[PATCH] D43306: [X86] Add pass to infer required-vector-width attribute based on size of function arguments and use of intrinsics

[Eric Christopher via Phabricator via llvm-commits]

echristo added a comment.

In https://reviews.llvm.org/D43306#1050892, @chandlerc wrote:

> One thing I should clarify here:
>
> What I'm imagining is reasonable to be per-subtarget and inform the legalizer are the *set of legal vector types*. This is subtly different from the *size of legal registers*. Does that still make sense?


This split-hair makes sense to me and works for what I'm imagining we do.


https://reviews.llvm.org/D43306