > Not really, do you use the latest LLVM/clang versions?
> I just pulled LLVM/clang and rerun the test with the up-to-date Release
> build, ... no change, I still PASS them ...
> 
> Could you get me the output of those tests on the buildbot by any
> chance?
> 
> -- 
> 
> Johannes Doerfert
> Researcher / PhD Student
> 
> Compiler Design Lab (Prof. Hack)
> Saarland University, Computer Science
> Building E1.3, Room 4.31
> 
> Tel. +49 (0)681 302-57521 : doerfert at cs.uni-saarland.de
> Fax. +49 (0)681 302-3065  : http://www.cdl.uni-saarland.de/people/doerfert



> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits


-- 

Johannes Doerfert
Researcher / PhD Student

Compiler Design Lab (Prof. Hack)
Saarland University, Computer Science
Building E1.3, Room 4.31

Tel. +49 (0)681 302-57521 : doerfert at cs.uni-saarland.de
Fax. +49 (0)681 302-3065  : http://www.cdl.uni-saarland.de/people/doerfert
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 213 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20150226/e67f4aa3/attachment.sig>