[PATCH] D82551: [AMDGPU] Don't combine DPP if DPP register is used more than once per instruction

[Jay Foad via Phabricator via llvm-commits]

foad accepted this revision.
foad added inline comments.
This revision is now accepted and ready to land.


================
Comment at: llvm/lib/Target/AMDGPU/GCNDPPCombine.cpp:517-520
+
+    if (Src0 && Src1 && Src0->isReg() && Src1->isReg() &&
+        Src0->getReg() == Src1->getReg() &&
+        Src0->getSubReg() == Src1->getSubReg()) {
----------------
This looks technically OK now. I still think it would be cleaner to structure the code as:
```
if (Use == Src0) {
  // do it
} else if (Use == Src1 && commutable && Src0 not identical to Src1) {
  // commute and do it
} else {
  // fail
}
```
But I'll leave that up to your judgement.


================
Comment at: llvm/lib/Target/AMDGPU/GCNDPPCombine.cpp:518-520
+    if (Src0 && Src1 && Src0->isReg() && Src1->isReg() &&
+        Src0->getReg() == Src1->getReg() &&
+        Src0->getSubReg() == Src1->getSubReg()) {
----------------
Could use MachineOperand::isIdenticalTo ?


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D82551/new/

https://reviews.llvm.org/D82551