[llvm] [AMDGPU][SIInsertWaitcnts] Cleanup: Remove WaitEventMaskForInst member variable (PR #178030)

[via llvm-commits]

================
@@ -3328,9 +3328,7 @@ bool SIInsertWaitcnts::run(MachineFunction &MF) {
   for (auto T : inst_counter_types())
     ForceEmitWaitcnt[T] = false;
 
-  WaitEventMaskForInst = WCG->getWaitEventMask();
-
-  SmemAccessCounter = eventCounter(WaitEventMaskForInst, SMEM_ACCESS);
+  SmemAccessCounter = eventCounter(WCG->getWaitEventMask(), SMEM_ACCESS);
----------------
vporpo wrote:

Anyway, I think I will merge this to unblock the follow-up patch, because we may need to rethink whether having `SIInsertWaitcnt::getWaitEventMask()` in the first place is the best implementation, as there seem to be more WCG functions that would need to be exposed in a similar way. So perhaps providing public access to WCG with something like `getWCG()` may be the better option.

https://github.com/llvm/llvm-project/pull/178030