[PATCH] D128656: [AMDGPU] gfx11 Generate VOPD Instructions

[Stanislav Mekhanoshin via Phabricator via llvm-commits]

rampitec added inline comments.


================
Comment at: llvm/lib/Target/AMDGPU/GCNCreateVOPD.cpp:85
+      break;
+    case AMDGPU::V_FMAAK_F32:
+      VOPDInst.add(FirstMI->getOperand(2));
----------------
Could merge this case with the previous.


================
Comment at: llvm/lib/Target/AMDGPU/GCNCreateVOPD.cpp:103
+      break;
+    case AMDGPU::V_FMAAK_F32:
+      VOPDInst.add(SecondMI->getOperand(2));
----------------
Here as well.


================
Comment at: llvm/lib/Target/AMDGPU/GCNVOPDUtils.cpp:172
+/// Greedily pairs instruction candidates. O(n^2) algorithm.
+struct VOPDPairingMutation : ScheduleDAGMutation {
+  ShouldSchedulePredTy shouldScheduleAdjacent; // NOLINT: function pointer
----------------
Isn't it the same as MacroFusion?


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D128656/new/

https://reviews.llvm.org/D128656