[PATCH] D17543: AMDGPU/SI: Handle wait states required for DPP instructions

[Matt Arsenault via llvm-commits]

arsenm added inline comments.

================
Comment at: lib/Target/AMDGPU/SIInsertWaits.cpp:507
@@ +506,3 @@
+
+    // If we have reached the strt of the block, we need to check predecessors.
+    if (SearchI == E) {
----------------
Typo strt

================
Comment at: lib/Target/AMDGPU/SIInsertWaits.cpp:509-510
@@ +508,4 @@
+    if (SearchI == E) {
+      for (MachineBasicBlock::pred_iterator PI = SearchMBB->pred_begin(),
+           PE = SearchMBB->pred_end(); PI != PE; ++PI) {
+        MachineBasicBlock *Pred = *PI;
----------------
You should be able to use range loop with predecessors()

================
Comment at: lib/Target/AMDGPU/SIInsertWaits.cpp:516
@@ +515,3 @@
+          continue;
+        insertDPPWaitStates(DPP, Pred->rbegin(), Pred, WaitStates);
+      }
----------------
There could be a break here since there will only ever be one fall through block, although it doesn't really matter because with structured control flow there is at most one predecessor


http://reviews.llvm.org/D17543