[llvm] [AMDGPU][SIInsertWaitcnts] Do not add s_waitcnt when the counters are known to be 0 already (PR #72830)

[Jay Foad via llvm-commits]

================
@@ -292,6 +292,11 @@ class WaitcntBrackets {
     VgprVmemTypes[GprNo] = 0;
   }
 
+  void setNonKernelFunctionInitialState() {
+    setScoreUB(VS_CNT, getWaitCountMax(VS_CNT));
+    PendingEvents |= WaitEventMaskForInst[VS_CNT];
+  }
----------------
jayfoad wrote:

Could you please split out a prerequisite patch that just adds this functionality? I think there are only a couple of tests that it will affect, to do with barriers or release vgprs.

https://github.com/llvm/llvm-project/pull/72830