[PATCH] D112064: [GlobalISel] Combine (build_vector_trunc x, undef) -> (bitcast x)

[Matt Arsenault via Phabricator via llvm-commits]

arsenm added a comment.

Should add some MIR tests specifically for this in case we decide to stop using G_BUILD_VECTOR_TRUNC someday



================
Comment at: llvm/lib/Target/AMDGPU/AMDGPUInstructionSelector.cpp:630-637
-  // (build_vector_trunc $src0, undef -> copy $src0
-  MachineInstr *Src1Def = getDefIgnoringCopies(Src1, *MRI);
-  if (Src1Def && Src1Def->getOpcode() == AMDGPU::G_IMPLICIT_DEF) {
-    MI.setDesc(TII.get(AMDGPU::COPY));
-    MI.RemoveOperand(2);
-    return RBI.constrainGenericRegister(Dst, AMDGPU::SReg_32RegClass, *MRI) &&
-           RBI.constrainGenericRegister(Src0, AMDGPU::SReg_32RegClass, *MRI);
----------------
I think deleting this should be split into a separate patch


================
Comment at: llvm/test/CodeGen/AMDGPU/GlobalISel/inst-select-build-vector-trunc.v2s16.mir:157
 
     ; GFX9-LABEL: name: test_build_vector_trunc_s_v2s16_s_s32_s_undef_s32
     ; GFX9: liveins: $sgpr0
----------------
foad wrote:
> I could just remove all these "undef" test cases, since there is no longer any code in AMDGPUInstructionSelector that handles IMPLICIT_DEF specially.
Yes


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D112064/new/

https://reviews.llvm.org/D112064