[llvm] r270478 - [LoopUnroll] Enable advanced unrolling analysis by default.
Mikhail Zolotukhin via llvm-commits
llvm-commits at lists.llvm.org
Mon May 23 17:58:57 PDT 2016
> On May 23, 2016, at 4:52 PM, Mikhail Zolotukhin via llvm-commits <llvm-commits at lists.llvm.org> wrote:
>
>
>> On May 23, 2016, at 4:50 PM, Hans Wennborg <hans at chromium.org <mailto:hans at chromium.org>> wrote:
>>
>> On Mon, May 23, 2016 at 12:10 PM, Michael Zolotukhin via llvm-commits
>> <llvm-commits at lists.llvm.org <mailto:llvm-commits at lists.llvm.org>> wrote:
>>> Author: mzolotukhin
>>> Date: Mon May 23 14:10:19 2016
>>> New Revision: 270478
>>>
>>> URL: http://llvm.org/viewvc/llvm-project?rev=270478&view=rev <http://llvm.org/viewvc/llvm-project?rev=270478&view=rev>
>>> Log:
>>> [LoopUnroll] Enable advanced unrolling analysis by default.
>>>
>>> Summary:
>>> This patch turns on LoopUnrollAnalyzer by default. To mitigate compile
>>> time regressions, I chose very conservative thresholds for now. Later we
>>> can make them more aggressive, but it might require being smarter in
>>> which loops we're optimizing. E.g. currently the biggest issue is that
>>> with more agressive thresholds we unroll many cold loops, which
>>> increases compile time for no performance benefit (performance of those
>>> loops is improved, but it doesn't matter since they are cold).
>> [..]
>>> Differential Revision: http://reviews.llvm.org/D20482 <http://reviews.llvm.org/D20482>
>>
>> I've reverted this in r270512, as it caused PR27847.
It should be fixed in r270517. I'm going to reapply r270478 - please let me know if you see any failures after it.
Thanks,
Michael
> Thanks, I'm looking into it!
>
> Michael
>
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at lists.llvm.org
> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-commits
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20160523/0db4710a/attachment.html>
More information about the llvm-commits
mailing list