[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#954525, @craig.topper wrote:

> Ok so now that we have a plan. Can we review this patch on its content?


My person preference is for additive features, not subtractive ones. Can you split AVX-512 into the 512-bit-register part and the rest? Do you prefer to do it this way? They'll need to be a follow-up patch to adjust the calls to addRegisterClass, etc. in X86ISelLowering, and I suspect the code will look cleaner in the end with additive features.


https://reviews.llvm.org/D41096