[PATCH] D26506: Compiler-rt part of D26230: Add (constant) masked load/store support

[Filipe Cabecinhas via llvm-commits]

Are we setting up that feature?

I'll check when I get home, but didn't find it before. Checking for avx on
lit/cm ale Will be annoying, but won't be a bad thing.

On Fri, 2 Dec 2016 at 19:59, Kuba (Brecka) Mracek via Phabricator <
reviews at reviews.llvm.org> wrote:

> kubabrecka added a comment.
>
>
>
> In https://reviews.llvm.org/D26506#611883, @filcab wrote:
>
>
>
> > I've tried pushing this test, but one of the buildbots (at least)
> doesn't have AVX support. And AVX/AVX2 are the only things on X86 that will
> generate those masked loads/stores.
>
> >  I'd go with either:
>
> >
>
> >   No testing on compiler-rt (ASan doesn't behave differently in the face
> of masked ops, so we should be able to go by with only the tests from llvm)
>
> >   Testing on compiler-rt using IR, *assuming* we can lower calls to
> masked.loads/stores even on non-AVX CPUs
>
> >   Testing on compiler-rt using some other arch that's not X86 (I don't
> really have the hardware to try that out, though)
>
>
>
>
>
> Can we just add "REQUIRES: avx" to the test and lit?
>
>
>
>
>
> Repository:
>
>   rL LLVM
>
>
>
> https://reviews.llvm.org/D26506
>
>
>
>
>
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20161202/891aa7bc/attachment.html>