[AArch64][GlobalISel] Add isel support for vector G_ICMP and G_ASHR & G_LSR

[aemerson]

Summary

The selection for G_ICMP is unfortunately not currently importable from SDAG due to the use of custom SDNodes. To support this, this selection method has an opcode table which has been generated by a script, indexed by various instruction properties. Ideally in future we will have a GISel native selection patterns that we can write in tablegen to improve on this.

For selection of some types we also need support for G_ASHR and G_LSR which are generated as a result of legalization. This patch also adds support for them, generating the same code as SelectionDAG currently does.

[paquette]

LGTM

llvm/lib/Target/AArch64/AArch64InstructionSelector.cpp

private:
   bool selectVaStartAAPCS(MachineInstr &I, MachineFunction &MF,
                           MachineRegisterInfo &MRI) const;
   bool selectVaStartDarwin(MachineInstr &I, MachineFunction &MF,
                            MachineRegisterInfo &MRI) const;
 
   bool selectCompareBranch(MachineInstr &I, MachineFunction &MF,
                            MachineRegisterInfo &MRI) const;
 
+  bool selectVectorASHR(MachineInstr &I, MachineRegisterInfo &MRI) const;
+  bool selectVectorSHL(MachineInstr &I, MachineRegisterInfo &MRI) const;
+
   // Helper to generate an equivalent of scalar_to_vector into a new register,
   // returned via 'Dst'.
   MachineInstr *emitScalarToVector(unsigned EltSize,
                                    const TargetRegisterClass *DstRC,
                                    unsigned Scalar,
                                    MachineIRBuilder &MIRBuilder) const;
 
   /// Emit a lane insert into \p DstReg, or a new vector register if None is
void collectShuffleMaskIndices(MachineInstr &I, MachineRegisterInfo &MRI,
                                  SmallVectorImpl<int> &Idxs) const;
   bool selectShuffleVector(MachineInstr &I, MachineRegisterInfo &MRI) const;
   bool selectExtractElt(MachineInstr &I, MachineRegisterInfo &MRI) const;
   bool selectConcatVectors(MachineInstr &I, MachineRegisterInfo &MRI) const;
   bool selectSplitVectorUnmerge(MachineInstr &I,
                                 MachineRegisterInfo &MRI) const;
   bool selectIntrinsicWithSideEffects(MachineInstr &I,
                                       MachineRegisterInfo &MRI) const;
+  bool selectVectorICmp(MachineInstr &I, MachineRegisterInfo &MRI) const;
 
   unsigned emitConstantPoolEntry(Constant *CPVal, MachineFunction &MF) const;
   MachineInstr *emitLoadFromConstantPool(Constant *CPVal,
                                          MachineIRBuilder &MIRBuilder) const;
 
   // Emit a vector concat operation.
   MachineInstr *emitVectorConcat(Optional<unsigned> Dst, unsigned Op1,
                                  unsigned Op2,
BuildMI(*I.getParent(), I, I.getDebugLoc(), TII.get(CBOpc))
       .addUse(LHS)
       .addMBB(DestMBB)
       .constrainAllUses(TII, TRI, RBI);
 
   I.eraseFromParent();
   return true;
 }
 
+bool AArch64InstructionSelector::selectVectorSHL(
+    MachineInstr &I, MachineRegisterInfo &MRI) const {
+  assert(I.getOpcode() == TargetOpcode::G_SHL);
+  unsigned DstReg = I.getOperand(0).getReg();
+  const LLT Ty = MRI.getType(DstReg);
+  unsigned Src1Reg = I.getOperand(1).getReg();
+  unsigned Src2Reg = I.getOperand(2).getReg();
+
+  if (!Ty.isVector())
+    return false;
+
+  unsigned Opc = 0;
+  const TargetRegisterClass *RC = nullptr;
+  if (Ty == LLT::vector(4, 32)) {
+    Opc = AArch64::USHLv4i32;
+    RC = &AArch64::FPR128RegClass;
+  } else if (Ty == LLT::vector(2, 32)) {
+    Opc = AArch64::USHLv2i32;
+    RC = &AArch64::FPR64RegClass;
+  } else {
+    LLVM_DEBUG(dbgs() << "Unhandled G_SHL type");
+    return false;
+  }
+
+  MachineIRBuilder MIB(I);
+  auto UShl = MIB.buildInstr(Opc, {DstReg}, {Src1Reg, Src2Reg});
+  constrainSelectedInstRegOperands(*UShl, TII, TRI, RBI);
+  I.eraseFromParent();
+  return true;
+}
+
+bool AArch64InstructionSelector::selectVectorASHR(
+    MachineInstr &I, MachineRegisterInfo &MRI) const {
+  assert(I.getOpcode() == TargetOpcode::G_ASHR);
+  unsigned DstReg = I.getOperand(0).getReg();
+  const LLT Ty = MRI.getType(DstReg);
+  unsigned Src1Reg = I.getOperand(1).getReg();
+  unsigned Src2Reg = I.getOperand(2).getReg();
+
+  if (!Ty.isVector())
+    return false;
+
+  // There is not a shift right register instruction, but the shift left
+  // register instruction takes a signed value, where negative numbers specify a
+  // right shift.
+
+  unsigned Opc = 0;
+  unsigned NegOpc = 0;
+  const TargetRegisterClass *RC = nullptr;
+  if (Ty == LLT::vector(4, 32)) {
+    Opc = AArch64::SSHLv4i32;
+    NegOpc = AArch64::NEGv4i32;
+    RC = &AArch64::FPR128RegClass;
+  } else if (Ty == LLT::vector(2, 32)) {
+    Opc = AArch64::SSHLv2i32;
+    NegOpc = AArch64::NEGv2i32;
+    RC = &AArch64::FPR64RegClass;
+  } else {
+    LLVM_DEBUG(dbgs() << "Unhandled G_ASHR type");
+    return false;
+  }
+
+  MachineIRBuilder MIB(I);
+  auto Neg = MIB.buildInstr(NegOpc, {RC}, {Src2Reg});
+  constrainSelectedInstRegOperands(*Neg, TII, TRI, RBI);
+  auto SShl = MIB.buildInstr(Opc, {DstReg}, {Src1Reg, Neg});
+  constrainSelectedInstRegOperands(*SShl, TII, TRI, RBI);
+  I.eraseFromParent();
+  return true;
+}
+
 bool AArch64InstructionSelector::selectVaStartAAPCS(
     MachineInstr &I, MachineFunction &MF, MachineRegisterInfo &MRI) const {
   return false;
 }
 
 bool AArch64InstructionSelector::selectVaStartDarwin(
     MachineInstr &I, MachineFunction &MF, MachineRegisterInfo &MRI) const {
   AArch64FunctionInfo *FuncInfo = MF.getInfo<AArch64FunctionInfo>();
case TargetOpcode::G_UMULH: {
     // operands to use appropriate classes.
     return constrainSelectedInstRegOperands(I, TII, TRI, RBI);
   }
   case TargetOpcode::G_FADD:
   case TargetOpcode::G_FSUB:
   case TargetOpcode::G_FMUL:
   case TargetOpcode::G_FDIV:
 
-  case TargetOpcode::G_OR:
+  case TargetOpcode::G_ASHR:
+    if (MRI.getType(I.getOperand(0).getReg()).isVector())
+      return selectVectorASHR(I, MRI);
+    LLVM_FALLTHROUGH;
   case TargetOpcode::G_SHL:
+    if (Opcode == TargetOpcode::G_SHL &&
+        MRI.getType(I.getOperand(0).getReg()).isVector())
+      return selectVectorSHL(I, MRI);
+    LLVM_FALLTHROUGH;
+  case TargetOpcode::G_OR:
   case TargetOpcode::G_LSHR:
-  case TargetOpcode::G_ASHR:
   case TargetOpcode::G_GEP: {
     // Reject the various things we don't support yet.
     if (unsupportedBinOp(I, RBI, MRI, TRI))
       return false;
 
     const unsigned OpSize = Ty.getSizeInBits();
 
     const unsigned DefReg = I.getOperand(0).getReg();
case TargetOpcode::G_SELECT: {
 
     constrainSelectedInstRegOperands(TstMI, TII, TRI, RBI);
     constrainSelectedInstRegOperands(CSelMI, TII, TRI, RBI);
 
     I.eraseFromParent();
     return true;
   }
   case TargetOpcode::G_ICMP: {
+    if (Ty.isVector())
+      return selectVectorICmp(I, MRI);
+
     if (Ty != LLT::scalar(32)) {
       LLVM_DEBUG(dbgs() << "G_ICMP result has type: " << Ty
                         << ", expected: " << LLT::scalar(32) << '\n');
       return false;
     }
 
     unsigned CmpOpc = 0;
     unsigned ZReg = 0;
case TargetOpcode::G_INSERT_VECTOR_ELT:
     return selectInsertElt(I, MRI);
   case TargetOpcode::G_CONCAT_VECTORS:
     return selectConcatVectors(I, MRI);
   }
 
   return false;
 }
 
+bool AArch64InstructionSelector::selectVectorICmp(
+    MachineInstr &I, MachineRegisterInfo &MRI) const {
+  unsigned DstReg = I.getOperand(0).getReg();
+  LLT DstTy = MRI.getType(DstReg);
+  unsigned SrcReg = I.getOperand(2).getReg();
+  unsigned Src2Reg = I.getOperand(3).getReg();
+  LLT SrcTy = MRI.getType(SrcReg);
+
+  unsigned SrcEltSize = SrcTy.getElementType().getSizeInBits();
+  unsigned NumElts = DstTy.getNumElements();
+
+  // First index is element size, 0 == 8b, 1 == 16b, 2 == 32b, 3 == 64b
+  // Second index is num elts, 0 == v2, 1 == v4, 2 == v8, 3 == v16
+  // Third index is cc opcode:
+  // 0 == eq
+  // 1 == ugt
+  // 2 == uge
+  // 3 == ult
+  // 4 == ule
+  // 5 == sgt
+  // 6 == sge
+  // 7 == slt
+  // 8 == sle
+  // ne is done by negating 'eq' result.
+
+  // This table below assumes that for some comparisons the operands will be
+  // commuted.
+  // ult op == commute + ugt op
+  // ule op == commute + uge op
+  // slt op == commute + sgt op
+  // sle op == commute + sge op
+  unsigned PredIdx = 0;
+  bool SwapOperands = false;
+  CmpInst::Predicate Pred = (CmpInst::Predicate)I.getOperand(1).getPredicate();
+  switch (Pred) {
+  case CmpInst::ICMP_EQ:
+    PredIdx = 0;
+    break;
+  case CmpInst::ICMP_UGT:
+    PredIdx = 1;
+    break;
+  case CmpInst::ICMP_UGE:
+    PredIdx = 2;
+    break;
+  case CmpInst::ICMP_ULT:
+    PredIdx = 3;
+    SwapOperands = true;
+    break;
+  case CmpInst::ICMP_ULE:
+    PredIdx = 4;
+    SwapOperands = true;
+    break;
+  case CmpInst::ICMP_SGT:
+    PredIdx = 5;
+    break;
+  case CmpInst::ICMP_SGE:
+    PredIdx = 6;
+    break;
+  case CmpInst::ICMP_SLT:
+    PredIdx = 7;
+    SwapOperands = true;
+    break;
+  case CmpInst::ICMP_SLE:
+    PredIdx = 8;
+    SwapOperands = true;
+    break;
+  default:
+    llvm_unreachable("Unhandled icmp predicate");
+    return false;
+  }
+
+  // This table obviously should be tablegen'd when we have our GISel native
+  // tablegen selector.
+
+  static const unsigned OpcTable[4][4][9] = {
+      {
+          {0 /* invalid */, 0 /* invalid */, 0 /* invalid */, 0 /* invalid */,
+           0 /* invalid */, 0 /* invalid */, 0 /* invalid */, 0 /* invalid */,
+           0 /* invalid */},
+          {0 /* invalid */, 0 /* invalid */, 0 /* invalid */, 0 /* invalid */,
+           0 /* invalid */, 0 /* invalid */, 0 /* invalid */, 0 /* invalid */,
+           0 /* invalid */},
+          {AArch64::CMEQv8i8, AArch64::CMHIv8i8, AArch64::CMHSv8i8,
+           AArch64::CMHIv8i8, AArch64::CMHSv8i8, AArch64::CMGTv8i8,
+           AArch64::CMGEv8i8, AArch64::CMGTv8i8, AArch64::CMGEv8i8},
+          {AArch64::CMEQv16i8, AArch64::CMHIv16i8, AArch64::CMHSv16i8,
+           AArch64::CMHIv16i8, AArch64::CMHSv16i8, AArch64::CMGTv16i8,
+           AArch64::CMGEv16i8, AArch64::CMGTv16i8, AArch64::CMGEv16i8}
+      },
+      {
+          {0 /* invalid */, 0 /* invalid */, 0 /* invalid */, 0 /* invalid */,
+           0 /* invalid */, 0 /* invalid */, 0 /* invalid */, 0 /* invalid */,
+           0 /* invalid */},
+          {AArch64::CMEQv4i16, AArch64::CMHIv4i16, AArch64::CMHSv4i16,
+           AArch64::CMHIv4i16, AArch64::CMHSv4i16, AArch64::CMGTv4i16,
+           AArch64::CMGEv4i16, AArch64::CMGTv4i16, AArch64::CMGEv4i16},
+          {AArch64::CMEQv8i16, AArch64::CMHIv8i16, AArch64::CMHSv8i16,
+           AArch64::CMHIv8i16, AArch64::CMHSv8i16, AArch64::CMGTv8i16,
+           AArch64::CMGEv8i16, AArch64::CMGTv8i16, AArch64::CMGEv8i16},
+          {0 /* invalid */, 0 /* invalid */, 0 /* invalid */, 0 /* invalid */,
+           0 /* invalid */, 0 /* invalid */, 0 /* invalid */, 0 /* invalid */,
+           0 /* invalid */}
+      },
+      {
+          {AArch64::CMEQv2i32, AArch64::CMHIv2i32, AArch64::CMHSv2i32,
+           AArch64::CMHIv2i32, AArch64::CMHSv2i32, AArch64::CMGTv2i32,
+           AArch64::CMGEv2i32, AArch64::CMGTv2i32, AArch64::CMGEv2i32},
+          {AArch64::CMEQv4i32, AArch64::CMHIv4i32, AArch64::CMHSv4i32,
+           AArch64::CMHIv4i32, AArch64::CMHSv4i32, AArch64::CMGTv4i32,
+           AArch64::CMGEv4i32, AArch64::CMGTv4i32, AArch64::CMGEv4i32},
+          {0 /* invalid */, 0 /* invalid */, 0 /* invalid */, 0 /* invalid */,
+           0 /* invalid */, 0 /* invalid */, 0 /* invalid */, 0 /* invalid */,
+           0 /* invalid */},
+          {0 /* invalid */, 0 /* invalid */, 0 /* invalid */, 0 /* invalid */,
+           0 /* invalid */, 0 /* invalid */, 0 /* invalid */, 0 /* invalid */,
+           0 /* invalid */}
+      },
+      {
+          {AArch64::CMEQv2i64, AArch64::CMHIv2i64, AArch64::CMHSv2i64,
+           AArch64::CMHIv2i64, AArch64::CMHSv2i64, AArch64::CMGTv2i64,
+           AArch64::CMGEv2i64, AArch64::CMGTv2i64, AArch64::CMGEv2i64},
+          {0 /* invalid */, 0 /* invalid */, 0 /* invalid */, 0 /* invalid */,
+           0 /* invalid */, 0 /* invalid */, 0 /* invalid */, 0 /* invalid */,
+           0 /* invalid */},
+          {0 /* invalid */, 0 /* invalid */, 0 /* invalid */, 0 /* invalid */,
+           0 /* invalid */, 0 /* invalid */, 0 /* invalid */, 0 /* invalid */,
+           0 /* invalid */},
+          {0 /* invalid */, 0 /* invalid */, 0 /* invalid */, 0 /* invalid */,
+           0 /* invalid */, 0 /* invalid */, 0 /* invalid */, 0 /* invalid */,
+           0 /* invalid */}
+      },
+  };
+  unsigned EltIdx = Log2_32(SrcEltSize / 8);
+  unsigned NumEltsIdx = Log2_32(NumElts / 2);
+  unsigned Opc = OpcTable[EltIdx][NumEltsIdx][PredIdx];
+  if (!Opc) {
+    LLVM_DEBUG(dbgs() << "Could not map G_ICMP to cmp opcode");
+    return false;
+  }
+
+  const RegisterBank &VecRB = *RBI.getRegBank(SrcReg, MRI, TRI);
+  const TargetRegisterClass *SrcRC =
+      getRegClassForTypeOnBank(SrcTy, VecRB, RBI, true);
+  if (!SrcRC) {
+    LLVM_DEBUG(dbgs() << "Could not determine source register class.\n");
+    return false;
+  }
+
+  unsigned NotOpc = Pred == ICmpInst::ICMP_NE ? AArch64::NOTv8i8 : 0;
+  if (SrcTy.getSizeInBits() == 128)
+    NotOpc = NotOpc ? AArch64::NOTv16i8 : 0;
+
+  if (SwapOperands)
+    std::swap(SrcReg, Src2Reg);
+
+  MachineIRBuilder MIB(I);
+  auto Cmp = MIB.buildInstr(Opc, {SrcRC}, {SrcReg, Src2Reg});
+  constrainSelectedInstRegOperands(*Cmp, TII, TRI, RBI);
+
+  // Invert if we had a 'ne' cc.
+  if (NotOpc) {
+    Cmp = MIB.buildInstr(NotOpc, {DstReg}, {Cmp});
+    constrainSelectedInstRegOperands(*Cmp, TII, TRI, RBI);
+  } else {
+    MIB.buildCopy(DstReg, Cmp.getReg(0));
+  }
+  RBI.constrainGenericRegister(DstReg, *SrcRC, MRI);
+  I.eraseFromParent();
+  return true;
+}
+
 MachineInstr *AArch64InstructionSelector::emitScalarToVector(
     unsigned EltSize, const TargetRegisterClass *DstRC, unsigned Scalar,
     MachineIRBuilder &MIRBuilder) const {
   auto Undef = MIRBuilder.buildInstr(TargetOpcode::IMPLICIT_DEF, {DstRC}, {});
 
   auto BuildFn = [&](unsigned SubregIndex) {
     auto Ins =
         MIRBuilder

llvm/test/CodeGen/AArch64/GlobalISel/select-vector-icmp.mir

+# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
+# RUN: llc -mtriple=aarch64-- -run-pass=instruction-select -verify-machineinstrs %s -o - | FileCheck %s
+
+--- |
+  target datalayout = "e-m:e-i8:8:32-i16:16:32-i64:64-i128:128-n32:64-S128"
+  target triple = "aarch64"
+
+  define <2 x i1> @test_v2i64_eq(<2 x i64> %v1, <2 x i64> %v2) {
+    %cmp = icmp eq <2 x i64> %v1, %v2
+    ret <2 x i1> %cmp
+  }
+
+  define <4 x i1> @test_v4i32_eq(<4 x i32> %v1, <4 x i32> %v2) {
+    %cmp = icmp eq <4 x i32> %v1, %v2
+    ret <4 x i1> %cmp
+  }
+
+  define <2 x i1> @test_v2i32_eq(<2 x i32> %v1, <2 x i32> %v2) {
+    %cmp = icmp eq <2 x i32> %v1, %v2
+    ret <2 x i1> %cmp
+  }
+
+  define <2 x i1> @test_v2i16_eq(<2 x i16> %v1, <2 x i16> %v2) {
+    %cmp = icmp eq <2 x i16> %v1, %v2
+    ret <2 x i1> %cmp
+  }
+
+  define <8 x i1> @test_v8i16_eq(<8 x i16> %v1, <8 x i16> %v2) {
+    %cmp = icmp eq <8 x i16> %v1, %v2
+    ret <8 x i1> %cmp
+  }
+
+  define <4 x i1> @test_v4i16_eq(<4 x i16> %v1, <4 x i16> %v2) {
+    %cmp = icmp eq <4 x i16> %v1, %v2
+    ret <4 x i1> %cmp
+  }
+
+  define <16 x i1> @test_v16i8_eq(<16 x i8> %v1, <16 x i8> %v2) {
+    %cmp = icmp eq <16 x i8> %v1, %v2
+    ret <16 x i1> %cmp
+  }
+
+  define <8 x i1> @test_v8i8_eq(<8 x i8> %v1, <8 x i8> %v2) {
+    %cmp = icmp eq <8 x i8> %v1, %v2
+    ret <8 x i1> %cmp
+  }
+
+  define <2 x i1> @test_v2i64_ugt(<2 x i64> %v1, <2 x i64> %v2) {
+    %cmp = icmp ugt <2 x i64> %v1, %v2
+    ret <2 x i1> %cmp
+  }
+
+  define <4 x i1> @test_v4i32_ugt(<4 x i32> %v1, <4 x i32> %v2) {
+    %cmp = icmp ugt <4 x i32> %v1, %v2
+    ret <4 x i1> %cmp
+  }
+
+  define <2 x i1> @test_v2i32_ugt(<2 x i32> %v1, <2 x i32> %v2) {
+    %cmp = icmp ugt <2 x i32> %v1, %v2
+    ret <2 x i1> %cmp
+  }
+
+  define <2 x i1> @test_v2i16_ugt(<2 x i16> %v1, <2 x i16> %v2) {
+    %cmp = icmp ugt <2 x i16> %v1, %v2
+    ret <2 x i1> %cmp
+  }
+
+  define <8 x i1> @test_v8i16_ugt(<8 x i16> %v1, <8 x i16> %v2) {
+    %cmp = icmp ugt <8 x i16> %v1, %v2
+    ret <8 x i1> %cmp
+  }
+
+  define <4 x i1> @test_v4i16_ugt(<4 x i16> %v1, <4 x i16> %v2) {
+    %cmp = icmp ugt <4 x i16> %v1, %v2
+    ret <4 x i1> %cmp
+  }
+
+  define <16 x i1> @test_v16i8_ugt(<16 x i8> %v1, <16 x i8> %v2) {
+    %cmp = icmp ugt <16 x i8> %v1, %v2
+    ret <16 x i1> %cmp
+  }
+
+  define <8 x i1> @test_v8i8_ugt(<8 x i8> %v1, <8 x i8> %v2) {
+    %cmp = icmp ugt <8 x i8> %v1, %v2
+    ret <8 x i1> %cmp
+  }
+
+  define <2 x i1> @test_v2i64_uge(<2 x i64> %v1, <2 x i64> %v2) {
+    %cmp = icmp uge <2 x i64> %v1, %v2
+    ret <2 x i1> %cmp
+  }
+
+  define <4 x i1> @test_v4i32_uge(<4 x i32> %v1, <4 x i32> %v2) {
+    %cmp = icmp uge <4 x i32> %v1, %v2
+    ret <4 x i1> %cmp
+  }
+
+  define <2 x i1> @test_v2i32_uge(<2 x i32> %v1, <2 x i32> %v2) {
+    %cmp = icmp uge <2 x i32> %v1, %v2
+    ret <2 x i1> %cmp
+  }
+
+  define <2 x i1> @test_v2i16_uge(<2 x i16> %v1, <2 x i16> %v2) {
+    %cmp = icmp uge <2 x i16> %v1, %v2
+    ret <2 x i1> %cmp
+  }
+
+  define <8 x i1> @test_v8i16_uge(<8 x i16> %v1, <8 x i16> %v2) {
+    %cmp = icmp uge <8 x i16> %v1, %v2
+    ret <8 x i1> %cmp
+  }
+
+  define <4 x i1> @test_v4i16_uge(<4 x i16> %v1, <4 x i16> %v2) {
+    %cmp = icmp uge <4 x i16> %v1, %v2
+    ret <4 x i1> %cmp
+  }
+
+  define <16 x i1> @test_v16i8_uge(<16 x i8> %v1, <16 x i8> %v2) {
+    %cmp = icmp uge <16 x i8> %v1, %v2
+    ret <16 x i1> %cmp
+  }
+
+  define <8 x i1> @test_v8i8_uge(<8 x i8> %v1, <8 x i8> %v2) {
+    %cmp = icmp uge <8 x i8> %v1, %v2
+    ret <8 x i1> %cmp
+  }
+
+  define <2 x i1> @test_v2i64_ult(<2 x i64> %v1, <2 x i64> %v2) {
+    %cmp = icmp ult <2 x i64> %v1, %v2
+    ret <2 x i1> %cmp
+  }
+
+  define <4 x i1> @test_v4i32_ult(<4 x i32> %v1, <4 x i32> %v2) {
+    %cmp = icmp ult <4 x i32> %v1, %v2
+    ret <4 x i1> %cmp
+  }
+
+  define <2 x i1> @test_v2i32_ult(<2 x i32> %v1, <2 x i32> %v2) {
+    %cmp = icmp ult <2 x i32> %v1, %v2
+    ret <2 x i1> %cmp
+  }
+
+  define <2 x i1> @test_v2i16_ult(<2 x i16> %v1, <2 x i16> %v2) {
+    %cmp = icmp ult <2 x i16> %v1, %v2
+    ret <2 x i1> %cmp
+  }
+
+  define <8 x i1> @test_v8i16_ult(<8 x i16> %v1, <8 x i16> %v2) {
+    %cmp = icmp ult <8 x i16> %v1, %v2
+    ret <8 x i1> %cmp
+  }
+
+  define <4 x i1> @test_v4i16_ult(<4 x i16> %v1, <4 x i16> %v2) {
+    %cmp = icmp ult <4 x i16> %v1, %v2
+    ret <4 x i1> %cmp
+  }
+
+  define <16 x i1> @test_v16i8_ult(<16 x i8> %v1, <16 x i8> %v2) {
+    %cmp = icmp ult <16 x i8> %v1, %v2
+    ret <16 x i1> %cmp
+  }
+
+  define <8 x i1> @test_v8i8_ult(<8 x i8> %v1, <8 x i8> %v2) {
+    %cmp = icmp ult <8 x i8> %v1, %v2
+    ret <8 x i1> %cmp
+  }
+
+  define <2 x i1> @test_v2i64_ule(<2 x i64> %v1, <2 x i64> %v2) {
+    %cmp = icmp ule <2 x i64> %v1, %v2
+    ret <2 x i1> %cmp
+  }
+
+  define <4 x i1> @test_v4i32_ule(<4 x i32> %v1, <4 x i32> %v2) {
+    %cmp = icmp ule <4 x i32> %v1, %v2
+    ret <4 x i1> %cmp
+  }
+
+  define <2 x i1> @test_v2i32_ule(<2 x i32> %v1, <2 x i32> %v2) {
+    %cmp = icmp ule <2 x i32> %v1, %v2
+    ret <2 x i1> %cmp
+  }
+
+  define <2 x i1> @test_v2i16_ule(<2 x i16> %v1, <2 x i16> %v2) {
+    %cmp = icmp ule <2 x i16> %v1, %v2
+    ret <2 x i1> %cmp
+  }
+
+  define <8 x i1> @test_v8i16_ule(<8 x i16> %v1, <8 x i16> %v2) {
+    %cmp = icmp ule <8 x i16> %v1, %v2
+    ret <8 x i1> %cmp
+  }
+
+  define <4 x i1> @test_v4i16_ule(<4 x i16> %v1, <4 x i16> %v2) {
+    %cmp = icmp ule <4 x i16> %v1, %v2
+    ret <4 x i1> %cmp
+  }
+
+  define <16 x i1> @test_v16i8_ule(<16 x i8> %v1, <16 x i8> %v2) {
+    %cmp = icmp ule <16 x i8> %v1, %v2
+    ret <16 x i1> %cmp
+  }
+
+  define <8 x i1> @test_v8i8_ule(<8 x i8> %v1, <8 x i8> %v2) {
+    %cmp = icmp ule <8 x i8> %v1, %v2
+    ret <8 x i1> %cmp
+  }
+
+  define <2 x i1> @test_v2i64_sgt(<2 x i64> %v1, <2 x i64> %v2) {
+    %cmp = icmp sgt <2 x i64> %v1, %v2
+    ret <2 x i1> %cmp
+  }
+
+  define <4 x i1> @test_v4i32_sgt(<4 x i32> %v1, <4 x i32> %v2) {
+    %cmp = icmp sgt <4 x i32> %v1, %v2
+    ret <4 x i1> %cmp
+  }
+
+  define <2 x i1> @test_v2i32_sgt(<2 x i32> %v1, <2 x i32> %v2) {
+    %cmp = icmp sgt <2 x i32> %v1, %v2
+    ret <2 x i1> %cmp
+  }
+
+  define <2 x i1> @test_v2i16_sgt(<2 x i16> %v1, <2 x i16> %v2) {
+    %cmp = icmp sgt <2 x i16> %v1, %v2
+    ret <2 x i1> %cmp
+  }
+
+  define <8 x i1> @test_v8i16_sgt(<8 x i16> %v1, <8 x i16> %v2) {
+    %cmp = icmp sgt <8 x i16> %v1, %v2
+    ret <8 x i1> %cmp
+  }
+
+  define <4 x i1> @test_v4i16_sgt(<4 x i16> %v1, <4 x i16> %v2) {
+    %cmp = icmp sgt <4 x i16> %v1, %v2
+    ret <4 x i1> %cmp
+  }
+
+  define <16 x i1> @test_v16i8_sgt(<16 x i8> %v1, <16 x i8> %v2) {
+    %cmp = icmp sgt <16 x i8> %v1, %v2
+    ret <16 x i1> %cmp
+  }
+
+  define <8 x i1> @test_v8i8_sgt(<8 x i8> %v1, <8 x i8> %v2) {
+    %cmp = icmp sgt <8 x i8> %v1, %v2
+    ret <8 x i1> %cmp
+  }
+
+  define <2 x i1> @test_v2i64_sge(<2 x i64> %v1, <2 x i64> %v2) {
+    %cmp = icmp sge <2 x i64> %v1, %v2
+    ret <2 x i1> %cmp
+  }
+
+  define <4 x i1> @test_v4i32_sge(<4 x i32> %v1, <4 x i32> %v2) {
+    %cmp = icmp sge <4 x i32> %v1, %v2
+    ret <4 x i1> %cmp
+  }
+
+  define <2 x i1> @test_v2i32_sge(<2 x i32> %v1, <2 x i32> %v2) {
+    %cmp = icmp sge <2 x i32> %v1, %v2
+    ret <2 x i1> %cmp
+  }
+
+  define <2 x i1> @test_v2i16_sge(<2 x i16> %v1, <2 x i16> %v2) {
+    %cmp = icmp sge <2 x i16> %v1, %v2
+    ret <2 x i1> %cmp
+  }
+
+  define <8 x i1> @test_v8i16_sge(<8 x i16> %v1, <8 x i16> %v2) {
+    %cmp = icmp sge <8 x i16> %v1, %v2
+    ret <8 x i1> %cmp
+  }
+
+  define <4 x i1> @test_v4i16_sge(<4 x i16> %v1, <4 x i16> %v2) {
+    %cmp = icmp sge <4 x i16> %v1, %v2
+    ret <4 x i1> %cmp
+  }
+
+  define <16 x i1> @test_v16i8_sge(<16 x i8> %v1, <16 x i8> %v2) {
+    %cmp = icmp sge <16 x i8> %v1, %v2
+    ret <16 x i1> %cmp
+  }
+
+  define <8 x i1> @test_v8i8_sge(<8 x i8> %v1, <8 x i8> %v2) {
+    %cmp = icmp sge <8 x i8> %v1, %v2
+    ret <8 x i1> %cmp
+  }
+
+  define <2 x i1> @test_v2i64_slt(<2 x i64> %v1, <2 x i64> %v2) {
+    %cmp = icmp slt <2 x i64> %v1, %v2
+    ret <2 x i1> %cmp
+  }
+
+  define <4 x i1> @test_v4i32_slt(<4 x i32> %v1, <4 x i32> %v2) {
+    %cmp = icmp slt <4 x i32> %v1, %v2
+    ret <4 x i1> %cmp
+  }
+
+  define <2 x i1> @test_v2i32_slt(<2 x i32> %v1, <2 x i32> %v2) {
+    %cmp = icmp slt <2 x i32> %v1, %v2
+    ret <2 x i1> %cmp
+  }
+
+  define <2 x i1> @test_v2i16_slt(<2 x i16> %v1, <2 x i16> %v2) {
+    %cmp = icmp slt <2 x i16> %v1, %v2
+    ret <2 x i1> %cmp
+  }
+
+  define <8 x i1> @test_v8i16_slt(<8 x i16> %v1, <8 x i16> %v2) {
+    %cmp = icmp slt <8 x i16> %v1, %v2
+    ret <8 x i1> %cmp
+  }
+
+  define <4 x i1> @test_v4i16_slt(<4 x i16> %v1, <4 x i16> %v2) {
+    %cmp = icmp slt <4 x i16> %v1, %v2
+    ret <4 x i1> %cmp
+  }
+
+  define <16 x i1> @test_v16i8_slt(<16 x i8> %v1, <16 x i8> %v2) {
+    %cmp = icmp slt <16 x i8> %v1, %v2
+    ret <16 x i1> %cmp
+  }
+
+  define <8 x i1> @test_v8i8_slt(<8 x i8> %v1, <8 x i8> %v2) {
+    %cmp = icmp slt <8 x i8> %v1, %v2
+    ret <8 x i1> %cmp
+  }
+
+  define <2 x i1> @test_v2i64_sle(<2 x i64> %v1, <2 x i64> %v2) {
+    %cmp = icmp sle <2 x i64> %v1, %v2
+    ret <2 x i1> %cmp
+  }
+
+  define <4 x i1> @test_v4i32_sle(<4 x i32> %v1, <4 x i32> %v2) {
+    %cmp = icmp sle <4 x i32> %v1, %v2
+    ret <4 x i1> %cmp
+  }
+
+  define <2 x i1> @test_v2i32_sle(<2 x i32> %v1, <2 x i32> %v2) {
+    %cmp = icmp sle <2 x i32> %v1, %v2
+    ret <2 x i1> %cmp
+  }
+
+  define <2 x i1> @test_v2i16_sle(<2 x i16> %v1, <2 x i16> %v2) {
+    %cmp = icmp sle <2 x i16> %v1, %v2
+    ret <2 x i1> %cmp
+  }
+
+  define <8 x i1> @test_v8i16_sle(<8 x i16> %v1, <8 x i16> %v2) {
+    %cmp = icmp sle <8 x i16> %v1, %v2
+    ret <8 x i1> %cmp
+  }
+
+  define <4 x i1> @test_v4i16_sle(<4 x i16> %v1, <4 x i16> %v2) {
+    %cmp = icmp sle <4 x i16> %v1, %v2
+    ret <4 x i1> %cmp
+  }
+
+  define <16 x i1> @test_v16i8_sle(<16 x i8> %v1, <16 x i8> %v2) {
+    %cmp = icmp sle <16 x i8> %v1, %v2
+    ret <16 x i1> %cmp
+  }
+
+  define <8 x i1> @test_v8i8_sle(<8 x i8> %v1, <8 x i8> %v2) {
+    %cmp = icmp sle <8 x i8> %v1, %v2
+    ret <8 x i1> %cmp
+  }
+
+...
+---
+name:            test_v2i64_eq
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: fpr }
+  - { id: 1, class: fpr }
+  - { id: 2, class: _ }
+  - { id: 3, class: fpr }
+  - { id: 4, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $q0, $q1
+
+    ; CHECK-LABEL: name: test_v2i64_eq
+    ; CHECK: liveins: $q0, $q1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr128 = COPY $q0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr128 = COPY $q1
+    ; CHECK: [[CMEQv2i64_:%[0-9]+]]:fpr128 = CMEQv2i64 [[COPY]], [[COPY1]]
+    ; CHECK: [[XTNv2i32_:%[0-9]+]]:fpr64 = XTNv2i32 [[CMEQv2i64_]]
+    ; CHECK: $d0 = COPY [[XTNv2i32_]]
+    ; CHECK: RET_ReallyLR implicit $d0
+    %0:fpr(<2 x s64>) = COPY $q0
+    %1:fpr(<2 x s64>) = COPY $q1
+    %4:fpr(<2 x s64>) = G_ICMP intpred(eq), %0(<2 x s64>), %1
+    %3:fpr(<2 x s32>) = G_TRUNC %4(<2 x s64>)
+    $d0 = COPY %3(<2 x s32>)
+    RET_ReallyLR implicit $d0
+
+...
+---
+name:            test_v4i32_eq
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: fpr }
+  - { id: 1, class: fpr }
+  - { id: 2, class: _ }
+  - { id: 3, class: fpr }
+  - { id: 4, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $q0, $q1
+
+    ; CHECK-LABEL: name: test_v4i32_eq
+    ; CHECK: liveins: $q0, $q1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr128 = COPY $q0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr128 = COPY $q1
+    ; CHECK: [[CMEQv4i32_:%[0-9]+]]:fpr128 = CMEQv4i32 [[COPY]], [[COPY1]]
+    ; CHECK: [[XTNv4i16_:%[0-9]+]]:fpr64 = XTNv4i16 [[CMEQv4i32_]]
+    ; CHECK: $d0 = COPY [[XTNv4i16_]]
+    ; CHECK: RET_ReallyLR implicit $d0
+    %0:fpr(<4 x s32>) = COPY $q0
+    %1:fpr(<4 x s32>) = COPY $q1
+    %4:fpr(<4 x s32>) = G_ICMP intpred(eq), %0(<4 x s32>), %1
+    %3:fpr(<4 x s16>) = G_TRUNC %4(<4 x s32>)
+    $d0 = COPY %3(<4 x s16>)
+    RET_ReallyLR implicit $d0
+
+...
+---
+name:            test_v2i32_eq
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: fpr }
+  - { id: 1, class: fpr }
+  - { id: 2, class: _ }
+  - { id: 3, class: fpr }
+  - { id: 4, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $d0, $d1
+
+    ; CHECK-LABEL: name: test_v2i32_eq
+    ; CHECK: liveins: $d0, $d1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr64 = COPY $d0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr64 = COPY $d1
+    ; CHECK: [[CMEQv2i32_:%[0-9]+]]:fpr64 = CMEQv2i32 [[COPY]], [[COPY1]]
+    ; CHECK: $d0 = COPY [[CMEQv2i32_]]
+    ; CHECK: RET_ReallyLR implicit $d0
+    %0:fpr(<2 x s32>) = COPY $d0
+    %1:fpr(<2 x s32>) = COPY $d1
+    %4:fpr(<2 x s32>) = G_ICMP intpred(eq), %0(<2 x s32>), %1
+    %3:fpr(<2 x s32>) = COPY %4(<2 x s32>)
+    $d0 = COPY %3(<2 x s32>)
+    RET_ReallyLR implicit $d0
+
+...
+---
+name:            test_v2i16_eq
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: _ }
+  - { id: 1, class: _ }
+  - { id: 2, class: fpr }
+  - { id: 3, class: fpr }
+  - { id: 4, class: _ }
+  - { id: 5, class: fpr }
+  - { id: 6, class: _ }
+  - { id: 7, class: fpr }
+  - { id: 8, class: fpr }
+  - { id: 9, class: fpr }
+  - { id: 10, class: gpr }
+  - { id: 11, class: fpr }
+  - { id: 12, class: fpr }
+  - { id: 13, class: gpr }
+  - { id: 14, class: fpr }
+  - { id: 15, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $d0, $d1
+
+    ; CHECK-LABEL: name: test_v2i16_eq
+    ; CHECK: liveins: $d0, $d1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr64 = COPY $d0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr64 = COPY $d1
+    ; CHECK: [[MOVi32imm:%[0-9]+]]:gpr32 = MOVi32imm 65535
+    ; CHECK: [[DEF:%[0-9]+]]:fpr128 = IMPLICIT_DEF
+    ; CHECK: [[INSERT_SUBREG:%[0-9]+]]:fpr128 = INSERT_SUBREG [[DEF]], [[MOVi32imm]], %subreg.ssub
+    ; CHECK: [[INSvi32gpr:%[0-9]+]]:fpr128 = INSvi32gpr [[INSERT_SUBREG]], 1, [[MOVi32imm]]
+    ; CHECK: [[COPY2:%[0-9]+]]:fpr64 = COPY [[INSvi32gpr]].dsub
+    ; CHECK: [[ANDv8i8_:%[0-9]+]]:fpr64 = ANDv8i8 [[COPY]], [[COPY2]]
+    ; CHECK: [[MOVi32imm1:%[0-9]+]]:gpr32 = MOVi32imm 65535
+    ; CHECK: [[DEF1:%[0-9]+]]:fpr128 = IMPLICIT_DEF
+    ; CHECK: [[INSERT_SUBREG1:%[0-9]+]]:fpr128 = INSERT_SUBREG [[DEF1]], [[MOVi32imm1]], %subreg.ssub
+    ; CHECK: [[INSvi32gpr1:%[0-9]+]]:fpr128 = INSvi32gpr [[INSERT_SUBREG1]], 1, [[MOVi32imm1]]
+    ; CHECK: [[COPY3:%[0-9]+]]:fpr64 = COPY [[INSvi32gpr1]].dsub
+    ; CHECK: [[ANDv8i8_1:%[0-9]+]]:fpr64 = ANDv8i8 [[COPY1]], [[COPY3]]
+    ; CHECK: [[CMEQv2i32_:%[0-9]+]]:fpr64 = CMEQv2i32 [[ANDv8i8_]], [[ANDv8i8_1]]
+    ; CHECK: $d0 = COPY [[CMEQv2i32_]]
+    ; CHECK: RET_ReallyLR implicit $d0
+    %2:fpr(<2 x s32>) = COPY $d0
+    %3:fpr(<2 x s32>) = COPY $d1
+    %13:gpr(s32) = G_CONSTANT i32 65535
+    %14:fpr(<2 x s32>) = G_BUILD_VECTOR %13(s32), %13(s32)
+    %15:fpr(<2 x s32>) = COPY %2(<2 x s32>)
+    %7:fpr(<2 x s32>) = G_AND %15, %14
+    %10:gpr(s32) = G_CONSTANT i32 65535
+    %11:fpr(<2 x s32>) = G_BUILD_VECTOR %10(s32), %10(s32)
+    %12:fpr(<2 x s32>) = COPY %3(<2 x s32>)
+    %8:fpr(<2 x s32>) = G_AND %12, %11
+    %9:fpr(<2 x s32>) = G_ICMP intpred(eq), %7(<2 x s32>), %8
+    %5:fpr(<2 x s32>) = COPY %9(<2 x s32>)
+    $d0 = COPY %5(<2 x s32>)
+    RET_ReallyLR implicit $d0
+
+...
+---
+name:            test_v8i16_eq
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: fpr }
+  - { id: 1, class: fpr }
+  - { id: 2, class: _ }
+  - { id: 3, class: fpr }
+  - { id: 4, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $q0, $q1
+
+    ; CHECK-LABEL: name: test_v8i16_eq
+    ; CHECK: liveins: $q0, $q1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr128 = COPY $q0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr128 = COPY $q1
+    ; CHECK: [[CMEQv8i16_:%[0-9]+]]:fpr128 = CMEQv8i16 [[COPY]], [[COPY1]]
+    ; CHECK: [[XTNv8i8_:%[0-9]+]]:fpr64 = XTNv8i8 [[CMEQv8i16_]]
+    ; CHECK: $d0 = COPY [[XTNv8i8_]]
+    ; CHECK: RET_ReallyLR implicit $d0
+    %0:fpr(<8 x s16>) = COPY $q0
+    %1:fpr(<8 x s16>) = COPY $q1
+    %4:fpr(<8 x s16>) = G_ICMP intpred(eq), %0(<8 x s16>), %1
+    %3:fpr(<8 x s8>) = G_TRUNC %4(<8 x s16>)
+    $d0 = COPY %3(<8 x s8>)
+    RET_ReallyLR implicit $d0
+
+...
+---
+name:            test_v4i16_eq
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: fpr }
+  - { id: 1, class: fpr }
+  - { id: 2, class: _ }
+  - { id: 3, class: fpr }
+  - { id: 4, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $d0, $d1
+
+    ; CHECK-LABEL: name: test_v4i16_eq
+    ; CHECK: liveins: $d0, $d1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr64 = COPY $d0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr64 = COPY $d1
+    ; CHECK: [[CMEQv4i16_:%[0-9]+]]:fpr64 = CMEQv4i16 [[COPY]], [[COPY1]]
+    ; CHECK: $d0 = COPY [[CMEQv4i16_]]
+    ; CHECK: RET_ReallyLR implicit $d0
+    %0:fpr(<4 x s16>) = COPY $d0
+    %1:fpr(<4 x s16>) = COPY $d1
+    %4:fpr(<4 x s16>) = G_ICMP intpred(eq), %0(<4 x s16>), %1
+    %3:fpr(<4 x s16>) = COPY %4(<4 x s16>)
+    $d0 = COPY %3(<4 x s16>)
+    RET_ReallyLR implicit $d0
+
+...
+---
+name:            test_v16i8_eq
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: fpr }
+  - { id: 1, class: fpr }
+  - { id: 2, class: _ }
+  - { id: 3, class: fpr }
+  - { id: 4, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $q0, $q1
+
+    ; CHECK-LABEL: name: test_v16i8_eq
+    ; CHECK: liveins: $q0, $q1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr128 = COPY $q0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr128 = COPY $q1
+    ; CHECK: [[CMEQv16i8_:%[0-9]+]]:fpr128 = CMEQv16i8 [[COPY]], [[COPY1]]
+    ; CHECK: $q0 = COPY [[CMEQv16i8_]]
+    ; CHECK: RET_ReallyLR implicit $q0
+    %0:fpr(<16 x s8>) = COPY $q0
+    %1:fpr(<16 x s8>) = COPY $q1
+    %4:fpr(<16 x s8>) = G_ICMP intpred(eq), %0(<16 x s8>), %1
+    %3:fpr(<16 x s8>) = COPY %4(<16 x s8>)
+    $q0 = COPY %3(<16 x s8>)
+    RET_ReallyLR implicit $q0
+
+...
+---
+name:            test_v8i8_eq
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: fpr }
+  - { id: 1, class: fpr }
+  - { id: 2, class: _ }
+  - { id: 3, class: fpr }
+  - { id: 4, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $d0, $d1
+
+    ; CHECK-LABEL: name: test_v8i8_eq
+    ; CHECK: liveins: $d0, $d1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr64 = COPY $d0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr64 = COPY $d1
+    ; CHECK: [[CMEQv8i8_:%[0-9]+]]:fpr64 = CMEQv8i8 [[COPY]], [[COPY1]]
+    ; CHECK: $d0 = COPY [[CMEQv8i8_]]
+    ; CHECK: RET_ReallyLR implicit $d0
+    %0:fpr(<8 x s8>) = COPY $d0
+    %1:fpr(<8 x s8>) = COPY $d1
+    %4:fpr(<8 x s8>) = G_ICMP intpred(eq), %0(<8 x s8>), %1
+    %3:fpr(<8 x s8>) = COPY %4(<8 x s8>)
+    $d0 = COPY %3(<8 x s8>)
+    RET_ReallyLR implicit $d0
+
+...
+---
+name:            test_v2i64_ugt
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: fpr }
+  - { id: 1, class: fpr }
+  - { id: 2, class: _ }
+  - { id: 3, class: fpr }
+  - { id: 4, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $q0, $q1
+
+    ; CHECK-LABEL: name: test_v2i64_ugt
+    ; CHECK: liveins: $q0, $q1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr128 = COPY $q0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr128 = COPY $q1
+    ; CHECK: [[CMHIv2i64_:%[0-9]+]]:fpr128 = CMHIv2i64 [[COPY]], [[COPY1]]
+    ; CHECK: [[XTNv2i32_:%[0-9]+]]:fpr64 = XTNv2i32 [[CMHIv2i64_]]
+    ; CHECK: $d0 = COPY [[XTNv2i32_]]
+    ; CHECK: RET_ReallyLR implicit $d0
+    %0:fpr(<2 x s64>) = COPY $q0
+    %1:fpr(<2 x s64>) = COPY $q1
+    %4:fpr(<2 x s64>) = G_ICMP intpred(ugt), %0(<2 x s64>), %1
+    %3:fpr(<2 x s32>) = G_TRUNC %4(<2 x s64>)
+    $d0 = COPY %3(<2 x s32>)
+    RET_ReallyLR implicit $d0
+
+...
+---
+name:            test_v4i32_ugt
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: fpr }
+  - { id: 1, class: fpr }
+  - { id: 2, class: _ }
+  - { id: 3, class: fpr }
+  - { id: 4, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $q0, $q1
+
+    ; CHECK-LABEL: name: test_v4i32_ugt
+    ; CHECK: liveins: $q0, $q1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr128 = COPY $q0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr128 = COPY $q1
+    ; CHECK: [[CMHIv4i32_:%[0-9]+]]:fpr128 = CMHIv4i32 [[COPY]], [[COPY1]]
+    ; CHECK: [[XTNv4i16_:%[0-9]+]]:fpr64 = XTNv4i16 [[CMHIv4i32_]]
+    ; CHECK: $d0 = COPY [[XTNv4i16_]]
+    ; CHECK: RET_ReallyLR implicit $d0
+    %0:fpr(<4 x s32>) = COPY $q0
+    %1:fpr(<4 x s32>) = COPY $q1
+    %4:fpr(<4 x s32>) = G_ICMP intpred(ugt), %0(<4 x s32>), %1
+    %3:fpr(<4 x s16>) = G_TRUNC %4(<4 x s32>)
+    $d0 = COPY %3(<4 x s16>)
+    RET_ReallyLR implicit $d0
+
+...
+---
+name:            test_v2i32_ugt
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: fpr }
+  - { id: 1, class: fpr }
+  - { id: 2, class: _ }
+  - { id: 3, class: fpr }
+  - { id: 4, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $d0, $d1
+
+    ; CHECK-LABEL: name: test_v2i32_ugt
+    ; CHECK: liveins: $d0, $d1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr64 = COPY $d0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr64 = COPY $d1
+    ; CHECK: [[CMHIv2i32_:%[0-9]+]]:fpr64 = CMHIv2i32 [[COPY]], [[COPY1]]
+    ; CHECK: $d0 = COPY [[CMHIv2i32_]]
+    ; CHECK: RET_ReallyLR implicit $d0
+    %0:fpr(<2 x s32>) = COPY $d0
+    %1:fpr(<2 x s32>) = COPY $d1
+    %4:fpr(<2 x s32>) = G_ICMP intpred(ugt), %0(<2 x s32>), %1
+    %3:fpr(<2 x s32>) = COPY %4(<2 x s32>)
+    $d0 = COPY %3(<2 x s32>)
+    RET_ReallyLR implicit $d0
+
+...
+---
+name:            test_v2i16_ugt
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: _ }
+  - { id: 1, class: _ }
+  - { id: 2, class: fpr }
+  - { id: 3, class: fpr }
+  - { id: 4, class: _ }
+  - { id: 5, class: fpr }
+  - { id: 6, class: _ }
+  - { id: 7, class: fpr }
+  - { id: 8, class: fpr }
+  - { id: 9, class: fpr }
+  - { id: 10, class: gpr }
+  - { id: 11, class: fpr }
+  - { id: 12, class: fpr }
+  - { id: 13, class: gpr }
+  - { id: 14, class: fpr }
+  - { id: 15, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $d0, $d1
+
+    ; CHECK-LABEL: name: test_v2i16_ugt
+    ; CHECK: liveins: $d0, $d1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr64 = COPY $d0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr64 = COPY $d1
+    ; CHECK: [[MOVi32imm:%[0-9]+]]:gpr32 = MOVi32imm 65535
+    ; CHECK: [[DEF:%[0-9]+]]:fpr128 = IMPLICIT_DEF
+    ; CHECK: [[INSERT_SUBREG:%[0-9]+]]:fpr128 = INSERT_SUBREG [[DEF]], [[MOVi32imm]], %subreg.ssub
+    ; CHECK: [[INSvi32gpr:%[0-9]+]]:fpr128 = INSvi32gpr [[INSERT_SUBREG]], 1, [[MOVi32imm]]
+    ; CHECK: [[COPY2:%[0-9]+]]:fpr64 = COPY [[INSvi32gpr]].dsub
+    ; CHECK: [[ANDv8i8_:%[0-9]+]]:fpr64 = ANDv8i8 [[COPY]], [[COPY2]]
+    ; CHECK: [[MOVi32imm1:%[0-9]+]]:gpr32 = MOVi32imm 65535
+    ; CHECK: [[DEF1:%[0-9]+]]:fpr128 = IMPLICIT_DEF
+    ; CHECK: [[INSERT_SUBREG1:%[0-9]+]]:fpr128 = INSERT_SUBREG [[DEF1]], [[MOVi32imm1]], %subreg.ssub
+    ; CHECK: [[INSvi32gpr1:%[0-9]+]]:fpr128 = INSvi32gpr [[INSERT_SUBREG1]], 1, [[MOVi32imm1]]
+    ; CHECK: [[COPY3:%[0-9]+]]:fpr64 = COPY [[INSvi32gpr1]].dsub
+    ; CHECK: [[ANDv8i8_1:%[0-9]+]]:fpr64 = ANDv8i8 [[COPY1]], [[COPY3]]
+    ; CHECK: [[CMHIv2i32_:%[0-9]+]]:fpr64 = CMHIv2i32 [[ANDv8i8_]], [[ANDv8i8_1]]
+    ; CHECK: $d0 = COPY [[CMHIv2i32_]]
+    ; CHECK: RET_ReallyLR implicit $d0
+    %2:fpr(<2 x s32>) = COPY $d0
+    %3:fpr(<2 x s32>) = COPY $d1
+    %13:gpr(s32) = G_CONSTANT i32 65535
+    %14:fpr(<2 x s32>) = G_BUILD_VECTOR %13(s32), %13(s32)
+    %15:fpr(<2 x s32>) = COPY %2(<2 x s32>)
+    %7:fpr(<2 x s32>) = G_AND %15, %14
+    %10:gpr(s32) = G_CONSTANT i32 65535
+    %11:fpr(<2 x s32>) = G_BUILD_VECTOR %10(s32), %10(s32)
+    %12:fpr(<2 x s32>) = COPY %3(<2 x s32>)
+    %8:fpr(<2 x s32>) = G_AND %12, %11
+    %9:fpr(<2 x s32>) = G_ICMP intpred(ugt), %7(<2 x s32>), %8
+    %5:fpr(<2 x s32>) = COPY %9(<2 x s32>)
+    $d0 = COPY %5(<2 x s32>)
+    RET_ReallyLR implicit $d0
+
+...
+---
+name:            test_v8i16_ugt
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: fpr }
+  - { id: 1, class: fpr }
+  - { id: 2, class: _ }
+  - { id: 3, class: fpr }
+  - { id: 4, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $q0, $q1
+
+    ; CHECK-LABEL: name: test_v8i16_ugt
+    ; CHECK: liveins: $q0, $q1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr128 = COPY $q0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr128 = COPY $q1
+    ; CHECK: [[CMHIv8i16_:%[0-9]+]]:fpr128 = CMHIv8i16 [[COPY]], [[COPY1]]
+    ; CHECK: [[XTNv8i8_:%[0-9]+]]:fpr64 = XTNv8i8 [[CMHIv8i16_]]
+    ; CHECK: $d0 = COPY [[XTNv8i8_]]
+    ; CHECK: RET_ReallyLR implicit $d0
+    %0:fpr(<8 x s16>) = COPY $q0
+    %1:fpr(<8 x s16>) = COPY $q1
+    %4:fpr(<8 x s16>) = G_ICMP intpred(ugt), %0(<8 x s16>), %1
+    %3:fpr(<8 x s8>) = G_TRUNC %4(<8 x s16>)
+    $d0 = COPY %3(<8 x s8>)
+    RET_ReallyLR implicit $d0
+
+...
+---
+name:            test_v4i16_ugt
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: fpr }
+  - { id: 1, class: fpr }
+  - { id: 2, class: _ }
+  - { id: 3, class: fpr }
+  - { id: 4, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $d0, $d1
+
+    ; CHECK-LABEL: name: test_v4i16_ugt
+    ; CHECK: liveins: $d0, $d1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr64 = COPY $d0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr64 = COPY $d1
+    ; CHECK: [[CMHIv4i16_:%[0-9]+]]:fpr64 = CMHIv4i16 [[COPY]], [[COPY1]]
+    ; CHECK: $d0 = COPY [[CMHIv4i16_]]
+    ; CHECK: RET_ReallyLR implicit $d0
+    %0:fpr(<4 x s16>) = COPY $d0
+    %1:fpr(<4 x s16>) = COPY $d1
+    %4:fpr(<4 x s16>) = G_ICMP intpred(ugt), %0(<4 x s16>), %1
+    %3:fpr(<4 x s16>) = COPY %4(<4 x s16>)
+    $d0 = COPY %3(<4 x s16>)
+    RET_ReallyLR implicit $d0
+
+...
+---
+name:            test_v16i8_ugt
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: fpr }
+  - { id: 1, class: fpr }
+  - { id: 2, class: _ }
+  - { id: 3, class: fpr }
+  - { id: 4, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $q0, $q1
+
+    ; CHECK-LABEL: name: test_v16i8_ugt
+    ; CHECK: liveins: $q0, $q1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr128 = COPY $q0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr128 = COPY $q1
+    ; CHECK: [[CMHIv16i8_:%[0-9]+]]:fpr128 = CMHIv16i8 [[COPY]], [[COPY1]]
+    ; CHECK: $q0 = COPY [[CMHIv16i8_]]
+    ; CHECK: RET_ReallyLR implicit $q0
+    %0:fpr(<16 x s8>) = COPY $q0
+    %1:fpr(<16 x s8>) = COPY $q1
+    %4:fpr(<16 x s8>) = G_ICMP intpred(ugt), %0(<16 x s8>), %1
+    %3:fpr(<16 x s8>) = COPY %4(<16 x s8>)
+    $q0 = COPY %3(<16 x s8>)
+    RET_ReallyLR implicit $q0
+
+...
+---
+name:            test_v8i8_ugt
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: fpr }
+  - { id: 1, class: fpr }
+  - { id: 2, class: _ }
+  - { id: 3, class: fpr }
+  - { id: 4, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $d0, $d1
+
+    ; CHECK-LABEL: name: test_v8i8_ugt
+    ; CHECK: liveins: $d0, $d1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr64 = COPY $d0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr64 = COPY $d1
+    ; CHECK: [[CMHIv8i8_:%[0-9]+]]:fpr64 = CMHIv8i8 [[COPY]], [[COPY1]]
+    ; CHECK: $d0 = COPY [[CMHIv8i8_]]
+    ; CHECK: RET_ReallyLR implicit $d0
+    %0:fpr(<8 x s8>) = COPY $d0
+    %1:fpr(<8 x s8>) = COPY $d1
+    %4:fpr(<8 x s8>) = G_ICMP intpred(ugt), %0(<8 x s8>), %1
+    %3:fpr(<8 x s8>) = COPY %4(<8 x s8>)
+    $d0 = COPY %3(<8 x s8>)
+    RET_ReallyLR implicit $d0
+
+...
+---
+name:            test_v2i64_uge
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: fpr }
+  - { id: 1, class: fpr }
+  - { id: 2, class: _ }
+  - { id: 3, class: fpr }
+  - { id: 4, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $q0, $q1
+
+    ; CHECK-LABEL: name: test_v2i64_uge
+    ; CHECK: liveins: $q0, $q1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr128 = COPY $q0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr128 = COPY $q1
+    ; CHECK: [[CMHSv2i64_:%[0-9]+]]:fpr128 = CMHSv2i64 [[COPY]], [[COPY1]]
+    ; CHECK: [[XTNv2i32_:%[0-9]+]]:fpr64 = XTNv2i32 [[CMHSv2i64_]]
+    ; CHECK: $d0 = COPY [[XTNv2i32_]]
+    ; CHECK: RET_ReallyLR implicit $d0
+    %0:fpr(<2 x s64>) = COPY $q0
+    %1:fpr(<2 x s64>) = COPY $q1
+    %4:fpr(<2 x s64>) = G_ICMP intpred(uge), %0(<2 x s64>), %1
+    %3:fpr(<2 x s32>) = G_TRUNC %4(<2 x s64>)
+    $d0 = COPY %3(<2 x s32>)
+    RET_ReallyLR implicit $d0
+
+...
+---
+name:            test_v4i32_uge
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: fpr }
+  - { id: 1, class: fpr }
+  - { id: 2, class: _ }
+  - { id: 3, class: fpr }
+  - { id: 4, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $q0, $q1
+
+    ; CHECK-LABEL: name: test_v4i32_uge
+    ; CHECK: liveins: $q0, $q1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr128 = COPY $q0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr128 = COPY $q1
+    ; CHECK: [[CMHSv4i32_:%[0-9]+]]:fpr128 = CMHSv4i32 [[COPY]], [[COPY1]]
+    ; CHECK: [[XTNv4i16_:%[0-9]+]]:fpr64 = XTNv4i16 [[CMHSv4i32_]]
+    ; CHECK: $d0 = COPY [[XTNv4i16_]]
+    ; CHECK: RET_ReallyLR implicit $d0
+    %0:fpr(<4 x s32>) = COPY $q0
+    %1:fpr(<4 x s32>) = COPY $q1
+    %4:fpr(<4 x s32>) = G_ICMP intpred(uge), %0(<4 x s32>), %1
+    %3:fpr(<4 x s16>) = G_TRUNC %4(<4 x s32>)
+    $d0 = COPY %3(<4 x s16>)
+    RET_ReallyLR implicit $d0
+
+...
+---
+name:            test_v2i32_uge
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: fpr }
+  - { id: 1, class: fpr }
+  - { id: 2, class: _ }
+  - { id: 3, class: fpr }
+  - { id: 4, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $d0, $d1
+
+    ; CHECK-LABEL: name: test_v2i32_uge
+    ; CHECK: liveins: $d0, $d1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr64 = COPY $d0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr64 = COPY $d1
+    ; CHECK: [[CMHSv2i32_:%[0-9]+]]:fpr64 = CMHSv2i32 [[COPY]], [[COPY1]]
+    ; CHECK: $d0 = COPY [[CMHSv2i32_]]
+    ; CHECK: RET_ReallyLR implicit $d0
+    %0:fpr(<2 x s32>) = COPY $d0
+    %1:fpr(<2 x s32>) = COPY $d1
+    %4:fpr(<2 x s32>) = G_ICMP intpred(uge), %0(<2 x s32>), %1
+    %3:fpr(<2 x s32>) = COPY %4(<2 x s32>)
+    $d0 = COPY %3(<2 x s32>)
+    RET_ReallyLR implicit $d0
+
+...
+---
+name:            test_v2i16_uge
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: _ }
+  - { id: 1, class: _ }
+  - { id: 2, class: fpr }
+  - { id: 3, class: fpr }
+  - { id: 4, class: _ }
+  - { id: 5, class: fpr }
+  - { id: 6, class: _ }
+  - { id: 7, class: fpr }
+  - { id: 8, class: fpr }
+  - { id: 9, class: fpr }
+  - { id: 10, class: gpr }
+  - { id: 11, class: fpr }
+  - { id: 12, class: fpr }
+  - { id: 13, class: gpr }
+  - { id: 14, class: fpr }
+  - { id: 15, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $d0, $d1
+
+    ; CHECK-LABEL: name: test_v2i16_uge
+    ; CHECK: liveins: $d0, $d1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr64 = COPY $d0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr64 = COPY $d1
+    ; CHECK: [[MOVi32imm:%[0-9]+]]:gpr32 = MOVi32imm 65535
+    ; CHECK: [[DEF:%[0-9]+]]:fpr128 = IMPLICIT_DEF
+    ; CHECK: [[INSERT_SUBREG:%[0-9]+]]:fpr128 = INSERT_SUBREG [[DEF]], [[MOVi32imm]], %subreg.ssub
+    ; CHECK: [[INSvi32gpr:%[0-9]+]]:fpr128 = INSvi32gpr [[INSERT_SUBREG]], 1, [[MOVi32imm]]
+    ; CHECK: [[COPY2:%[0-9]+]]:fpr64 = COPY [[INSvi32gpr]].dsub
+    ; CHECK: [[ANDv8i8_:%[0-9]+]]:fpr64 = ANDv8i8 [[COPY]], [[COPY2]]
+    ; CHECK: [[MOVi32imm1:%[0-9]+]]:gpr32 = MOVi32imm 65535
+    ; CHECK: [[DEF1:%[0-9]+]]:fpr128 = IMPLICIT_DEF
+    ; CHECK: [[INSERT_SUBREG1:%[0-9]+]]:fpr128 = INSERT_SUBREG [[DEF1]], [[MOVi32imm1]], %subreg.ssub
+    ; CHECK: [[INSvi32gpr1:%[0-9]+]]:fpr128 = INSvi32gpr [[INSERT_SUBREG1]], 1, [[MOVi32imm1]]
+    ; CHECK: [[COPY3:%[0-9]+]]:fpr64 = COPY [[INSvi32gpr1]].dsub
+    ; CHECK: [[ANDv8i8_1:%[0-9]+]]:fpr64 = ANDv8i8 [[COPY1]], [[COPY3]]
+    ; CHECK: [[CMHSv2i32_:%[0-9]+]]:fpr64 = CMHSv2i32 [[ANDv8i8_]], [[ANDv8i8_1]]
+    ; CHECK: $d0 = COPY [[CMHSv2i32_]]
+    ; CHECK: RET_ReallyLR implicit $d0
+    %2:fpr(<2 x s32>) = COPY $d0
+    %3:fpr(<2 x s32>) = COPY $d1
+    %13:gpr(s32) = G_CONSTANT i32 65535
+    %14:fpr(<2 x s32>) = G_BUILD_VECTOR %13(s32), %13(s32)
+    %15:fpr(<2 x s32>) = COPY %2(<2 x s32>)
+    %7:fpr(<2 x s32>) = G_AND %15, %14
+    %10:gpr(s32) = G_CONSTANT i32 65535
+    %11:fpr(<2 x s32>) = G_BUILD_VECTOR %10(s32), %10(s32)
+    %12:fpr(<2 x s32>) = COPY %3(<2 x s32>)
+    %8:fpr(<2 x s32>) = G_AND %12, %11
+    %9:fpr(<2 x s32>) = G_ICMP intpred(uge), %7(<2 x s32>), %8
+    %5:fpr(<2 x s32>) = COPY %9(<2 x s32>)
+    $d0 = COPY %5(<2 x s32>)
+    RET_ReallyLR implicit $d0
+
+...
+---
+name:            test_v8i16_uge
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: fpr }
+  - { id: 1, class: fpr }
+  - { id: 2, class: _ }
+  - { id: 3, class: fpr }
+  - { id: 4, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $q0, $q1
+
+    ; CHECK-LABEL: name: test_v8i16_uge
+    ; CHECK: liveins: $q0, $q1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr128 = COPY $q0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr128 = COPY $q1
+    ; CHECK: [[CMHSv8i16_:%[0-9]+]]:fpr128 = CMHSv8i16 [[COPY]], [[COPY1]]
+    ; CHECK: [[XTNv8i8_:%[0-9]+]]:fpr64 = XTNv8i8 [[CMHSv8i16_]]
+    ; CHECK: $d0 = COPY [[XTNv8i8_]]
+    ; CHECK: RET_ReallyLR implicit $d0
+    %0:fpr(<8 x s16>) = COPY $q0
+    %1:fpr(<8 x s16>) = COPY $q1
+    %4:fpr(<8 x s16>) = G_ICMP intpred(uge), %0(<8 x s16>), %1
+    %3:fpr(<8 x s8>) = G_TRUNC %4(<8 x s16>)
+    $d0 = COPY %3(<8 x s8>)
+    RET_ReallyLR implicit $d0
+
+...
+---
+name:            test_v4i16_uge
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: fpr }
+  - { id: 1, class: fpr }
+  - { id: 2, class: _ }
+  - { id: 3, class: fpr }
+  - { id: 4, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $d0, $d1
+
+    ; CHECK-LABEL: name: test_v4i16_uge
+    ; CHECK: liveins: $d0, $d1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr64 = COPY $d0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr64 = COPY $d1
+    ; CHECK: [[CMHSv4i16_:%[0-9]+]]:fpr64 = CMHSv4i16 [[COPY]], [[COPY1]]
+    ; CHECK: $d0 = COPY [[CMHSv4i16_]]
+    ; CHECK: RET_ReallyLR implicit $d0
+    %0:fpr(<4 x s16>) = COPY $d0
+    %1:fpr(<4 x s16>) = COPY $d1
+    %4:fpr(<4 x s16>) = G_ICMP intpred(uge), %0(<4 x s16>), %1
+    %3:fpr(<4 x s16>) = COPY %4(<4 x s16>)
+    $d0 = COPY %3(<4 x s16>)
+    RET_ReallyLR implicit $d0
+
+...
+---
+name:            test_v16i8_uge
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: fpr }
+  - { id: 1, class: fpr }
+  - { id: 2, class: _ }
+  - { id: 3, class: fpr }
+  - { id: 4, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $q0, $q1
+
+    ; CHECK-LABEL: name: test_v16i8_uge
+    ; CHECK: liveins: $q0, $q1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr128 = COPY $q0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr128 = COPY $q1
+    ; CHECK: [[CMHSv16i8_:%[0-9]+]]:fpr128 = CMHSv16i8 [[COPY]], [[COPY1]]
+    ; CHECK: $q0 = COPY [[CMHSv16i8_]]
+    ; CHECK: RET_ReallyLR implicit $q0
+    %0:fpr(<16 x s8>) = COPY $q0
+    %1:fpr(<16 x s8>) = COPY $q1
+    %4:fpr(<16 x s8>) = G_ICMP intpred(uge), %0(<16 x s8>), %1
+    %3:fpr(<16 x s8>) = COPY %4(<16 x s8>)
+    $q0 = COPY %3(<16 x s8>)
+    RET_ReallyLR implicit $q0
+
+...
+---
+name:            test_v8i8_uge
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: fpr }
+  - { id: 1, class: fpr }
+  - { id: 2, class: _ }
+  - { id: 3, class: fpr }
+  - { id: 4, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $d0, $d1
+
+    ; CHECK-LABEL: name: test_v8i8_uge
+    ; CHECK: liveins: $d0, $d1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr64 = COPY $d0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr64 = COPY $d1
+    ; CHECK: [[CMHSv8i8_:%[0-9]+]]:fpr64 = CMHSv8i8 [[COPY]], [[COPY1]]
+    ; CHECK: $d0 = COPY [[CMHSv8i8_]]
+    ; CHECK: RET_ReallyLR implicit $d0
+    %0:fpr(<8 x s8>) = COPY $d0
+    %1:fpr(<8 x s8>) = COPY $d1
+    %4:fpr(<8 x s8>) = G_ICMP intpred(uge), %0(<8 x s8>), %1
+    %3:fpr(<8 x s8>) = COPY %4(<8 x s8>)
+    $d0 = COPY %3(<8 x s8>)
+    RET_ReallyLR implicit $d0
+
+...
+---
+name:            test_v2i64_ult
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: fpr }
+  - { id: 1, class: fpr }
+  - { id: 2, class: _ }
+  - { id: 3, class: fpr }
+  - { id: 4, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $q0, $q1
+
+    ; CHECK-LABEL: name: test_v2i64_ult
+    ; CHECK: liveins: $q0, $q1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr128 = COPY $q0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr128 = COPY $q1
+    ; CHECK: [[CMHIv2i64_:%[0-9]+]]:fpr128 = CMHIv2i64 [[COPY1]], [[COPY]]
+    ; CHECK: [[XTNv2i32_:%[0-9]+]]:fpr64 = XTNv2i32 [[CMHIv2i64_]]
+    ; CHECK: $d0 = COPY [[XTNv2i32_]]
+    ; CHECK: RET_ReallyLR implicit $d0
+    %0:fpr(<2 x s64>) = COPY $q0
+    %1:fpr(<2 x s64>) = COPY $q1
+    %4:fpr(<2 x s64>) = G_ICMP intpred(ult), %0(<2 x s64>), %1
+    %3:fpr(<2 x s32>) = G_TRUNC %4(<2 x s64>)
+    $d0 = COPY %3(<2 x s32>)
+    RET_ReallyLR implicit $d0
+
+...
+---
+name:            test_v4i32_ult
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: fpr }
+  - { id: 1, class: fpr }
+  - { id: 2, class: _ }
+  - { id: 3, class: fpr }
+  - { id: 4, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $q0, $q1
+
+    ; CHECK-LABEL: name: test_v4i32_ult
+    ; CHECK: liveins: $q0, $q1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr128 = COPY $q0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr128 = COPY $q1
+    ; CHECK: [[CMHIv4i32_:%[0-9]+]]:fpr128 = CMHIv4i32 [[COPY1]], [[COPY]]
+    ; CHECK: [[XTNv4i16_:%[0-9]+]]:fpr64 = XTNv4i16 [[CMHIv4i32_]]
+    ; CHECK: $d0 = COPY [[XTNv4i16_]]
+    ; CHECK: RET_ReallyLR implicit $d0
+    %0:fpr(<4 x s32>) = COPY $q0
+    %1:fpr(<4 x s32>) = COPY $q1
+    %4:fpr(<4 x s32>) = G_ICMP intpred(ult), %0(<4 x s32>), %1
+    %3:fpr(<4 x s16>) = G_TRUNC %4(<4 x s32>)
+    $d0 = COPY %3(<4 x s16>)
+    RET_ReallyLR implicit $d0
+
+...
+---
+name:            test_v2i32_ult
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: fpr }
+  - { id: 1, class: fpr }
+  - { id: 2, class: _ }
+  - { id: 3, class: fpr }
+  - { id: 4, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $d0, $d1
+
+    ; CHECK-LABEL: name: test_v2i32_ult
+    ; CHECK: liveins: $d0, $d1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr64 = COPY $d0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr64 = COPY $d1
+    ; CHECK: [[CMHIv2i32_:%[0-9]+]]:fpr64 = CMHIv2i32 [[COPY1]], [[COPY]]
+    ; CHECK: $d0 = COPY [[CMHIv2i32_]]
+    ; CHECK: RET_ReallyLR implicit $d0
+    %0:fpr(<2 x s32>) = COPY $d0
+    %1:fpr(<2 x s32>) = COPY $d1
+    %4:fpr(<2 x s32>) = G_ICMP intpred(ult), %0(<2 x s32>), %1
+    %3:fpr(<2 x s32>) = COPY %4(<2 x s32>)
+    $d0 = COPY %3(<2 x s32>)
+    RET_ReallyLR implicit $d0
+
+...
+---
+name:            test_v2i16_ult
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: _ }
+  - { id: 1, class: _ }
+  - { id: 2, class: fpr }
+  - { id: 3, class: fpr }
+  - { id: 4, class: _ }
+  - { id: 5, class: fpr }
+  - { id: 6, class: _ }
+  - { id: 7, class: fpr }
+  - { id: 8, class: fpr }
+  - { id: 9, class: fpr }
+  - { id: 10, class: gpr }
+  - { id: 11, class: fpr }
+  - { id: 12, class: fpr }
+  - { id: 13, class: gpr }
+  - { id: 14, class: fpr }
+  - { id: 15, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $d0, $d1
+
+    ; CHECK-LABEL: name: test_v2i16_ult
+    ; CHECK: liveins: $d0, $d1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr64 = COPY $d0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr64 = COPY $d1
+    ; CHECK: [[MOVi32imm:%[0-9]+]]:gpr32 = MOVi32imm 65535
+    ; CHECK: [[DEF:%[0-9]+]]:fpr128 = IMPLICIT_DEF
+    ; CHECK: [[INSERT_SUBREG:%[0-9]+]]:fpr128 = INSERT_SUBREG [[DEF]], [[MOVi32imm]], %subreg.ssub
+    ; CHECK: [[INSvi32gpr:%[0-9]+]]:fpr128 = INSvi32gpr [[INSERT_SUBREG]], 1, [[MOVi32imm]]
+    ; CHECK: [[COPY2:%[0-9]+]]:fpr64 = COPY [[INSvi32gpr]].dsub
+    ; CHECK: [[ANDv8i8_:%[0-9]+]]:fpr64 = ANDv8i8 [[COPY]], [[COPY2]]
+    ; CHECK: [[MOVi32imm1:%[0-9]+]]:gpr32 = MOVi32imm 65535
+    ; CHECK: [[DEF1:%[0-9]+]]:fpr128 = IMPLICIT_DEF
+    ; CHECK: [[INSERT_SUBREG1:%[0-9]+]]:fpr128 = INSERT_SUBREG [[DEF1]], [[MOVi32imm1]], %subreg.ssub
+    ; CHECK: [[INSvi32gpr1:%[0-9]+]]:fpr128 = INSvi32gpr [[INSERT_SUBREG1]], 1, [[MOVi32imm1]]
+    ; CHECK: [[COPY3:%[0-9]+]]:fpr64 = COPY [[INSvi32gpr1]].dsub
+    ; CHECK: [[ANDv8i8_1:%[0-9]+]]:fpr64 = ANDv8i8 [[COPY1]], [[COPY3]]
+    ; CHECK: [[CMHIv2i32_:%[0-9]+]]:fpr64 = CMHIv2i32 [[ANDv8i8_1]], [[ANDv8i8_]]
+    ; CHECK: $d0 = COPY [[CMHIv2i32_]]
+    ; CHECK: RET_ReallyLR implicit $d0
+    %2:fpr(<2 x s32>) = COPY $d0
+    %3:fpr(<2 x s32>) = COPY $d1
+    %13:gpr(s32) = G_CONSTANT i32 65535
+    %14:fpr(<2 x s32>) = G_BUILD_VECTOR %13(s32), %13(s32)
+    %15:fpr(<2 x s32>) = COPY %2(<2 x s32>)
+    %7:fpr(<2 x s32>) = G_AND %15, %14
+    %10:gpr(s32) = G_CONSTANT i32 65535
+    %11:fpr(<2 x s32>) = G_BUILD_VECTOR %10(s32), %10(s32)
+    %12:fpr(<2 x s32>) = COPY %3(<2 x s32>)
+    %8:fpr(<2 x s32>) = G_AND %12, %11
+    %9:fpr(<2 x s32>) = G_ICMP intpred(ult), %7(<2 x s32>), %8
+    %5:fpr(<2 x s32>) = COPY %9(<2 x s32>)
+    $d0 = COPY %5(<2 x s32>)
+    RET_ReallyLR implicit $d0
+
+...
+---
+name:            test_v8i16_ult
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: fpr }
+  - { id: 1, class: fpr }
+  - { id: 2, class: _ }
+  - { id: 3, class: fpr }
+  - { id: 4, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $q0, $q1
+
+    ; CHECK-LABEL: name: test_v8i16_ult
+    ; CHECK: liveins: $q0, $q1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr128 = COPY $q0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr128 = COPY $q1
+    ; CHECK: [[CMHIv8i16_:%[0-9]+]]:fpr128 = CMHIv8i16 [[COPY1]], [[COPY]]
+    ; CHECK: [[XTNv8i8_:%[0-9]+]]:fpr64 = XTNv8i8 [[CMHIv8i16_]]
+    ; CHECK: $d0 = COPY [[XTNv8i8_]]
+    ; CHECK: RET_ReallyLR implicit $d0
+    %0:fpr(<8 x s16>) = COPY $q0
+    %1:fpr(<8 x s16>) = COPY $q1
+    %4:fpr(<8 x s16>) = G_ICMP intpred(ult), %0(<8 x s16>), %1
+    %3:fpr(<8 x s8>) = G_TRUNC %4(<8 x s16>)
+    $d0 = COPY %3(<8 x s8>)
+    RET_ReallyLR implicit $d0
+
+...
+---
+name:            test_v4i16_ult
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: fpr }
+  - { id: 1, class: fpr }
+  - { id: 2, class: _ }
+  - { id: 3, class: fpr }
+  - { id: 4, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $d0, $d1
+
+    ; CHECK-LABEL: name: test_v4i16_ult
+    ; CHECK: liveins: $d0, $d1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr64 = COPY $d0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr64 = COPY $d1
+    ; CHECK: [[CMHIv4i16_:%[0-9]+]]:fpr64 = CMHIv4i16 [[COPY1]], [[COPY]]
+    ; CHECK: $d0 = COPY [[CMHIv4i16_]]
+    ; CHECK: RET_ReallyLR implicit $d0
+    %0:fpr(<4 x s16>) = COPY $d0
+    %1:fpr(<4 x s16>) = COPY $d1
+    %4:fpr(<4 x s16>) = G_ICMP intpred(ult), %0(<4 x s16>), %1
+    %3:fpr(<4 x s16>) = COPY %4(<4 x s16>)
+    $d0 = COPY %3(<4 x s16>)
+    RET_ReallyLR implicit $d0
+
+...
+---
+name:            test_v16i8_ult
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: fpr }
+  - { id: 1, class: fpr }
+  - { id: 2, class: _ }
+  - { id: 3, class: fpr }
+  - { id: 4, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $q0, $q1
+
+    ; CHECK-LABEL: name: test_v16i8_ult
+    ; CHECK: liveins: $q0, $q1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr128 = COPY $q0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr128 = COPY $q1
+    ; CHECK: [[CMHIv16i8_:%[0-9]+]]:fpr128 = CMHIv16i8 [[COPY1]], [[COPY]]
+    ; CHECK: $q0 = COPY [[CMHIv16i8_]]
+    ; CHECK: RET_ReallyLR implicit $q0
+    %0:fpr(<16 x s8>) = COPY $q0
+    %1:fpr(<16 x s8>) = COPY $q1
+    %4:fpr(<16 x s8>) = G_ICMP intpred(ult), %0(<16 x s8>), %1
+    %3:fpr(<16 x s8>) = COPY %4(<16 x s8>)
+    $q0 = COPY %3(<16 x s8>)
+    RET_ReallyLR implicit $q0
+
+...
+---
+name:            test_v8i8_ult
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: fpr }
+  - { id: 1, class: fpr }
+  - { id: 2, class: _ }
+  - { id: 3, class: fpr }
+  - { id: 4, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $d0, $d1
+
+    ; CHECK-LABEL: name: test_v8i8_ult
+    ; CHECK: liveins: $d0, $d1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr64 = COPY $d0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr64 = COPY $d1
+    ; CHECK: [[CMHIv8i8_:%[0-9]+]]:fpr64 = CMHIv8i8 [[COPY1]], [[COPY]]
+    ; CHECK: $d0 = COPY [[CMHIv8i8_]]
+    ; CHECK: RET_ReallyLR implicit $d0
+    %0:fpr(<8 x s8>) = COPY $d0
+    %1:fpr(<8 x s8>) = COPY $d1
+    %4:fpr(<8 x s8>) = G_ICMP intpred(ult), %0(<8 x s8>), %1
+    %3:fpr(<8 x s8>) = COPY %4(<8 x s8>)
+    $d0 = COPY %3(<8 x s8>)
+    RET_ReallyLR implicit $d0
+
+...
+---
+name:            test_v2i64_ule
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: fpr }
+  - { id: 1, class: fpr }
+  - { id: 2, class: _ }
+  - { id: 3, class: fpr }
+  - { id: 4, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $q0, $q1
+
+    ; CHECK-LABEL: name: test_v2i64_ule
+    ; CHECK: liveins: $q0, $q1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr128 = COPY $q0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr128 = COPY $q1
+    ; CHECK: [[CMHSv2i64_:%[0-9]+]]:fpr128 = CMHSv2i64 [[COPY1]], [[COPY]]
+    ; CHECK: [[XTNv2i32_:%[0-9]+]]:fpr64 = XTNv2i32 [[CMHSv2i64_]]
+    ; CHECK: $d0 = COPY [[XTNv2i32_]]
+    ; CHECK: RET_ReallyLR implicit $d0
+    %0:fpr(<2 x s64>) = COPY $q0
+    %1:fpr(<2 x s64>) = COPY $q1
+    %4:fpr(<2 x s64>) = G_ICMP intpred(ule), %0(<2 x s64>), %1
+    %3:fpr(<2 x s32>) = G_TRUNC %4(<2 x s64>)
+    $d0 = COPY %3(<2 x s32>)
+    RET_ReallyLR implicit $d0
+
+...
+---
+name:            test_v4i32_ule
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: fpr }
+  - { id: 1, class: fpr }
+  - { id: 2, class: _ }
+  - { id: 3, class: fpr }
+  - { id: 4, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $q0, $q1
+
+    ; CHECK-LABEL: name: test_v4i32_ule
+    ; CHECK: liveins: $q0, $q1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr128 = COPY $q0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr128 = COPY $q1
+    ; CHECK: [[CMHSv4i32_:%[0-9]+]]:fpr128 = CMHSv4i32 [[COPY1]], [[COPY]]
+    ; CHECK: [[XTNv4i16_:%[0-9]+]]:fpr64 = XTNv4i16 [[CMHSv4i32_]]
+    ; CHECK: $d0 = COPY [[XTNv4i16_]]
+    ; CHECK: RET_ReallyLR implicit $d0
+    %0:fpr(<4 x s32>) = COPY $q0
+    %1:fpr(<4 x s32>) = COPY $q1
+    %4:fpr(<4 x s32>) = G_ICMP intpred(ule), %0(<4 x s32>), %1
+    %3:fpr(<4 x s16>) = G_TRUNC %4(<4 x s32>)
+    $d0 = COPY %3(<4 x s16>)
+    RET_ReallyLR implicit $d0
+
+...
+---
+name:            test_v2i32_ule
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: fpr }
+  - { id: 1, class: fpr }
+  - { id: 2, class: _ }
+  - { id: 3, class: fpr }
+  - { id: 4, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $d0, $d1
+
+    ; CHECK-LABEL: name: test_v2i32_ule
+    ; CHECK: liveins: $d0, $d1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr64 = COPY $d0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr64 = COPY $d1
+    ; CHECK: [[CMHSv2i32_:%[0-9]+]]:fpr64 = CMHSv2i32 [[COPY1]], [[COPY]]
+    ; CHECK: $d0 = COPY [[CMHSv2i32_]]
+    ; CHECK: RET_ReallyLR implicit $d0
+    %0:fpr(<2 x s32>) = COPY $d0
+    %1:fpr(<2 x s32>) = COPY $d1
+    %4:fpr(<2 x s32>) = G_ICMP intpred(ule), %0(<2 x s32>), %1
+    %3:fpr(<2 x s32>) = COPY %4(<2 x s32>)
+    $d0 = COPY %3(<2 x s32>)
+    RET_ReallyLR implicit $d0
+
+...
+---
+name:            test_v2i16_ule
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: _ }
+  - { id: 1, class: _ }
+  - { id: 2, class: fpr }
+  - { id: 3, class: fpr }
+  - { id: 4, class: _ }
+  - { id: 5, class: fpr }
+  - { id: 6, class: _ }
+  - { id: 7, class: fpr }
+  - { id: 8, class: fpr }
+  - { id: 9, class: fpr }
+  - { id: 10, class: gpr }
+  - { id: 11, class: fpr }
+  - { id: 12, class: fpr }
+  - { id: 13, class: gpr }
+  - { id: 14, class: fpr }
+  - { id: 15, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $d0, $d1
+
+    ; CHECK-LABEL: name: test_v2i16_ule
+    ; CHECK: liveins: $d0, $d1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr64 = COPY $d0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr64 = COPY $d1
+    ; CHECK: [[MOVi32imm:%[0-9]+]]:gpr32 = MOVi32imm 65535
+    ; CHECK: [[DEF:%[0-9]+]]:fpr128 = IMPLICIT_DEF
+    ; CHECK: [[INSERT_SUBREG:%[0-9]+]]:fpr128 = INSERT_SUBREG [[DEF]], [[MOVi32imm]], %subreg.ssub
+    ; CHECK: [[INSvi32gpr:%[0-9]+]]:fpr128 = INSvi32gpr [[INSERT_SUBREG]], 1, [[MOVi32imm]]
+    ; CHECK: [[COPY2:%[0-9]+]]:fpr64 = COPY [[INSvi32gpr]].dsub
+    ; CHECK: [[ANDv8i8_:%[0-9]+]]:fpr64 = ANDv8i8 [[COPY]], [[COPY2]]
+    ; CHECK: [[MOVi32imm1:%[0-9]+]]:gpr32 = MOVi32imm 65535
+    ; CHECK: [[DEF1:%[0-9]+]]:fpr128 = IMPLICIT_DEF
+    ; CHECK: [[INSERT_SUBREG1:%[0-9]+]]:fpr128 = INSERT_SUBREG [[DEF1]], [[MOVi32imm1]], %subreg.ssub
+    ; CHECK: [[INSvi32gpr1:%[0-9]+]]:fpr128 = INSvi32gpr [[INSERT_SUBREG1]], 1, [[MOVi32imm1]]
+    ; CHECK: [[COPY3:%[0-9]+]]:fpr64 = COPY [[INSvi32gpr1]].dsub
+    ; CHECK: [[ANDv8i8_1:%[0-9]+]]:fpr64 = ANDv8i8 [[COPY1]], [[COPY3]]
+    ; CHECK: [[CMHSv2i32_:%[0-9]+]]:fpr64 = CMHSv2i32 [[ANDv8i8_1]], [[ANDv8i8_]]
+    ; CHECK: $d0 = COPY [[CMHSv2i32_]]
+    ; CHECK: RET_ReallyLR implicit $d0
+    %2:fpr(<2 x s32>) = COPY $d0
+    %3:fpr(<2 x s32>) = COPY $d1
+    %13:gpr(s32) = G_CONSTANT i32 65535
+    %14:fpr(<2 x s32>) = G_BUILD_VECTOR %13(s32), %13(s32)
+    %15:fpr(<2 x s32>) = COPY %2(<2 x s32>)
+    %7:fpr(<2 x s32>) = G_AND %15, %14
+    %10:gpr(s32) = G_CONSTANT i32 65535
+    %11:fpr(<2 x s32>) = G_BUILD_VECTOR %10(s32), %10(s32)
+    %12:fpr(<2 x s32>) = COPY %3(<2 x s32>)
+    %8:fpr(<2 x s32>) = G_AND %12, %11
+    %9:fpr(<2 x s32>) = G_ICMP intpred(ule), %7(<2 x s32>), %8
+    %5:fpr(<2 x s32>) = COPY %9(<2 x s32>)
+    $d0 = COPY %5(<2 x s32>)
+    RET_ReallyLR implicit $d0
+
+...
+---
+name:            test_v8i16_ule
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: fpr }
+  - { id: 1, class: fpr }
+  - { id: 2, class: _ }
+  - { id: 3, class: fpr }
+  - { id: 4, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $q0, $q1
+
+    ; CHECK-LABEL: name: test_v8i16_ule
+    ; CHECK: liveins: $q0, $q1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr128 = COPY $q0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr128 = COPY $q1
+    ; CHECK: [[CMHSv8i16_:%[0-9]+]]:fpr128 = CMHSv8i16 [[COPY1]], [[COPY]]
+    ; CHECK: [[XTNv8i8_:%[0-9]+]]:fpr64 = XTNv8i8 [[CMHSv8i16_]]
+    ; CHECK: $d0 = COPY [[XTNv8i8_]]
+    ; CHECK: RET_ReallyLR implicit $d0
+    %0:fpr(<8 x s16>) = COPY $q0
+    %1:fpr(<8 x s16>) = COPY $q1
+    %4:fpr(<8 x s16>) = G_ICMP intpred(ule), %0(<8 x s16>), %1
+    %3:fpr(<8 x s8>) = G_TRUNC %4(<8 x s16>)
+    $d0 = COPY %3(<8 x s8>)
+    RET_ReallyLR implicit $d0
+
+...
+---
+name:            test_v4i16_ule
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: fpr }
+  - { id: 1, class: fpr }
+  - { id: 2, class: _ }
+  - { id: 3, class: fpr }
+  - { id: 4, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $d0, $d1
+
+    ; CHECK-LABEL: name: test_v4i16_ule
+    ; CHECK: liveins: $d0, $d1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr64 = COPY $d0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr64 = COPY $d1
+    ; CHECK: [[CMHSv4i16_:%[0-9]+]]:fpr64 = CMHSv4i16 [[COPY1]], [[COPY]]
+    ; CHECK: $d0 = COPY [[CMHSv4i16_]]
+    ; CHECK: RET_ReallyLR implicit $d0
+    %0:fpr(<4 x s16>) = COPY $d0
+    %1:fpr(<4 x s16>) = COPY $d1
+    %4:fpr(<4 x s16>) = G_ICMP intpred(ule), %0(<4 x s16>), %1
+    %3:fpr(<4 x s16>) = COPY %4(<4 x s16>)
+    $d0 = COPY %3(<4 x s16>)
+    RET_ReallyLR implicit $d0
+
+...
+---
+name:            test_v16i8_ule
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: fpr }
+  - { id: 1, class: fpr }
+  - { id: 2, class: _ }
+  - { id: 3, class: fpr }
+  - { id: 4, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $q0, $q1
+
+    ; CHECK-LABEL: name: test_v16i8_ule
+    ; CHECK: liveins: $q0, $q1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr128 = COPY $q0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr128 = COPY $q1
+    ; CHECK: [[CMHSv16i8_:%[0-9]+]]:fpr128 = CMHSv16i8 [[COPY1]], [[COPY]]
+    ; CHECK: $q0 = COPY [[CMHSv16i8_]]
+    ; CHECK: RET_ReallyLR implicit $q0
+    %0:fpr(<16 x s8>) = COPY $q0
+    %1:fpr(<16 x s8>) = COPY $q1
+    %4:fpr(<16 x s8>) = G_ICMP intpred(ule), %0(<16 x s8>), %1
+    %3:fpr(<16 x s8>) = COPY %4(<16 x s8>)
+    $q0 = COPY %3(<16 x s8>)
+    RET_ReallyLR implicit $q0
+
+...
+---
+name:            test_v8i8_ule
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: fpr }
+  - { id: 1, class: fpr }
+  - { id: 2, class: _ }
+  - { id: 3, class: fpr }
+  - { id: 4, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $d0, $d1
+
+    ; CHECK-LABEL: name: test_v8i8_ule
+    ; CHECK: liveins: $d0, $d1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr64 = COPY $d0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr64 = COPY $d1
+    ; CHECK: [[CMHSv8i8_:%[0-9]+]]:fpr64 = CMHSv8i8 [[COPY1]], [[COPY]]
+    ; CHECK: $d0 = COPY [[CMHSv8i8_]]
+    ; CHECK: RET_ReallyLR implicit $d0
+    %0:fpr(<8 x s8>) = COPY $d0
+    %1:fpr(<8 x s8>) = COPY $d1
+    %4:fpr(<8 x s8>) = G_ICMP intpred(ule), %0(<8 x s8>), %1
+    %3:fpr(<8 x s8>) = COPY %4(<8 x s8>)
+    $d0 = COPY %3(<8 x s8>)
+    RET_ReallyLR implicit $d0
+
+...
+---
+name:            test_v2i64_sgt
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: fpr }
+  - { id: 1, class: fpr }
+  - { id: 2, class: _ }
+  - { id: 3, class: fpr }
+  - { id: 4, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $q0, $q1
+
+    ; CHECK-LABEL: name: test_v2i64_sgt
+    ; CHECK: liveins: $q0, $q1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr128 = COPY $q0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr128 = COPY $q1
+    ; CHECK: [[CMGTv2i64_:%[0-9]+]]:fpr128 = CMGTv2i64 [[COPY]], [[COPY1]]
+    ; CHECK: [[XTNv2i32_:%[0-9]+]]:fpr64 = XTNv2i32 [[CMGTv2i64_]]
+    ; CHECK: $d0 = COPY [[XTNv2i32_]]
+    ; CHECK: RET_ReallyLR implicit $d0
+    %0:fpr(<2 x s64>) = COPY $q0
+    %1:fpr(<2 x s64>) = COPY $q1
+    %4:fpr(<2 x s64>) = G_ICMP intpred(sgt), %0(<2 x s64>), %1
+    %3:fpr(<2 x s32>) = G_TRUNC %4(<2 x s64>)
+    $d0 = COPY %3(<2 x s32>)
+    RET_ReallyLR implicit $d0
+
+...
+---
+name:            test_v4i32_sgt
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: fpr }
+  - { id: 1, class: fpr }
+  - { id: 2, class: _ }
+  - { id: 3, class: fpr }
+  - { id: 4, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $q0, $q1
+
+    ; CHECK-LABEL: name: test_v4i32_sgt
+    ; CHECK: liveins: $q0, $q1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr128 = COPY $q0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr128 = COPY $q1
+    ; CHECK: [[CMGTv4i32_:%[0-9]+]]:fpr128 = CMGTv4i32 [[COPY]], [[COPY1]]
+    ; CHECK: [[XTNv4i16_:%[0-9]+]]:fpr64 = XTNv4i16 [[CMGTv4i32_]]
+    ; CHECK: $d0 = COPY [[XTNv4i16_]]
+    ; CHECK: RET_ReallyLR implicit $d0
+    %0:fpr(<4 x s32>) = COPY $q0
+    %1:fpr(<4 x s32>) = COPY $q1
+    %4:fpr(<4 x s32>) = G_ICMP intpred(sgt), %0(<4 x s32>), %1
+    %3:fpr(<4 x s16>) = G_TRUNC %4(<4 x s32>)
+    $d0 = COPY %3(<4 x s16>)
+    RET_ReallyLR implicit $d0
+
+...
+---
+name:            test_v2i32_sgt
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: fpr }
+  - { id: 1, class: fpr }
+  - { id: 2, class: _ }
+  - { id: 3, class: fpr }
+  - { id: 4, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $d0, $d1
+
+    ; CHECK-LABEL: name: test_v2i32_sgt
+    ; CHECK: liveins: $d0, $d1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr64 = COPY $d0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr64 = COPY $d1
+    ; CHECK: [[CMGTv2i32_:%[0-9]+]]:fpr64 = CMGTv2i32 [[COPY]], [[COPY1]]
+    ; CHECK: $d0 = COPY [[CMGTv2i32_]]
+    ; CHECK: RET_ReallyLR implicit $d0
+    %0:fpr(<2 x s32>) = COPY $d0
+    %1:fpr(<2 x s32>) = COPY $d1
+    %4:fpr(<2 x s32>) = G_ICMP intpred(sgt), %0(<2 x s32>), %1
+    %3:fpr(<2 x s32>) = COPY %4(<2 x s32>)
+    $d0 = COPY %3(<2 x s32>)
+    RET_ReallyLR implicit $d0
+
+...
+---
+name:            test_v2i16_sgt
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: _ }
+  - { id: 1, class: _ }
+  - { id: 2, class: fpr }
+  - { id: 3, class: fpr }
+  - { id: 4, class: _ }
+  - { id: 5, class: fpr }
+  - { id: 6, class: _ }
+  - { id: 7, class: fpr }
+  - { id: 8, class: fpr }
+  - { id: 9, class: fpr }
+  - { id: 10, class: gpr }
+  - { id: 11, class: fpr }
+  - { id: 12, class: fpr }
+  - { id: 13, class: fpr }
+  - { id: 14, class: gpr }
+  - { id: 15, class: fpr }
+  - { id: 16, class: fpr }
+  - { id: 17, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $d0, $d1
+
+    ; CHECK-LABEL: name: test_v2i16_sgt
+    ; CHECK: liveins: $d0, $d1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr64 = COPY $d0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr64 = COPY $d1
+    ; CHECK: [[MOVi32imm:%[0-9]+]]:gpr32 = MOVi32imm 16
+    ; CHECK: [[DEF:%[0-9]+]]:fpr128 = IMPLICIT_DEF
+    ; CHECK: [[INSERT_SUBREG:%[0-9]+]]:fpr128 = INSERT_SUBREG [[DEF]], [[MOVi32imm]], %subreg.ssub
+    ; CHECK: [[INSvi32gpr:%[0-9]+]]:fpr128 = INSvi32gpr [[INSERT_SUBREG]], 1, [[MOVi32imm]]
+    ; CHECK: [[COPY2:%[0-9]+]]:fpr64 = COPY [[INSvi32gpr]].dsub
+    ; CHECK: [[USHLv2i32_:%[0-9]+]]:fpr64 = USHLv2i32 [[COPY]], [[COPY2]]
+    ; CHECK: [[NEGv2i32_:%[0-9]+]]:fpr64 = NEGv2i32 [[COPY2]]
+    ; CHECK: [[SSHLv2i32_:%[0-9]+]]:fpr64 = SSHLv2i32 [[USHLv2i32_]], [[NEGv2i32_]]
+    ; CHECK: [[MOVi32imm1:%[0-9]+]]:gpr32 = MOVi32imm 16
+    ; CHECK: [[DEF1:%[0-9]+]]:fpr128 = IMPLICIT_DEF
+    ; CHECK: [[INSERT_SUBREG1:%[0-9]+]]:fpr128 = INSERT_SUBREG [[DEF1]], [[MOVi32imm1]], %subreg.ssub
+    ; CHECK: [[INSvi32gpr1:%[0-9]+]]:fpr128 = INSvi32gpr [[INSERT_SUBREG1]], 1, [[MOVi32imm1]]
+    ; CHECK: [[COPY3:%[0-9]+]]:fpr64 = COPY [[INSvi32gpr1]].dsub
+    ; CHECK: [[USHLv2i32_1:%[0-9]+]]:fpr64 = USHLv2i32 [[COPY1]], [[COPY3]]
+    ; CHECK: [[NEGv2i32_1:%[0-9]+]]:fpr64 = NEGv2i32 [[COPY3]]
+    ; CHECK: [[SSHLv2i32_1:%[0-9]+]]:fpr64 = SSHLv2i32 [[USHLv2i32_1]], [[NEGv2i32_1]]
+    ; CHECK: [[CMGTv2i32_:%[0-9]+]]:fpr64 = CMGTv2i32 [[SSHLv2i32_]], [[SSHLv2i32_1]]
+    ; CHECK: $d0 = COPY [[CMGTv2i32_]]
+    ; CHECK: RET_ReallyLR implicit $d0
+    %2:fpr(<2 x s32>) = COPY $d0
+    %3:fpr(<2 x s32>) = COPY $d1
+    %14:gpr(s32) = G_CONSTANT i32 16
+    %15:fpr(<2 x s32>) = G_BUILD_VECTOR %14(s32), %14(s32)
+    %16:fpr(<2 x s32>) = COPY %2(<2 x s32>)
+    %17:fpr(<2 x s32>) = G_SHL %16, %15(<2 x s32>)
+    %7:fpr(<2 x s32>) = G_ASHR %17, %15(<2 x s32>)
+    %10:gpr(s32) = G_CONSTANT i32 16
+    %11:fpr(<2 x s32>) = G_BUILD_VECTOR %10(s32), %10(s32)
+    %12:fpr(<2 x s32>) = COPY %3(<2 x s32>)
+    %13:fpr(<2 x s32>) = G_SHL %12, %11(<2 x s32>)
+    %8:fpr(<2 x s32>) = G_ASHR %13, %11(<2 x s32>)
+    %9:fpr(<2 x s32>) = G_ICMP intpred(sgt), %7(<2 x s32>), %8
+    %5:fpr(<2 x s32>) = COPY %9(<2 x s32>)
+    $d0 = COPY %5(<2 x s32>)
+    RET_ReallyLR implicit $d0
+
+...
+---
+name:            test_v8i16_sgt
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: fpr }
+  - { id: 1, class: fpr }
+  - { id: 2, class: _ }
+  - { id: 3, class: fpr }
+  - { id: 4, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $q0, $q1
+
+    ; CHECK-LABEL: name: test_v8i16_sgt
+    ; CHECK: liveins: $q0, $q1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr128 = COPY $q0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr128 = COPY $q1
+    ; CHECK: [[CMGTv8i16_:%[0-9]+]]:fpr128 = CMGTv8i16 [[COPY]], [[COPY1]]
+    ; CHECK: [[XTNv8i8_:%[0-9]+]]:fpr64 = XTNv8i8 [[CMGTv8i16_]]
+    ; CHECK: $d0 = COPY [[XTNv8i8_]]
+    ; CHECK: RET_ReallyLR implicit $d0
+    %0:fpr(<8 x s16>) = COPY $q0
+    %1:fpr(<8 x s16>) = COPY $q1
+    %4:fpr(<8 x s16>) = G_ICMP intpred(sgt), %0(<8 x s16>), %1
+    %3:fpr(<8 x s8>) = G_TRUNC %4(<8 x s16>)
+    $d0 = COPY %3(<8 x s8>)
+    RET_ReallyLR implicit $d0
+
+...
+---
+name:            test_v4i16_sgt
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: fpr }
+  - { id: 1, class: fpr }
+  - { id: 2, class: _ }
+  - { id: 3, class: fpr }
+  - { id: 4, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $d0, $d1
+
+    ; CHECK-LABEL: name: test_v4i16_sgt
+    ; CHECK: liveins: $d0, $d1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr64 = COPY $d0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr64 = COPY $d1
+    ; CHECK: [[CMGTv4i16_:%[0-9]+]]:fpr64 = CMGTv4i16 [[COPY]], [[COPY1]]
+    ; CHECK: $d0 = COPY [[CMGTv4i16_]]
+    ; CHECK: RET_ReallyLR implicit $d0
+    %0:fpr(<4 x s16>) = COPY $d0
+    %1:fpr(<4 x s16>) = COPY $d1
+    %4:fpr(<4 x s16>) = G_ICMP intpred(sgt), %0(<4 x s16>), %1
+    %3:fpr(<4 x s16>) = COPY %4(<4 x s16>)
+    $d0 = COPY %3(<4 x s16>)
+    RET_ReallyLR implicit $d0
+
+...
+---
+name:            test_v16i8_sgt
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: fpr }
+  - { id: 1, class: fpr }
+  - { id: 2, class: _ }
+  - { id: 3, class: fpr }
+  - { id: 4, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $q0, $q1
+
+    ; CHECK-LABEL: name: test_v16i8_sgt
+    ; CHECK: liveins: $q0, $q1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr128 = COPY $q0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr128 = COPY $q1
+    ; CHECK: [[CMGTv16i8_:%[0-9]+]]:fpr128 = CMGTv16i8 [[COPY]], [[COPY1]]
+    ; CHECK: $q0 = COPY [[CMGTv16i8_]]
+    ; CHECK: RET_ReallyLR implicit $q0
+    %0:fpr(<16 x s8>) = COPY $q0
+    %1:fpr(<16 x s8>) = COPY $q1
+    %4:fpr(<16 x s8>) = G_ICMP intpred(sgt), %0(<16 x s8>), %1
+    %3:fpr(<16 x s8>) = COPY %4(<16 x s8>)
+    $q0 = COPY %3(<16 x s8>)
+    RET_ReallyLR implicit $q0
+
+...
+---
+name:            test_v8i8_sgt
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: fpr }
+  - { id: 1, class: fpr }
+  - { id: 2, class: _ }
+  - { id: 3, class: fpr }
+  - { id: 4, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $d0, $d1
+
+    ; CHECK-LABEL: name: test_v8i8_sgt
+    ; CHECK: liveins: $d0, $d1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr64 = COPY $d0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr64 = COPY $d1
+    ; CHECK: [[CMGTv8i8_:%[0-9]+]]:fpr64 = CMGTv8i8 [[COPY]], [[COPY1]]
+    ; CHECK: $d0 = COPY [[CMGTv8i8_]]
+    ; CHECK: RET_ReallyLR implicit $d0
+    %0:fpr(<8 x s8>) = COPY $d0
+    %1:fpr(<8 x s8>) = COPY $d1
+    %4:fpr(<8 x s8>) = G_ICMP intpred(sgt), %0(<8 x s8>), %1
+    %3:fpr(<8 x s8>) = COPY %4(<8 x s8>)
+    $d0 = COPY %3(<8 x s8>)
+    RET_ReallyLR implicit $d0
+
+...
+---
+name:            test_v2i64_sge
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: fpr }
+  - { id: 1, class: fpr }
+  - { id: 2, class: _ }
+  - { id: 3, class: fpr }
+  - { id: 4, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $q0, $q1
+
+    ; CHECK-LABEL: name: test_v2i64_sge
+    ; CHECK: liveins: $q0, $q1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr128 = COPY $q0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr128 = COPY $q1
+    ; CHECK: [[CMGEv2i64_:%[0-9]+]]:fpr128 = CMGEv2i64 [[COPY]], [[COPY1]]
+    ; CHECK: [[XTNv2i32_:%[0-9]+]]:fpr64 = XTNv2i32 [[CMGEv2i64_]]
+    ; CHECK: $d0 = COPY [[XTNv2i32_]]
+    ; CHECK: RET_ReallyLR implicit $d0
+    %0:fpr(<2 x s64>) = COPY $q0
+    %1:fpr(<2 x s64>) = COPY $q1
+    %4:fpr(<2 x s64>) = G_ICMP intpred(sge), %0(<2 x s64>), %1
+    %3:fpr(<2 x s32>) = G_TRUNC %4(<2 x s64>)
+    $d0 = COPY %3(<2 x s32>)
+    RET_ReallyLR implicit $d0
+
+...
+---
+name:            test_v4i32_sge
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: fpr }
+  - { id: 1, class: fpr }
+  - { id: 2, class: _ }
+  - { id: 3, class: fpr }
+  - { id: 4, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $q0, $q1
+
+    ; CHECK-LABEL: name: test_v4i32_sge
+    ; CHECK: liveins: $q0, $q1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr128 = COPY $q0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr128 = COPY $q1
+    ; CHECK: [[CMGEv4i32_:%[0-9]+]]:fpr128 = CMGEv4i32 [[COPY]], [[COPY1]]
+    ; CHECK: [[XTNv4i16_:%[0-9]+]]:fpr64 = XTNv4i16 [[CMGEv4i32_]]
+    ; CHECK: $d0 = COPY [[XTNv4i16_]]
+    ; CHECK: RET_ReallyLR implicit $d0
+    %0:fpr(<4 x s32>) = COPY $q0
+    %1:fpr(<4 x s32>) = COPY $q1
+    %4:fpr(<4 x s32>) = G_ICMP intpred(sge), %0(<4 x s32>), %1
+    %3:fpr(<4 x s16>) = G_TRUNC %4(<4 x s32>)
+    $d0 = COPY %3(<4 x s16>)
+    RET_ReallyLR implicit $d0
+
+...
+---
+name:            test_v2i32_sge
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: fpr }
+  - { id: 1, class: fpr }
+  - { id: 2, class: _ }
+  - { id: 3, class: fpr }
+  - { id: 4, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $d0, $d1
+
+    ; CHECK-LABEL: name: test_v2i32_sge
+    ; CHECK: liveins: $d0, $d1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr64 = COPY $d0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr64 = COPY $d1
+    ; CHECK: [[CMGEv2i32_:%[0-9]+]]:fpr64 = CMGEv2i32 [[COPY]], [[COPY1]]
+    ; CHECK: $d0 = COPY [[CMGEv2i32_]]
+    ; CHECK: RET_ReallyLR implicit $d0
+    %0:fpr(<2 x s32>) = COPY $d0
+    %1:fpr(<2 x s32>) = COPY $d1
+    %4:fpr(<2 x s32>) = G_ICMP intpred(sge), %0(<2 x s32>), %1
+    %3:fpr(<2 x s32>) = COPY %4(<2 x s32>)
+    $d0 = COPY %3(<2 x s32>)
+    RET_ReallyLR implicit $d0
+
+...
+---
+name:            test_v2i16_sge
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: _ }
+  - { id: 1, class: _ }
+  - { id: 2, class: fpr }
+  - { id: 3, class: fpr }
+  - { id: 4, class: _ }
+  - { id: 5, class: fpr }
+  - { id: 6, class: _ }
+  - { id: 7, class: fpr }
+  - { id: 8, class: fpr }
+  - { id: 9, class: fpr }
+  - { id: 10, class: gpr }
+  - { id: 11, class: fpr }
+  - { id: 12, class: fpr }
+  - { id: 13, class: fpr }
+  - { id: 14, class: gpr }
+  - { id: 15, class: fpr }
+  - { id: 16, class: fpr }
+  - { id: 17, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $d0, $d1
+
+    ; CHECK-LABEL: name: test_v2i16_sge
+    ; CHECK: liveins: $d0, $d1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr64 = COPY $d0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr64 = COPY $d1
+    ; CHECK: [[MOVi32imm:%[0-9]+]]:gpr32 = MOVi32imm 16
+    ; CHECK: [[DEF:%[0-9]+]]:fpr128 = IMPLICIT_DEF
+    ; CHECK: [[INSERT_SUBREG:%[0-9]+]]:fpr128 = INSERT_SUBREG [[DEF]], [[MOVi32imm]], %subreg.ssub
+    ; CHECK: [[INSvi32gpr:%[0-9]+]]:fpr128 = INSvi32gpr [[INSERT_SUBREG]], 1, [[MOVi32imm]]
+    ; CHECK: [[COPY2:%[0-9]+]]:fpr64 = COPY [[INSvi32gpr]].dsub
+    ; CHECK: [[USHLv2i32_:%[0-9]+]]:fpr64 = USHLv2i32 [[COPY]], [[COPY2]]
+    ; CHECK: [[NEGv2i32_:%[0-9]+]]:fpr64 = NEGv2i32 [[COPY2]]
+    ; CHECK: [[SSHLv2i32_:%[0-9]+]]:fpr64 = SSHLv2i32 [[USHLv2i32_]], [[NEGv2i32_]]
+    ; CHECK: [[MOVi32imm1:%[0-9]+]]:gpr32 = MOVi32imm 16
+    ; CHECK: [[DEF1:%[0-9]+]]:fpr128 = IMPLICIT_DEF
+    ; CHECK: [[INSERT_SUBREG1:%[0-9]+]]:fpr128 = INSERT_SUBREG [[DEF1]], [[MOVi32imm1]], %subreg.ssub
+    ; CHECK: [[INSvi32gpr1:%[0-9]+]]:fpr128 = INSvi32gpr [[INSERT_SUBREG1]], 1, [[MOVi32imm1]]
+    ; CHECK: [[COPY3:%[0-9]+]]:fpr64 = COPY [[INSvi32gpr1]].dsub
+    ; CHECK: [[USHLv2i32_1:%[0-9]+]]:fpr64 = USHLv2i32 [[COPY1]], [[COPY3]]
+    ; CHECK: [[NEGv2i32_1:%[0-9]+]]:fpr64 = NEGv2i32 [[COPY3]]
+    ; CHECK: [[SSHLv2i32_1:%[0-9]+]]:fpr64 = SSHLv2i32 [[USHLv2i32_1]], [[NEGv2i32_1]]
+    ; CHECK: [[CMGEv2i32_:%[0-9]+]]:fpr64 = CMGEv2i32 [[SSHLv2i32_]], [[SSHLv2i32_1]]
+    ; CHECK: $d0 = COPY [[CMGEv2i32_]]
+    ; CHECK: RET_ReallyLR implicit $d0
+    %2:fpr(<2 x s32>) = COPY $d0
+    %3:fpr(<2 x s32>) = COPY $d1
+    %14:gpr(s32) = G_CONSTANT i32 16
+    %15:fpr(<2 x s32>) = G_BUILD_VECTOR %14(s32), %14(s32)
+    %16:fpr(<2 x s32>) = COPY %2(<2 x s32>)
+    %17:fpr(<2 x s32>) = G_SHL %16, %15(<2 x s32>)
+    %7:fpr(<2 x s32>) = G_ASHR %17, %15(<2 x s32>)
+    %10:gpr(s32) = G_CONSTANT i32 16
+    %11:fpr(<2 x s32>) = G_BUILD_VECTOR %10(s32), %10(s32)
+    %12:fpr(<2 x s32>) = COPY %3(<2 x s32>)
+    %13:fpr(<2 x s32>) = G_SHL %12, %11(<2 x s32>)
+    %8:fpr(<2 x s32>) = G_ASHR %13, %11(<2 x s32>)
+    %9:fpr(<2 x s32>) = G_ICMP intpred(sge), %7(<2 x s32>), %8
+    %5:fpr(<2 x s32>) = COPY %9(<2 x s32>)
+    $d0 = COPY %5(<2 x s32>)
+    RET_ReallyLR implicit $d0
+
+...
+---
+name:            test_v8i16_sge
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: fpr }
+  - { id: 1, class: fpr }
+  - { id: 2, class: _ }
+  - { id: 3, class: fpr }
+  - { id: 4, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $q0, $q1
+
+    ; CHECK-LABEL: name: test_v8i16_sge
+    ; CHECK: liveins: $q0, $q1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr128 = COPY $q0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr128 = COPY $q1
+    ; CHECK: [[CMGEv8i16_:%[0-9]+]]:fpr128 = CMGEv8i16 [[COPY]], [[COPY1]]
+    ; CHECK: [[XTNv8i8_:%[0-9]+]]:fpr64 = XTNv8i8 [[CMGEv8i16_]]
+    ; CHECK: $d0 = COPY [[XTNv8i8_]]
+    ; CHECK: RET_ReallyLR implicit $d0
+    %0:fpr(<8 x s16>) = COPY $q0
+    %1:fpr(<8 x s16>) = COPY $q1
+    %4:fpr(<8 x s16>) = G_ICMP intpred(sge), %0(<8 x s16>), %1
+    %3:fpr(<8 x s8>) = G_TRUNC %4(<8 x s16>)
+    $d0 = COPY %3(<8 x s8>)
+    RET_ReallyLR implicit $d0
+
+...
+---
+name:            test_v4i16_sge
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: fpr }
+  - { id: 1, class: fpr }
+  - { id: 2, class: _ }
+  - { id: 3, class: fpr }
+  - { id: 4, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $d0, $d1
+
+    ; CHECK-LABEL: name: test_v4i16_sge
+    ; CHECK: liveins: $d0, $d1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr64 = COPY $d0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr64 = COPY $d1
+    ; CHECK: [[CMGEv4i16_:%[0-9]+]]:fpr64 = CMGEv4i16 [[COPY]], [[COPY1]]
+    ; CHECK: $d0 = COPY [[CMGEv4i16_]]
+    ; CHECK: RET_ReallyLR implicit $d0
+    %0:fpr(<4 x s16>) = COPY $d0
+    %1:fpr(<4 x s16>) = COPY $d1
+    %4:fpr(<4 x s16>) = G_ICMP intpred(sge), %0(<4 x s16>), %1
+    %3:fpr(<4 x s16>) = COPY %4(<4 x s16>)
+    $d0 = COPY %3(<4 x s16>)
+    RET_ReallyLR implicit $d0
+
+...
+---
+name:            test_v16i8_sge
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: fpr }
+  - { id: 1, class: fpr }
+  - { id: 2, class: _ }
+  - { id: 3, class: fpr }
+  - { id: 4, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $q0, $q1
+
+    ; CHECK-LABEL: name: test_v16i8_sge
+    ; CHECK: liveins: $q0, $q1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr128 = COPY $q0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr128 = COPY $q1
+    ; CHECK: [[CMGEv16i8_:%[0-9]+]]:fpr128 = CMGEv16i8 [[COPY]], [[COPY1]]
+    ; CHECK: $q0 = COPY [[CMGEv16i8_]]
+    ; CHECK: RET_ReallyLR implicit $q0
+    %0:fpr(<16 x s8>) = COPY $q0
+    %1:fpr(<16 x s8>) = COPY $q1
+    %4:fpr(<16 x s8>) = G_ICMP intpred(sge), %0(<16 x s8>), %1
+    %3:fpr(<16 x s8>) = COPY %4(<16 x s8>)
+    $q0 = COPY %3(<16 x s8>)
+    RET_ReallyLR implicit $q0
+
+...
+---
+name:            test_v8i8_sge
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: fpr }
+  - { id: 1, class: fpr }
+  - { id: 2, class: _ }
+  - { id: 3, class: fpr }
+  - { id: 4, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $d0, $d1
+
+    ; CHECK-LABEL: name: test_v8i8_sge
+    ; CHECK: liveins: $d0, $d1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr64 = COPY $d0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr64 = COPY $d1
+    ; CHECK: [[CMGEv8i8_:%[0-9]+]]:fpr64 = CMGEv8i8 [[COPY]], [[COPY1]]
+    ; CHECK: $d0 = COPY [[CMGEv8i8_]]
+    ; CHECK: RET_ReallyLR implicit $d0
+    %0:fpr(<8 x s8>) = COPY $d0
+    %1:fpr(<8 x s8>) = COPY $d1
+    %4:fpr(<8 x s8>) = G_ICMP intpred(sge), %0(<8 x s8>), %1
+    %3:fpr(<8 x s8>) = COPY %4(<8 x s8>)
+    $d0 = COPY %3(<8 x s8>)
+    RET_ReallyLR implicit $d0
+
+...
+---
+name:            test_v2i64_slt
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: fpr }
+  - { id: 1, class: fpr }
+  - { id: 2, class: _ }
+  - { id: 3, class: fpr }
+  - { id: 4, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $q0, $q1
+
+    ; CHECK-LABEL: name: test_v2i64_slt
+    ; CHECK: liveins: $q0, $q1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr128 = COPY $q0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr128 = COPY $q1
+    ; CHECK: [[CMGTv2i64_:%[0-9]+]]:fpr128 = CMGTv2i64 [[COPY1]], [[COPY]]
+    ; CHECK: [[XTNv2i32_:%[0-9]+]]:fpr64 = XTNv2i32 [[CMGTv2i64_]]
+    ; CHECK: $d0 = COPY [[XTNv2i32_]]
+    ; CHECK: RET_ReallyLR implicit $d0
+    %0:fpr(<2 x s64>) = COPY $q0
+    %1:fpr(<2 x s64>) = COPY $q1
+    %4:fpr(<2 x s64>) = G_ICMP intpred(slt), %0(<2 x s64>), %1
+    %3:fpr(<2 x s32>) = G_TRUNC %4(<2 x s64>)
+    $d0 = COPY %3(<2 x s32>)
+    RET_ReallyLR implicit $d0
+
+...
+---
+name:            test_v4i32_slt
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: fpr }
+  - { id: 1, class: fpr }
+  - { id: 2, class: _ }
+  - { id: 3, class: fpr }
+  - { id: 4, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $q0, $q1
+
+    ; CHECK-LABEL: name: test_v4i32_slt
+    ; CHECK: liveins: $q0, $q1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr128 = COPY $q0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr128 = COPY $q1
+    ; CHECK: [[CMGTv4i32_:%[0-9]+]]:fpr128 = CMGTv4i32 [[COPY1]], [[COPY]]
+    ; CHECK: [[XTNv4i16_:%[0-9]+]]:fpr64 = XTNv4i16 [[CMGTv4i32_]]
+    ; CHECK: $d0 = COPY [[XTNv4i16_]]
+    ; CHECK: RET_ReallyLR implicit $d0
+    %0:fpr(<4 x s32>) = COPY $q0
+    %1:fpr(<4 x s32>) = COPY $q1
+    %4:fpr(<4 x s32>) = G_ICMP intpred(slt), %0(<4 x s32>), %1
+    %3:fpr(<4 x s16>) = G_TRUNC %4(<4 x s32>)
+    $d0 = COPY %3(<4 x s16>)
+    RET_ReallyLR implicit $d0
+
+...
+---
+name:            test_v2i32_slt
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: fpr }
+  - { id: 1, class: fpr }
+  - { id: 2, class: _ }
+  - { id: 3, class: fpr }
+  - { id: 4, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $d0, $d1
+
+    ; CHECK-LABEL: name: test_v2i32_slt
+    ; CHECK: liveins: $d0, $d1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr64 = COPY $d0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr64 = COPY $d1
+    ; CHECK: [[CMGTv2i32_:%[0-9]+]]:fpr64 = CMGTv2i32 [[COPY1]], [[COPY]]
+    ; CHECK: $d0 = COPY [[CMGTv2i32_]]
+    ; CHECK: RET_ReallyLR implicit $d0
+    %0:fpr(<2 x s32>) = COPY $d0
+    %1:fpr(<2 x s32>) = COPY $d1
+    %4:fpr(<2 x s32>) = G_ICMP intpred(slt), %0(<2 x s32>), %1
+    %3:fpr(<2 x s32>) = COPY %4(<2 x s32>)
+    $d0 = COPY %3(<2 x s32>)
+    RET_ReallyLR implicit $d0
+
+...
+---
+name:            test_v2i16_slt
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: _ }
+  - { id: 1, class: _ }
+  - { id: 2, class: fpr }
+  - { id: 3, class: fpr }
+  - { id: 4, class: _ }
+  - { id: 5, class: fpr }
+  - { id: 6, class: _ }
+  - { id: 7, class: fpr }
+  - { id: 8, class: fpr }
+  - { id: 9, class: fpr }
+  - { id: 10, class: gpr }
+  - { id: 11, class: fpr }
+  - { id: 12, class: fpr }
+  - { id: 13, class: fpr }
+  - { id: 14, class: gpr }
+  - { id: 15, class: fpr }
+  - { id: 16, class: fpr }
+  - { id: 17, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $d0, $d1
+
+    ; CHECK-LABEL: name: test_v2i16_slt
+    ; CHECK: liveins: $d0, $d1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr64 = COPY $d0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr64 = COPY $d1
+    ; CHECK: [[MOVi32imm:%[0-9]+]]:gpr32 = MOVi32imm 16
+    ; CHECK: [[DEF:%[0-9]+]]:fpr128 = IMPLICIT_DEF
+    ; CHECK: [[INSERT_SUBREG:%[0-9]+]]:fpr128 = INSERT_SUBREG [[DEF]], [[MOVi32imm]], %subreg.ssub
+    ; CHECK: [[INSvi32gpr:%[0-9]+]]:fpr128 = INSvi32gpr [[INSERT_SUBREG]], 1, [[MOVi32imm]]
+    ; CHECK: [[COPY2:%[0-9]+]]:fpr64 = COPY [[INSvi32gpr]].dsub
+    ; CHECK: [[USHLv2i32_:%[0-9]+]]:fpr64 = USHLv2i32 [[COPY]], [[COPY2]]
+    ; CHECK: [[NEGv2i32_:%[0-9]+]]:fpr64 = NEGv2i32 [[COPY2]]
+    ; CHECK: [[SSHLv2i32_:%[0-9]+]]:fpr64 = SSHLv2i32 [[USHLv2i32_]], [[NEGv2i32_]]
+    ; CHECK: [[MOVi32imm1:%[0-9]+]]:gpr32 = MOVi32imm 16
+    ; CHECK: [[DEF1:%[0-9]+]]:fpr128 = IMPLICIT_DEF
+    ; CHECK: [[INSERT_SUBREG1:%[0-9]+]]:fpr128 = INSERT_SUBREG [[DEF1]], [[MOVi32imm1]], %subreg.ssub
+    ; CHECK: [[INSvi32gpr1:%[0-9]+]]:fpr128 = INSvi32gpr [[INSERT_SUBREG1]], 1, [[MOVi32imm1]]
+    ; CHECK: [[COPY3:%[0-9]+]]:fpr64 = COPY [[INSvi32gpr1]].dsub
+    ; CHECK: [[USHLv2i32_1:%[0-9]+]]:fpr64 = USHLv2i32 [[COPY1]], [[COPY3]]
+    ; CHECK: [[NEGv2i32_1:%[0-9]+]]:fpr64 = NEGv2i32 [[COPY3]]
+    ; CHECK: [[SSHLv2i32_1:%[0-9]+]]:fpr64 = SSHLv2i32 [[USHLv2i32_1]], [[NEGv2i32_1]]
+    ; CHECK: [[CMGTv2i32_:%[0-9]+]]:fpr64 = CMGTv2i32 [[SSHLv2i32_1]], [[SSHLv2i32_]]
+    ; CHECK: $d0 = COPY [[CMGTv2i32_]]
+    ; CHECK: RET_ReallyLR implicit $d0
+    %2:fpr(<2 x s32>) = COPY $d0
+    %3:fpr(<2 x s32>) = COPY $d1
+    %14:gpr(s32) = G_CONSTANT i32 16
+    %15:fpr(<2 x s32>) = G_BUILD_VECTOR %14(s32), %14(s32)
+    %16:fpr(<2 x s32>) = COPY %2(<2 x s32>)
+    %17:fpr(<2 x s32>) = G_SHL %16, %15(<2 x s32>)
+    %7:fpr(<2 x s32>) = G_ASHR %17, %15(<2 x s32>)
+    %10:gpr(s32) = G_CONSTANT i32 16
+    %11:fpr(<2 x s32>) = G_BUILD_VECTOR %10(s32), %10(s32)
+    %12:fpr(<2 x s32>) = COPY %3(<2 x s32>)
+    %13:fpr(<2 x s32>) = G_SHL %12, %11(<2 x s32>)
+    %8:fpr(<2 x s32>) = G_ASHR %13, %11(<2 x s32>)
+    %9:fpr(<2 x s32>) = G_ICMP intpred(slt), %7(<2 x s32>), %8
+    %5:fpr(<2 x s32>) = COPY %9(<2 x s32>)
+    $d0 = COPY %5(<2 x s32>)
+    RET_ReallyLR implicit $d0
+
+...
+---
+name:            test_v8i16_slt
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: fpr }
+  - { id: 1, class: fpr }
+  - { id: 2, class: _ }
+  - { id: 3, class: fpr }
+  - { id: 4, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $q0, $q1
+
+    ; CHECK-LABEL: name: test_v8i16_slt
+    ; CHECK: liveins: $q0, $q1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr128 = COPY $q0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr128 = COPY $q1
+    ; CHECK: [[CMGTv8i16_:%[0-9]+]]:fpr128 = CMGTv8i16 [[COPY1]], [[COPY]]
+    ; CHECK: [[XTNv8i8_:%[0-9]+]]:fpr64 = XTNv8i8 [[CMGTv8i16_]]
+    ; CHECK: $d0 = COPY [[XTNv8i8_]]
+    ; CHECK: RET_ReallyLR implicit $d0
+    %0:fpr(<8 x s16>) = COPY $q0
+    %1:fpr(<8 x s16>) = COPY $q1
+    %4:fpr(<8 x s16>) = G_ICMP intpred(slt), %0(<8 x s16>), %1
+    %3:fpr(<8 x s8>) = G_TRUNC %4(<8 x s16>)
+    $d0 = COPY %3(<8 x s8>)
+    RET_ReallyLR implicit $d0
+
+...
+---
+name:            test_v4i16_slt
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: fpr }
+  - { id: 1, class: fpr }
+  - { id: 2, class: _ }
+  - { id: 3, class: fpr }
+  - { id: 4, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $d0, $d1
+
+    ; CHECK-LABEL: name: test_v4i16_slt
+    ; CHECK: liveins: $d0, $d1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr64 = COPY $d0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr64 = COPY $d1
+    ; CHECK: [[CMGTv4i16_:%[0-9]+]]:fpr64 = CMGTv4i16 [[COPY1]], [[COPY]]
+    ; CHECK: $d0 = COPY [[CMGTv4i16_]]
+    ; CHECK: RET_ReallyLR implicit $d0
+    %0:fpr(<4 x s16>) = COPY $d0
+    %1:fpr(<4 x s16>) = COPY $d1
+    %4:fpr(<4 x s16>) = G_ICMP intpred(slt), %0(<4 x s16>), %1
+    %3:fpr(<4 x s16>) = COPY %4(<4 x s16>)
+    $d0 = COPY %3(<4 x s16>)
+    RET_ReallyLR implicit $d0
+
+...
+---
+name:            test_v16i8_slt
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: fpr }
+  - { id: 1, class: fpr }
+  - { id: 2, class: _ }
+  - { id: 3, class: fpr }
+  - { id: 4, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $q0, $q1
+
+    ; CHECK-LABEL: name: test_v16i8_slt
+    ; CHECK: liveins: $q0, $q1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr128 = COPY $q0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr128 = COPY $q1
+    ; CHECK: [[CMGTv16i8_:%[0-9]+]]:fpr128 = CMGTv16i8 [[COPY1]], [[COPY]]
+    ; CHECK: $q0 = COPY [[CMGTv16i8_]]
+    ; CHECK: RET_ReallyLR implicit $q0
+    %0:fpr(<16 x s8>) = COPY $q0
+    %1:fpr(<16 x s8>) = COPY $q1
+    %4:fpr(<16 x s8>) = G_ICMP intpred(slt), %0(<16 x s8>), %1
+    %3:fpr(<16 x s8>) = COPY %4(<16 x s8>)
+    $q0 = COPY %3(<16 x s8>)
+    RET_ReallyLR implicit $q0
+
+...
+---
+name:            test_v8i8_slt
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: fpr }
+  - { id: 1, class: fpr }
+  - { id: 2, class: _ }
+  - { id: 3, class: fpr }
+  - { id: 4, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $d0, $d1
+
+    ; CHECK-LABEL: name: test_v8i8_slt
+    ; CHECK: liveins: $d0, $d1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr64 = COPY $d0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr64 = COPY $d1
+    ; CHECK: [[CMGTv8i8_:%[0-9]+]]:fpr64 = CMGTv8i8 [[COPY1]], [[COPY]]
+    ; CHECK: $d0 = COPY [[CMGTv8i8_]]
+    ; CHECK: RET_ReallyLR implicit $d0
+    %0:fpr(<8 x s8>) = COPY $d0
+    %1:fpr(<8 x s8>) = COPY $d1
+    %4:fpr(<8 x s8>) = G_ICMP intpred(slt), %0(<8 x s8>), %1
+    %3:fpr(<8 x s8>) = COPY %4(<8 x s8>)
+    $d0 = COPY %3(<8 x s8>)
+    RET_ReallyLR implicit $d0
+
+...
+---
+name:            test_v2i64_sle
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: fpr }
+  - { id: 1, class: fpr }
+  - { id: 2, class: _ }
+  - { id: 3, class: fpr }
+  - { id: 4, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $q0, $q1
+
+    ; CHECK-LABEL: name: test_v2i64_sle
+    ; CHECK: liveins: $q0, $q1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr128 = COPY $q0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr128 = COPY $q1
+    ; CHECK: [[CMGEv2i64_:%[0-9]+]]:fpr128 = CMGEv2i64 [[COPY1]], [[COPY]]
+    ; CHECK: [[XTNv2i32_:%[0-9]+]]:fpr64 = XTNv2i32 [[CMGEv2i64_]]
+    ; CHECK: $d0 = COPY [[XTNv2i32_]]
+    ; CHECK: RET_ReallyLR implicit $d0
+    %0:fpr(<2 x s64>) = COPY $q0
+    %1:fpr(<2 x s64>) = COPY $q1
+    %4:fpr(<2 x s64>) = G_ICMP intpred(sle), %0(<2 x s64>), %1
+    %3:fpr(<2 x s32>) = G_TRUNC %4(<2 x s64>)
+    $d0 = COPY %3(<2 x s32>)
+    RET_ReallyLR implicit $d0
+
+...
+---
+name:            test_v4i32_sle
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: fpr }
+  - { id: 1, class: fpr }
+  - { id: 2, class: _ }
+  - { id: 3, class: fpr }
+  - { id: 4, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $q0, $q1
+
+    ; CHECK-LABEL: name: test_v4i32_sle
+    ; CHECK: liveins: $q0, $q1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr128 = COPY $q0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr128 = COPY $q1
+    ; CHECK: [[CMGEv4i32_:%[0-9]+]]:fpr128 = CMGEv4i32 [[COPY1]], [[COPY]]
+    ; CHECK: [[XTNv4i16_:%[0-9]+]]:fpr64 = XTNv4i16 [[CMGEv4i32_]]
+    ; CHECK: $d0 = COPY [[XTNv4i16_]]
+    ; CHECK: RET_ReallyLR implicit $d0
+    %0:fpr(<4 x s32>) = COPY $q0
+    %1:fpr(<4 x s32>) = COPY $q1
+    %4:fpr(<4 x s32>) = G_ICMP intpred(sle), %0(<4 x s32>), %1
+    %3:fpr(<4 x s16>) = G_TRUNC %4(<4 x s32>)
+    $d0 = COPY %3(<4 x s16>)
+    RET_ReallyLR implicit $d0
+
+...
+---
+name:            test_v2i32_sle
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: fpr }
+  - { id: 1, class: fpr }
+  - { id: 2, class: _ }
+  - { id: 3, class: fpr }
+  - { id: 4, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $d0, $d1
+
+    ; CHECK-LABEL: name: test_v2i32_sle
+    ; CHECK: liveins: $d0, $d1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr64 = COPY $d0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr64 = COPY $d1
+    ; CHECK: [[CMGEv2i32_:%[0-9]+]]:fpr64 = CMGEv2i32 [[COPY1]], [[COPY]]
+    ; CHECK: $d0 = COPY [[CMGEv2i32_]]
+    ; CHECK: RET_ReallyLR implicit $d0
+    %0:fpr(<2 x s32>) = COPY $d0
+    %1:fpr(<2 x s32>) = COPY $d1
+    %4:fpr(<2 x s32>) = G_ICMP intpred(sle), %0(<2 x s32>), %1
+    %3:fpr(<2 x s32>) = COPY %4(<2 x s32>)
+    $d0 = COPY %3(<2 x s32>)
+    RET_ReallyLR implicit $d0
+
+...
+---
+name:            test_v2i16_sle
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: _ }
+  - { id: 1, class: _ }
+  - { id: 2, class: fpr }
+  - { id: 3, class: fpr }
+  - { id: 4, class: _ }
+  - { id: 5, class: fpr }
+  - { id: 6, class: _ }
+  - { id: 7, class: fpr }
+  - { id: 8, class: fpr }
+  - { id: 9, class: fpr }
+  - { id: 10, class: gpr }
+  - { id: 11, class: fpr }
+  - { id: 12, class: fpr }
+  - { id: 13, class: fpr }
+  - { id: 14, class: gpr }
+  - { id: 15, class: fpr }
+  - { id: 16, class: fpr }
+  - { id: 17, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $d0, $d1
+
+    ; CHECK-LABEL: name: test_v2i16_sle
+    ; CHECK: liveins: $d0, $d1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr64 = COPY $d0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr64 = COPY $d1
+    ; CHECK: [[MOVi32imm:%[0-9]+]]:gpr32 = MOVi32imm 16
+    ; CHECK: [[DEF:%[0-9]+]]:fpr128 = IMPLICIT_DEF
+    ; CHECK: [[INSERT_SUBREG:%[0-9]+]]:fpr128 = INSERT_SUBREG [[DEF]], [[MOVi32imm]], %subreg.ssub
+    ; CHECK: [[INSvi32gpr:%[0-9]+]]:fpr128 = INSvi32gpr [[INSERT_SUBREG]], 1, [[MOVi32imm]]
+    ; CHECK: [[COPY2:%[0-9]+]]:fpr64 = COPY [[INSvi32gpr]].dsub
+    ; CHECK: [[USHLv2i32_:%[0-9]+]]:fpr64 = USHLv2i32 [[COPY]], [[COPY2]]
+    ; CHECK: [[NEGv2i32_:%[0-9]+]]:fpr64 = NEGv2i32 [[COPY2]]
+    ; CHECK: [[SSHLv2i32_:%[0-9]+]]:fpr64 = SSHLv2i32 [[USHLv2i32_]], [[NEGv2i32_]]
+    ; CHECK: [[MOVi32imm1:%[0-9]+]]:gpr32 = MOVi32imm 16
+    ; CHECK: [[DEF1:%[0-9]+]]:fpr128 = IMPLICIT_DEF
+    ; CHECK: [[INSERT_SUBREG1:%[0-9]+]]:fpr128 = INSERT_SUBREG [[DEF1]], [[MOVi32imm1]], %subreg.ssub
+    ; CHECK: [[INSvi32gpr1:%[0-9]+]]:fpr128 = INSvi32gpr [[INSERT_SUBREG1]], 1, [[MOVi32imm1]]
+    ; CHECK: [[COPY3:%[0-9]+]]:fpr64 = COPY [[INSvi32gpr1]].dsub
+    ; CHECK: [[USHLv2i32_1:%[0-9]+]]:fpr64 = USHLv2i32 [[COPY1]], [[COPY3]]
+    ; CHECK: [[NEGv2i32_1:%[0-9]+]]:fpr64 = NEGv2i32 [[COPY3]]
+    ; CHECK: [[SSHLv2i32_1:%[0-9]+]]:fpr64 = SSHLv2i32 [[USHLv2i32_1]], [[NEGv2i32_1]]
+    ; CHECK: [[CMGEv2i32_:%[0-9]+]]:fpr64 = CMGEv2i32 [[SSHLv2i32_1]], [[SSHLv2i32_]]
+    ; CHECK: $d0 = COPY [[CMGEv2i32_]]
+    ; CHECK: RET_ReallyLR implicit $d0
+    %2:fpr(<2 x s32>) = COPY $d0
+    %3:fpr(<2 x s32>) = COPY $d1
+    %14:gpr(s32) = G_CONSTANT i32 16
+    %15:fpr(<2 x s32>) = G_BUILD_VECTOR %14(s32), %14(s32)
+    %16:fpr(<2 x s32>) = COPY %2(<2 x s32>)
+    %17:fpr(<2 x s32>) = G_SHL %16, %15(<2 x s32>)
+    %7:fpr(<2 x s32>) = G_ASHR %17, %15(<2 x s32>)
+    %10:gpr(s32) = G_CONSTANT i32 16
+    %11:fpr(<2 x s32>) = G_BUILD_VECTOR %10(s32), %10(s32)
+    %12:fpr(<2 x s32>) = COPY %3(<2 x s32>)
+    %13:fpr(<2 x s32>) = G_SHL %12, %11(<2 x s32>)
+    %8:fpr(<2 x s32>) = G_ASHR %13, %11(<2 x s32>)
+    %9:fpr(<2 x s32>) = G_ICMP intpred(sle), %7(<2 x s32>), %8
+    %5:fpr(<2 x s32>) = COPY %9(<2 x s32>)
+    $d0 = COPY %5(<2 x s32>)
+    RET_ReallyLR implicit $d0
+
+...
+---
+name:            test_v8i16_sle
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: fpr }
+  - { id: 1, class: fpr }
+  - { id: 2, class: _ }
+  - { id: 3, class: fpr }
+  - { id: 4, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $q0, $q1
+
+    ; CHECK-LABEL: name: test_v8i16_sle
+    ; CHECK: liveins: $q0, $q1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr128 = COPY $q0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr128 = COPY $q1
+    ; CHECK: [[CMGEv8i16_:%[0-9]+]]:fpr128 = CMGEv8i16 [[COPY1]], [[COPY]]
+    ; CHECK: [[XTNv8i8_:%[0-9]+]]:fpr64 = XTNv8i8 [[CMGEv8i16_]]
+    ; CHECK: $d0 = COPY [[XTNv8i8_]]
+    ; CHECK: RET_ReallyLR implicit $d0
+    %0:fpr(<8 x s16>) = COPY $q0
+    %1:fpr(<8 x s16>) = COPY $q1
+    %4:fpr(<8 x s16>) = G_ICMP intpred(sle), %0(<8 x s16>), %1
+    %3:fpr(<8 x s8>) = G_TRUNC %4(<8 x s16>)
+    $d0 = COPY %3(<8 x s8>)
+    RET_ReallyLR implicit $d0
+
+...
+---
+name:            test_v4i16_sle
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: fpr }
+  - { id: 1, class: fpr }
+  - { id: 2, class: _ }
+  - { id: 3, class: fpr }
+  - { id: 4, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $d0, $d1
+
+    ; CHECK-LABEL: name: test_v4i16_sle
+    ; CHECK: liveins: $d0, $d1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr64 = COPY $d0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr64 = COPY $d1
+    ; CHECK: [[CMGEv4i16_:%[0-9]+]]:fpr64 = CMGEv4i16 [[COPY1]], [[COPY]]
+    ; CHECK: $d0 = COPY [[CMGEv4i16_]]
+    ; CHECK: RET_ReallyLR implicit $d0
+    %0:fpr(<4 x s16>) = COPY $d0
+    %1:fpr(<4 x s16>) = COPY $d1
+    %4:fpr(<4 x s16>) = G_ICMP intpred(sle), %0(<4 x s16>), %1
+    %3:fpr(<4 x s16>) = COPY %4(<4 x s16>)
+    $d0 = COPY %3(<4 x s16>)
+    RET_ReallyLR implicit $d0
+
+...
+---
+name:            test_v16i8_sle
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: fpr }
+  - { id: 1, class: fpr }
+  - { id: 2, class: _ }
+  - { id: 3, class: fpr }
+  - { id: 4, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $q0, $q1
+
+    ; CHECK-LABEL: name: test_v16i8_sle
+    ; CHECK: liveins: $q0, $q1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr128 = COPY $q0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr128 = COPY $q1
+    ; CHECK: [[CMGEv16i8_:%[0-9]+]]:fpr128 = CMGEv16i8 [[COPY1]], [[COPY]]
+    ; CHECK: $q0 = COPY [[CMGEv16i8_]]
+    ; CHECK: RET_ReallyLR implicit $q0
+    %0:fpr(<16 x s8>) = COPY $q0
+    %1:fpr(<16 x s8>) = COPY $q1
+    %4:fpr(<16 x s8>) = G_ICMP intpred(sle), %0(<16 x s8>), %1
+    %3:fpr(<16 x s8>) = COPY %4(<16 x s8>)
+    $q0 = COPY %3(<16 x s8>)
+    RET_ReallyLR implicit $q0
+
+...
+---
+name:            test_v8i8_sle
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: fpr }
+  - { id: 1, class: fpr }
+  - { id: 2, class: _ }
+  - { id: 3, class: fpr }
+  - { id: 4, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1 (%ir-block.0):
+    liveins: $d0, $d1
+
+    ; CHECK-LABEL: name: test_v8i8_sle
+    ; CHECK: liveins: $d0, $d1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr64 = COPY $d0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr64 = COPY $d1
+    ; CHECK: [[CMGEv8i8_:%[0-9]+]]:fpr64 = CMGEv8i8 [[COPY1]], [[COPY]]
+    ; CHECK: $d0 = COPY [[CMGEv8i8_]]
+    ; CHECK: RET_ReallyLR implicit $d0
+    %0:fpr(<8 x s8>) = COPY $d0
+    %1:fpr(<8 x s8>) = COPY $d1
+    %4:fpr(<8 x s8>) = G_ICMP intpred(sle), %0(<8 x s8>), %1
+    %3:fpr(<8 x s8>) = COPY %4(<8 x s8>)
+    $d0 = COPY %3(<8 x s8>)
+    RET_ReallyLR implicit $d0
+
+...

llvm/test/CodeGen/AArch64/GlobalISel/select-vector-shift.mir

+# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
+# RUN: llc -mtriple=aarch64-- -run-pass=instruction-select -verify-machineinstrs %s -o - | FileCheck %s
+---
+name:            shl_v2i32
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: fpr }
+  - { id: 1, class: fpr }
+  - { id: 2, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1:
+    liveins: $d0, $d1
+
+    ; CHECK-LABEL: name: shl_v2i32
+    ; CHECK: liveins: $d0, $d1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr64 = COPY $d0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr64 = COPY $d1
+    ; CHECK: [[USHLv2i32_:%[0-9]+]]:fpr64 = USHLv2i32 [[COPY]], [[COPY1]]
+    ; CHECK: $d0 = COPY [[USHLv2i32_]]
+    ; CHECK: RET_ReallyLR implicit $d0
+    %0:fpr(<2 x s32>) = COPY $d0
+    %1:fpr(<2 x s32>) = COPY $d1
+    %2:fpr(<2 x s32>) = G_SHL %0, %1(<2 x s32>)
+    $d0 = COPY %2(<2 x s32>)
+    RET_ReallyLR implicit $d0
+
+...
+---
+name:            shl_v4i32
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: fpr }
+  - { id: 1, class: fpr }
+  - { id: 2, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1:
+    liveins: $q0, $q1
+
+    ; CHECK-LABEL: name: shl_v4i32
+    ; CHECK: liveins: $q0, $q1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr128 = COPY $q0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr128 = COPY $q1
+    ; CHECK: [[USHLv4i32_:%[0-9]+]]:fpr128 = USHLv4i32 [[COPY]], [[COPY1]]
+    ; CHECK: $q0 = COPY [[USHLv4i32_]]
+    ; CHECK: RET_ReallyLR implicit $q0
+    %0:fpr(<4 x s32>) = COPY $q0
+    %1:fpr(<4 x s32>) = COPY $q1
+    %2:fpr(<4 x s32>) = G_SHL %0, %1(<4 x s32>)
+    $q0 = COPY %2(<4 x s32>)
+    RET_ReallyLR implicit $q0
+
+...
+---
+name:            ashr_v2i32
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: fpr }
+  - { id: 1, class: fpr }
+  - { id: 2, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1:
+    liveins: $d0, $d1
+
+    ; CHECK-LABEL: name: ashr_v2i32
+    ; CHECK: liveins: $d0, $d1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr64 = COPY $d0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr64 = COPY $d1
+    ; CHECK: [[NEGv2i32_:%[0-9]+]]:fpr64 = NEGv2i32 [[COPY1]]
+    ; CHECK: [[SSHLv2i32_:%[0-9]+]]:fpr64 = SSHLv2i32 [[COPY]], [[NEGv2i32_]]
+    ; CHECK: $d0 = COPY [[SSHLv2i32_]]
+    ; CHECK: RET_ReallyLR implicit $d0
+    %0:fpr(<2 x s32>) = COPY $d0
+    %1:fpr(<2 x s32>) = COPY $d1
+    %2:fpr(<2 x s32>) = G_ASHR %0, %1(<2 x s32>)
+    $d0 = COPY %2(<2 x s32>)
+    RET_ReallyLR implicit $d0
+
+...
+---
+name:            ashr_v4i32
+alignment:       2
+legalized:       true
+regBankSelected: true
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: fpr }
+  - { id: 1, class: fpr }
+  - { id: 2, class: fpr }
+machineFunctionInfo: {}
+body:             |
+  bb.1:
+    liveins: $q0, $q1
+
+    ; CHECK-LABEL: name: ashr_v4i32
+    ; CHECK: liveins: $q0, $q1
+    ; CHECK: [[COPY:%[0-9]+]]:fpr128 = COPY $q0
+    ; CHECK: [[COPY1:%[0-9]+]]:fpr128 = COPY $q1
+    ; CHECK: [[NEGv4i32_:%[0-9]+]]:fpr128 = NEGv4i32 [[COPY1]]
+    ; CHECK: [[SSHLv4i32_:%[0-9]+]]:fpr128 = SSHLv4i32 [[COPY]], [[NEGv4i32_]]
+    ; CHECK: $q0 = COPY [[SSHLv4i32_]]
+    ; CHECK: RET_ReallyLR implicit $q0
+    %0:fpr(<4 x s32>) = COPY $q0
+    %1:fpr(<4 x s32>) = COPY $q1
+    %2:fpr(<4 x s32>) = G_ASHR %0, %1(<4 x s32>)
+    $q0 = COPY %2(<4 x s32>)
+    RET_ReallyLR implicit $q0
+
+...