[all-commits] [llvm/llvm-project] 81fae0: [Clang][AMDGPU] Stop defaulting to `one-as` for al...

[Joseph Huber via All-commits]

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: 81fae0d5e3d3378959483ccd7709a212731bff3f
      https://github.com/llvm/llvm-project/commit/81fae0d5e3d3378959483ccd7709a212731bff3f
  Author: Joseph Huber <huberjn at outlook.com>
  Date:   2025-01-06 (Mon, 06 Jan 2025)

  Changed paths:
    M clang/lib/CodeGen/Targets/AMDGPU.cpp
    M clang/test/CodeGen/scoped-atomic-ops.c
    M clang/test/CodeGen/scoped-fence-ops.c
    M clang/test/CodeGenCUDA/amdgpu-atomic-ops.cu
    M clang/test/CodeGenCUDA/atomic-ops.cu

  Log Message:
  -----------
  [Clang][AMDGPU] Stop defaulting to `one-as` for all atomic scopes (#120095)

Summary:
The documentation at
https://llvm.org/docs/AMDGPUUsage.html#memory-scopes states that these
'one-as' modifiers are more specific versions of the scopes that only
apply to a specific address space. This doesn't make sense for fences
which have no associated address space to use, and it's a more
restrictive version the normal scope. This should not tbe the default
behavior, but it is currently emitted in all cases except for
sequentially consistent.



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