[all-commits] [llvm/llvm-project] 012d21: [LV] Use scalar CMP for active-lane-mask with scal...

[Cameron McInally via All-commits]

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: 012d217174e28dc000ba1f437827cd23ec5ef606
      https://github.com/llvm/llvm-project/commit/012d217174e28dc000ba1f437827cd23ec5ef606
  Author: Cameron McInally <cameron.mcinally at nyu.edu>
  Date:   2024-03-06 (Wed, 06 Mar 2024)

  Changed paths:
    M llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp
    M llvm/test/Transforms/LoopVectorize/AArch64/uniform-args-call-variants.ll
    A llvm/test/Transforms/LoopVectorize/pr73894.ll
    M llvm/test/Transforms/LoopVectorize/strict-fadd-interleave-only.ll

  Log Message:
  -----------
  [LV] Use scalar CMP for active-lane-mask with scalar VF (#83902)

Instead of generating a <1 x i1> active lane mask intrinsic, generate
the equivalent scalar ICMP instead. This allows us to avoid
unnecessarily extracting the scalar part from the vector mask.

Fixes llvm#73894.



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