3 files changed, 13 insertions, 55 deletions
diff --git a/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp b/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp
index 980e585..d35a022 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp
@@ -3117,30 +3117,20 @@ static bool isCttzOpc(unsigned Opc) {
 SDValue AMDGPUTargetLowering::lowerCTLZResults(SDValue Op,
                                                SelectionDAG &DAG) const {
   auto SL = SDLoc(Op);
-  auto Opc = Op.getOpcode();
   auto Arg = Op.getOperand(0u);
   auto ResultVT = Op.getValueType();
 
   if (ResultVT != MVT::i8 && ResultVT != MVT::i16)
     return {};
 
-  assert(isCtlzOpc(Opc));
+  assert(isCtlzOpc(Op.getOpcode()));
   assert(ResultVT == Arg.getValueType());
 
-  const uint64_t NumBits = ResultVT.getFixedSizeInBits();
-  SDValue NumExtBits = DAG.getConstant(32u - NumBits, SL, MVT::i32);
-  SDValue NewOp;
-
-  if (Opc == ISD::CTLZ_ZERO_UNDEF) {
-    NewOp = DAG.getNode(ISD::ANY_EXTEND, SL, MVT::i32, Arg);
-    NewOp = DAG.getNode(ISD::SHL, SL, MVT::i32, NewOp, NumExtBits);
-    NewOp = DAG.getNode(Opc, SL, MVT::i32, NewOp);
-  } else {
-    NewOp = DAG.getNode(ISD::ZERO_EXTEND, SL, MVT::i32, Arg);
-    NewOp = DAG.getNode(Opc, SL, MVT::i32, NewOp);
-    NewOp = DAG.getNode(ISD::SUB, SL, MVT::i32, NewOp, NumExtBits);
-  }
-
+  auto const LeadingZeroes = 32u - ResultVT.getFixedSizeInBits();
+  auto SubVal = DAG.getConstant(LeadingZeroes, SL, MVT::i32);
+  auto NewOp = DAG.getNode(ISD::ZERO_EXTEND, SL, MVT::i32, Arg);
+  NewOp = DAG.getNode(Op.getOpcode(), SL, MVT::i32, NewOp);
+  NewOp = DAG.getNode(ISD::SUB, SL, MVT::i32, NewOp, SubVal);
   return DAG.getNode(ISD::TRUNCATE, SL, ResultVT, NewOp);
 }
 
diff --git a/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp b/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp
index 15a4b67..bd7bf78 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp
@@ -1270,22 +1270,13 @@ AMDGPULegalizerInfo::AMDGPULegalizerInfo(const GCNSubtarget &ST_,
     .custom();
 
   // The 64-bit versions produce 32-bit results, but only on the SALU.
-  getActionDefinitionsBuilder(G_CTLZ_ZERO_UNDEF)
-      .legalFor({{S32, S32}, {S32, S64}})
-      .customIf(scalarNarrowerThan(1, 32))
-      .clampScalar(0, S32, S32)
-      .clampScalar(1, S32, S64)
-      .scalarize(0)
-      .widenScalarToNextPow2(0, 32)
-      .widenScalarToNextPow2(1, 32);
-
-  getActionDefinitionsBuilder(G_CTTZ_ZERO_UNDEF)
-      .legalFor({{S32, S32}, {S32, S64}})
-      .clampScalar(0, S32, S32)
-      .clampScalar(1, S32, S64)
-      .scalarize(0)
-      .widenScalarToNextPow2(0, 32)
-      .widenScalarToNextPow2(1, 32);
+  getActionDefinitionsBuilder({G_CTLZ_ZERO_UNDEF, G_CTTZ_ZERO_UNDEF})
+    .legalFor({{S32, S32}, {S32, S64}})
+    .clampScalar(0, S32, S32)
+    .clampScalar(1, S32, S64)
+    .scalarize(0)
+    .widenScalarToNextPow2(0, 32)
+    .widenScalarToNextPow2(1, 32);
 
   // S64 is only legal on SALU, and needs to be broken into 32-bit elements in
   // RegBankSelect.
@@ -2137,8 +2128,6 @@ bool AMDGPULegalizerInfo::legalizeCustom(
   case TargetOpcode::G_CTLZ:
   case TargetOpcode::G_CTTZ:
     return legalizeCTLZ_CTTZ(MI, MRI, B);
-  case TargetOpcode::G_CTLZ_ZERO_UNDEF:
-    return legalizeCTLZ_ZERO_UNDEF(MI, MRI, B);
   case TargetOpcode::G_INTRINSIC_FPTRUNC_ROUND:
     return legalizeFPTruncRound(MI, B);
   case TargetOpcode::G_STACKSAVE:
@@ -4156,25 +4145,6 @@ bool AMDGPULegalizerInfo::legalizeCTLZ_CTTZ(MachineInstr &MI,
   return true;
 }
 
-bool AMDGPULegalizerInfo::legalizeCTLZ_ZERO_UNDEF(MachineInstr &MI,
-                                                  MachineRegisterInfo &MRI,
-                                                  MachineIRBuilder &B) const {
-  Register Dst = MI.getOperand(0).getReg();
-  Register Src = MI.getOperand(1).getReg();
-  LLT SrcTy = MRI.getType(Src);
-  TypeSize NumBits = SrcTy.getSizeInBits();
-
-  assert(NumBits < 32u);
-
-  auto ShiftAmt = B.buildConstant(S32, 32u - NumBits);
-  auto Extend = B.buildAnyExt(S32, {Src}).getReg(0u);
-  auto Shift = B.buildLShr(S32, {Extend}, ShiftAmt);
-  auto Ctlz = B.buildInstr(AMDGPU::G_AMDGPU_FFBH_U32, {S32}, {Shift});
-  B.buildTrunc(Dst, Ctlz);
-  MI.eraseFromParent();
-  return true;
-}
-
 // Check that this is a G_XOR x, -1
 static bool isNot(const MachineRegisterInfo &MRI, const MachineInstr &MI) {
   if (MI.getOpcode() != TargetOpcode::G_XOR)
diff --git a/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.h b/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.h
index 4b1d821..e5ba84a 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.h
+++ b/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.h
@@ -108,8 +108,6 @@ public:
   bool legalizeMul(LegalizerHelper &Helper, MachineInstr &MI) const;
   bool legalizeCTLZ_CTTZ(MachineInstr &MI, MachineRegisterInfo &MRI,
                          MachineIRBuilder &B) const;
-  bool legalizeCTLZ_ZERO_UNDEF(MachineInstr &MI, MachineRegisterInfo &MRI,
-                               MachineIRBuilder &B) const;
 
   bool loadInputValue(Register DstReg, MachineIRBuilder &B,
                       const ArgDescriptor *Arg,