[PATCH] D98047: [AMDGPU][GlobalISel] Transform (fadd (fma x, y, (fpext (fmul u, v))), z) -> (fma x, y, (fma (fpext u), (fpext v), z))

[Mirko Brkusanin via Phabricator via llvm-commits]

mbrkusanin accepted this revision.
mbrkusanin added a comment.
This revision is now accepted and ready to land.

LGTM with a few nits



================
Comment at: llvm/lib/CodeGen/GlobalISel/CombinerHelper.cpp:4640-4643
+  const auto &TLI = *MI.getMF()->getSubtarget().getTargetLowering();
+  LLT DstType = MRI.getType(MI.getOperand(0).getReg());
+  MachineInstr *LHS = MRI.getVRegDef(MI.getOperand(1).getReg());
+  MachineInstr *RHS = MRI.getVRegDef(MI.getOperand(2).getReg());
----------------
Put this after canCombineFMadOrFMA call and !Agressive check


================
Comment at: llvm/lib/CodeGen/GlobalISel/CombinerHelper.cpp:4663
+
+  auto buildMatchInfo = [=, &MI](Register U, Register V, Register Z,
+                                 Register X, Register Y, MachineIRBuilder &B) {
----------------
Maybe add a comment for what this builds


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D98047/new/

https://reviews.llvm.org/D98047