[PATCH] D54267: [X86][SSE] Add lowerVectorShuffleAsByteRotateAndPermute (PR39387)

[Simon Pilgrim via Phabricator via llvm-commits]

RKSimon added a comment.

In https://reviews.llvm.org/D54267#1291801, @craig.topper wrote:

> Does this change the code generated for rgbcmyk here https://godbolt.org/z/cot3xT    I filed the original PR based on what happened from trying to vectorize it for sse4.2 which we don't currently do, but I think the two vblendvbs in the avx2 output are similar.


At the moment I haven't added this to v32i8 shuffle lowering as its still causing some regressions (a lot of the lowering patterns in lowerV32I8VectorShuffle are hidden in sub functions making it more difficult to use lowerVectorShuffleAsByteRotateAndPermute in the right cicrumstances). I could add it but I'd have to bail out if I see certain patterns - I'll update the patch for you to see the effect.


Repository:
  rL LLVM

https://reviews.llvm.org/D54267