[PATCH] D41096: [X86] Initial support for prefer-vector-width function attribute

[Hal Finkel via Phabricator via llvm-commits]

hfinkel added a comment.

In https://reviews.llvm.org/D41096#953951, @spatel wrote:

> In https://reviews.llvm.org/D41096#953922, @craig.topper wrote:
>
> > How/where do you propose detecting the presence or absense of 512-bit instructions to change the CPU name?
>
>
> I think the path through clang's CodeGenFunction::checkTargetFeatures() is where we tell the user if they've used an intrinsic without the required CPU attribute. So somewhere around there might work?


I think this would go in sys::getHostCPUName in lib/Support/Host.cpp (or in sys::getHostCPUFeatures as appropriate).


https://reviews.llvm.org/D41096