[PATCH] D69581: AMDGPU/GlobalISel: Legalize FDIV32

[Matt Arsenault via Phabricator via llvm-commits]

arsenm added inline comments.


================
Comment at: llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp:1977
+                               MachineIRBuilder &B,
+                               unsigned Flags) {
+  LLT S32 = LLT::scalar(32);
----------------
You don't need to propagate the flags here. You should remove the argument and not bother setting it on the instructions. These don't actually do anything with FP values


================
Comment at: llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp:1993
+    B.buildInstr(AMDGPU::S_DENORM_MODE)
+      .addUse(NewDenormModeValue.getReg(0))
+      .setMIFlags(Flags);
----------------
This is an immediate field. You shouldn't be materializing it in a register 


================
Comment at: llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp:2006
+      .addUse(NewDenormModeValue.getReg(0))
+      .addUse(ModeRegBitField.getReg(0))
+      .setMIFlags(Flags);
----------------
Ditto


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D69581/new/

https://reviews.llvm.org/D69581