[all-commits] [llvm/llvm-project] 56eb12: AMDGPU: Do not create phi user for atomicrmw with ...

[Matt Arsenault via All-commits]

  Branch: refs/heads/users/arsenm/amdgpu-avoid-introducing-rmw-uses
  Home:   https://github.com/llvm/llvm-project
  Commit: 56eb1227ab639b6483e7c62716c9b4bd5e589fec
      https://github.com/llvm/llvm-project/commit/56eb1227ab639b6483e7c62716c9b4bd5e589fec
  Author: Matt Arsenault <Matthew.Arsenault at amd.com>
  Date:   2024-08-13 (Tue, 13 Aug 2024)

  Changed paths:
    M llvm/lib/Target/AMDGPU/SIISelLowering.cpp
    M llvm/test/Transforms/AtomicExpand/AMDGPU/expand-atomic-mmra.ll
    M llvm/test/Transforms/AtomicExpand/AMDGPU/expand-atomic-rmw-fadd-flat-specialization.ll

  Log Message:
  -----------
  AMDGPU: Do not create phi user for atomicrmw with no uses

Currently the custom expansion introducing address space tests
always introduces a phi for the return value, even if it is unused.

Avoid introducing the phi use if we don't need the return value.
This avoids unnecessarily expanding the global sequence on
targets that only support the no return version in a future
patch that enables incremental legalization of atomics.



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