[llvm-branch-commits] [clang] [llvm] [AMDGPU] Introduce asyncmark/wait intrinsics (PR #180467)

[Sameer Sahasrabuddhe via llvm-branch-commits]

================
@@ -710,6 +710,24 @@ class SIInsertWaitcnts {
   std::optional<WaitEventType>
   getExpertSchedulingEventType(const MachineInstr &Inst) const;
 
+  bool isAsync(const MachineInstr &MI) const {
----------------
ssahasra wrote:

This particular function can't go to TII, because it computes the result from a dynamic operand of the instruction. That stops the other two functions since they use this function.

https://github.com/llvm/llvm-project/pull/180467