[compiler-rt] r253483 - [PGO] Runtime support for value profiling.

Sun Nov 22 11:08:35 PST 2015

On Sun, Nov 22, 2015 at 10:31 AM, Daniel Sanders <Daniel.Sanders at imgtec.com>
wrote:

> > The error messages or the warnings are not related to r253681.
>
> Do you mean r253483 - r253681? I was trying to say that allowing the
> buildbot to be red for as long as it was means that there's a ~400 commit
> range to search for the cause of the current failures. Any regressions in
> this range will not have caused notification emails to be sent to the
> authors of the affected commits so they'd be unaware that they had caused
> regressions and therefore wouldn't attempt to fix them.
>

Agreed.

>
> After looking further, it turns out that the buildbot was red prior to
> r253483 and failed with the same failures as it currently does so things
> are back to normal for this bot. Of course, 'normal' should be green but
> that's another issue.
>

Good to hear that!

thanks,

David

>
> ------------------------------
> *From:* Xinliang David Li [xinliangli at gmail.com]
> *Sent:* 22 November 2015 18:14
> *To:* Betul Buyukkurt
> *Cc:* Daniel Sanders; llvm-commits
>
> *Subject:* Re: [compiler-rt] r253483 - [PGO] Runtime support for value
> profiling.
>
> yes, they are sanitizer related runtime failures.
>
> On Sun, Nov 22, 2015 at 9:45 AM, Betul Buyukkurt <betulb at codeaurora.org>
> wrote:
>
>> The error messages or the warnings are not related to r253681. They seem
>> to be purely asan/tsan test failures.
>>
>>
>>
>> -Betul
>>
>>
>>
>> *From:* Daniel Sanders [mailto:Daniel.Sanders at imgtec.com]
>> *Sent:* Sunday, November 22, 2015 8:58 AM
>> *To:* Xinliang David Li
>> *Cc:* Betul Buyukkurt; llvm-commits at lists.llvm.org
>>
>> *Subject:* RE: [compiler-rt] r253483 - [PGO] Runtime support for value
>> profiling.
>>
>>
>>
>> Just to let you know: As of r253681 the failure changed. It gets past the
>> compilation failure now but I can't tell if the current failure is related
>> to your change or something else that happened in the few days the build
>> was red.
>>
>> See
>> http://lab.llvm.org:8011/builders/sanitizer_x86_64-freebsd/builds/6638
>> for the log that tested that commit.
>> ------------------------------
>>
>> *From:* Xinliang David Li [xinliangli at gmail.com]
>> *Sent:* 20 November 2015 19:21
>> *To:* Daniel Sanders
>> *Cc:* Betul Buyukkurt; llvm-commits at lists.llvm.org
>> *Subject:* Re: [compiler-rt] r253483 - [PGO] Runtime support for value
>> profiling.
>>
>> I happen to have a pending clean up patch that should fix the problem.
>> Patch coming soon ..
>>
>>
>>
>> David
>>
>>
>>
>>
>>
>> On Fri, Nov 20, 2015 at 11:16 AM, Xinliang David Li <xinliangli at gmail.com>
>> wrote:
>>
>> Let me take a look.
>>
>>
>>
>> David
>>
>>
>>
>> On Fri, Nov 20, 2015 at 11:05 AM, Daniel Sanders <
>> Daniel.Sanders at imgtec.com> wrote:
>>
>> r253687 has done the trick. Thanks.
>>
>> I think that just leaves sanitizer_x86_64-freebsd which is failing (
>> http://lab.llvm.org:8011/builders/sanitizer_x86_64-freebsd/builds/6637/steps/compile/logs/stdio)
>> with errors of the form:
>>
>> In file included from /usr/home/buildslave/slave_as-bldslv5/sanitizer_x86_64-freebsd/llvm.src/projects/compiler-rt/lib/profile/InstrProfilingBuffer.c:10:
>>
>> /usr/home/buildslave/slave_as-bldslv5/sanitizer_x86_64-freebsd/llvm.src/projects/compiler-rt/lib/profile/InstrProfiling.h:58:9: error: unknown type name 'uint8_t'
>>
>>   const uint8_t *FunctionPointer;
>>
>> In file included from /usr/home/buildslave/slave_as-bldslv5/sanitizer_x86_64-freebsd/llvm.src/projects/compiler-rt/lib/profile/InstrProfilingFile.c:10:
>>
>> /usr/home/buildslave/slave_as-bldslv5/sanitizer_x86_64-freebsd/llvm.src/projects/compiler-rt/lib/profile/InstrProfiling.h:58:9: error: unknown type name 'uint8_t'
>>
>>         ^
>>
>>   const uint8_t *FunctionPointer;
>>
>>          ^
>>
>> I'm not sure who can help you with this one.
>>
>> ------------------------------
>>
>> *From:* Daniel Sanders
>> *Sent:* 20 November 2015 18:47
>> *To:* Xinliang David Li; Betul Buyukkurt; llvm-commits at lists.llvm.org
>>
>>
>> *Subject:* RE: [compiler-rt] r253483 - [PGO] Runtime support for value
>> profiling.
>>
>>
>>
>> Sure, I see you've committed it and I'd test it on the same machine as
>> that builder so I'll keep an eye on that build for now.
>> ------------------------------
>>
>> *From:* Xinliang David Li [xinliangli at gmail.com]
>> *Sent:* 20 November 2015 17:42
>> *To:* Betul Buyukkurt; Daniel Sanders
>> *Subject:* Re: [compiler-rt] r253483 - [PGO] Runtime support for value
>> profiling.
>>
>> Daniel, can you help test the following small patch?
>>
>>
>>
>> thanks,
>>
>>
>>
>> David
>>
>>
>>
>> On Fri, Nov 20, 2015 at 9:36 AM, Xinliang David Li <xinliangli at gmail.com>
>> wrote:
>>
>> Or disable value profiling for those targets -- such as stub out the
>> __llvm_profile_instrument_target..
>>
>>
>>
>> I can help with a patch for Daniel to test.
>>
>>
>>
>>
>>
>> David
>>
>>
>>
>> On Fri, Nov 20, 2015 at 9:20 AM, Betul Buyukkurt <betulb at codeaurora.org>
>> wrote:
>>
>> Hi Daniel,
>>
>> This was the merging of a runtime support of the value profiling feature
>> which was merged along w/ the LLVM and clang counterparts of this work.
>> Those were r253484 and r253485 respectively.
>>
>> It looks like that the references to  __sycn_fetch_and_add operations are
>> not defined for your target. Is there an alternative that you'd like us to
>> use (mutex lock/unlock  based operation)?
>>
>> -Betul
>>
>>
>> -----Original Message-----
>> From: Daniel Sanders [mailto:Daniel.Sanders at imgtec.com]
>> Sent: Friday, November 20, 2015 8:29 AM
>> To: Daniel Sanders; Renato Golin; Betul Buyukkurt; Xinliang David Li;
>> llvm-commits at lists.llvm.org
>> Subject: RE: [compiler-rt] r253483 - [PGO] Runtime support for value
>> profiling.
>>
>> Not sure why I re-added (Outlook removes the list) llvm-dev instead of
>> llvm-commits on that last one, sorry.
>>
>> I missed a few fixup commits in the last list. Reverting r253545,
>> r253510, r253508, r253503, r253500, r253496, r253492, r253489, and r253483
>> (all of which are part of this commit and the corrections) fixes the
>> failing tests in my working copy. I'm going to do a full 'ninja check-all'
>> before I revert.
>>
>> > -----Original Message-----
>> > From: llvm-dev [mailto:llvm-dev-bounces at lists.llvm.org] On Behalf Of
>> > Daniel Sanders via llvm-dev
>> > Sent: 20 November 2015 16:05
>> > To: Renato Golin; Betul Buyukkurt; llvm-dev at lists.llvm.org; Xinliang
>> > David Li
>> > Subject: Re: [llvm-dev] [compiler-rt] r253483 - [PGO] Runtime support
>> > for value profiling.
>> >
>> > Hi,
>> >
>> > I'm currently bisecting a buildbot regression in the clang-cmake-mips
>> > builder and I've ended up with a range of commits, all of them are
>> > either this commit or one of its corrections. The commits are r253483,
>> > r253489, r253492, and r253496. I'm going to try reverting these in my
>> > working copy and if that works I'll commit the revert.
>> >
>> > The first build to fail is
>> > http://lab.llvm.org:8011/builders/clang-cmake-
>> > mips/builds/10816 and the first one to show the current failures is
>> > http://lab.llvm.org:8011/builders/clang-cmake-mips/builds/10819.
>> >
>> > > -----Original Message-----
>> > > From: llvm-commits [mailto:llvm-commits-bounces at lists.llvm.org] On
>> > Behalf
>> > > Of Renato Golin via llvm-commits
>> > > Sent: 18 November 2015 21:46
>> > > To: Betul Buyukkurt
>> > > Cc: LLVM Commits
>> > > Subject: Re: [compiler-rt] r253483 - [PGO] Runtime support for value
>> > > profiling.
>> > >
>> > > On 18 November 2015 at 18:12, Betul Buyukkurt via llvm-commits
>> > > <llvm-commits at lists.llvm.org> wrote:
>> > > > Author: betulb
>> > > > Date: Wed Nov 18 12:12:35 2015
>> > > > New Revision: 253483
>> > > >
>> > > > URL: http://llvm.org/viewvc/llvm-project?rev=253483&view=rev
>> > > > Log:
>> > > > [PGO] Runtime support for value profiling.
>> > > >
>> > > > This change adds extends the data structures and adds in the
>> > > > routines for handling runtime calls for value profiling. The
>> > > > profile data format is modified and the version number is
>> incremented.
>> > >
>> > > Hi Betul,
>> > >
>> > > There has been a good number of commits and the self-hosting bots
>> > > are still broken, shouldn't we revert everything and try again,
>> > > *after* you make it self-host locally?
>> > >
>> > > http://lab.llvm.org:8011/builders/clang-cmake-armv7-a15-full/builds/
>> > > 9175
>> > >
>> > > You're making it very hard to revert this patch already...
>> > >
>> > > cheers,
>> > > --renato
>> > > _______________________________________________
>> > > llvm-commits mailing list
>> > > llvm-commits at lists.llvm.org
>> > > http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-commits
>> > _______________________________________________
>> > LLVM Developers mailing list
>> > llvm-dev at lists.llvm.org
>> > http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev
>>
>>
>>
>>
>>
>>
>>
>>
>>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20151122/92d59eb9/attachment.html>