[AMDGPU] Simplify selection of llvm.amdgcn.inverse.ballot. NFCI. (#99345)

Author: jayfoad

llvm/lib/Target/AMDGPU/AMDGPUISelDAGToDAG.cpp

@@ -2775,18 +2775,6 @@ void AMDGPUDAGToDAGISel::SelectINTRINSIC_WO_CHAIN(SDNode *N) {
   case Intrinsic::amdgcn_interp_p1_f16:
     SelectInterpP1F16(N);
     return;
-  case Intrinsic::amdgcn_inverse_ballot:
-    switch (N->getOperand(1).getValueSizeInBits()) {
-    case 32:
-      Opcode = AMDGPU::S_INVERSE_BALLOT_U32;
-      break;
-    case 64:
-      Opcode = AMDGPU::S_INVERSE_BALLOT_U64;
-      break;
-    default:
-      llvm_unreachable("Unsupported size for inverse ballot mask.");
-    }
-    break;
   default:
     SelectCode(N);
     break;

llvm/lib/Target/AMDGPU/AMDGPUInstructionSelector.cpp

@@ -1055,8 +1055,6 @@ bool AMDGPUInstructionSelector::selectG_INTRINSIC(MachineInstr &I) const {
     return selectIntrinsicCmp(I);
   case Intrinsic::amdgcn_ballot:
     return selectBallot(I);
-  case Intrinsic::amdgcn_inverse_ballot:
-    return selectInverseBallot(I);
   case Intrinsic::amdgcn_reloc_constant:
     return selectRelocConstant(I);
   case Intrinsic::amdgcn_groupstaticsize:
@@ -1449,17 +1447,6 @@ bool AMDGPUInstructionSelector::selectBallot(MachineInstr &I) const {
   return true;
 }
 
-bool AMDGPUInstructionSelector::selectInverseBallot(MachineInstr &I) const {
-  MachineBasicBlock *BB = I.getParent();
-  const DebugLoc &DL = I.getDebugLoc();
-  const Register DstReg = I.getOperand(0).getReg();
-  const Register MaskReg = I.getOperand(2).getReg();
-
-  BuildMI(*BB, &I, DL, TII.get(AMDGPU::COPY), DstReg).addReg(MaskReg);
-  I.eraseFromParent();
-  return true;
-}
-
 bool AMDGPUInstructionSelector::selectRelocConstant(MachineInstr &I) const {
   Register DstReg = I.getOperand(0).getReg();
   const RegisterBank *DstBank = RBI.getRegBank(DstReg, *MRI, TRI);

llvm/lib/Target/AMDGPU/AMDGPUInstructionSelector.h

@@ -112,7 +112,6 @@ class AMDGPUInstructionSelector final : public InstructionSelector {
   bool selectDivScale(MachineInstr &MI) const;
   bool selectIntrinsicCmp(MachineInstr &MI) const;
   bool selectBallot(MachineInstr &I) const;
-  bool selectInverseBallot(MachineInstr &I) const;
   bool selectRelocConstant(MachineInstr &I) const;
   bool selectGroupStaticSize(MachineInstr &I) const;
   bool selectReturnAddress(MachineInstr &I) const;

llvm/lib/Target/AMDGPU/SIISelLowering.cpp

@@ -5480,24 +5480,11 @@ MachineBasicBlock *SITargetLowering::EmitInstrWithCustomInserter(
     return BB;
   }
   case AMDGPU::S_INVERSE_BALLOT_U32:
-  case AMDGPU::S_INVERSE_BALLOT_U64: {
-    MachineRegisterInfo &MRI = BB->getParent()->getRegInfo();
-    const GCNSubtarget &ST = MF->getSubtarget<GCNSubtarget>();
-    const SIRegisterInfo *TRI = ST.getRegisterInfo();
-    const DebugLoc &DL = MI.getDebugLoc();
-    const Register DstReg = MI.getOperand(0).getReg();
-    Register MaskReg = MI.getOperand(1).getReg();
-
-    const bool IsVALU = TRI->isVectorRegister(MRI, MaskReg);
-
-    if (IsVALU) {
-      MaskReg = TII->readlaneVGPRToSGPR(MaskReg, MI, MRI);
-    }
-
-    BuildMI(*BB, &MI, DL, TII->get(AMDGPU::COPY), DstReg).addReg(MaskReg);
-    MI.eraseFromParent();
+  case AMDGPU::S_INVERSE_BALLOT_U64:
+    // These opcodes only exist to let SIFixSGPRCopies insert a readfirstlane if
+    // necessary. After that they are equivalent to a COPY.
+    MI.setDesc(TII->get(AMDGPU::COPY));
     return BB;
-  }
   case AMDGPU::ENDPGM_TRAP: {
     const DebugLoc &DL = MI.getDebugLoc();
     if (BB->succ_empty() && std::next(MI.getIterator()) == BB->end()) {

llvm/lib/Target/AMDGPU/SIInstrInfo.cpp

@@ -6686,7 +6686,9 @@ SIInstrInfo::legalizeOperands(MachineInstr &MI,
       MI.getOpcode() == AMDGPU::S_QUADMASK_B32 ||
       MI.getOpcode() == AMDGPU::S_QUADMASK_B64 ||
       MI.getOpcode() == AMDGPU::S_WQM_B32 ||
-      MI.getOpcode() == AMDGPU::S_WQM_B64) {
+      MI.getOpcode() == AMDGPU::S_WQM_B64 ||
+      MI.getOpcode() == AMDGPU::S_INVERSE_BALLOT_U32 ||
+      MI.getOpcode() == AMDGPU::S_INVERSE_BALLOT_U64) {
     MachineOperand &Src = MI.getOperand(1);
     if (Src.isReg() && RI.hasVectorRegisters(MRI.getRegClass(Src.getReg())))
       Src.setReg(readlaneVGPRToSGPR(Src.getReg(), MI, MRI));

llvm/lib/Target/AMDGPU/SIInstructions.td

@@ -212,9 +212,15 @@ def EXIT_STRICT_WQM : SPseudoInstSI <(outs SReg_1:$sdst), (ins SReg_1:$src0)> {
 }
 
 let usesCustomInserter = 1 in {
-def S_INVERSE_BALLOT_U32 : SPseudoInstSI <(outs SReg_32:$sdst), (ins SSrc_b32:$mask)>;
+def S_INVERSE_BALLOT_U32 : SPseudoInstSI<
+  (outs SReg_32:$sdst), (ins SSrc_b32:$mask),
+  [(set i1:$sdst, (int_amdgcn_inverse_ballot i32:$mask))]
+>;
 
-def S_INVERSE_BALLOT_U64 : SPseudoInstSI <(outs SReg_64:$sdst), (ins SSrc_b64:$mask)>;
+def S_INVERSE_BALLOT_U64 : SPseudoInstSI<
+  (outs SReg_64:$sdst), (ins SSrc_b64:$mask),
+  [(set i1:$sdst, (int_amdgcn_inverse_ballot i64:$mask))]
+>;
 } // End usesCustomInserter = 1
 
 // Pseudo instructions used for @llvm.fptrunc.round upward