LegalizeVectorTypes.cpp

Go to the documentation of this file.

//===------- LegalizeVectorTypes.cpp - Legalization of vector types -------===//
//
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
// See https://llvm.org/LICENSE.txt for license information.
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
//
//===----------------------------------------------------------------------===//
//
// This file performs vector type splitting and scalarization for LegalizeTypes.
// Scalarization is the act of changing a computation in an illegal one-element
// vector type to be a computation in its scalar element type.  For example,
// implementing <1 x f32> arithmetic in a scalar f32 register.  This is needed
// as a base case when scalarizing vector arithmetic like <4 x f32>, which
// eventually decomposes to scalars if the target doesn't support v4f32 or v2f32
// types.
// Splitting is the act of changing a computation in an invalid vector type to
// be a computation in two vectors of half the size.  For example, implementing
// <128 x f32> operations in terms of two <64 x f32> operations.
//
//===----------------------------------------------------------------------===//
 
#include "LegalizeTypes.h"
#include "llvm/ADT/SmallBitVector.h"
#include "llvm/Analysis/MemoryLocation.h"
#include "llvm/Analysis/VectorUtils.h"
#include "llvm/CodeGen/ISDOpcodes.h"
#include "llvm/IR/DataLayout.h"
#include "llvm/Support/ErrorHandling.h"
#include "llvm/Support/TypeSize.h"
#include "llvm/Support/raw_ostream.h"
#include <numeric>
 
using namespace llvm;
 
#define DEBUG_TYPE "legalize-types"
 
//===----------------------------------------------------------------------===//
//  Result Vector Scalarization: <1 x ty> -> ty.
//===----------------------------------------------------------------------===//
 
void DAGTypeLegalizer::ScalarizeVectorResult(SDNode *N, unsigned ResNo) {
  LLVM_DEBUG(dbgs() << "Scalarize node result " << ResNo << ": ";
             N->dump(&DAG));
  SDValue R = SDValue();
 
  switch (N->getOpcode()) {
  default:
#ifndef NDEBUG
    dbgs() << "ScalarizeVectorResult #" << ResNo << ": ";
    N->dump(&DAG);
    dbgs() << "\n";
#endif
    report_fatal_error("Do not know how to scalarize the result of this "
                       "operator!\n");
 
  case ISD::MERGE_VALUES:      R = ScalarizeVecRes_MERGE_VALUES(N, ResNo);break;
  case ISD::BITCAST:           R = ScalarizeVecRes_BITCAST(N); break;
  case ISD::BUILD_VECTOR:      R = ScalarizeVecRes_BUILD_VECTOR(N); break;
  case ISD::EXTRACT_SUBVECTOR: R = ScalarizeVecRes_EXTRACT_SUBVECTOR(N); break;
  case ISD::FP_ROUND:          R = ScalarizeVecRes_FP_ROUND(N); break;
  case ISD::FPOWI:             R = ScalarizeVecRes_ExpOp(N); break;
  case ISD::INSERT_VECTOR_ELT: R = ScalarizeVecRes_INSERT_VECTOR_ELT(N); break;
  case ISD::LOAD:           R = ScalarizeVecRes_LOAD(cast<LoadSDNode>(N));break;
  case ISD::SCALAR_TO_VECTOR:  R = ScalarizeVecRes_SCALAR_TO_VECTOR(N); break;
  case ISD::SIGN_EXTEND_INREG: R = ScalarizeVecRes_InregOp(N); break;
  case ISD::VSELECT:           R = ScalarizeVecRes_VSELECT(N); break;
  case ISD::SELECT:            R = ScalarizeVecRes_SELECT(N); break;
  case ISD::SELECT_CC:         R = ScalarizeVecRes_SELECT_CC(N); break;
  case ISD::SETCC:             R = ScalarizeVecRes_SETCC(N); break;
  case ISD::UNDEF:             R = ScalarizeVecRes_UNDEF(N); break;
  case ISD::VECTOR_SHUFFLE:    R = ScalarizeVecRes_VECTOR_SHUFFLE(N); break;
  case ISD::IS_FPCLASS:        R = ScalarizeVecRes_IS_FPCLASS(N); break;
  case ISD::ANY_EXTEND_VECTOR_INREG:
  case ISD::SIGN_EXTEND_VECTOR_INREG:
  case ISD::ZERO_EXTEND_VECTOR_INREG:
    R = ScalarizeVecRes_VecInregOp(N);
    break;
  case ISD::ABS:
  case ISD::ANY_EXTEND:
  case ISD::BITREVERSE:
  case ISD::BSWAP:
  case ISD::CTLZ:
  case ISD::CTLZ_ZERO_UNDEF:
  case ISD::CTPOP:
  case ISD::CTTZ:
  case ISD::CTTZ_ZERO_UNDEF:
  case ISD::FABS:
  case ISD::FACOS:
  case ISD::FASIN:
  case ISD::FATAN:
  case ISD::FCEIL:
  case ISD::FCOS:
  case ISD::FCOSH:
  case ISD::FEXP:
  case ISD::FEXP2:
  case ISD::FEXP10:
  case ISD::FFLOOR:
  case ISD::FLOG:
  case ISD::FLOG10:
  case ISD::FLOG2:
  case ISD::FNEARBYINT:
  case ISD::FNEG:
  case ISD::FREEZE:
  case ISD::ARITH_FENCE:
  case ISD::FP_EXTEND:
  case ISD::FP_TO_SINT:
  case ISD::FP_TO_UINT:
  case ISD::FRINT:
  case ISD::LRINT:
  case ISD::LLRINT:
  case ISD::FROUND:
  case ISD::FROUNDEVEN:
  case ISD::LROUND:
  case ISD::LLROUND:
  case ISD::FSIN:
  case ISD::FSINH:
  case ISD::FSQRT:
  case ISD::FTAN:
  case ISD::FTANH:
  case ISD::FTRUNC:
  case ISD::SIGN_EXTEND:
  case ISD::SINT_TO_FP:
  case ISD::TRUNCATE:
  case ISD::UINT_TO_FP:
  case ISD::ZERO_EXTEND:
  case ISD::FCANONICALIZE:
    R = ScalarizeVecRes_UnaryOp(N);
    break;
  case ISD::ADDRSPACECAST:
    R = ScalarizeVecRes_ADDRSPACECAST(N);
    break;
  case ISD::FFREXP:
    R = ScalarizeVecRes_FFREXP(N, ResNo);
    break;
  case ISD::ADD:
  case ISD::AND:
  case ISD::AVGCEILS:
  case ISD::AVGCEILU:
  case ISD::AVGFLOORS:
  case ISD::AVGFLOORU:
  case ISD::FADD:
  case ISD::FCOPYSIGN:
  case ISD::FDIV:
  case ISD::FMUL:
  case ISD::FMINNUM:
  case ISD::FMAXNUM:
  case ISD::FMINNUM_IEEE:
  case ISD::FMAXNUM_IEEE:
  case ISD::FMINIMUM:
  case ISD::FMAXIMUM:
  case ISD::FLDEXP:
  case ISD::ABDS:
  case ISD::ABDU:
  case ISD::SMIN:
  case ISD::SMAX:
  case ISD::UMIN:
  case ISD::UMAX:
 
  case ISD::SADDSAT:
  case ISD::UADDSAT:
  case ISD::SSUBSAT:
  case ISD::USUBSAT:
  case ISD::SSHLSAT:
  case ISD::USHLSAT:
 
  case ISD::FPOW:
  case ISD::FREM:
  case ISD::FSUB:
  case ISD::MUL:
  case ISD::MULHS:
  case ISD::MULHU:
  case ISD::OR:
  case ISD::SDIV:
  case ISD::SREM:
  case ISD::SUB:
  case ISD::UDIV:
  case ISD::UREM:
  case ISD::XOR:
  case ISD::SHL:
  case ISD::SRA:
  case ISD::SRL:
  case ISD::ROTL:
  case ISD::ROTR:
    R = ScalarizeVecRes_BinOp(N);
    break;
 
  case ISD::SCMP:
  case ISD::UCMP:
    R = ScalarizeVecRes_CMP(N);
    break;
 
  case ISD::FMA:
  case ISD::FSHL:
  case ISD::FSHR:
    R = ScalarizeVecRes_TernaryOp(N);
    break;
 
#define DAG_INSTRUCTION(NAME, NARG, ROUND_MODE, INTRINSIC, DAGN)               \
  case ISD::STRICT_##DAGN:
#include "llvm/IR/ConstrainedOps.def"
    R = ScalarizeVecRes_StrictFPOp(N);
    break;
 
  case ISD::FP_TO_UINT_SAT:
  case ISD::FP_TO_SINT_SAT:
    R = ScalarizeVecRes_FP_TO_XINT_SAT(N);
    break;
 
  case ISD::UADDO:
  case ISD::SADDO:
  case ISD::USUBO:
  case ISD::SSUBO:
  case ISD::UMULO:
  case ISD::SMULO:
    R = ScalarizeVecRes_OverflowOp(N, ResNo);
    break;
  case ISD::SMULFIX:
  case ISD::SMULFIXSAT:
  case ISD::UMULFIX:
  case ISD::UMULFIXSAT:
  case ISD::SDIVFIX:
  case ISD::SDIVFIXSAT:
  case ISD::UDIVFIX:
  case ISD::UDIVFIXSAT:
    R = ScalarizeVecRes_FIX(N);
    break;
  }
 
  // If R is null, the sub-method took care of registering the result.
  if (R.getNode())
    SetScalarizedVector(SDValue(N, ResNo), R);
}
 
SDValue DAGTypeLegalizer::ScalarizeVecRes_BinOp(SDNode *N) {
  SDValue LHS = GetScalarizedVector(N->getOperand(0));
  SDValue RHS = GetScalarizedVector(N->getOperand(1));
  return DAG.getNode(N->getOpcode(), SDLoc(N),
                     LHS.getValueType(), LHS, RHS, N->getFlags());
}
 
SDValue DAGTypeLegalizer::ScalarizeVecRes_CMP(SDNode *N) {
  SDLoc DL(N);
 
  SDValue LHS = N->getOperand(0);
  SDValue RHS = N->getOperand(1);
  if (getTypeAction(LHS.getValueType()) ==
      TargetLowering::TypeScalarizeVector) {
    LHS = GetScalarizedVector(LHS);
    RHS = GetScalarizedVector(RHS);
  } else {
    EVT VT = LHS.getValueType().getVectorElementType();
    LHS = DAG.getNode(ISD::EXTRACT_VECTOR_ELT, DL, VT, LHS,
                      DAG.getVectorIdxConstant(0, DL));
    RHS = DAG.getNode(ISD::EXTRACT_VECTOR_ELT, DL, VT, RHS,
                      DAG.getVectorIdxConstant(0, DL));
  }
 
  return DAG.getNode(N->getOpcode(), SDLoc(N),
                     N->getValueType(0).getVectorElementType(), LHS, RHS);
}
 
SDValue DAGTypeLegalizer::ScalarizeVecRes_TernaryOp(SDNode *N) {
  SDValue Op0 = GetScalarizedVector(N->getOperand(0));
  SDValue Op1 = GetScalarizedVector(N->getOperand(1));
  SDValue Op2 = GetScalarizedVector(N->getOperand(2));
  return DAG.getNode(N->getOpcode(), SDLoc(N), Op0.getValueType(), Op0, Op1,
                     Op2, N->getFlags());
}
 
SDValue DAGTypeLegalizer::ScalarizeVecRes_FIX(SDNode *N) {
  SDValue Op0 = GetScalarizedVector(N->getOperand(0));
  SDValue Op1 = GetScalarizedVector(N->getOperand(1));
  SDValue Op2 = N->getOperand(2);
  return DAG.getNode(N->getOpcode(), SDLoc(N), Op0.getValueType(), Op0, Op1,
                     Op2, N->getFlags());
}
 
SDValue DAGTypeLegalizer::ScalarizeVecRes_FFREXP(SDNode *N, unsigned ResNo) {
  assert(N->getValueType(0).getVectorNumElements() == 1 &&
         "Unexpected vector type!");
  SDValue Elt = GetScalarizedVector(N->getOperand(0));
 
  EVT VT0 = N->getValueType(0);
  EVT VT1 = N->getValueType(1);
  SDLoc dl(N);
 
  SDNode *ScalarNode =
      DAG.getNode(N->getOpcode(), dl,
                  {VT0.getScalarType(), VT1.getScalarType()}, Elt)
          .getNode();
 
  // Replace the other vector result not being explicitly scalarized here.
  unsigned OtherNo = 1 - ResNo;
  EVT OtherVT = N->getValueType(OtherNo);
  if (getTypeAction(OtherVT) == TargetLowering::TypeScalarizeVector) {
    SetScalarizedVector(SDValue(N, OtherNo), SDValue(ScalarNode, OtherNo));
  } else {
    SDValue OtherVal = DAG.getNode(ISD::SCALAR_TO_VECTOR, dl, OtherVT,
                                   SDValue(ScalarNode, OtherNo));
    ReplaceValueWith(SDValue(N, OtherNo), OtherVal);
  }
 
  return SDValue(ScalarNode, ResNo);
}
 
SDValue DAGTypeLegalizer::ScalarizeVecRes_StrictFPOp(SDNode *N) {
  EVT VT = N->getValueType(0).getVectorElementType();
  unsigned NumOpers = N->getNumOperands();
  SDValue Chain = N->getOperand(0);
  EVT ValueVTs[] = {VT, MVT::Other};
  SDLoc dl(N);
 
  SmallVector<SDValue, 4> Opers(NumOpers);
 
  // The Chain is the first operand.
  Opers[0] = Chain;
 
  // Now process the remaining operands.
  for (unsigned i = 1; i < NumOpers; ++i) {
    SDValue Oper = N->getOperand(i);
    EVT OperVT = Oper.getValueType();
 
    if (OperVT.isVector()) {
      if (getTypeAction(OperVT) == TargetLowering::TypeScalarizeVector)
        Oper = GetScalarizedVector(Oper);
      else
        Oper = DAG.getNode(ISD::EXTRACT_VECTOR_ELT, dl,
                           OperVT.getVectorElementType(), Oper,
                           DAG.getVectorIdxConstant(0, dl));
    }
 
    Opers[i] = Oper;
  }
 
  SDValue Result = DAG.getNode(N->getOpcode(), dl, DAG.getVTList(ValueVTs),
                               Opers, N->getFlags());
 
  // Legalize the chain result - switch anything that used the old chain to
  // use the new one.
  ReplaceValueWith(SDValue(N, 1), Result.getValue(1));
  return Result;
}
 
SDValue DAGTypeLegalizer::ScalarizeVecRes_OverflowOp(SDNode *N,
                                                     unsigned ResNo) {
  SDLoc DL(N);
  EVT ResVT = N->getValueType(0);
  EVT OvVT = N->getValueType(1);
 
  SDValue ScalarLHS, ScalarRHS;
  if (getTypeAction(ResVT) == TargetLowering::TypeScalarizeVector) {
    ScalarLHS = GetScalarizedVector(N->getOperand(0));
    ScalarRHS = GetScalarizedVector(N->getOperand(1));
  } else {
    SmallVector<SDValue, 1> ElemsLHS, ElemsRHS;
    DAG.ExtractVectorElements(N->getOperand(0), ElemsLHS);
    DAG.ExtractVectorElements(N->getOperand(1), ElemsRHS);
    ScalarLHS = ElemsLHS[0];
    ScalarRHS = ElemsRHS[0];
  }
 
  SDVTList ScalarVTs = DAG.getVTList(
      ResVT.getVectorElementType(), OvVT.getVectorElementType());
  SDNode *ScalarNode = DAG.getNode(
      N->getOpcode(), DL, ScalarVTs, ScalarLHS, ScalarRHS).getNode();
  ScalarNode->setFlags(N->getFlags());
 
  // Replace the other vector result not being explicitly scalarized here.
  unsigned OtherNo = 1 - ResNo;
  EVT OtherVT = N->getValueType(OtherNo);
  if (getTypeAction(OtherVT) == TargetLowering::TypeScalarizeVector) {
    SetScalarizedVector(SDValue(N, OtherNo), SDValue(ScalarNode, OtherNo));
  } else {
    SDValue OtherVal = DAG.getNode(
        ISD::SCALAR_TO_VECTOR, DL, OtherVT, SDValue(ScalarNode, OtherNo));
    ReplaceValueWith(SDValue(N, OtherNo), OtherVal);
  }
 
  return SDValue(ScalarNode, ResNo);
}
 
SDValue DAGTypeLegalizer::ScalarizeVecRes_MERGE_VALUES(SDNode *N,
                                                       unsigned ResNo) {
  SDValue Op = DisintegrateMERGE_VALUES(N, ResNo);
  return GetScalarizedVector(Op);
}
 
SDValue DAGTypeLegalizer::ScalarizeVecRes_BITCAST(SDNode *N) {
  SDValue Op = N->getOperand(0);
  if (Op.getValueType().isVector()
      && Op.getValueType().getVectorNumElements() == 1
      && !isSimpleLegalType(Op.getValueType()))
    Op = GetScalarizedVector(Op);
  EVT NewVT = N->getValueType(0).getVectorElementType();
  return DAG.getNode(ISD::BITCAST, SDLoc(N),
                     NewVT, Op);
}
 
SDValue DAGTypeLegalizer::ScalarizeVecRes_BUILD_VECTOR(SDNode *N) {
  EVT EltVT = N->getValueType(0).getVectorElementType();
  SDValue InOp = N->getOperand(0);
  // The BUILD_VECTOR operands may be of wider element types and
  // we may need to truncate them back to the requested return type.
  if (EltVT.isInteger())
    return DAG.getNode(ISD::TRUNCATE, SDLoc(N), EltVT, InOp);
  return InOp;
}
 
SDValue DAGTypeLegalizer::ScalarizeVecRes_EXTRACT_SUBVECTOR(SDNode *N) {
  return DAG.getNode(ISD::EXTRACT_VECTOR_ELT, SDLoc(N),
                     N->getValueType(0).getVectorElementType(),
                     N->getOperand(0), N->getOperand(1));
}
 
SDValue DAGTypeLegalizer::ScalarizeVecRes_FP_ROUND(SDNode *N) {
  SDLoc DL(N);
  SDValue Op = N->getOperand(0);
  EVT OpVT = Op.getValueType();
  // The result needs scalarizing, but it's not a given that the source does.
  // See similar logic in ScalarizeVecRes_UnaryOp.
  if (getTypeAction(OpVT) == TargetLowering::TypeScalarizeVector) {
    Op = GetScalarizedVector(Op);
  } else {
    EVT VT = OpVT.getVectorElementType();
    Op = DAG.getNode(ISD::EXTRACT_VECTOR_ELT, DL, VT, Op,
                     DAG.getVectorIdxConstant(0, DL));
  }
  return DAG.getNode(ISD::FP_ROUND, DL,
                     N->getValueType(0).getVectorElementType(), Op,
                     N->getOperand(1));
}
 
SDValue DAGTypeLegalizer::ScalarizeVecRes_ExpOp(SDNode *N) {
  SDValue Op = GetScalarizedVector(N->getOperand(0));
  return DAG.getNode(N->getOpcode(), SDLoc(N), Op.getValueType(), Op,
                     N->getOperand(1));
}
 
SDValue DAGTypeLegalizer::ScalarizeVecRes_INSERT_VECTOR_ELT(SDNode *N) {
  // The value to insert may have a wider type than the vector element type,
  // so be sure to truncate it to the element type if necessary.
  SDValue Op = N->getOperand(1);
  EVT EltVT = N->getValueType(0).getVectorElementType();
  if (Op.getValueType() != EltVT)
    // FIXME: Can this happen for floating point types?
    Op = DAG.getNode(ISD::TRUNCATE, SDLoc(N), EltVT, Op);
  return Op;
}
 
SDValue DAGTypeLegalizer::ScalarizeVecRes_LOAD(LoadSDNode *N) {
  assert(N->isUnindexed() && "Indexed vector load?");
 
  SDValue Result = DAG.getLoad(
      ISD::UNINDEXED, N->getExtensionType(),
      N->getValueType(0).getVectorElementType(), SDLoc(N), N->getChain(),
      N->getBasePtr(), DAG.getUNDEF(N->getBasePtr().getValueType()),
      N->getPointerInfo(), N->getMemoryVT().getVectorElementType(),
      N->getOriginalAlign(), N->getMemOperand()->getFlags(), N->getAAInfo());
 
  // Legalize the chain result - switch anything that used the old chain to
  // use the new one.
  ReplaceValueWith(SDValue(N, 1), Result.getValue(1));
  return Result;
}
 
SDValue DAGTypeLegalizer::ScalarizeVecRes_UnaryOp(SDNode *N) {
  // Get the dest type - it doesn't always match the input type, e.g. int_to_fp.
  EVT DestVT = N->getValueType(0).getVectorElementType();
  SDValue Op = N->getOperand(0);
  EVT OpVT = Op.getValueType();
  SDLoc DL(N);
  // The result needs scalarizing, but it's not a given that the source does.
  // This is a workaround for targets where it's impossible to scalarize the
  // result of a conversion, because the source type is legal.
  // For instance, this happens on AArch64: v1i1 is illegal but v1i{8,16,32}
  // are widened to v8i8, v4i16, and v2i32, which is legal, because v1i64 is
  // legal and was not scalarized.
  // See the similar logic in ScalarizeVecRes_SETCC
  if (getTypeAction(OpVT) == TargetLowering::TypeScalarizeVector) {
    Op = GetScalarizedVector(Op);
  } else {
    EVT VT = OpVT.getVectorElementType();
    Op = DAG.getNode(ISD::EXTRACT_VECTOR_ELT, DL, VT, Op,
                     DAG.getVectorIdxConstant(0, DL));
  }
  return DAG.getNode(N->getOpcode(), SDLoc(N), DestVT, Op, N->getFlags());
}
 
SDValue DAGTypeLegalizer::ScalarizeVecRes_InregOp(SDNode *N) {
  EVT EltVT = N->getValueType(0).getVectorElementType();
  EVT ExtVT = cast<VTSDNode>(N->getOperand(1))->getVT().getVectorElementType();
  SDValue LHS = GetScalarizedVector(N->getOperand(0));
  return DAG.getNode(N->getOpcode(), SDLoc(N), EltVT,
                     LHS, DAG.getValueType(ExtVT));
}
 
SDValue DAGTypeLegalizer::ScalarizeVecRes_VecInregOp(SDNode *N) {
  SDLoc DL(N);
  SDValue Op = N->getOperand(0);
 
  EVT OpVT = Op.getValueType();
  EVT OpEltVT = OpVT.getVectorElementType();
  EVT EltVT = N->getValueType(0).getVectorElementType();
 
  if (getTypeAction(OpVT) == TargetLowering::TypeScalarizeVector) {
    Op = GetScalarizedVector(Op);
  } else {
    Op = DAG.getNode(ISD::EXTRACT_VECTOR_ELT, DL, OpEltVT, Op,
                     DAG.getVectorIdxConstant(0, DL));
  }
 
  switch (N->getOpcode()) {
  case ISD::ANY_EXTEND_VECTOR_INREG:
    return DAG.getNode(ISD::ANY_EXTEND, DL, EltVT, Op);
  case ISD::SIGN_EXTEND_VECTOR_INREG:
    return DAG.getNode(ISD::SIGN_EXTEND, DL, EltVT, Op);
  case ISD::ZERO_EXTEND_VECTOR_INREG:
    return DAG.getNode(ISD::ZERO_EXTEND, DL, EltVT, Op);
  }
 
  llvm_unreachable("Illegal extend_vector_inreg opcode");
}
 
SDValue DAGTypeLegalizer::ScalarizeVecRes_ADDRSPACECAST(SDNode *N) {
  EVT DestVT = N->getValueType(0).getVectorElementType();
  SDValue Op = N->getOperand(0);
  EVT OpVT = Op.getValueType();
  SDLoc DL(N);
  // The result needs scalarizing, but it's not a given that the source does.
  // This is a workaround for targets where it's impossible to scalarize the
  // result of a conversion, because the source type is legal.
  // For instance, this happens on AArch64: v1i1 is illegal but v1i{8,16,32}
  // are widened to v8i8, v4i16, and v2i32, which is legal, because v1i64 is
  // legal and was not scalarized.
  // See the similar logic in ScalarizeVecRes_SETCC
  if (getTypeAction(OpVT) == TargetLowering::TypeScalarizeVector) {
    Op = GetScalarizedVector(Op);
  } else {
    EVT VT = OpVT.getVectorElementType();
    Op = DAG.getNode(ISD::EXTRACT_VECTOR_ELT, DL, VT, Op,
                     DAG.getVectorIdxConstant(0, DL));
  }
  auto *AddrSpaceCastN = cast<AddrSpaceCastSDNode>(N);
  unsigned SrcAS = AddrSpaceCastN->getSrcAddressSpace();
  unsigned DestAS = AddrSpaceCastN->getDestAddressSpace();
  return DAG.getAddrSpaceCast(DL, DestVT, Op, SrcAS, DestAS);
}
 
SDValue DAGTypeLegalizer::ScalarizeVecRes_SCALAR_TO_VECTOR(SDNode *N) {
  // If the operand is wider than the vector element type then it is implicitly
  // truncated.  Make that explicit here.
  EVT EltVT = N->getValueType(0).getVectorElementType();
  SDValue InOp = N->getOperand(0);
  if (InOp.getValueType() != EltVT)
    return DAG.getNode(ISD::TRUNCATE, SDLoc(N), EltVT, InOp);
  return InOp;
}
 
SDValue DAGTypeLegalizer::ScalarizeVecRes_VSELECT(SDNode *N) {
  SDValue Cond = N->getOperand(0);
  EVT OpVT = Cond.getValueType();
  SDLoc DL(N);
  // The vselect result and true/value operands needs scalarizing, but it's
  // not a given that the Cond does. For instance, in AVX512 v1i1 is legal.
  // See the similar logic in ScalarizeVecRes_SETCC
  if (getTypeAction(OpVT) == TargetLowering::TypeScalarizeVector) {
    Cond = GetScalarizedVector(Cond);
  } else {
    EVT VT = OpVT.getVectorElementType();
    Cond = DAG.getNode(ISD::EXTRACT_VECTOR_ELT, DL, VT, Cond,
                       DAG.getVectorIdxConstant(0, DL));
  }
 
  SDValue LHS = GetScalarizedVector(N->getOperand(1));
  TargetLowering::BooleanContent ScalarBool =
      TLI.getBooleanContents(false, false);
  TargetLowering::BooleanContent VecBool = TLI.getBooleanContents(true, false);
 
  // If integer and float booleans have different contents then we can't
  // reliably optimize in all cases. There is a full explanation for this in
  // DAGCombiner::visitSELECT() where the same issue affects folding
  // (select C, 0, 1) to (xor C, 1).
  if (TLI.getBooleanContents(false, false) !=
      TLI.getBooleanContents(false, true)) {
    // At least try the common case where the boolean is generated by a
    // comparison.
    if (Cond->getOpcode() == ISD::SETCC) {
      EVT OpVT = Cond->getOperand(0).getValueType();
      ScalarBool = TLI.getBooleanContents(OpVT.getScalarType());
      VecBool = TLI.getBooleanContents(OpVT);
    } else
      ScalarBool = TargetLowering::UndefinedBooleanContent;
  }
 
  EVT CondVT = Cond.getValueType();
  if (ScalarBool != VecBool) {
    switch (ScalarBool) {
      case TargetLowering::UndefinedBooleanContent:
        break;
      case TargetLowering::ZeroOrOneBooleanContent:
        assert(VecBool == TargetLowering::UndefinedBooleanContent ||
               VecBool == TargetLowering::ZeroOrNegativeOneBooleanContent);
        // Vector read from all ones, scalar expects a single 1 so mask.
        Cond = DAG.getNode(ISD::AND, SDLoc(N), CondVT,
                           Cond, DAG.getConstant(1, SDLoc(N), CondVT));
        break;
      case TargetLowering::ZeroOrNegativeOneBooleanContent:
        assert(VecBool == TargetLowering::UndefinedBooleanContent ||
               VecBool == TargetLowering::ZeroOrOneBooleanContent);
        // Vector reads from a one, scalar from all ones so sign extend.
        Cond = DAG.getNode(ISD::SIGN_EXTEND_INREG, SDLoc(N), CondVT,
                           Cond, DAG.getValueType(MVT::i1));
        break;
    }
  }
 
  // Truncate the condition if needed
  auto BoolVT = getSetCCResultType(CondVT);
  if (BoolVT.bitsLT(CondVT))
    Cond = DAG.getNode(ISD::TRUNCATE, SDLoc(N), BoolVT, Cond);
 
  return DAG.getSelect(SDLoc(N),
                       LHS.getValueType(), Cond, LHS,
                       GetScalarizedVector(N->getOperand(2)));
}
 
SDValue DAGTypeLegalizer::ScalarizeVecRes_SELECT(SDNode *N) {
  SDValue LHS = GetScalarizedVector(N->getOperand(1));
  return DAG.getSelect(SDLoc(N),
                       LHS.getValueType(), N->getOperand(0), LHS,
                       GetScalarizedVector(N->getOperand(2)));
}
 
SDValue DAGTypeLegalizer::ScalarizeVecRes_SELECT_CC(SDNode *N) {
  SDValue LHS = GetScalarizedVector(N->getOperand(2));
  return DAG.getNode(ISD::SELECT_CC, SDLoc(N), LHS.getValueType(),
                     N->getOperand(0), N->getOperand(1),
                     LHS, GetScalarizedVector(N->getOperand(3)),
                     N->getOperand(4));
}
 
SDValue DAGTypeLegalizer::ScalarizeVecRes_UNDEF(SDNode *N) {
  return DAG.getUNDEF(N->getValueType(0).getVectorElementType());
}
 
SDValue DAGTypeLegalizer::ScalarizeVecRes_VECTOR_SHUFFLE(SDNode *N) {
  // Figure out if the scalar is the LHS or RHS and return it.
  SDValue Arg = N->getOperand(2).getOperand(0);
  if (Arg.isUndef())
    return DAG.getUNDEF(N->getValueType(0).getVectorElementType());
  unsigned Op = !cast<ConstantSDNode>(Arg)->isZero();
  return GetScalarizedVector(N->getOperand(Op));
}
 
SDValue DAGTypeLegalizer::ScalarizeVecRes_FP_TO_XINT_SAT(SDNode *N) {
  SDValue Src = N->getOperand(0);
  EVT SrcVT = Src.getValueType();
  SDLoc dl(N);
 
  // Handle case where result is scalarized but operand is not
  if (getTypeAction(SrcVT) == TargetLowering::TypeScalarizeVector)
    Src = GetScalarizedVector(Src);
  else
    Src = DAG.getNode(
        ISD::EXTRACT_VECTOR_ELT, dl, SrcVT.getVectorElementType(), Src,
        DAG.getConstant(0, dl, TLI.getVectorIdxTy(DAG.getDataLayout())));
 
  EVT DstVT = N->getValueType(0).getVectorElementType();
  return DAG.getNode(N->getOpcode(), dl, DstVT, Src, N->getOperand(1));
}
 
SDValue DAGTypeLegalizer::ScalarizeVecRes_SETCC(SDNode *N) {
  assert(N->getValueType(0).isVector() &&
         N->getOperand(0).getValueType().isVector() &&
         "Operand types must be vectors");
  SDValue LHS = N->getOperand(0);
  SDValue RHS = N->getOperand(1);
  EVT OpVT = LHS.getValueType();
  EVT NVT = N->getValueType(0).getVectorElementType();
  SDLoc DL(N);
 
  // The result needs scalarizing, but it's not a given that the source does.
  if (getTypeAction(OpVT) == TargetLowering::TypeScalarizeVector) {
    LHS = GetScalarizedVector(LHS);
    RHS = GetScalarizedVector(RHS);
  } else {
    EVT VT = OpVT.getVectorElementType();
    LHS = DAG.getNode(ISD::EXTRACT_VECTOR_ELT, DL, VT, LHS,
                      DAG.getVectorIdxConstant(0, DL));
    RHS = DAG.getNode(ISD::EXTRACT_VECTOR_ELT, DL, VT, RHS,
                      DAG.getVectorIdxConstant(0, DL));
  }
 
  // Turn it into a scalar SETCC.
  SDValue Res = DAG.getNode(ISD::SETCC, DL, MVT::i1, LHS, RHS,
                            N->getOperand(2));
  // Vectors may have a different boolean contents to scalars.  Promote the
  // value appropriately.
  ISD::NodeType ExtendCode =
      TargetLowering::getExtendForContent(TLI.getBooleanContents(OpVT));
  return DAG.getNode(ExtendCode, DL, NVT, Res);
}
 
SDValue DAGTypeLegalizer::ScalarizeVecRes_IS_FPCLASS(SDNode *N) {
  SDLoc DL(N);
  SDValue Arg = N->getOperand(0);
  SDValue Test = N->getOperand(1);
  EVT ArgVT = Arg.getValueType();
  EVT ResultVT = N->getValueType(0).getVectorElementType();
 
  if (getTypeAction(ArgVT) == TargetLowering::TypeScalarizeVector) {
    Arg = GetScalarizedVector(Arg);
  } else {
    EVT VT = ArgVT.getVectorElementType();
    Arg = DAG.getNode(ISD::EXTRACT_VECTOR_ELT, DL, VT, Arg,
                      DAG.getVectorIdxConstant(0, DL));
  }
 
  SDValue Res =
      DAG.getNode(ISD::IS_FPCLASS, DL, MVT::i1, {Arg, Test}, N->getFlags());
  // Vectors may have a different boolean contents to scalars.  Promote the
  // value appropriately.
  ISD::NodeType ExtendCode =
      TargetLowering::getExtendForContent(TLI.getBooleanContents(ArgVT));
  return DAG.getNode(ExtendCode, DL, ResultVT, Res);
}
 
//===----------------------------------------------------------------------===//
//  Operand Vector Scalarization <1 x ty> -> ty.
//===----------------------------------------------------------------------===//
 
bool DAGTypeLegalizer::ScalarizeVectorOperand(SDNode *N, unsigned OpNo) {
  LLVM_DEBUG(dbgs() << "Scalarize node operand " << OpNo << ": ";
             N->dump(&DAG));
  SDValue Res = SDValue();
 
  switch (N->getOpcode()) {
  default:
#ifndef NDEBUG
    dbgs() << "ScalarizeVectorOperand Op #" << OpNo << ": ";
    N->dump(&DAG);
    dbgs() << "\n";
#endif
    report_fatal_error("Do not know how to scalarize this operator's "
                       "operand!\n");
  case ISD::BITCAST:
    Res = ScalarizeVecOp_BITCAST(N);
    break;
  case ISD::ANY_EXTEND:
  case ISD::ZERO_EXTEND:
  case ISD::SIGN_EXTEND:
  case ISD::TRUNCATE:
  case ISD::FP_TO_SINT:
  case ISD::FP_TO_UINT:
  case ISD::SINT_TO_FP:
  case ISD::UINT_TO_FP:
  case ISD::LROUND:
  case ISD::LLROUND:
  case ISD::LRINT:
  case ISD::LLRINT:
    Res = ScalarizeVecOp_UnaryOp(N);
    break;
  case ISD::STRICT_SINT_TO_FP:
  case ISD::STRICT_UINT_TO_FP:
  case ISD::STRICT_FP_TO_SINT:
  case ISD::STRICT_FP_TO_UINT:
    Res = ScalarizeVecOp_UnaryOp_StrictFP(N);
    break;
  case ISD::CONCAT_VECTORS:
    Res = ScalarizeVecOp_CONCAT_VECTORS(N);
    break;
  case ISD::INSERT_SUBVECTOR:
    Res = ScalarizeVecOp_INSERT_SUBVECTOR(N, OpNo);
    break;
  case ISD::EXTRACT_VECTOR_ELT:
    Res = ScalarizeVecOp_EXTRACT_VECTOR_ELT(N);
    break;
  case ISD::VSELECT:
    Res = ScalarizeVecOp_VSELECT(N);
    break;
  case ISD::SETCC:
    Res = ScalarizeVecOp_VSETCC(N);
    break;
  case ISD::STORE:
    Res = ScalarizeVecOp_STORE(cast<StoreSDNode>(N), OpNo);
    break;
  case ISD::STRICT_FP_ROUND:
    Res = ScalarizeVecOp_STRICT_FP_ROUND(N, OpNo);
    break;
  case ISD::FP_ROUND:
    Res = ScalarizeVecOp_FP_ROUND(N, OpNo);
    break;
  case ISD::STRICT_FP_EXTEND:
    Res = ScalarizeVecOp_STRICT_FP_EXTEND(N);
    break;
  case ISD::FP_EXTEND:
    Res = ScalarizeVecOp_FP_EXTEND(N);
    break;
  case ISD::VECREDUCE_FADD:
  case ISD::VECREDUCE_FMUL:
  case ISD::VECREDUCE_ADD:
  case ISD::VECREDUCE_MUL:
  case ISD::VECREDUCE_AND:
  case ISD::VECREDUCE_OR:
  case ISD::VECREDUCE_XOR:
  case ISD::VECREDUCE_SMAX:
  case ISD::VECREDUCE_SMIN:
  case ISD::VECREDUCE_UMAX:
  case ISD::VECREDUCE_UMIN:
  case ISD::VECREDUCE_FMAX:
  case ISD::VECREDUCE_FMIN:
  case ISD::VECREDUCE_FMAXIMUM:
  case ISD::VECREDUCE_FMINIMUM:
    Res = ScalarizeVecOp_VECREDUCE(N);
    break;
  case ISD::VECREDUCE_SEQ_FADD:
  case ISD::VECREDUCE_SEQ_FMUL:
    Res = ScalarizeVecOp_VECREDUCE_SEQ(N);
    break;
  case ISD::SCMP:
  case ISD::UCMP:
    Res = ScalarizeVecOp_CMP(N);
    break;
  }
 
  // If the result is null, the sub-method took care of registering results etc.
  if (!Res.getNode()) return false;
 
  // If the result is N, the sub-method updated N in place.  Tell the legalizer
  // core about this.
  if (Res.getNode() == N)
    return true;
 
  assert(Res.getValueType() == N->getValueType(0) && N->getNumValues() == 1 &&
         "Invalid operand expansion");
 
  ReplaceValueWith(SDValue(N, 0), Res);
  return false;
}
 
/// If the value to convert is a vector that needs to be scalarized, it must be
/// <1 x ty>. Convert the element instead.
SDValue DAGTypeLegalizer::ScalarizeVecOp_BITCAST(SDNode *N) {
  SDValue Elt = GetScalarizedVector(N->getOperand(0));
  return DAG.getNode(ISD::BITCAST, SDLoc(N),
                     N->getValueType(0), Elt);
}
 
/// If the input is a vector that needs to be scalarized, it must be <1 x ty>.
/// Do the operation on the element instead.
SDValue DAGTypeLegalizer::ScalarizeVecOp_UnaryOp(SDNode *N) {
  assert(N->getValueType(0).getVectorNumElements() == 1 &&
         "Unexpected vector type!");
  SDValue Elt = GetScalarizedVector(N->getOperand(0));
  SDValue Op = DAG.getNode(N->getOpcode(), SDLoc(N),
                           N->getValueType(0).getScalarType(), Elt);
  // Revectorize the result so the types line up with what the uses of this
  // expression expect.
  return DAG.getNode(ISD::SCALAR_TO_VECTOR, SDLoc(N), N->getValueType(0), Op);
}
 
/// If the input is a vector that needs to be scalarized, it must be <1 x ty>.
/// Do the strict FP operation on the element instead.
SDValue DAGTypeLegalizer::ScalarizeVecOp_UnaryOp_StrictFP(SDNode *N) {
  assert(N->getValueType(0).getVectorNumElements() == 1 &&
         "Unexpected vector type!");
  SDValue Elt = GetScalarizedVector(N->getOperand(1));
  SDValue Res = DAG.getNode(N->getOpcode(), SDLoc(N),
                            { N->getValueType(0).getScalarType(), MVT::Other },
                            { N->getOperand(0), Elt });
  // Legalize the chain result - switch anything that used the old chain to
  // use the new one.
  ReplaceValueWith(SDValue(N, 1), Res.getValue(1));
  // Revectorize the result so the types line up with what the uses of this
  // expression expect.
  Res = DAG.getNode(ISD::SCALAR_TO_VECTOR, SDLoc(N), N->getValueType(0), Res);
 
  // Do our own replacement and return SDValue() to tell the caller that we
  // handled all replacements since caller can only handle a single result.
  ReplaceValueWith(SDValue(N, 0), Res);
  return SDValue();
}
 
/// The vectors to concatenate have length one - use a BUILD_VECTOR instead.
SDValue DAGTypeLegalizer::ScalarizeVecOp_CONCAT_VECTORS(SDNode *N) {
  SmallVector<SDValue, 8> Ops(N->getNumOperands());
  for (unsigned i = 0, e = N->getNumOperands(); i < e; ++i)
    Ops[i] = GetScalarizedVector(N->getOperand(i));
  return DAG.getBuildVector(N->getValueType(0), SDLoc(N), Ops);
}
 
/// The inserted subvector is to be scalarized - use insert vector element
/// instead.
SDValue DAGTypeLegalizer::ScalarizeVecOp_INSERT_SUBVECTOR(SDNode *N,
                                                          unsigned OpNo) {
  // We should not be attempting to scalarize the containing vector
  assert(OpNo == 1);
  SDValue Elt = GetScalarizedVector(N->getOperand(1));
  SDValue ContainingVec = N->getOperand(0);
  return DAG.getNode(ISD::INSERT_VECTOR_ELT, SDLoc(N),
                     ContainingVec.getValueType(), ContainingVec, Elt,
                     N->getOperand(2));
}
 
/// If the input is a vector that needs to be scalarized, it must be <1 x ty>,
/// so just return the element, ignoring the index.
SDValue DAGTypeLegalizer::ScalarizeVecOp_EXTRACT_VECTOR_ELT(SDNode *N) {
  EVT VT = N->getValueType(0);
  SDValue Res = GetScalarizedVector(N->getOperand(0));
  if (Res.getValueType() != VT)
    Res = VT.isFloatingPoint()
              ? DAG.getNode(ISD::FP_EXTEND, SDLoc(N), VT, Res)
              : DAG.getNode(ISD::ANY_EXTEND, SDLoc(N), VT, Res);
  return Res;
}
 
/// If the input condition is a vector that needs to be scalarized, it must be
/// <1 x i1>, so just convert to a normal ISD::SELECT
/// (still with vector output type since that was acceptable if we got here).
SDValue DAGTypeLegalizer::ScalarizeVecOp_VSELECT(SDNode *N) {
  SDValue ScalarCond = GetScalarizedVector(N->getOperand(0));
  EVT VT = N->getValueType(0);
 
  return DAG.getNode(ISD::SELECT, SDLoc(N), VT, ScalarCond, N->getOperand(1),
                     N->getOperand(2));
}
 
/// If the operand is a vector that needs to be scalarized then the
/// result must be v1i1, so just convert to a scalar SETCC and wrap
/// with a scalar_to_vector since the res type is legal if we got here
SDValue DAGTypeLegalizer::ScalarizeVecOp_VSETCC(SDNode *N) {
  assert(N->getValueType(0).isVector() &&
         N->getOperand(0).getValueType().isVector() &&
         "Operand types must be vectors");
  assert(N->getValueType(0) == MVT::v1i1 && "Expected v1i1 type");
 
  EVT VT = N->getValueType(0);
  SDValue LHS = GetScalarizedVector(N->getOperand(0));
  SDValue RHS = GetScalarizedVector(N->getOperand(1));
 
  EVT OpVT = N->getOperand(0).getValueType();
  EVT NVT = VT.getVectorElementType();
  SDLoc DL(N);
  // Turn it into a scalar SETCC.
  SDValue Res = DAG.getNode(ISD::SETCC, DL, MVT::i1, LHS, RHS,
      N->getOperand(2));
 
  // Vectors may have a different boolean contents to scalars.  Promote the
  // value appropriately.
  ISD::NodeType ExtendCode =
      TargetLowering::getExtendForContent(TLI.getBooleanContents(OpVT));
 
  Res = DAG.getNode(ExtendCode, DL, NVT, Res);
 
  return DAG.getNode(ISD::SCALAR_TO_VECTOR, DL, VT, Res);
}
 
/// If the value to store is a vector that needs to be scalarized, it must be
/// <1 x ty>. Just store the element.
SDValue DAGTypeLegalizer::ScalarizeVecOp_STORE(StoreSDNode *N, unsigned OpNo){
  assert(N->isUnindexed() && "Indexed store of one-element vector?");
  assert(OpNo == 1 && "Do not know how to scalarize this operand!");
  SDLoc dl(N);
 
  if (N->isTruncatingStore())
    return DAG.getTruncStore(
        N->getChain(), dl, GetScalarizedVector(N->getOperand(1)),
        N->getBasePtr(), N->getPointerInfo(),
        N->getMemoryVT().getVectorElementType(), N->getOriginalAlign(),
        N->getMemOperand()->getFlags(), N->getAAInfo());
 
  return DAG.getStore(N->getChain(), dl, GetScalarizedVector(N->getOperand(1)),
                      N->getBasePtr(), N->getPointerInfo(),
                      N->getOriginalAlign(), N->getMemOperand()->getFlags(),
                      N->getAAInfo());
}
 
/// If the value to round is a vector that needs to be scalarized, it must be
/// <1 x ty>. Convert the element instead.
SDValue DAGTypeLegalizer::ScalarizeVecOp_FP_ROUND(SDNode *N, unsigned OpNo) {
  assert(OpNo == 0 && "Wrong operand for scalarization!");
  SDValue Elt = GetScalarizedVector(N->getOperand(0));
  SDValue Res = DAG.getNode(ISD::FP_ROUND, SDLoc(N),
                            N->getValueType(0).getVectorElementType(), Elt,
                            N->getOperand(1));
  return DAG.getNode(ISD::SCALAR_TO_VECTOR, SDLoc(N), N->getValueType(0), Res);
}
 
SDValue DAGTypeLegalizer::ScalarizeVecOp_STRICT_FP_ROUND(SDNode *N, 
                                                         unsigned OpNo) {
  assert(OpNo == 1 && "Wrong operand for scalarization!");
  SDValue Elt = GetScalarizedVector(N->getOperand(1));
  SDValue Res = DAG.getNode(ISD::STRICT_FP_ROUND, SDLoc(N),
                            { N->getValueType(0).getVectorElementType(), 
                              MVT::Other },
                            { N->getOperand(0), Elt, N->getOperand(2) });
  // Legalize the chain result - switch anything that used the old chain to
  // use the new one.
  ReplaceValueWith(SDValue(N, 1), Res.getValue(1));
 
  Res = DAG.getNode(ISD::SCALAR_TO_VECTOR, SDLoc(N), N->getValueType(0), Res);
 
  // Do our own replacement and return SDValue() to tell the caller that we
  // handled all replacements since caller can only handle a single result.
  ReplaceValueWith(SDValue(N, 0), Res);
  return SDValue();
}
 
/// If the value to extend is a vector that needs to be scalarized, it must be
/// <1 x ty>. Convert the element instead.
SDValue DAGTypeLegalizer::ScalarizeVecOp_FP_EXTEND(SDNode *N) {
  SDValue Elt = GetScalarizedVector(N->getOperand(0));
  SDValue Res = DAG.getNode(ISD::FP_EXTEND, SDLoc(N),
                            N->getValueType(0).getVectorElementType(), Elt);
  return DAG.getNode(ISD::SCALAR_TO_VECTOR, SDLoc(N), N->getValueType(0), Res);
}
 
/// If the value to extend is a vector that needs to be scalarized, it must be
/// <1 x ty>. Convert the element instead.
SDValue DAGTypeLegalizer::ScalarizeVecOp_STRICT_FP_EXTEND(SDNode *N) {
  SDValue Elt = GetScalarizedVector(N->getOperand(1));
  SDValue Res =
      DAG.getNode(ISD::STRICT_FP_EXTEND, SDLoc(N),
                  {N->getValueType(0).getVectorElementType(), MVT::Other},
                  {N->getOperand(0), Elt});
  // Legalize the chain result - switch anything that used the old chain to
  // use the new one.
  ReplaceValueWith(SDValue(N, 1), Res.getValue(1));
 
  Res = DAG.getNode(ISD::SCALAR_TO_VECTOR, SDLoc(N), N->getValueType(0), Res);
 
  // Do our own replacement and return SDValue() to tell the caller that we
  // handled all replacements since caller can only handle a single result.
  ReplaceValueWith(SDValue(N, 0), Res);
  return SDValue();
}
 
SDValue DAGTypeLegalizer::ScalarizeVecOp_VECREDUCE(SDNode *N) {
  SDValue Res = GetScalarizedVector(N->getOperand(0));
  // Result type may be wider than element type.
  if (Res.getValueType() != N->getValueType(0))
    Res = DAG.getNode(ISD::ANY_EXTEND, SDLoc(N), N->getValueType(0), Res);
  return Res;
}
 
SDValue DAGTypeLegalizer::ScalarizeVecOp_VECREDUCE_SEQ(SDNode *N) {
  SDValue AccOp = N->getOperand(0);
  SDValue VecOp = N->getOperand(1);
 
  unsigned BaseOpc = ISD::getVecReduceBaseOpcode(N->getOpcode());
 
  SDValue Op = GetScalarizedVector(VecOp);
  return DAG.getNode(BaseOpc, SDLoc(N), N->getValueType(0),
                     AccOp, Op, N->getFlags());
}
 
SDValue DAGTypeLegalizer::ScalarizeVecOp_CMP(SDNode *N) {
  SDValue LHS = GetScalarizedVector(N->getOperand(0));
  SDValue RHS = GetScalarizedVector(N->getOperand(1));
 
  EVT ResVT = N->getValueType(0).getVectorElementType();
  SDValue Cmp = DAG.getNode(N->getOpcode(), SDLoc(N), ResVT, LHS, RHS);
  return DAG.getNode(ISD::SCALAR_TO_VECTOR, SDLoc(N), N->getValueType(0), Cmp);
}
 
//===----------------------------------------------------------------------===//
//  Result Vector Splitting
//===----------------------------------------------------------------------===//
 
/// This method is called when the specified result of the specified node is
/// found to need vector splitting. At this point, the node may also have
/// invalid operands or may have other results that need legalization, we just
/// know that (at least) one result needs vector splitting.
void DAGTypeLegalizer::SplitVectorResult(SDNode *N, unsigned ResNo) {
  LLVM_DEBUG(dbgs() << "Split node result: "; N->dump(&DAG));
  SDValue Lo, Hi;
 
  // See if the target wants to custom expand this node.
  if (CustomLowerNode(N, N->getValueType(ResNo), true))
    return;
 
  switch (N->getOpcode()) {
  default:
#ifndef NDEBUG
    dbgs() << "SplitVectorResult #" << ResNo << ": ";
    N->dump(&DAG);
    dbgs() << "\n";
#endif
    report_fatal_error("Do not know how to split the result of this "
                       "operator!\n");
 
  case ISD::MERGE_VALUES: SplitRes_MERGE_VALUES(N, ResNo, Lo, Hi); break;
  case ISD::AssertZext:   SplitVecRes_AssertZext(N, Lo, Hi); break;
  case ISD::VSELECT:
  case ISD::SELECT:
  case ISD::VP_MERGE:
  case ISD::VP_SELECT:    SplitRes_Select(N, Lo, Hi); break;
  case ISD::SELECT_CC:    SplitRes_SELECT_CC(N, Lo, Hi); break;
  case ISD::UNDEF:        SplitRes_UNDEF(N, Lo, Hi); break;
  case ISD::BITCAST:           SplitVecRes_BITCAST(N, Lo, Hi); break;
  case ISD::BUILD_VECTOR:      SplitVecRes_BUILD_VECTOR(N, Lo, Hi); break;
  case ISD::CONCAT_VECTORS:    SplitVecRes_CONCAT_VECTORS(N, Lo, Hi); break;
  case ISD::EXTRACT_SUBVECTOR: SplitVecRes_EXTRACT_SUBVECTOR(N, Lo, Hi); break;
  case ISD::INSERT_SUBVECTOR:  SplitVecRes_INSERT_SUBVECTOR(N, Lo, Hi); break;
  case ISD::FPOWI:
  case ISD::FLDEXP:
  case ISD::FCOPYSIGN:         SplitVecRes_FPOp_MultiType(N, Lo, Hi); break;
  case ISD::IS_FPCLASS:        SplitVecRes_IS_FPCLASS(N, Lo, Hi); break;
  case ISD::INSERT_VECTOR_ELT: SplitVecRes_INSERT_VECTOR_ELT(N, Lo, Hi); break;
  case ISD::EXPERIMENTAL_VP_SPLAT: SplitVecRes_VP_SPLAT(N, Lo, Hi); break;
  case ISD::SPLAT_VECTOR:
  case ISD::SCALAR_TO_VECTOR:
    SplitVecRes_ScalarOp(N, Lo, Hi);
    break;
  case ISD::STEP_VECTOR:
    SplitVecRes_STEP_VECTOR(N, Lo, Hi);
    break;
  case ISD::SIGN_EXTEND_INREG: SplitVecRes_InregOp(N, Lo, Hi); break;
  case ISD::LOAD:
    SplitVecRes_LOAD(cast<LoadSDNode>(N), Lo, Hi);
    break;
  case ISD::VP_LOAD:
    SplitVecRes_VP_LOAD(cast<VPLoadSDNode>(N), Lo, Hi);
    break;
  case ISD::EXPERIMENTAL_VP_STRIDED_LOAD:
    SplitVecRes_VP_STRIDED_LOAD(cast<VPStridedLoadSDNode>(N), Lo, Hi);
    break;
  case ISD::MLOAD:
    SplitVecRes_MLOAD(cast<MaskedLoadSDNode>(N), Lo, Hi);
    break;
  case ISD::MGATHER:
  case ISD::VP_GATHER:
    SplitVecRes_Gather(cast<MemSDNode>(N), Lo, Hi, /*SplitSETCC*/ true);
    break;
  case ISD::VECTOR_COMPRESS:
    SplitVecRes_VECTOR_COMPRESS(N, Lo, Hi);
    break;
  case ISD::SETCC:
  case ISD::VP_SETCC:
    SplitVecRes_SETCC(N, Lo, Hi);
    break;
  case ISD::VECTOR_REVERSE:
    SplitVecRes_VECTOR_REVERSE(N, Lo, Hi);
    break;
  case ISD::VECTOR_SHUFFLE:
    SplitVecRes_VECTOR_SHUFFLE(cast<ShuffleVectorSDNode>(N), Lo, Hi);
    break;
  case ISD::VECTOR_SPLICE:
    SplitVecRes_VECTOR_SPLICE(N, Lo, Hi);
    break;
  case ISD::VECTOR_DEINTERLEAVE:
    SplitVecRes_VECTOR_DEINTERLEAVE(N);
    return;
  case ISD::VECTOR_INTERLEAVE:
    SplitVecRes_VECTOR_INTERLEAVE(N);
    return;
  case ISD::VAARG:
    SplitVecRes_VAARG(N, Lo, Hi);
    break;
 
  case ISD::ANY_EXTEND_VECTOR_INREG:
  case ISD::SIGN_EXTEND_VECTOR_INREG:
  case ISD::ZERO_EXTEND_VECTOR_INREG:
    SplitVecRes_ExtVecInRegOp(N, Lo, Hi);
    break;
 
  case ISD::ABS:
  case ISD::VP_ABS:
  case ISD::BITREVERSE:
  case ISD::VP_BITREVERSE:
  case ISD::BSWAP:
  case ISD::VP_BSWAP:
  case ISD::CTLZ:
  case ISD::VP_CTLZ:
  case ISD::CTTZ:
  case ISD::VP_CTTZ:
  case ISD::CTLZ_ZERO_UNDEF:
  case ISD::VP_CTLZ_ZERO_UNDEF:
  case ISD::CTTZ_ZERO_UNDEF:
  case ISD::VP_CTTZ_ZERO_UNDEF:
  case ISD::CTPOP:
  case ISD::VP_CTPOP:
  case ISD::FABS: case ISD::VP_FABS:
  case ISD::FACOS:
  case ISD::FASIN:
  case ISD::FATAN:
  case ISD::FCEIL:
  case ISD::VP_FCEIL:
  case ISD::FCOS:
  case ISD::FCOSH:
  case ISD::FEXP:
  case ISD::FEXP2:
  case ISD::FEXP10:
  case ISD::FFLOOR:
  case ISD::VP_FFLOOR:
  case ISD::FLOG:
  case ISD::FLOG10:
  case ISD::FLOG2:
  case ISD::FNEARBYINT:
  case ISD::VP_FNEARBYINT:
  case ISD::FNEG: case ISD::VP_FNEG:
  case ISD::FREEZE:
  case ISD::ARITH_FENCE:
  case ISD::FP_EXTEND:
  case ISD::VP_FP_EXTEND:
  case ISD::FP_ROUND:
  case ISD::VP_FP_ROUND:
  case ISD::FP_TO_SINT:
  case ISD::VP_FP_TO_SINT:
  case ISD::FP_TO_UINT:
  case ISD::VP_FP_TO_UINT:
  case ISD::FRINT:
  case ISD::VP_FRINT:
  case ISD::LRINT:
  case ISD::VP_LRINT:
  case ISD::LLRINT:
  case ISD::VP_LLRINT:
  case ISD::FROUND:
  case ISD::VP_FROUND:
  case ISD::FROUNDEVEN:
  case ISD::VP_FROUNDEVEN:
  case ISD::LROUND:
  case ISD::LLROUND:
  case ISD::FSIN:
  case ISD::FSINH:
  case ISD::FSQRT: case ISD::VP_SQRT:
  case ISD::FTAN:
  case ISD::FTANH:
  case ISD::FTRUNC:
  case ISD::VP_FROUNDTOZERO:
  case ISD::SINT_TO_FP:
  case ISD::VP_SINT_TO_FP:
  case ISD::TRUNCATE:
  case ISD::VP_TRUNCATE:
  case ISD::UINT_TO_FP:
  case ISD::VP_UINT_TO_FP:
  case ISD::FCANONICALIZE:
    SplitVecRes_UnaryOp(N, Lo, Hi);
    break;
  case ISD::ADDRSPACECAST:
    SplitVecRes_ADDRSPACECAST(N, Lo, Hi);
    break;
  case ISD::FFREXP:
    SplitVecRes_FFREXP(N, ResNo, Lo, Hi);
    break;
 
  case ISD::ANY_EXTEND:
  case ISD::SIGN_EXTEND:
  case ISD::ZERO_EXTEND:
  case ISD::VP_SIGN_EXTEND:
  case ISD::VP_ZERO_EXTEND:
    SplitVecRes_ExtendOp(N, Lo, Hi);
    break;
 
  case ISD::ADD: case ISD::VP_ADD:
  case ISD::SUB: case ISD::VP_SUB:
  case ISD::MUL: case ISD::VP_MUL:
  case ISD::MULHS:
  case ISD::MULHU:
  case ISD::ABDS:
  case ISD::ABDU:
  case ISD::AVGCEILS:
  case ISD::AVGCEILU:
  case ISD::AVGFLOORS:
  case ISD::AVGFLOORU:
  case ISD::FADD: case ISD::VP_FADD:
  case ISD::FSUB: case ISD::VP_FSUB:
  case ISD::FMUL: case ISD::VP_FMUL:
  case ISD::FMINNUM:
  case ISD::FMINNUM_IEEE:
  case ISD::VP_FMINNUM:
  case ISD::FMAXNUM:
  case ISD::FMAXNUM_IEEE:
  case ISD::VP_FMAXNUM:
  case ISD::FMINIMUM:
  case ISD::VP_FMINIMUM:
  case ISD::FMAXIMUM:
  case ISD::VP_FMAXIMUM:
  case ISD::SDIV: case ISD::VP_SDIV:
  case ISD::UDIV: case ISD::VP_UDIV:
  case ISD::FDIV: case ISD::VP_FDIV:
  case ISD::FPOW:
  case ISD::AND: case ISD::VP_AND:
  case ISD::OR: case ISD::VP_OR:
  case ISD::XOR: case ISD::VP_XOR:
  case ISD::SHL: case ISD::VP_SHL:
  case ISD::SRA: case ISD::VP_SRA:
  case ISD::SRL: case ISD::VP_SRL:
  case ISD::UREM: case ISD::VP_UREM:
  case ISD::SREM: case ISD::VP_SREM:
  case ISD::FREM: case ISD::VP_FREM:
  case ISD::SMIN: case ISD::VP_SMIN:
  case ISD::SMAX: case ISD::VP_SMAX:
  case ISD::UMIN: case ISD::VP_UMIN:
  case ISD::UMAX: case ISD::VP_UMAX:
  case ISD::SADDSAT: case ISD::VP_SADDSAT:
  case ISD::UADDSAT: case ISD::VP_UADDSAT:
  case ISD::SSUBSAT: case ISD::VP_SSUBSAT:
  case ISD::USUBSAT: case ISD::VP_USUBSAT:
  case ISD::SSHLSAT:
  case ISD::USHLSAT:
  case ISD::ROTL:
  case ISD::ROTR:
  case ISD::VP_FCOPYSIGN:
    SplitVecRes_BinOp(N, Lo, Hi);
    break;
  case ISD::FMA: case ISD::VP_FMA:
  case ISD::FSHL:
  case ISD::VP_FSHL:
  case ISD::FSHR:
  case ISD::VP_FSHR:
    SplitVecRes_TernaryOp(N, Lo, Hi);
    break;
 
  case ISD::SCMP: case ISD::UCMP:
    SplitVecRes_CMP(N, Lo, Hi);
    break;
 
#define DAG_INSTRUCTION(NAME, NARG, ROUND_MODE, INTRINSIC, DAGN)               \
  case ISD::STRICT_##DAGN:
#include "llvm/IR/ConstrainedOps.def"
    SplitVecRes_StrictFPOp(N, Lo, Hi);
    break;
 
  case ISD::FP_TO_UINT_SAT:
  case ISD::FP_TO_SINT_SAT:
    SplitVecRes_FP_TO_XINT_SAT(N, Lo, Hi);
    break;
 
  case ISD::UADDO:
  case ISD::SADDO:
  case ISD::USUBO:
  case ISD::SSUBO:
  case ISD::UMULO:
  case ISD::SMULO:
    SplitVecRes_OverflowOp(N, ResNo, Lo, Hi);
    break;
  case ISD::SMULFIX:
  case ISD::SMULFIXSAT:
  case ISD::UMULFIX:
  case ISD::UMULFIXSAT:
  case ISD::SDIVFIX:
  case ISD::SDIVFIXSAT:
  case ISD::UDIVFIX:
  case ISD::UDIVFIXSAT:
    SplitVecRes_FIX(N, Lo, Hi);
    break;
  case ISD::EXPERIMENTAL_VP_REVERSE:
    SplitVecRes_VP_REVERSE(N, Lo, Hi);
    break;
  }
 
  // If Lo/Hi is null, the sub-method took care of registering results etc.
  if (Lo.getNode())
    SetSplitVector(SDValue(N, ResNo), Lo, Hi);
}
 
void DAGTypeLegalizer::IncrementPointer(MemSDNode *N, EVT MemVT,
                                        MachinePointerInfo &MPI, SDValue &Ptr,
                                        uint64_t *ScaledOffset) {
  SDLoc DL(N);
  unsigned IncrementSize = MemVT.getSizeInBits().getKnownMinValue() / 8;
 
  if (MemVT.isScalableVector()) {
    SDNodeFlags Flags;
    SDValue BytesIncrement = DAG.getVScale(
        DL, Ptr.getValueType(),
        APInt(Ptr.getValueSizeInBits().getFixedValue(), IncrementSize));
    MPI = MachinePointerInfo(N->getPointerInfo().getAddrSpace());
    Flags.setNoUnsignedWrap(true);
    if (ScaledOffset)
      *ScaledOffset += IncrementSize;
    Ptr = DAG.getNode(ISD::ADD, DL, Ptr.getValueType(), Ptr, BytesIncrement,
                      Flags);
  } else {
    MPI = N->getPointerInfo().getWithOffset(IncrementSize);
    // Increment the pointer to the other half.
    Ptr = DAG.getObjectPtrOffset(DL, Ptr, TypeSize::getFixed(IncrementSize));
  }
}
 
std::pair<SDValue, SDValue> DAGTypeLegalizer::SplitMask(SDValue Mask) {
  return SplitMask(Mask, SDLoc(Mask));
}
 
std::pair<SDValue, SDValue> DAGTypeLegalizer::SplitMask(SDValue Mask,
                                                        const SDLoc &DL) {
  SDValue MaskLo, MaskHi;
  EVT MaskVT = Mask.getValueType();
  if (getTypeAction(MaskVT) == TargetLowering::TypeSplitVector)
    GetSplitVector(Mask, MaskLo, MaskHi);
  else
    std::tie(MaskLo, MaskHi) = DAG.SplitVector(Mask, DL);
  return std::make_pair(MaskLo, MaskHi);
}
 
void DAGTypeLegalizer::SplitVecRes_BinOp(SDNode *N, SDValue &Lo, SDValue &Hi) {
  SDValue LHSLo, LHSHi;
  GetSplitVector(N->getOperand(0), LHSLo, LHSHi);
  SDValue RHSLo, RHSHi;
  GetSplitVector(N->getOperand(1), RHSLo, RHSHi);
  SDLoc dl(N);
 
  const SDNodeFlags Flags = N->getFlags();
  unsigned Opcode = N->getOpcode();
  if (N->getNumOperands() == 2) {
    Lo = DAG.getNode(Opcode, dl, LHSLo.getValueType(), LHSLo, RHSLo, Flags);
    Hi = DAG.getNode(Opcode, dl, LHSHi.getValueType(), LHSHi, RHSHi, Flags);
    return;
  }
 
  assert(N->getNumOperands() == 4 && "Unexpected number of operands!");
  assert(N->isVPOpcode() && "Expected VP opcode");
 
  SDValue MaskLo, MaskHi;
  std::tie(MaskLo, MaskHi) = SplitMask(N->getOperand(2));
 
  SDValue EVLLo, EVLHi;
  std::tie(EVLLo, EVLHi) =
      DAG.SplitEVL(N->getOperand(3), N->getValueType(0), dl);
 
  Lo = DAG.getNode(Opcode, dl, LHSLo.getValueType(),
                   {LHSLo, RHSLo, MaskLo, EVLLo}, Flags);
  Hi = DAG.getNode(Opcode, dl, LHSHi.getValueType(),
                   {LHSHi, RHSHi, MaskHi, EVLHi}, Flags);
}
 
void DAGTypeLegalizer::SplitVecRes_TernaryOp(SDNode *N, SDValue &Lo,
                                             SDValue &Hi) {
  SDValue Op0Lo, Op0Hi;
  GetSplitVector(N->getOperand(0), Op0Lo, Op0Hi);
  SDValue Op1Lo, Op1Hi;
  GetSplitVector(N->getOperand(1), Op1Lo, Op1Hi);
  SDValue Op2Lo, Op2Hi;
  GetSplitVector(N->getOperand(2), Op2Lo, Op2Hi);
  SDLoc dl(N);
 
  const SDNodeFlags Flags = N->getFlags();
  unsigned Opcode = N->getOpcode();
  if (N->getNumOperands() == 3) {
    Lo = DAG.getNode(Opcode, dl, Op0Lo.getValueType(), Op0Lo, Op1Lo, Op2Lo, Flags);
    Hi = DAG.getNode(Opcode, dl, Op0Hi.getValueType(), Op0Hi, Op1Hi, Op2Hi, Flags);
    return;
  }
 
  assert(N->getNumOperands() == 5 && "Unexpected number of operands!");
  assert(N->isVPOpcode() && "Expected VP opcode");
 
  SDValue MaskLo, MaskHi;
  std::tie(MaskLo, MaskHi) = SplitMask(N->getOperand(3));
 
  SDValue EVLLo, EVLHi;
  std::tie(EVLLo, EVLHi) =
      DAG.SplitEVL(N->getOperand(4), N->getValueType(0), dl);
 
  Lo = DAG.getNode(Opcode, dl, Op0Lo.getValueType(),
                   {Op0Lo, Op1Lo, Op2Lo, MaskLo, EVLLo}, Flags);
  Hi = DAG.getNode(Opcode, dl, Op0Hi.getValueType(),
                   {Op0Hi, Op1Hi, Op2Hi, MaskHi, EVLHi}, Flags);
}
 
void DAGTypeLegalizer::SplitVecRes_CMP(SDNode *N, SDValue &Lo, SDValue &Hi) {
  LLVMContext &Ctxt = *DAG.getContext();
  SDLoc dl(N);
 
  SDValue LHS = N->getOperand(0);
  SDValue RHS = N->getOperand(1);
 
  SDValue LHSLo, LHSHi, RHSLo, RHSHi;
  if (getTypeAction(LHS.getValueType()) == TargetLowering::TypeSplitVector) {
    GetSplitVector(LHS, LHSLo, LHSHi);
    GetSplitVector(RHS, RHSLo, RHSHi);
  } else {
    std::tie(LHSLo, LHSHi) = DAG.SplitVector(LHS, dl);
    std::tie(RHSLo, RHSHi) = DAG.SplitVector(RHS, dl);
  }
 
  EVT SplitResVT = N->getValueType(0).getHalfNumVectorElementsVT(Ctxt);
  Lo = DAG.getNode(N->getOpcode(), dl, SplitResVT, LHSLo, RHSLo);
  Hi = DAG.getNode(N->getOpcode(), dl, SplitResVT, LHSHi, RHSHi);
}
 
void DAGTypeLegalizer::SplitVecRes_FIX(SDNode *N, SDValue &Lo, SDValue &Hi) {
  SDValue LHSLo, LHSHi;
  GetSplitVector(N->getOperand(0), LHSLo, LHSHi);
  SDValue RHSLo, RHSHi;
  GetSplitVector(N->getOperand(1), RHSLo, RHSHi);
  SDLoc dl(N);
  SDValue Op2 = N->getOperand(2);
 
  unsigned Opcode = N->getOpcode();
  Lo = DAG.getNode(Opcode, dl, LHSLo.getValueType(), LHSLo, RHSLo, Op2,
                   N->getFlags());
  Hi = DAG.getNode(Opcode, dl, LHSHi.getValueType(), LHSHi, RHSHi, Op2,
                   N->getFlags());
}
 
void DAGTypeLegalizer::SplitVecRes_BITCAST(SDNode *N, SDValue &Lo,
                                           SDValue &Hi) {
  // We know the result is a vector.  The input may be either a vector or a
  // scalar value.
  EVT LoVT, HiVT;
  std::tie(LoVT, HiVT) = DAG.GetSplitDestVTs(N->getValueType(0));
  SDLoc dl(N);
 
  SDValue InOp = N->getOperand(0);
  EVT InVT = InOp.getValueType();
 
  // Handle some special cases efficiently.
  switch (getTypeAction(InVT)) {
  case TargetLowering::TypeLegal:
  case TargetLowering::TypePromoteInteger:
  case TargetLowering::TypePromoteFloat:
  case TargetLowering::TypeSoftPromoteHalf:
  case TargetLowering::TypeSoftenFloat:
  case TargetLowering::TypeScalarizeVector:
  case TargetLowering::TypeWidenVector:
    break;
  case TargetLowering::TypeExpandInteger:
  case TargetLowering::TypeExpandFloat:
    // A scalar to vector conversion, where the scalar needs expansion.
    // If the vector is being split in two then we can just convert the
    // expanded pieces.
    if (LoVT == HiVT) {
      GetExpandedOp(InOp, Lo, Hi);
      if (DAG.getDataLayout().isBigEndian())
        std::swap(Lo, Hi);
      Lo = DAG.getNode(ISD::BITCAST, dl, LoVT, Lo);
      Hi = DAG.getNode(ISD::BITCAST, dl, HiVT, Hi);
      return;
    }
    break;
  case TargetLowering::TypeSplitVector:
    // If the input is a vector that needs to be split, convert each split
    // piece of the input now.
    GetSplitVector(InOp, Lo, Hi);
    Lo = DAG.getNode(ISD::BITCAST, dl, LoVT, Lo);
    Hi = DAG.getNode(ISD::BITCAST, dl, HiVT, Hi);
    return;
  case TargetLowering::TypeScalarizeScalableVector:
    report_fatal_error("Scalarization of scalable vectors is not supported.");
  }
 
  if (LoVT.isScalableVector()) {
    auto [InLo, InHi] = DAG.SplitVectorOperand(N, 0);
    Lo = DAG.getNode(ISD::BITCAST, dl, LoVT, InLo);
    Hi = DAG.getNode(ISD::BITCAST, dl, HiVT, InHi);
    return;
  }
 
  // In the general case, convert the input to an integer and split it by hand.
  EVT LoIntVT = EVT::getIntegerVT(*DAG.getContext(), LoVT.getSizeInBits());
  EVT HiIntVT = EVT::getIntegerVT(*DAG.getContext(), HiVT.getSizeInBits());
  if (DAG.getDataLayout().isBigEndian())
    std::swap(LoIntVT, HiIntVT);
 
  SplitInteger(BitConvertToInteger(InOp), LoIntVT, HiIntVT, Lo, Hi);
 
  if (DAG.getDataLayout().isBigEndian())
    std::swap(Lo, Hi);
  Lo = DAG.getNode(ISD::BITCAST, dl, LoVT, Lo);
  Hi = DAG.getNode(ISD::BITCAST, dl, HiVT, Hi);
}
 
void DAGTypeLegalizer::SplitVecRes_BUILD_VECTOR(SDNode *N, SDValue &Lo,
                                                SDValue &Hi) {
  EVT LoVT, HiVT;
  SDLoc dl(N);
  std::tie(LoVT, HiVT) = DAG.GetSplitDestVTs(N->getValueType(0));
  unsigned LoNumElts = LoVT.getVectorNumElements();
  SmallVector<SDValue, 8> LoOps(N->op_begin(), N->op_begin()+LoNumElts);
  Lo = DAG.getBuildVector(LoVT, dl, LoOps);
 
  SmallVector<SDValue, 8> HiOps(N->op_begin()+LoNumElts, N->op_end());
  Hi = DAG.getBuildVector(HiVT, dl, HiOps);
}
 
void DAGTypeLegalizer::SplitVecRes_CONCAT_VECTORS(SDNode *N, SDValue &Lo,
                                                  SDValue &Hi) {
  assert(!(N->getNumOperands() & 1) && "Unsupported CONCAT_VECTORS");
  SDLoc dl(N);
  unsigned NumSubvectors = N->getNumOperands() / 2;
  if (NumSubvectors == 1) {
    Lo = N->getOperand(0);
    Hi = N->getOperand(1);
    return;
  }
 
  EVT LoVT, HiVT;
  std::tie(LoVT, HiVT) = DAG.GetSplitDestVTs(N->getValueType(0));
 
  SmallVector<SDValue, 8> LoOps(N->op_begin(), N->op_begin()+NumSubvectors);
  Lo = DAG.getNode(ISD::CONCAT_VECTORS, dl, LoVT, LoOps);
 
  SmallVector<SDValue, 8> HiOps(N->op_begin()+NumSubvectors, N->op_end());
  Hi = DAG.getNode(ISD::CONCAT_VECTORS, dl, HiVT, HiOps);
}
 
void DAGTypeLegalizer::SplitVecRes_EXTRACT_SUBVECTOR(SDNode *N, SDValue &Lo,
                                                     SDValue &Hi) {
  SDValue Vec = N->getOperand(0);
  SDValue Idx = N->getOperand(1);
  SDLoc dl(N);
 
  EVT LoVT, HiVT;
  std::tie(LoVT, HiVT) = DAG.GetSplitDestVTs(N->getValueType(0));
 
  Lo = DAG.getNode(ISD::EXTRACT_SUBVECTOR, dl, LoVT, Vec, Idx);
  uint64_t IdxVal = Idx->getAsZExtVal();
  Hi = DAG.getNode(
      ISD::EXTRACT_SUBVECTOR, dl, HiVT, Vec,
      DAG.getVectorIdxConstant(IdxVal + LoVT.getVectorMinNumElements(), dl));
}
 
void DAGTypeLegalizer::SplitVecRes_INSERT_SUBVECTOR(SDNode *N, SDValue &Lo,
                                                    SDValue &Hi) {
  SDValue Vec = N->getOperand(0);
  SDValue SubVec = N->getOperand(1);
  SDValue Idx = N->getOperand(2);
  SDLoc dl(N);
  GetSplitVector(Vec, Lo, Hi);
 
  EVT VecVT = Vec.getValueType();
  EVT LoVT = Lo.getValueType();
  EVT SubVecVT = SubVec.getValueType();
  unsigned VecElems = VecVT.getVectorMinNumElements();
  unsigned SubElems = SubVecVT.getVectorMinNumElements();
  unsigned LoElems = LoVT.getVectorMinNumElements();
 
  // If we know the index is in the first half, and we know the subvector
  // doesn't cross the boundary between the halves, we can avoid spilling the
  // vector, and insert into the lower half of the split vector directly.
  unsigned IdxVal = Idx->getAsZExtVal();
  if (IdxVal + SubElems <= LoElems) {
    Lo = DAG.getNode(ISD::INSERT_SUBVECTOR, dl, LoVT, Lo, SubVec, Idx);
    return;
  }
  // Similarly if the subvector is fully in the high half, but mind that we
  // can't tell whether a fixed-length subvector is fully within the high half
  // of a scalable vector.
  if (VecVT.isScalableVector() == SubVecVT.isScalableVector() &&
      IdxVal >= LoElems && IdxVal + SubElems <= VecElems) {
    Hi = DAG.getNode(ISD::INSERT_SUBVECTOR, dl, Hi.getValueType(), Hi, SubVec,
                     DAG.getVectorIdxConstant(IdxVal - LoElems, dl));
    return;
  }
 
  // Spill the vector to the stack.
  // In cases where the vector is illegal it will be broken down into parts
  // and stored in parts - we should use the alignment for the smallest part.
  Align SmallestAlign = DAG.getReducedAlign(VecVT, /*UseABI=*/false);
  SDValue StackPtr =
      DAG.CreateStackTemporary(VecVT.getStoreSize(), SmallestAlign);
  auto &MF = DAG.getMachineFunction();
  auto FrameIndex = cast<FrameIndexSDNode>(StackPtr.getNode())->getIndex();
  auto PtrInfo = MachinePointerInfo::getFixedStack(MF, FrameIndex);
 
  SDValue Store = DAG.getStore(DAG.getEntryNode(), dl, Vec, StackPtr, PtrInfo,
                               SmallestAlign);
 
  // Store the new subvector into the specified index.
  SDValue SubVecPtr =
      TLI.getVectorSubVecPointer(DAG, StackPtr, VecVT, SubVecVT, Idx);
  Store = DAG.getStore(Store, dl, SubVec, SubVecPtr,
                       MachinePointerInfo::getUnknownStack(MF));
 
  // Load the Lo part from the stack slot.
  Lo = DAG.getLoad(Lo.getValueType(), dl, Store, StackPtr, PtrInfo,
                   SmallestAlign);
 
  // Increment the pointer to the other part.
  auto *Load = cast<LoadSDNode>(Lo);
  MachinePointerInfo MPI = Load->getPointerInfo();
  IncrementPointer(Load, LoVT, MPI, StackPtr);
 
  // Load the Hi part from the stack slot.
  Hi = DAG.getLoad(Hi.getValueType(), dl, Store, StackPtr, MPI, SmallestAlign);
}
 
// Handle splitting an FP where the second operand does not match the first
// type. The second operand may be a scalar, or a vector that has exactly as
// many elements as the first
void DAGTypeLegalizer::SplitVecRes_FPOp_MultiType(SDNode *N, SDValue &Lo,
                                                  SDValue &Hi) {
  SDValue LHSLo, LHSHi;
  GetSplitVector(N->getOperand(0), LHSLo, LHSHi);
  SDLoc DL(N);
 
  SDValue RHSLo, RHSHi;
  SDValue RHS = N->getOperand(1);
  EVT RHSVT = RHS.getValueType();
  if (RHSVT.isVector()) {
    if (getTypeAction(RHSVT) == TargetLowering::TypeSplitVector)
      GetSplitVector(RHS, RHSLo, RHSHi);
    else
      std::tie(RHSLo, RHSHi) = DAG.SplitVector(RHS, SDLoc(RHS));
 
    Lo = DAG.getNode(N->getOpcode(), DL, LHSLo.getValueType(), LHSLo, RHSLo);
    Hi = DAG.getNode(N->getOpcode(), DL, LHSHi.getValueType(), LHSHi, RHSHi);
  } else {
    Lo = DAG.getNode(N->getOpcode(), DL, LHSLo.getValueType(), LHSLo, RHS);
    Hi = DAG.getNode(N->getOpcode(), DL, LHSHi.getValueType(), LHSHi, RHS);
  }
}
 
void DAGTypeLegalizer::SplitVecRes_IS_FPCLASS(SDNode *N, SDValue &Lo,
                                              SDValue &Hi) {
  SDLoc DL(N);
  SDValue ArgLo, ArgHi;
  SDValue Test = N->getOperand(1);
  SDValue FpValue = N->getOperand(0);
  if (getTypeAction(FpValue.getValueType()) == TargetLowering::TypeSplitVector)
    GetSplitVector(FpValue, ArgLo, ArgHi);
  else
    std::tie(ArgLo, ArgHi) = DAG.SplitVector(FpValue, SDLoc(FpValue));
  EVT LoVT, HiVT;
  std::tie(LoVT, HiVT) = DAG.GetSplitDestVTs(N->getValueType(0));
 
  Lo = DAG.getNode(ISD::IS_FPCLASS, DL, LoVT, ArgLo, Test, N->getFlags());
  Hi = DAG.getNode(ISD::IS_FPCLASS, DL, HiVT, ArgHi, Test, N->getFlags());
}
 
void DAGTypeLegalizer::SplitVecRes_InregOp(SDNode *N, SDValue &Lo,
                                           SDValue &Hi) {
  SDValue LHSLo, LHSHi;
  GetSplitVector(N->getOperand(0), LHSLo, LHSHi);
  SDLoc dl(N);
 
  EVT LoVT, HiVT;
  std::tie(LoVT, HiVT) =
    DAG.GetSplitDestVTs(cast<VTSDNode>(N->getOperand(1))->getVT());
 
  Lo = DAG.getNode(N->getOpcode(), dl, LHSLo.getValueType(), LHSLo,
                   DAG.getValueType(LoVT));
  Hi = DAG.getNode(N->getOpcode(), dl, LHSHi.getValueType(), LHSHi,
                   DAG.getValueType(HiVT));
}
 
void DAGTypeLegalizer::SplitVecRes_ExtVecInRegOp(SDNode *N, SDValue &Lo,
                                                 SDValue &Hi) {
  unsigned Opcode = N->getOpcode();
  SDValue N0 = N->getOperand(0);
 
  SDLoc dl(N);
  SDValue InLo, InHi;
 
  if (getTypeAction(N0.getValueType()) == TargetLowering::TypeSplitVector)
    GetSplitVector(N0, InLo, InHi);
  else
    std::tie(InLo, InHi) = DAG.SplitVectorOperand(N, 0);
 
  EVT InLoVT = InLo.getValueType();
  unsigned InNumElements = InLoVT.getVectorNumElements();
 
  EVT OutLoVT, OutHiVT;
  std::tie(OutLoVT, OutHiVT) = DAG.GetSplitDestVTs(N->getValueType(0));
  unsigned OutNumElements = OutLoVT.getVectorNumElements();
  assert((2 * OutNumElements) <= InNumElements &&
         "Illegal extend vector in reg split");
 
  // *_EXTEND_VECTOR_INREG instructions extend the lowest elements of the
  // input vector (i.e. we only use InLo):
  // OutLo will extend the first OutNumElements from InLo.
  // OutHi will extend the next OutNumElements from InLo.
 
  // Shuffle the elements from InLo for OutHi into the bottom elements to
  // create a 'fake' InHi.
  SmallVector<int, 8> SplitHi(InNumElements, -1);
  for (unsigned i = 0; i != OutNumElements; ++i)
    SplitHi[i] = i + OutNumElements;
  InHi = DAG.getVectorShuffle(InLoVT, dl, InLo, DAG.getUNDEF(InLoVT), SplitHi);
 
  Lo = DAG.getNode(Opcode, dl, OutLoVT, InLo);
  Hi = DAG.getNode(Opcode, dl, OutHiVT, InHi);
}
 
void DAGTypeLegalizer::SplitVecRes_StrictFPOp(SDNode *N, SDValue &Lo,
                                              SDValue &Hi) {
  unsigned NumOps = N->getNumOperands();
  SDValue Chain = N->getOperand(0);
  EVT LoVT, HiVT;
  SDLoc dl(N);
  std::tie(LoVT, HiVT) = DAG.GetSplitDestVTs(N->getValueType(0));
 
  SmallVector<SDValue, 4> OpsLo(NumOps);
  SmallVector<SDValue, 4> OpsHi(NumOps);
 
  // The Chain is the first operand.
  OpsLo[0] = Chain;
  OpsHi[0] = Chain;
 
  // Now process the remaining operands.
  for (unsigned i = 1; i < NumOps; ++i) {
    SDValue Op = N->getOperand(i);
    SDValue OpLo = Op;
    SDValue OpHi = Op;
 
    EVT InVT = Op.getValueType();
    if (InVT.isVector()) {
      // If the input also splits, handle it directly for a
      // compile time speedup. Otherwise split it by hand.
      if (getTypeAction(InVT) == TargetLowering::TypeSplitVector)
        GetSplitVector(Op, OpLo, OpHi);
      else
        std::tie(OpLo, OpHi) = DAG.SplitVectorOperand(N, i);
    }
 
    OpsLo[i] = OpLo;
    OpsHi[i] = OpHi;
  }
 
  EVT LoValueVTs[] = {LoVT, MVT::Other};
  EVT HiValueVTs[] = {HiVT, MVT::Other};
  Lo = DAG.getNode(N->getOpcode(), dl, DAG.getVTList(LoValueVTs), OpsLo,
                   N->getFlags());
  Hi = DAG.getNode(N->getOpcode(), dl, DAG.getVTList(HiValueVTs), OpsHi,
                   N->getFlags());
 
  // Build a factor node to remember that this Op is independent of the
  // other one.
  Chain = DAG.getNode(ISD::TokenFactor, dl, MVT::Other,
                      Lo.getValue(1), Hi.getValue(1));
 
  // Legalize the chain result - switch anything that used the old chain to
  // use the new one.
  ReplaceValueWith(SDValue(N, 1), Chain);
}
 
SDValue DAGTypeLegalizer::UnrollVectorOp_StrictFP(SDNode *N, unsigned ResNE) {
  SDValue Chain = N->getOperand(0);
  EVT VT = N->getValueType(0);
  unsigned NE = VT.getVectorNumElements();
  EVT EltVT = VT.getVectorElementType();
  SDLoc dl(N);
 
  SmallVector<SDValue, 8> Scalars;
  SmallVector<SDValue, 4> Operands(N->getNumOperands());
 
  // If ResNE is 0, fully unroll the vector op.
  if (ResNE == 0)
    ResNE = NE;
  else if (NE > ResNE)
    NE = ResNE;
 
  //The results of each unrolled operation, including the chain.
  EVT ChainVTs[] = {EltVT, MVT::Other};
  SmallVector<SDValue, 8> Chains;
 
  unsigned i;
  for (i = 0; i != NE; ++i) {
    Operands[0] = Chain;
    for (unsigned j = 1, e = N->getNumOperands(); j != e; ++j) {
      SDValue Operand = N->getOperand(j);
      EVT OperandVT = Operand.getValueType();
      if (OperandVT.isVector()) {
        EVT OperandEltVT = OperandVT.getVectorElementType();
        Operands[j] = DAG.getNode(ISD::EXTRACT_VECTOR_ELT, dl, OperandEltVT,
                                  Operand, DAG.getVectorIdxConstant(i, dl));
      } else {
        Operands[j] = Operand;
      }
    }
    SDValue Scalar = DAG.getNode(N->getOpcode(), dl, ChainVTs, Operands);
    Scalar.getNode()->setFlags(N->getFlags());
 
    //Add in the scalar as well as its chain value to the
    //result vectors.
    Scalars.push_back(Scalar);
    Chains.push_back(Scalar.getValue(1));
  }
 
  for (; i < ResNE; ++i)
    Scalars.push_back(DAG.getUNDEF(EltVT));
 
  // Build a new factor node to connect the chain back together.
  Chain = DAG.getNode(ISD::TokenFactor, dl, MVT::Other, Chains);
  ReplaceValueWith(SDValue(N, 1), Chain);
 
  // Create a new BUILD_VECTOR node
  EVT VecVT = EVT::getVectorVT(*DAG.getContext(), EltVT, ResNE);
  return DAG.getBuildVector(VecVT, dl, Scalars);
}
 
void DAGTypeLegalizer::SplitVecRes_OverflowOp(SDNode *N, unsigned ResNo,
                                              SDValue &Lo, SDValue &Hi) {
  SDLoc dl(N);
  EVT ResVT = N->getValueType(0);
  EVT OvVT = N->getValueType(1);
  EVT LoResVT, HiResVT, LoOvVT, HiOvVT;
  std::tie(LoResVT, HiResVT) = DAG.GetSplitDestVTs(ResVT);
  std::tie(LoOvVT, HiOvVT) = DAG.GetSplitDestVTs(OvVT);
 
  SDValue LoLHS, HiLHS, LoRHS, HiRHS;
  if (getTypeAction(ResVT) == TargetLowering::TypeSplitVector) {
    GetSplitVector(N->getOperand(0), LoLHS, HiLHS);
    GetSplitVector(N->getOperand(1), LoRHS, HiRHS);
  } else {
    std::tie(LoLHS, HiLHS) = DAG.SplitVectorOperand(N, 0);
    std::tie(LoRHS, HiRHS) = DAG.SplitVectorOperand(N, 1);
  }
 
  unsigned Opcode = N->getOpcode();
  SDVTList LoVTs = DAG.getVTList(LoResVT, LoOvVT);
  SDVTList HiVTs = DAG.getVTList(HiResVT, HiOvVT);
  SDNode *LoNode = DAG.getNode(Opcode, dl, LoVTs, LoLHS, LoRHS).getNode();
  SDNode *HiNode = DAG.getNode(Opcode, dl, HiVTs, HiLHS, HiRHS).getNode();
  LoNode->setFlags(N->getFlags());
  HiNode->setFlags(N->getFlags());
 
  Lo = SDValue(LoNode, ResNo);
  Hi = SDValue(HiNode, ResNo);
 
  // Replace the other vector result not being explicitly split here.
  unsigned OtherNo = 1 - ResNo;
  EVT OtherVT = N->getValueType(OtherNo);
  if (getTypeAction(OtherVT) == TargetLowering::TypeSplitVector) {
    SetSplitVector(SDValue(N, OtherNo),
                   SDValue(LoNode, OtherNo), SDValue(HiNode, OtherNo));
  } else {
    SDValue OtherVal = DAG.getNode(
        ISD::CONCAT_VECTORS, dl, OtherVT,
        SDValue(LoNode, OtherNo), SDValue(HiNode, OtherNo));
    ReplaceValueWith(SDValue(N, OtherNo), OtherVal);
  }
}
 
void DAGTypeLegalizer::SplitVecRes_INSERT_VECTOR_ELT(SDNode *N, SDValue &Lo,
                                                     SDValue &Hi) {
  SDValue Vec = N->getOperand(0);
  SDValue Elt = N->getOperand(1);
  SDValue Idx = N->getOperand(2);
  SDLoc dl(N);
  GetSplitVector(Vec, Lo, Hi);
 
  if (ConstantSDNode *CIdx = dyn_cast<ConstantSDNode>(Idx)) {
    unsigned IdxVal = CIdx->getZExtValue();
    unsigned LoNumElts = Lo.getValueType().getVectorMinNumElements();
    if (IdxVal < LoNumElts) {
      Lo = DAG.getNode(ISD::INSERT_VECTOR_ELT, dl,
                       Lo.getValueType(), Lo, Elt, Idx);
      return;
    } else if (!Vec.getValueType().isScalableVector()) {
      Hi = DAG.getNode(ISD::INSERT_VECTOR_ELT, dl, Hi.getValueType(), Hi, Elt,
                       DAG.getVectorIdxConstant(IdxVal - LoNumElts, dl));
      return;
    }
  }
 
  // Make the vector elements byte-addressable if they aren't already.
  EVT VecVT = Vec.getValueType();
  EVT EltVT = VecVT.getVectorElementType();
  if (!EltVT.isByteSized()) {
    EltVT = EltVT.changeTypeToInteger().getRoundIntegerType(*DAG.getContext());
    VecVT = VecVT.changeElementType(EltVT);
    Vec = DAG.getNode(ISD::ANY_EXTEND, dl, VecVT, Vec);
    // Extend the element type to match if needed.
    if (EltVT.bitsGT(Elt.getValueType()))
      Elt = DAG.getNode(ISD::ANY_EXTEND, dl, EltVT, Elt);
  }
 
  // Spill the vector to the stack.
  // In cases where the vector is illegal it will be broken down into parts
  // and stored in parts - we should use the alignment for the smallest part.
  Align SmallestAlign = DAG.getReducedAlign(VecVT, /*UseABI=*/false);
  SDValue StackPtr =
      DAG.CreateStackTemporary(VecVT.getStoreSize(), SmallestAlign);
  auto &MF = DAG.getMachineFunction();
  auto FrameIndex = cast<FrameIndexSDNode>(StackPtr.getNode())->getIndex();
  auto PtrInfo = MachinePointerInfo::getFixedStack(MF, FrameIndex);
 
  SDValue Store = DAG.getStore(DAG.getEntryNode(), dl, Vec, StackPtr, PtrInfo,
                               SmallestAlign);
 
  // Store the new element.  This may be larger than the vector element type,
  // so use a truncating store.
  SDValue EltPtr = TLI.getVectorElementPointer(DAG, StackPtr, VecVT, Idx);
  Store = DAG.getTruncStore(
      Store, dl, Elt, EltPtr, MachinePointerInfo::getUnknownStack(MF), EltVT,
      commonAlignment(SmallestAlign,
                      EltVT.getFixedSizeInBits() / 8));
 
  EVT LoVT, HiVT;
  std::tie(LoVT, HiVT) = DAG.GetSplitDestVTs(VecVT);
 
  // Load the Lo part from the stack slot.
  Lo = DAG.getLoad(LoVT, dl, Store, StackPtr, PtrInfo, SmallestAlign);
 
  // Increment the pointer to the other part.
  auto Load = cast<LoadSDNode>(Lo);
  MachinePointerInfo MPI = Load->getPointerInfo();
  IncrementPointer(Load, LoVT, MPI, StackPtr);
 
  Hi = DAG.getLoad(HiVT, dl, Store, StackPtr, MPI, SmallestAlign);
 
  // If we adjusted the original type, we need to truncate the results.
  std::tie(LoVT, HiVT) = DAG.GetSplitDestVTs(N->getValueType(0));
  if (LoVT != Lo.getValueType())
    Lo = DAG.getNode(ISD::TRUNCATE, dl, LoVT, Lo);
  if (HiVT != Hi.getValueType())
    Hi = DAG.getNode(ISD::TRUNCATE, dl, HiVT, Hi);
}
 
void DAGTypeLegalizer::SplitVecRes_STEP_VECTOR(SDNode *N, SDValue &Lo,
                                               SDValue &Hi) {
  EVT LoVT, HiVT;
  SDLoc dl(N);
  assert(N->getValueType(0).isScalableVector() &&
         "Only scalable vectors are supported for STEP_VECTOR");
  std::tie(LoVT, HiVT) = DAG.GetSplitDestVTs(N->getValueType(0));
  SDValue Step = N->getOperand(0);
 
  Lo = DAG.getNode(ISD::STEP_VECTOR, dl, LoVT, Step);
 
  // Hi = Lo + (EltCnt * Step)
  EVT EltVT = Step.getValueType();
  APInt StepVal = Step->getAsAPIntVal();
  SDValue StartOfHi =
      DAG.getVScale(dl, EltVT, StepVal * LoVT.getVectorMinNumElements());
  StartOfHi = DAG.getSExtOrTrunc(StartOfHi, dl, HiVT.getVectorElementType());
  StartOfHi = DAG.getNode(ISD::SPLAT_VECTOR, dl, HiVT, StartOfHi);
 
  Hi = DAG.getNode(ISD::STEP_VECTOR, dl, HiVT, Step);
  Hi = DAG.getNode(ISD::ADD, dl, HiVT, Hi, StartOfHi);
}
 
void DAGTypeLegalizer::SplitVecRes_ScalarOp(SDNode *N, SDValue &Lo,
                                            SDValue &Hi) {
  EVT LoVT, HiVT;
  SDLoc dl(N);
  std::tie(LoVT, HiVT) = DAG.GetSplitDestVTs(N->getValueType(0));
  Lo = DAG.getNode(N->getOpcode(), dl, LoVT, N->getOperand(0));
  if (N->getOpcode() == ISD::SCALAR_TO_VECTOR) {
    Hi = DAG.getUNDEF(HiVT);
  } else {
    assert(N->getOpcode() == ISD::SPLAT_VECTOR && "Unexpected opcode");
    Hi = Lo;
  }
}
 
void DAGTypeLegalizer::SplitVecRes_VP_SPLAT(SDNode *N, SDValue &Lo,
                                            SDValue &Hi) {
  SDLoc dl(N);
  auto [LoVT, HiVT] = DAG.GetSplitDestVTs(N->getValueType(0));
  auto [MaskLo, MaskHi] = SplitMask(N->getOperand(1));
  auto [EVLLo, EVLHi] = DAG.SplitEVL(N->getOperand(2), N->getValueType(0), dl);
  Lo = DAG.getNode(N->getOpcode(), dl, LoVT, N->getOperand(0), MaskLo, EVLLo);
  Hi = DAG.getNode(N->getOpcode(), dl, HiVT, N->getOperand(0), MaskHi, EVLHi);
}
 
void DAGTypeLegalizer::SplitVecRes_LOAD(LoadSDNode *LD, SDValue &Lo,
                                        SDValue &Hi) {
  assert(ISD::isUNINDEXEDLoad(LD) && "Indexed load during type legalization!");
  EVT LoVT, HiVT;
  SDLoc dl(LD);
  std::tie(LoVT, HiVT) = DAG.GetSplitDestVTs(LD->getValueType(0));
 
  ISD::LoadExtType ExtType = LD->getExtensionType();
  SDValue Ch = LD->getChain();
  SDValue Ptr = LD->getBasePtr();
  SDValue Offset = DAG.getUNDEF(Ptr.getValueType());
  EVT MemoryVT = LD->getMemoryVT();
  MachineMemOperand::Flags MMOFlags = LD->getMemOperand()->getFlags();
  AAMDNodes AAInfo = LD->getAAInfo();
 
  EVT LoMemVT, HiMemVT;
  std::tie(LoMemVT, HiMemVT) = DAG.GetSplitDestVTs(MemoryVT);
 
  if (!LoMemVT.isByteSized() || !HiMemVT.isByteSized()) {
    SDValue Value, NewChain;
    std::tie(Value, NewChain) = TLI.scalarizeVectorLoad(LD, DAG);
    std::tie(Lo, Hi) = DAG.SplitVector(Value, dl);
    ReplaceValueWith(SDValue(LD, 1), NewChain);
    return;
  }
 
  Lo = DAG.getLoad(ISD::UNINDEXED, ExtType, LoVT, dl, Ch, Ptr, Offset,
                   LD->getPointerInfo(), LoMemVT, LD->getOriginalAlign(),
                   MMOFlags, AAInfo);
 
  MachinePointerInfo MPI;
  IncrementPointer(LD, LoMemVT, MPI, Ptr);
 
  Hi = DAG.getLoad(ISD::UNINDEXED, ExtType, HiVT, dl, Ch, Ptr, Offset, MPI,
                   HiMemVT, LD->getOriginalAlign(), MMOFlags, AAInfo);
 
  // Build a factor node to remember that this load is independent of the
  // other one.
  Ch = DAG.getNode(ISD::TokenFactor, dl, MVT::Other, Lo.getValue(1),
                   Hi.getValue(1));
 
  // Legalize the chain result - switch anything that used the old chain to
  // use the new one.
  ReplaceValueWith(SDValue(LD, 1), Ch);
}
 
void DAGTypeLegalizer::SplitVecRes_VP_LOAD(VPLoadSDNode *LD, SDValue &Lo,
                                           SDValue &Hi) {
  assert(LD->isUnindexed() && "Indexed VP load during type legalization!");
  EVT LoVT, HiVT;
  SDLoc dl(LD);
  std::tie(LoVT, HiVT) = DAG.GetSplitDestVTs(LD->getValueType(0));
 
  ISD::LoadExtType ExtType = LD->getExtensionType();
  SDValue Ch = LD->getChain();
  SDValue Ptr = LD->getBasePtr();
  SDValue Offset = LD->getOffset();
  assert(Offset.isUndef() && "Unexpected indexed variable-length load offset");
  Align Alignment = LD->getOriginalAlign();
  SDValue Mask = LD->getMask();
  SDValue EVL = LD->getVectorLength();
  EVT MemoryVT = LD->getMemoryVT();
 
  EVT LoMemVT, HiMemVT;
  bool HiIsEmpty = false;
  std::tie(LoMemVT, HiMemVT) =
      DAG.GetDependentSplitDestVTs(MemoryVT, LoVT, &HiIsEmpty);
 
  // Split Mask operand
  SDValue MaskLo, MaskHi;
  if (Mask.getOpcode() == ISD::SETCC) {
    SplitVecRes_SETCC(Mask.getNode(), MaskLo, MaskHi);
  } else {
    if (getTypeAction(Mask.getValueType()) == TargetLowering::TypeSplitVector)
      GetSplitVector(Mask, MaskLo, MaskHi);
    else
      std::tie(MaskLo, MaskHi) = DAG.SplitVector(Mask, dl);
  }
 
  // Split EVL operand
  SDValue EVLLo, EVLHi;
  std::tie(EVLLo, EVLHi) = DAG.SplitEVL(EVL, LD->getValueType(0), dl);
 
  MachineMemOperand *MMO = DAG.getMachineFunction().getMachineMemOperand(
      LD->getPointerInfo(), MachineMemOperand::MOLoad,
      LocationSize::beforeOrAfterPointer(), Alignment, LD->getAAInfo(),
      LD->getRanges());
 
  Lo =
      DAG.getLoadVP(LD->getAddressingMode(), ExtType, LoVT, dl, Ch, Ptr, Offset,
                    MaskLo, EVLLo, LoMemVT, MMO, LD->isExpandingLoad());
 
  if (HiIsEmpty) {
    // The hi vp_load has zero storage size. We therefore simply set it to
    // the low vp_load and rely on subsequent removal from the chain.
    Hi = Lo;
  } else {
    // Generate hi vp_load.
    Ptr = TLI.IncrementMemoryAddress(Ptr, MaskLo, dl, LoMemVT, DAG,
                                     LD->isExpandingLoad());
 
    MachinePointerInfo MPI;
    if (LoMemVT.isScalableVector())
      MPI = MachinePointerInfo(LD->getPointerInfo().getAddrSpace());
    else
      MPI = LD->getPointerInfo().getWithOffset(
          LoMemVT.getStoreSize().getFixedValue());
 
    MMO = DAG.getMachineFunction().getMachineMemOperand(
        MPI, MachineMemOperand::MOLoad, LocationSize::beforeOrAfterPointer(),
        Alignment, LD->getAAInfo(), LD->getRanges());
 
    Hi = DAG.getLoadVP(LD->getAddressingMode(), ExtType, HiVT, dl, Ch, Ptr,
                       Offset, MaskHi, EVLHi, HiMemVT, MMO,
                       LD->isExpandingLoad());
  }
 
  // Build a factor node to remember that this load is independent of the
  // other one.
  Ch = DAG.getNode(ISD::TokenFactor, dl, MVT::Other, Lo.getValue(1),
                   Hi.getValue(1));
 
  // Legalize the chain result - switch anything that used the old chain to
  // use the new one.
  ReplaceValueWith(SDValue(LD, 1), Ch);
}
 
void DAGTypeLegalizer::SplitVecRes_VP_STRIDED_LOAD(VPStridedLoadSDNode *SLD,
                                                   SDValue &Lo, SDValue &Hi) {
  assert(SLD->isUnindexed() &&
         "Indexed VP strided load during type legalization!");
  assert(SLD->getOffset().isUndef() &&
         "Unexpected indexed variable-length load offset");
 
  SDLoc DL(SLD);
 
  EVT LoVT, HiVT;
  std::tie(LoVT, HiVT) = DAG.GetSplitDestVTs(SLD->getValueType(0));
 
  EVT LoMemVT, HiMemVT;
  bool HiIsEmpty = false;
  std::tie(LoMemVT, HiMemVT) =
      DAG.GetDependentSplitDestVTs(SLD->getMemoryVT(), LoVT, &HiIsEmpty);
 
  SDValue Mask = SLD->getMask();
  SDValue LoMask, HiMask;
  if (Mask.getOpcode() == ISD::SETCC) {
    SplitVecRes_SETCC(Mask.getNode(), LoMask, HiMask);
  } else {
    if (getTypeAction(Mask.getValueType()) == TargetLowering::TypeSplitVector)
      GetSplitVector(Mask, LoMask, HiMask);
    else
      std::tie(LoMask, HiMask) = DAG.SplitVector(Mask, DL);
  }
 
  SDValue LoEVL, HiEVL;
  std::tie(LoEVL, HiEVL) =
      DAG.SplitEVL(SLD->getVectorLength(), SLD->getValueType(0), DL);
 
  // Generate the low vp_strided_load
  Lo = DAG.getStridedLoadVP(
      SLD->getAddressingMode(), SLD->getExtensionType(), LoVT, DL,
      SLD->getChain(), SLD->getBasePtr(), SLD->getOffset(), SLD->getStride(),
      LoMask, LoEVL, LoMemVT, SLD->getMemOperand(), SLD->isExpandingLoad());
 
  if (HiIsEmpty) {
    // The high vp_strided_load has zero storage size. We therefore simply set
    // it to the low vp_strided_load and rely on subsequent removal from the
    // chain.
    Hi = Lo;
  } else {
    // Generate the high vp_strided_load.
    // To calculate the high base address, we need to sum to the low base
    // address stride number of bytes for each element already loaded by low,
    // that is: Ptr = Ptr + (LoEVL * Stride)
    EVT PtrVT = SLD->getBasePtr().getValueType();
    SDValue Increment =
        DAG.getNode(ISD::MUL, DL, PtrVT, LoEVL,
                    DAG.getSExtOrTrunc(SLD->getStride(), DL, PtrVT));
    SDValue Ptr =
        DAG.getNode(ISD::ADD, DL, PtrVT, SLD->getBasePtr(), Increment);
 
    Align Alignment = SLD->getOriginalAlign();
    if (LoMemVT.isScalableVector())
      Alignment = commonAlignment(
          Alignment, LoMemVT.getSizeInBits().getKnownMinValue() / 8);
 
    MachineMemOperand *MMO = DAG.getMachineFunction().getMachineMemOperand(
        MachinePointerInfo(SLD->getPointerInfo().getAddrSpace()),
        MachineMemOperand::MOLoad, LocationSize::beforeOrAfterPointer(),
        Alignment, SLD->getAAInfo(), SLD->getRanges());
 
    Hi = DAG.getStridedLoadVP(SLD->getAddressingMode(), SLD->getExtensionType(),
                              HiVT, DL, SLD->getChain(), Ptr, SLD->getOffset(),
                              SLD->getStride(), HiMask, HiEVL, HiMemVT, MMO,
                              SLD->isExpandingLoad());
  }
 
  // Build a factor node to remember that this load is independent of the
  // other one.
  SDValue Ch = DAG.getNode(ISD::TokenFactor, DL, MVT::Other, Lo.getValue(1),
                           Hi.getValue(1));
 
  // Legalize the chain result - switch anything that used the old chain to
  // use the new one.
  ReplaceValueWith(SDValue(SLD, 1), Ch);
}
 
void DAGTypeLegalizer::SplitVecRes_MLOAD(MaskedLoadSDNode *MLD,
                                         SDValue &Lo, SDValue &Hi) {
  assert(MLD->isUnindexed() && "Indexed masked load during type legalization!");
  EVT LoVT, HiVT;
  SDLoc dl(MLD);
  std::tie(LoVT, HiVT) = DAG.GetSplitDestVTs(MLD->getValueType(0));
 
  SDValue Ch = MLD->getChain();
  SDValue Ptr = MLD->getBasePtr();
  SDValue Offset = MLD->getOffset();
  assert(Offset.isUndef() && "Unexpected indexed masked load offset");
  SDValue Mask = MLD->getMask();
  SDValue PassThru = MLD->getPassThru();
  Align Alignment = MLD->getOriginalAlign();
  ISD::LoadExtType ExtType = MLD->getExtensionType();
 
  // Split Mask operand
  SDValue MaskLo, MaskHi;
  if (Mask.getOpcode() == ISD::SETCC) {
    SplitVecRes_SETCC(Mask.getNode(), MaskLo, MaskHi);
  } else {
    if (getTypeAction(Mask.getValueType()) == TargetLowering::TypeSplitVector)
      GetSplitVector(Mask, MaskLo, MaskHi);
    else
      std::tie(MaskLo, MaskHi) = DAG.SplitVector(Mask, dl);
  }
 
  EVT MemoryVT = MLD->getMemoryVT();
  EVT LoMemVT, HiMemVT;
  bool HiIsEmpty = false;
  std::tie(LoMemVT, HiMemVT) =
      DAG.GetDependentSplitDestVTs(MemoryVT, LoVT, &HiIsEmpty);
 
  SDValue PassThruLo, PassThruHi;
  if (getTypeAction(PassThru.getValueType()) == TargetLowering::TypeSplitVector)
    GetSplitVector(PassThru, PassThruLo, PassThruHi);
  else
    std::tie(PassThruLo, PassThruHi) = DAG.SplitVector(PassThru, dl);
 
  MachineMemOperand *MMO = DAG.getMachineFunction().getMachineMemOperand(
      MLD->getPointerInfo(), MachineMemOperand::MOLoad,
      LocationSize::beforeOrAfterPointer(), Alignment, MLD->getAAInfo(),
      MLD->getRanges());
 
  Lo = DAG.getMaskedLoad(LoVT, dl, Ch, Ptr, Offset, MaskLo, PassThruLo, LoMemVT,
                         MMO, MLD->getAddressingMode(), ExtType,
                         MLD->isExpandingLoad());
 
  if (HiIsEmpty) {
    // The hi masked load has zero storage size. We therefore simply set it to
    // the low masked load and rely on subsequent removal from the chain.
    Hi = Lo;
  } else {
    // Generate hi masked load.
    Ptr = TLI.IncrementMemoryAddress(Ptr, MaskLo, dl, LoMemVT, DAG,
                                     MLD->isExpandingLoad());
 
    MachinePointerInfo MPI;
    if (LoMemVT.isScalableVector())
      MPI = MachinePointerInfo(MLD->getPointerInfo().getAddrSpace());
    else
      MPI = MLD->getPointerInfo().getWithOffset(
          LoMemVT.getStoreSize().getFixedValue());
 
    MMO = DAG.getMachineFunction().getMachineMemOperand(
        MPI, MachineMemOperand::MOLoad, LocationSize::beforeOrAfterPointer(),
        Alignment, MLD->getAAInfo(), MLD->getRanges());
 
    Hi = DAG.getMaskedLoad(HiVT, dl, Ch, Ptr, Offset, MaskHi, PassThruHi,
                           HiMemVT, MMO, MLD->getAddressingMode(), ExtType,
                           MLD->isExpandingLoad());
  }
 
  // Build a factor node to remember that this load is independent of the
  // other one.
  Ch = DAG.getNode(ISD::TokenFactor, dl, MVT::Other, Lo.getValue(1),
                   Hi.getValue(1));
 
  // Legalize the chain result - switch anything that used the old chain to
  // use the new one.
  ReplaceValueWith(SDValue(MLD, 1), Ch);
 
}
 
void DAGTypeLegalizer::SplitVecRes_Gather(MemSDNode *N, SDValue &Lo,
                                          SDValue &Hi, bool SplitSETCC) {
  EVT LoVT, HiVT;
  SDLoc dl(N);
  std::tie(LoVT, HiVT) = DAG.GetSplitDestVTs(N->getValueType(0));
 
  SDValue Ch = N->getChain();
  SDValue Ptr = N->getBasePtr();
  struct Operands {
    SDValue Mask;
    SDValue Index;
    SDValue Scale;
  } Ops = [&]() -> Operands {
    if (auto *MSC = dyn_cast<MaskedGatherSDNode>(N)) {
      return {MSC->getMask(), MSC->getIndex(), MSC->getScale()};
    }
    auto *VPSC = cast<VPGatherSDNode>(N);
    return {VPSC->getMask(), VPSC->getIndex(), VPSC->getScale()};
  }();
 
  EVT MemoryVT = N->getMemoryVT();
  Align Alignment = N->getOriginalAlign();
 
  // Split Mask operand
  SDValue MaskLo, MaskHi;
  if (SplitSETCC && Ops.Mask.getOpcode() == ISD::SETCC) {
    SplitVecRes_SETCC(Ops.Mask.getNode(), MaskLo, MaskHi);
  } else {
    std::tie(MaskLo, MaskHi) = SplitMask(Ops.Mask, dl);
  }
 
  EVT LoMemVT, HiMemVT;
  // Split MemoryVT
  std::tie(LoMemVT, HiMemVT) = DAG.GetSplitDestVTs(MemoryVT);
 
  SDValue IndexHi, IndexLo;
  if (getTypeAction(Ops.Index.getValueType()) ==
      TargetLowering::TypeSplitVector)
    GetSplitVector(Ops.Index, IndexLo, IndexHi);
  else
    std::tie(IndexLo, IndexHi) = DAG.SplitVector(Ops.Index, dl);
 
  MachineMemOperand *MMO = DAG.getMachineFunction().getMachineMemOperand(
      N->getPointerInfo(), MachineMemOperand::MOLoad,
      LocationSize::beforeOrAfterPointer(), Alignment, N->getAAInfo(),
      N->getRanges());
 
  if (auto *MGT = dyn_cast<MaskedGatherSDNode>(N)) {
    SDValue PassThru = MGT->getPassThru();
    SDValue PassThruLo, PassThruHi;
    if (getTypeAction(PassThru.getValueType()) ==
        TargetLowering::TypeSplitVector)
      GetSplitVector(PassThru, PassThruLo, PassThruHi);
    else
      std::tie(PassThruLo, PassThruHi) = DAG.SplitVector(PassThru, dl);
 
    ISD::LoadExtType ExtType = MGT->getExtensionType();
    ISD::MemIndexType IndexTy = MGT->getIndexType();
 
    SDValue OpsLo[] = {Ch, PassThruLo, MaskLo, Ptr, IndexLo, Ops.Scale};
    Lo = DAG.getMaskedGather(DAG.getVTList(LoVT, MVT::Other), LoMemVT, dl,
                             OpsLo, MMO, IndexTy, ExtType);
 
    SDValue OpsHi[] = {Ch, PassThruHi, MaskHi, Ptr, IndexHi, Ops.Scale};
    Hi = DAG.getMaskedGather(DAG.getVTList(HiVT, MVT::Other), HiMemVT, dl,
                             OpsHi, MMO, IndexTy, ExtType);
  } else {
    auto *VPGT = cast<VPGatherSDNode>(N);
    SDValue EVLLo, EVLHi;
    std::tie(EVLLo, EVLHi) =
        DAG.SplitEVL(VPGT->getVectorLength(), MemoryVT, dl);
 
    SDValue OpsLo[] = {Ch, Ptr, IndexLo, Ops.Scale, MaskLo, EVLLo};
    Lo = DAG.getGatherVP(DAG.getVTList(LoVT, MVT::Other), LoMemVT, dl, OpsLo,
                         MMO, VPGT->getIndexType());
 
    SDValue OpsHi[] = {Ch, Ptr, IndexHi, Ops.Scale, MaskHi, EVLHi};
    Hi = DAG.getGatherVP(DAG.getVTList(HiVT, MVT::Other), HiMemVT, dl, OpsHi,
                         MMO, VPGT->getIndexType());
  }
 
  // Build a factor node to remember that this load is independent of the
  // other one.
  Ch = DAG.getNode(ISD::TokenFactor, dl, MVT::Other, Lo.getValue(1),
                   Hi.getValue(1));
 
  // Legalize the chain result - switch anything that used the old chain to
  // use the new one.
  ReplaceValueWith(SDValue(N, 1), Ch);
}
 
void DAGTypeLegalizer::SplitVecRes_VECTOR_COMPRESS(SDNode *N, SDValue &Lo,
                                                   SDValue &Hi) {
  // This is not "trivial", as there is a dependency between the two subvectors.
  // Depending on the number of 1s in the mask, the elements from the Hi vector
  // need to be moved to the Lo vector. Passthru values make this even harder.
  // We try to use VECTOR_COMPRESS if the target has custom lowering with
  // smaller types and passthru is undef, as it is most likely faster than the
  // fully expand path. Otherwise, just do the full expansion as one "big"
  // operation and then extract the Lo and Hi vectors from that. This gets
  // rid of VECTOR_COMPRESS and all other operands can be legalized later.
  SDLoc DL(N);
  EVT VecVT = N->getValueType(0);
 
  auto [LoVT, HiVT] = DAG.GetSplitDestVTs(VecVT);
  bool HasCustomLowering = false;
  EVT CheckVT = LoVT;
  while (CheckVT.getVectorMinNumElements() > 1) {
    // TLI.isOperationLegalOrCustom requires a legal type, but we could have a
    // custom lowering for illegal types. So we do the checks separately.
    if (TLI.isOperationLegal(ISD::VECTOR_COMPRESS, CheckVT) ||
        TLI.isOperationCustom(ISD::VECTOR_COMPRESS, CheckVT)) {
      HasCustomLowering = true;
      break;
    }
    CheckVT = CheckVT.getHalfNumVectorElementsVT(*DAG.getContext());
  }
 
  SDValue Passthru = N->getOperand(2);
  if (!HasCustomLowering || !Passthru.isUndef()) {
    SDValue Compressed = TLI.expandVECTOR_COMPRESS(N, DAG);
    std::tie(Lo, Hi) = DAG.SplitVector(Compressed, DL, LoVT, HiVT);
    return;
  }
 
  // Try to VECTOR_COMPRESS smaller vectors and combine via a stack store+load.
  SDValue LoMask, HiMask;
  std::tie(Lo, Hi) = DAG.SplitVectorOperand(N, 0);
  std::tie(LoMask, HiMask) = SplitMask(N->getOperand(1));
 
  SDValue UndefPassthru = DAG.getUNDEF(LoVT);
  Lo = DAG.getNode(ISD::VECTOR_COMPRESS, DL, LoVT, Lo, LoMask, UndefPassthru);
  Hi = DAG.getNode(ISD::VECTOR_COMPRESS, DL, HiVT, Hi, HiMask, UndefPassthru);
 
  SDValue StackPtr = DAG.CreateStackTemporary(
      VecVT.getStoreSize(), DAG.getReducedAlign(VecVT, /*UseABI=*/false));
  MachineFunction &MF = DAG.getMachineFunction();
  MachinePointerInfo PtrInfo = MachinePointerInfo::getFixedStack(
      MF, cast<FrameIndexSDNode>(StackPtr.getNode())->getIndex());
 
  // We store LoVec and then insert HiVec starting at offset=|1s| in LoMask.
  SDValue WideMask =
      DAG.getNode(ISD::ZERO_EXTEND, DL, LoMask.getValueType(), LoMask);
  SDValue Offset = DAG.getNode(ISD::VECREDUCE_ADD, DL, MVT::i32, WideMask);
  Offset = TLI.getVectorElementPointer(DAG, StackPtr, VecVT, Offset);
 
  SDValue Chain = DAG.getEntryNode();
  Chain = DAG.getStore(Chain, DL, Lo, StackPtr, PtrInfo);
  Chain = DAG.getStore(Chain, DL, Hi, Offset,
                       MachinePointerInfo::getUnknownStack(MF));
 
  SDValue Compressed = DAG.getLoad(VecVT, DL, Chain, StackPtr, PtrInfo);
  std::tie(Lo, Hi) = DAG.SplitVector(Compressed, DL);
}
 
void DAGTypeLegalizer::SplitVecRes_SETCC(SDNode *N, SDValue &Lo, SDValue &Hi) {
  assert(N->getValueType(0).isVector() &&
         N->getOperand(0).getValueType().isVector() &&
         "Operand types must be vectors");
 
  EVT LoVT, HiVT;
  SDLoc DL(N);
  std::tie(LoVT, HiVT) = DAG.GetSplitDestVTs(N->getValueType(0));
 
  // If the input also splits, handle it directly. Otherwise split it by hand.
  SDValue LL, LH, RL, RH;
  if (getTypeAction(N->getOperand(0).getValueType()) ==
      TargetLowering::TypeSplitVector)
    GetSplitVector(N->getOperand(0), LL, LH);
  else
    std::tie(LL, LH) = DAG.SplitVectorOperand(N, 0);
 
  if (getTypeAction(N->getOperand(1).getValueType()) ==
      TargetLowering::TypeSplitVector)
    GetSplitVector(N->getOperand(1), RL, RH);
  else
    std::tie(RL, RH) = DAG.SplitVectorOperand(N, 1);
 
  if (N->getOpcode() == ISD::SETCC) {
    Lo = DAG.getNode(N->getOpcode(), DL, LoVT, LL, RL, N->getOperand(2));
    Hi = DAG.getNode(N->getOpcode(), DL, HiVT, LH, RH, N->getOperand(2));
  } else {
    assert(N->getOpcode() == ISD::VP_SETCC && "Expected VP_SETCC opcode");
    SDValue MaskLo, MaskHi, EVLLo, EVLHi;
    std::tie(MaskLo, MaskHi) = SplitMask(N->getOperand(3));
    std::tie(EVLLo, EVLHi) =
        DAG.SplitEVL(N->getOperand(4), N->getValueType(0), DL);
    Lo = DAG.getNode(N->getOpcode(), DL, LoVT, LL, RL, N->getOperand(2), MaskLo,
                     EVLLo);
    Hi = DAG.getNode(N->getOpcode(), DL, HiVT, LH, RH, N->getOperand(2), MaskHi,
                     EVLHi);
  }
}
 
void DAGTypeLegalizer::SplitVecRes_UnaryOp(SDNode *N, SDValue &Lo,
                                           SDValue &Hi) {
  // Get the dest types - they may not match the input types, e.g. int_to_fp.
  EVT LoVT, HiVT;
  SDLoc dl(N);
  std::tie(LoVT, HiVT) = DAG.GetSplitDestVTs(N->getValueType(0));
 
  // If the input also splits, handle it directly for a compile time speedup.
  // Otherwise split it by hand.
  EVT InVT = N->getOperand(0).getValueType();
  if (getTypeAction(InVT) == TargetLowering::TypeSplitVector)
    GetSplitVector(N->getOperand(0), Lo, Hi);
  else
    std::tie(Lo, Hi) = DAG.SplitVectorOperand(N, 0);
 
  const SDNodeFlags Flags = N->getFlags();
  unsigned Opcode = N->getOpcode();
  if (N->getNumOperands() <= 2) {
    if (Opcode == ISD::FP_ROUND) {
      Lo = DAG.getNode(Opcode, dl, LoVT, Lo, N->getOperand(1), Flags);
      Hi = DAG.getNode(Opcode, dl, HiVT, Hi, N->getOperand(1), Flags);
    } else {
      Lo = DAG.getNode(Opcode, dl, LoVT, Lo, Flags);
      Hi = DAG.getNode(Opcode, dl, HiVT, Hi, Flags);
    }
    return;
  }
 
  assert(N->getNumOperands() == 3 && "Unexpected number of operands!");
  assert(N->isVPOpcode() && "Expected VP opcode");
 
  SDValue MaskLo, MaskHi;
  std::tie(MaskLo, MaskHi) = SplitMask(N->getOperand(1));
 
  SDValue EVLLo, EVLHi;
  std::tie(EVLLo, EVLHi) =
      DAG.SplitEVL(N->getOperand(2), N->getValueType(0), dl);
 
  Lo = DAG.getNode(Opcode, dl, LoVT, {Lo, MaskLo, EVLLo}, Flags);
  Hi = DAG.getNode(Opcode, dl, HiVT, {Hi, MaskHi, EVLHi}, Flags);
}
 
void DAGTypeLegalizer::SplitVecRes_ADDRSPACECAST(SDNode *N, SDValue &Lo,
                                                 SDValue &Hi) {
  SDLoc dl(N);
  auto [LoVT, HiVT] = DAG.GetSplitDestVTs(N->getValueType(0));
 
  // If the input also splits, handle it directly for a compile time speedup.
  // Otherwise split it by hand.
  EVT InVT = N->getOperand(0).getValueType();
  if (getTypeAction(InVT) == TargetLowering::TypeSplitVector)
    GetSplitVector(N->getOperand(0), Lo, Hi);
  else
    std::tie(Lo, Hi) = DAG.SplitVectorOperand(N, 0);
 
  auto *AddrSpaceCastN = cast<AddrSpaceCastSDNode>(N);
  unsigned SrcAS = AddrSpaceCastN->getSrcAddressSpace();
  unsigned DestAS = AddrSpaceCastN->getDestAddressSpace();
  Lo = DAG.getAddrSpaceCast(dl, LoVT, Lo, SrcAS, DestAS);
  Hi = DAG.getAddrSpaceCast(dl, HiVT, Hi, SrcAS, DestAS);
}
 
void DAGTypeLegalizer::SplitVecRes_FFREXP(SDNode *N, unsigned ResNo,
                                          SDValue &Lo, SDValue &Hi) {
  SDLoc dl(N);
  auto [LoVT, HiVT] = DAG.GetSplitDestVTs(N->getValueType(0));
  auto [LoVT1, HiVT1] = DAG.GetSplitDestVTs(N->getValueType(1));
 
  // If the input also splits, handle it directly for a compile time speedup.
  // Otherwise split it by hand.
  EVT InVT = N->getOperand(0).getValueType();
  if (getTypeAction(InVT) == TargetLowering::TypeSplitVector)
    GetSplitVector(N->getOperand(0), Lo, Hi);
  else
    std::tie(Lo, Hi) = DAG.SplitVectorOperand(N, 0);
 
  Lo = DAG.getNode(N->getOpcode(), dl, {LoVT, LoVT1}, Lo);
  Hi = DAG.getNode(N->getOpcode(), dl, {HiVT, HiVT1}, Hi);
  Lo->setFlags(N->getFlags());
  Hi->setFlags(N->getFlags());
 
  SDNode *HiNode = Hi.getNode();
  SDNode *LoNode = Lo.getNode();
 
  // Replace the other vector result not being explicitly split here.
  unsigned OtherNo = 1 - ResNo;
  EVT OtherVT = N->getValueType(OtherNo);
  if (getTypeAction(OtherVT) == TargetLowering::TypeSplitVector) {
    SetSplitVector(SDValue(N, OtherNo), SDValue(LoNode, OtherNo),
                   SDValue(HiNode, OtherNo));
  } else {
    SDValue OtherVal =
        DAG.getNode(ISD::CONCAT_VECTORS, dl, OtherVT, SDValue(LoNode, OtherNo),
                    SDValue(HiNode, OtherNo));
    ReplaceValueWith(SDValue(N, OtherNo), OtherVal);
  }
}
 
void DAGTypeLegalizer::SplitVecRes_ExtendOp(SDNode *N, SDValue &Lo,
                                            SDValue &Hi) {
  SDLoc dl(N);
  EVT SrcVT = N->getOperand(0).getValueType();
  EVT DestVT = N->getValueType(0);
  EVT LoVT, HiVT;
  std::tie(LoVT, HiVT) = DAG.GetSplitDestVTs(DestVT);
 
  // We can do better than a generic split operation if the extend is doing
  // more than just doubling the width of the elements and the following are
  // true:
  //   - The number of vector elements is even,
  //   - the source type is legal,
  //   - the type of a split source is illegal,
  //   - the type of an extended (by doubling element size) source is legal, and
  //   - the type of that extended source when split is legal.
  //
  // This won't necessarily completely legalize the operation, but it will
  // more effectively move in the right direction and prevent falling down
  // to scalarization in many cases due to the input vector being split too
  // far.
  if (SrcVT.getVectorElementCount().isKnownEven() &&
      SrcVT.getScalarSizeInBits() * 2 < DestVT.getScalarSizeInBits()) {
    LLVMContext &Ctx = *DAG.getContext();
    EVT NewSrcVT = SrcVT.widenIntegerVectorElementType(Ctx);
    EVT SplitSrcVT = SrcVT.getHalfNumVectorElementsVT(Ctx);
 
    EVT SplitLoVT, SplitHiVT;
    std::tie(SplitLoVT, SplitHiVT) = DAG.GetSplitDestVTs(NewSrcVT);
    if (TLI.isTypeLegal(SrcVT) && !TLI.isTypeLegal(SplitSrcVT) &&
        TLI.isTypeLegal(NewSrcVT) && TLI.isTypeLegal(SplitLoVT)) {
      LLVM_DEBUG(dbgs() << "Split vector extend via incremental extend:";
                 N->dump(&DAG); dbgs() << "\n");
      if (!N->isVPOpcode()) {
        // Extend the source vector by one step.
        SDValue NewSrc =
            DAG.getNode(N->getOpcode(), dl, NewSrcVT, N->getOperand(0));
        // Get the low and high halves of the new, extended one step, vector.
        std::tie(Lo, Hi) = DAG.SplitVector(NewSrc, dl);
        // Extend those vector halves the rest of the way.
        Lo = DAG.getNode(N->getOpcode(), dl, LoVT, Lo);
        Hi = DAG.getNode(N->getOpcode(), dl, HiVT, Hi);
        return;
      }
 
      // Extend the source vector by one step.
      SDValue NewSrc =
          DAG.getNode(N->getOpcode(), dl, NewSrcVT, N->getOperand(0),
                      N->getOperand(1), N->getOperand(2));
      // Get the low and high halves of the new, extended one step, vector.
      std::tie(Lo, Hi) = DAG.SplitVector(NewSrc, dl);
 
      SDValue MaskLo, MaskHi;
      std::tie(MaskLo, MaskHi) = SplitMask(N->getOperand(1));
 
      SDValue EVLLo, EVLHi;
      std::tie(EVLLo, EVLHi) =
          DAG.SplitEVL(N->getOperand(2), N->getValueType(0), dl);
      // Extend those vector halves the rest of the way.
      Lo = DAG.getNode(N->getOpcode(), dl, LoVT, {Lo, MaskLo, EVLLo});
      Hi = DAG.getNode(N->getOpcode(), dl, HiVT, {Hi, MaskHi, EVLHi});
      return;
    }
  }
  // Fall back to the generic unary operator splitting otherwise.
  SplitVecRes_UnaryOp(N, Lo, Hi);
}
 
void DAGTypeLegalizer::SplitVecRes_VECTOR_SHUFFLE(ShuffleVectorSDNode *N,
                                                  SDValue &Lo, SDValue &Hi) {
  // The low and high parts of the original input give four input vectors.
  SDValue Inputs[4];
  SDLoc DL(N);
  GetSplitVector(N->getOperand(0), Inputs[0], Inputs[1]);
  GetSplitVector(N->getOperand(1), Inputs[2], Inputs[3]);
  EVT NewVT = Inputs[0].getValueType();
  unsigned NewElts = NewVT.getVectorNumElements();
 
  auto &&IsConstant = [](const SDValue &N) {
    APInt SplatValue;
    return N.getResNo() == 0 &&
           (ISD::isConstantSplatVector(N.getNode(), SplatValue) ||
            ISD::isBuildVectorOfConstantSDNodes(N.getNode()));
  };
  auto &&BuildVector = [NewElts, &DAG = DAG, NewVT, &DL](SDValue &Input1,
                                                         SDValue &Input2,
                                                         ArrayRef<int> Mask) {
    assert(Input1->getOpcode() == ISD::BUILD_VECTOR &&
           Input2->getOpcode() == ISD::BUILD_VECTOR &&
           "Expected build vector node.");
    EVT EltVT = NewVT.getVectorElementType();
    SmallVector<SDValue> Ops(NewElts, DAG.getUNDEF(EltVT));
    for (unsigned I = 0; I < NewElts; ++I) {
      if (Mask[I] == PoisonMaskElem)
        continue;
      unsigned Idx = Mask[I];
      if (Idx >= NewElts)
        Ops[I] = Input2.getOperand(Idx - NewElts);
      else
        Ops[I] = Input1.getOperand(Idx);
      // Make the type of all elements the same as the element type.
      if (Ops[I].getValueType().bitsGT(EltVT))
        Ops[I] = DAG.getNode(ISD::TRUNCATE, DL, EltVT, Ops[I]);
    }
    return DAG.getBuildVector(NewVT, DL, Ops);
  };
 
  // If Lo or Hi uses elements from at most two of the four input vectors, then
  // express it as a vector shuffle of those two inputs.  Otherwise extract the
  // input elements by hand and construct the Lo/Hi output using a BUILD_VECTOR.
  SmallVector<int> OrigMask(N->getMask());
  // Try to pack incoming shuffles/inputs.
  auto &&TryPeekThroughShufflesInputs = [&Inputs, &NewVT, this, NewElts,
                                         &DL](SmallVectorImpl<int> &Mask) {
    // Check if all inputs are shuffles of the same operands or non-shuffles.
    MapVector<std::pair<SDValue, SDValue>, SmallVector<unsigned>> ShufflesIdxs;
    for (unsigned Idx = 0; Idx < std::size(Inputs); ++Idx) {
      SDValue Input = Inputs[Idx];
      auto *Shuffle = dyn_cast<ShuffleVectorSDNode>(Input.getNode());
      if (!Shuffle ||
          Input.getOperand(0).getValueType() != Input.getValueType())
        continue;
      ShufflesIdxs[std::make_pair(Input.getOperand(0), Input.getOperand(1))]
          .push_back(Idx);
      ShufflesIdxs[std::make_pair(Input.getOperand(1), Input.getOperand(0))]
          .push_back(Idx);
    }
    for (auto &P : ShufflesIdxs) {
      if (P.second.size() < 2)
        continue;
      // Use shuffles operands instead of shuffles themselves.
      // 1. Adjust mask.
      for (int &Idx : Mask) {
        if (Idx == PoisonMaskElem)
          continue;
        unsigned SrcRegIdx = Idx / NewElts;
        if (Inputs[SrcRegIdx].isUndef()) {
          Idx = PoisonMaskElem;
          continue;
        }
        auto *Shuffle =
            dyn_cast<ShuffleVectorSDNode>(Inputs[SrcRegIdx].getNode());
        if (!Shuffle || !is_contained(P.second, SrcRegIdx))
          continue;
        int MaskElt = Shuffle->getMaskElt(Idx % NewElts);
        if (MaskElt == PoisonMaskElem) {
          Idx = PoisonMaskElem;
          continue;
        }
        Idx = MaskElt % NewElts +
              P.second[Shuffle->getOperand(MaskElt / NewElts) == P.first.first
                           ? 0
                           : 1] *
                  NewElts;
      }
      // 2. Update inputs.
      Inputs[P.second[0]] = P.first.first;
      Inputs[P.second[1]] = P.first.second;
      // Clear the pair data.
      P.second.clear();
      ShufflesIdxs[std::make_pair(P.first.second, P.first.first)].clear();
    }
    // Check if any concat_vectors can be simplified.
    SmallBitVector UsedSubVector(2 * std::size(Inputs));
    for (int &Idx : Mask) {
      if (Idx == PoisonMaskElem)
        continue;
      unsigned SrcRegIdx = Idx / NewElts;
      if (Inputs[SrcRegIdx].isUndef()) {
        Idx = PoisonMaskElem;
        continue;
      }
      TargetLowering::LegalizeTypeAction TypeAction =
          getTypeAction(Inputs[SrcRegIdx].getValueType());
      if (Inputs[SrcRegIdx].getOpcode() == ISD::CONCAT_VECTORS &&
          Inputs[SrcRegIdx].getNumOperands() == 2 &&
          !Inputs[SrcRegIdx].getOperand(1).isUndef() &&
          (TypeAction == TargetLowering::TypeLegal ||
           TypeAction == TargetLowering::TypeWidenVector))
        UsedSubVector.set(2 * SrcRegIdx + (Idx % NewElts) / (NewElts / 2));
    }
    if (UsedSubVector.count() > 1) {
      SmallVector<SmallVector<std::pair<unsigned, int>, 2>> Pairs;
      for (unsigned I = 0; I < std::size(Inputs); ++I) {
        if (UsedSubVector.test(2 * I) == UsedSubVector.test(2 * I + 1))
          continue;
        if (Pairs.empty() || Pairs.back().size() == 2)
          Pairs.emplace_back();
        if (UsedSubVector.test(2 * I)) {
          Pairs.back().emplace_back(I, 0);
        } else {
          assert(UsedSubVector.test(2 * I + 1) &&
                 "Expected to be used one of the subvectors.");
          Pairs.back().emplace_back(I, 1);
        }
      }
      if (!Pairs.empty() && Pairs.front().size() > 1) {
        // Adjust mask.
        for (int &Idx : Mask) {
          if (Idx == PoisonMaskElem)
            continue;
          unsigned SrcRegIdx = Idx / NewElts;
          auto *It = find_if(
              Pairs, [SrcRegIdx](ArrayRef<std::pair<unsigned, int>> Idxs) {
                return Idxs.front().first == SrcRegIdx ||
                       Idxs.back().first == SrcRegIdx;
              });
          if (It == Pairs.end())
            continue;
          Idx = It->front().first * NewElts + (Idx % NewElts) % (NewElts / 2) +
                (SrcRegIdx == It->front().first ? 0 : (NewElts / 2));
        }
        // Adjust inputs.
        for (ArrayRef<std::pair<unsigned, int>> Idxs : Pairs) {
          Inputs[Idxs.front().first] = DAG.getNode(
              ISD::CONCAT_VECTORS, DL,
              Inputs[Idxs.front().first].getValueType(),
              Inputs[Idxs.front().first].getOperand(Idxs.front().second),
              Inputs[Idxs.back().first].getOperand(Idxs.back().second));
        }
      }
    }
    bool Changed;
    do {
      // Try to remove extra shuffles (except broadcasts) and shuffles with the
      // reused operands.
      Changed = false;
      for (unsigned I = 0; I < std::size(Inputs); ++I) {
        auto *Shuffle = dyn_cast<ShuffleVectorSDNode>(Inputs[I].getNode());
        if (!Shuffle)
          continue;
        if (Shuffle->getOperand(0).getValueType() != NewVT)
          continue;
        int Op = -1;
        if (!Inputs[I].hasOneUse() && Shuffle->getOperand(1).isUndef() &&
            !Shuffle->isSplat()) {
          Op = 0;
        } else if (!Inputs[I].hasOneUse() &&
                   !Shuffle->getOperand(1).isUndef()) {
          // Find the only used operand, if possible.
          for (int &Idx : Mask) {
            if (Idx == PoisonMaskElem)
              continue;
            unsigned SrcRegIdx = Idx / NewElts;
            if (SrcRegIdx != I)
              continue;
            int MaskElt = Shuffle->getMaskElt(Idx % NewElts);
            if (MaskElt == PoisonMaskElem) {
              Idx = PoisonMaskElem;
              continue;
            }
            int OpIdx = MaskElt / NewElts;
            if (Op == -1) {
              Op = OpIdx;
              continue;
            }
            if (Op != OpIdx) {
              Op = -1;
              break;
            }
          }
        }
        if (Op < 0) {
          // Try to check if one of the shuffle operands is used already.
          for (int OpIdx = 0; OpIdx < 2; ++OpIdx) {
            if (Shuffle->getOperand(OpIdx).isUndef())
              continue;
            auto *It = find(Inputs, Shuffle->getOperand(OpIdx));
            if (It == std::end(Inputs))
              continue;
            int FoundOp = std::distance(std::begin(Inputs), It);
            // Found that operand is used already.
            // 1. Fix the mask for the reused operand.
            for (int &Idx : Mask) {
              if (Idx == PoisonMaskElem)
                continue;
              unsigned SrcRegIdx = Idx / NewElts;
              if (SrcRegIdx != I)
                continue;
              int MaskElt = Shuffle->getMaskElt(Idx % NewElts);
              if (MaskElt == PoisonMaskElem) {
                Idx = PoisonMaskElem;
                continue;
              }
              int MaskIdx = MaskElt / NewElts;
              if (OpIdx == MaskIdx)
                Idx = MaskElt % NewElts + FoundOp * NewElts;
            }
            // 2. Set Op to the unused OpIdx.
            Op = (OpIdx + 1) % 2;
            break;
          }
        }
        if (Op >= 0) {
          Changed = true;
          Inputs[I] = Shuffle->getOperand(Op);
          // Adjust mask.
          for (int &Idx : Mask) {
            if (Idx == PoisonMaskElem)
              continue;
            unsigned SrcRegIdx = Idx / NewElts;
            if (SrcRegIdx != I)
              continue;
            int MaskElt = Shuffle->getMaskElt(Idx % NewElts);
            int OpIdx = MaskElt / NewElts;
            if (OpIdx != Op)
              continue;
            Idx = MaskElt % NewElts + SrcRegIdx * NewElts;
          }
        }
      }
    } while (Changed);
  };
  TryPeekThroughShufflesInputs(OrigMask);
  // Proces unique inputs.
  auto &&MakeUniqueInputs = [&Inputs, &IsConstant,
                             NewElts](SmallVectorImpl<int> &Mask) {
    SetVector<SDValue> UniqueInputs;
    SetVector<SDValue> UniqueConstantInputs;
    for (const auto &I : Inputs) {
      if (IsConstant(I))
        UniqueConstantInputs.insert(I);
      else if (!I.isUndef())
        UniqueInputs.insert(I);
    }
    // Adjust mask in case of reused inputs. Also, need to insert constant
    // inputs at first, otherwise it affects the final outcome.
    if (UniqueInputs.size() != std::size(Inputs)) {
      auto &&UniqueVec = UniqueInputs.takeVector();
      auto &&UniqueConstantVec = UniqueConstantInputs.takeVector();
      unsigned ConstNum = UniqueConstantVec.size();
      for (int &Idx : Mask) {
        if (Idx == PoisonMaskElem)
          continue;
        unsigned SrcRegIdx = Idx / NewElts;
        if (Inputs[SrcRegIdx].isUndef()) {
          Idx = PoisonMaskElem;
          continue;
        }
        const auto It = find(UniqueConstantVec, Inputs[SrcRegIdx]);
        if (It != UniqueConstantVec.end()) {
          Idx = (Idx % NewElts) +
                NewElts * std::distance(UniqueConstantVec.begin(), It);
          assert(Idx >= 0 && "Expected defined mask idx.");
          continue;
        }
        const auto RegIt = find(UniqueVec, Inputs[SrcRegIdx]);
        assert(RegIt != UniqueVec.end() && "Cannot find non-const value.");
        Idx = (Idx % NewElts) +
              NewElts * (std::distance(UniqueVec.begin(), RegIt) + ConstNum);
        assert(Idx >= 0 && "Expected defined mask idx.");
      }
      copy(UniqueConstantVec, std::begin(Inputs));
      copy(UniqueVec, std::next(std::begin(Inputs), ConstNum));
    }
  };
  MakeUniqueInputs(OrigMask);
  SDValue OrigInputs[4];
  copy(Inputs, std::begin(OrigInputs));
  for (unsigned High = 0; High < 2; ++High) {
    SDValue &Output = High ? Hi : Lo;
 
    // Build a shuffle mask for the output, discovering on the fly which
    // input vectors to use as shuffle operands.
    unsigned FirstMaskIdx = High * NewElts;
    SmallVector<int> Mask(NewElts * std::size(Inputs), PoisonMaskElem);
    copy(ArrayRef(OrigMask).slice(FirstMaskIdx, NewElts), Mask.begin());
    assert(!Output && "Expected default initialized initial value.");
    TryPeekThroughShufflesInputs(Mask);
    MakeUniqueInputs(Mask);
    SDValue TmpInputs[4];
    copy(Inputs, std::begin(TmpInputs));
    // Track changes in the output registers.
    int UsedIdx = -1;
    bool SecondIteration = false;
    auto &&AccumulateResults = [&UsedIdx, &SecondIteration](unsigned Idx) {
      if (UsedIdx < 0) {
        UsedIdx = Idx;
        return false;
      }
      if (UsedIdx >= 0 && static_cast<unsigned>(UsedIdx) == Idx)
        SecondIteration = true;
      return SecondIteration;
    };
    processShuffleMasks(
        Mask, std::size(Inputs), std::size(Inputs),
        /*NumOfUsedRegs=*/1,
        [&Output, &DAG = DAG, NewVT]() { Output = DAG.getUNDEF(NewVT); },
        [&Output, &DAG = DAG, NewVT, &DL, &Inputs,
         &BuildVector](ArrayRef<int> Mask, unsigned Idx, unsigned /*Unused*/) {
          if (Inputs[Idx]->getOpcode() == ISD::BUILD_VECTOR)
            Output = BuildVector(Inputs[Idx], Inputs[Idx], Mask);
          else
            Output = DAG.getVectorShuffle(NewVT, DL, Inputs[Idx],
                                          DAG.getUNDEF(NewVT), Mask);
          Inputs[Idx] = Output;
        },
        [&AccumulateResults, &Output, &DAG = DAG, NewVT, &DL, &Inputs,
         &TmpInputs,
         &BuildVector](ArrayRef<int> Mask, unsigned Idx1, unsigned Idx2) {
          if (AccumulateResults(Idx1)) {
            if (Inputs[Idx1]->getOpcode() == ISD::BUILD_VECTOR &&
                Inputs[Idx2]->getOpcode() == ISD::BUILD_VECTOR)
              Output = BuildVector(Inputs[Idx1], Inputs[Idx2], Mask);
            else
              Output = DAG.getVectorShuffle(NewVT, DL, Inputs[Idx1],
                                            Inputs[Idx2], Mask);
          } else {
            if (TmpInputs[Idx1]->getOpcode() == ISD::BUILD_VECTOR &&
                TmpInputs[Idx2]->getOpcode() == ISD::BUILD_VECTOR)
              Output = BuildVector(TmpInputs[Idx1], TmpInputs[Idx2], Mask);
            else
              Output = DAG.getVectorShuffle(NewVT, DL, TmpInputs[Idx1],
                                            TmpInputs[Idx2], Mask);
          }
          Inputs[Idx1] = Output;
        });
    copy(OrigInputs, std::begin(Inputs));
  }
}
 
void DAGTypeLegalizer::SplitVecRes_VAARG(SDNode *N, SDValue &Lo, SDValue &Hi) {
  EVT OVT = N->getValueType(0);
  EVT NVT = OVT.getHalfNumVectorElementsVT(*DAG.getContext());
  SDValue Chain = N->getOperand(0);
  SDValue Ptr = N->getOperand(1);
  SDValue SV = N->getOperand(2);
  SDLoc dl(N);
 
  const Align Alignment =
      DAG.getDataLayout().getABITypeAlign(NVT.getTypeForEVT(*DAG.getContext()));
 
  Lo = DAG.getVAArg(NVT, dl, Chain, Ptr, SV, Alignment.value());
  Hi = DAG.getVAArg(NVT, dl, Lo.getValue(1), Ptr, SV, Alignment.value());
  Chain = Hi.getValue(1);
 
  // Modified the chain - switch anything that used the old chain to use
  // the new one.
  ReplaceValueWith(SDValue(N, 1), Chain);
}
 
void DAGTypeLegalizer::SplitVecRes_FP_TO_XINT_SAT(SDNode *N, SDValue &Lo,
                                                  SDValue &Hi) {
  EVT DstVTLo, DstVTHi;
  std::tie(DstVTLo, DstVTHi) = DAG.GetSplitDestVTs(N->getValueType(0));
  SDLoc dl(N);
 
  SDValue SrcLo, SrcHi;
  EVT SrcVT = N->getOperand(0).getValueType();
  if (getTypeAction(SrcVT) == TargetLowering::TypeSplitVector)
    GetSplitVector(N->getOperand(0), SrcLo, SrcHi);
  else
    std::tie(SrcLo, SrcHi) = DAG.SplitVectorOperand(N, 0);
 
  Lo = DAG.getNode(N->getOpcode(), dl, DstVTLo, SrcLo, N->getOperand(1));
  Hi = DAG.getNode(N->getOpcode(), dl, DstVTHi, SrcHi, N->getOperand(1));
}
 
void DAGTypeLegalizer::SplitVecRes_VECTOR_REVERSE(SDNode *N, SDValue &Lo,
                                                  SDValue &Hi) {
  SDValue InLo, InHi;
  GetSplitVector(N->getOperand(0), InLo, InHi);
  SDLoc DL(N);
 
  Lo = DAG.getNode(ISD::VECTOR_REVERSE, DL, InHi.getValueType(), InHi);
  Hi = DAG.getNode(ISD::VECTOR_REVERSE, DL, InLo.getValueType(), InLo);
}
 
void DAGTypeLegalizer::SplitVecRes_VECTOR_SPLICE(SDNode *N, SDValue &Lo,
                                                 SDValue &Hi) {
  SDLoc DL(N);
 
  SDValue Expanded = TLI.expandVectorSplice(N, DAG);
  std::tie(Lo, Hi) = DAG.SplitVector(Expanded, DL);
}
 
void DAGTypeLegalizer::SplitVecRes_VP_REVERSE(SDNode *N, SDValue &Lo,
                                              SDValue &Hi) {
  EVT VT = N->getValueType(0);
  SDValue Val = N->getOperand(0);
  SDValue Mask = N->getOperand(1);
  SDValue EVL = N->getOperand(2);
  SDLoc DL(N);
 
  // Fallback to VP_STRIDED_STORE to stack followed by VP_LOAD.
  Align Alignment = DAG.getReducedAlign(VT, /*UseABI=*/false);
 
  EVT MemVT = EVT::getVectorVT(*DAG.getContext(), VT.getVectorElementType(),
                               VT.getVectorElementCount());
  SDValue StackPtr = DAG.CreateStackTemporary(MemVT.getStoreSize(), Alignment);
  EVT PtrVT = StackPtr.getValueType();
  auto &MF = DAG.getMachineFunction();
  auto FrameIndex = cast<FrameIndexSDNode>(StackPtr.getNode())->getIndex();
  auto PtrInfo = MachinePointerInfo::getFixedStack(MF, FrameIndex);
 
  MachineMemOperand *StoreMMO = DAG.getMachineFunction().getMachineMemOperand(
      PtrInfo, MachineMemOperand::MOStore, LocationSize::beforeOrAfterPointer(),
      Alignment);
  MachineMemOperand *LoadMMO = DAG.getMachineFunction().getMachineMemOperand(
      PtrInfo, MachineMemOperand::MOLoad, LocationSize::beforeOrAfterPointer(),
      Alignment);
 
  unsigned EltWidth = VT.getScalarSizeInBits() / 8;
  SDValue NumElemMinus1 =
      DAG.getNode(ISD::SUB, DL, PtrVT, DAG.getZExtOrTrunc(EVL, DL, PtrVT),
                  DAG.getConstant(1, DL, PtrVT));
  SDValue StartOffset = DAG.getNode(ISD::MUL, DL, PtrVT, NumElemMinus1,
                                    DAG.getConstant(EltWidth, DL, PtrVT));
  SDValue StorePtr = DAG.getNode(ISD::ADD, DL, PtrVT, StackPtr, StartOffset);
  SDValue Stride = DAG.getConstant(-(int64_t)EltWidth, DL, PtrVT);
 
  SDValue TrueMask = DAG.getBoolConstant(true, DL, Mask.getValueType(), VT);
  SDValue Store = DAG.getStridedStoreVP(DAG.getEntryNode(), DL, Val, StorePtr,
                                        DAG.getUNDEF(PtrVT), Stride, TrueMask,
                                        EVL, MemVT, StoreMMO, ISD::UNINDEXED);
 
  SDValue Load = DAG.getLoadVP(VT, DL, Store, StackPtr, Mask, EVL, LoadMMO);
 
  std::tie(Lo, Hi) = DAG.SplitVector(Load, DL);
}
 
void DAGTypeLegalizer::SplitVecRes_VECTOR_DEINTERLEAVE(SDNode *N) {
 
  SDValue Op0Lo, Op0Hi, Op1Lo, Op1Hi;
  GetSplitVector(N->getOperand(0), Op0Lo, Op0Hi);
  GetSplitVector(N->getOperand(1), Op1Lo, Op1Hi);
  EVT VT = Op0Lo.getValueType();
  SDLoc DL(N);
  SDValue ResLo = DAG.getNode(ISD::VECTOR_DEINTERLEAVE, DL,
                              DAG.getVTList(VT, VT), Op0Lo, Op0Hi);
  SDValue ResHi = DAG.getNode(ISD::VECTOR_DEINTERLEAVE, DL,
                              DAG.getVTList(VT, VT), Op1Lo, Op1Hi);
 
  SetSplitVector(SDValue(N, 0), ResLo.getValue(0), ResHi.getValue(0));
  SetSplitVector(SDValue(N, 1), ResLo.getValue(1), ResHi.getValue(1));
}
 
void DAGTypeLegalizer::SplitVecRes_VECTOR_INTERLEAVE(SDNode *N) {
  SDValue Op0Lo, Op0Hi, Op1Lo, Op1Hi;
  GetSplitVector(N->getOperand(0), Op0Lo, Op0Hi);
  GetSplitVector(N->getOperand(1), Op1Lo, Op1Hi);
  EVT VT = Op0Lo.getValueType();
  SDLoc DL(N);
  SDValue Res[] = {DAG.getNode(ISD::VECTOR_INTERLEAVE, DL,
                               DAG.getVTList(VT, VT), Op0Lo, Op1Lo),
                   DAG.getNode(ISD::VECTOR_INTERLEAVE, DL,
                               DAG.getVTList(VT, VT), Op0Hi, Op1Hi)};
 
  SetSplitVector(SDValue(N, 0), Res[0].getValue(0), Res[0].getValue(1));
  SetSplitVector(SDValue(N, 1), Res[1].getValue(0), Res[1].getValue(1));
}
 
//===----------------------------------------------------------------------===//
//  Operand Vector Splitting
//===----------------------------------------------------------------------===//
 
/// This method is called when the specified operand of the specified node is
/// found to need vector splitting. At this point, all of the result types of
/// the node are known to be legal, but other operands of the node may need
/// legalization as well as the specified one.
bool DAGTypeLegalizer::SplitVectorOperand(SDNode *N, unsigned OpNo) {
  LLVM_DEBUG(dbgs() << "Split node operand: "; N->dump(&DAG));
  SDValue Res = SDValue();
 
  // See if the target wants to custom split this node.
  if (CustomLowerNode(N, N->getOperand(OpNo).getValueType(), false))
    return false;
 
  switch (N->getOpcode()) {
  default:
#ifndef NDEBUG
    dbgs() << "SplitVectorOperand Op #" << OpNo << ": ";
    N->dump(&DAG);
    dbgs() << "\n";
#endif
    report_fatal_error("Do not know how to split this operator's "
                       "operand!\n");
 
  case ISD::VP_SETCC:
  case ISD::STRICT_FSETCC:
  case ISD::SETCC:             Res = SplitVecOp_VSETCC(N); break;
  case ISD::BITCAST:           Res = SplitVecOp_BITCAST(N); break;
  case ISD::EXTRACT_SUBVECTOR: Res = SplitVecOp_EXTRACT_SUBVECTOR(N); break;
  case ISD::INSERT_SUBVECTOR:  Res = SplitVecOp_INSERT_SUBVECTOR(N, OpNo); break;
  case ISD::EXTRACT_VECTOR_ELT:Res = SplitVecOp_EXTRACT_VECTOR_ELT(N); break;
  case ISD::CONCAT_VECTORS:    Res = SplitVecOp_CONCAT_VECTORS(N); break;
  case ISD::VP_TRUNCATE:
  case ISD::TRUNCATE:
    Res = SplitVecOp_TruncateHelper(N);
    break;
  case ISD::STRICT_FP_ROUND:
  case ISD::VP_FP_ROUND:
  case ISD::FP_ROUND:          Res = SplitVecOp_FP_ROUND(N); break;
  case ISD::FCOPYSIGN:         Res = SplitVecOp_FPOpDifferentTypes(N); break;
  case ISD::STORE:
    Res = SplitVecOp_STORE(cast<StoreSDNode>(N), OpNo);
    break;
  case ISD::VP_STORE:
    Res = SplitVecOp_VP_STORE(cast<VPStoreSDNode>(N), OpNo);
    break;
  case ISD::EXPERIMENTAL_VP_STRIDED_STORE:
    Res = SplitVecOp_VP_STRIDED_STORE(cast<VPStridedStoreSDNode>(N), OpNo);
    break;
  case ISD::MSTORE:
    Res = SplitVecOp_MSTORE(cast<MaskedStoreSDNode>(N), OpNo);
    break;
  case ISD::MSCATTER:
  case ISD::VP_SCATTER:
    Res = SplitVecOp_Scatter(cast<MemSDNode>(N), OpNo);
    break;
  case ISD::MGATHER:
  case ISD::VP_GATHER:
    Res = SplitVecOp_Gather(cast<MemSDNode>(N), OpNo);
    break;
  case ISD::VSELECT:
    Res = SplitVecOp_VSELECT(N, OpNo);
    break;
  case ISD::STRICT_SINT_TO_FP:
  case ISD::STRICT_UINT_TO_FP:
  case ISD::SINT_TO_FP:
  case ISD::UINT_TO_FP:
  case ISD::VP_SINT_TO_FP:
  case ISD::VP_UINT_TO_FP:
    if (N->getValueType(0).bitsLT(
            N->getOperand(N->isStrictFPOpcode() ? 1 : 0).getValueType()))
      Res = SplitVecOp_TruncateHelper(N);
    else
      Res = SplitVecOp_UnaryOp(N);
    break;
  case ISD::FP_TO_SINT_SAT:
  case ISD::FP_TO_UINT_SAT:
    Res = SplitVecOp_FP_TO_XINT_SAT(N);
    break;
  case ISD::FP_TO_SINT:
  case ISD::FP_TO_UINT:
  case ISD::VP_FP_TO_SINT:
  case ISD::VP_FP_TO_UINT:
  case ISD::STRICT_FP_TO_SINT:
  case ISD::STRICT_FP_TO_UINT:
  case ISD::STRICT_FP_EXTEND:
  case ISD::FP_EXTEND:
  case ISD::SIGN_EXTEND:
  case ISD::ZERO_EXTEND:
  case ISD::ANY_EXTEND:
  case ISD::FTRUNC:
  case ISD::LROUND:
  case ISD::LLROUND:
  case ISD::LRINT:
  case ISD::LLRINT:
    Res = SplitVecOp_UnaryOp(N);
    break;
  case ISD::FLDEXP:
    Res = SplitVecOp_FPOpDifferentTypes(N);
    break;
 
  case ISD::SCMP:
  case ISD::UCMP:
    Res = SplitVecOp_CMP(N);
    break;
 
  case ISD::ANY_EXTEND_VECTOR_INREG:
  case ISD::SIGN_EXTEND_VECTOR_INREG:
  case ISD::ZERO_EXTEND_VECTOR_INREG:
    Res = SplitVecOp_ExtVecInRegOp(N);
    break;
 
  case ISD::VECREDUCE_FADD:
  case ISD::VECREDUCE_FMUL:
  case ISD::VECREDUCE_ADD:
  case ISD::VECREDUCE_MUL:
  case ISD::VECREDUCE_AND:
  case ISD::VECREDUCE_OR:
  case ISD::VECREDUCE_XOR:
  case ISD::VECREDUCE_SMAX:
  case ISD::VECREDUCE_SMIN:
  case ISD::VECREDUCE_UMAX:
  case ISD::VECREDUCE_UMIN:
  case ISD::VECREDUCE_FMAX:
  case ISD::VECREDUCE_FMIN:
  case ISD::VECREDUCE_FMAXIMUM:
  case ISD::VECREDUCE_FMINIMUM:
    Res = SplitVecOp_VECREDUCE(N, OpNo);
    break;
  case ISD::VECREDUCE_SEQ_FADD:
  case ISD::VECREDUCE_SEQ_FMUL:
    Res = SplitVecOp_VECREDUCE_SEQ(N);
    break;
  case ISD::VP_REDUCE_FADD:
  case ISD::VP_REDUCE_SEQ_FADD:
  case ISD::VP_REDUCE_FMUL:
  case ISD::VP_REDUCE_SEQ_FMUL:
  case ISD::VP_REDUCE_ADD:
  case ISD::VP_REDUCE_MUL:
  case ISD::VP_REDUCE_AND:
  case ISD::VP_REDUCE_OR:
  case ISD::VP_REDUCE_XOR:
  case ISD::VP_REDUCE_SMAX:
  case ISD::VP_REDUCE_SMIN:
  case ISD::VP_REDUCE_UMAX:
  case ISD::VP_REDUCE_UMIN:
  case ISD::VP_REDUCE_FMAX:
  case ISD::VP_REDUCE_FMIN:
  case ISD::VP_REDUCE_FMAXIMUM:
  case ISD::VP_REDUCE_FMINIMUM:
    Res = SplitVecOp_VP_REDUCE(N, OpNo);
    break;
  case ISD::VP_CTTZ_ELTS:
  case ISD::VP_CTTZ_ELTS_ZERO_UNDEF:
    Res = SplitVecOp_VP_CttzElements(N);
    break;
  }
 
  // If the result is null, the sub-method took care of registering results etc.
  if (!Res.getNode()) return false;
 
  // If the result is N, the sub-method updated N in place.  Tell the legalizer
  // core about this.
  if (Res.getNode() == N)
    return true;
 
  if (N->isStrictFPOpcode())
    assert(Res.getValueType() == N->getValueType(0) && N->getNumValues() == 2 &&
           "Invalid operand expansion");
  else
    assert(Res.getValueType() == N->getValueType(0) && N->getNumValues() == 1 &&
         "Invalid operand expansion");
 
  ReplaceValueWith(SDValue(N, 0), Res);
  return false;
}
 
SDValue DAGTypeLegalizer::SplitVecOp_VSELECT(SDNode *N, unsigned OpNo) {
  // The only possibility for an illegal operand is the mask, since result type
  // legalization would have handled this node already otherwise.
  assert(OpNo == 0 && "Illegal operand must be mask");
 
  SDValue Mask = N->getOperand(0);
  SDValue Src0 = N->getOperand(1);
  SDValue Src1 = N->getOperand(2);
  EVT Src0VT = Src0.getValueType();
  SDLoc DL(N);
  assert(Mask.getValueType().isVector() && "VSELECT without a vector mask?");
 
  SDValue Lo, Hi;
  GetSplitVector(N->getOperand(0), Lo, Hi);
  assert(Lo.getValueType() == Hi.getValueType() &&
         "Lo and Hi have differing types");
 
  EVT LoOpVT, HiOpVT;
  std::tie(LoOpVT, HiOpVT) = DAG.GetSplitDestVTs(Src0VT);
  assert(LoOpVT == HiOpVT && "Asymmetric vector split?");
 
  SDValue LoOp0, HiOp0, LoOp1, HiOp1, LoMask, HiMask;
  std::tie(LoOp0, HiOp0) = DAG.SplitVector(Src0, DL);
  std::tie(LoOp1, HiOp1) = DAG.SplitVector(Src1, DL);
  std::tie(LoMask, HiMask) = DAG.SplitVector(Mask, DL);
 
  SDValue LoSelect =
    DAG.getNode(ISD::VSELECT, DL, LoOpVT, LoMask, LoOp0, LoOp1);
  SDValue HiSelect =
    DAG.getNode(ISD::VSELECT, DL, HiOpVT, HiMask, HiOp0, HiOp1);
 
  return DAG.getNode(ISD::CONCAT_VECTORS, DL, Src0VT, LoSelect, HiSelect);
}
 
SDValue DAGTypeLegalizer::SplitVecOp_VECREDUCE(SDNode *N, unsigned OpNo) {
  EVT ResVT = N->getValueType(0);
  SDValue Lo, Hi;
  SDLoc dl(N);
 
  SDValue VecOp = N->getOperand(OpNo);
  EVT VecVT = VecOp.getValueType();
  assert(VecVT.isVector() && "Can only split reduce vector operand");
  GetSplitVector(VecOp, Lo, Hi);
  EVT LoOpVT, HiOpVT;
  std::tie(LoOpVT, HiOpVT) = DAG.GetSplitDestVTs(VecVT);
 
  // Use the appropriate scalar instruction on the split subvectors before
  // reducing the now partially reduced smaller vector.
  unsigned CombineOpc = ISD::getVecReduceBaseOpcode(N->getOpcode());
  SDValue Partial = DAG.getNode(CombineOpc, dl, LoOpVT, Lo, Hi, N->getFlags());
  return DAG.getNode(N->getOpcode(), dl, ResVT, Partial, N->getFlags());
}
 
SDValue DAGTypeLegalizer::SplitVecOp_VECREDUCE_SEQ(SDNode *N) {
  EVT ResVT = N->getValueType(0);
  SDValue Lo, Hi;
  SDLoc dl(N);
 
  SDValue AccOp = N->getOperand(0);
  SDValue VecOp = N->getOperand(1);
  SDNodeFlags Flags = N->getFlags();
 
  EVT VecVT = VecOp.getValueType();
  assert(VecVT.isVector() && "Can only split reduce vector operand");
  GetSplitVector(VecOp, Lo, Hi);
  EVT LoOpVT, HiOpVT;
  std::tie(LoOpVT, HiOpVT) = DAG.GetSplitDestVTs(VecVT);
 
  // Reduce low half.
  SDValue Partial = DAG.getNode(N->getOpcode(), dl, ResVT, AccOp, Lo, Flags);
 
  // Reduce high half, using low half result as initial value.
  return DAG.getNode(N->getOpcode(), dl, ResVT, Partial, Hi, Flags);
}
 
SDValue DAGTypeLegalizer::SplitVecOp_VP_REDUCE(SDNode *N, unsigned OpNo) {
  assert(N->isVPOpcode() && "Expected VP opcode");
  assert(OpNo == 1 && "Can only split reduce vector operand");
 
  unsigned Opc = N->getOpcode();
  EVT ResVT = N->getValueType(0);
  SDValue Lo, Hi;
  SDLoc dl(N);
 
  SDValue VecOp = N->getOperand(OpNo);
  EVT VecVT = VecOp.getValueType();
  assert(VecVT.isVector() && "Can only split reduce vector operand");
  GetSplitVector(VecOp, Lo, Hi);
 
  SDValue MaskLo, MaskHi;
  std::tie(MaskLo, MaskHi) = SplitMask(N->getOperand(2));
 
  SDValue EVLLo, EVLHi;
  std::tie(EVLLo, EVLHi) = DAG.SplitEVL(N->getOperand(3), VecVT, dl);
 
  const SDNodeFlags Flags = N->getFlags();
 
  SDValue ResLo =
      DAG.getNode(Opc, dl, ResVT, {N->getOperand(0), Lo, MaskLo, EVLLo}, Flags);
  return DAG.getNode(Opc, dl, ResVT, {ResLo, Hi, MaskHi, EVLHi}, Flags);
}
 
SDValue DAGTypeLegalizer::SplitVecOp_UnaryOp(SDNode *N) {
  // The result has a legal vector type, but the input needs splitting.
  EVT ResVT = N->getValueType(0);
  SDValue Lo, Hi;
  SDLoc dl(N);
  GetSplitVector(N->getOperand(N->isStrictFPOpcode() ? 1 : 0), Lo, Hi);
  EVT InVT = Lo.getValueType();
 
  EVT OutVT = EVT::getVectorVT(*DAG.getContext(), ResVT.getVectorElementType(),
                               InVT.getVectorElementCount());
 
  if (N->isStrictFPOpcode()) {
    Lo = DAG.getNode(N->getOpcode(), dl, { OutVT, MVT::Other }, 
                     { N->getOperand(0), Lo });
    Hi = DAG.getNode(N->getOpcode(), dl, { OutVT, MVT::Other }, 
                     { N->getOperand(0), Hi });
 
    // Build a factor node to remember that this operation is independent
    // of the other one.
    SDValue Ch = DAG.getNode(ISD::TokenFactor, dl, MVT::Other, Lo.getValue(1),
                             Hi.getValue(1));
  
    // Legalize the chain result - switch anything that used the old chain to
    // use the new one.
    ReplaceValueWith(SDValue(N, 1), Ch);
  } else if (N->getNumOperands() == 3) {
    assert(N->isVPOpcode() && "Expected VP opcode");
    SDValue MaskLo, MaskHi, EVLLo, EVLHi;
    std::tie(MaskLo, MaskHi) = SplitMask(N->getOperand(1));
    std::tie(EVLLo, EVLHi) =
        DAG.SplitEVL(N->getOperand(2), N->getValueType(0), dl);
    Lo = DAG.getNode(N->getOpcode(), dl, OutVT, Lo, MaskLo, EVLLo);
    Hi = DAG.getNode(N->getOpcode(), dl, OutVT, Hi, MaskHi, EVLHi);
  } else {
    Lo = DAG.getNode(N->getOpcode(), dl, OutVT, Lo);
    Hi = DAG.getNode(N->getOpcode(), dl, OutVT, Hi);
  }
 
  return DAG.getNode(ISD::CONCAT_VECTORS, dl, ResVT, Lo, Hi);
}
 
SDValue DAGTypeLegalizer::SplitVecOp_BITCAST(SDNode *N) {
  // For example, i64 = BITCAST v4i16 on alpha.  Typically the vector will
  // end up being split all the way down to individual components.  Convert the
  // split pieces into integers and reassemble.
  EVT ResVT = N->getValueType(0);
  SDValue Lo, Hi;
  GetSplitVector(N->getOperand(0), Lo, Hi);
  SDLoc dl(N);
 
  if (ResVT.isScalableVector()) {
    auto [LoVT, HiVT] = DAG.GetSplitDestVTs(ResVT);
    Lo = DAG.getNode(ISD::BITCAST, dl, LoVT, Lo);
    Hi = DAG.getNode(ISD::BITCAST, dl, HiVT, Hi);
    return DAG.getNode(ISD::CONCAT_VECTORS, dl, ResVT, Lo, Hi);
  }
 
  Lo = BitConvertToInteger(Lo);
  Hi = BitConvertToInteger(Hi);
 
  if (DAG.getDataLayout().isBigEndian())
    std::swap(Lo, Hi);
 
  return DAG.getNode(ISD::BITCAST, dl, ResVT, JoinIntegers(Lo, Hi));
}
 
SDValue DAGTypeLegalizer::SplitVecOp_INSERT_SUBVECTOR(SDNode *N,
                                                      unsigned OpNo) {
  assert(OpNo == 1 && "Invalid OpNo; can only split SubVec.");
  // We know that the result type is legal.
  EVT ResVT = N->getValueType(0);
 
  SDValue Vec = N->getOperand(0);
  SDValue SubVec = N->getOperand(1);
  SDValue Idx = N->getOperand(2);
  SDLoc dl(N);
 
  SDValue Lo, Hi;
  GetSplitVector(SubVec, Lo, Hi);
 
  uint64_t IdxVal = Idx->getAsZExtVal();
  uint64_t LoElts = Lo.getValueType().getVectorMinNumElements();
 
  SDValue FirstInsertion =
      DAG.getNode(ISD::INSERT_SUBVECTOR, dl, ResVT, Vec, Lo, Idx);
  SDValue SecondInsertion =
      DAG.getNode(ISD::INSERT_SUBVECTOR, dl, ResVT, FirstInsertion, Hi,
                  DAG.getVectorIdxConstant(IdxVal + LoElts, dl));
 
  return SecondInsertion;
}
 
SDValue DAGTypeLegalizer::SplitVecOp_EXTRACT_SUBVECTOR(SDNode *N) {
  // We know that the extracted result type is legal.
  EVT SubVT = N->getValueType(0);
  SDValue Idx = N->getOperand(1);
  SDLoc dl(N);
  SDValue Lo, Hi;
 
  GetSplitVector(N->getOperand(0), Lo, Hi);
 
  uint64_t LoEltsMin = Lo.getValueType().getVectorMinNumElements();
  uint64_t IdxVal = Idx->getAsZExtVal();
 
  if (IdxVal < LoEltsMin) {
    assert(IdxVal + SubVT.getVectorMinNumElements() <= LoEltsMin &&
           "Extracted subvector crosses vector split!");
    return DAG.getNode(ISD::EXTRACT_SUBVECTOR, dl, SubVT, Lo, Idx);
  } else if (SubVT.isScalableVector() ==
             N->getOperand(0).getValueType().isScalableVector())
    return DAG.getNode(ISD::EXTRACT_SUBVECTOR, dl, SubVT, Hi,
                       DAG.getVectorIdxConstant(IdxVal - LoEltsMin, dl));
 
  // After this point the DAG node only permits extracting fixed-width
  // subvectors from scalable vectors.
  assert(SubVT.isFixedLengthVector() &&
         "Extracting scalable subvector from fixed-width unsupported");
 
  // If the element type is i1 and we're not promoting the result, then we may
  // end up loading the wrong data since the bits are packed tightly into
  // bytes. For example, if we extract a v4i1 (legal) from a nxv4i1 (legal)
  // type at index 4, then we will load a byte starting at index 0.
  if (SubVT.getScalarType() == MVT::i1)
    report_fatal_error("Don't know how to extract fixed-width predicate "
                       "subvector from a scalable predicate vector");
 
  // Spill the vector to the stack. We should use the alignment for
  // the smallest part.
  SDValue Vec = N->getOperand(0);
  EVT VecVT = Vec.getValueType();
  Align SmallestAlign = DAG.getReducedAlign(VecVT, /*UseABI=*/false);
  SDValue StackPtr =
      DAG.CreateStackTemporary(VecVT.getStoreSize(), SmallestAlign);
  auto &MF = DAG.getMachineFunction();
  auto FrameIndex = cast<FrameIndexSDNode>(StackPtr.getNode())->getIndex();
  auto PtrInfo = MachinePointerInfo::getFixedStack(MF, FrameIndex);
 
  SDValue Store = DAG.getStore(DAG.getEntryNode(), dl, Vec, StackPtr, PtrInfo,
                               SmallestAlign);
 
  // Extract the subvector by loading the correct part.
  StackPtr = TLI.getVectorSubVecPointer(DAG, StackPtr, VecVT, SubVT, Idx);
 
  return DAG.getLoad(
      SubVT, dl, Store, StackPtr,
      MachinePointerInfo::getUnknownStack(DAG.getMachineFunction()));
}
 
SDValue DAGTypeLegalizer::SplitVecOp_EXTRACT_VECTOR_ELT(SDNode *N) {
  SDValue Vec = N->getOperand(0);
  SDValue Idx = N->getOperand(1);
  EVT VecVT = Vec.getValueType();
 
  if (const ConstantSDNode *Index = dyn_cast<ConstantSDNode>(Idx)) {
    uint64_t IdxVal = Index->getZExtValue();
 
    SDValue Lo, Hi;
    GetSplitVector(Vec, Lo, Hi);
 
    uint64_t LoElts = Lo.getValueType().getVectorMinNumElements();
 
    if (IdxVal < LoElts)
      return SDValue(DAG.UpdateNodeOperands(N, Lo, Idx), 0);
    else if (!Vec.getValueType().isScalableVector())
      return SDValue(DAG.UpdateNodeOperands(N, Hi,
                                    DAG.getConstant(IdxVal - LoElts, SDLoc(N),
                                                    Idx.getValueType())), 0);
  }
 
  // See if the target wants to custom expand this node.
  if (CustomLowerNode(N, N->getValueType(0), true))
    return SDValue();
 
  // Make the vector elements byte-addressable if they aren't already.
  SDLoc dl(N);
  EVT EltVT = VecVT.getVectorElementType();
  if (!EltVT.isByteSized()) {
    EltVT = EltVT.changeTypeToInteger().getRoundIntegerType(*DAG.getContext());
    VecVT = VecVT.changeElementType(EltVT);
    Vec = DAG.getNode(ISD::ANY_EXTEND, dl, VecVT, Vec);
    SDValue NewExtract =
        DAG.getNode(ISD::EXTRACT_VECTOR_ELT, dl, EltVT, Vec, Idx);
    return DAG.getAnyExtOrTrunc(NewExtract, dl, N->getValueType(0));
  }
 
  // Store the vector to the stack.
  // In cases where the vector is illegal it will be broken down into parts
  // and stored in parts - we should use the alignment for the smallest part.
  Align SmallestAlign = DAG.getReducedAlign(VecVT, /*UseABI=*/false);
  SDValue StackPtr =
      DAG.CreateStackTemporary(VecVT.getStoreSize(), SmallestAlign);
  auto &MF = DAG.getMachineFunction();
  auto FrameIndex = cast<FrameIndexSDNode>(StackPtr.getNode())->getIndex();
  auto PtrInfo = MachinePointerInfo::getFixedStack(MF, FrameIndex);
  SDValue Store = DAG.getStore(DAG.getEntryNode(), dl, Vec, StackPtr, PtrInfo,
                               SmallestAlign);
 
  // Load back the required element.
  StackPtr = TLI.getVectorElementPointer(DAG, StackPtr, VecVT, Idx);
 
  // EXTRACT_VECTOR_ELT can extend the element type to the width of the return
  // type, leaving the high bits undefined. But it can't truncate.
  assert(N->getValueType(0).bitsGE(EltVT) && "Illegal EXTRACT_VECTOR_ELT.");
 
  return DAG.getExtLoad(
      ISD::EXTLOAD, dl, N->getValueType(0), Store, StackPtr,
      MachinePointerInfo::getUnknownStack(DAG.getMachineFunction()), EltVT,
      commonAlignment(SmallestAlign, EltVT.getFixedSizeInBits() / 8));
}
 
SDValue DAGTypeLegalizer::SplitVecOp_ExtVecInRegOp(SDNode *N) {
  SDValue Lo, Hi;
 
  // *_EXTEND_VECTOR_INREG only reference the lower half of the input, so
  // splitting the result has the same effect as splitting the input operand.
  SplitVecRes_ExtVecInRegOp(N, Lo, Hi);
 
  return DAG.getNode(ISD::CONCAT_VECTORS, SDLoc(N), N->getValueType(0), Lo, Hi);
}
 
SDValue DAGTypeLegalizer::SplitVecOp_Gather(MemSDNode *N, unsigned OpNo) {
  (void)OpNo;
  SDValue Lo, Hi;
  SplitVecRes_Gather(N, Lo, Hi);
 
  SDValue Res = DAG.getNode(ISD::CONCAT_VECTORS, N, N->getValueType(0), Lo, Hi);
  ReplaceValueWith(SDValue(N, 0), Res);
  return SDValue();
}
 
SDValue DAGTypeLegalizer::SplitVecOp_VP_STORE(VPStoreSDNode *N, unsigned OpNo) {
  assert(N->isUnindexed() && "Indexed vp_store of vector?");
  SDValue Ch = N->getChain();
  SDValue Ptr = N->getBasePtr();
  SDValue Offset = N->getOffset();
  assert(Offset.isUndef() && "Unexpected VP store offset");
  SDValue Mask = N->getMask();
  SDValue EVL = N->getVectorLength();
  SDValue Data = N->getValue();
  Align Alignment = N->getOriginalAlign();
  SDLoc DL(N);
 
  SDValue DataLo, DataHi;
  if (getTypeAction(Data.getValueType()) == TargetLowering::TypeSplitVector)
    // Split Data operand
    GetSplitVector(Data, DataLo, DataHi);
  else
    std::tie(DataLo, DataHi) = DAG.SplitVector(Data, DL);
 
  // Split Mask operand
  SDValue MaskLo, MaskHi;
  if (OpNo == 1 && Mask.getOpcode() == ISD::SETCC) {
    SplitVecRes_SETCC(Mask.getNode(), MaskLo, MaskHi);
  } else {
    if (getTypeAction(Mask.getValueType()) == TargetLowering::TypeSplitVector)
      GetSplitVector(Mask, MaskLo, MaskHi);
    else
      std::tie(MaskLo, MaskHi) = DAG.SplitVector(Mask, DL);
  }
 
  EVT MemoryVT = N->getMemoryVT();
  EVT LoMemVT, HiMemVT;
  bool HiIsEmpty = false;
  std::tie(LoMemVT, HiMemVT) =
      DAG.GetDependentSplitDestVTs(MemoryVT, DataLo.getValueType(), &HiIsEmpty);
 
  // Split EVL
  SDValue EVLLo, EVLHi;
  std::tie(EVLLo, EVLHi) = DAG.SplitEVL(EVL, Data.getValueType(), DL);
 
  SDValue Lo, Hi;
  MachineMemOperand *MMO = DAG.getMachineFunction().getMachineMemOperand(
      N->getPointerInfo(), MachineMemOperand::MOStore,
      LocationSize::beforeOrAfterPointer(), Alignment, N->getAAInfo(),
      N->getRanges());
 
  Lo = DAG.getStoreVP(Ch, DL, DataLo, Ptr, Offset, MaskLo, EVLLo, LoMemVT, MMO,
                      N->getAddressingMode(), N->isTruncatingStore(),
                      N->isCompressingStore());
 
  // If the hi vp_store has zero storage size, only the lo vp_store is needed.
  if (HiIsEmpty)
    return Lo;
 
  Ptr = TLI.IncrementMemoryAddress(Ptr, MaskLo, DL, LoMemVT, DAG,
                                   N->isCompressingStore());
 
  MachinePointerInfo MPI;
  if (LoMemVT.isScalableVector()) {
    Alignment = commonAlignment(Alignment,
                                LoMemVT.getSizeInBits().getKnownMinValue() / 8);
    MPI = MachinePointerInfo(N->getPointerInfo().getAddrSpace());
  } else
    MPI = N->getPointerInfo().getWithOffset(
        LoMemVT.getStoreSize().getFixedValue());
 
  MMO = DAG.getMachineFunction().getMachineMemOperand(
      MPI, MachineMemOperand::MOStore, LocationSize::beforeOrAfterPointer(),
      Alignment, N->getAAInfo(), N->getRanges());
 
  Hi = DAG.getStoreVP(Ch, DL, DataHi, Ptr, Offset, MaskHi, EVLHi, HiMemVT, MMO,
                      N->getAddressingMode(), N->isTruncatingStore(),
                      N->isCompressingStore());
 
  // Build a factor node to remember that this store is independent of the
  // other one.
  return DAG.getNode(ISD::TokenFactor, DL, MVT::Other, Lo, Hi);
}
 
SDValue DAGTypeLegalizer::SplitVecOp_VP_STRIDED_STORE(VPStridedStoreSDNode *N,
                                                      unsigned OpNo) {
  assert(N->isUnindexed() && "Indexed vp_strided_store of a vector?");
  assert(N->getOffset().isUndef() && "Unexpected VP strided store offset");
 
  SDLoc DL(N);
 
  SDValue Data = N->getValue();
  SDValue LoData, HiData;
  if (getTypeAction(Data.getValueType()) == TargetLowering::TypeSplitVector)
    GetSplitVector(Data, LoData, HiData);
  else
    std::tie(LoData, HiData) = DAG.SplitVector(Data, DL);
 
  EVT LoMemVT, HiMemVT;
  bool HiIsEmpty = false;
  std::tie(LoMemVT, HiMemVT) = DAG.GetDependentSplitDestVTs(
      N->getMemoryVT(), LoData.getValueType(), &HiIsEmpty);
 
  SDValue Mask = N->getMask();
  SDValue LoMask, HiMask;
  if (OpNo == 1 && Mask.getOpcode() == ISD::SETCC)
    SplitVecRes_SETCC(Mask.getNode(), LoMask, HiMask);
  else if (getTypeAction(Mask.getValueType()) ==
           TargetLowering::TypeSplitVector)
    GetSplitVector(Mask, LoMask, HiMask);
  else
    std::tie(LoMask, HiMask) = DAG.SplitVector(Mask, DL);
 
  SDValue LoEVL, HiEVL;
  std::tie(LoEVL, HiEVL) =
      DAG.SplitEVL(N->getVectorLength(), Data.getValueType(), DL);
 
  // Generate the low vp_strided_store
  SDValue Lo = DAG.getStridedStoreVP(
      N->getChain(), DL, LoData, N->getBasePtr(), N->getOffset(),
      N->getStride(), LoMask, LoEVL, LoMemVT, N->getMemOperand(),
      N->getAddressingMode(), N->isTruncatingStore(), N->isCompressingStore());
 
  // If the high vp_strided_store has zero storage size, only the low
  // vp_strided_store is needed.
  if (HiIsEmpty)
    return Lo;
 
  // Generate the high vp_strided_store.
  // To calculate the high base address, we need to sum to the low base
  // address stride number of bytes for each element already stored by low,
  // that is: Ptr = Ptr + (LoEVL * Stride)
  EVT PtrVT = N->getBasePtr().getValueType();
  SDValue Increment =
      DAG.getNode(ISD::MUL, DL, PtrVT, LoEVL,
                  DAG.getSExtOrTrunc(N->getStride(), DL, PtrVT));
  SDValue Ptr = DAG.getNode(ISD::ADD, DL, PtrVT, N->getBasePtr(), Increment);
 
  Align Alignment = N->getOriginalAlign();
  if (LoMemVT.isScalableVector())
    Alignment = commonAlignment(Alignment,
                                LoMemVT.getSizeInBits().getKnownMinValue() / 8);
 
  MachineMemOperand *MMO = DAG.getMachineFunction().getMachineMemOperand(
      MachinePointerInfo(N->getPointerInfo().getAddrSpace()),
      MachineMemOperand::MOStore, LocationSize::beforeOrAfterPointer(),
      Alignment, N->getAAInfo(), N->getRanges());
 
  SDValue Hi = DAG.getStridedStoreVP(
      N->getChain(), DL, HiData, Ptr, N->getOffset(), N->getStride(), HiMask,
      HiEVL, HiMemVT, MMO, N->getAddressingMode(), N->isTruncatingStore(),
      N->isCompressingStore());
 
  // Build a factor node to remember that this store is independent of the
  // other one.
  return DAG.getNode(ISD::TokenFactor, DL, MVT::Other, Lo, Hi);
}
 
SDValue DAGTypeLegalizer::SplitVecOp_MSTORE(MaskedStoreSDNode *N,
                                            unsigned OpNo) {
  assert(N->isUnindexed() && "Indexed masked store of vector?");
  SDValue Ch  = N->getChain();
  SDValue Ptr = N->getBasePtr();
  SDValue Offset = N->getOffset();
  assert(Offset.isUndef() && "Unexpected indexed masked store offset");
  SDValue Mask = N->getMask();
  SDValue Data = N->getValue();
  Align Alignment = N->getOriginalAlign();
  SDLoc DL(N);
 
  SDValue DataLo, DataHi;
  if (getTypeAction(Data.getValueType()) == TargetLowering::TypeSplitVector)
    // Split Data operand
    GetSplitVector(Data, DataLo, DataHi);
  else
    std::tie(DataLo, DataHi) = DAG.SplitVector(Data, DL);
 
  // Split Mask operand
  SDValue MaskLo, MaskHi;
  if (OpNo == 1 && Mask.getOpcode() == ISD::SETCC) {
    SplitVecRes_SETCC(Mask.getNode(), MaskLo, MaskHi);
  } else {
    if (getTypeAction(Mask.getValueType()) == TargetLowering::TypeSplitVector)
      GetSplitVector(Mask, MaskLo, MaskHi);
    else
      std::tie(MaskLo, MaskHi) = DAG.SplitVector(Mask, DL);
  }
 
  EVT MemoryVT = N->getMemoryVT();
  EVT LoMemVT, HiMemVT;
  bool HiIsEmpty = false;
  std::tie(LoMemVT, HiMemVT) =
      DAG.GetDependentSplitDestVTs(MemoryVT, DataLo.getValueType(), &HiIsEmpty);
 
  SDValue Lo, Hi, Res;
  MachineMemOperand *MMO = DAG.getMachineFunction().getMachineMemOperand(
      N->getPointerInfo(), MachineMemOperand::MOStore,
      LocationSize::beforeOrAfterPointer(), Alignment, N->getAAInfo(),
      N->getRanges());
 
  Lo = DAG.getMaskedStore(Ch, DL, DataLo, Ptr, Offset, MaskLo, LoMemVT, MMO,
                          N->getAddressingMode(), N->isTruncatingStore(),
                          N->isCompressingStore());
 
  if (HiIsEmpty) {
    // The hi masked store has zero storage size.
    // Only the lo masked store is needed.
    Res = Lo;
  } else {
 
    Ptr = TLI.IncrementMemoryAddress(Ptr, MaskLo, DL, LoMemVT, DAG,
                                     N->isCompressingStore());
 
    MachinePointerInfo MPI;
    if (LoMemVT.isScalableVector()) {
      Alignment = commonAlignment(
          Alignment, LoMemVT.getSizeInBits().getKnownMinValue() / 8);
      MPI = MachinePointerInfo(N->getPointerInfo().getAddrSpace());
    } else
      MPI = N->getPointerInfo().getWithOffset(
          LoMemVT.getStoreSize().getFixedValue());
 
    MMO = DAG.getMachineFunction().getMachineMemOperand(
        MPI, MachineMemOperand::MOStore, LocationSize::beforeOrAfterPointer(),
        Alignment, N->getAAInfo(), N->getRanges());
 
    Hi = DAG.getMaskedStore(Ch, DL, DataHi, Ptr, Offset, MaskHi, HiMemVT, MMO,
                            N->getAddressingMode(), N->isTruncatingStore(),
                            N->isCompressingStore());
 
    // Build a factor node to remember that this store is independent of the
    // other one.
    Res = DAG.getNode(ISD::TokenFactor, DL, MVT::Other, Lo, Hi);
  }
 
  return Res;
}
 
SDValue DAGTypeLegalizer::SplitVecOp_Scatter(MemSDNode *N, unsigned OpNo) {
  SDValue Ch = N->getChain();
  SDValue Ptr = N->getBasePtr();
  EVT MemoryVT = N->getMemoryVT();
  Align Alignment = N->getOriginalAlign();
  SDLoc DL(N);
  struct Operands {
    SDValue Mask;
    SDValue Index;
    SDValue Scale;
    SDValue Data;
  } Ops = [&]() -> Operands {
    if (auto *MSC = dyn_cast<MaskedScatterSDNode>(N)) {
      return {MSC->getMask(), MSC->getIndex(), MSC->getScale(),
              MSC->getValue()};
    }
    auto *VPSC = cast<VPScatterSDNode>(N);
    return {VPSC->getMask(), VPSC->getIndex(), VPSC->getScale(),
            VPSC->getValue()};
  }();
  // Split all operands
 
  EVT LoMemVT, HiMemVT;
  std::tie(LoMemVT, HiMemVT) = DAG.GetSplitDestVTs(MemoryVT);
 
  SDValue DataLo, DataHi;
  if (getTypeAction(Ops.Data.getValueType()) == TargetLowering::TypeSplitVector)
    // Split Data operand
    GetSplitVector(Ops.Data, DataLo, DataHi);
  else
    std::tie(DataLo, DataHi) = DAG.SplitVector(Ops.Data, DL);
 
  // Split Mask operand
  SDValue MaskLo, MaskHi;
  if (OpNo == 1 && Ops.Mask.getOpcode() == ISD::SETCC) {
    SplitVecRes_SETCC(Ops.Mask.getNode(), MaskLo, MaskHi);
  } else {
    std::tie(MaskLo, MaskHi) = SplitMask(Ops.Mask, DL);
  }
 
  SDValue IndexHi, IndexLo;
  if (getTypeAction(Ops.Index.getValueType()) ==
      TargetLowering::TypeSplitVector)
    GetSplitVector(Ops.Index, IndexLo, IndexHi);
  else
    std::tie(IndexLo, IndexHi) = DAG.SplitVector(Ops.Index, DL);
 
  SDValue Lo;
  MachineMemOperand *MMO = DAG.getMachineFunction().getMachineMemOperand(
      N->getPointerInfo(), MachineMemOperand::MOStore,
      LocationSize::beforeOrAfterPointer(), Alignment, N->getAAInfo(),
      N->getRanges());
 
  if (auto *MSC = dyn_cast<MaskedScatterSDNode>(N)) {
    SDValue OpsLo[] = {Ch, DataLo, MaskLo, Ptr, IndexLo, Ops.Scale};
    Lo =
        DAG.getMaskedScatter(DAG.getVTList(MVT::Other), LoMemVT, DL, OpsLo, MMO,
                             MSC->getIndexType(), MSC->isTruncatingStore());
 
    // The order of the Scatter operation after split is well defined. The "Hi"
    // part comes after the "Lo". So these two operations should be chained one
    // after another.
    SDValue OpsHi[] = {Lo, DataHi, MaskHi, Ptr, IndexHi, Ops.Scale};
    return DAG.getMaskedScatter(DAG.getVTList(MVT::Other), HiMemVT, DL, OpsHi,
                                MMO, MSC->getIndexType(),
                                MSC->isTruncatingStore());
  }
  auto *VPSC = cast<VPScatterSDNode>(N);
  SDValue EVLLo, EVLHi;
  std::tie(EVLLo, EVLHi) =
      DAG.SplitEVL(VPSC->getVectorLength(), Ops.Data.getValueType(), DL);
 
  SDValue OpsLo[] = {Ch, DataLo, Ptr, IndexLo, Ops.Scale, MaskLo, EVLLo};
  Lo = DAG.getScatterVP(DAG.getVTList(MVT::Other), LoMemVT, DL, OpsLo, MMO,
                        VPSC->getIndexType());
 
  // The order of the Scatter operation after split is well defined. The "Hi"
  // part comes after the "Lo". So these two operations should be chained one
  // after another.
  SDValue OpsHi[] = {Lo, DataHi, Ptr, IndexHi, Ops.Scale, MaskHi, EVLHi};
  return DAG.getScatterVP(DAG.getVTList(MVT::Other), HiMemVT, DL, OpsHi, MMO,
                          VPSC->getIndexType());
}
 
SDValue DAGTypeLegalizer::SplitVecOp_STORE(StoreSDNode *N, unsigned OpNo) {
  assert(N->isUnindexed() && "Indexed store of vector?");
  assert(OpNo == 1 && "Can only split the stored value");
  SDLoc DL(N);
 
  bool isTruncating = N->isTruncatingStore();
  SDValue Ch  = N->getChain();
  SDValue Ptr = N->getBasePtr();
  EVT MemoryVT = N->getMemoryVT();
  Align Alignment = N->getOriginalAlign();
  MachineMemOperand::Flags MMOFlags = N->getMemOperand()->getFlags();
  AAMDNodes AAInfo = N->getAAInfo();
  SDValue Lo, Hi;
  GetSplitVector(N->getOperand(1), Lo, Hi);
 
  EVT LoMemVT, HiMemVT;
  std::tie(LoMemVT, HiMemVT) = DAG.GetSplitDestVTs(MemoryVT);
 
  // Scalarize if the split halves are not byte-sized.
  if (!LoMemVT.isByteSized() || !HiMemVT.isByteSized())
    return TLI.scalarizeVectorStore(N, DAG);
 
  if (isTruncating)
    Lo = DAG.getTruncStore(Ch, DL, Lo, Ptr, N->getPointerInfo(), LoMemVT,
                           Alignment, MMOFlags, AAInfo);
  else
    Lo = DAG.getStore(Ch, DL, Lo, Ptr, N->getPointerInfo(), Alignment, MMOFlags,
                      AAInfo);
 
  MachinePointerInfo MPI;
  IncrementPointer(N, LoMemVT, MPI, Ptr);
 
  if (isTruncating)
    Hi = DAG.getTruncStore(Ch, DL, Hi, Ptr, MPI,
                           HiMemVT, Alignment, MMOFlags, AAInfo);
  else
    Hi = DAG.getStore(Ch, DL, Hi, Ptr, MPI, Alignment, MMOFlags, AAInfo);
 
  return DAG.getNode(ISD::TokenFactor, DL, MVT::Other, Lo, Hi);
}
 
SDValue DAGTypeLegalizer::SplitVecOp_CONCAT_VECTORS(SDNode *N) {
  SDLoc DL(N);
 
  // The input operands all must have the same type, and we know the result
  // type is valid.  Convert this to a buildvector which extracts all the
  // input elements.
  // TODO: If the input elements are power-two vectors, we could convert this to
  // a new CONCAT_VECTORS node with elements that are half-wide.
  SmallVector<SDValue, 32> Elts;
  EVT EltVT = N->getValueType(0).getVectorElementType();
  for (const SDValue &Op : N->op_values()) {
    for (unsigned i = 0, e = Op.getValueType().getVectorNumElements();
         i != e; ++i) {
      Elts.push_back(DAG.getNode(ISD::EXTRACT_VECTOR_ELT, DL, EltVT, Op,
                                 DAG.getVectorIdxConstant(i, DL)));
    }
  }
 
  return DAG.getBuildVector(N->getValueType(0), DL, Elts);
}
 
SDValue DAGTypeLegalizer::SplitVecOp_TruncateHelper(SDNode *N) {
  // The result type is legal, but the input type is illegal.  If splitting
  // ends up with the result type of each half still being legal, just
  // do that.  If, however, that would result in an illegal result type,
  // we can try to get more clever with power-two vectors. Specifically,
  // split the input type, but also widen the result element size, then
  // concatenate the halves and truncate again.  For example, consider a target
  // where v8i8 is legal and v8i32 is not (ARM, which doesn't have 256-bit
  // vectors). To perform a "%res = v8i8 trunc v8i32 %in" we do:
  //   %inlo = v4i32 extract_subvector %in, 0
  //   %inhi = v4i32 extract_subvector %in, 4
  //   %lo16 = v4i16 trunc v4i32 %inlo
  //   %hi16 = v4i16 trunc v4i32 %inhi
  //   %in16 = v8i16 concat_vectors v4i16 %lo16, v4i16 %hi16
  //   %res = v8i8 trunc v8i16 %in16
  //
  // Without this transform, the original truncate would end up being
  // scalarized, which is pretty much always a last resort.
  unsigned OpNo = N->isStrictFPOpcode() ? 1 : 0;
  SDValue InVec = N->getOperand(OpNo);
  EVT InVT = InVec->getValueType(0);
  EVT OutVT = N->getValueType(0);
  ElementCount NumElements = OutVT.getVectorElementCount();
  bool IsFloat = OutVT.isFloatingPoint();
 
  unsigned InElementSize = InVT.getScalarSizeInBits();
  unsigned OutElementSize = OutVT.getScalarSizeInBits();