[PATCH] D22782: Added 'inline' attribute to __init to inline the basic_string's constructor

[Eric Fiselier via cfe-commits]

On Mon, Aug 8, 2016 at 9:32 AM, David Blaikie via cfe-commits <
cfe-commits at lists.llvm.org> wrote:

> I'm still (as per another similar thread) a bit concerned this is working
> around a compiler optimizer bug - I'd love to see a standalone example of
> this behavior (that adding the inline keyword to an already
> inline/available definition is what's causing the inlining) so we can look
> at what the compiler might be doing wrong (& consider fixing that as the
> root cause instead of having to change every instance of this problem) -
> preferably an example without std::string/headers/etc (so, preprocessed and
> reduced).
>

Normally I would agree that this is an optimizer bug, but std::string is
externally instantiated.
It seems reasonable for Clang *not* to inline externally instantiated
templates unless requested.
Simply removing the extern template declaration will resolve these inlining
issues.

Here is the standalone example you requested: https://godbolt.org/g/aVwEMR



>
> But up to you folks maintaining libc++ whether this goes in of course,
> just my 2c.
>
> On Wed, Aug 3, 2016 at 5:23 AM Aditya Kumar via cfe-commits <
> cfe-commits at lists.llvm.org> wrote:
>
>> hiraditya added a comment.
>>
>> In https://reviews.llvm.org/D22782#504416, @EricWF wrote:
>>
>> > The change itself LGTM, although we probably want to inline the
>> forward/input iterator __init's as well.
>> >
>> > However I would like to see a small benchmark that demonstrates the
>> performance change. Please try and write the benchmark using Google
>> Benchmark.
>> >  Some helpful links:
>> >
>> > - http://libcxx.llvm.org/docs/TestingLibcxx.html#building-benchmarks
>> > - http://github.com/google/benchmark
>>
>>
>> Sure,
>> We'll come up with a synthetic benchmark to expose performance
>> improvements.
>>
>> Thanks,
>>
>>
>> https://reviews.llvm.org/D22782
>>
>>
>>
>> _______________________________________________
>> cfe-commits mailing list
>> cfe-commits at lists.llvm.org
>> http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
>>
>
> _______________________________________________
> cfe-commits mailing list
> cfe-commits at lists.llvm.org
> http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-commits/attachments/20160810/00fe9181/attachment.html>