[llvm] r303850 - [GVNSink] GVNSink pass
Galina Kistanova via llvm-commits
llvm-commits at lists.llvm.org
Tue Jun 6 19:15:19 PDT 2017
Hello James.
Here is the bug:
+ std::stable_sort(
+ Candidates.begin(), Candidates.end(),
+ [](const SinkingInstructionCandidate &A,
+ const SinkingInstructionCandidate &B) { return A >= B; });
This comparison operator is invalid, because it does not implement strict
weak ordering. With this logic, if 2 candidates A and B have the same cost,
the comparison operator will return true, but must return false.
This fix would fix the
http://lab.llvm.org:8011/builders/llvm-clang-x86_64-expensive-checks-win
bot and https://bugs.llvm.org/show_bug.cgi?id=33210.
Could you take care of this, please?
Thanks
Galina
On Fri, Jun 2, 2017 at 12:18 PM, Galina Kistanova <gkistanova at gmail.com>
wrote:
> The http://lab.llvm.org:8011/builders/llvm-clang-x86_64-
> expensive-checks-win bot builds the debug configuration now.
>
> James, could you fix the GVNSink.cpp, line 749, please?
>
> DEBUG(dbgs() << " -- Sinking candidates:\n"; for (auto &C
> : Candidates) dbgs()
> << " " << C << "\n";);
> It triggers a crash.
> http://lab.llvm.org:8011/builders/llvm-clang-x86_64-
> expensive-checks-win/builds/2891/steps/test-check-all/logs/stdio
>
> Thanks
>
> Galina
>
>
> On Tue, May 30, 2017 at 12:14 PM, Davide Italiano <davide at freebsd.org>
> wrote:
>
>> On Tue, May 30, 2017 at 11:34 AM, Galina Kistanova <gkistanova at gmail.com>
>> wrote:
>> > I'll see if I could find few spare CPU circles on one of the MSVC
>> builder.
>> >
>> > What targets we are after here? Just LLVM or LLVM+Clang?
>> > If this is LLVM only, I might be able to reconfigure one of the existing
>> > builders to build debug instead of the release.
>> >
>>
>> I think the crash was triggered by LLVM, but having LLVM + clang would
>> be better.
>>
>> --
>> Davide
>>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20170606/6654cfa0/attachment.html>
More information about the llvm-commits
mailing list