[clang] [llvm] [AMDGPU] Introduce asyncmark/wait intrinsics (PR #173259)

[Jay Foad via llvm-commits]

================
@@ -2213,8 +2384,10 @@ bool SIInsertWaitcnts::generateWaitcntInstBefore(MachineInstr &MI,
   if (ForceEmitZeroLoadFlag && Wait.LoadCnt != ~0u)
     Wait.LoadCnt = 0;
 
-  return generateWaitcnt(Wait, MI.getIterator(), *MI.getParent(), ScoreBrackets,
-                         OldWaitcntInstr);
+  bool Changed = generateWaitcnt(Wait, MI.getIterator(), *MI.getParent(),
----------------
jayfoad wrote:

Remove this change?

https://github.com/llvm/llvm-project/pull/173259