[PATCH] D16725: AMDGPU/SI: Implement a work-around for smrd corrupting vccz bit

[Matt Arsenault via llvm-commits]

arsenm added inline comments.

================
Comment at: lib/Target/AMDGPU/SIInsertWaits.cpp:523
@@ +522,3 @@
+          VCCZCorrupt = true;
+        } else if (!hasOutstandingLGKM() && I->modifiesRegister(AMDGPU::VCC, TRI)) {
+          // FIXME: We only care about SMRD instructions here, not LDS or GDS. 
----------------
Do we need to worry about modifications to vcc_lo/vcc_hi?

================
Comment at: lib/Target/AMDGPU/SIInsertWaits.cpp:536
@@ +535,3 @@
+          // terminators, and we always wait on everything at the end of the
+          // block, so if we only wait on LGKM heere, we might end up with
+          // another s_waitcnt inserted right after this if there are non-LGKM
----------------
Typo heere


http://reviews.llvm.org/D16725