[all-commits] [llvm/llvm-project] 21aec4: [X86][SLM] Fix vector PSHUFB + variable shift reso...

[Simon Pilgrim via All-commits]

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: 21aec4fdc5de213a1009bc684adff988c290c6b3
      https://github.com/llvm/llvm-project/commit/21aec4fdc5de213a1009bc684adff988c290c6b3
  Author: Simon Pilgrim <llvm-dev at redking.me.uk>
  Date:   2021-05-26 (Wed, 26 May 2021)

  Changed paths:
    M llvm/lib/Target/X86/X86ScheduleSLM.td
    M llvm/test/tools/llvm-mca/X86/SLM/resources-mmx.s
    M llvm/test/tools/llvm-mca/X86/SLM/resources-sse2.s
    M llvm/test/tools/llvm-mca/X86/SLM/resources-ssse3.s

  Log Message:
  -----------
  [X86][SLM] Fix vector PSHUFB + variable shift resource/throughputs

Match whats documented in the Intel AOM (+Agner) - PSHUFB xmm is really slow, and mmx/xmm vector shifts are half rate.

Noticed while working to get the cost tables to more closely match llvm-mca analysis, in this case for shifts and truncations.