[PATCH] D46266: [X86] Introduce X86SchedWriteWidths schedule wrapper for different vector widths.

[Simon Pilgrim via Phabricator via llvm-commits]

RKSimon added a comment.

In https://reviews.llvm.org/D46266#1082964, @craig.topper wrote:

> This seems reasonsable to me. For the scalar question, do any CPUs have different scalar performance? Except for divide/sqrt instructions.


There are a few models that have different fmul values as well - Atom + Btver2 for instance - they can be just split into float/double classes like I intend to for fsqrt/fdiv.


Repository:
  rL LLVM

https://reviews.llvm.org/D46266