[PATCH] D44434: [AMDGPU] Waitcnt pass: propagate info in the case of a single basic block loop

[Stanislav Mekhanoshin via Phabricator via llvm-commits]

rampitec added inline comments.


================
Comment at: lib/Target/AMDGPU/SIInsertWaitcnts.cpp:1313
 
-  // Clear the score bracket state.
-  ScoreBrackets->clear();
+  for (MachineBasicBlock *Pred : Block.predecessors()) {
+    if (Pred == &Block) {
----------------
I think you can use std::find().


================
Comment at: lib/Target/AMDGPU/SIInsertWaitcnts.cpp:1328
+  if (IsSelfPred) {
+    ScoreBrackets = S.get();
+  }
----------------
Do you need to make a copy if block is not a self predecessor? You are not going to reuse it then as far as I understand.


https://reviews.llvm.org/D44434