[PATCH] D113609: [X86][Costmodel] `getReplicationShuffleCost()`: promote 16 bit-wide elements to 32 bit when no AVX512BW

[Roman Lebedev via Phabricator via llvm-commits]

lebedev.ri added a comment.

In D113609#3127794 <https://reviews.llvm.org/D113609#3127794>, @RKSimon wrote:

> Can you give more examples of the bad cast costs?  trunc <32 x i32> to <32 x i8> -> https://wide.godbolt.org/z/Y3rznPEnY seems to be OK

I wasn't actually saying that `trunc <32 x i32> to <32 x i8>` is bad, i was only describing that the cost should include not only the extensions/truncation, but vector extractions/insertions.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D113609/new/

https://reviews.llvm.org/D113609