[llvm] [AMDGPU][SIInsertWaitCnts] Use RegUnits-based tracking (PR #162077)

[Pierre van Houtryve via llvm-commits]

================
@@ -1150,27 +1110,31 @@ void WaitcntBrackets::print(raw_ostream &OS) const {
       // Print vgpr scores.
       unsigned LB = getScoreLB(T);
 
-      for (int J = 0; J <= VgprUB; J++) {
-        unsigned RegScore = getRegScore(J, T);
+      for (auto &[ID, Info] : VMem) {
----------------
Pierre-vh wrote:

I sorted the keys, I think quality of debug output is important.

https://github.com/llvm/llvm-project/pull/162077