[PATCH] D150324: [X86] Make constant `mul` -> `shl` + `add`/`sub` work for vector types

[Noah Goldstein via Phabricator via llvm-commits]

goldstein.w.n added a comment.

In D150324#4333880 <https://reviews.llvm.org/D150324#4333880>, @efriedma wrote:

> On current chips, pmullw has the same throughput as psllw; it seems unlikely to me that it's profitable to replace a pmullw with a four-instruction sequence.

On SKX and newer there is an additional shuffle unit (p01). So will make the arbitrary `Pow2A + Pow2B` only for targets with fast shift.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D150324/new/

https://reviews.llvm.org/D150324