SelectionDAGBuilder.cpp

Go to the documentation of this file.

//===- SelectionDAGBuilder.cpp - Selection-DAG building -------------------===//
//
// 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 implements routines for translating from LLVM IR into SelectionDAG IR.
//
//===----------------------------------------------------------------------===//
 
#include "SelectionDAGBuilder.h"
#include "SDNodeDbgValue.h"
#include "llvm/ADT/APFloat.h"
#include "llvm/ADT/APInt.h"
#include "llvm/ADT/BitVector.h"
#include "llvm/ADT/None.h"
#include "llvm/ADT/Optional.h"
#include "llvm/ADT/STLExtras.h"
#include "llvm/ADT/SmallPtrSet.h"
#include "llvm/ADT/SmallSet.h"
#include "llvm/ADT/StringRef.h"
#include "llvm/ADT/Triple.h"
#include "llvm/ADT/Twine.h"
#include "llvm/Analysis/AliasAnalysis.h"
#include "llvm/Analysis/BlockFrequencyInfo.h"
#include "llvm/Analysis/BranchProbabilityInfo.h"
#include "llvm/Analysis/ConstantFolding.h"
#include "llvm/Analysis/EHPersonalities.h"
#include "llvm/Analysis/Loads.h"
#include "llvm/Analysis/MemoryLocation.h"
#include "llvm/Analysis/ProfileSummaryInfo.h"
#include "llvm/Analysis/TargetLibraryInfo.h"
#include "llvm/Analysis/ValueTracking.h"
#include "llvm/Analysis/VectorUtils.h"
#include "llvm/CodeGen/Analysis.h"
#include "llvm/CodeGen/FunctionLoweringInfo.h"
#include "llvm/CodeGen/GCMetadata.h"
#include "llvm/CodeGen/MachineBasicBlock.h"
#include "llvm/CodeGen/MachineFrameInfo.h"
#include "llvm/CodeGen/MachineFunction.h"
#include "llvm/CodeGen/MachineInstr.h"
#include "llvm/CodeGen/MachineInstrBuilder.h"
#include "llvm/CodeGen/MachineJumpTableInfo.h"
#include "llvm/CodeGen/MachineMemOperand.h"
#include "llvm/CodeGen/MachineModuleInfo.h"
#include "llvm/CodeGen/MachineOperand.h"
#include "llvm/CodeGen/MachineRegisterInfo.h"
#include "llvm/CodeGen/RuntimeLibcalls.h"
#include "llvm/CodeGen/SelectionDAG.h"
#include "llvm/CodeGen/SelectionDAGTargetInfo.h"
#include "llvm/CodeGen/StackMaps.h"
#include "llvm/CodeGen/SwiftErrorValueTracking.h"
#include "llvm/CodeGen/TargetFrameLowering.h"
#include "llvm/CodeGen/TargetInstrInfo.h"
#include "llvm/CodeGen/TargetOpcodes.h"
#include "llvm/CodeGen/TargetRegisterInfo.h"
#include "llvm/CodeGen/TargetSubtargetInfo.h"
#include "llvm/CodeGen/WinEHFuncInfo.h"
#include "llvm/IR/Argument.h"
#include "llvm/IR/Attributes.h"
#include "llvm/IR/BasicBlock.h"
#include "llvm/IR/CFG.h"
#include "llvm/IR/CallingConv.h"
#include "llvm/IR/Constant.h"
#include "llvm/IR/ConstantRange.h"
#include "llvm/IR/Constants.h"
#include "llvm/IR/DataLayout.h"
#include "llvm/IR/DebugInfoMetadata.h"
#include "llvm/IR/DerivedTypes.h"
#include "llvm/IR/DiagnosticInfo.h"
#include "llvm/IR/Function.h"
#include "llvm/IR/GetElementPtrTypeIterator.h"
#include "llvm/IR/InlineAsm.h"
#include "llvm/IR/InstrTypes.h"
#include "llvm/IR/Instructions.h"
#include "llvm/IR/IntrinsicInst.h"
#include "llvm/IR/Intrinsics.h"
#include "llvm/IR/IntrinsicsAArch64.h"
#include "llvm/IR/IntrinsicsWebAssembly.h"
#include "llvm/IR/LLVMContext.h"
#include "llvm/IR/Metadata.h"
#include "llvm/IR/Module.h"
#include "llvm/IR/Operator.h"
#include "llvm/IR/PatternMatch.h"
#include "llvm/IR/Statepoint.h"
#include "llvm/IR/Type.h"
#include "llvm/IR/User.h"
#include "llvm/IR/Value.h"
#include "llvm/MC/MCContext.h"
#include "llvm/MC/MCSymbol.h"
#include "llvm/Support/AtomicOrdering.h"
#include "llvm/Support/Casting.h"
#include "llvm/Support/CommandLine.h"
#include "llvm/Support/Compiler.h"
#include "llvm/Support/Debug.h"
#include "llvm/Support/MathExtras.h"
#include "llvm/Support/raw_ostream.h"
#include "llvm/Target/TargetIntrinsicInfo.h"
#include "llvm/Target/TargetMachine.h"
#include "llvm/Target/TargetOptions.h"
#include "llvm/Transforms/Utils/Local.h"
#include <cstddef>
#include <cstring>
#include <iterator>
#include <limits>
#include <numeric>
#include <tuple>
 
using namespace llvm;
using namespace PatternMatch;
using namespace SwitchCG;
 
#define DEBUG_TYPE "isel"
 
/// LimitFloatPrecision - Generate low-precision inline sequences for
/// some float libcalls (6, 8 or 12 bits).
static unsigned LimitFloatPrecision;
 
static cl::opt<bool>
    InsertAssertAlign("insert-assert-align", cl::init(true),
                      cl::desc("Insert the experimental `assertalign` node."),
                      cl::ReallyHidden);
 
static cl::opt<unsigned, true>
    LimitFPPrecision("limit-float-precision",
                     cl::desc("Generate low-precision inline sequences "
                              "for some float libcalls"),
                     cl::location(LimitFloatPrecision), cl::Hidden,
                     cl::init(0));
 
static cl::opt<unsigned> SwitchPeelThreshold(
    "switch-peel-threshold", cl::Hidden, cl::init(66),
    cl::desc("Set the case probability threshold for peeling the case from a "
             "switch statement. A value greater than 100 will void this "
             "optimization"));
 
// Limit the width of DAG chains. This is important in general to prevent
// DAG-based analysis from blowing up. For example, alias analysis and
// load clustering may not complete in reasonable time. It is difficult to
// recognize and avoid this situation within each individual analysis, and
// future analyses are likely to have the same behavior. Limiting DAG width is
// the safe approach and will be especially important with global DAGs.
//
// MaxParallelChains default is arbitrarily high to avoid affecting
// optimization, but could be lowered to improve compile time. Any ld-ld-st-st
// sequence over this should have been converted to llvm.memcpy by the
// frontend. It is easy to induce this behavior with .ll code such as:
// %buffer = alloca [4096 x i8]
// %data = load [4096 x i8]* %argPtr
// store [4096 x i8] %data, [4096 x i8]* %buffer
static const unsigned MaxParallelChains = 64;
 
static SDValue getCopyFromPartsVector(SelectionDAG &DAG, const SDLoc &DL,
                                      const SDValue *Parts, unsigned NumParts,
                                      MVT PartVT, EVT ValueVT, const Value *V,
                                      Optional<CallingConv::ID> CC);
 
/// getCopyFromParts - Create a value that contains the specified legal parts
/// combined into the value they represent.  If the parts combine to a type
/// larger than ValueVT then AssertOp can be used to specify whether the extra
/// bits are known to be zero (ISD::AssertZext) or sign extended from ValueVT
/// (ISD::AssertSext).
static SDValue getCopyFromParts(SelectionDAG &DAG, const SDLoc &DL,
                                const SDValue *Parts, unsigned NumParts,
                                MVT PartVT, EVT ValueVT, const Value *V,
                                Optional<CallingConv::ID> CC = None,
                                Optional<ISD::NodeType> AssertOp = None) {
  // Let the target assemble the parts if it wants to
  const TargetLowering &TLI = DAG.getTargetLoweringInfo();
  if (SDValue Val = TLI.joinRegisterPartsIntoValue(DAG, DL, Parts, NumParts,
                                                   PartVT, ValueVT, CC))
    return Val;
 
  if (ValueVT.isVector())
    return getCopyFromPartsVector(DAG, DL, Parts, NumParts, PartVT, ValueVT, V,
                                  CC);
 
  assert(NumParts > 0 && "No parts to assemble!");
  SDValue Val = Parts[0];
 
  if (NumParts > 1) {
    // Assemble the value from multiple parts.
    if (ValueVT.isInteger()) {
      unsigned PartBits = PartVT.getSizeInBits();
      unsigned ValueBits = ValueVT.getSizeInBits();
 
      // Assemble the power of 2 part.
      unsigned RoundParts =
          (NumParts & (NumParts - 1)) ? 1 << Log2_32(NumParts) : NumParts;
      unsigned RoundBits = PartBits * RoundParts;
      EVT RoundVT = RoundBits == ValueBits ?
        ValueVT : EVT::getIntegerVT(*DAG.getContext(), RoundBits);
      SDValue Lo, Hi;
 
      EVT HalfVT = EVT::getIntegerVT(*DAG.getContext(), RoundBits/2);
 
      if (RoundParts > 2) {
        Lo = getCopyFromParts(DAG, DL, Parts, RoundParts / 2,
                              PartVT, HalfVT, V);
        Hi = getCopyFromParts(DAG, DL, Parts + RoundParts / 2,
                              RoundParts / 2, PartVT, HalfVT, V);
      } else {
        Lo = DAG.getNode(ISD::BITCAST, DL, HalfVT, Parts[0]);
        Hi = DAG.getNode(ISD::BITCAST, DL, HalfVT, Parts[1]);
      }
 
      if (DAG.getDataLayout().isBigEndian())
        std::swap(Lo, Hi);
 
      Val = DAG.getNode(ISD::BUILD_PAIR, DL, RoundVT, Lo, Hi);
 
      if (RoundParts < NumParts) {
        // Assemble the trailing non-power-of-2 part.
        unsigned OddParts = NumParts - RoundParts;
        EVT OddVT = EVT::getIntegerVT(*DAG.getContext(), OddParts * PartBits);
        Hi = getCopyFromParts(DAG, DL, Parts + RoundParts, OddParts, PartVT,
                              OddVT, V, CC);
 
        // Combine the round and odd parts.
        Lo = Val;
        if (DAG.getDataLayout().isBigEndian())
          std::swap(Lo, Hi);
        EVT TotalVT = EVT::getIntegerVT(*DAG.getContext(), NumParts * PartBits);
        Hi = DAG.getNode(ISD::ANY_EXTEND, DL, TotalVT, Hi);
        Hi =
            DAG.getNode(ISD::SHL, DL, TotalVT, Hi,
                        DAG.getConstant(Lo.getValueSizeInBits(), DL,
                                        TLI.getPointerTy(DAG.getDataLayout())));
        Lo = DAG.getNode(ISD::ZERO_EXTEND, DL, TotalVT, Lo);
        Val = DAG.getNode(ISD::OR, DL, TotalVT, Lo, Hi);
      }
    } else if (PartVT.isFloatingPoint()) {
      // FP split into multiple FP parts (for ppcf128)
      assert(ValueVT == EVT(MVT::ppcf128) && PartVT == MVT::f64 &&
             "Unexpected split");
      SDValue Lo, Hi;
      Lo = DAG.getNode(ISD::BITCAST, DL, EVT(MVT::f64), Parts[0]);
      Hi = DAG.getNode(ISD::BITCAST, DL, EVT(MVT::f64), Parts[1]);
      if (TLI.hasBigEndianPartOrdering(ValueVT, DAG.getDataLayout()))
        std::swap(Lo, Hi);
      Val = DAG.getNode(ISD::BUILD_PAIR, DL, ValueVT, Lo, Hi);
    } else {
      // FP split into integer parts (soft fp)
      assert(ValueVT.isFloatingPoint() && PartVT.isInteger() &&
             !PartVT.isVector() && "Unexpected split");
      EVT IntVT = EVT::getIntegerVT(*DAG.getContext(), ValueVT.getSizeInBits());
      Val = getCopyFromParts(DAG, DL, Parts, NumParts, PartVT, IntVT, V, CC);
    }
  }
 
  // There is now one part, held in Val.  Correct it to match ValueVT.
  // PartEVT is the type of the register class that holds the value.
  // ValueVT is the type of the inline asm operation.
  EVT PartEVT = Val.getValueType();
 
  if (PartEVT == ValueVT)
    return Val;
 
  if (PartEVT.isInteger() && ValueVT.isFloatingPoint() &&
      ValueVT.bitsLT(PartEVT)) {
    // For an FP value in an integer part, we need to truncate to the right
    // width first.
    PartEVT = EVT::getIntegerVT(*DAG.getContext(),  ValueVT.getSizeInBits());
    Val = DAG.getNode(ISD::TRUNCATE, DL, PartEVT, Val);
  }
 
  // Handle types that have the same size.
  if (PartEVT.getSizeInBits() == ValueVT.getSizeInBits())
    return DAG.getNode(ISD::BITCAST, DL, ValueVT, Val);
 
  // Handle types with different sizes.
  if (PartEVT.isInteger() && ValueVT.isInteger()) {
    if (ValueVT.bitsLT(PartEVT)) {
      // For a truncate, see if we have any information to
      // indicate whether the truncated bits will always be
      // zero or sign-extension.
      if (AssertOp.hasValue())
        Val = DAG.getNode(*AssertOp, DL, PartEVT, Val,
                          DAG.getValueType(ValueVT));
      return DAG.getNode(ISD::TRUNCATE, DL, ValueVT, Val);
    }
    return DAG.getNode(ISD::ANY_EXTEND, DL, ValueVT, Val);
  }
 
  if (PartEVT.isFloatingPoint() && ValueVT.isFloatingPoint()) {
    // FP_ROUND's are always exact here.
    if (ValueVT.bitsLT(Val.getValueType()))
      return DAG.getNode(
          ISD::FP_ROUND, DL, ValueVT, Val,
          DAG.getTargetConstant(1, DL, TLI.getPointerTy(DAG.getDataLayout())));
 
    return DAG.getNode(ISD::FP_EXTEND, DL, ValueVT, Val);
  }
 
  // Handle MMX to a narrower integer type by bitcasting MMX to integer and
  // then truncating.
  if (PartEVT == MVT::x86mmx && ValueVT.isInteger() &&
      ValueVT.bitsLT(PartEVT)) {
    Val = DAG.getNode(ISD::BITCAST, DL, MVT::i64, Val);
    return DAG.getNode(ISD::TRUNCATE, DL, ValueVT, Val);
  }
 
  report_fatal_error("Unknown mismatch in getCopyFromParts!");
}
 
static void diagnosePossiblyInvalidConstraint(LLVMContext &Ctx, const Value *V,
                                              const Twine &ErrMsg) {
  const Instruction *I = dyn_cast_or_null<Instruction>(V);
  if (!V)
    return Ctx.emitError(ErrMsg);
 
  const char *AsmError = ", possible invalid constraint for vector type";
  if (const CallInst *CI = dyn_cast<CallInst>(I))
    if (CI->isInlineAsm())
      return Ctx.emitError(I, ErrMsg + AsmError);
 
  return Ctx.emitError(I, ErrMsg);
}
 
/// getCopyFromPartsVector - Create a value that contains the specified legal
/// parts combined into the value they represent.  If the parts combine to a
/// type larger than ValueVT then AssertOp can be used to specify whether the
/// extra bits are known to be zero (ISD::AssertZext) or sign extended from
/// ValueVT (ISD::AssertSext).
static SDValue getCopyFromPartsVector(SelectionDAG &DAG, const SDLoc &DL,
                                      const SDValue *Parts, unsigned NumParts,
                                      MVT PartVT, EVT ValueVT, const Value *V,
                                      Optional<CallingConv::ID> CallConv) {
  assert(ValueVT.isVector() && "Not a vector value");
  assert(NumParts > 0 && "No parts to assemble!");
  const bool IsABIRegCopy = CallConv.hasValue();
 
  const TargetLowering &TLI = DAG.getTargetLoweringInfo();
  SDValue Val = Parts[0];
 
  // Handle a multi-element vector.
  if (NumParts > 1) {
    EVT IntermediateVT;
    MVT RegisterVT;
    unsigned NumIntermediates;
    unsigned NumRegs;
 
    if (IsABIRegCopy) {
      NumRegs = TLI.getVectorTypeBreakdownForCallingConv(
          *DAG.getContext(), CallConv.getValue(), ValueVT, IntermediateVT,
          NumIntermediates, RegisterVT);
    } else {
      NumRegs =
          TLI.getVectorTypeBreakdown(*DAG.getContext(), ValueVT, IntermediateVT,
                                     NumIntermediates, RegisterVT);
    }
 
    assert(NumRegs == NumParts && "Part count doesn't match vector breakdown!");
    NumParts = NumRegs; // Silence a compiler warning.
    assert(RegisterVT == PartVT && "Part type doesn't match vector breakdown!");
    assert(RegisterVT.getSizeInBits() ==
           Parts[0].getSimpleValueType().getSizeInBits() &&
           "Part type sizes don't match!");
 
    // Assemble the parts into intermediate operands.
    SmallVector<SDValue, 8> Ops(NumIntermediates);
    if (NumIntermediates == NumParts) {
      // If the register was not expanded, truncate or copy the value,
      // as appropriate.
      for (unsigned i = 0; i != NumParts; ++i)
        Ops[i] = getCopyFromParts(DAG, DL, &Parts[i], 1,
                                  PartVT, IntermediateVT, V, CallConv);
    } else if (NumParts > 0) {
      // If the intermediate type was expanded, build the intermediate
      // operands from the parts.
      assert(NumParts % NumIntermediates == 0 &&
             "Must expand into a divisible number of parts!");
      unsigned Factor = NumParts / NumIntermediates;
      for (unsigned i = 0; i != NumIntermediates; ++i)
        Ops[i] = getCopyFromParts(DAG, DL, &Parts[i * Factor], Factor,
                                  PartVT, IntermediateVT, V, CallConv);
    }
 
    // Build a vector with BUILD_VECTOR or CONCAT_VECTORS from the
    // intermediate operands.
    EVT BuiltVectorTy =
        IntermediateVT.isVector()
            ? EVT::getVectorVT(
                  *DAG.getContext(), IntermediateVT.getScalarType(),
                  IntermediateVT.getVectorElementCount() * NumParts)
            : EVT::getVectorVT(*DAG.getContext(),
                               IntermediateVT.getScalarType(),
                               NumIntermediates);
    Val = DAG.getNode(IntermediateVT.isVector() ? ISD::CONCAT_VECTORS
                                                : ISD::BUILD_VECTOR,
                      DL, BuiltVectorTy, Ops);
  }
 
  // There is now one part, held in Val.  Correct it to match ValueVT.
  EVT PartEVT = Val.getValueType();
 
  if (PartEVT == ValueVT)
    return Val;
 
  if (PartEVT.isVector()) {
    // Vector/Vector bitcast.
    if (ValueVT.getSizeInBits() == PartEVT.getSizeInBits())
      return DAG.getNode(ISD::BITCAST, DL, ValueVT, Val);
 
    // If the element type of the source/dest vectors are the same, but the
    // parts vector has more elements than the value vector, then we have a
    // vector widening case (e.g. <2 x float> -> <4 x float>).  Extract the
    // elements we want.
    if (PartEVT.getVectorElementCount() != ValueVT.getVectorElementCount()) {
      assert((PartEVT.getVectorElementCount().getKnownMinValue() >
              ValueVT.getVectorElementCount().getKnownMinValue()) &&
             (PartEVT.getVectorElementCount().isScalable() ==
              ValueVT.getVectorElementCount().isScalable()) &&
             "Cannot narrow, it would be a lossy transformation");
      PartEVT =
          EVT::getVectorVT(*DAG.getContext(), PartEVT.getVectorElementType(),
                           ValueVT.getVectorElementCount());
      Val = DAG.getNode(ISD::EXTRACT_SUBVECTOR, DL, PartEVT, Val,
                        DAG.getVectorIdxConstant(0, DL));
      if (PartEVT == ValueVT)
        return Val;
    }
 
    // Promoted vector extract
    return DAG.getAnyExtOrTrunc(Val, DL, ValueVT);
  }
 
  // Trivial bitcast if the types are the same size and the destination
  // vector type is legal.
  if (PartEVT.getSizeInBits() == ValueVT.getSizeInBits() &&
      TLI.isTypeLegal(ValueVT))
    return DAG.getNode(ISD::BITCAST, DL, ValueVT, Val);
 
  if (ValueVT.getVectorNumElements() != 1) {
     // Certain ABIs require that vectors are passed as integers. For vectors
     // are the same size, this is an obvious bitcast.
     if (ValueVT.getSizeInBits() == PartEVT.getSizeInBits()) {
       return DAG.getNode(ISD::BITCAST, DL, ValueVT, Val);
     } else if (ValueVT.bitsLT(PartEVT)) {
       const uint64_t ValueSize = ValueVT.getFixedSizeInBits();
       EVT IntermediateType = EVT::getIntegerVT(*DAG.getContext(), ValueSize);
       // Drop the extra bits.
       Val = DAG.getNode(ISD::TRUNCATE, DL, IntermediateType, Val);
       return DAG.getBitcast(ValueVT, Val);
     }
 
     diagnosePossiblyInvalidConstraint(
         *DAG.getContext(), V, "non-trivial scalar-to-vector conversion");
     return DAG.getUNDEF(ValueVT);
  }
 
  // Handle cases such as i8 -> <1 x i1>
  EVT ValueSVT = ValueVT.getVectorElementType();
  if (ValueVT.getVectorNumElements() == 1 && ValueSVT != PartEVT) {
    if (ValueSVT.getSizeInBits() == PartEVT.getSizeInBits())
      Val = DAG.getNode(ISD::BITCAST, DL, ValueSVT, Val);
    else
      Val = ValueVT.isFloatingPoint()
                ? DAG.getFPExtendOrRound(Val, DL, ValueSVT)
                : DAG.getAnyExtOrTrunc(Val, DL, ValueSVT);
  }
 
  return DAG.getBuildVector(ValueVT, DL, Val);
}
 
static void getCopyToPartsVector(SelectionDAG &DAG, const SDLoc &dl,
                                 SDValue Val, SDValue *Parts, unsigned NumParts,
                                 MVT PartVT, const Value *V,
                                 Optional<CallingConv::ID> CallConv);
 
/// getCopyToParts - Create a series of nodes that contain the specified value
/// split into legal parts.  If the parts contain more bits than Val, then, for
/// integers, ExtendKind can be used to specify how to generate the extra bits.
static void getCopyToParts(SelectionDAG &DAG, const SDLoc &DL, SDValue Val,
                           SDValue *Parts, unsigned NumParts, MVT PartVT,
                           const Value *V,
                           Optional<CallingConv::ID> CallConv = None,
                           ISD::NodeType ExtendKind = ISD::ANY_EXTEND) {
  // Let the target split the parts if it wants to
  const TargetLowering &TLI = DAG.getTargetLoweringInfo();
  if (TLI.splitValueIntoRegisterParts(DAG, DL, Val, Parts, NumParts, PartVT,
                                      CallConv))
    return;
  EVT ValueVT = Val.getValueType();
 
  // Handle the vector case separately.
  if (ValueVT.isVector())
    return getCopyToPartsVector(DAG, DL, Val, Parts, NumParts, PartVT, V,
                                CallConv);
 
  unsigned PartBits = PartVT.getSizeInBits();
  unsigned OrigNumParts = NumParts;
  assert(DAG.getTargetLoweringInfo().isTypeLegal(PartVT) &&
         "Copying to an illegal type!");
 
  if (NumParts == 0)
    return;
 
  assert(!ValueVT.isVector() && "Vector case handled elsewhere");
  EVT PartEVT = PartVT;
  if (PartEVT == ValueVT) {
    assert(NumParts == 1 && "No-op copy with multiple parts!");
    Parts[0] = Val;
    return;
  }
 
  if (NumParts * PartBits > ValueVT.getSizeInBits()) {
    // If the parts cover more bits than the value has, promote the value.
    if (PartVT.isFloatingPoint() && ValueVT.isFloatingPoint()) {
      assert(NumParts == 1 && "Do not know what to promote to!");
      Val = DAG.getNode(ISD::FP_EXTEND, DL, PartVT, Val);
    } else {
      if (ValueVT.isFloatingPoint()) {
        // FP values need to be bitcast, then extended if they are being put
        // into a larger container.
        ValueVT = EVT::getIntegerVT(*DAG.getContext(),  ValueVT.getSizeInBits());
        Val = DAG.getNode(ISD::BITCAST, DL, ValueVT, Val);
      }
      assert((PartVT.isInteger() || PartVT == MVT::x86mmx) &&
             ValueVT.isInteger() &&
             "Unknown mismatch!");
      ValueVT = EVT::getIntegerVT(*DAG.getContext(), NumParts * PartBits);
      Val = DAG.getNode(ExtendKind, DL, ValueVT, Val);
      if (PartVT == MVT::x86mmx)
        Val = DAG.getNode(ISD::BITCAST, DL, PartVT, Val);
    }
  } else if (PartBits == ValueVT.getSizeInBits()) {
    // Different types of the same size.
    assert(NumParts == 1 && PartEVT != ValueVT);
    Val = DAG.getNode(ISD::BITCAST, DL, PartVT, Val);
  } else if (NumParts * PartBits < ValueVT.getSizeInBits()) {
    // If the parts cover less bits than value has, truncate the value.
    assert((PartVT.isInteger() || PartVT == MVT::x86mmx) &&
           ValueVT.isInteger() &&
           "Unknown mismatch!");
    ValueVT = EVT::getIntegerVT(*DAG.getContext(), NumParts * PartBits);
    Val = DAG.getNode(ISD::TRUNCATE, DL, ValueVT, Val);
    if (PartVT == MVT::x86mmx)
      Val = DAG.getNode(ISD::BITCAST, DL, PartVT, Val);
  }
 
  // The value may have changed - recompute ValueVT.
  ValueVT = Val.getValueType();
  assert(NumParts * PartBits == ValueVT.getSizeInBits() &&
         "Failed to tile the value with PartVT!");
 
  if (NumParts == 1) {
    if (PartEVT != ValueVT) {
      diagnosePossiblyInvalidConstraint(*DAG.getContext(), V,
                                        "scalar-to-vector conversion failed");
      Val = DAG.getNode(ISD::BITCAST, DL, PartVT, Val);
    }
 
    Parts[0] = Val;
    return;
  }
 
  // Expand the value into multiple parts.
  if (NumParts & (NumParts - 1)) {
    // The number of parts is not a power of 2.  Split off and copy the tail.
    assert(PartVT.isInteger() && ValueVT.isInteger() &&
           "Do not know what to expand to!");
    unsigned RoundParts = 1 << Log2_32(NumParts);
    unsigned RoundBits = RoundParts * PartBits;
    unsigned OddParts = NumParts - RoundParts;
    SDValue OddVal = DAG.getNode(ISD::SRL, DL, ValueVT, Val,
      DAG.getShiftAmountConstant(RoundBits, ValueVT, DL, /*LegalTypes*/false));
 
    getCopyToParts(DAG, DL, OddVal, Parts + RoundParts, OddParts, PartVT, V,
                   CallConv);
 
    if (DAG.getDataLayout().isBigEndian())
      // The odd parts were reversed by getCopyToParts - unreverse them.
      std::reverse(Parts + RoundParts, Parts + NumParts);
 
    NumParts = RoundParts;
    ValueVT = EVT::getIntegerVT(*DAG.getContext(), NumParts * PartBits);
    Val = DAG.getNode(ISD::TRUNCATE, DL, ValueVT, Val);
  }
 
  // The number of parts is a power of 2.  Repeatedly bisect the value using
  // EXTRACT_ELEMENT.
  Parts[0] = DAG.getNode(ISD::BITCAST, DL,
                         EVT::getIntegerVT(*DAG.getContext(),
                                           ValueVT.getSizeInBits()),
                         Val);
 
  for (unsigned StepSize = NumParts; StepSize > 1; StepSize /= 2) {
    for (unsigned i = 0; i < NumParts; i += StepSize) {
      unsigned ThisBits = StepSize * PartBits / 2;
      EVT ThisVT = EVT::getIntegerVT(*DAG.getContext(), ThisBits);
      SDValue &Part0 = Parts[i];
      SDValue &Part1 = Parts[i+StepSize/2];
 
      Part1 = DAG.getNode(ISD::EXTRACT_ELEMENT, DL,
                          ThisVT, Part0, DAG.getIntPtrConstant(1, DL));
      Part0 = DAG.getNode(ISD::EXTRACT_ELEMENT, DL,
                          ThisVT, Part0, DAG.getIntPtrConstant(0, DL));
 
      if (ThisBits == PartBits && ThisVT != PartVT) {
        Part0 = DAG.getNode(ISD::BITCAST, DL, PartVT, Part0);
        Part1 = DAG.getNode(ISD::BITCAST, DL, PartVT, Part1);
      }
    }
  }
 
  if (DAG.getDataLayout().isBigEndian())
    std::reverse(Parts, Parts + OrigNumParts);
}
 
static SDValue widenVectorToPartType(SelectionDAG &DAG, SDValue Val,
                                     const SDLoc &DL, EVT PartVT) {
  if (!PartVT.isVector())
    return SDValue();
 
  EVT ValueVT = Val.getValueType();
  ElementCount PartNumElts = PartVT.getVectorElementCount();
  ElementCount ValueNumElts = ValueVT.getVectorElementCount();
 
  // We only support widening vectors with equivalent element types and
  // fixed/scalable properties. If a target needs to widen a fixed-length type
  // to a scalable one, it should be possible to use INSERT_SUBVECTOR below.
  if (ElementCount::isKnownLE(PartNumElts, ValueNumElts) ||
      PartNumElts.isScalable() != ValueNumElts.isScalable() ||
      PartVT.getVectorElementType() != ValueVT.getVectorElementType())
    return SDValue();
 
  // Widening a scalable vector to another scalable vector is done by inserting
  // the vector into a larger undef one.
  if (PartNumElts.isScalable())
    return DAG.getNode(ISD::INSERT_SUBVECTOR, DL, PartVT, DAG.getUNDEF(PartVT),
                       Val, DAG.getVectorIdxConstant(0, DL));
 
  EVT ElementVT = PartVT.getVectorElementType();
  // Vector widening case, e.g. <2 x float> -> <4 x float>.  Shuffle in
  // undef elements.
  SmallVector<SDValue, 16> Ops;
  DAG.ExtractVectorElements(Val, Ops);
  SDValue EltUndef = DAG.getUNDEF(ElementVT);
  Ops.append((PartNumElts - ValueNumElts).getFixedValue(), EltUndef);
 
  // FIXME: Use CONCAT for 2x -> 4x.
  return DAG.getBuildVector(PartVT, DL, Ops);
}
 
/// getCopyToPartsVector - Create a series of nodes that contain the specified
/// value split into legal parts.
static void getCopyToPartsVector(SelectionDAG &DAG, const SDLoc &DL,
                                 SDValue Val, SDValue *Parts, unsigned NumParts,
                                 MVT PartVT, const Value *V,
                                 Optional<CallingConv::ID> CallConv) {
  EVT ValueVT = Val.getValueType();
  assert(ValueVT.isVector() && "Not a vector");
  const TargetLowering &TLI = DAG.getTargetLoweringInfo();
  const bool IsABIRegCopy = CallConv.hasValue();
 
  if (NumParts == 1) {
    EVT PartEVT = PartVT;
    if (PartEVT == ValueVT) {
      // Nothing to do.
    } else if (PartVT.getSizeInBits() == ValueVT.getSizeInBits()) {
      // Bitconvert vector->vector case.
      Val = DAG.getNode(ISD::BITCAST, DL, PartVT, Val);
    } else if (SDValue Widened = widenVectorToPartType(DAG, Val, DL, PartVT)) {
      Val = Widened;
    } else if (PartVT.isVector() &&
               PartEVT.getVectorElementType().bitsGE(
                   ValueVT.getVectorElementType()) &&
               PartEVT.getVectorElementCount() ==
                   ValueVT.getVectorElementCount()) {
 
      // Promoted vector extract
      Val = DAG.getAnyExtOrTrunc(Val, DL, PartVT);
    } else {
      if (ValueVT.getVectorElementCount().isScalar()) {
        Val = DAG.getNode(ISD::EXTRACT_VECTOR_ELT, DL, PartVT, Val,
                          DAG.getVectorIdxConstant(0, DL));
      } else {
        uint64_t ValueSize = ValueVT.getFixedSizeInBits();
        assert(PartVT.getFixedSizeInBits() > ValueSize &&
               "lossy conversion of vector to scalar type");
        EVT IntermediateType = EVT::getIntegerVT(*DAG.getContext(), ValueSize);
        Val = DAG.getBitcast(IntermediateType, Val);
        Val = DAG.getAnyExtOrTrunc(Val, DL, PartVT);
      }
    }
 
    assert(Val.getValueType() == PartVT && "Unexpected vector part value type");
    Parts[0] = Val;
    return;
  }
 
  // Handle a multi-element vector.
  EVT IntermediateVT;
  MVT RegisterVT;
  unsigned NumIntermediates;
  unsigned NumRegs;
  if (IsABIRegCopy) {
    NumRegs = TLI.getVectorTypeBreakdownForCallingConv(
        *DAG.getContext(), CallConv.getValue(), ValueVT, IntermediateVT,
        NumIntermediates, RegisterVT);
  } else {
    NumRegs =
        TLI.getVectorTypeBreakdown(*DAG.getContext(), ValueVT, IntermediateVT,
                                   NumIntermediates, RegisterVT);
  }
 
  assert(NumRegs == NumParts && "Part count doesn't match vector breakdown!");
  NumParts = NumRegs; // Silence a compiler warning.
  assert(RegisterVT == PartVT && "Part type doesn't match vector breakdown!");
 
  assert(IntermediateVT.isScalableVector() == ValueVT.isScalableVector() &&
         "Mixing scalable and fixed vectors when copying in parts");
 
  Optional<ElementCount> DestEltCnt;
 
  if (IntermediateVT.isVector())
    DestEltCnt = IntermediateVT.getVectorElementCount() * NumIntermediates;
  else
    DestEltCnt = ElementCount::getFixed(NumIntermediates);
 
  EVT BuiltVectorTy = EVT::getVectorVT(
      *DAG.getContext(), IntermediateVT.getScalarType(), DestEltCnt.getValue());
 
  if (ValueVT == BuiltVectorTy) {
    // Nothing to do.
  } else if (ValueVT.getSizeInBits() == BuiltVectorTy.getSizeInBits()) {
    // Bitconvert vector->vector case.
    Val = DAG.getNode(ISD::BITCAST, DL, BuiltVectorTy, Val);
  } else {
    if (BuiltVectorTy.getVectorElementType().bitsGT(
            ValueVT.getVectorElementType())) {
      // Integer promotion.
      ValueVT = EVT::getVectorVT(*DAG.getContext(),
                                 BuiltVectorTy.getVectorElementType(),
                                 ValueVT.getVectorElementCount());
      Val = DAG.getNode(ISD::ANY_EXTEND, DL, ValueVT, Val);
    }
 
    if (SDValue Widened = widenVectorToPartType(DAG, Val, DL, BuiltVectorTy)) {
      Val = Widened;
    }
  }
 
  assert(Val.getValueType() == BuiltVectorTy && "Unexpected vector value type");
 
  // Split the vector into intermediate operands.
  SmallVector<SDValue, 8> Ops(NumIntermediates);
  for (unsigned i = 0; i != NumIntermediates; ++i) {
    if (IntermediateVT.isVector()) {
      // This does something sensible for scalable vectors - see the
      // definition of EXTRACT_SUBVECTOR for further details.
      unsigned IntermediateNumElts = IntermediateVT.getVectorMinNumElements();
      Ops[i] =
          DAG.getNode(ISD::EXTRACT_SUBVECTOR, DL, IntermediateVT, Val,
                      DAG.getVectorIdxConstant(i * IntermediateNumElts, DL));
    } else {
      Ops[i] = DAG.getNode(ISD::EXTRACT_VECTOR_ELT, DL, IntermediateVT, Val,
                           DAG.getVectorIdxConstant(i, DL));
    }
  }
 
  // Split the intermediate operands into legal parts.
  if (NumParts == NumIntermediates) {
    // If the register was not expanded, promote or copy the value,
    // as appropriate.
    for (unsigned i = 0; i != NumParts; ++i)
      getCopyToParts(DAG, DL, Ops[i], &Parts[i], 1, PartVT, V, CallConv);
  } else if (NumParts > 0) {
    // If the intermediate type was expanded, split each the value into
    // legal parts.
    assert(NumIntermediates != 0 && "division by zero");
    assert(NumParts % NumIntermediates == 0 &&
           "Must expand into a divisible number of parts!");
    unsigned Factor = NumParts / NumIntermediates;
    for (unsigned i = 0; i != NumIntermediates; ++i)
      getCopyToParts(DAG, DL, Ops[i], &Parts[i * Factor], Factor, PartVT, V,
                     CallConv);
  }
}
 
RegsForValue::RegsForValue(const SmallVector<unsigned, 4> &regs, MVT regvt,
                           EVT valuevt, Optional<CallingConv::ID> CC)
    : ValueVTs(1, valuevt), RegVTs(1, regvt), Regs(regs),
      RegCount(1, regs.size()), CallConv(CC) {}
 
RegsForValue::RegsForValue(LLVMContext &Context, const TargetLowering &TLI,
                           const DataLayout &DL, unsigned Reg, Type *Ty,
                           Optional<CallingConv::ID> CC) {
  ComputeValueVTs(TLI, DL, Ty, ValueVTs);
 
  CallConv = CC;
 
  for (EVT ValueVT : ValueVTs) {
    unsigned NumRegs =
        isABIMangled()
            ? TLI.getNumRegistersForCallingConv(Context, CC.getValue(), ValueVT)
            : TLI.getNumRegisters(Context, ValueVT);
    MVT RegisterVT =
        isABIMangled()
            ? TLI.getRegisterTypeForCallingConv(Context, CC.getValue(), ValueVT)
            : TLI.getRegisterType(Context, ValueVT);
    for (unsigned i = 0; i != NumRegs; ++i)
      Regs.push_back(Reg + i);
    RegVTs.push_back(RegisterVT);
    RegCount.push_back(NumRegs);
    Reg += NumRegs;
  }
}
 
SDValue RegsForValue::getCopyFromRegs(SelectionDAG &DAG,
                                      FunctionLoweringInfo &FuncInfo,
                                      const SDLoc &dl, SDValue &Chain,
                                      SDValue *Flag, const Value *V) const {
  // A Value with type {} or [0 x %t] needs no registers.
  if (ValueVTs.empty())
    return SDValue();
 
  const TargetLowering &TLI = DAG.getTargetLoweringInfo();
 
  // Assemble the legal parts into the final values.
  SmallVector<SDValue, 4> Values(ValueVTs.size());
  SmallVector<SDValue, 8> Parts;
  for (unsigned Value = 0, Part = 0, e = ValueVTs.size(); Value != e; ++Value) {
    // Copy the legal parts from the registers.
    EVT ValueVT = ValueVTs[Value];
    unsigned NumRegs = RegCount[Value];
    MVT RegisterVT = isABIMangled() ? TLI.getRegisterTypeForCallingConv(
                                          *DAG.getContext(),
                                          CallConv.getValue(), RegVTs[Value])
                                    : RegVTs[Value];
 
    Parts.resize(NumRegs);
    for (unsigned i = 0; i != NumRegs; ++i) {
      SDValue P;
      if (!Flag) {
        P = DAG.getCopyFromReg(Chain, dl, Regs[Part+i], RegisterVT);
      } else {
        P = DAG.getCopyFromReg(Chain, dl, Regs[Part+i], RegisterVT, *Flag);
        *Flag = P.getValue(2);
      }
 
      Chain = P.getValue(1);
      Parts[i] = P;
 
      // If the source register was virtual and if we know something about it,
      // add an assert node.
      if (!Register::isVirtualRegister(Regs[Part + i]) ||
          !RegisterVT.isInteger())
        continue;
 
      const FunctionLoweringInfo::LiveOutInfo *LOI =
        FuncInfo.GetLiveOutRegInfo(Regs[Part+i]);
      if (!LOI)
        continue;
 
      unsigned RegSize = RegisterVT.getScalarSizeInBits();
      unsigned NumSignBits = LOI->NumSignBits;
      unsigned NumZeroBits = LOI->Known.countMinLeadingZeros();
 
      if (NumZeroBits == RegSize) {
        // The current value is a zero.
        // Explicitly express that as it would be easier for
        // optimizations to kick in.
        Parts[i] = DAG.getConstant(0, dl, RegisterVT);
        continue;
      }
 
      // FIXME: We capture more information than the dag can represent.  For
      // now, just use the tightest assertzext/assertsext possible.
      bool isSExt;
      EVT FromVT(MVT::Other);
      if (NumZeroBits) {
        FromVT = EVT::getIntegerVT(*DAG.getContext(), RegSize - NumZeroBits);
        isSExt = false;
      } else if (NumSignBits > 1) {
        FromVT =
            EVT::getIntegerVT(*DAG.getContext(), RegSize - NumSignBits + 1);
        isSExt = true;
      } else {
        continue;
      }
      // Add an assertion node.
      assert(FromVT != MVT::Other);
      Parts[i] = DAG.getNode(isSExt ? ISD::AssertSext : ISD::AssertZext, dl,
                             RegisterVT, P, DAG.getValueType(FromVT));
    }
 
    Values[Value] = getCopyFromParts(DAG, dl, Parts.begin(), NumRegs,
                                     RegisterVT, ValueVT, V, CallConv);
    Part += NumRegs;
    Parts.clear();
  }
 
  return DAG.getNode(ISD::MERGE_VALUES, dl, DAG.getVTList(ValueVTs), Values);
}
 
void RegsForValue::getCopyToRegs(SDValue Val, SelectionDAG &DAG,
                                 const SDLoc &dl, SDValue &Chain, SDValue *Flag,
                                 const Value *V,
                                 ISD::NodeType PreferredExtendType) const {
  const TargetLowering &TLI = DAG.getTargetLoweringInfo();
  ISD::NodeType ExtendKind = PreferredExtendType;
 
  // Get the list of the values's legal parts.
  unsigned NumRegs = Regs.size();
  SmallVector<SDValue, 8> Parts(NumRegs);
  for (unsigned Value = 0, Part = 0, e = ValueVTs.size(); Value != e; ++Value) {
    unsigned NumParts = RegCount[Value];
 
    MVT RegisterVT = isABIMangled() ? TLI.getRegisterTypeForCallingConv(
                                          *DAG.getContext(),
                                          CallConv.getValue(), RegVTs[Value])
                                    : RegVTs[Value];
 
    if (ExtendKind == ISD::ANY_EXTEND && TLI.isZExtFree(Val, RegisterVT))
      ExtendKind = ISD::ZERO_EXTEND;
 
    getCopyToParts(DAG, dl, Val.getValue(Val.getResNo() + Value), &Parts[Part],
                   NumParts, RegisterVT, V, CallConv, ExtendKind);
    Part += NumParts;
  }
 
  // Copy the parts into the registers.
  SmallVector<SDValue, 8> Chains(NumRegs);
  for (unsigned i = 0; i != NumRegs; ++i) {
    SDValue Part;
    if (!Flag) {
      Part = DAG.getCopyToReg(Chain, dl, Regs[i], Parts[i]);
    } else {
      Part = DAG.getCopyToReg(Chain, dl, Regs[i], Parts[i], *Flag);
      *Flag = Part.getValue(1);
    }
 
    Chains[i] = Part.getValue(0);
  }
 
  if (NumRegs == 1 || Flag)
    // If NumRegs > 1 && Flag is used then the use of the last CopyToReg is
    // flagged to it. That is the CopyToReg nodes and the user are considered
    // a single scheduling unit. If we create a TokenFactor and return it as
    // chain, then the TokenFactor is both a predecessor (operand) of the
    // user as well as a successor (the TF operands are flagged to the user).
    // c1, f1 = CopyToReg
    // c2, f2 = CopyToReg
    // c3     = TokenFactor c1, c2
    // ...
    //        = op c3, ..., f2
    Chain = Chains[NumRegs-1];
  else
    Chain = DAG.getNode(ISD::TokenFactor, dl, MVT::Other, Chains);
}
 
void RegsForValue::AddInlineAsmOperands(unsigned Code, bool HasMatching,
                                        unsigned MatchingIdx, const SDLoc &dl,
                                        SelectionDAG &DAG,
                                        std::vector<SDValue> &Ops) const {
  const TargetLowering &TLI = DAG.getTargetLoweringInfo();
 
  unsigned Flag = InlineAsm::getFlagWord(Code, Regs.size());
  if (HasMatching)
    Flag = InlineAsm::getFlagWordForMatchingOp(Flag, MatchingIdx);
  else if (!Regs.empty() && Register::isVirtualRegister(Regs.front())) {
    // Put the register class of the virtual registers in the flag word.  That
    // way, later passes can recompute register class constraints for inline
    // assembly as well as normal instructions.
    // Don't do this for tied operands that can use the regclass information
    // from the def.
    const MachineRegisterInfo &MRI = DAG.getMachineFunction().getRegInfo();
    const TargetRegisterClass *RC = MRI.getRegClass(Regs.front());
    Flag = InlineAsm::getFlagWordForRegClass(Flag, RC->getID());
  }
 
  SDValue Res = DAG.getTargetConstant(Flag, dl, MVT::i32);
  Ops.push_back(Res);
 
  if (Code == InlineAsm::Kind_Clobber) {
    // Clobbers should always have a 1:1 mapping with registers, and may
    // reference registers that have illegal (e.g. vector) types. Hence, we
    // shouldn't try to apply any sort of splitting logic to them.
    assert(Regs.size() == RegVTs.size() && Regs.size() == ValueVTs.size() &&
           "No 1:1 mapping from clobbers to regs?");
    Register SP = TLI.getStackPointerRegisterToSaveRestore();
    (void)SP;
    for (unsigned I = 0, E = ValueVTs.size(); I != E; ++I) {
      Ops.push_back(DAG.getRegister(Regs[I], RegVTs[I]));
      assert(
          (Regs[I] != SP ||
           DAG.getMachineFunction().getFrameInfo().hasOpaqueSPAdjustment()) &&
          "If we clobbered the stack pointer, MFI should know about it.");
    }
    return;
  }
 
  for (unsigned Value = 0, Reg = 0, e = ValueVTs.size(); Value != e; ++Value) {
    MVT RegisterVT = RegVTs[Value];
    unsigned NumRegs = TLI.getNumRegisters(*DAG.getContext(), ValueVTs[Value],
                                           RegisterVT);
    for (unsigned i = 0; i != NumRegs; ++i) {
      assert(Reg < Regs.size() && "Mismatch in # registers expected");
      unsigned TheReg = Regs[Reg++];
      Ops.push_back(DAG.getRegister(TheReg, RegisterVT));
    }
  }
}
 
SmallVector<std::pair<unsigned, TypeSize>, 4>
RegsForValue::getRegsAndSizes() const {
  SmallVector<std::pair<unsigned, TypeSize>, 4> OutVec;
  unsigned I = 0;
  for (auto CountAndVT : zip_first(RegCount, RegVTs)) {
    unsigned RegCount = std::get<0>(CountAndVT);
    MVT RegisterVT = std::get<1>(CountAndVT);
    TypeSize RegisterSize = RegisterVT.getSizeInBits();
    for (unsigned E = I + RegCount; I != E; ++I)
      OutVec.push_back(std::make_pair(Regs[I], RegisterSize));
  }
  return OutVec;
}
 
void SelectionDAGBuilder::init(GCFunctionInfo *gfi, AliasAnalysis *aa,
                               const TargetLibraryInfo *li) {
  AA = aa;
  GFI = gfi;
  LibInfo = li;
  DL = &DAG.getDataLayout();
  Context = DAG.getContext();
  LPadToCallSiteMap.clear();
  SL->init(DAG.getTargetLoweringInfo(), TM, DAG.getDataLayout());
}
 
void SelectionDAGBuilder::clear() {
  NodeMap.clear();
  UnusedArgNodeMap.clear();
  PendingLoads.clear();
  PendingExports.clear();
  PendingConstrainedFP.clear();
  PendingConstrainedFPStrict.clear();
  CurInst = nullptr;
  HasTailCall = false;
  SDNodeOrder = LowestSDNodeOrder;
  StatepointLowering.clear();
}
 
void SelectionDAGBuilder::clearDanglingDebugInfo() {
  DanglingDebugInfoMap.clear();
}
 
// Update DAG root to include dependencies on Pending chains.
SDValue SelectionDAGBuilder::updateRoot(SmallVectorImpl<SDValue> &Pending) {
  SDValue Root = DAG.getRoot();
 
  if (Pending.empty())
    return Root;
 
  // Add current root to PendingChains, unless we already indirectly
  // depend on it.
  if (Root.getOpcode() != ISD::EntryToken) {
    unsigned i = 0, e = Pending.size();
    for (; i != e; ++i) {
      assert(Pending[i].getNode()->getNumOperands() > 1);
      if (Pending[i].getNode()->getOperand(0) == Root)
        break;  // Don't add the root if we already indirectly depend on it.
    }
 
    if (i == e)
      Pending.push_back(Root);
  }
 
  if (Pending.size() == 1)
    Root = Pending[0];
  else
    Root = DAG.getTokenFactor(getCurSDLoc(), Pending);
 
  DAG.setRoot(Root);
  Pending.clear();
  return Root;
}
 
SDValue SelectionDAGBuilder::getMemoryRoot() {
  return updateRoot(PendingLoads);
}
 
SDValue SelectionDAGBuilder::getRoot() {
  // Chain up all pending constrained intrinsics together with all
  // pending loads, by simply appending them to PendingLoads and
  // then calling getMemoryRoot().
  PendingLoads.reserve(PendingLoads.size() +
                       PendingConstrainedFP.size() +
                       PendingConstrainedFPStrict.size());
  PendingLoads.append(PendingConstrainedFP.begin(),
                      PendingConstrainedFP.end());
  PendingLoads.append(PendingConstrainedFPStrict.begin(),
                      PendingConstrainedFPStrict.end());
  PendingConstrainedFP.clear();
  PendingConstrainedFPStrict.clear();
  return getMemoryRoot();
}
 
SDValue SelectionDAGBuilder::getControlRoot() {
  // We need to emit pending fpexcept.strict constrained intrinsics,
  // so append them to the PendingExports list.
  PendingExports.append(PendingConstrainedFPStrict.begin(),
                        PendingConstrainedFPStrict.end());
  PendingConstrainedFPStrict.clear();
  return updateRoot(PendingExports);
}
 
void SelectionDAGBuilder::visit(const Instruction &I) {
  // Set up outgoing PHI node register values before emitting the terminator.
  if (I.isTerminator()) {
    HandlePHINodesInSuccessorBlocks(I.getParent());
  }
 
  // Increase the SDNodeOrder if dealing with a non-debug instruction.
  if (!isa<DbgInfoIntrinsic>(I))
    ++SDNodeOrder;
 
  CurInst = &I;
 
  visit(I.getOpcode(), I);
 
  if (!I.isTerminator() && !HasTailCall &&
      !isa<GCStatepointInst>(I)) // statepoints handle their exports internally
    CopyToExportRegsIfNeeded(&I);
 
  CurInst = nullptr;
}
 
void SelectionDAGBuilder::visitPHI(const PHINode &) {
  llvm_unreachable("SelectionDAGBuilder shouldn't visit PHI nodes!");
}
 
void SelectionDAGBuilder::visit(unsigned Opcode, const User &I) {
  // Note: this doesn't use InstVisitor, because it has to work with
  // ConstantExpr's in addition to instructions.
  switch (Opcode) {
  default: llvm_unreachable("Unknown instruction type encountered!");
    // Build the switch statement using the Instruction.def file.
#define HANDLE_INST(NUM, OPCODE, CLASS) \
    case Instruction::OPCODE: visit##OPCODE((const CLASS&)I); break;
#include "llvm/IR/Instruction.def"
  }
}
 
void SelectionDAGBuilder::addDanglingDebugInfo(const DbgValueInst *DI,
                                               DebugLoc DL, unsigned Order) {
  // We treat variadic dbg_values differently at this stage.
  if (DI->hasArgList()) {
    // For variadic dbg_values we will now insert an undef.
    // FIXME: We can potentially recover these!
    SmallVector<SDDbgOperand, 2> Locs;
    for (const Value *V : DI->getValues()) {
      auto Undef = UndefValue::get(V->getType());
      Locs.push_back(SDDbgOperand::fromConst(Undef));
    }
    SDDbgValue *SDV = DAG.getDbgValueList(
        DI->getVariable(), DI->getExpression(), Locs, {},
        /*IsIndirect=*/false, DL, Order, /*IsVariadic=*/true);
    DAG.AddDbgValue(SDV, /*isParameter=*/false);
  } else {
    // TODO: Dangling debug info will eventually either be resolved or produce
    // an Undef DBG_VALUE. However in the resolution case, a gap may appear
    // between the original dbg.value location and its resolved DBG_VALUE,
    // which we should ideally fill with an extra Undef DBG_VALUE.
    assert(DI->getNumVariableLocationOps() == 1 &&
           "DbgValueInst without an ArgList should have a single location "
           "operand.");
    DanglingDebugInfoMap[DI->getValue(0)].emplace_back(DI, DL, Order);
  }
}
 
void SelectionDAGBuilder::dropDanglingDebugInfo(const DILocalVariable *Variable,
                                                const DIExpression *Expr) {
  auto isMatchingDbgValue = [&](DanglingDebugInfo &DDI) {
    const DbgValueInst *DI = DDI.getDI();
    DIVariable *DanglingVariable = DI->getVariable();
    DIExpression *DanglingExpr = DI->getExpression();
    if (DanglingVariable == Variable && Expr->fragmentsOverlap(DanglingExpr)) {
      LLVM_DEBUG(dbgs() << "Dropping dangling debug info for " << *DI << "\n");
      return true;
    }
    return false;
  };
 
  for (auto &DDIMI : DanglingDebugInfoMap) {
    DanglingDebugInfoVector &DDIV = DDIMI.second;
 
    // If debug info is to be dropped, run it through final checks to see
    // whether it can be salvaged.
    for (auto &DDI : DDIV)
      if (isMatchingDbgValue(DDI))
        salvageUnresolvedDbgValue(DDI);
 
    erase_if(DDIV, isMatchingDbgValue);
  }
}
 
// resolveDanglingDebugInfo - if we saw an earlier dbg_value referring to V,
// generate the debug data structures now that we've seen its definition.
void SelectionDAGBuilder::resolveDanglingDebugInfo(const Value *V,
                                                   SDValue Val) {
  auto DanglingDbgInfoIt = DanglingDebugInfoMap.find(V);
  if (DanglingDbgInfoIt == DanglingDebugInfoMap.end())
    return;
 
  DanglingDebugInfoVector &DDIV = DanglingDbgInfoIt->second;
  for (auto &DDI : DDIV) {
    const DbgValueInst *DI = DDI.getDI();
    assert(!DI->hasArgList() && "Not implemented for variadic dbg_values");
    assert(DI && "Ill-formed DanglingDebugInfo");
    DebugLoc dl = DDI.getdl();
    unsigned ValSDNodeOrder = Val.getNode()->getIROrder();
    unsigned DbgSDNodeOrder = DDI.getSDNodeOrder();
    DILocalVariable *Variable = DI->getVariable();
    DIExpression *Expr = DI->getExpression();
    assert(Variable->isValidLocationForIntrinsic(dl) &&
           "Expected inlined-at fields to agree");
    SDDbgValue *SDV;
    if (Val.getNode()) {
      // FIXME: I doubt that it is correct to resolve a dangling DbgValue as a
      // FuncArgumentDbgValue (it would be hoisted to the function entry, and if
      // we couldn't resolve it directly when examining the DbgValue intrinsic
      // in the first place we should not be more successful here). Unless we
      // have some test case that prove this to be correct we should avoid
      // calling EmitFuncArgumentDbgValue here.
      if (!EmitFuncArgumentDbgValue(V, Variable, Expr, dl, false, Val)) {
        LLVM_DEBUG(dbgs() << "Resolve dangling debug info [order="
                          << DbgSDNodeOrder << "] for:\n  " << *DI << "\n");
        LLVM_DEBUG(dbgs() << "  By mapping to:\n    "; Val.dump());
        // Increase the SDNodeOrder for the DbgValue here to make sure it is
        // inserted after the definition of Val when emitting the instructions
        // after ISel. An alternative could be to teach
        // ScheduleDAGSDNodes::EmitSchedule to delay the insertion properly.
        LLVM_DEBUG(if (ValSDNodeOrder > DbgSDNodeOrder) dbgs()
                   << "changing SDNodeOrder from " << DbgSDNodeOrder << " to "
                   << ValSDNodeOrder << "\n");
        SDV = getDbgValue(Val, Variable, Expr, dl,
                          std::max(DbgSDNodeOrder, ValSDNodeOrder));
        DAG.AddDbgValue(SDV, false);
      } else
        LLVM_DEBUG(dbgs() << "Resolved dangling debug info for " << *DI
                          << "in EmitFuncArgumentDbgValue\n");
    } else {
      LLVM_DEBUG(dbgs() << "Dropping debug info for " << *DI << "\n");
      auto Undef = UndefValue::get(DDI.getDI()->getValue(0)->getType());
      auto SDV =
          DAG.getConstantDbgValue(Variable, Expr, Undef, dl, DbgSDNodeOrder);
      DAG.AddDbgValue(SDV, false);
    }
  }
  DDIV.clear();
}
 
void SelectionDAGBuilder::salvageUnresolvedDbgValue(DanglingDebugInfo &DDI) {
  // TODO: For the variadic implementation, instead of only checking the fail
  // state of `handleDebugValue`, we need know specifically which values were
  // invalid, so that we attempt to salvage only those values when processing
  // a DIArgList.
  assert(!DDI.getDI()->hasArgList() &&
         "Not implemented for variadic dbg_values");
  Value *V = DDI.getDI()->getValue(0);
  DILocalVariable *Var = DDI.getDI()->getVariable();
  DIExpression *Expr = DDI.getDI()->getExpression();
  DebugLoc DL = DDI.getdl();
  DebugLoc InstDL = DDI.getDI()->getDebugLoc();
  unsigned SDOrder = DDI.getSDNodeOrder();
  // Currently we consider only dbg.value intrinsics -- we tell the salvager
  // that DW_OP_stack_value is desired.
  assert(isa<DbgValueInst>(DDI.getDI()));
  bool StackValue = true;
 
  // Can this Value can be encoded without any further work?
  if (handleDebugValue(V, Var, Expr, DL, InstDL, SDOrder, /*IsVariadic=*/false))
    return;
 
  // Attempt to salvage back through as many instructions as possible. Bail if
  // a non-instruction is seen, such as a constant expression or global
  // variable. FIXME: Further work could recover those too.
  while (isa<Instruction>(V)) {
    Instruction &VAsInst = *cast<Instruction>(V);
    // Temporary "0", awaiting real implementation.
    SmallVector<uint64_t, 16> Ops;
    SmallVector<Value *, 4> AdditionalValues;
    V = salvageDebugInfoImpl(VAsInst, Expr->getNumLocationOperands(), Ops,
                             AdditionalValues);
    // If we cannot salvage any further, and haven't yet found a suitable debug
    // expression, bail out.
    if (!V)
      break;
 
    // TODO: If AdditionalValues isn't empty, then the salvage can only be
    // represented with a DBG_VALUE_LIST, so we give up. When we have support
    // here for variadic dbg_values, remove that condition.
    if (!AdditionalValues.empty())
      break;
 
    // New value and expr now represent this debuginfo.
    Expr = DIExpression::appendOpsToArg(Expr, Ops, 0, StackValue);
 
    // Some kind of simplification occurred: check whether the operand of the
    // salvaged debug expression can be encoded in this DAG.
    if (handleDebugValue(V, Var, Expr, DL, InstDL, SDOrder,
                         /*IsVariadic=*/false)) {
      LLVM_DEBUG(dbgs() << "Salvaged debug location info for:\n  "
                        << DDI.getDI() << "\nBy stripping back to:\n  " << V);
      return;
    }
  }
 
  // This was the final opportunity to salvage this debug information, and it
  // couldn't be done. Place an undef DBG_VALUE at this location to terminate
  // any earlier variable location.
  auto Undef = UndefValue::get(DDI.getDI()->getValue(0)->getType());
  auto SDV = DAG.getConstantDbgValue(Var, Expr, Undef, DL, SDNodeOrder);
  DAG.AddDbgValue(SDV, false);
 
  LLVM_DEBUG(dbgs() << "Dropping debug value info for:\n  " << DDI.getDI()
                    << "\n");
  LLVM_DEBUG(dbgs() << "  Last seen at:\n    " << *DDI.getDI()->getOperand(0)
                    << "\n");
}
 
bool SelectionDAGBuilder::handleDebugValue(ArrayRef<const Value *> Values,
                                           DILocalVariable *Var,
                                           DIExpression *Expr, DebugLoc dl,
                                           DebugLoc InstDL, unsigned Order,
                                           bool IsVariadic) {
  if (Values.empty())
    return true;
  SmallVector<SDDbgOperand> LocationOps;
  SmallVector<SDNode *> Dependencies;
  for (const Value *V : Values) {
    // Constant value.
    if (isa<ConstantInt>(V) || isa<ConstantFP>(V) || isa<UndefValue>(V) ||
        isa<ConstantPointerNull>(V)) {
      LocationOps.emplace_back(SDDbgOperand::fromConst(V));
      continue;
    }
 
    // If the Value is a frame index, we can create a FrameIndex debug value
    // without relying on the DAG at all.
    if (const AllocaInst *AI = dyn_cast<AllocaInst>(V)) {
      auto SI = FuncInfo.StaticAllocaMap.find(AI);
      if (SI != FuncInfo.StaticAllocaMap.end()) {
        LocationOps.emplace_back(SDDbgOperand::fromFrameIdx(SI->second));
        continue;
      }
    }
 
    // Do not use getValue() in here; we don't want to generate code at
    // this point if it hasn't been done yet.
    SDValue N = NodeMap[V];
    if (!N.getNode() && isa<Argument>(V)) // Check unused arguments map.
      N = UnusedArgNodeMap[V];
    if (N.getNode()) {
      // Only emit func arg dbg value for non-variadic dbg.values for now.
      if (!IsVariadic && EmitFuncArgumentDbgValue(V, Var, Expr, dl, false, N))
        return true;
      if (auto *FISDN = dyn_cast<FrameIndexSDNode>(N.getNode())) {
        // Construct a FrameIndexDbgValue for FrameIndexSDNodes so we can
        // describe stack slot locations.
        //
        // Consider "int x = 0; int *px = &x;". There are two kinds of
        // interesting debug values here after optimization:
        //
        //   dbg.value(i32* %px, !"int *px", !DIExpression()), and
        //   dbg.value(i32* %px, !"int x", !DIExpression(DW_OP_deref))
        //
        // Both describe the direct values of their associated variables.
        Dependencies.push_back(N.getNode());
        LocationOps.emplace_back(SDDbgOperand::fromFrameIdx(FISDN->getIndex()));
        continue;
      }
      LocationOps.emplace_back(
          SDDbgOperand::fromNode(N.getNode(), N.getResNo()));
      continue;
    }
 
    const TargetLowering &TLI = DAG.getTargetLoweringInfo();
    // Special rules apply for the first dbg.values of parameter variables in a
    // function. Identify them by the fact they reference Argument Values, that
    // they're parameters, and they are parameters of the current function. We
    // need to let them dangle until they get an SDNode.
    bool IsParamOfFunc =
        isa<Argument>(V) && Var->isParameter() && !InstDL.getInlinedAt();
    if (IsParamOfFunc)
      return false;
 
    // The value is not used in this block yet (or it would have an SDNode).
    // We still want the value to appear for the user if possible -- if it has
    // an associated VReg, we can refer to that instead.
    auto VMI = FuncInfo.ValueMap.find(V);
    if (VMI != FuncInfo.ValueMap.end()) {
      unsigned Reg = VMI->second;
      // If this is a PHI node, it may be split up into several MI PHI nodes
      // (in FunctionLoweringInfo::set).
      RegsForValue RFV(V->getContext(), TLI, DAG.getDataLayout(), Reg,
                       V->getType(), None);
      if (RFV.occupiesMultipleRegs()) {
        // FIXME: We could potentially support variadic dbg_values here.
        if (IsVariadic)
          return false;
        unsigned Offset = 0;
        unsigned BitsToDescribe = 0;
        if (auto VarSize = Var->getSizeInBits())
          BitsToDescribe = *VarSize;
        if (auto Fragment = Expr->getFragmentInfo())
          BitsToDescribe = Fragment->SizeInBits;
        for (const auto &RegAndSize : RFV.getRegsAndSizes()) {
          // Bail out if all bits are described already.
          if (Offset >= BitsToDescribe)
            break;
          // TODO: handle scalable vectors.
          unsigned RegisterSize = RegAndSize.second;
          unsigned FragmentSize = (Offset + RegisterSize > BitsToDescribe)
                                      ? BitsToDescribe - Offset
                                      : RegisterSize;
          auto FragmentExpr = DIExpression::createFragmentExpression(
              Expr, Offset, FragmentSize);
          if (!FragmentExpr)
            continue;
          SDDbgValue *SDV = DAG.getVRegDbgValue(
              Var, *FragmentExpr, RegAndSize.first, false, dl, SDNodeOrder);
          DAG.AddDbgValue(SDV, false);
          Offset += RegisterSize;
        }
        return true;
      }
      // We can use simple vreg locations for variadic dbg_values as well.
      LocationOps.emplace_back(SDDbgOperand::fromVReg(Reg));
      continue;
    }
    // We failed to create a SDDbgOperand for V.
    return false;
  }
 
  // We have created a SDDbgOperand for each Value in Values.
  // Should use Order instead of SDNodeOrder?
  assert(!LocationOps.empty());
  SDDbgValue *SDV =
      DAG.getDbgValueList(Var, Expr, LocationOps, Dependencies,
                          /*IsIndirect=*/false, dl, SDNodeOrder, IsVariadic);
  DAG.AddDbgValue(SDV, /*isParameter=*/false);
  return true;
}
 
void SelectionDAGBuilder::resolveOrClearDbgInfo() {
  // Try to fixup any remaining dangling debug info -- and drop it if we can't.
  for (auto &Pair : DanglingDebugInfoMap)
    for (auto &DDI : Pair.second)
      salvageUnresolvedDbgValue(DDI);
  clearDanglingDebugInfo();
}
 
/// getCopyFromRegs - If there was virtual register allocated for the value V
/// emit CopyFromReg of the specified type Ty. Return empty SDValue() otherwise.
SDValue SelectionDAGBuilder::getCopyFromRegs(const Value *V, Type *Ty) {
  DenseMap<const Value *, Register>::iterator It = FuncInfo.ValueMap.find(V);
  SDValue Result;
 
  if (It != FuncInfo.ValueMap.end()) {
    Register InReg = It->second;
 
    RegsForValue RFV(*DAG.getContext(), DAG.getTargetLoweringInfo(),
                     DAG.getDataLayout(), InReg, Ty,
                     None); // This is not an ABI copy.
    SDValue Chain = DAG.getEntryNode();
    Result = RFV.getCopyFromRegs(DAG, FuncInfo, getCurSDLoc(), Chain, nullptr,
                                 V);
    resolveDanglingDebugInfo(V, Result);
  }
 
  return Result;
}
 
/// getValue - Return an SDValue for the given Value.
SDValue SelectionDAGBuilder::getValue(const Value *V) {
  // If we already have an SDValue for this value, use it. It's important
  // to do this first, so that we don't create a CopyFromReg if we already
  // have a regular SDValue.
  SDValue &N = NodeMap[V];
  if (N.getNode()) return N;
 
  // If there's a virtual register allocated and initialized for this
  // value, use it.
  if (SDValue copyFromReg = getCopyFromRegs(V, V->getType()))
    return copyFromReg;
 
  // Otherwise create a new SDValue and remember it.
  SDValue Val = getValueImpl(V);
  NodeMap[V] = Val;
  resolveDanglingDebugInfo(V, Val);
  return Val;
}
 
/// getNonRegisterValue - Return an SDValue for the given Value, but
/// don't look in FuncInfo.ValueMap for a virtual register.
SDValue SelectionDAGBuilder::getNonRegisterValue(const Value *V) {
  // If we already have an SDValue for this value, use it.
  SDValue &N = NodeMap[V];
  if (N.getNode()) {
    if (isa<ConstantSDNode>(N) || isa<ConstantFPSDNode>(N)) {
      // Remove the debug location from the node as the node is about to be used
      // in a location which may differ from the original debug location.  This
      // is relevant to Constant and ConstantFP nodes because they can appear
      // as constant expressions inside PHI nodes.
      N->setDebugLoc(DebugLoc());
    }
    return N;
  }
 
  // Otherwise create a new SDValue and remember it.
  SDValue Val = getValueImpl(V);
  NodeMap[V] = Val;
  resolveDanglingDebugInfo(V, Val);
  return Val;
}
 
/// getValueImpl - Helper function for getValue and getNonRegisterValue.
/// Create an SDValue for the given value.
SDValue SelectionDAGBuilder::getValueImpl(const Value *V) {
  const TargetLowering &TLI = DAG.getTargetLoweringInfo();
 
  if (const Constant *C = dyn_cast<Constant>(V)) {
    EVT VT = TLI.getValueType(DAG.getDataLayout(), V->getType(), true);
 
    if (const ConstantInt *CI = dyn_cast<ConstantInt>(C))
      return DAG.getConstant(*CI, getCurSDLoc(), VT);
 
    if (const GlobalValue *GV = dyn_cast<GlobalValue>(C))
      return DAG.getGlobalAddress(GV, getCurSDLoc(), VT);
 
    if (isa<ConstantPointerNull>(C)) {
      unsigned AS = V->getType()->getPointerAddressSpace();
      return DAG.getConstant(0, getCurSDLoc(),
                             TLI.getPointerTy(DAG.getDataLayout(), AS));
    }
 
    if (match(C, m_VScale(DAG.getDataLayout())))
      return DAG.getVScale(getCurSDLoc(), VT, APInt(VT.getSizeInBits(), 1));
 
    if (const ConstantFP *CFP = dyn_cast<ConstantFP>(C))
      return DAG.getConstantFP(*CFP, getCurSDLoc(), VT);
 
    if (isa<UndefValue>(C) && !V->getType()->isAggregateType())
      return DAG.getUNDEF(VT);
 
    if (const ConstantExpr *CE = dyn_cast<ConstantExpr>(C)) {
      visit(CE->getOpcode(), *CE);
      SDValue N1 = NodeMap[V];
      assert(N1.getNode() && "visit didn't populate the NodeMap!");
      return N1;
    }
 
    if (isa<ConstantStruct>(C) || isa<ConstantArray>(C)) {
      SmallVector<SDValue, 4> Constants;
      for (const Use &U : C->operands()) {
        SDNode *Val = getValue(U).getNode();
        // If the operand is an empty aggregate, there are no values.
        if (!Val) continue;
        // Add each leaf value from the operand to the Constants list
        // to form a flattened list of all the values.
        for (unsigned i = 0, e = Val->getNumValues(); i != e; ++i)
          Constants.push_back(SDValue(Val, i));
      }
 
      return DAG.getMergeValues(Constants, getCurSDLoc());
    }
 
    if (const ConstantDataSequential *CDS =
          dyn_cast<ConstantDataSequential>(C)) {
      SmallVector<SDValue, 4> Ops;
      for (unsigned i = 0, e = CDS->getNumElements(); i != e; ++i) {
        SDNode *Val = getValue(CDS->getElementAsConstant(i)).getNode();
        // Add each leaf value from the operand to the Constants list
        // to form a flattened list of all the values.
        for (unsigned i = 0, e = Val->getNumValues(); i != e; ++i)
          Ops.push_back(SDValue(Val, i));
      }
 
      if (isa<ArrayType>(CDS->getType()))
        return DAG.getMergeValues(Ops, getCurSDLoc());
      return NodeMap[V] = DAG.getBuildVector(VT, getCurSDLoc(), Ops);
    }
 
    if (C->getType()->isStructTy() || C->getType()->isArrayTy()) {
      assert((isa<ConstantAggregateZero>(C) || isa<UndefValue>(C)) &&
             "Unknown struct or array constant!");
 
      SmallVector<EVT, 4> ValueVTs;
      ComputeValueVTs(TLI, DAG.getDataLayout(), C->getType(), ValueVTs);
      unsigned NumElts = ValueVTs.size();
      if (NumElts == 0)
        return SDValue(); // empty struct
      SmallVector<SDValue, 4> Constants(NumElts);
      for (unsigned i = 0; i != NumElts; ++i) {
        EVT EltVT = ValueVTs[i];
        if (isa<UndefValue>(C))
          Constants[i] = DAG.getUNDEF(EltVT);
        else if (EltVT.isFloatingPoint())
          Constants[i] = DAG.getConstantFP(0, getCurSDLoc(), EltVT);
        else
          Constants[i] = DAG.getConstant(0, getCurSDLoc(), EltVT);
      }
 
      return DAG.getMergeValues(Constants, getCurSDLoc());
    }
 
    if (const BlockAddress *BA = dyn_cast<BlockAddress>(C))
      return DAG.getBlockAddress(BA, VT);
 
    if (const auto *Equiv = dyn_cast<DSOLocalEquivalent>(C))
      return getValue(Equiv->getGlobalValue());
 
    VectorType *VecTy = cast<VectorType>(V->getType());
 
    // Now that we know the number and type of the elements, get that number of
    // elements into the Ops array based on what kind of constant it is.
    if (const ConstantVector *CV = dyn_cast<ConstantVector>(C)) {
      SmallVector<SDValue, 16> Ops;
      unsigned NumElements = cast<FixedVectorType>(VecTy)->getNumElements();
      for (unsigned i = 0; i != NumElements; ++i)
        Ops.push_back(getValue(CV->getOperand(i)));
 
      return NodeMap[V] = DAG.getBuildVector(VT, getCurSDLoc(), Ops);
    } else if (isa<ConstantAggregateZero>(C)) {
      EVT EltVT =
          TLI.getValueType(DAG.getDataLayout(), VecTy->getElementType());
 
      SDValue Op;
      if (EltVT.isFloatingPoint())
        Op = DAG.getConstantFP(0, getCurSDLoc(), EltVT);
      else
        Op = DAG.getConstant(0, getCurSDLoc(), EltVT);
 
      if (isa<ScalableVectorType>(VecTy))
        return NodeMap[V] = DAG.getSplatVector(VT, getCurSDLoc(), Op);
      else {
        SmallVector<SDValue, 16> Ops;
        Ops.assign(cast<FixedVectorType>(VecTy)->getNumElements(), Op);
        return NodeMap[V] = DAG.getBuildVector(VT, getCurSDLoc(), Ops);
      }
    }
    llvm_unreachable("Unknown vector constant");
  }
 
  // If this is a static alloca, generate it as the frameindex instead of
  // computation.
  if (const AllocaInst *AI = dyn_cast<AllocaInst>(V)) {
    DenseMap<const AllocaInst*, int>::iterator SI =
      FuncInfo.StaticAllocaMap.find(AI);
    if (SI != FuncInfo.StaticAllocaMap.end())
      return DAG.getFrameIndex(SI->second,
                               TLI.getFrameIndexTy(DAG.getDataLayout()));
  }
 
  // If this is an instruction which fast-isel has deferred, select it now.
  if (const Instruction *Inst = dyn_cast<Instruction>(V)) {
    unsigned InReg = FuncInfo.InitializeRegForValue(Inst);
 
    RegsForValue RFV(*DAG.getContext(), TLI, DAG.getDataLayout(), InReg,
                     Inst->getType(), None);
    SDValue Chain = DAG.getEntryNode();
    return RFV.getCopyFromRegs(DAG, FuncInfo, getCurSDLoc(), Chain, nullptr, V);
  }
 
  if (const MetadataAsValue *MD = dyn_cast<MetadataAsValue>(V)) {
    return DAG.getMDNode(cast<MDNode>(MD->getMetadata()));
  }
  llvm_unreachable("Can't get register for value!");
}
 
void SelectionDAGBuilder::visitCatchPad(const CatchPadInst &I) {
  auto Pers = classifyEHPersonality(FuncInfo.Fn->getPersonalityFn());
  bool IsMSVCCXX = Pers == EHPersonality::MSVC_CXX;
  bool IsCoreCLR = Pers == EHPersonality::CoreCLR;
  bool IsSEH = isAsynchronousEHPersonality(Pers);
  MachineBasicBlock *CatchPadMBB = FuncInfo.MBB;
  if (!IsSEH)
    CatchPadMBB->setIsEHScopeEntry();
  // In MSVC C++ and CoreCLR, catchblocks are funclets and need prologues.
  if (IsMSVCCXX || IsCoreCLR)
    CatchPadMBB->setIsEHFuncletEntry();
}
 
void SelectionDAGBuilder::visitCatchRet(const CatchReturnInst &I) {
  // Update machine-CFG edge.
  MachineBasicBlock *TargetMBB = FuncInfo.MBBMap[I.getSuccessor()];
  FuncInfo.MBB->addSuccessor(TargetMBB);
  TargetMBB->setIsEHCatchretTarget(true);
  DAG.getMachineFunction().setHasEHCatchret(true);
 
  auto Pers = classifyEHPersonality(FuncInfo.Fn->getPersonalityFn());
  bool IsSEH = isAsynchronousEHPersonality(Pers);
  if (IsSEH) {
    // If this is not a fall-through branch or optimizations are switched off,
    // emit the branch.
    if (TargetMBB != NextBlock(FuncInfo.MBB) ||
        TM.getOptLevel() == CodeGenOpt::None)
      DAG.setRoot(DAG.getNode(ISD::BR, getCurSDLoc(), MVT::Other,
                              getControlRoot(), DAG.getBasicBlock(TargetMBB)));
    return;
  }
 
  // Figure out the funclet membership for the catchret's successor.
  // This will be used by the FuncletLayout pass to determine how to order the
  // BB's.
  // A 'catchret' returns to the outer scope's color.
  Value *ParentPad = I.getCatchSwitchParentPad();
  const BasicBlock *SuccessorColor;
  if (isa<ConstantTokenNone>(ParentPad))
    SuccessorColor = &FuncInfo.Fn->getEntryBlock();
  else
    SuccessorColor = cast<Instruction>(ParentPad)->getParent();
  assert(SuccessorColor && "No parent funclet for catchret!");
  MachineBasicBlock *SuccessorColorMBB = FuncInfo.MBBMap[SuccessorColor];
  assert(SuccessorColorMBB && "No MBB for SuccessorColor!");
 
  // Create the terminator node.
  SDValue Ret = DAG.getNode(ISD::CATCHRET, getCurSDLoc(), MVT::Other,
                            getControlRoot(), DAG.getBasicBlock(TargetMBB),
                            DAG.getBasicBlock(SuccessorColorMBB));
  DAG.setRoot(Ret);
}
 
void SelectionDAGBuilder::visitCleanupPad(const CleanupPadInst &CPI) {
  // Don't emit any special code for the cleanuppad instruction. It just marks
  // the start of an EH scope/funclet.
  FuncInfo.MBB->setIsEHScopeEntry();
  auto Pers = classifyEHPersonality(FuncInfo.Fn->getPersonalityFn());
  if (Pers != EHPersonality::Wasm_CXX) {
    FuncInfo.MBB->setIsEHFuncletEntry();
    FuncInfo.MBB->setIsCleanupFuncletEntry();
  }
}
 
// In wasm EH, even though a catchpad may not catch an exception if a tag does
// not match, it is OK to add only the first unwind destination catchpad to the
// successors, because there will be at least one invoke instruction within the
// catch scope that points to the next unwind destination, if one exists, so
// CFGSort cannot mess up with BB sorting order.
// (All catchpads with 'catch (type)' clauses have a 'llvm.rethrow' intrinsic
// call within them, and catchpads only consisting of 'catch (...)' have a
// '__cxa_end_catch' call within them, both of which generate invokes in case
// the next unwind destination exists, i.e., the next unwind destination is not
// the caller.)
//
// Having at most one EH pad successor is also simpler and helps later
// transformations.
//
// For example,
// current:
//   invoke void @foo to ... unwind label %catch.dispatch
// catch.dispatch:
//   %0 = catchswitch within ... [label %catch.start] unwind label %next
// catch.start:
//   ...
//   ... in this BB or some other child BB dominated by this BB there will be an
//   invoke that points to 'next' BB as an unwind destination
//
// next: ; We don't need to add this to 'current' BB's successor
//   ...
static void findWasmUnwindDestinations(
    FunctionLoweringInfo &FuncInfo, const BasicBlock *EHPadBB,
    BranchProbability Prob,
    SmallVectorImpl<std::pair<MachineBasicBlock *, BranchProbability>>
        &UnwindDests) {
  while (EHPadBB) {
    const Instruction *Pad = EHPadBB->getFirstNonPHI();
    if (isa<CleanupPadInst>(Pad)) {
      // Stop on cleanup pads.
      UnwindDests.emplace_back(FuncInfo.MBBMap[EHPadBB], Prob);
      UnwindDests.back().first->setIsEHScopeEntry();
      break;
    } else if (auto *CatchSwitch = dyn_cast<CatchSwitchInst>(Pad)) {
      // Add the catchpad handlers to the possible destinations. We don't
      // continue to the unwind destination of the catchswitch for wasm.
      for (const BasicBlock *CatchPadBB : CatchSwitch->handlers()) {
        UnwindDests.emplace_back(FuncInfo.MBBMap[CatchPadBB], Prob);
        UnwindDests.back().first->setIsEHScopeEntry();
      }
      break;
    } else {
      continue;
    }
  }
}
 
/// When an invoke or a cleanupret unwinds to the next EH pad, there are
/// many places it could ultimately go. In the IR, we have a single unwind
/// destination, but in the machine CFG, we enumerate all the possible blocks.
/// This function skips over imaginary basic blocks that hold catchswitch
/// instructions, and finds all the "real" machine
/// basic block destinations. As those destinations may not be successors of
/// EHPadBB, here we also calculate the edge probability to those destinations.
/// The passed-in Prob is the edge probability to EHPadBB.
static void findUnwindDestinations(
    FunctionLoweringInfo &FuncInfo, const BasicBlock *EHPadBB,
    BranchProbability Prob,
    SmallVectorImpl<std::pair<MachineBasicBlock *, BranchProbability>>
        &UnwindDests) {
  EHPersonality Personality =
    classifyEHPersonality(FuncInfo.Fn->getPersonalityFn());
  bool IsMSVCCXX = Personality == EHPersonality::MSVC_CXX;
  bool IsCoreCLR = Personality == EHPersonality::CoreCLR;
  bool IsWasmCXX = Personality == EHPersonality::Wasm_CXX;
  bool IsSEH = isAsynchronousEHPersonality(Personality);
 
  if (IsWasmCXX) {
    findWasmUnwindDestinations(FuncInfo, EHPadBB, Prob, UnwindDests);
    assert(UnwindDests.size() <= 1 &&
           "There should be at most one unwind destination for wasm");
    return;
  }
 
  while (EHPadBB) {
    const Instruction *Pad = EHPadBB->getFirstNonPHI();
    BasicBlock *NewEHPadBB = nullptr;
    if (isa<LandingPadInst>(Pad)) {
      // Stop on landingpads. They are not funclets.
      UnwindDests.emplace_back(FuncInfo.MBBMap[EHPadBB], Prob);
      break;
    } else if (isa<CleanupPadInst>(Pad)) {
      // Stop on cleanup pads. Cleanups are always funclet entries for all known
      // personalities.
      UnwindDests.emplace_back(FuncInfo.MBBMap[EHPadBB], Prob);
      UnwindDests.back().first->setIsEHScopeEntry();
      UnwindDests.back().first->setIsEHFuncletEntry();
      break;
    } else if (auto *CatchSwitch = dyn_cast<CatchSwitchInst>(Pad)) {
      // Add the catchpad handlers to the possible destinations.
      for (const BasicBlock *CatchPadBB : CatchSwitch->handlers()) {
        UnwindDests.emplace_back(FuncInfo.MBBMap[CatchPadBB], Prob);
        // For MSVC++ and the CLR, catchblocks are funclets and need prologues.
        if (IsMSVCCXX || IsCoreCLR)
          UnwindDests.back().first->setIsEHFuncletEntry();
        if (!IsSEH)
          UnwindDests.back().first->setIsEHScopeEntry();
      }
      NewEHPadBB = CatchSwitch->getUnwindDest();
    } else {
      continue;
    }
 
    BranchProbabilityInfo *BPI = FuncInfo.BPI;
    if (BPI && NewEHPadBB)
      Prob *= BPI->getEdgeProbability(EHPadBB, NewEHPadBB);
    EHPadBB = NewEHPadBB;
  }
}
 
void SelectionDAGBuilder::visitCleanupRet(const CleanupReturnInst &I) {
  // Update successor info.
  SmallVector<std::pair<MachineBasicBlock *, BranchProbability>, 1> UnwindDests;
  auto UnwindDest = I.getUnwindDest();
  BranchProbabilityInfo *BPI = FuncInfo.BPI;
  BranchProbability UnwindDestProb =
      (BPI && UnwindDest)
          ? BPI->getEdgeProbability(FuncInfo.MBB->getBasicBlock(), UnwindDest)
          : BranchProbability::getZero();
  findUnwindDestinations(FuncInfo, UnwindDest, UnwindDestProb, UnwindDests);
  for (auto &UnwindDest : UnwindDests) {
    UnwindDest.first->setIsEHPad();
    addSuccessorWithProb(FuncInfo.MBB, UnwindDest.first, UnwindDest.second);
  }
  FuncInfo.MBB->normalizeSuccProbs();
 
  // Create the terminator node.
  SDValue Ret =
      DAG.getNode(ISD::CLEANUPRET, getCurSDLoc(), MVT::Other, getControlRoot());
  DAG.setRoot(Ret);
}
 
void SelectionDAGBuilder::visitCatchSwitch(const CatchSwitchInst &CSI) {
  report_fatal_error("visitCatchSwitch not yet implemented!");
}
 
void SelectionDAGBuilder::visitRet(const ReturnInst &I) {
  const TargetLowering &TLI = DAG.getTargetLoweringInfo();
  auto &DL = DAG.getDataLayout();
  SDValue Chain = getControlRoot();
  SmallVector<ISD::OutputArg, 8> Outs;
  SmallVector<SDValue, 8> OutVals;
 
  // Calls to @llvm.experimental.deoptimize don't generate a return value, so
  // lower
  //
  //   %val = call <ty> @llvm.experimental.deoptimize()
  //   ret <ty> %val
  //
  // differently.
  if (I.getParent()->getTerminatingDeoptimizeCall()) {
    LowerDeoptimizingReturn();
    return;
  }
 
  if (!FuncInfo.CanLowerReturn) {
    unsigned DemoteReg = FuncInfo.DemoteRegister;
    const Function *F = I.getParent()->getParent();
 
    // Emit a store of the return value through the virtual register.
    // Leave Outs empty so that LowerReturn won't try to load return
    // registers the usual way.
    SmallVector<EVT, 1> PtrValueVTs;
    ComputeValueVTs(TLI, DL,
                    F->getReturnType()->getPointerTo(
                        DAG.getDataLayout().getAllocaAddrSpace()),
                    PtrValueVTs);
 
    SDValue RetPtr = DAG.getCopyFromReg(DAG.getEntryNode(), getCurSDLoc(),
                                        DemoteReg, PtrValueVTs[0]);
    SDValue RetOp = getValue(I.getOperand(0));
 
    SmallVector<EVT, 4> ValueVTs, MemVTs;
    SmallVector<uint64_t, 4> Offsets;
    ComputeValueVTs(TLI, DL, I.getOperand(0)->getType(), ValueVTs, &MemVTs,
                    &Offsets);
    unsigned NumValues = ValueVTs.size();
 
    SmallVector<SDValue, 4> Chains(NumValues);
    Align BaseAlign = DL.getPrefTypeAlign(I.getOperand(0)->getType());
    for (unsigned i = 0; i != NumValues; ++i) {
      // An aggregate return value cannot wrap around the address space, so
      // offsets to its parts don't wrap either.
      SDValue Ptr = DAG.getObjectPtrOffset(getCurSDLoc(), RetPtr,
                                           TypeSize::Fixed(Offsets[i]));
 
      SDValue Val = RetOp.getValue(RetOp.getResNo() + i);
      if (MemVTs[i] != ValueVTs[i])
        Val = DAG.getPtrExtOrTrunc(Val, getCurSDLoc(), MemVTs[i]);
      Chains[i] = DAG.getStore(
          Chain, getCurSDLoc(), Val,
          // FIXME: better loc info would be nice.
          Ptr, MachinePointerInfo::getUnknownStack(DAG.getMachineFunction()),
          commonAlignment(BaseAlign, Offsets[i]));
    }
 
    Chain = DAG.getNode(ISD::TokenFactor, getCurSDLoc(),
                        MVT::Other, Chains);
  } else if (I.getNumOperands() != 0) {
    SmallVector<EVT, 4> ValueVTs;
    ComputeValueVTs(TLI, DL, I.getOperand(0)->getType(), ValueVTs);
    unsigned NumValues = ValueVTs.size();
    if (NumValues) {
      SDValue RetOp = getValue(I.getOperand(0));
 
      const Function *F = I.getParent()->getParent();
 
      bool NeedsRegBlock = TLI.functionArgumentNeedsConsecutiveRegisters(
          I.getOperand(0)->getType(), F->getCallingConv(),
          /*IsVarArg*/ false, DL);
 
      ISD::NodeType ExtendKind = ISD::ANY_EXTEND;
      if (F->getAttributes().hasRetAttr(Attribute::SExt))
        ExtendKind = ISD::SIGN_EXTEND;
      else if (F->getAttributes().hasRetAttr(Attribute::ZExt))
        ExtendKind = ISD::ZERO_EXTEND;
 
      LLVMContext &Context = F->getContext();
      bool RetInReg = F->getAttributes().hasRetAttr(Attribute::InReg);
 
      for (unsigned j = 0; j != NumValues; ++j) {
        EVT VT = ValueVTs[j];
 
        if (ExtendKind != ISD::ANY_EXTEND && VT.isInteger())
          VT = TLI.getTypeForExtReturn(Context, VT, ExtendKind);
 
        CallingConv::ID CC = F->getCallingConv();
 
        unsigned NumParts = TLI.getNumRegistersForCallingConv(Context, CC, VT);
        MVT PartVT = TLI.getRegisterTypeForCallingConv(Context, CC, VT);
        SmallVector<SDValue, 4> Parts(NumParts);
        getCopyToParts(DAG, getCurSDLoc(),
                       SDValue(RetOp.getNode(), RetOp.getResNo() + j),
                       &Parts[0], NumParts, PartVT, &I, CC, ExtendKind);
 
        // 'inreg' on function refers to return value
        ISD::ArgFlagsTy Flags = ISD::ArgFlagsTy();
        if (RetInReg)
          Flags.setInReg();
 
        if (I.getOperand(0)->getType()->isPointerTy()) {
          Flags.setPointer();
          Flags.setPointerAddrSpace(
              cast<PointerType>(I.getOperand(0)->getType())->getAddressSpace());
        }
 
        if (NeedsRegBlock) {
          Flags.setInConsecutiveRegs();
          if (j == NumValues - 1)
            Flags.setInConsecutiveRegsLast();
        }
 
        // Propagate extension type if any
        if (ExtendKind == ISD::SIGN_EXTEND)
          Flags.setSExt();
        else if (ExtendKind == ISD::ZERO_EXTEND)
          Flags.setZExt();
 
        for (unsigned i = 0; i < NumParts; ++i) {
          Outs.push_back(ISD::OutputArg(Flags,
                                        Parts[i].getValueType().getSimpleVT(),
                                        VT, /*isfixed=*/true, 0, 0));
          OutVals.push_back(Parts[i]);
        }
      }
    }
  }
 
  // Push in swifterror virtual register as the last element of Outs. This makes
  // sure swifterror virtual register will be returned in the swifterror
  // physical register.
  const Function *F = I.getParent()->getParent();
  if (TLI.supportSwiftError() &&
      F->getAttributes().hasAttrSomewhere(Attribute::SwiftError)) {
    assert(SwiftError.getFunctionArg() && "Need a swift error argument");
    ISD::ArgFlagsTy Flags = ISD::ArgFlagsTy();
    Flags.setSwiftError();
    Outs.push_back(ISD::OutputArg(
        Flags, /*vt=*/TLI.getPointerTy(DL), /*argvt=*/EVT(TLI.getPointerTy(DL)),
        /*isfixed=*/true, /*origidx=*/1, /*partOffs=*/0));
    // Create SDNode for the swifterror virtual register.
    OutVals.push_back(
        DAG.getRegister(SwiftError.getOrCreateVRegUseAt(
                            &I, FuncInfo.MBB, SwiftError.getFunctionArg()),
                        EVT(TLI.getPointerTy(DL))));
  }
 
  bool isVarArg = DAG.getMachineFunction().getFunction().isVarArg();
  CallingConv::ID CallConv =
    DAG.getMachineFunction().getFunction().getCallingConv();
  Chain = DAG.getTargetLoweringInfo().LowerReturn(
      Chain, CallConv, isVarArg, Outs, OutVals, getCurSDLoc(), DAG);
 
  // Verify that the target's LowerReturn behaved as expected.
  assert(Chain.getNode() && Chain.getValueType() == MVT::Other &&
         "LowerReturn didn't return a valid chain!");
 
  // Update the DAG with the new chain value resulting from return lowering.
  DAG.setRoot(Chain);
}
 
/// CopyToExportRegsIfNeeded - If the given value has virtual registers
/// created for it, emit nodes to copy the value into the virtual
/// registers.
void SelectionDAGBuilder::CopyToExportRegsIfNeeded(const Value *V) {
  // Skip empty types
  if (V->getType()->isEmptyTy())
    return;
 
  DenseMap<const Value *, Register>::iterator VMI = FuncInfo.ValueMap.find(V);
  if (VMI != FuncInfo.ValueMap.end()) {
    assert(!V->use_empty() && "Unused value assigned virtual registers!");
    CopyValueToVirtualRegister(V, VMI->second);
  }
}
 
/// ExportFromCurrentBlock - If this condition isn't known to be exported from
/// the current basic block, add it to ValueMap now so that we'll get a
/// CopyTo/FromReg.
void SelectionDAGBuilder::ExportFromCurrentBlock(const Value *V) {
  // No need to export constants.
  if (!isa<Instruction>(V) && !isa<Argument>(V)) return;
 
  // Already exported?
  if (FuncInfo.isExportedInst(V)) return;
 
  unsigned Reg = FuncInfo.InitializeRegForValue(V);
  CopyValueToVirtualRegister(V, Reg);
}
 
bool SelectionDAGBuilder::isExportableFromCurrentBlock(const Value *V,
                                                     const BasicBlock *FromBB) {
  // The operands of the setcc have to be in this block.  We don't know
  // how to export them from some other block.
  if (const Instruction *VI = dyn_cast<Instruction>(V)) {
    // Can export from current BB.
    if (VI->getParent() == FromBB)
      return true;
 
    // Is already exported, noop.
    return FuncInfo.isExportedInst(V);
  }
 
  // If this is an argument, we can export it if the BB is the entry block or
  // if it is already exported.
  if (isa<Argument>(V)) {
    if (FromBB->isEntryBlock())
      return true;
 
    // Otherwise, can only export this if it is already exported.
    return FuncInfo.isExportedInst(V);
  }
 
  // Otherwise, constants can always be exported.
  return true;
}
 
/// Return branch probability calculated by BranchProbabilityInfo for IR blocks.
BranchProbability
SelectionDAGBuilder::getEdgeProbability(const MachineBasicBlock *Src,
                                        const MachineBasicBlock *Dst) const {
  BranchProbabilityInfo *BPI = FuncInfo.BPI;
  const BasicBlock *SrcBB = Src->getBasicBlock();
  const BasicBlock *DstBB = Dst->getBasicBlock();
  if (!BPI) {
    // If BPI is not available, set the default probability as 1 / N, where N is
    // the number of successors.
    auto SuccSize = std::max<uint32_t>(succ_size(SrcBB), 1);
    return BranchProbability(1, SuccSize);
  }
  return BPI->getEdgeProbability(SrcBB, DstBB);
}
 
void SelectionDAGBuilder::addSuccessorWithProb(MachineBasicBlock *Src,
                                               MachineBasicBlock *Dst,
                                               BranchProbability Prob) {
  if (!FuncInfo.BPI)
    Src->addSuccessorWithoutProb(Dst);
  else {
    if (Prob.isUnknown())
      Prob = getEdgeProbability(Src, Dst);
    Src->addSuccessor(Dst, Prob);
  }
}
 
static bool InBlock(const Value *V, const BasicBlock *BB) {
  if (const Instruction *I = dyn_cast<Instruction>(V))
    return I->getParent() == BB;
  return true;
}
 
/// EmitBranchForMergedCondition - Helper method for FindMergedConditions.
/// This function emits a branch and is used at the leaves of an OR or an
/// AND operator tree.
void
SelectionDAGBuilder::EmitBranchForMergedCondition(const Value *Cond,
                                                  MachineBasicBlock *TBB,
                                                  MachineBasicBlock *FBB,
                                                  MachineBasicBlock *CurBB,
                                                  MachineBasicBlock *SwitchBB,
                                                  BranchProbability TProb,
                                                  BranchProbability FProb,
                                                  bool InvertCond) {
  const BasicBlock *BB = CurBB->getBasicBlock();
 
  // If the leaf of the tree is a comparison, merge the condition into
  // the caseblock.
  if (const CmpInst *BOp = dyn_cast<CmpInst>(Cond)) {
    // The operands of the cmp have to be in this block.  We don't know
    // how to export them from some other block.  If this is the first block
    // of the sequence, no exporting is needed.
    if (CurBB == SwitchBB ||
        (isExportableFromCurrentBlock(BOp->getOperand(0), BB) &&
         isExportableFromCurrentBlock(BOp->getOperand(1), BB))) {
      ISD::CondCode Condition;
      if (const ICmpInst *IC = dyn_cast<ICmpInst>(Cond)) {
        ICmpInst::Predicate Pred =
            InvertCond ? IC->getInversePredicate() : IC->getPredicate();
        Condition = getICmpCondCode(Pred);
      } else {
        const FCmpInst *FC = cast<FCmpInst>(Cond);
        FCmpInst::Predicate Pred =
            InvertCond ? FC->getInversePredicate() : FC->getPredicate();
        Condition = getFCmpCondCode(Pred);
        if (TM.Options.NoNaNsFPMath)
          Condition = getFCmpCodeWithoutNaN(Condition);
      }
 
      CaseBlock CB(Condition, BOp->getOperand(0), BOp->getOperand(1), nullptr,
                   TBB, FBB, CurBB, getCurSDLoc(), TProb, FProb);
      SL->SwitchCases.push_back(CB);
      return;
    }
  }
 
  // Create a CaseBlock record representing this branch.
  ISD::CondCode Opc = InvertCond ? ISD::SETNE : ISD::SETEQ;
  CaseBlock CB(Opc, Cond, ConstantInt::getTrue(*DAG.getContext()),
               nullptr, TBB, FBB, CurBB, getCurSDLoc(), TProb, FProb);
  SL->SwitchCases.push_back(CB);
}
 
void SelectionDAGBuilder::FindMergedConditions(const Value *Cond,
                                               MachineBasicBlock *TBB,
                                               MachineBasicBlock *FBB,
                                               MachineBasicBlock *CurBB,
                                               MachineBasicBlock *SwitchBB,
                                               Instruction::BinaryOps Opc,
                                               BranchProbability TProb,
                                               BranchProbability FProb,
                                               bool InvertCond) {
  // Skip over not part of the tree and remember to invert op and operands at
  // next level.
  Value *NotCond;
  if (match(Cond, m_OneUse(m_Not(m_Value(NotCond)))) &&
      InBlock(NotCond, CurBB->getBasicBlock())) {
    FindMergedConditions(NotCond, TBB, FBB, CurBB, SwitchBB, Opc, TProb, FProb,
                         !InvertCond);
    return;
  }
 
  const Instruction *BOp = dyn_cast<Instruction>(Cond);
  const Value *BOpOp0, *BOpOp1;
  // Compute the effective opcode for Cond, taking into account whether it needs
  // to be inverted, e.g.
  //   and (not (or A, B)), C
  // gets lowered as
  //   and (and (not A, not B), C)
  Instruction::BinaryOps BOpc = (Instruction::BinaryOps)0;
  if (BOp) {
    BOpc = match(BOp, m_LogicalAnd(m_Value(BOpOp0), m_Value(BOpOp1)))
               ? Instruction::And
               : (match(BOp, m_LogicalOr(m_Value(BOpOp0), m_Value(BOpOp1)))
                      ? Instruction::Or
                      : (Instruction::BinaryOps)0);
    if (InvertCond) {
      if (BOpc == Instruction::And)
        BOpc = Instruction::Or;
      else if (BOpc == Instruction::Or)
        BOpc = Instruction::And;
    }
  }
 
  // If this node is not part of the or/and tree, emit it as a branch.
  // Note that all nodes in the tree should have same opcode.
  bool BOpIsInOrAndTree = BOpc && BOpc == Opc && BOp->hasOneUse();
  if (!BOpIsInOrAndTree || BOp->getParent() != CurBB->getBasicBlock() ||
      !InBlock(BOpOp0, CurBB->getBasicBlock()) ||
      !InBlock(BOpOp1, CurBB->getBasicBlock())) {
    EmitBranchForMergedCondition(Cond, TBB, FBB, CurBB, SwitchBB,
                                 TProb, FProb, InvertCond);
    return;
  }
 
  //  Create TmpBB after CurBB.
  MachineFunction::iterator BBI(CurBB);
  MachineFunction &MF = DAG.getMachineFunction();
  MachineBasicBlock *TmpBB = MF.CreateMachineBasicBlock(CurBB->getBasicBlock());
  CurBB->getParent()->insert(++BBI, TmpBB);
 
  if (Opc == Instruction::Or) {
    // Codegen X | Y as:
    // BB1:
    //   jmp_if_X TBB
    //   jmp TmpBB
    // TmpBB:
    //   jmp_if_Y TBB
    //   jmp FBB
    //
 
    // We have flexibility in setting Prob for BB1 and Prob for TmpBB.
    // The requirement is that
    //   TrueProb for BB1 + (FalseProb for BB1 * TrueProb for TmpBB)
    //     = TrueProb for original BB.
    // Assuming the original probabilities are A and B, one choice is to set
    // BB1's probabilities to A/2 and A/2+B, and set TmpBB's probabilities to
    // A/(1+B) and 2B/(1+B). This choice assumes that
    //   TrueProb for BB1 == FalseProb for BB1 * TrueProb for TmpBB.
    // Another choice is to assume TrueProb for BB1 equals to TrueProb for
    // TmpBB, but the math is more complicated.
 
    auto NewTrueProb = TProb / 2;
    auto NewFalseProb = TProb / 2 + FProb;
    // Emit the LHS condition.
    FindMergedConditions(BOpOp0, TBB, TmpBB, CurBB, SwitchBB, Opc, NewTrueProb,
                         NewFalseProb, InvertCond);
 
    // Normalize A/2 and B to get A/(1+B) and 2B/(1+B).
    SmallVector<BranchProbability, 2> Probs{TProb / 2, FProb};
    BranchProbability::normalizeProbabilities(Probs.begin(), Probs.end());
    // Emit the RHS condition into TmpBB.
    FindMergedConditions(BOpOp1, TBB, FBB, TmpBB, SwitchBB, Opc, Probs[0],
                         Probs[1], InvertCond);
  } else {
    assert(Opc == Instruction::And && "Unknown merge op!");
    // Codegen X & Y as:
    // BB1:
    //   jmp_if_X TmpBB
    //   jmp FBB
    // TmpBB:
    //   jmp_if_Y TBB
    //   jmp FBB
    //
    //  This requires creation of TmpBB after CurBB.
 
    // We have flexibility in setting Prob for BB1 and Prob for TmpBB.
    // The requirement is that
    //   FalseProb for BB1 + (TrueProb for BB1 * FalseProb for TmpBB)
    //     = FalseProb for original BB.
    // Assuming the original probabilities are A and B, one choice is to set
    // BB1's probabilities to A+B/2 and B/2, and set TmpBB's probabilities to
    // 2A/(1+A) and B/(1+A). This choice assumes that FalseProb for BB1 ==
    // TrueProb for BB1 * FalseProb for TmpBB.
 
    auto NewTrueProb = TProb + FProb / 2;
    auto NewFalseProb = FProb / 2;
    // Emit the LHS condition.
    FindMergedConditions(BOpOp0, TmpBB, FBB, CurBB, SwitchBB, Opc, NewTrueProb,
                         NewFalseProb, InvertCond);
 
    // Normalize A and B/2 to get 2A/(1+A) and B/(1+A).
    SmallVector<BranchProbability, 2> Probs{TProb, FProb / 2};
    BranchProbability::normalizeProbabilities(Probs.begin(), Probs.end());
    // Emit the RHS condition into TmpBB.
    FindMergedConditions(BOpOp1, TBB, FBB, TmpBB, SwitchBB, Opc, Probs[0],
                         Probs[1], InvertCond);
  }
}
 
/// If the set of cases should be emitted as a series of branches, return true.
/// If we should emit this as a bunch of and/or'd together conditions, return
/// false.
bool
SelectionDAGBuilder::ShouldEmitAsBranches(const std::vector<CaseBlock> &Cases) {
  if (Cases.size() != 2) return true;
 
  // If this is two comparisons of the same values or'd or and'd together, they
  // will get folded into a single comparison, so don't emit two blocks.
  if ((Cases[0].CmpLHS == Cases[1].CmpLHS &&
       Cases[0].CmpRHS == Cases[1].CmpRHS) ||
      (Cases[0].CmpRHS == Cases[1].CmpLHS &&
       Cases[0].CmpLHS == Cases[1].CmpRHS)) {
    return false;
  }
 
  // Handle: (X != null) | (Y != null) --> (X|Y) != 0
  // Handle: (X == null) & (Y == null) --> (X|Y) == 0
  if (Cases[0].CmpRHS == Cases[1].CmpRHS &&
      Cases[0].CC == Cases[1].CC &&
      isa<Constant>(Cases[0].CmpRHS) &&
      cast<Constant>(Cases[0].CmpRHS)->isNullValue()) {
    if (Cases[0].CC == ISD::SETEQ && Cases[0].TrueBB == Cases[1].ThisBB)
      return false;
    if (Cases[0].CC == ISD::SETNE && Cases[0].FalseBB == Cases[1].ThisBB)
      return false;
  }
 
  return true;
}
 
void SelectionDAGBuilder::visitBr(const BranchInst &I) {
  MachineBasicBlock *BrMBB = FuncInfo.MBB;
 
  // Update machine-CFG edges.
  MachineBasicBlock *Succ0MBB = FuncInfo.MBBMap[I.getSuccessor(0)];
 
  if (I.isUnconditional()) {
    // Update machine-CFG edges.
    BrMBB->addSuccessor(Succ0MBB);
 
    // If this is not a fall-through branch or optimizations are switched off,
    // emit the branch.
    if (Succ0MBB != NextBlock(BrMBB) || TM.getOptLevel() == CodeGenOpt::None)
      DAG.setRoot(DAG.getNode(ISD::BR, getCurSDLoc(),
                              MVT::Other, getControlRoot(),
                              DAG.getBasicBlock(Succ0MBB)));
 
    return;
  }
 
  // If this condition is one of the special cases we handle, do special stuff
  // now.
  const Value *CondVal = I.getCondition();
  MachineBasicBlock *Succ1MBB = FuncInfo.MBBMap[I.getSuccessor(1)];
 
  // If this is a series of conditions that are or'd or and'd together, emit
  // this as a sequence of branches instead of setcc's with and/or operations.
  // As long as jumps are not expensive (exceptions for multi-use logic ops,
  // unpredictable branches, and vector extracts because those jumps are likely
  // expensive for any target), this should improve performance.
  // For example, instead of something like:
  //     cmp A, B
  //     C = seteq
  //     cmp D, E
  //     F = setle
  //     or C, F
  //     jnz foo
  // Emit:
  //     cmp A, B
  //     je foo
  //     cmp D, E
  //     jle foo
  const Instruction *BOp = dyn_cast<Instruction>(CondVal);
  if (!DAG.getTargetLoweringInfo().isJumpExpensive() && BOp &&
      BOp->hasOneUse() && !I.hasMetadata(LLVMContext::MD_unpredictable)) {
    Value *Vec;
    const Value *BOp0, *BOp1;
    Instruction::BinaryOps Opcode = (Instruction::BinaryOps)0;
    if (match(BOp, m_LogicalAnd(m_Value(BOp0), m_Value(BOp1))))
      Opcode = Instruction::And;
    else if (match(BOp, m_LogicalOr(m_Value(BOp0), m_Value(BOp1))))
      Opcode = Instruction::Or;
 
    if (Opcode && !(match(BOp0, m_ExtractElt(m_Value(Vec), m_Value())) &&
                    match(BOp1, m_ExtractElt(m_Specific(Vec), m_Value())))) {
      FindMergedConditions(BOp, Succ0MBB, Succ1MBB, BrMBB, BrMBB, Opcode,
                           getEdgeProbability(BrMBB, Succ0MBB),
                           getEdgeProbability(BrMBB, Succ1MBB),
                           /*InvertCond=*/false);
      // If the compares in later blocks need to use values not currently
      // exported from this block, export them now.  This block should always
      // be the first entry.
      assert(SL->SwitchCases[0].ThisBB == BrMBB && "Unexpected lowering!");
 
      // Allow some cases to be rejected.
      if (ShouldEmitAsBranches(SL->SwitchCases)) {
        for (unsigned i = 1, e = SL->SwitchCases.size(); i != e; ++i) {
          ExportFromCurrentBlock(SL->SwitchCases[i].CmpLHS);
          ExportFromCurrentBlock(SL->SwitchCases[i].CmpRHS);
        }
 
        // Emit the branch for this block.
        visitSwitchCase(SL->SwitchCases[0], BrMBB);
        SL->SwitchCases.erase(SL->SwitchCases.begin());
        return;
      }
 
      // Okay, we decided not to do this, remove any inserted MBB's and clear
      // SwitchCases.
      for (unsigned i = 1, e = SL->SwitchCases.size(); i != e; ++i)
        FuncInfo.MF->erase(SL->SwitchCases[i].ThisBB);
 
      SL->SwitchCases.clear();
    }
  }
 
  // Create a CaseBlock record representing this branch.
  CaseBlock CB(ISD::SETEQ, CondVal, ConstantInt::getTrue(*DAG.getContext()),
               nullptr, Succ0MBB, Succ1MBB, BrMBB, getCurSDLoc());
 
  // Use visitSwitchCase to actually insert the fast branch sequence for this
  // cond branch.
  visitSwitchCase(CB, BrMBB);
}
 
/// visitSwitchCase - Emits the necessary code to represent a single node in
/// the binary search tree resulting from lowering a switch instruction.
void SelectionDAGBuilder::visitSwitchCase(CaseBlock &CB,
                                          MachineBasicBlock *SwitchBB) {
  SDValue Cond;
  SDValue CondLHS = getValue(CB.CmpLHS);
  SDLoc dl = CB.DL;
 
  if (CB.CC == ISD::SETTRUE) {
    // Branch or fall through to TrueBB.
    addSuccessorWithProb(SwitchBB, CB.TrueBB, CB.TrueProb);
    SwitchBB->normalizeSuccProbs();
    if (CB.TrueBB != NextBlock(SwitchBB)) {
      DAG.setRoot(DAG.getNode(ISD::BR, dl, MVT::Other, getControlRoot(),
                              DAG.getBasicBlock(CB.TrueBB)));
    }
    return;
  }
 
  auto &TLI = DAG.getTargetLoweringInfo();
  EVT MemVT = TLI.getMemValueType(DAG.getDataLayout(), CB.CmpLHS->getType());
 
  // Build the setcc now.
  if (!CB.CmpMHS) {
    // Fold "(X == true)" to X and "(X == false)" to !X to
    // handle common cases produced by branch lowering.
    if (CB.CmpRHS == ConstantInt::getTrue(*DAG.getContext()) &&
        CB.CC == ISD::SETEQ)
      Cond = CondLHS;
    else if (CB.CmpRHS == ConstantInt::getFalse(*DAG.getContext()) &&
             CB.CC == ISD::SETEQ) {
      SDValue True = DAG.getConstant(1, dl, CondLHS.getValueType());
      Cond = DAG.getNode(ISD::XOR, dl, CondLHS.getValueType(), CondLHS, True);
    } else {
      SDValue CondRHS = getValue(CB.CmpRHS);
 
      // If a pointer's DAG type is larger than its memory type then the DAG
      // values are zero-extended. This breaks signed comparisons so truncate
      // back to the underlying type before doing the compare.
      if (CondLHS.getValueType() != MemVT) {
        CondLHS = DAG.getPtrExtOrTrunc(CondLHS, getCurSDLoc(), MemVT);
        CondRHS = DAG.getPtrExtOrTrunc(CondRHS, getCurSDLoc(), MemVT);
      }
      Cond = DAG.getSetCC(dl, MVT::i1, CondLHS, CondRHS, CB.CC);
    }
  } else {
    assert(CB.CC == ISD::SETLE && "Can handle only LE ranges now");
 
    const APInt& Low = cast<ConstantInt>(CB.CmpLHS)->getValue();
    const APInt& High = cast<ConstantInt>(CB.CmpRHS)->getValue();
 
    SDValue CmpOp = getValue(CB.CmpMHS);
    EVT VT = CmpOp.getValueType();
 
    if (cast<ConstantInt>(CB.CmpLHS)->isMinValue(true)) {
      Cond = DAG.getSetCC(dl, MVT::i1, CmpOp, DAG.getConstant(High, dl, VT),
                          ISD::SETLE);
    } else {
      SDValue SUB = DAG.getNode(ISD::SUB, dl,
                                VT, CmpOp, DAG.getConstant(Low, dl, VT));
      Cond = DAG.getSetCC(dl, MVT::i1, SUB,
                          DAG.getConstant(High-Low, dl, VT), ISD::SETULE);
    }
  }
 
  // Update successor info
  addSuccessorWithProb(SwitchBB, CB.TrueBB, CB.TrueProb);
  // TrueBB and FalseBB are always different unless the incoming IR is
  // degenerate. This only happens when running llc on weird IR.
  if (CB.TrueBB != CB.FalseBB)
    addSuccessorWithProb(SwitchBB, CB.FalseBB, CB.FalseProb);
  SwitchBB->normalizeSuccProbs();
 
  // If the lhs block is the next block, invert the condition so that we can
  // fall through to the lhs instead of the rhs block.
  if (CB.TrueBB == NextBlock(SwitchBB)) {
    std::swap(CB.TrueBB, CB.FalseBB);
    SDValue True = DAG.getConstant(1, dl, Cond.getValueType());
    Cond = DAG.getNode(ISD::XOR, dl, Cond.getValueType(), Cond, True);
  }
 
  SDValue BrCond = DAG.getNode(ISD::BRCOND, dl,
                               MVT::Other, getControlRoot(), Cond,
                               DAG.getBasicBlock(CB.TrueBB));
 
  // Insert the false branch. Do this even if it's a fall through branch,
  // this makes it easier to do DAG optimizations which require inverting
  // the branch condition.
  BrCond = DAG.getNode(ISD::BR, dl, MVT::Other, BrCond,
                       DAG.getBasicBlock(CB.FalseBB));
 
  DAG.setRoot(BrCond);
}
 
/// visitJumpTable - Emit JumpTable node in the current MBB
void SelectionDAGBuilder::visitJumpTable(SwitchCG::JumpTable &JT) {
  // Emit the code for the jump table
  assert(JT.Reg != -1U && "Should lower JT Header first!");
  EVT PTy = DAG.getTargetLoweringInfo().getPointerTy(DAG.getDataLayout());
  SDValue Index = DAG.getCopyFromReg(getControlRoot(), getCurSDLoc(),
                                     JT.Reg, PTy);
  SDValue Table = DAG.getJumpTable(JT.JTI, PTy);
  SDValue BrJumpTable = DAG.getNode(ISD::BR_JT, getCurSDLoc(),
                                    MVT::Other, Index.getValue(1),
                                    Table, Index);
  DAG.setRoot(BrJumpTable);
}
 
/// visitJumpTableHeader - This function emits necessary code to produce index
/// in the JumpTable from switch case.
void SelectionDAGBuilder::visitJumpTableHeader(SwitchCG::JumpTable &JT,
                                               JumpTableHeader &JTH,
                                               MachineBasicBlock *SwitchBB) {
  SDLoc dl = getCurSDLoc();
 
  // Subtract the lowest switch case value from the value being switched on.
  SDValue SwitchOp = getValue(JTH.SValue);
  EVT VT = SwitchOp.getValueType();
  SDValue Sub = DAG.getNode(ISD::SUB, dl, VT, SwitchOp,
                            DAG.getConstant(JTH.First, dl, VT));
 
  // The SDNode we just created, which holds the value being switched on minus
  // the smallest case value, needs to be copied to a virtual register so it
  // can be used as an index into the jump table in a subsequent basic block.
  // This value may be smaller or larger than the target's pointer type, and
  // therefore require extension or truncating.
  const TargetLowering &TLI = DAG.getTargetLoweringInfo();
  SwitchOp = DAG.getZExtOrTrunc(Sub, dl, TLI.getPointerTy(DAG.getDataLayout()));
 
  unsigned JumpTableReg =
      FuncInfo.CreateReg(TLI.getPointerTy(DAG.getDataLayout()));
  SDValue CopyTo = DAG.getCopyToReg(getControlRoot(), dl,
                                    JumpTableReg, SwitchOp);
  JT.Reg = JumpTableReg;
 
  if (!JTH.FallthroughUnreachable) {
    // Emit the range check for the jump table, and branch to the default block
    // for the switch statement if the value being switched on exceeds the
    // largest case in the switch.
    SDValue CMP = DAG.getSetCC(
        dl, TLI.getSetCCResultType(DAG.getDataLayout(), *DAG.getContext(),
                                   Sub.getValueType()),
        Sub, DAG.getConstant(JTH.Last - JTH.First, dl, VT), ISD::SETUGT);
 
    SDValue BrCond = DAG.getNode(ISD::BRCOND, dl,
                                 MVT::Other, CopyTo, CMP,
                                 DAG.getBasicBlock(JT.Default));
 
    // Avoid emitting unnecessary branches to the next block.
    if (JT.MBB != NextBlock(SwitchBB))
      BrCond = DAG.getNode(ISD::BR, dl, MVT::Other, BrCond,
                           DAG.getBasicBlock(JT.MBB));
 
    DAG.setRoot(BrCond);
  } else {
    // Avoid emitting unnecessary branches to the next block.
    if (JT.MBB != NextBlock(SwitchBB))
      DAG.setRoot(DAG.getNode(ISD::BR, dl, MVT::Other, CopyTo,
                              DAG.getBasicBlock(JT.MBB)));
    else
      DAG.setRoot(CopyTo);
  }
}
 
/// Create a LOAD_STACK_GUARD node, and let it carry the target specific global
/// variable if there exists one.
static SDValue getLoadStackGuard(SelectionDAG &DAG, const SDLoc &DL,
                                 SDValue &Chain) {
  const TargetLowering &TLI = DAG.getTargetLoweringInfo();
  EVT PtrTy = TLI.getPointerTy(DAG.getDataLayout());
  EVT PtrMemTy = TLI.getPointerMemTy(DAG.getDataLayout());
  MachineFunction &MF = DAG.getMachineFunction();
  Value *Global = TLI.getSDagStackGuard(*MF.getFunction().getParent());
  MachineSDNode *Node =
      DAG.getMachineNode(TargetOpcode::LOAD_STACK_GUARD, DL, PtrTy, Chain);
  if (Global) {
    MachinePointerInfo MPInfo(Global);
    auto Flags = MachineMemOperand::MOLoad | MachineMemOperand::MOInvariant |
                 MachineMemOperand::MODereferenceable;
    MachineMemOperand *MemRef = MF.getMachineMemOperand(
        MPInfo, Flags, PtrTy.getSizeInBits() / 8, DAG.getEVTAlign(PtrTy));
    DAG.setNodeMemRefs(Node, {MemRef});
  }
  if (PtrTy != PtrMemTy)
    return DAG.getPtrExtOrTrunc(SDValue(Node, 0), DL, PtrMemTy);
  return SDValue(Node, 0);
}
 
/// Codegen a new tail for a stack protector check ParentMBB which has had its
/// tail spliced into a stack protector check success bb.
///
/// For a high level explanation of how this fits into the stack protector
/// generation see the comment on the declaration of class
/// StackProtectorDescriptor.
void SelectionDAGBuilder::visitSPDescriptorParent(StackProtectorDescriptor &SPD,
                                                  MachineBasicBlock *ParentBB) {
 
  // First create the loads to the guard/stack slot for the comparison.
  const TargetLowering &TLI = DAG.getTargetLoweringInfo();
  EVT PtrTy = TLI.getPointerTy(DAG.getDataLayout());
  EVT PtrMemTy = TLI.getPointerMemTy(DAG.getDataLayout());
 
  MachineFrameInfo &MFI = ParentBB->getParent()->getFrameInfo();
  int FI = MFI.getStackProtectorIndex();
 
  SDValue Guard;
  SDLoc dl = getCurSDLoc();
  SDValue StackSlotPtr = DAG.getFrameIndex(FI, PtrTy);
  const Module &M = *ParentBB->getParent()->getFunction().getParent();
  Align Align = DL->getPrefTypeAlign(Type::getInt8PtrTy(M.getContext()));
 
  // Generate code to load the content of the guard slot.
  SDValue GuardVal = DAG.getLoad(
      PtrMemTy, dl, DAG.getEntryNode(), StackSlotPtr,
      MachinePointerInfo::getFixedStack(DAG.getMachineFunction(), FI), Align,
      MachineMemOperand::MOVolatile);
 
  if (TLI.useStackGuardXorFP())
    GuardVal = TLI.emitStackGuardXorFP(DAG, GuardVal, dl);
 
  // Retrieve guard check function, nullptr if instrumentation is inlined.
  if (const Function *GuardCheckFn = TLI.getSSPStackGuardCheck(M)) {
    // The target provides a guard check function to validate the guard value.
    // Generate a call to that function with the content of the guard slot as
    // argument.
    FunctionType *FnTy = GuardCheckFn->getFunctionType();
    assert(FnTy->getNumParams() == 1 && "Invalid function signature");
 
    TargetLowering::ArgListTy Args;
    TargetLowering::ArgListEntry Entry;
    Entry.Node = GuardVal;
    Entry.Ty = FnTy->getParamType(0);
    if (GuardCheckFn->hasParamAttribute(0, Attribute::AttrKind::InReg))
      Entry.IsInReg = true;
    Args.push_back(Entry);
 
    TargetLowering::CallLoweringInfo CLI(DAG);
    CLI.setDebugLoc(getCurSDLoc())
        .setChain(DAG.getEntryNode())
        .setCallee(GuardCheckFn->getCallingConv(), FnTy->getReturnType(),
                   getValue(GuardCheckFn), std::move(Args));
 
    std::pair<SDValue, SDValue> Result = TLI.LowerCallTo(CLI);
    DAG.setRoot(Result.second);
    return;
  }
 
  // If useLoadStackGuardNode returns true, generate LOAD_STACK_GUARD.
  // Otherwise, emit a volatile load to retrieve the stack guard value.
  SDValue Chain = DAG.getEntryNode();
  if (TLI.useLoadStackGuardNode()) {
    Guard = getLoadStackGuard(DAG, dl, Chain);
  } else {
    const Value *IRGuard = TLI.getSDagStackGuard(M);
    SDValue GuardPtr = getValue(IRGuard);
 
    Guard = DAG.getLoad(PtrMemTy, dl, Chain, GuardPtr,
                        MachinePointerInfo(IRGuard, 0), Align,
                        MachineMemOperand::MOVolatile);
  }
 
  // Perform the comparison via a getsetcc.
  SDValue Cmp = DAG.getSetCC(dl, TLI.getSetCCResultType(DAG.getDataLayout(),
                                                        *DAG.getContext(),
                                                        Guard.getValueType()),
                             Guard, GuardVal, ISD::SETNE);
 
  // If the guard/stackslot do not equal, branch to failure MBB.
  SDValue BrCond = DAG.getNode(ISD::BRCOND, dl,
                               MVT::Other, GuardVal.getOperand(0),
                               Cmp, DAG.getBasicBlock(SPD.getFailureMBB()));
  // Otherwise branch to success MBB.
  SDValue Br = DAG.getNode(ISD::BR, dl,
                           MVT::Other, BrCond,
                           DAG.getBasicBlock(SPD.getSuccessMBB()));
 
  DAG.setRoot(Br);
}
 
/// Codegen the failure basic block for a stack protector check.
///
/// A failure stack protector machine basic block consists simply of a call to
/// __stack_chk_fail().
///
/// For a high level explanation of how this fits into the stack protector
/// generation see the comment on the declaration of class
/// StackProtectorDescriptor.
void
SelectionDAGBuilder::visitSPDescriptorFailure(StackProtectorDescriptor &SPD) {
  const TargetLowering &TLI = DAG.getTargetLoweringInfo();
  TargetLowering::MakeLibCallOptions CallOptions;
  CallOptions.setDiscardResult(true);
  SDValue Chain =
      TLI.makeLibCall(DAG, RTLIB::STACKPROTECTOR_CHECK_FAIL, MVT::isVoid,
                      None, CallOptions, getCurSDLoc()).second;
  // On PS4, the "return address" must still be within the calling function,
  // even if it's at the very end, so emit an explicit TRAP here.
  // Passing 'true' for doesNotReturn above won't generate the trap for us.
  if (TM.getTargetTriple().isPS4CPU())
    Chain = DAG.getNode(ISD::TRAP, getCurSDLoc(), MVT::Other, Chain);
  // WebAssembly needs an unreachable instruction after a non-returning call,
  // because the function return type can be different from __stack_chk_fail's
  // return type (void).
  if (TM.getTargetTriple().isWasm())
    Chain = DAG.getNode(ISD::TRAP, getCurSDLoc(), MVT::Other, Chain);
 
  DAG.setRoot(Chain);
}
 
/// visitBitTestHeader - This function emits necessary code to produce value
/// suitable for "bit tests"
void SelectionDAGBuilder::visitBitTestHeader(BitTestBlock &B,
                                             MachineBasicBlock *SwitchBB) {
  SDLoc dl = getCurSDLoc();
 
  // Subtract the minimum value.
  SDValue SwitchOp = getValue(B.SValue);
  EVT VT = SwitchOp.getValueType();
  SDValue RangeSub =
      DAG.getNode(ISD::SUB, dl, VT, SwitchOp, DAG.getConstant(B.First, dl, VT));
 
  // Determine the type of the test operands.
  const TargetLowering &TLI = DAG.getTargetLoweringInfo();
  bool UsePtrType = false;
  if (!TLI.isTypeLegal(VT)) {
    UsePtrType = true;
  } else {
    for (unsigned i = 0, e = B.Cases.size(); i != e; ++i)
      if (!isUIntN(VT.getSizeInBits(), B.Cases[i].Mask)) {
        // Switch table case range are encoded into series of masks.
        // Just use pointer type, it's guaranteed to fit.
        UsePtrType = true;
        break;
      }
  }
  SDValue Sub = RangeSub;
  if (UsePtrType) {
    VT = TLI.getPointerTy(DAG.getDataLayout());
    Sub = DAG.getZExtOrTrunc(Sub, dl, VT);
  }
 
  B.RegVT = VT.getSimpleVT();
  B.Reg = FuncInfo.CreateReg(B.RegVT);
  SDValue CopyTo = DAG.getCopyToReg(getControlRoot(), dl, B.Reg, Sub);
 
  MachineBasicBlock* MBB = B.Cases[0].ThisBB;
 
  if (!B.FallthroughUnreachable)
    addSuccessorWithProb(SwitchBB, B.Default, B.DefaultProb);
  addSuccessorWithProb(SwitchBB, MBB, B.Prob);
  SwitchBB->normalizeSuccProbs();
 
  SDValue Root = CopyTo;
  if (!B.FallthroughUnreachable) {
    // Conditional branch to the default block.
    SDValue RangeCmp = DAG.getSetCC(dl,
        TLI.getSetCCResultType(DAG.getDataLayout(), *DAG.getContext(),
                               RangeSub.getValueType()),
        RangeSub, DAG.getConstant(B.Range, dl, RangeSub.getValueType()),
        ISD::SETUGT);
 
    Root = DAG.getNode(ISD::BRCOND, dl, MVT::Other, Root, RangeCmp,
                       DAG.getBasicBlock(B.Default));
  }
 
  // Avoid emitting unnecessary branches to the next block.
  if (MBB != NextBlock(SwitchBB))
    Root = DAG.getNode(ISD::BR, dl, MVT::Other, Root, DAG.getBasicBlock(MBB));
 
  DAG.setRoot(Root);
}
 
/// visitBitTestCase - this function produces one "bit test"
void SelectionDAGBuilder::visitBitTestCase(BitTestBlock &BB,
                                           MachineBasicBlock* NextMBB,
                                           BranchProbability BranchProbToNext,
                                           unsigned Reg,
                                           BitTestCase &B,
                                           MachineBasicBlock *SwitchBB) {
  SDLoc dl = getCurSDLoc();
  MVT VT = BB.RegVT;
  SDValue ShiftOp = DAG.getCopyFromReg(getControlRoot(), dl, Reg, VT);
  SDValue Cmp;
  unsigned PopCount = countPopulation(B.Mask);
  const TargetLowering &TLI = DAG.getTargetLoweringInfo();
  if (PopCount == 1) {
    // Testing for a single bit; just compare the shift count with what it
    // would need to be to shift a 1 bit in that position.
    Cmp = DAG.getSetCC(
        dl, TLI.getSetCCResultType(DAG.getDataLayout(), *DAG.getContext(), VT),
        ShiftOp, DAG.getConstant(countTrailingZeros(B.Mask), dl, VT),
        ISD::SETEQ);
  } else if (PopCount == BB.Range) {
    // There is only one zero bit in the range, test for it directly.
    Cmp = DAG.getSetCC(
        dl, TLI.getSetCCResultType(DAG.getDataLayout(), *DAG.getContext(), VT),
        ShiftOp, DAG.getConstant(countTrailingOnes(B.Mask), dl, VT),
        ISD::SETNE);
  } else {
    // Make desired shift
    SDValue SwitchVal = DAG.getNode(ISD::SHL, dl, VT,
                                    DAG.getConstant(1, dl, VT), ShiftOp);
 
    // Emit bit tests and jumps
    SDValue AndOp = DAG.getNode(ISD::AND, dl,
                                VT, SwitchVal, DAG.getConstant(B.Mask, dl, VT));
    Cmp = DAG.getSetCC(
        dl, TLI.getSetCCResultType(DAG.getDataLayout(), *DAG.getContext(), VT),
        AndOp, DAG.getConstant(0, dl, VT), ISD::SETNE);
  }
 
  // The branch probability from SwitchBB to B.TargetBB is B.ExtraProb.
  addSuccessorWithProb(SwitchBB, B.TargetBB, B.ExtraProb);
  // The branch probability from SwitchBB to NextMBB is BranchProbToNext.
  addSuccessorWithProb(SwitchBB, NextMBB, BranchProbToNext);
  // It is not guaranteed that the sum of B.ExtraProb and BranchProbToNext is
  // one as they are relative probabilities (and thus work more like weights),
  // and hence we need to normalize them to let the sum of them become one.
  SwitchBB->normalizeSuccProbs();
 
  SDValue BrAnd = DAG.getNode(ISD::BRCOND, dl,
                              MVT::Other, getControlRoot(),
                              Cmp, DAG.getBasicBlock(B.TargetBB));
 
  // Avoid emitting unnecessary branches to the next block.
  if (NextMBB != NextBlock(SwitchBB))
    BrAnd = DAG.getNode(ISD::BR, dl, MVT::Other, BrAnd,
                        DAG.getBasicBlock(NextMBB));
 
  DAG.setRoot(BrAnd);
}
 
void SelectionDAGBuilder::visitInvoke(const InvokeInst &I) {
  MachineBasicBlock *InvokeMBB = FuncInfo.MBB;
 
  // Retrieve successors. Look through artificial IR level blocks like
  // catchswitch for successors.
  MachineBasicBlock *Return = FuncInfo.MBBMap[I.getSuccessor(0)];
  const BasicBlock *EHPadBB = I.getSuccessor(1);
 
  // Deopt bundles are lowered in LowerCallSiteWithDeoptBundle, and we don't
  // have to do anything here to lower funclet bundles.
  assert(!I.hasOperandBundlesOtherThan(
             {LLVMContext::OB_deopt, LLVMContext::OB_gc_transition,
              LLVMContext::OB_gc_live, LLVMContext::OB_funclet,
              LLVMContext::OB_cfguardtarget,
              LLVMContext::OB_clang_arc_attachedcall}) &&
         "Cannot lower invokes with arbitrary operand bundles yet!");
 
  const Value *Callee(I.getCalledOperand());
  const Function *Fn = dyn_cast<Function>(Callee);
  if (isa<InlineAsm>(Callee))
    visitInlineAsm(I, EHPadBB);
  else if (Fn && Fn->isIntrinsic()) {
    switch (Fn->getIntrinsicID()) {
    default:
      llvm_unreachable("Cannot invoke this intrinsic");
    case Intrinsic::donothing:
      // Ignore invokes to @llvm.donothing: jump directly to the next BB.
    case Intrinsic::seh_try_begin:
    case Intrinsic::seh_scope_begin:
    case Intrinsic::seh_try_end:
    case Intrinsic::seh_scope_end:
      break;
    case Intrinsic::experimental_patchpoint_void:
    case Intrinsic::experimental_patchpoint_i64:
      visitPatchpoint(I, EHPadBB);
      break;
    case Intrinsic::experimental_gc_statepoint:
      LowerStatepoint(cast<GCStatepointInst>(I), EHPadBB);
      break;
    case Intrinsic::wasm_rethrow: {
      // This is usually done in visitTargetIntrinsic, but this intrinsic is
      // special because it can be invoked, so we manually lower it to a DAG
      // node here.
      SmallVector<SDValue, 8> Ops;
      Ops.push_back(getRoot()); // inchain
      const TargetLowering &TLI = DAG.getTargetLoweringInfo();
      Ops.push_back(
          DAG.getTargetConstant(Intrinsic::wasm_rethrow, getCurSDLoc(),
                                TLI.getPointerTy(DAG.getDataLayout())));
      SDVTList VTs = DAG.getVTList(ArrayRef<EVT>({MVT::Other})); // outchain
      DAG.setRoot(DAG.getNode(ISD::INTRINSIC_VOID, getCurSDLoc(), VTs, Ops));
      break;
    }
    }
  } else if (I.countOperandBundlesOfType(LLVMContext::OB_deopt)) {
    // Currently we do not lower any intrinsic calls with deopt operand bundles.
    // Eventually we will support lowering the @llvm.experimental.deoptimize
    // intrinsic, and right now there are no plans to support other intrinsics
    // with deopt state.
    LowerCallSiteWithDeoptBundle(&I, getValue(Callee), EHPadBB);
  } else {
    LowerCallTo(I, getValue(Callee), false, false, EHPadBB);
  }
 
  // If the value of the invoke is used outside of its defining block, make it
  // available as a virtual register.
  // We already took care of the exported value for the statepoint instruction
  // during call to the LowerStatepoint.
  if (!isa<GCStatepointInst>(I)) {
    CopyToExportRegsIfNeeded(&I);
  }
 
  SmallVector<std::pair<MachineBasicBlock *, BranchProbability>, 1> UnwindDests;
  BranchProbabilityInfo *BPI = FuncInfo.BPI;
  BranchProbability EHPadBBProb =
      BPI ? BPI->getEdgeProbability(InvokeMBB->getBasicBlock(), EHPadBB)
          : BranchProbability::getZero();
  findUnwindDestinations(FuncInfo, EHPadBB, EHPadBBProb, UnwindDests);
 
  // Update successor info.
  addSuccessorWithProb(InvokeMBB, Return);
  for (auto &UnwindDest : UnwindDests) {
    UnwindDest.first->setIsEHPad();
    addSuccessorWithProb(InvokeMBB, UnwindDest.first, UnwindDest.second);
  }
  InvokeMBB->normalizeSuccProbs();
 
  // Drop into normal successor.
  DAG.setRoot(DAG.getNode(ISD::BR, getCurSDLoc(), MVT::Other, getControlRoot(),
                          DAG.getBasicBlock(Return)));
}
 
void SelectionDAGBuilder::visitCallBr(const CallBrInst &I) {
  MachineBasicBlock *CallBrMBB = FuncInfo.MBB;
 
  // Deopt bundles are lowered in LowerCallSiteWithDeoptBundle, and we don't
  // have to do anything here to lower funclet bundles.
  assert(!I.hasOperandBundlesOtherThan(
             {LLVMContext::OB_deopt, LLVMContext::OB_funclet}) &&
         "Cannot lower callbrs with arbitrary operand bundles yet!");
 
  assert(I.isInlineAsm() && "Only know how to handle inlineasm callbr");
  visitInlineAsm(I);
  CopyToExportRegsIfNeeded(&I);
 
  // Retrieve successors.
  MachineBasicBlock *Return = FuncInfo.MBBMap[I.getDefaultDest()];
 
  // Update successor info.
  addSuccessorWithProb(CallBrMBB, Return, BranchProbability::getOne());
  for (unsigned i = 0, e = I.getNumIndirectDests(); i < e; ++i) {
    MachineBasicBlock *Target = FuncInfo.MBBMap[I.getIndirectDest(i)];
    addSuccessorWithProb(CallBrMBB, Target, BranchProbability::getZero());
    Target->setIsInlineAsmBrIndirectTarget();
  }
  CallBrMBB->normalizeSuccProbs();
 
  // Drop into default successor.
  DAG.setRoot(DAG.getNode(ISD::BR, getCurSDLoc(),
                          MVT::Other, getControlRoot(),
                          DAG.getBasicBlock(Return)));
}
 
void SelectionDAGBuilder::visitResume(const ResumeInst &RI) {
  llvm_unreachable("SelectionDAGBuilder shouldn't visit resume instructions!");
}
 
void SelectionDAGBuilder::visitLandingPad(const LandingPadInst &LP) {
  assert(FuncInfo.MBB->isEHPad() &&
         "Call to landingpad not in landing pad!");
 
  // If there aren't registers to copy the values into (e.g., during SjLj
  // exceptions), then don't bother to create these DAG nodes.
  const TargetLowering &TLI = DAG.getTargetLoweringInfo();
  const Constant *PersonalityFn = FuncInfo.Fn->getPersonalityFn();
  if (TLI.getExceptionPointerRegister(PersonalityFn) == 0 &&
      TLI.getExceptionSelectorRegister(PersonalityFn) == 0)
    return;
 
  // If landingpad's return type is token type, we don't create DAG nodes
  // for its exception pointer and selector value. The extraction of exception
  // pointer or selector value from token type landingpads is not currently
  // supported.
  if (LP.getType()->isTokenTy())
    return;
 
  SmallVector<EVT, 2> ValueVTs;
  SDLoc dl = getCurSDLoc();
  ComputeValueVTs(TLI, DAG.getDataLayout(), LP.getType(), ValueVTs);
  assert(ValueVTs.size() == 2 && "Only two-valued landingpads are supported");
 
  // Get the two live-in registers as SDValues. The physregs have already been
  // copied into virtual registers.
  SDValue Ops[2];
  if (FuncInfo.ExceptionPointerVirtReg) {
    Ops[0] = DAG.getZExtOrTrunc(
        DAG.getCopyFromReg(DAG.getEntryNode(), dl,
                           FuncInfo.ExceptionPointerVirtReg,
                           TLI.getPointerTy(DAG.getDataLayout())),
        dl, ValueVTs[0]);
  } else {
    Ops[0] = DAG.getConstant(0, dl, TLI.getPointerTy(DAG.getDataLayout()));
  }
  Ops[1] = DAG.getZExtOrTrunc(
      DAG.getCopyFromReg(DAG.getEntryNode(), dl,
                         FuncInfo.ExceptionSelectorVirtReg,
                         TLI.getPointerTy(DAG.getDataLayout())),
      dl, ValueVTs[1]);
 
  // Merge into one.
  SDValue Res = DAG.getNode(ISD::MERGE_VALUES, dl,
                            DAG.getVTList(ValueVTs), Ops);
  setValue(&LP, Res);
}
 
void SelectionDAGBuilder::UpdateSplitBlock(MachineBasicBlock *First,
                                           MachineBasicBlock *Last) {
  // Update JTCases.
  for (unsigned i = 0, e = SL->JTCases.size(); i != e; ++i)
    if (SL->JTCases[i].first.HeaderBB == First)
      SL->JTCases[i].first.HeaderBB = Last;
 
  // Update BitTestCases.
  for (unsigned i = 0, e = SL->BitTestCases.size(); i != e; ++i)
    if (SL->BitTestCases[i].Parent == First)
      SL->BitTestCases[i].Parent = Last;
}
 
void SelectionDAGBuilder::visitIndirectBr(const IndirectBrInst &I) {
  MachineBasicBlock *IndirectBrMBB = FuncInfo.MBB;
 
  // Update machine-CFG edges with unique successors.
  SmallSet<BasicBlock*, 32> Done;
  for (unsigned i = 0, e = I.getNumSuccessors(); i != e; ++i) {
    BasicBlock *BB = I.getSuccessor(i);
    bool Inserted = Done.insert(BB).second;
    if (!Inserted)
        continue;
 
    MachineBasicBlock *Succ = FuncInfo.MBBMap[BB];
    addSuccessorWithProb(IndirectBrMBB, Succ);
  }
  IndirectBrMBB->normalizeSuccProbs();
 
  DAG.setRoot(DAG.getNode(ISD::BRIND, getCurSDLoc(),
                          MVT::Other, getControlRoot(),
                          getValue(I.getAddress())));
}
 
void SelectionDAGBuilder::visitUnreachable(const UnreachableInst &I) {
  if (!DAG.getTarget().Options.TrapUnreachable)
    return;
 
  // We may be able to ignore unreachable behind a noreturn call.
  if (DAG.getTarget().Options.NoTrapAfterNoreturn) {
    const BasicBlock &BB = *I.getParent();
    if (&I != &BB.front()) {
      BasicBlock::const_iterator PredI =
        std::prev(BasicBlock::const_iterator(&I));
      if (const CallInst *Call = dyn_cast<CallInst>(&*PredI)) {
        if (Call->doesNotReturn())
          return;
      }
    }
  }
 
  DAG.setRoot(DAG.getNode(ISD::TRAP, getCurSDLoc(), MVT::Other, DAG.getRoot()));
}
 
void SelectionDAGBuilder::visitUnary(const User &I, unsigned Opcode) {
  SDNodeFlags Flags;
 
  SDValue Op = getValue(I.getOperand(0));
  SDValue UnNodeValue = DAG.getNode(Opcode, getCurSDLoc(), Op.getValueType(),
                                    Op, Flags);
  setValue(&I, UnNodeValue);
}
 
void SelectionDAGBuilder::visitBinary(const User &I, unsigned Opcode) {
  SDNodeFlags Flags;
  if (auto *OFBinOp = dyn_cast<OverflowingBinaryOperator>(&I)) {
    Flags.setNoSignedWrap(OFBinOp->hasNoSignedWrap());
    Flags.setNoUnsignedWrap(OFBinOp->hasNoUnsignedWrap());
  }
  if (auto *ExactOp = dyn_cast<PossiblyExactOperator>(&I))
    Flags.setExact(ExactOp->isExact());
  if (auto *FPOp = dyn_cast<FPMathOperator>(&I))
    Flags.copyFMF(*FPOp);
 
  SDValue Op1 = getValue(I.getOperand(0));
  SDValue Op2 = getValue(I.getOperand(1));
  SDValue BinNodeValue = DAG.getNode(Opcode, getCurSDLoc(), Op1.getValueType(),
                                     Op1, Op2, Flags);
  setValue(&I, BinNodeValue);
}
 
void SelectionDAGBuilder::visitShift(const User &I, unsigned Opcode) {
  SDValue Op1 = getValue(I.getOperand(0));
  SDValue Op2 = getValue(I.getOperand(1));
 
  EVT ShiftTy = DAG.getTargetLoweringInfo().getShiftAmountTy(
      Op1.getValueType(), DAG.getDataLayout());
 
  // Coerce the shift amount to the right type if we can.
  if (!I.getType()->isVectorTy() && Op2.getValueType() != ShiftTy) {
    unsigned ShiftSize = ShiftTy.getSizeInBits();
    unsigned Op2Size = Op2.getValueSizeInBits();
    SDLoc DL = getCurSDLoc();
 
    // If the operand is smaller than the shift count type, promote it.
    if (ShiftSize > Op2Size)
      Op2 = DAG.getNode(ISD::ZERO_EXTEND, DL, ShiftTy, Op2);
 
    // If the operand is larger than the shift count type but the shift
    // count type has enough bits to represent any shift value, truncate
    // it now. This is a common case and it exposes the truncate to
    // optimization early.
    else if (ShiftSize >= Log2_32_Ceil(Op2.getValueSizeInBits()))
      Op2 = DAG.getNode(ISD::TRUNCATE, DL, ShiftTy, Op2);
    // Otherwise we'll need to temporarily settle for some other convenient
    // type.  Type legalization will make adjustments once the shiftee is split.
    else
      Op2 = DAG.getZExtOrTrunc(Op2, DL, MVT::i32);
  }
 
  bool nuw = false;
  bool nsw = false;
  bool exact = false;
 
  if (Opcode == ISD::SRL || Opcode == ISD::SRA || Opcode == ISD::SHL) {
 
    if (const OverflowingBinaryOperator *OFBinOp =
            dyn_cast<const OverflowingBinaryOperator>(&I)) {
      nuw = OFBinOp->hasNoUnsignedWrap();
      nsw = OFBinOp->hasNoSignedWrap();
    }
    if (const PossiblyExactOperator *ExactOp =
            dyn_cast<const PossiblyExactOperator>(&I))
      exact = ExactOp->isExact();
  }
  SDNodeFlags Flags;
  Flags.setExact(exact);
  Flags.setNoSignedWrap(nsw);
  Flags.setNoUnsignedWrap(nuw);
  SDValue Res = DAG.getNode(Opcode, getCurSDLoc(), Op1.getValueType(), Op1, Op2,
                            Flags);
  setValue(&I, Res);
}
 
void SelectionDAGBuilder::visitSDiv(const User &I) {
  SDValue Op1 = getValue(I.getOperand(0));
  SDValue Op2 = getValue(I.getOperand(1));
 
  SDNodeFlags Flags;
  Flags.setExact(isa<PossiblyExactOperator>(&I) &&
                 cast<PossiblyExactOperator>(&I)->isExact());
  setValue(&I, DAG.getNode(ISD::SDIV, getCurSDLoc(), Op1.getValueType(), Op1,
                           Op2, Flags));
}
 
void SelectionDAGBuilder::visitICmp(const User &I) {
  ICmpInst::Predicate predicate = ICmpInst::BAD_ICMP_PREDICATE;
  if (const ICmpInst *IC = dyn_cast<ICmpInst>(&I))
    predicate = IC->getPredicate();
  else if (const ConstantExpr *IC = dyn_cast<ConstantExpr>(&I))
    predicate = ICmpInst::Predicate(IC->getPredicate());
  SDValue Op1 = getValue(I.getOperand(0));
  SDValue Op2 = getValue(I.getOperand(1));
  ISD::CondCode Opcode = getICmpCondCode(predicate);
 
  auto &TLI = DAG.getTargetLoweringInfo();
  EVT MemVT =
      TLI.getMemValueType(DAG.getDataLayout(), I.getOperand(0)->getType());
 
  // If a pointer's DAG type is larger than its memory type then the DAG values
  // are zero-extended. This breaks signed comparisons so truncate back to the
  // underlying type before doing the compare.
  if (Op1.getValueType() != MemVT) {
    Op1 = DAG.getPtrExtOrTrunc(Op1, getCurSDLoc(), MemVT);
    Op2 = DAG.getPtrExtOrTrunc(Op2, getCurSDLoc(), MemVT);
  }
 
  EVT DestVT = DAG.getTargetLoweringInfo().getValueType(DAG.getDataLayout(),
                                                        I.getType());
  setValue(&I, DAG.getSetCC(getCurSDLoc(), DestVT, Op1, Op2, Opcode));
}
 
void SelectionDAGBuilder::visitFCmp(const User &I) {
  FCmpInst::Predicate predicate = FCmpInst::BAD_FCMP_PREDICATE;
  if (const FCmpInst *FC = dyn_cast<FCmpInst>(&I))
    predicate = FC->getPredicate();
  else if (const ConstantExpr *FC = dyn_cast<ConstantExpr>(&I))
    predicate = FCmpInst::Predicate(FC->getPredicate());
  SDValue Op1 = getValue(I.getOperand(0));
  SDValue Op2 = getValue(I.getOperand(1));
 
  ISD::CondCode Condition = getFCmpCondCode(predicate);
  auto *FPMO = cast<FPMathOperator>(&I);
  if (FPMO->hasNoNaNs() || TM.Options.NoNaNsFPMath)
    Condition = getFCmpCodeWithoutNaN(Condition);
 
  SDNodeFlags Flags;
  Flags.copyFMF(*FPMO);
  SelectionDAG::FlagInserter FlagsInserter(DAG, Flags);
 
  EVT DestVT = DAG.getTargetLoweringInfo().getValueType(DAG.getDataLayout(),
                                                        I.getType());
  setValue(&I, DAG.getSetCC(getCurSDLoc(), DestVT, Op1, Op2, Condition));
}
 
// Check if the condition of the select has one use or two users that are both
// selects with the same condition.
static bool hasOnlySelectUsers(const Value *Cond) {
  return llvm::all_of(Cond->users(), [](const Value *V) {
    return isa<SelectInst>(V);
  });
}
 
void SelectionDAGBuilder::visitSelect(const User &I) {
  SmallVector<EVT, 4> ValueVTs;
  ComputeValueVTs(DAG.getTargetLoweringInfo(), DAG.getDataLayout(), I.getType(),
                  ValueVTs);
  unsigned NumValues = ValueVTs.size();
  if (NumValues == 0) return;
 
  SmallVector<SDValue, 4> Values(NumValues);
  SDValue Cond     = getValue(I.getOperand(0));
  SDValue LHSVal   = getValue(I.getOperand(1));
  SDValue RHSVal   = getValue(I.getOperand(2));
  SmallVector<SDValue, 1> BaseOps(1, Cond);
  ISD::NodeType OpCode =
      Cond.getValueType().isVector() ? ISD::VSELECT : ISD::SELECT;
 
  bool IsUnaryAbs = false;
  bool Negate = false;
 
  SDNodeFlags Flags;
  if (auto *FPOp = dyn_cast<FPMathOperator>(&I))
    Flags.copyFMF(*FPOp);
 
  // Min/max matching is only viable if all output VTs are the same.
  if (is_splat(ValueVTs)) {
    EVT VT = ValueVTs[0];
    LLVMContext &Ctx = *DAG.getContext();
    auto &TLI = DAG.getTargetLoweringInfo();
 
    // We care about the legality of the operation after it has been type
    // legalized.
    while (TLI.getTypeAction(Ctx, VT) != TargetLoweringBase::TypeLegal)
      VT = TLI.getTypeToTransformTo(Ctx, VT);
 
    // If the vselect is legal, assume we want to leave this as a vector setcc +
    // vselect. Otherwise, if this is going to be scalarized, we want to see if
    // min/max is legal on the scalar type.
    bool UseScalarMinMax = VT.isVector() &&
      !TLI.isOperationLegalOrCustom(ISD::VSELECT, VT);
 
    Value *LHS, *RHS;
    auto SPR = matchSelectPattern(const_cast<User*>(&I), LHS, RHS);
    ISD::NodeType Opc = ISD::DELETED_NODE;
    switch (SPR.Flavor) {
    case SPF_UMAX:    Opc = ISD::UMAX; break;
    case SPF_UMIN:    Opc = ISD::UMIN; break;
    case SPF_SMAX:    Opc = ISD::SMAX; break;
    case SPF_SMIN:    Opc = ISD::SMIN; break;
    case SPF_FMINNUM:
      switch (SPR.NaNBehavior) {
      case SPNB_NA: llvm_unreachable("No NaN behavior for FP op?");
      case SPNB_RETURNS_NAN:   Opc = ISD::FMINIMUM; break;
      case SPNB_RETURNS_OTHER: Opc = ISD::FMINNUM; break;
      case SPNB_RETURNS_ANY: {
        if (TLI.isOperationLegalOrCustom(ISD::FMINNUM, VT))
          Opc = ISD::FMINNUM;
        else if (TLI.isOperationLegalOrCustom(ISD::FMINIMUM, VT))
          Opc = ISD::FMINIMUM;
        else if (UseScalarMinMax)
          Opc = TLI.isOperationLegalOrCustom(ISD::FMINNUM, VT.getScalarType()) ?
            ISD::FMINNUM : ISD::FMINIMUM;
        break;
      }
      }
      break;
    case SPF_FMAXNUM:
      switch (SPR.NaNBehavior) {
      case SPNB_NA: llvm_unreachable("No NaN behavior for FP op?");
      case SPNB_RETURNS_NAN:   Opc = ISD::FMAXIMUM; break;
      case SPNB_RETURNS_OTHER: Opc = ISD::FMAXNUM; break;
      case SPNB_RETURNS_ANY:
 
        if (TLI.isOperationLegalOrCustom(ISD::FMAXNUM, VT))
          Opc = ISD::FMAXNUM;
        else if (TLI.isOperationLegalOrCustom(ISD::FMAXIMUM, VT))
          Opc = ISD::FMAXIMUM;
        else if (UseScalarMinMax)
          Opc = TLI.isOperationLegalOrCustom(ISD::FMAXNUM, VT.getScalarType()) ?
            ISD::FMAXNUM : ISD::FMAXIMUM;
        break;
      }
      break;
    case SPF_NABS:
      Negate = true;
      LLVM_FALLTHROUGH;
    case SPF_ABS:
      IsUnaryAbs = true;
      Opc = ISD::ABS;
      break;
    default: break;
    }
 
    if (!IsUnaryAbs && Opc != ISD::DELETED_NODE &&
        (TLI.isOperationLegalOrCustom(Opc, VT) ||
         (UseScalarMinMax &&
          TLI.isOperationLegalOrCustom(Opc, VT.getScalarType()))) &&
        // If the underlying comparison instruction is used by any other
        // instruction, the consumed instructions won't be destroyed, so it is
        // not profitable to convert to a min/max.
        hasOnlySelectUsers(cast<SelectInst>(I).getCondition())) {
      OpCode = Opc;
      LHSVal = getValue(LHS);
      RHSVal = getValue(RHS);
      BaseOps.clear();
    }
 
    if (IsUnaryAbs) {
      OpCode = Opc;
      LHSVal = getValue(LHS);
      BaseOps.clear();
    }
  }
 
  if (IsUnaryAbs) {
    for (unsigned i = 0; i != NumValues; ++i) {
      SDLoc dl = getCurSDLoc();
      EVT VT = LHSVal.getNode()->getValueType(LHSVal.getResNo() + i);
      Values[i] =
          DAG.getNode(OpCode, dl, VT, LHSVal.getValue(LHSVal.getResNo() + i));
      if (Negate)
        Values[i] = DAG.getNode(ISD::SUB, dl, VT, DAG.getConstant(0, dl, VT),
                                Values[i]);
    }
  } else {
    for (unsigned i = 0; i != NumValues; ++i) {
      SmallVector<SDValue, 3> Ops(BaseOps.begin(), BaseOps.end());
      Ops.push_back(SDValue(LHSVal.getNode(), LHSVal.getResNo() + i));
      Ops.push_back(SDValue(RHSVal.getNode(), RHSVal.getResNo() + i));
      Values[i] = DAG.getNode(
          OpCode, getCurSDLoc(),
          LHSVal.getNode()->getValueType(LHSVal.getResNo() + i), Ops, Flags);
    }
  }
 
  setValue(&I, DAG.getNode(ISD::MERGE_VALUES, getCurSDLoc(),
                           DAG.getVTList(ValueVTs), Values));
}
 
void SelectionDAGBuilder::visitTrunc(const User &I) {
  // TruncInst cannot be a no-op cast because sizeof(src) > sizeof(dest).
  SDValue N = getValue(I.getOperand(0));
  EVT DestVT = DAG.getTargetLoweringInfo().getValueType(DAG.getDataLayout(),
                                                        I.getType());
  setValue(&I, DAG.getNode(ISD::TRUNCATE, getCurSDLoc(), DestVT, N));
}
 
void SelectionDAGBuilder::visitZExt(const User &I) {
  // ZExt cannot be a no-op cast because sizeof(src) < sizeof(dest).
  // ZExt also can't be a cast to bool for same reason. So, nothing much to do
  SDValue N = getValue(I.getOperand(0));
  EVT DestVT = DAG.getTargetLoweringInfo().getValueType(DAG.getDataLayout(),
                                                        I.getType());
  setValue(&I, DAG.getNode(ISD::ZERO_EXTEND, getCurSDLoc(), DestVT, N));
}
 
void SelectionDAGBuilder::visitSExt(const User &I) {
  // SExt cannot be a no-op cast because sizeof(src) < sizeof(dest).
  // SExt also can't be a cast to bool for same reason. So, nothing much to do
  SDValue N = getValue(I.getOperand(0));
  EVT DestVT = DAG.getTargetLoweringInfo().getValueType(DAG.getDataLayout(),
                                                        I.getType());
  setValue(&I, DAG.getNode(ISD::SIGN_EXTEND, getCurSDLoc(), DestVT, N));
}
 
void SelectionDAGBuilder::visitFPTrunc(const User &I) {
  // FPTrunc is never a no-op cast, no need to check
  SDValue N = getValue(I.getOperand(0));
  SDLoc dl = getCurSDLoc();
  const TargetLowering &TLI = DAG.getTargetLoweringInfo();
  EVT DestVT = TLI.getValueType(DAG.getDataLayout(), I.getType());
  setValue(&I, DAG.getNode(ISD::FP_ROUND, dl, DestVT, N,
                           DAG.getTargetConstant(
                               0, dl, TLI.getPointerTy(DAG.getDataLayout()))));
}
 
void SelectionDAGBuilder::visitFPExt(const User &I) {
  // FPExt is never a no-op cast, no need to check
  SDValue N = getValue(I.getOperand(0));
  EVT DestVT = DAG.getTargetLoweringInfo().getValueType(DAG.getDataLayout(),
                                                        I.getType());
  setValue(&I, DAG.getNode(ISD::FP_EXTEND, getCurSDLoc(), DestVT, N));
}
 
void SelectionDAGBuilder::visitFPToUI(const User &I) {
  // FPToUI is never a no-op cast, no need to check
  SDValue N = getValue(I.getOperand(0));
  EVT DestVT = DAG.getTargetLoweringInfo().getValueType(DAG.getDataLayout(),
                                                        I.getType());
  setValue(&I, DAG.getNode(ISD::FP_TO_UINT, getCurSDLoc(), DestVT, N));
}
 
void SelectionDAGBuilder::visitFPToSI(const User &I) {
  // FPToSI is never a no-op cast, no need to check
  SDValue N = getValue(I.getOperand(0));
  EVT DestVT = DAG.getTargetLoweringInfo().getValueType(DAG.getDataLayout(),
                                                        I.getType());
  setValue(&I, DAG.getNode(ISD::FP_TO_SINT, getCurSDLoc(), DestVT, N));
}
 
void SelectionDAGBuilder::visitUIToFP(const User &I) {
  // UIToFP is never a no-op cast, no need to check
  SDValue N = getValue(I.getOperand(0));
  EVT DestVT = DAG.getTargetLoweringInfo().getValueType(DAG.getDataLayout(),
                                                        I.getType());
  setValue(&I, DAG.getNode(ISD::UINT_TO_FP, getCurSDLoc(), DestVT, N));
}
 
void SelectionDAGBuilder::visitSIToFP(const User &I) {
  // SIToFP is never a no-op cast, no need to check
  SDValue N = getValue(I.getOperand(0));
  EVT DestVT = DAG.getTargetLoweringInfo().getValueType(DAG.getDataLayout(),
                                                        I.getType());
  setValue(&I, DAG.getNode(ISD::SINT_TO_FP, getCurSDLoc(), DestVT, N));
}
 
void SelectionDAGBuilder::visitPtrToInt(const User &I) {
  // What to do depends on the size of the integer and the size of the pointer.
  // We can either truncate, zero extend, or no-op, accordingly.
  SDValue N = getValue(I.getOperand(0));
  auto &TLI = DAG.getTargetLoweringInfo();
  EVT DestVT = DAG.getTargetLoweringInfo().getValueType(DAG.getDataLayout(),
                                                        I.getType());
  EVT PtrMemVT =
      TLI.getMemValueType(DAG.getDataLayout(), I.getOperand(0)->getType());
  N = DAG.getPtrExtOrTrunc(N, getCurSDLoc(), PtrMemVT);
  N = DAG.getZExtOrTrunc(N, getCurSDLoc(), DestVT);
  setValue(&I, N);
}
 
void SelectionDAGBuilder::visitIntToPtr(const User &I) {
  // What to do depends on the size of the integer and the size of the pointer.
  // We can either truncate, zero extend, or no-op, accordingly.
  SDValue N = getValue(I.getOperand(0));
  auto &TLI = DAG.getTargetLoweringInfo();
  EVT DestVT = TLI.getValueType(DAG.getDataLayout(), I.getType());
  EVT PtrMemVT = TLI.getMemValueType(DAG.getDataLayout(), I.getType());
  N = DAG.getZExtOrTrunc(N, getCurSDLoc(), PtrMemVT);
  N = DAG.getPtrExtOrTrunc(N, getCurSDLoc(), DestVT);
  setValue(&I, N);
}
 
void SelectionDAGBuilder::visitBitCast(const User &I) {
  SDValue N = getValue(I.getOperand(0));
  SDLoc dl = getCurSDLoc();
  EVT DestVT = DAG.getTargetLoweringInfo().getValueType(DAG.getDataLayout(),
                                                        I.getType());
 
  // BitCast assures us that source and destination are the same size so this is
  // either a BITCAST or a no-op.
  if (DestVT != N.getValueType())
    setValue(&I, DAG.getNode(ISD::BITCAST, dl,
                             DestVT, N)); // convert types.
  // Check if the original LLVM IR Operand was a ConstantInt, because getValue()
  // might fold any kind of constant expression to an integer constant and that
  // is not what we are looking for. Only recognize a bitcast of a genuine
  // constant integer as an opaque constant.
  else if(ConstantInt *C = dyn_cast<ConstantInt>(I.getOperand(0)))
    setValue(&I, DAG.getConstant(C->getValue(), dl, DestVT, /*isTarget=*/false,
                                 /*isOpaque*/true));
  else
    setValue(&I, N);            // noop cast.
}
 
void SelectionDAGBuilder::visitAddrSpaceCast(const User &I) {
  const TargetLowering &TLI = DAG.getTargetLoweringInfo();
  const Value *SV = I.getOperand(0);
  SDValue N = getValue(SV);
  EVT DestVT = TLI.getValueType(DAG.getDataLayout(), I.getType());
 
  unsigned SrcAS = SV->getType()->getPointerAddressSpace();
  unsigned DestAS = I.getType()->getPointerAddressSpace();
 
  if (!TM.isNoopAddrSpaceCast(SrcAS, DestAS))
    N = DAG.getAddrSpaceCast(getCurSDLoc(), DestVT, N, SrcAS, DestAS);
 
  setValue(&I, N);
}
 
void SelectionDAGBuilder::visitInsertElement(const User &I) {
  const TargetLowering &TLI = DAG.getTargetLoweringInfo();
  SDValue InVec = getValue(I.getOperand(0));
  SDValue InVal = getValue(I.getOperand(1));
  SDValue InIdx = DAG.getSExtOrTrunc(getValue(I.getOperand(2)), getCurSDLoc(),
                                     TLI.getVectorIdxTy(DAG.getDataLayout()));
  setValue(&I, DAG.getNode(ISD::INSERT_VECTOR_ELT, getCurSDLoc(),
                           TLI.getValueType(DAG.getDataLayout(), I.getType()),
                           InVec, InVal, InIdx));
}
 
void SelectionDAGBuilder::visitExtractElement(const User &I) {
  const TargetLowering &TLI = DAG.getTargetLoweringInfo();
  SDValue InVec = getValue(I.getOperand(0));
  SDValue InIdx = DAG.getSExtOrTrunc(getValue(I.getOperand(1)), getCurSDLoc(),
                                     TLI.getVectorIdxTy(DAG.getDataLayout()));
  setValue(&I, DAG.getNode(ISD::EXTRACT_VECTOR_ELT, getCurSDLoc(),
                           TLI.getValueType(DAG.getDataLayout(), I.getType()),
                           InVec, InIdx));
}
 
void SelectionDAGBuilder::visitShuffleVector(const User &I) {
  SDValue Src1 = getValue(I.getOperand(0));
  SDValue Src2 = getValue(I.getOperand(1));
  ArrayRef<int> Mask;
  if (auto *SVI = dyn_cast<ShuffleVectorInst>(&I))
    Mask = SVI->getShuffleMask();
  else
    Mask = cast<ConstantExpr>(I).getShuffleMask();
  SDLoc DL = getCurSDLoc();
  const TargetLowering &TLI = DAG.getTargetLoweringInfo();
  EVT VT = TLI.getValueType(DAG.getDataLayout(), I.getType());
  EVT SrcVT = Src1.getValueType();
 
  if (all_of(Mask, [](int Elem) { return Elem == 0; }) &&
      VT.isScalableVector()) {
    // Canonical splat form of first element of first input vector.
    SDValue FirstElt =
        DAG.getNode(ISD::EXTRACT_VECTOR_ELT, DL, SrcVT.getScalarType(), Src1,
                    DAG.getVectorIdxConstant(0, DL));
    setValue(&I, DAG.getNode(ISD::SPLAT_VECTOR, DL, VT, FirstElt));
    return;
  }
 
  // For now, we only handle splats for scalable vectors.
  // The DAGCombiner will perform a BUILD_VECTOR -> SPLAT_VECTOR transformation
  // for targets that support a SPLAT_VECTOR for non-scalable vector types.
  assert(!VT.isScalableVector() && "Unsupported scalable vector shuffle");
 
  unsigned SrcNumElts = SrcVT.getVectorNumElements();
  unsigned MaskNumElts = Mask.size();
 
  if (SrcNumElts == MaskNumElts) {
    setValue(&I, DAG.getVectorShuffle(VT, DL, Src1, Src2, Mask));
    return;
  }
 
  // Normalize the shuffle vector since mask and vector length don't match.
  if (SrcNumElts < MaskNumElts) {
    // Mask is longer than the source vectors. We can use concatenate vector to
    // make the mask and vectors lengths match.
 
    if (MaskNumElts % SrcNumElts == 0) {
      // Mask length is a multiple of the source vector length.
      // Check if the shuffle is some kind of concatenation of the input
      // vectors.
      unsigned NumConcat = MaskNumElts / SrcNumElts;
      bool IsConcat = true;
      SmallVector<int, 8> ConcatSrcs(NumConcat, -1);
      for (unsigned i = 0; i != MaskNumElts; ++i) {
        int Idx = Mask[i];
        if (Idx < 0)
          continue;
        // Ensure the indices in each SrcVT sized piece are sequential and that
        // the same source is used for the whole piece.
        if ((Idx % SrcNumElts != (i % SrcNumElts)) ||
            (ConcatSrcs[i / SrcNumElts] >= 0 &&
             ConcatSrcs[i / SrcNumElts] != (int)(Idx / SrcNumElts))) {
          IsConcat = false;
          break;
        }
        // Remember which source this index came from.
        ConcatSrcs[i / SrcNumElts] = Idx / SrcNumElts;
      }
 
      // The shuffle is concatenating multiple vectors together. Just emit
      // a CONCAT_VECTORS operation.
      if (IsConcat) {
        SmallVector<SDValue, 8> ConcatOps;
        for (auto Src : ConcatSrcs) {
          if (Src < 0)
            ConcatOps.push_back(DAG.getUNDEF(SrcVT));
          else if (Src == 0)
            ConcatOps.push_back(Src1);
          else
            ConcatOps.push_back(Src2);
        }
        setValue(&I, DAG.getNode(ISD::CONCAT_VECTORS, DL, VT, ConcatOps));
        return;
      }
    }
 
    unsigned PaddedMaskNumElts = alignTo(MaskNumElts, SrcNumElts);
    unsigned NumConcat = PaddedMaskNumElts / SrcNumElts;
    EVT PaddedVT = EVT::getVectorVT(*DAG.getContext(), VT.getScalarType(),
                                    PaddedMaskNumElts);
 
    // Pad both vectors with undefs to make them the same length as the mask.
    SDValue UndefVal = DAG.getUNDEF(SrcVT);
 
    SmallVector<SDValue, 8> MOps1(NumConcat, UndefVal);
    SmallVector<SDValue, 8> MOps2(NumConcat, UndefVal);
    MOps1[0] = Src1;
    MOps2[0] = Src2;
 
    Src1 = DAG.getNode(ISD::CONCAT_VECTORS, DL, PaddedVT, MOps1);
    Src2 = DAG.getNode(ISD::CONCAT_VECTORS, DL, PaddedVT, MOps2);
 
    // Readjust mask for new input vector length.
    SmallVector<int, 8> MappedOps(PaddedMaskNumElts, -1);
    for (unsigned i = 0; i != MaskNumElts; ++i) {
      int Idx = Mask[i];
      if (Idx >= (int)SrcNumElts)
        Idx -= SrcNumElts - PaddedMaskNumElts;
      MappedOps[i] = Idx;
    }
 
    SDValue Result = DAG.getVectorShuffle(PaddedVT, DL, Src1, Src2, MappedOps);
 
    // If the concatenated vector was padded, extract a subvector with the
    // correct number of elements.
    if (MaskNumElts != PaddedMaskNumElts)
      Result = DAG.getNode(ISD::EXTRACT_SUBVECTOR, DL, VT, Result,
                           DAG.getVectorIdxConstant(0, DL));
 
    setValue(&I, Result);
    return;
  }
 
  if (SrcNumElts > MaskNumElts) {
    // Analyze the access pattern of the vector to see if we can extract
    // two subvectors and do the shuffle.
    int StartIdx[2] = { -1, -1 };  // StartIdx to extract from
    bool CanExtract = true;
    for (int Idx : Mask) {
      unsigned Input = 0;
      if (Idx < 0)
        continue;
 
      if (Idx >= (int)SrcNumElts) {
        Input = 1;
        Idx -= SrcNumElts;
      }
 
      // If all the indices come from the same MaskNumElts sized portion of
      // the sources we can use extract. Also make sure the extract wouldn't
      // extract past the end of the source.
      int NewStartIdx = alignDown(Idx, MaskNumElts);
      if (NewStartIdx + MaskNumElts > SrcNumElts ||
          (StartIdx[Input] >= 0 && StartIdx[Input] != NewStartIdx))
        CanExtract = false;
      // Make sure we always update StartIdx as we use it to track if all
      // elements are undef.
      StartIdx[Input] = NewStartIdx;
    }
 
    if (StartIdx[0] < 0 && StartIdx[1] < 0) {
      setValue(&I, DAG.getUNDEF(VT)); // Vectors are not used.
      return;
    }
    if (CanExtract) {
      // Extract appropriate subvector and generate a vector shuffle
      for (unsigned Input = 0; Input < 2; ++Input) {
        SDValue &Src = Input == 0 ? Src1 : Src2;
        if (StartIdx[Input] < 0)
          Src = DAG.getUNDEF(VT);
        else {
          Src = DAG.getNode(ISD::EXTRACT_SUBVECTOR, DL, VT, Src,
                            DAG.getVectorIdxConstant(StartIdx[Input], DL));
        }
      }
 
      // Calculate new mask.
      SmallVector<int, 8> MappedOps(Mask.begin(), Mask.end());
      for (int &Idx : MappedOps) {
        if (Idx >= (int)SrcNumElts)
          Idx -= SrcNumElts + StartIdx[1] - MaskNumElts;
        else if (Idx >= 0)
          Idx -= StartIdx[0];
      }
 
      setValue(&I, DAG.getVectorShuffle(VT, DL, Src1, Src2, MappedOps));
      return;
    }
  }
 
  // We can't use either concat vectors or extract subvectors so fall back to
  // replacing the shuffle with extract and build vector.
  // to insert and build vector.
  EVT EltVT = VT.getVectorElementType();
  SmallVector<SDValue,8> Ops;
  for (int Idx : Mask) {
    SDValue Res;
 
    if (Idx < 0) {
      Res = DAG.getUNDEF(EltVT);
    } else {
      SDValue &Src = Idx < (int)SrcNumElts ? Src1 : Src2;
      if (Idx >= (int)SrcNumElts) Idx -= SrcNumElts;
 
      Res = DAG.getNode(ISD::EXTRACT_VECTOR_ELT, DL, EltVT, Src,
                        DAG.getVectorIdxConstant(Idx, DL));
    }
 
    Ops.push_back(Res);
  }
 
  setValue(&I, DAG.getBuildVector(VT, DL, Ops));
}
 
void SelectionDAGBuilder::visitInsertValue(const User &I) {
  ArrayRef<unsigned> Indices;
  if (const InsertValueInst *IV = dyn_cast<InsertValueInst>(&I))
    Indices = IV->getIndices();
  else
    Indices = cast<ConstantExpr>(&I)->getIndices();
 
  const Value *Op0 = I.getOperand(0);
  const Value *Op1 = I.getOperand(1);
  Type *AggTy = I.getType();
  Type *ValTy = Op1->getType();
  bool IntoUndef = isa<UndefValue>(Op0);
  bool FromUndef = isa<UndefValue>(Op1);
 
  unsigned LinearIndex = ComputeLinearIndex(AggTy, Indices);
 
  const TargetLowering &TLI = DAG.getTargetLoweringInfo();
  SmallVector<EVT, 4> AggValueVTs;
  ComputeValueVTs(TLI, DAG.getDataLayout(), AggTy, AggValueVTs);
  SmallVector<EVT, 4> ValValueVTs;
  ComputeValueVTs(TLI, DAG.getDataLayout(), ValTy, ValValueVTs);
 
  unsigned NumAggValues = AggValueVTs.size();
  unsigned NumValValues = ValValueVTs.size();
  SmallVector<SDValue, 4> Values(NumAggValues);
 
  // Ignore an insertvalue that produces an empty object
  if (!NumAggValues) {
    setValue(&I, DAG.getUNDEF(MVT(MVT::Other)));
    return;
  }
 
  SDValue Agg = getValue(Op0);
  unsigned i = 0;
  // Copy the beginning value(s) from the original aggregate.
  for (; i != LinearIndex; ++i)
    Values[i] = IntoUndef ? DAG.getUNDEF(AggValueVTs[i]) :
                SDValue(Agg.getNode(), Agg.getResNo() + i);
  // Copy values from the inserted value(s).
  if (NumValValues) {
    SDValue Val = getValue(Op1);
    for (; i != LinearIndex + NumValValues; ++i)
      Values[i] = FromUndef ? DAG.getUNDEF(AggValueVTs[i]) :
                  SDValue(Val.getNode(), Val.getResNo() + i - LinearIndex);
  }
  // Copy remaining value(s) from the original aggregate.
  for (; i != NumAggValues; ++i)
    Values[i] = IntoUndef ? DAG.getUNDEF(AggValueVTs[i]) :
                SDValue(Agg.getNode(), Agg.getResNo() + i);
 
  setValue(&I, DAG.getNode(ISD::MERGE_VALUES, getCurSDLoc(),
                           DAG.getVTList(AggValueVTs), Values));
}
 
void SelectionDAGBuilder::visitExtractValue(const User &I) {
  ArrayRef<unsigned> Indices;
  if (const ExtractValueInst *EV = dyn_cast<ExtractValueInst>(&I))
    Indices = EV->getIndices();
  else
    Indices = cast<ConstantExpr>(&I)->getIndices();
 
  const Value *Op0 = I.getOperand(0);
  Type *AggTy = Op0->getType();
  Type *ValTy = I.getType();
  bool OutOfUndef = isa<UndefValue>(Op0);
 
  unsigned LinearIndex = ComputeLinearIndex(AggTy, Indices);
 
  const TargetLowering &TLI = DAG.getTargetLoweringInfo();
  SmallVector<EVT, 4> ValValueVTs;
  ComputeValueVTs(TLI, DAG.getDataLayout(), ValTy, ValValueVTs);
 
  unsigned NumValValues = ValValueVTs.size();
 
  // Ignore a extractvalue that produces an empty object
  if (!NumValValues) {
    setValue(&I, DAG.getUNDEF(MVT(MVT::Other)));
    return;
  }
 
  SmallVector<SDValue, 4> Values(NumValValues);
 
  SDValue Agg = getValue(Op0);
  // Copy out the selected value(s).
  for (unsigned i = LinearIndex; i != LinearIndex + NumValValues; ++i)
    Values[i - LinearIndex] =
      OutOfUndef ?
        DAG.getUNDEF(Agg.getNode()->getValueType(Agg.getResNo() + i)) :
        SDValue(Agg.getNode(), Agg.getResNo() + i);
 
  setValue(&I, DAG.getNode(ISD::MERGE_VALUES, getCurSDLoc(),
                           DAG.getVTList(ValValueVTs), Values));
}
 
void SelectionDAGBuilder::visitGetElementPtr(const User &I) {
  Value *Op0 = I.getOperand(0);
  // Note that the pointer operand may be a vector of pointers. Take the scalar
  // element which holds a pointer.
  unsigned AS = Op0->getType()->getScalarType()->getPointerAddressSpace();
  SDValue N = getValue(Op0);
  SDLoc dl = getCurSDLoc();
  auto &TLI = DAG.getTargetLoweringInfo();
 
  // Normalize Vector GEP - all scalar operands should be converted to the
  // splat vector.
  bool IsVectorGEP = I.getType()->isVectorTy();
  ElementCount VectorElementCount =
      IsVectorGEP ? cast<VectorType>(I.getType())->getElementCount()
                  : ElementCount::getFixed(0);
 
  if (IsVectorGEP && !N.getValueType().isVector()) {
    LLVMContext &Context = *DAG.getContext();
    EVT VT = EVT::getVectorVT(Context, N.getValueType(), VectorElementCount);
    if (VectorElementCount.isScalable())
      N = DAG.getSplatVector(VT, dl, N);
    else
      N = DAG.getSplatBuildVector(VT, dl, N);
  }
 
  for (gep_type_iterator GTI = gep_type_begin(&I), E = gep_type_end(&I);
       GTI != E; ++GTI) {
    const Value *Idx = GTI.getOperand();
    if (StructType *StTy = GTI.getStructTypeOrNull()) {
      unsigned Field = cast<Constant>(Idx)->getUniqueInteger().getZExtValue();
      if (Field) {
        // N = N + Offset
        uint64_t Offset = DL->getStructLayout(StTy)->getElementOffset(Field);
 
        // In an inbounds GEP with an offset that is nonnegative even when
        // interpreted as signed, assume there is no unsigned overflow.
        SDNodeFlags Flags;
        if (int64_t(Offset) >= 0 && cast<GEPOperator>(I).isInBounds())
          Flags.setNoUnsignedWrap(true);
 
        N = DAG.getNode(ISD::ADD, dl, N.getValueType(), N,
                        DAG.getConstant(Offset, dl, N.getValueType()), Flags);
      }
    } else {
      // IdxSize is the width of the arithmetic according to IR semantics.
      // In SelectionDAG, we may prefer to do arithmetic in a wider bitwidth
      // (and fix up the result later).
      unsigned IdxSize = DAG.getDataLayout().getIndexSizeInBits(AS);
      MVT IdxTy = MVT::getIntegerVT(IdxSize);
      TypeSize ElementSize = DL->getTypeAllocSize(GTI.getIndexedType());
      // We intentionally mask away the high bits here; ElementSize may not
      // fit in IdxTy.
      APInt ElementMul(IdxSize, ElementSize.getKnownMinSize());
      bool ElementScalable = ElementSize.isScalable();
 
      // If this is a scalar constant or a splat vector of constants,
      // handle it quickly.
      const auto *C = dyn_cast<Constant>(Idx);
      if (C && isa<VectorType>(C->getType()))
        C = C->getSplatValue();
 
      const auto *CI = dyn_cast_or_null<ConstantInt>(C);
      if (CI && CI->isZero())
        continue;
      if (CI && !ElementScalable) {
        APInt Offs = ElementMul * CI->getValue().sextOrTrunc(IdxSize);
        LLVMContext &Context = *DAG.getContext();
        SDValue OffsVal;
        if (IsVectorGEP)
          OffsVal = DAG.getConstant(
              Offs, dl, EVT::getVectorVT(Context, IdxTy, VectorElementCount));
        else
          OffsVal = DAG.getConstant(Offs, dl, IdxTy);
 
        // In an inbounds GEP with an offset that is nonnegative even when
        // interpreted as signed, assume there is no unsigned overflow.
        SDNodeFlags Flags;
        if (Offs.isNonNegative() && cast<GEPOperator>(I).isInBounds())
          Flags.setNoUnsignedWrap(true);
 
        OffsVal = DAG.getSExtOrTrunc(OffsVal, dl, N.getValueType());
 
        N = DAG.getNode(ISD::ADD, dl, N.getValueType(), N, OffsVal, Flags);
        continue;
      }
 
      // N = N + Idx * ElementMul;
      SDValue IdxN = getValue(Idx);
 
      if (!IdxN.getValueType().isVector() && IsVectorGEP) {
        EVT VT = EVT::getVectorVT(*Context, IdxN.getValueType(),
                                  VectorElementCount);
        if (VectorElementCount.isScalable())
          IdxN = DAG.getSplatVector(VT, dl, IdxN);
        else
          IdxN = DAG.getSplatBuildVector(VT, dl, IdxN);
      }
 
      // If the index is smaller or larger than intptr_t, truncate or extend
      // it.
      IdxN = DAG.getSExtOrTrunc(IdxN, dl, N.getValueType());
 
      if (ElementScalable) {
        EVT VScaleTy = N.getValueType().getScalarType();
        SDValue VScale = DAG.getNode(
            ISD::VSCALE, dl, VScaleTy,
            DAG.getConstant(ElementMul.getZExtValue(), dl, VScaleTy));
        if (IsVectorGEP)
          VScale = DAG.getSplatVector(N.getValueType(), dl, VScale);
        IdxN = DAG.getNode(ISD::MUL, dl, N.getValueType(), IdxN, VScale);
      } else {
        // If this is a multiply by a power of two, turn it into a shl
        // immediately.  This is a very common case.
        if (ElementMul != 1) {
          if (ElementMul.isPowerOf2()) {
            unsigned Amt = ElementMul.logBase2();
            IdxN = DAG.getNode(ISD::SHL, dl,
                               N.getValueType(), IdxN,
                               DAG.getConstant(Amt, dl, IdxN.getValueType()));
          } else {
            SDValue Scale = DAG.getConstant(ElementMul.getZExtValue(), dl,
                                            IdxN.getValueType());
            IdxN = DAG.getNode(ISD::MUL, dl,
                               N.getValueType(), IdxN, Scale);
          }
        }
      }
 
      N = DAG.getNode(ISD::ADD, dl,
                      N.getValueType(), N, IdxN);
    }
  }
 
  MVT PtrTy = TLI.getPointerTy(DAG.getDataLayout(), AS);
  MVT PtrMemTy = TLI.getPointerMemTy(DAG.getDataLayout(), AS);
  if (IsVectorGEP) {
    PtrTy = MVT::getVectorVT(PtrTy, VectorElementCount);
    PtrMemTy = MVT::getVectorVT(PtrMemTy, VectorElementCount);
  }
 
  if (PtrMemTy != PtrTy && !cast<GEPOperator>(I).isInBounds())
    N = DAG.getPtrExtendInReg(N, dl, PtrMemTy);
 
  setValue(&I, N);
}
 
void SelectionDAGBuilder::visitAlloca(const AllocaInst &I) {
  // If this is a fixed sized alloca in the entry block of the function,
  // allocate it statically on the stack.
  if (FuncInfo.StaticAllocaMap.count(&I))
    return;   // getValue will auto-populate this.
 
  SDLoc dl = getCurSDLoc();
  Type *Ty = I.getAllocatedType();
  const TargetLowering &TLI = DAG.getTargetLoweringInfo();
  auto &DL = DAG.getDataLayout();
  uint64_t TySize = DL.getTypeAllocSize(Ty);
  MaybeAlign Alignment = std::max(DL.getPrefTypeAlign(Ty), I.getAlign());
 
  SDValue AllocSize = getValue(I.getArraySize());
 
  EVT IntPtr = TLI.getPointerTy(DAG.getDataLayout(), DL.getAllocaAddrSpace());
  if (AllocSize.getValueType() != IntPtr)
    AllocSize = DAG.getZExtOrTrunc(AllocSize, dl, IntPtr);
 
  AllocSize = DAG.getNode(ISD::MUL, dl, IntPtr,
                          AllocSize,
                          DAG.getConstant(TySize, dl, IntPtr));
 
  // Handle alignment.  If the requested alignment is less than or equal to
  // the stack alignment, ignore it.  If the size is greater than or equal to
  // the stack alignment, we note this in the DYNAMIC_STACKALLOC node.
  Align StackAlign = DAG.getSubtarget().getFrameLowering()->getStackAlign();
  if (*Alignment <= StackAlign)
    Alignment = None;
 
  const uint64_t StackAlignMask = StackAlign.value() - 1U;
  // Round the size of the allocation up to the stack alignment size
  // by add SA-1 to the size. This doesn't overflow because we're computing
  // an address inside an alloca.
  SDNodeFlags Flags;
  Flags.setNoUnsignedWrap(true);
  AllocSize = DAG.getNode(ISD::ADD, dl, AllocSize.getValueType(), AllocSize,
                          DAG.getConstant(StackAlignMask, dl, IntPtr), Flags);
 
  // Mask out the low bits for alignment purposes.
  AllocSize = DAG.getNode(ISD::AND, dl, AllocSize.getValueType(), AllocSize,
                          DAG.getConstant(~StackAlignMask, dl, IntPtr));
 
  SDValue Ops[] = {
      getRoot(), AllocSize,
      DAG.getConstant(Alignment ? Alignment->value() : 0, dl, IntPtr)};
  SDVTList VTs = DAG.getVTList(AllocSize.getValueType(), MVT::Other);
  SDValue DSA = DAG.getNode(ISD::DYNAMIC_STACKALLOC, dl, VTs, Ops);
  setValue(&I, DSA);
  DAG.setRoot(DSA.getValue(1));
 
  assert(FuncInfo.MF->getFrameInfo().hasVarSizedObjects());
}
 
void SelectionDAGBuilder::visitLoad(const LoadInst &I) {
  if (I.isAtomic())
    return visitAtomicLoad(I);
 
  const TargetLowering &TLI = DAG.getTargetLoweringInfo();
  const Value *SV = I.getOperand(0);
  if (TLI.supportSwiftError()) {
    // Swifterror values can come from either a function parameter with
    // swifterror attribute or an alloca with swifterror attribute.
    if (const Argument *Arg = dyn_cast<Argument>(SV)) {
      if (Arg->hasSwiftErrorAttr())
        return visitLoadFromSwiftError(I);
    }
 
    if (const AllocaInst *Alloca = dyn_cast<AllocaInst>(SV)) {
      if (Alloca->isSwiftError())
        return visitLoadFromSwiftError(I);
    }
  }
 
  SDValue Ptr = getValue(SV);
 
  Type *Ty = I.getType();
  Align Alignment = I.getAlign();
 
  AAMDNodes AAInfo = I.getAAMetadata();
  const MDNode *Ranges = I.getMetadata(LLVMContext::MD_range);
 
  SmallVector<EVT, 4> ValueVTs, MemVTs;
  SmallVector<uint64_t, 4> Offsets;
  ComputeValueVTs(TLI, DAG.getDataLayout(), Ty, ValueVTs, &MemVTs, &Offsets);
  unsigned NumValues = ValueVTs.size();
  if (NumValues == 0)
    return;
 
  bool isVolatile = I.isVolatile();
 
  SDValue Root;
  bool ConstantMemory = false;
  if (isVolatile)
    // Serialize volatile loads with other side effects.
    Root = getRoot();
  else if (NumValues > MaxParallelChains)
    Root = getMemoryRoot();
  else if (AA &&
           AA->pointsToConstantMemory(MemoryLocation(
               SV,
               LocationSize::precise(DAG.getDataLayout().getTypeStoreSize(Ty)),
               AAInfo))) {
    // Do not serialize (non-volatile) loads of constant memory with anything.
    Root = DAG.getEntryNode();
    ConstantMemory = true;
  } else {
    // Do not serialize non-volatile loads against each other.
    Root = DAG.getRoot();
  }
 
  SDLoc dl = getCurSDLoc();
 
  if (isVolatile)
    Root = TLI.prepareVolatileOrAtomicLoad(Root, dl, DAG);
 
  // An aggregate load cannot wrap around the address space, so offsets to its
  // parts don't wrap either.
  SDNodeFlags Flags;
  Flags.setNoUnsignedWrap(true);
 
  SmallVector<SDValue, 4> Values(NumValues);
  SmallVector<SDValue, 4> Chains(std::min(MaxParallelChains, NumValues));
  EVT PtrVT = Ptr.getValueType();
 
  MachineMemOperand::Flags MMOFlags
    = TLI.getLoadMemOperandFlags(I, DAG.getDataLayout());
 
  unsigned ChainI = 0;
  for (unsigned i = 0; i != NumValues; ++i, ++ChainI) {
    // Serializing loads here may result in excessive register pressure, and
    // TokenFactor places arbitrary choke points on the scheduler. SD scheduling
    // could recover a bit by hoisting nodes upward in the chain by recognizing
    // they are side-effect free or do not alias. The optimizer should really
    // avoid this case by converting large object/array copies to llvm.memcpy
    // (MaxParallelChains should always remain as failsafe).
    if (ChainI == MaxParallelChains) {
      assert(PendingLoads.empty() && "PendingLoads must be serialized first");
      SDValue Chain = DAG.getNode(ISD::TokenFactor, dl, MVT::Other,
                                  makeArrayRef(Chains.data(), ChainI));
      Root = Chain;
      ChainI = 0;
    }
    SDValue A = DAG.getNode(ISD::ADD, dl,
                            PtrVT, Ptr,
                            DAG.getConstant(Offsets[i], dl, PtrVT),
                            Flags);
 
    SDValue L = DAG.getLoad(MemVTs[i], dl, Root, A,
                            MachinePointerInfo(SV, Offsets[i]), Alignment,
                            MMOFlags, AAInfo, Ranges);
    Chains[ChainI] = L.getValue(1);
 
    if (MemVTs[i] != ValueVTs[i])
      L = DAG.getZExtOrTrunc(L, dl, ValueVTs[i]);
 
    Values[i] = L;
  }
 
  if (!ConstantMemory) {
    SDValue Chain = DAG.getNode(ISD::TokenFactor, dl, MVT::Other,
                                makeArrayRef(Chains.data(), ChainI));
    if (isVolatile)
      DAG.setRoot(Chain);
    else
      PendingLoads.push_back(Chain);
  }
 
  setValue(&I, DAG.getNode(ISD::MERGE_VALUES, dl,
                           DAG.getVTList(ValueVTs), Values));
}
 
void SelectionDAGBuilder::visitStoreToSwiftError(const StoreInst &I) {
  assert(DAG.getTargetLoweringInfo().supportSwiftError() &&
         "call visitStoreToSwiftError when backend supports swifterror");
 
  SmallVector<EVT, 4> ValueVTs;
  SmallVector<uint64_t, 4> Offsets;
  const Value *SrcV = I.getOperand(0);
  ComputeValueVTs(DAG.getTargetLoweringInfo(), DAG.getDataLayout(),
                  SrcV->getType(), ValueVTs, &Offsets);
  assert(ValueVTs.size() == 1 && Offsets[0] == 0 &&
         "expect a single EVT for swifterror");
 
  SDValue Src = getValue(SrcV);
  // Create a virtual register, then update the virtual register.
  Register VReg =
      SwiftError.getOrCreateVRegDefAt(&I, FuncInfo.MBB, I.getPointerOperand());
  // Chain, DL, Reg, N or Chain, DL, Reg, N, Glue
  // Chain can be getRoot or getControlRoot.
  SDValue CopyNode = DAG.getCopyToReg(getRoot(), getCurSDLoc(), VReg,
                                      SDValue(Src.getNode(), Src.getResNo()));
  DAG.setRoot(CopyNode);
}
 
void SelectionDAGBuilder::visitLoadFromSwiftError(const LoadInst &I) {
  assert(DAG.getTargetLoweringInfo().supportSwiftError() &&
         "call visitLoadFromSwiftError when backend supports swifterror");
 
  assert(!I.isVolatile() &&
         !I.hasMetadata(LLVMContext::MD_nontemporal) &&
         !I.hasMetadata(LLVMContext::MD_invariant_load) &&
         "Support volatile, non temporal, invariant for load_from_swift_error");
 
  const Value *SV = I.getOperand(0);
  Type *Ty = I.getType();
  assert(
      (!AA ||
       !AA->pointsToConstantMemory(MemoryLocation(
           SV, LocationSize::precise(DAG.getDataLayout().getTypeStoreSize(Ty)),
           I.getAAMetadata()))) &&
      "load_from_swift_error should not be constant memory");
 
  SmallVector<EVT, 4> ValueVTs;
  SmallVector<uint64_t, 4> Offsets;
  ComputeValueVTs(DAG.getTargetLoweringInfo(), DAG.getDataLayout(), Ty,
                  ValueVTs, &Offsets);
  assert(ValueVTs.size() == 1 && Offsets[0] == 0 &&
         "expect a single EVT for swifterror");
 
  // Chain, DL, Reg, VT, Glue or Chain, DL, Reg, VT
  SDValue L = DAG.getCopyFromReg(
      getRoot(), getCurSDLoc(),
      SwiftError.getOrCreateVRegUseAt(&I, FuncInfo.MBB, SV), ValueVTs[0]);
 
  setValue(&I, L);
}
 
void SelectionDAGBuilder::visitStore(const StoreInst &I) {
  if (I.isAtomic())
    return visitAtomicStore(I);
 
  const Value *SrcV = I.getOperand(0);
  const Value *PtrV = I.getOperand(1);
 
  const TargetLowering &TLI = DAG.getTargetLoweringInfo();
  if (TLI.supportSwiftError()) {
    // Swifterror values can come from either a function parameter with
    // swifterror attribute or an alloca with swifterror attribute.
    if (const Argument *Arg = dyn_cast<Argument>(PtrV)) {
      if (Arg->hasSwiftErrorAttr())
        return visitStoreToSwiftError(I);
    }
 
    if (const AllocaInst *Alloca = dyn_cast<AllocaInst>(PtrV)) {
      if (Alloca->isSwiftError())
        return visitStoreToSwiftError(I);
    }
  }
 
  SmallVector<EVT, 4> ValueVTs, MemVTs;
  SmallVector<uint64_t, 4> Offsets;
  ComputeValueVTs(DAG.getTargetLoweringInfo(), DAG.getDataLayout(),
                  SrcV->getType(), ValueVTs, &MemVTs, &Offsets);
  unsigned NumValues = ValueVTs.size();
  if (NumValues == 0)
    return;
 
  // Get the lowered operands. Note that we do this after
  // checking if NumResults is zero, because with zero results
  // the operands won't have values in the map.
  SDValue Src = getValue(SrcV);
  SDValue Ptr = getValue(PtrV);
 
  SDValue Root = I.isVolatile() ? getRoot() : getMemoryRoot();
  SmallVector<SDValue, 4> Chains(std::min(MaxParallelChains, NumValues));
  SDLoc dl = getCurSDLoc();
  Align Alignment = I.getAlign();
  AAMDNodes AAInfo = I.getAAMetadata();
 
  auto MMOFlags = TLI.getStoreMemOperandFlags(I, DAG.getDataLayout());
 
  // An aggregate load cannot wrap around the address space, so offsets to its
  // parts don't wrap either.
  SDNodeFlags Flags;
  Flags.setNoUnsignedWrap(true);
 
  unsigned ChainI = 0;
  for (unsigned i = 0; i != NumValues; ++i, ++ChainI) {
    // See visitLoad comments.
    if (ChainI == MaxParallelChains) {
      SDValue Chain = DAG.getNode(ISD::TokenFactor, dl, MVT::Other,
                                  makeArrayRef(Chains.data(), ChainI));
      Root = Chain;
      ChainI = 0;
    }
    SDValue Add =
        DAG.getMemBasePlusOffset(Ptr, TypeSize::Fixed(Offsets[i]), dl, Flags);
    SDValue Val = SDValue(Src.getNode(), Src.getResNo() + i);
    if (MemVTs[i] != ValueVTs[i])
      Val = DAG.getPtrExtOrTrunc(Val, dl, MemVTs[i]);
    SDValue St =
        DAG.getStore(Root, dl, Val, Add, MachinePointerInfo(PtrV, Offsets[i]),
                     Alignment, MMOFlags, AAInfo);
    Chains[ChainI] = St;
  }
 
  SDValue StoreNode = DAG.getNode(ISD::TokenFactor, dl, MVT::Other,
                                  makeArrayRef(Chains.data(), ChainI));
  DAG.setRoot(StoreNode);
}
 
void SelectionDAGBuilder::visitMaskedStore(const CallInst &I,
                                           bool IsCompressing) {
  SDLoc sdl = getCurSDLoc();
 
  auto getMaskedStoreOps = [&](Value *&Ptr, Value *&Mask, Value *&Src0,
                               MaybeAlign &Alignment) {
    // llvm.masked.store.*(Src0, Ptr, alignment, Mask)
    Src0 = I.getArgOperand(0);
    Ptr = I.getArgOperand(1);
    Alignment = cast<ConstantInt>(I.getArgOperand(2))->getMaybeAlignValue();
    Mask = I.getArgOperand(3);
  };
  auto getCompressingStoreOps = [&](Value *&Ptr, Value *&Mask, Value *&Src0,
                                    MaybeAlign &Alignment) {
    // llvm.masked.compressstore.*(Src0, Ptr, Mask)
    Src0 = I.getArgOperand(0);
    Ptr = I.getArgOperand(1);
    Mask = I.getArgOperand(2);
    Alignment = None;
  };
 
  Value  *PtrOperand, *MaskOperand, *Src0Operand;
  MaybeAlign Alignment;
  if (IsCompressing)
    getCompressingStoreOps(PtrOperand, MaskOperand, Src0Operand, Alignment);
  else
    getMaskedStoreOps(PtrOperand, MaskOperand, Src0Operand, Alignment);
 
  SDValue Ptr = getValue(PtrOperand);
  SDValue Src0 = getValue(Src0Operand);
  SDValue Mask = getValue(MaskOperand);
  SDValue Offset = DAG.getUNDEF(Ptr.getValueType());
 
  EVT VT = Src0.getValueType();
  if (!Alignment)
    Alignment = DAG.getEVTAlign(VT);
 
  MachineMemOperand *MMO = DAG.getMachineFunction().getMachineMemOperand(
      MachinePointerInfo(PtrOperand), MachineMemOperand::MOStore,
      // TODO: Make MachineMemOperands aware of scalable
      // vectors.
      VT.getStoreSize().getKnownMinSize(), *Alignment, I.getAAMetadata());
  SDValue StoreNode =
      DAG.getMaskedStore(getMemoryRoot(), sdl, Src0, Ptr, Offset, Mask, VT, MMO,
                         ISD::UNINDEXED, false /* Truncating */, IsCompressing);
  DAG.setRoot(StoreNode);
  setValue(&I, StoreNode);
}
 
// Get a uniform base for the Gather/Scatter intrinsic.
// The first argument of the Gather/Scatter intrinsic is a vector of pointers.
// We try to represent it as a base pointer + vector of indices.
// Usually, the vector of pointers comes from a 'getelementptr' instruction.
// The first operand of the GEP may be a single pointer or a vector of pointers
// Example:
//   %gep.ptr = getelementptr i32, <8 x i32*> %vptr, <8 x i32> %ind
//  or
//   %gep.ptr = getelementptr i32, i32* %ptr,        <8 x i32> %ind
// %res = call <8 x i32> @llvm.masked.gather.v8i32(<8 x i32*> %gep.ptr, ..
//
// When the first GEP operand is a single pointer - it is the uniform base we
// are looking for. If first operand of the GEP is a splat vector - we
// extract the splat value and use it as a uniform base.
// In all other cases the function returns 'false'.
static bool getUniformBase(const Value *Ptr, SDValue &Base, SDValue &Index,
                           ISD::MemIndexType &IndexType, SDValue &Scale,
                           SelectionDAGBuilder *SDB, const BasicBlock *CurBB) {
  SelectionDAG& DAG = SDB->DAG;
  const TargetLowering &TLI = DAG.getTargetLoweringInfo();
  const DataLayout &DL = DAG.getDataLayout();
 
  assert(Ptr->getType()->isVectorTy() && "Unexpected pointer type");
 
  // Handle splat constant pointer.
  if (auto *C = dyn_cast<Constant>(Ptr)) {
    C = C->getSplatValue();
    if (!C)
      return false;
 
    Base = SDB->getValue(C);
 
    ElementCount NumElts = cast<VectorType>(Ptr->getType())->getElementCount();
    EVT VT = EVT::getVectorVT(*DAG.getContext(), TLI.getPointerTy(DL), NumElts);
    Index = DAG.getConstant(0, SDB->getCurSDLoc(), VT);
    IndexType = ISD::SIGNED_SCALED;
    Scale = DAG.getTargetConstant(1, SDB->getCurSDLoc(), TLI.getPointerTy(DL));
    return true;
  }
 
  const GetElementPtrInst *GEP = dyn_cast<GetElementPtrInst>(Ptr);
  if (!GEP || GEP->getParent() != CurBB)
    return false;
 
  if (GEP->getNumOperands() != 2)
    return false;
 
  const Value *BasePtr = GEP->getPointerOperand();
  const Value *IndexVal = GEP->getOperand(GEP->getNumOperands() - 1);
 
  // Make sure the base is scalar and the index is a vector.
  if (BasePtr->getType()->isVectorTy() || !IndexVal->getType()->isVectorTy())
    return false;