[llvm-branch-commits] [llvm] [AMDGPU] Make amdgpu.noclobber respect acquire operations (PR #219991)

[Sameer Sahasrabuddhe via llvm-branch-commits]

ssahasra wrote:

The semantics and the implementation look correct to me. But the concern that @krzysz00 raised about fence address spaces via MMRA seems important. Would it be too difficult to have it ready for review either in this PR itself or as stacked PR?

https://github.com/llvm/llvm-project/pull/219991