[all-commits] [llvm/llvm-project] 905458: [X86] combineTargetShuffle - fold vpermv3(widen(x)...

[Simon Pilgrim via All-commits]

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: 90545897c07094221c202b959f6192cfb624f94b
      https://github.com/llvm/llvm-project/commit/90545897c07094221c202b959f6192cfb624f94b
  Author: Simon Pilgrim <llvm-dev at redking.me.uk>
  Date:   2026-06-10 (Wed, 10 Jun 2026)

  Changed paths:
    M llvm/lib/Target/X86/X86ISelLowering.cpp
    M llvm/test/CodeGen/X86/vector-fshr-128.ll
    M llvm/test/CodeGen/X86/vector-fshr-rot-128.ll
    M llvm/test/CodeGen/X86/vector-interleaved-load-i8-stride-8.ll
    M llvm/test/CodeGen/X86/vector-interleaved-store-i16-stride-4.ll
    M llvm/test/CodeGen/X86/vector-interleaved-store-i16-stride-7.ll
    M llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-5.ll
    M llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-8.ll

  Log Message:
  -----------
  [X86] combineTargetShuffle - fold vpermv3(widen(x),mask,widen(y)) -> vpermv(widen(concat(x,y)),mask') (#203031)

We already handle the case where the src vectors were half size, but we
can generalize this to widening from xmm to zmm as well - mainly to help
non-VLX builds

A couple of codesize increases in non-VLX builds - mostly from
additional asm / kill comments, but also due to a couple of poor folds
in combineConcatVectorOps that need further yak shaving.



To unsubscribe from these emails, change your notification settings at https://github.com/llvm/llvm-project/settings/notifications