[PATCH] D149332: [AMDGPU] Also consider global and scratch instructions when flushing vmcnt counter in loop preheader

[Jay Foad via Phabricator via llvm-commits]

foad added inline comments.


================
Comment at: llvm/lib/Target/AMDGPU/SIInsertWaitcnts.cpp:1724
     for (MachineInstr &MI : *MBB) {
-      if (SIInstrInfo::isVMEM(MI)) {
+      if (updateVMCntOnly(MI)) {
         if (MI.mayLoad())
----------------
My only slight concern is whether we should also accept FLAT instructions here? They update vmcnt but not //only// vmcnt. I'm not sure what the answer is.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D149332/new/

https://reviews.llvm.org/D149332