[all-commits] [llvm/llvm-project] 2f11b3: [MachineCombiner] Don't compute the latency of tra...

[weiguozhi via All-commits]

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: 2f11b3a6d7b8083be570f79f4ca18dfb2648eda3
      https://github.com/llvm/llvm-project/commit/2f11b3a6d7b8083be570f79f4ca18dfb2648eda3
  Author: Guozhi Wei <carrot at google.com>
  Date:   2022-07-14 (Thu, 14 Jul 2022)

  Changed paths:
    M llvm/lib/CodeGen/MachineCombiner.cpp
    M llvm/test/CodeGen/AArch64/aarch64-combine-fmul-fsub.mir
    A llvm/test/CodeGen/AArch64/machine-combiner-transient.ll
    M llvm/test/CodeGen/AArch64/neon-mla-mls.ll
    M llvm/test/CodeGen/X86/machine-combiner-int-vec.ll
    M llvm/test/CodeGen/X86/machine-combiner-int.ll

  Log Message:
  -----------
  [MachineCombiner] Don't compute the latency of transient instructions

If an MI will not generate a target instruction, we should not compute its
latency. Then we can compute more precise instruction sequence cost, and get
better result.

Differential Revision: https://reviews.llvm.org/D129615