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

[Matt Arsenault via llvm-branch-commits]

================
@@ -77,6 +77,13 @@ void AMDGPUAnnotateUniformValues::visitLoadInst(LoadInst &I) {
   // for memory operations that are live in to entry points only.
   if (!isEntryFunc)
     return;
+
+  // If I is atomic, it might see the effects of concurrent clobbering accesses,
+  // so local clobber analysis is not sufficient to ensure that it sees the
+  // initial value.
+  if (I.isAtomic())
----------------
arsenm wrote:

What about volatile? I thought we already handled that correctly, and would expect this was an isSimple check. Where is volatile filtered out? 

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