[llvm] r325448 - [ThinLTO] Add GraphTraits for FunctionSummaries

[Charles Saternos via llvm-commits]

Looked into it a bit more. Turns out LLVM_DUMP_METHOD works fine, it was
just segfaulting on the function_ref in Release mode. Refactored it as a
static function, and it works, so I'll submit the patch again and look into
reproducing the function_ref bug to submit a bug report.

Thanks,
Charles

On Sun, Feb 18, 2018 at 4:16 PM, Teresa Johnson <tejohnson at google.com>
wrote:

>
>
> On Sun, Feb 18, 2018 at 10:49 AM, Charles Saternos <
> charles.saternos at gmail.com> wrote:
>
>> Yup, it fails when I do a release build. Does that mean I should remove
>> LLVM_DUMP_METHOD or use a different qualifier?
>>
>
> Probably should remove LLVM_DUMP_METHOD - it looks like that is used
> typically for dump routines that are guarded by #ifndef DEBUG. However, it
> would be good to first understand why this isn't working in release mode -
> from my visual inspection LLVM_DUMP_METHOD should not be optimized out
> automatically under NDEBUG.
>
> Teresa
>
>
>>
>> - Charles
>>
>> On Sun, Feb 18, 2018 at 12:58 PM, Teresa Johnson <tejohnson at google.com>
>> wrote:
>>
>>>
>>>
>>> On Sat, Feb 17, 2018 at 4:43 PM, Charles Saternos via llvm-commits <
>>> llvm-commits at lists.llvm.org> wrote:
>>>
>>>> OK, I've been looking into the issue. Not sure why its happening since
>>>> the test passes when I build locally. I'll keep investigating.
>>>>
>>>
>>> Looked at a couple of the failures - looks like empty output. Initially
>>> thought you needed to add "REQUIRES: asserts", but neither dump method nor
>>> caller is qualified by an #ifndef NDEBUG. LLVM_DUMP_METHOD is defined
>>> somewhat differently for the NDEBUG case, but only to prevent it from being
>>> eliminated if there are no callers, and you have a caller. But as a sanity
>>> check, try reproducing locally with a compiler configured to build RELEASE.
>>>
>>> Teresa
>>>
>>>
>>>> Thanks,
>>>> Charles
>>>>
>>>> On Sat, Feb 17, 2018 at 7:35 PM, Simon Pilgrim <llvm-dev at redking.me.uk>
>>>> wrote:
>>>>
>>>>> Charles, I'm sorry but I had to revert this at rL325454 to try and fix
>>>>> the buildbot failures. Simon.
>>>>> On 17/02/2018 22:24, Charles Saternos via llvm-commits wrote:
>>>>>
>>>>> Hey Davide,
>>>>>
>>>>> It's been reported here: https://bugs.llvm.org/show_bug.cgi?id=32556
>>>>>
>>>>> The use of std::function rather than function_ref is related to some
>>>>> problem in clang.
>>>>>
>>>>> Thanks,
>>>>> Charles
>>>>>
>>>>> On Sat, Feb 17, 2018 at 5:04 PM, Davide Italiano <davide at freebsd.org>
>>>>> wrote:
>>>>>
>>>>>> On Sat, Feb 17, 2018 at 1:39 PM, Charles Saternos via llvm-commits
>>>>>> <llvm-commits at lists.llvm.org> wrote:
>>>>>> > Author: ncharlie
>>>>>> > Date: Sat Feb 17 13:39:24 2018
>>>>>> > New Revision: 325448
>>>>>> >
>>>>>> > URL: http://llvm.org/viewvc/llvm-project?rev=325448&view=rev
>>>>>> > Log:
>>>>>> > [ThinLTO] Add GraphTraits for FunctionSummaries
>>>>>> >
>>>>>> > Add GraphTraits definitions to the FunctionSummary and
>>>>>> ModuleSummaryIndex classes. These GraphTraits will be used to construct
>>>>>> find SCC's in ThinLTO analysis passes.
>>>>>> >
>>>>>> > Second attempt, since last patch caused stage2 build to fail (now
>>>>>> using function_ref rather than std::function).
>>>>>> >
>>>>>>
>>>>>> Charlie, what's the bug in clang? It would be good to fix it,
>>>>>> regardless.
>>>>>>
>>>>>> Thanks,
>>>>>>
>>>>>> --
>>>>>> Davide
>>>>>>
>>>>>
>>>>>
>>>>>
>>>>> _______________________________________________
>>>>> llvm-commits mailing listllvm-commits at lists.llvm.orghttp://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-commits
>>>>>
>>>>>
>>>>>
>>>>
>>>> _______________________________________________
>>>> llvm-commits mailing list
>>>> llvm-commits at lists.llvm.org
>>>> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-commits
>>>>
>>>>
>>>
>>>
>>> --
>>> Teresa Johnson |  Software Engineer |  tejohnson at google.com |
>>> 408-460-2413 <(408)%20460-2413>
>>>
>>
>>
>
>
> --
> Teresa Johnson |  Software Engineer |  tejohnson at google.com |
> 408-460-2413 <(408)%20460-2413>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20180218/e9521778/attachment.html>