[all-commits] [llvm/llvm-project] 639d19: [AMDGPU] Constant folding for wave-reduce intrinsi...

[Aaditya via All-commits]

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: 639d199326c1ae95d8b2fa26390d45ade0f20a19
      https://github.com/llvm/llvm-project/commit/639d199326c1ae95d8b2fa26390d45ade0f20a19
  Author: Aaditya <115080342+easyonaadit at users.noreply.github.com>
  Date:   2026-09-22 (Tue, 22 Sep 2026)

  Changed paths:
    M llvm/lib/Analysis/ConstantFolding.cpp
    M llvm/lib/Target/AMDGPU/SIISelLowering.cpp
    R llvm/test/CodeGen/AMDGPU/GlobalISel/atomic_optimizations_mul_one.ll
    A llvm/test/CodeGen/AMDGPU/atomic_optimizations_mul_one.ll

  Log Message:
  -----------
  [AMDGPU] Constant folding for wave-reduce intrinsics (#212755)

Relates to: https://github.com/llvm/llvm-project/pull/193142
A missed optimization opportunity.
While most constants to `add`, `sub` and `xor` wave reductions
cannot be folded, some identity constants(0, 1) can do with
some special handling.



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