ARMISelLowering.cpp

Go to the documentation of this file.

//===- ARMISelLowering.cpp - ARM DAG Lowering Implementation --------------===//
//
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
// See https://llvm.org/LICENSE.txt for license information.
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
//
//===----------------------------------------------------------------------===//
//
// This file defines the interfaces that ARM uses to lower LLVM code into a
// selection DAG.
//
//===----------------------------------------------------------------------===//
 
#include "ARMISelLowering.h"
#include "ARMBaseInstrInfo.h"
#include "ARMBaseRegisterInfo.h"
#include "ARMCallingConv.h"
#include "ARMConstantPoolValue.h"
#include "ARMMachineFunctionInfo.h"
#include "ARMPerfectShuffle.h"
#include "ARMRegisterInfo.h"
#include "ARMSelectionDAGInfo.h"
#include "ARMSubtarget.h"
#include "ARMTargetTransformInfo.h"
#include "MCTargetDesc/ARMAddressingModes.h"
#include "MCTargetDesc/ARMBaseInfo.h"
#include "Utils/ARMBaseInfo.h"
#include "llvm/ADT/APFloat.h"
#include "llvm/ADT/APInt.h"
#include "llvm/ADT/ArrayRef.h"
#include "llvm/ADT/BitVector.h"
#include "llvm/ADT/DenseMap.h"
#include "llvm/ADT/STLExtras.h"
#include "llvm/ADT/SmallPtrSet.h"
#include "llvm/ADT/SmallVector.h"
#include "llvm/ADT/Statistic.h"
#include "llvm/ADT/StringExtras.h"
#include "llvm/ADT/StringRef.h"
#include "llvm/ADT/StringSwitch.h"
#include "llvm/ADT/Triple.h"
#include "llvm/ADT/Twine.h"
#include "llvm/Analysis/VectorUtils.h"
#include "llvm/CodeGen/CallingConvLower.h"
#include "llvm/CodeGen/ISDOpcodes.h"
#include "llvm/CodeGen/IntrinsicLowering.h"
#include "llvm/CodeGen/MachineBasicBlock.h"
#include "llvm/CodeGen/MachineConstantPool.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/MachineOperand.h"
#include "llvm/CodeGen/MachineRegisterInfo.h"
#include "llvm/CodeGen/RuntimeLibcalls.h"
#include "llvm/CodeGen/SelectionDAG.h"
#include "llvm/CodeGen/SelectionDAGAddressAnalysis.h"
#include "llvm/CodeGen/SelectionDAGNodes.h"
#include "llvm/CodeGen/TargetInstrInfo.h"
#include "llvm/CodeGen/TargetLowering.h"
#include "llvm/CodeGen/TargetOpcodes.h"
#include "llvm/CodeGen/TargetRegisterInfo.h"
#include "llvm/CodeGen/TargetSubtargetInfo.h"
#include "llvm/CodeGen/ValueTypes.h"
#include "llvm/IR/Attributes.h"
#include "llvm/IR/CallingConv.h"
#include "llvm/IR/Constant.h"
#include "llvm/IR/Constants.h"
#include "llvm/IR/DataLayout.h"
#include "llvm/IR/DebugLoc.h"
#include "llvm/IR/DerivedTypes.h"
#include "llvm/IR/Function.h"
#include "llvm/IR/GlobalAlias.h"
#include "llvm/IR/GlobalValue.h"
#include "llvm/IR/GlobalVariable.h"
#include "llvm/IR/IRBuilder.h"
#include "llvm/IR/InlineAsm.h"
#include "llvm/IR/Instruction.h"
#include "llvm/IR/Instructions.h"
#include "llvm/IR/IntrinsicInst.h"
#include "llvm/IR/Intrinsics.h"
#include "llvm/IR/IntrinsicsARM.h"
#include "llvm/IR/Module.h"
#include "llvm/IR/PatternMatch.h"
#include "llvm/IR/Type.h"
#include "llvm/IR/User.h"
#include "llvm/IR/Value.h"
#include "llvm/MC/MCInstrDesc.h"
#include "llvm/MC/MCInstrItineraries.h"
#include "llvm/MC/MCRegisterInfo.h"
#include "llvm/MC/MCSchedule.h"
#include "llvm/Support/AtomicOrdering.h"
#include "llvm/Support/BranchProbability.h"
#include "llvm/Support/Casting.h"
#include "llvm/Support/CodeGen.h"
#include "llvm/Support/CommandLine.h"
#include "llvm/Support/Compiler.h"
#include "llvm/Support/Debug.h"
#include "llvm/Support/ErrorHandling.h"
#include "llvm/Support/KnownBits.h"
#include "llvm/Support/MachineValueType.h"
#include "llvm/Support/MathExtras.h"
#include "llvm/Support/raw_ostream.h"
#include "llvm/Target/TargetMachine.h"
#include "llvm/Target/TargetOptions.h"
#include <algorithm>
#include <cassert>
#include <cstdint>
#include <cstdlib>
#include <iterator>
#include <limits>
#include <optional>
#include <string>
#include <tuple>
#include <utility>
#include <vector>
 
using namespace llvm;
using namespace llvm::PatternMatch;
 
#define DEBUG_TYPE "arm-isel"
 
STATISTIC(NumTailCalls, "Number of tail calls");
STATISTIC(NumMovwMovt, "Number of GAs materialized with movw + movt");
STATISTIC(NumLoopByVals, "Number of loops generated for byval arguments");
STATISTIC(NumConstpoolPromoted,
  "Number of constants with their storage promoted into constant pools");
 
static cl::opt<bool>
ARMInterworking("arm-interworking", cl::Hidden,
  cl::desc("Enable / disable ARM interworking (for debugging only)"),
  cl::init(true));
 
static cl::opt<bool> EnableConstpoolPromotion(
    "arm-promote-constant", cl::Hidden,
    cl::desc("Enable / disable promotion of unnamed_addr constants into "
             "constant pools"),
    cl::init(false)); // FIXME: set to true by default once PR32780 is fixed
static cl::opt<unsigned> ConstpoolPromotionMaxSize(
    "arm-promote-constant-max-size", cl::Hidden,
    cl::desc("Maximum size of constant to promote into a constant pool"),
    cl::init(64));
static cl::opt<unsigned> ConstpoolPromotionMaxTotal(
    "arm-promote-constant-max-total", cl::Hidden,
    cl::desc("Maximum size of ALL constants to promote into a constant pool"),
    cl::init(128));
 
cl::opt<unsigned>
MVEMaxSupportedInterleaveFactor("mve-max-interleave-factor", cl::Hidden,
  cl::desc("Maximum interleave factor for MVE VLDn to generate."),
  cl::init(2));
 
// The APCS parameter registers.
static const MCPhysReg GPRArgRegs[] = {
  ARM::R0, ARM::R1, ARM::R2, ARM::R3
};
 
void ARMTargetLowering::addTypeForNEON(MVT VT, MVT PromotedLdStVT) {
  if (VT != PromotedLdStVT) {
    setOperationAction(ISD::LOAD, VT, Promote);
    AddPromotedToType (ISD::LOAD, VT, PromotedLdStVT);
 
    setOperationAction(ISD::STORE, VT, Promote);
    AddPromotedToType (ISD::STORE, VT, PromotedLdStVT);
  }
 
  MVT ElemTy = VT.getVectorElementType();
  if (ElemTy != MVT::f64)
    setOperationAction(ISD::SETCC, VT, Custom);
  setOperationAction(ISD::INSERT_VECTOR_ELT, VT, Custom);
  setOperationAction(ISD::EXTRACT_VECTOR_ELT, VT, Custom);
  if (ElemTy == MVT::i32) {
    setOperationAction(ISD::SINT_TO_FP, VT, Custom);
    setOperationAction(ISD::UINT_TO_FP, VT, Custom);
    setOperationAction(ISD::FP_TO_SINT, VT, Custom);
    setOperationAction(ISD::FP_TO_UINT, VT, Custom);
  } else {
    setOperationAction(ISD::SINT_TO_FP, VT, Expand);
    setOperationAction(ISD::UINT_TO_FP, VT, Expand);
    setOperationAction(ISD::FP_TO_SINT, VT, Expand);
    setOperationAction(ISD::FP_TO_UINT, VT, Expand);
  }
  setOperationAction(ISD::BUILD_VECTOR,      VT, Custom);
  setOperationAction(ISD::VECTOR_SHUFFLE,    VT, Custom);
  setOperationAction(ISD::CONCAT_VECTORS,    VT, Legal);
  setOperationAction(ISD::EXTRACT_SUBVECTOR, VT, Legal);
  setOperationAction(ISD::SELECT,            VT, Expand);
  setOperationAction(ISD::SELECT_CC,         VT, Expand);
  setOperationAction(ISD::VSELECT,           VT, Expand);
  setOperationAction(ISD::SIGN_EXTEND_INREG, VT, Expand);
  if (VT.isInteger()) {
    setOperationAction(ISD::SHL, VT, Custom);
    setOperationAction(ISD::SRA, VT, Custom);
    setOperationAction(ISD::SRL, VT, Custom);
  }
 
  // Neon does not support vector divide/remainder operations.
  setOperationAction(ISD::SDIV, VT, Expand);
  setOperationAction(ISD::UDIV, VT, Expand);
  setOperationAction(ISD::FDIV, VT, Expand);
  setOperationAction(ISD::SREM, VT, Expand);
  setOperationAction(ISD::UREM, VT, Expand);
  setOperationAction(ISD::FREM, VT, Expand);
  setOperationAction(ISD::SDIVREM, VT, Expand);
  setOperationAction(ISD::UDIVREM, VT, Expand);
 
  if (!VT.isFloatingPoint() &&
      VT != MVT::v2i64 && VT != MVT::v1i64)
    for (auto Opcode : {ISD::ABS, ISD::SMIN, ISD::SMAX, ISD::UMIN, ISD::UMAX})
      setOperationAction(Opcode, VT, Legal);
  if (!VT.isFloatingPoint())
    for (auto Opcode : {ISD::SADDSAT, ISD::UADDSAT, ISD::SSUBSAT, ISD::USUBSAT})
      setOperationAction(Opcode, VT, Legal);
}
 
void ARMTargetLowering::addDRTypeForNEON(MVT VT) {
  addRegisterClass(VT, &ARM::DPRRegClass);
  addTypeForNEON(VT, MVT::f64);
}
 
void ARMTargetLowering::addQRTypeForNEON(MVT VT) {
  addRegisterClass(VT, &ARM::DPairRegClass);
  addTypeForNEON(VT, MVT::v2f64);
}
 
void ARMTargetLowering::setAllExpand(MVT VT) {
  for (unsigned Opc = 0; Opc < ISD::BUILTIN_OP_END; ++Opc)
    setOperationAction(Opc, VT, Expand);
 
  // We support these really simple operations even on types where all
  // the actual arithmetic has to be broken down into simpler
  // operations or turned into library calls.
  setOperationAction(ISD::BITCAST, VT, Legal);
  setOperationAction(ISD::LOAD, VT, Legal);
  setOperationAction(ISD::STORE, VT, Legal);
  setOperationAction(ISD::UNDEF, VT, Legal);
}
 
void ARMTargetLowering::addAllExtLoads(const MVT From, const MVT To,
                                       LegalizeAction Action) {
  setLoadExtAction(ISD::EXTLOAD,  From, To, Action);
  setLoadExtAction(ISD::ZEXTLOAD, From, To, Action);
  setLoadExtAction(ISD::SEXTLOAD, From, To, Action);
}
 
void ARMTargetLowering::addMVEVectorTypes(bool HasMVEFP) {
  const MVT IntTypes[] = { MVT::v16i8, MVT::v8i16, MVT::v4i32 };
 
  for (auto VT : IntTypes) {
    addRegisterClass(VT, &ARM::MQPRRegClass);
    setOperationAction(ISD::VECTOR_SHUFFLE, VT, Custom);
    setOperationAction(ISD::INSERT_VECTOR_ELT, VT, Custom);
    setOperationAction(ISD::EXTRACT_VECTOR_ELT, VT, Custom);
    setOperationAction(ISD::BUILD_VECTOR, VT, Custom);
    setOperationAction(ISD::SHL, VT, Custom);
    setOperationAction(ISD::SRA, VT, Custom);
    setOperationAction(ISD::SRL, VT, Custom);
    setOperationAction(ISD::SMIN, VT, Legal);
    setOperationAction(ISD::SMAX, VT, Legal);
    setOperationAction(ISD::UMIN, VT, Legal);
    setOperationAction(ISD::UMAX, VT, Legal);
    setOperationAction(ISD::ABS, VT, Legal);
    setOperationAction(ISD::SETCC, VT, Custom);
    setOperationAction(ISD::MLOAD, VT, Custom);
    setOperationAction(ISD::MSTORE, VT, Legal);
    setOperationAction(ISD::CTLZ, VT, Legal);
    setOperationAction(ISD::CTTZ, VT, Custom);
    setOperationAction(ISD::BITREVERSE, VT, Legal);
    setOperationAction(ISD::BSWAP, VT, Legal);
    setOperationAction(ISD::SADDSAT, VT, Legal);
    setOperationAction(ISD::UADDSAT, VT, Legal);
    setOperationAction(ISD::SSUBSAT, VT, Legal);
    setOperationAction(ISD::USUBSAT, VT, Legal);
    setOperationAction(ISD::ABDS, VT, Legal);
    setOperationAction(ISD::ABDU, VT, Legal);
    setOperationAction(ISD::AVGFLOORS, VT, Legal);
    setOperationAction(ISD::AVGFLOORU, VT, Legal);
    setOperationAction(ISD::AVGCEILS, VT, Legal);
    setOperationAction(ISD::AVGCEILU, VT, Legal);
 
    // No native support for these.
    setOperationAction(ISD::UDIV, VT, Expand);
    setOperationAction(ISD::SDIV, VT, Expand);
    setOperationAction(ISD::UREM, VT, Expand);
    setOperationAction(ISD::SREM, VT, Expand);
    setOperationAction(ISD::UDIVREM, VT, Expand);
    setOperationAction(ISD::SDIVREM, VT, Expand);
    setOperationAction(ISD::CTPOP, VT, Expand);
    setOperationAction(ISD::SELECT, VT, Expand);
    setOperationAction(ISD::SELECT_CC, VT, Expand);
 
    // Vector reductions
    setOperationAction(ISD::VECREDUCE_ADD, VT, Legal);
    setOperationAction(ISD::VECREDUCE_SMAX, VT, Legal);
    setOperationAction(ISD::VECREDUCE_UMAX, VT, Legal);
    setOperationAction(ISD::VECREDUCE_SMIN, VT, Legal);
    setOperationAction(ISD::VECREDUCE_UMIN, VT, Legal);
    setOperationAction(ISD::VECREDUCE_MUL, VT, Custom);
    setOperationAction(ISD::VECREDUCE_AND, VT, Custom);
    setOperationAction(ISD::VECREDUCE_OR, VT, Custom);
    setOperationAction(ISD::VECREDUCE_XOR, VT, Custom);
 
    if (!HasMVEFP) {
      setOperationAction(ISD::SINT_TO_FP, VT, Expand);
      setOperationAction(ISD::UINT_TO_FP, VT, Expand);
      setOperationAction(ISD::FP_TO_SINT, VT, Expand);
      setOperationAction(ISD::FP_TO_UINT, VT, Expand);
    } else {
      setOperationAction(ISD::FP_TO_SINT_SAT, VT, Custom);
      setOperationAction(ISD::FP_TO_UINT_SAT, VT, Custom);
    }
 
    // Pre and Post inc are supported on loads and stores
    for (unsigned im = (unsigned)ISD::PRE_INC;
         im != (unsigned)ISD::LAST_INDEXED_MODE; ++im) {
      setIndexedLoadAction(im, VT, Legal);
      setIndexedStoreAction(im, VT, Legal);
      setIndexedMaskedLoadAction(im, VT, Legal);
      setIndexedMaskedStoreAction(im, VT, Legal);
    }
  }
 
  const MVT FloatTypes[] = { MVT::v8f16, MVT::v4f32 };
  for (auto VT : FloatTypes) {
    addRegisterClass(VT, &ARM::MQPRRegClass);
    if (!HasMVEFP)
      setAllExpand(VT);
 
    // These are legal or custom whether we have MVE.fp or not
    setOperationAction(ISD::VECTOR_SHUFFLE, VT, Custom);
    setOperationAction(ISD::INSERT_VECTOR_ELT, VT, Custom);
    setOperationAction(ISD::INSERT_VECTOR_ELT, VT.getVectorElementType(), Custom);
    setOperationAction(ISD::EXTRACT_VECTOR_ELT, VT, Custom);
    setOperationAction(ISD::BUILD_VECTOR, VT, Custom);
    setOperationAction(ISD::BUILD_VECTOR, VT.getVectorElementType(), Custom);
    setOperationAction(ISD::SCALAR_TO_VECTOR, VT, Legal);
    setOperationAction(ISD::SETCC, VT, Custom);
    setOperationAction(ISD::MLOAD, VT, Custom);
    setOperationAction(ISD::MSTORE, VT, Legal);
    setOperationAction(ISD::SELECT, VT, Expand);
    setOperationAction(ISD::SELECT_CC, VT, Expand);
 
    // Pre and Post inc are supported on loads and stores
    for (unsigned im = (unsigned)ISD::PRE_INC;
         im != (unsigned)ISD::LAST_INDEXED_MODE; ++im) {
      setIndexedLoadAction(im, VT, Legal);
      setIndexedStoreAction(im, VT, Legal);
      setIndexedMaskedLoadAction(im, VT, Legal);
      setIndexedMaskedStoreAction(im, VT, Legal);
    }
 
    if (HasMVEFP) {
      setOperationAction(ISD::FMINNUM, VT, Legal);
      setOperationAction(ISD::FMAXNUM, VT, Legal);
      setOperationAction(ISD::FROUND, VT, Legal);
      setOperationAction(ISD::VECREDUCE_FADD, VT, Custom);
      setOperationAction(ISD::VECREDUCE_FMUL, VT, Custom);
      setOperationAction(ISD::VECREDUCE_FMIN, VT, Custom);
      setOperationAction(ISD::VECREDUCE_FMAX, VT, Custom);
 
      // No native support for these.
      setOperationAction(ISD::FDIV, VT, Expand);
      setOperationAction(ISD::FREM, VT, Expand);
      setOperationAction(ISD::FSQRT, VT, Expand);
      setOperationAction(ISD::FSIN, VT, Expand);
      setOperationAction(ISD::FCOS, VT, Expand);
      setOperationAction(ISD::FPOW, VT, Expand);
      setOperationAction(ISD::FLOG, VT, Expand);
      setOperationAction(ISD::FLOG2, VT, Expand);
      setOperationAction(ISD::FLOG10, VT, Expand);
      setOperationAction(ISD::FEXP, VT, Expand);
      setOperationAction(ISD::FEXP2, VT, Expand);
      setOperationAction(ISD::FNEARBYINT, VT, Expand);
    }
  }
 
  // Custom Expand smaller than legal vector reductions to prevent false zero
  // items being added.
  setOperationAction(ISD::VECREDUCE_FADD, MVT::v4f16, Custom);
  setOperationAction(ISD::VECREDUCE_FMUL, MVT::v4f16, Custom);
  setOperationAction(ISD::VECREDUCE_FMIN, MVT::v4f16, Custom);
  setOperationAction(ISD::VECREDUCE_FMAX, MVT::v4f16, Custom);
  setOperationAction(ISD::VECREDUCE_FADD, MVT::v2f16, Custom);
  setOperationAction(ISD::VECREDUCE_FMUL, MVT::v2f16, Custom);
  setOperationAction(ISD::VECREDUCE_FMIN, MVT::v2f16, Custom);
  setOperationAction(ISD::VECREDUCE_FMAX, MVT::v2f16, Custom);
 
  // We 'support' these types up to bitcast/load/store level, regardless of
  // MVE integer-only / float support. Only doing FP data processing on the FP
  // vector types is inhibited at integer-only level.
  const MVT LongTypes[] = { MVT::v2i64, MVT::v2f64 };
  for (auto VT : LongTypes) {
    addRegisterClass(VT, &ARM::MQPRRegClass);
    setAllExpand(VT);
    setOperationAction(ISD::INSERT_VECTOR_ELT, VT, Custom);
    setOperationAction(ISD::EXTRACT_VECTOR_ELT, VT, Custom);
    setOperationAction(ISD::BUILD_VECTOR, VT, Custom);
    setOperationAction(ISD::VSELECT, VT, Legal);
    setOperationAction(ISD::VECTOR_SHUFFLE, VT, Custom);
  }
  setOperationAction(ISD::SCALAR_TO_VECTOR, MVT::v2f64, Legal);
 
  // We can do bitwise operations on v2i64 vectors
  setOperationAction(ISD::AND, MVT::v2i64, Legal);
  setOperationAction(ISD::OR, MVT::v2i64, Legal);
  setOperationAction(ISD::XOR, MVT::v2i64, Legal);
 
  // It is legal to extload from v4i8 to v4i16 or v4i32.
  addAllExtLoads(MVT::v8i16, MVT::v8i8, Legal);
  addAllExtLoads(MVT::v4i32, MVT::v4i16, Legal);
  addAllExtLoads(MVT::v4i32, MVT::v4i8, Legal);
 
  // It is legal to sign extend from v4i8/v4i16 to v4i32 or v8i8 to v8i16.
  setOperationAction(ISD::SIGN_EXTEND_INREG, MVT::v4i8,  Legal);
  setOperationAction(ISD::SIGN_EXTEND_INREG, MVT::v4i16, Legal);
  setOperationAction(ISD::SIGN_EXTEND_INREG, MVT::v4i32, Legal);
  setOperationAction(ISD::SIGN_EXTEND_INREG, MVT::v8i8,  Legal);
  setOperationAction(ISD::SIGN_EXTEND_INREG, MVT::v8i16, Legal);
 
  // Some truncating stores are legal too.
  setTruncStoreAction(MVT::v4i32, MVT::v4i16, Legal);
  setTruncStoreAction(MVT::v4i32, MVT::v4i8,  Legal);
  setTruncStoreAction(MVT::v8i16, MVT::v8i8,  Legal);
 
  // Pre and Post inc on these are legal, given the correct extends
  for (unsigned im = (unsigned)ISD::PRE_INC;
       im != (unsigned)ISD::LAST_INDEXED_MODE; ++im) {
    for (auto VT : {MVT::v8i8, MVT::v4i8, MVT::v4i16}) {
      setIndexedLoadAction(im, VT, Legal);
      setIndexedStoreAction(im, VT, Legal);
      setIndexedMaskedLoadAction(im, VT, Legal);
      setIndexedMaskedStoreAction(im, VT, Legal);
    }
  }
 
  // Predicate types
  const MVT pTypes[] = {MVT::v16i1, MVT::v8i1, MVT::v4i1, MVT::v2i1};
  for (auto VT : pTypes) {
    addRegisterClass(VT, &ARM::VCCRRegClass);
    setOperationAction(ISD::BUILD_VECTOR, VT, Custom);
    setOperationAction(ISD::VECTOR_SHUFFLE, VT, Custom);
    setOperationAction(ISD::EXTRACT_SUBVECTOR, VT, Custom);
    setOperationAction(ISD::CONCAT_VECTORS, VT, Custom);
    setOperationAction(ISD::INSERT_VECTOR_ELT, VT, Custom);
    setOperationAction(ISD::EXTRACT_VECTOR_ELT, VT, Custom);
    setOperationAction(ISD::SETCC, VT, Custom);
    setOperationAction(ISD::SCALAR_TO_VECTOR, VT, Expand);
    setOperationAction(ISD::LOAD, VT, Custom);
    setOperationAction(ISD::STORE, VT, Custom);
    setOperationAction(ISD::TRUNCATE, VT, Custom);
    setOperationAction(ISD::VSELECT, VT, Expand);
    setOperationAction(ISD::SELECT, VT, Expand);
    setOperationAction(ISD::SELECT_CC, VT, Expand);
 
    if (!HasMVEFP) {
      setOperationAction(ISD::SINT_TO_FP, VT, Expand);
      setOperationAction(ISD::UINT_TO_FP, VT, Expand);
      setOperationAction(ISD::FP_TO_SINT, VT, Expand);
      setOperationAction(ISD::FP_TO_UINT, VT, Expand);
    }
  }
  setOperationAction(ISD::SETCC, MVT::v2i1, Expand);
  setOperationAction(ISD::TRUNCATE, MVT::v2i1, Expand);
  setOperationAction(ISD::AND, MVT::v2i1, Expand);
  setOperationAction(ISD::OR, MVT::v2i1, Expand);
  setOperationAction(ISD::XOR, MVT::v2i1, Expand);
  setOperationAction(ISD::SINT_TO_FP, MVT::v2i1, Expand);
  setOperationAction(ISD::UINT_TO_FP, MVT::v2i1, Expand);
  setOperationAction(ISD::FP_TO_SINT, MVT::v2i1, Expand);
  setOperationAction(ISD::FP_TO_UINT, MVT::v2i1, Expand);
 
  setOperationAction(ISD::SIGN_EXTEND, MVT::v8i32, Custom);
  setOperationAction(ISD::SIGN_EXTEND, MVT::v16i16, Custom);
  setOperationAction(ISD::SIGN_EXTEND, MVT::v16i32, Custom);
  setOperationAction(ISD::ZERO_EXTEND, MVT::v8i32, Custom);
  setOperationAction(ISD::ZERO_EXTEND, MVT::v16i16, Custom);
  setOperationAction(ISD::ZERO_EXTEND, MVT::v16i32, Custom);
  setOperationAction(ISD::TRUNCATE, MVT::v8i32, Custom);
  setOperationAction(ISD::TRUNCATE, MVT::v16i16, Custom);
}
 
ARMTargetLowering::ARMTargetLowering(const TargetMachine &TM,
                                     const ARMSubtarget &STI)
    : TargetLowering(TM), Subtarget(&STI) {
  RegInfo = Subtarget->getRegisterInfo();
  Itins = Subtarget->getInstrItineraryData();
 
  setBooleanContents(ZeroOrOneBooleanContent);
  setBooleanVectorContents(ZeroOrNegativeOneBooleanContent);
 
  if (!Subtarget->isTargetDarwin() && !Subtarget->isTargetIOS() &&
      !Subtarget->isTargetWatchOS() && !Subtarget->isTargetDriverKit()) {
    bool IsHFTarget = TM.Options.FloatABIType == FloatABI::Hard;
    for (int LCID = 0; LCID < RTLIB::UNKNOWN_LIBCALL; ++LCID)
      setLibcallCallingConv(static_cast<RTLIB::Libcall>(LCID),
                            IsHFTarget ? CallingConv::ARM_AAPCS_VFP
                                       : CallingConv::ARM_AAPCS);
  }
 
  if (Subtarget->isTargetMachO()) {
    // Uses VFP for Thumb libfuncs if available.
    if (Subtarget->isThumb() && Subtarget->hasVFP2Base() &&
        Subtarget->hasARMOps() && !Subtarget->useSoftFloat()) {
      static const struct {
        const RTLIB::Libcall Op;
        const char * const Name;
        const ISD::CondCode Cond;
      } LibraryCalls[] = {
        // Single-precision floating-point arithmetic.
        { RTLIB::ADD_F32, "__addsf3vfp", ISD::SETCC_INVALID },
        { RTLIB::SUB_F32, "__subsf3vfp", ISD::SETCC_INVALID },
        { RTLIB::MUL_F32, "__mulsf3vfp", ISD::SETCC_INVALID },
        { RTLIB::DIV_F32, "__divsf3vfp", ISD::SETCC_INVALID },
 
        // Double-precision floating-point arithmetic.
        { RTLIB::ADD_F64, "__adddf3vfp", ISD::SETCC_INVALID },
        { RTLIB::SUB_F64, "__subdf3vfp", ISD::SETCC_INVALID },
        { RTLIB::MUL_F64, "__muldf3vfp", ISD::SETCC_INVALID },
        { RTLIB::DIV_F64, "__divdf3vfp", ISD::SETCC_INVALID },
 
        // Single-precision comparisons.
        { RTLIB::OEQ_F32, "__eqsf2vfp",    ISD::SETNE },
        { RTLIB::UNE_F32, "__nesf2vfp",    ISD::SETNE },
        { RTLIB::OLT_F32, "__ltsf2vfp",    ISD::SETNE },
        { RTLIB::OLE_F32, "__lesf2vfp",    ISD::SETNE },
        { RTLIB::OGE_F32, "__gesf2vfp",    ISD::SETNE },
        { RTLIB::OGT_F32, "__gtsf2vfp",    ISD::SETNE },
        { RTLIB::UO_F32,  "__unordsf2vfp", ISD::SETNE },
 
        // Double-precision comparisons.
        { RTLIB::OEQ_F64, "__eqdf2vfp",    ISD::SETNE },
        { RTLIB::UNE_F64, "__nedf2vfp",    ISD::SETNE },
        { RTLIB::OLT_F64, "__ltdf2vfp",    ISD::SETNE },
        { RTLIB::OLE_F64, "__ledf2vfp",    ISD::SETNE },
        { RTLIB::OGE_F64, "__gedf2vfp",    ISD::SETNE },
        { RTLIB::OGT_F64, "__gtdf2vfp",    ISD::SETNE },
        { RTLIB::UO_F64,  "__unorddf2vfp", ISD::SETNE },
 
        // Floating-point to integer conversions.
        // i64 conversions are done via library routines even when generating VFP
        // instructions, so use the same ones.
        { RTLIB::FPTOSINT_F64_I32, "__fixdfsivfp",    ISD::SETCC_INVALID },
        { RTLIB::FPTOUINT_F64_I32, "__fixunsdfsivfp", ISD::SETCC_INVALID },
        { RTLIB::FPTOSINT_F32_I32, "__fixsfsivfp",    ISD::SETCC_INVALID },
        { RTLIB::FPTOUINT_F32_I32, "__fixunssfsivfp", ISD::SETCC_INVALID },
 
        // Conversions between floating types.
        { RTLIB::FPROUND_F64_F32, "__truncdfsf2vfp",  ISD::SETCC_INVALID },
        { RTLIB::FPEXT_F32_F64,   "__extendsfdf2vfp", ISD::SETCC_INVALID },
 
        // Integer to floating-point conversions.
        // i64 conversions are done via library routines even when generating VFP
        // instructions, so use the same ones.
        // FIXME: There appears to be some naming inconsistency in ARM libgcc:
        // e.g., __floatunsidf vs. __floatunssidfvfp.
        { RTLIB::SINTTOFP_I32_F64, "__floatsidfvfp",    ISD::SETCC_INVALID },
        { RTLIB::UINTTOFP_I32_F64, "__floatunssidfvfp", ISD::SETCC_INVALID },
        { RTLIB::SINTTOFP_I32_F32, "__floatsisfvfp",    ISD::SETCC_INVALID },
        { RTLIB::UINTTOFP_I32_F32, "__floatunssisfvfp", ISD::SETCC_INVALID },
      };
 
      for (const auto &LC : LibraryCalls) {
        setLibcallName(LC.Op, LC.Name);
        if (LC.Cond != ISD::SETCC_INVALID)
          setCmpLibcallCC(LC.Op, LC.Cond);
      }
    }
  }
 
  // These libcalls are not available in 32-bit.
  setLibcallName(RTLIB::SHL_I128, nullptr);
  setLibcallName(RTLIB::SRL_I128, nullptr);
  setLibcallName(RTLIB::SRA_I128, nullptr);
  setLibcallName(RTLIB::MUL_I128, nullptr);
  setLibcallName(RTLIB::MULO_I64, nullptr);
  setLibcallName(RTLIB::MULO_I128, nullptr);
 
  // RTLIB
  if (Subtarget->isAAPCS_ABI() &&
      (Subtarget->isTargetAEABI() || Subtarget->isTargetGNUAEABI() ||
       Subtarget->isTargetMuslAEABI() || Subtarget->isTargetAndroid())) {
    static const struct {
      const RTLIB::Libcall Op;
      const char * const Name;
      const CallingConv::ID CC;
      const ISD::CondCode Cond;
    } LibraryCalls[] = {
      // Double-precision floating-point arithmetic helper functions
      // RTABI chapter 4.1.2, Table 2
      { RTLIB::ADD_F64, "__aeabi_dadd", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
      { RTLIB::DIV_F64, "__aeabi_ddiv", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
      { RTLIB::MUL_F64, "__aeabi_dmul", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
      { RTLIB::SUB_F64, "__aeabi_dsub", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
 
      // Double-precision floating-point comparison helper functions
      // RTABI chapter 4.1.2, Table 3
      { RTLIB::OEQ_F64, "__aeabi_dcmpeq", CallingConv::ARM_AAPCS, ISD::SETNE },
      { RTLIB::UNE_F64, "__aeabi_dcmpeq", CallingConv::ARM_AAPCS, ISD::SETEQ },
      { RTLIB::OLT_F64, "__aeabi_dcmplt", CallingConv::ARM_AAPCS, ISD::SETNE },
      { RTLIB::OLE_F64, "__aeabi_dcmple", CallingConv::ARM_AAPCS, ISD::SETNE },
      { RTLIB::OGE_F64, "__aeabi_dcmpge", CallingConv::ARM_AAPCS, ISD::SETNE },
      { RTLIB::OGT_F64, "__aeabi_dcmpgt", CallingConv::ARM_AAPCS, ISD::SETNE },
      { RTLIB::UO_F64,  "__aeabi_dcmpun", CallingConv::ARM_AAPCS, ISD::SETNE },
 
      // Single-precision floating-point arithmetic helper functions
      // RTABI chapter 4.1.2, Table 4
      { RTLIB::ADD_F32, "__aeabi_fadd", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
      { RTLIB::DIV_F32, "__aeabi_fdiv", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
      { RTLIB::MUL_F32, "__aeabi_fmul", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
      { RTLIB::SUB_F32, "__aeabi_fsub", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
 
      // Single-precision floating-point comparison helper functions
      // RTABI chapter 4.1.2, Table 5
      { RTLIB::OEQ_F32, "__aeabi_fcmpeq", CallingConv::ARM_AAPCS, ISD::SETNE },
      { RTLIB::UNE_F32, "__aeabi_fcmpeq", CallingConv::ARM_AAPCS, ISD::SETEQ },
      { RTLIB::OLT_F32, "__aeabi_fcmplt", CallingConv::ARM_AAPCS, ISD::SETNE },
      { RTLIB::OLE_F32, "__aeabi_fcmple", CallingConv::ARM_AAPCS, ISD::SETNE },
      { RTLIB::OGE_F32, "__aeabi_fcmpge", CallingConv::ARM_AAPCS, ISD::SETNE },
      { RTLIB::OGT_F32, "__aeabi_fcmpgt", CallingConv::ARM_AAPCS, ISD::SETNE },
      { RTLIB::UO_F32,  "__aeabi_fcmpun", CallingConv::ARM_AAPCS, ISD::SETNE },
 
      // Floating-point to integer conversions.
      // RTABI chapter 4.1.2, Table 6
      { RTLIB::FPTOSINT_F64_I32, "__aeabi_d2iz",  CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
      { RTLIB::FPTOUINT_F64_I32, "__aeabi_d2uiz", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
      { RTLIB::FPTOSINT_F64_I64, "__aeabi_d2lz",  CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
      { RTLIB::FPTOUINT_F64_I64, "__aeabi_d2ulz", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
      { RTLIB::FPTOSINT_F32_I32, "__aeabi_f2iz",  CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
      { RTLIB::FPTOUINT_F32_I32, "__aeabi_f2uiz", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
      { RTLIB::FPTOSINT_F32_I64, "__aeabi_f2lz",  CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
      { RTLIB::FPTOUINT_F32_I64, "__aeabi_f2ulz", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
 
      // Conversions between floating types.
      // RTABI chapter 4.1.2, Table 7
      { RTLIB::FPROUND_F64_F32, "__aeabi_d2f", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
      { RTLIB::FPROUND_F64_F16, "__aeabi_d2h", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
      { RTLIB::FPEXT_F32_F64,   "__aeabi_f2d", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
 
      // Integer to floating-point conversions.
      // RTABI chapter 4.1.2, Table 8
      { RTLIB::SINTTOFP_I32_F64, "__aeabi_i2d",  CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
      { RTLIB::UINTTOFP_I32_F64, "__aeabi_ui2d", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
      { RTLIB::SINTTOFP_I64_F64, "__aeabi_l2d",  CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
      { RTLIB::UINTTOFP_I64_F64, "__aeabi_ul2d", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
      { RTLIB::SINTTOFP_I32_F32, "__aeabi_i2f",  CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
      { RTLIB::UINTTOFP_I32_F32, "__aeabi_ui2f", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
      { RTLIB::SINTTOFP_I64_F32, "__aeabi_l2f",  CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
      { RTLIB::UINTTOFP_I64_F32, "__aeabi_ul2f", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
 
      // Long long helper functions
      // RTABI chapter 4.2, Table 9
      { RTLIB::MUL_I64, "__aeabi_lmul", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
      { RTLIB::SHL_I64, "__aeabi_llsl", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
      { RTLIB::SRL_I64, "__aeabi_llsr", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
      { RTLIB::SRA_I64, "__aeabi_lasr", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
 
      // Integer division functions
      // RTABI chapter 4.3.1
      { RTLIB::SDIV_I8,  "__aeabi_idiv",     CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
      { RTLIB::SDIV_I16, "__aeabi_idiv",     CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
      { RTLIB::SDIV_I32, "__aeabi_idiv",     CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
      { RTLIB::SDIV_I64, "__aeabi_ldivmod",  CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
      { RTLIB::UDIV_I8,  "__aeabi_uidiv",    CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
      { RTLIB::UDIV_I16, "__aeabi_uidiv",    CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
      { RTLIB::UDIV_I32, "__aeabi_uidiv",    CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
      { RTLIB::UDIV_I64, "__aeabi_uldivmod", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
    };
 
    for (const auto &LC : LibraryCalls) {
      setLibcallName(LC.Op, LC.Name);
      setLibcallCallingConv(LC.Op, LC.CC);
      if (LC.Cond != ISD::SETCC_INVALID)
        setCmpLibcallCC(LC.Op, LC.Cond);
    }
 
    // EABI dependent RTLIB
    if (TM.Options.EABIVersion == EABI::EABI4 ||
        TM.Options.EABIVersion == EABI::EABI5) {
      static const struct {
        const RTLIB::Libcall Op;
        const char *const Name;
        const CallingConv::ID CC;
        const ISD::CondCode Cond;
      } MemOpsLibraryCalls[] = {
        // Memory operations
        // RTABI chapter 4.3.4
        { RTLIB::MEMCPY,  "__aeabi_memcpy",  CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
        { RTLIB::MEMMOVE, "__aeabi_memmove", CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
        { RTLIB::MEMSET,  "__aeabi_memset",  CallingConv::ARM_AAPCS, ISD::SETCC_INVALID },
      };
 
      for (const auto &LC : MemOpsLibraryCalls) {
        setLibcallName(LC.Op, LC.Name);
        setLibcallCallingConv(LC.Op, LC.CC);
        if (LC.Cond != ISD::SETCC_INVALID)
          setCmpLibcallCC(LC.Op, LC.Cond);
      }
    }
  }
 
  if (Subtarget->isTargetWindows()) {
    static const struct {
      const RTLIB::Libcall Op;
      const char * const Name;
      const CallingConv::ID CC;
    } LibraryCalls[] = {
      { RTLIB::FPTOSINT_F32_I64, "__stoi64", CallingConv::ARM_AAPCS_VFP },
      { RTLIB::FPTOSINT_F64_I64, "__dtoi64", CallingConv::ARM_AAPCS_VFP },
      { RTLIB::FPTOUINT_F32_I64, "__stou64", CallingConv::ARM_AAPCS_VFP },
      { RTLIB::FPTOUINT_F64_I64, "__dtou64", CallingConv::ARM_AAPCS_VFP },
      { RTLIB::SINTTOFP_I64_F32, "__i64tos", CallingConv::ARM_AAPCS_VFP },
      { RTLIB::SINTTOFP_I64_F64, "__i64tod", CallingConv::ARM_AAPCS_VFP },
      { RTLIB::UINTTOFP_I64_F32, "__u64tos", CallingConv::ARM_AAPCS_VFP },
      { RTLIB::UINTTOFP_I64_F64, "__u64tod", CallingConv::ARM_AAPCS_VFP },
    };
 
    for (const auto &LC : LibraryCalls) {
      setLibcallName(LC.Op, LC.Name);
      setLibcallCallingConv(LC.Op, LC.CC);
    }
  }
 
  // Use divmod compiler-rt calls for iOS 5.0 and later.
  if (Subtarget->isTargetMachO() &&
      !(Subtarget->isTargetIOS() &&
        Subtarget->getTargetTriple().isOSVersionLT(5, 0))) {
    setLibcallName(RTLIB::SDIVREM_I32, "__divmodsi4");
    setLibcallName(RTLIB::UDIVREM_I32, "__udivmodsi4");
  }
 
  // The half <-> float conversion functions are always soft-float on
  // non-watchos platforms, but are needed for some targets which use a
  // hard-float calling convention by default.
  if (!Subtarget->isTargetWatchABI()) {
    if (Subtarget->isAAPCS_ABI()) {
      setLibcallCallingConv(RTLIB::FPROUND_F32_F16, CallingConv::ARM_AAPCS);
      setLibcallCallingConv(RTLIB::FPROUND_F64_F16, CallingConv::ARM_AAPCS);
      setLibcallCallingConv(RTLIB::FPEXT_F16_F32, CallingConv::ARM_AAPCS);
    } else {
      setLibcallCallingConv(RTLIB::FPROUND_F32_F16, CallingConv::ARM_APCS);
      setLibcallCallingConv(RTLIB::FPROUND_F64_F16, CallingConv::ARM_APCS);
      setLibcallCallingConv(RTLIB::FPEXT_F16_F32, CallingConv::ARM_APCS);
    }
  }
 
  // In EABI, these functions have an __aeabi_ prefix, but in GNUEABI they have
  // a __gnu_ prefix (which is the default).
  if (Subtarget->isTargetAEABI()) {
    static const struct {
      const RTLIB::Libcall Op;
      const char * const Name;
      const CallingConv::ID CC;
    } LibraryCalls[] = {
      { RTLIB::FPROUND_F32_F16, "__aeabi_f2h", CallingConv::ARM_AAPCS },
      { RTLIB::FPROUND_F64_F16, "__aeabi_d2h", CallingConv::ARM_AAPCS },
      { RTLIB::FPEXT_F16_F32, "__aeabi_h2f", CallingConv::ARM_AAPCS },
    };
 
    for (const auto &LC : LibraryCalls) {
      setLibcallName(LC.Op, LC.Name);
      setLibcallCallingConv(LC.Op, LC.CC);
    }
  }
 
  if (Subtarget->isThumb1Only())
    addRegisterClass(MVT::i32, &ARM::tGPRRegClass);
  else
    addRegisterClass(MVT::i32, &ARM::GPRRegClass);
 
  if (!Subtarget->useSoftFloat() && !Subtarget->isThumb1Only() &&
      Subtarget->hasFPRegs()) {
    addRegisterClass(MVT::f32, &ARM::SPRRegClass);
    addRegisterClass(MVT::f64, &ARM::DPRRegClass);
 
    setOperationAction(ISD::FP_TO_SINT_SAT, MVT::i32, Custom);
    setOperationAction(ISD::FP_TO_UINT_SAT, MVT::i32, Custom);
    setOperationAction(ISD::FP_TO_SINT_SAT, MVT::i64, Custom);
    setOperationAction(ISD::FP_TO_UINT_SAT, MVT::i64, Custom);
 
    if (!Subtarget->hasVFP2Base())
      setAllExpand(MVT::f32);
    if (!Subtarget->hasFP64())
      setAllExpand(MVT::f64);
  }
 
  if (Subtarget->hasFullFP16()) {
    addRegisterClass(MVT::f16, &ARM::HPRRegClass);
    setOperationAction(ISD::BITCAST, MVT::i16, Custom);
    setOperationAction(ISD::BITCAST, MVT::f16, Custom);
 
    setOperationAction(ISD::FMINNUM, MVT::f16, Legal);
    setOperationAction(ISD::FMAXNUM, MVT::f16, Legal);
  }
 
  if (Subtarget->hasBF16()) {
    addRegisterClass(MVT::bf16, &ARM::HPRRegClass);
    setAllExpand(MVT::bf16);
    if (!Subtarget->hasFullFP16())
      setOperationAction(ISD::BITCAST, MVT::bf16, Custom);
  }
 
  for (MVT VT : MVT::fixedlen_vector_valuetypes()) {
    for (MVT InnerVT : MVT::fixedlen_vector_valuetypes()) {
      setTruncStoreAction(VT, InnerVT, Expand);
      addAllExtLoads(VT, InnerVT, Expand);
    }
 
    setOperationAction(ISD::SMUL_LOHI, VT, Expand);
    setOperationAction(ISD::UMUL_LOHI, VT, Expand);
 
    setOperationAction(ISD::BSWAP, VT, Expand);
  }
 
  setOperationAction(ISD::ConstantFP, MVT::f32, Custom);
  setOperationAction(ISD::ConstantFP, MVT::f64, Custom);
 
  setOperationAction(ISD::READ_REGISTER, MVT::i64, Custom);
  setOperationAction(ISD::WRITE_REGISTER, MVT::i64, Custom);
 
  if (Subtarget->hasMVEIntegerOps())
    addMVEVectorTypes(Subtarget->hasMVEFloatOps());
 
  // Combine low-overhead loop intrinsics so that we can lower i1 types.
  if (Subtarget->hasLOB()) {
    setTargetDAGCombine({ISD::BRCOND, ISD::BR_CC});
  }
 
  if (Subtarget->hasNEON()) {
    addDRTypeForNEON(MVT::v2f32);
    addDRTypeForNEON(MVT::v8i8);
    addDRTypeForNEON(MVT::v4i16);
    addDRTypeForNEON(MVT::v2i32);
    addDRTypeForNEON(MVT::v1i64);
 
    addQRTypeForNEON(MVT::v4f32);
    addQRTypeForNEON(MVT::v2f64);
    addQRTypeForNEON(MVT::v16i8);
    addQRTypeForNEON(MVT::v8i16);
    addQRTypeForNEON(MVT::v4i32);
    addQRTypeForNEON(MVT::v2i64);
 
    if (Subtarget->hasFullFP16()) {
      addQRTypeForNEON(MVT::v8f16);
      addDRTypeForNEON(MVT::v4f16);
    }
 
    if (Subtarget->hasBF16()) {
      addQRTypeForNEON(MVT::v8bf16);
      addDRTypeForNEON(MVT::v4bf16);
    }
  }
 
  if (Subtarget->hasMVEIntegerOps() || Subtarget->hasNEON()) {
    // v2f64 is legal so that QR subregs can be extracted as f64 elements, but
    // none of Neon, MVE or VFP supports any arithmetic operations on it.
    setOperationAction(ISD::FADD, MVT::v2f64, Expand);
    setOperationAction(ISD::FSUB, MVT::v2f64, Expand);
    setOperationAction(ISD::FMUL, MVT::v2f64, Expand);
    // FIXME: Code duplication: FDIV and FREM are expanded always, see
    // ARMTargetLowering::addTypeForNEON method for details.
    setOperationAction(ISD::FDIV, MVT::v2f64, Expand);
    setOperationAction(ISD::FREM, MVT::v2f64, Expand);
    // FIXME: Create unittest.
    // In another words, find a way when "copysign" appears in DAG with vector
    // operands.
    setOperationAction(ISD::FCOPYSIGN, MVT::v2f64, Expand);
    // FIXME: Code duplication: SETCC has custom operation action, see
    // ARMTargetLowering::addTypeForNEON method for details.
    setOperationAction(ISD::SETCC, MVT::v2f64, Expand);
    // FIXME: Create unittest for FNEG and for FABS.
    setOperationAction(ISD::FNEG, MVT::v2f64, Expand);
    setOperationAction(ISD::FABS, MVT::v2f64, Expand);
    setOperationAction(ISD::FSQRT, MVT::v2f64, Expand);
    setOperationAction(ISD::FSIN, MVT::v2f64, Expand);
    setOperationAction(ISD::FCOS, MVT::v2f64, Expand);
    setOperationAction(ISD::FPOW, MVT::v2f64, Expand);
    setOperationAction(ISD::FLOG, MVT::v2f64, Expand);
    setOperationAction(ISD::FLOG2, MVT::v2f64, Expand);
    setOperationAction(ISD::FLOG10, MVT::v2f64, Expand);
    setOperationAction(ISD::FEXP, MVT::v2f64, Expand);
    setOperationAction(ISD::FEXP2, MVT::v2f64, Expand);
    // FIXME: Create unittest for FCEIL, FTRUNC, FRINT, FNEARBYINT, FFLOOR.
    setOperationAction(ISD::FCEIL, MVT::v2f64, Expand);
    setOperationAction(ISD::FTRUNC, MVT::v2f64, Expand);
    setOperationAction(ISD::FRINT, MVT::v2f64, Expand);
    setOperationAction(ISD::FNEARBYINT, MVT::v2f64, Expand);
    setOperationAction(ISD::FFLOOR, MVT::v2f64, Expand);
    setOperationAction(ISD::FMA, MVT::v2f64, Expand);
  }
 
  if (Subtarget->hasNEON()) {
    // The same with v4f32. But keep in mind that vadd, vsub, vmul are natively
    // supported for v4f32.
    setOperationAction(ISD::FSQRT, MVT::v4f32, Expand);
    setOperationAction(ISD::FSIN, MVT::v4f32, Expand);
    setOperationAction(ISD::FCOS, MVT::v4f32, Expand);
    setOperationAction(ISD::FPOW, MVT::v4f32, Expand);
    setOperationAction(ISD::FLOG, MVT::v4f32, Expand);
    setOperationAction(ISD::FLOG2, MVT::v4f32, Expand);
    setOperationAction(ISD::FLOG10, MVT::v4f32, Expand);
    setOperationAction(ISD::FEXP, MVT::v4f32, Expand);
    setOperationAction(ISD::FEXP2, MVT::v4f32, Expand);
    setOperationAction(ISD::FCEIL, MVT::v4f32, Expand);
    setOperationAction(ISD::FTRUNC, MVT::v4f32, Expand);
    setOperationAction(ISD::FRINT, MVT::v4f32, Expand);
    setOperationAction(ISD::FNEARBYINT, MVT::v4f32, Expand);
    setOperationAction(ISD::FFLOOR, MVT::v4f32, Expand);
 
    // Mark v2f32 intrinsics.
    setOperationAction(ISD::FSQRT, MVT::v2f32, Expand);
    setOperationAction(ISD::FSIN, MVT::v2f32, Expand);
    setOperationAction(ISD::FCOS, MVT::v2f32, Expand);
    setOperationAction(ISD::FPOW, MVT::v2f32, Expand);
    setOperationAction(ISD::FLOG, MVT::v2f32, Expand);
    setOperationAction(ISD::FLOG2, MVT::v2f32, Expand);
    setOperationAction(ISD::FLOG10, MVT::v2f32, Expand);
    setOperationAction(ISD::FEXP, MVT::v2f32, Expand);
    setOperationAction(ISD::FEXP2, MVT::v2f32, Expand);
    setOperationAction(ISD::FCEIL, MVT::v2f32, Expand);
    setOperationAction(ISD::FTRUNC, MVT::v2f32, Expand);
    setOperationAction(ISD::FRINT, MVT::v2f32, Expand);
    setOperationAction(ISD::FNEARBYINT, MVT::v2f32, Expand);
    setOperationAction(ISD::FFLOOR, MVT::v2f32, Expand);
 
    // Neon does not support some operations on v1i64 and v2i64 types.
    setOperationAction(ISD::MUL, MVT::v1i64, Expand);
    // Custom handling for some quad-vector types to detect VMULL.
    setOperationAction(ISD::MUL, MVT::v8i16, Custom);
    setOperationAction(ISD::MUL, MVT::v4i32, Custom);
    setOperationAction(ISD::MUL, MVT::v2i64, Custom);
    // Custom handling for some vector types to avoid expensive expansions
    setOperationAction(ISD::SDIV, MVT::v4i16, Custom);
    setOperationAction(ISD::SDIV, MVT::v8i8, Custom);
    setOperationAction(ISD::UDIV, MVT::v4i16, Custom);
    setOperationAction(ISD::UDIV, MVT::v8i8, Custom);
    // Neon does not have single instruction SINT_TO_FP and UINT_TO_FP with
    // a destination type that is wider than the source, and nor does
    // it have a FP_TO_[SU]INT instruction with a narrower destination than
    // source.
    setOperationAction(ISD::SINT_TO_FP, MVT::v4i16, Custom);
    setOperationAction(ISD::SINT_TO_FP, MVT::v8i16, Custom);
    setOperationAction(ISD::UINT_TO_FP, MVT::v4i16, Custom);
    setOperationAction(ISD::UINT_TO_FP, MVT::v8i16, Custom);
    setOperationAction(ISD::FP_TO_UINT, MVT::v4i16, Custom);
    setOperationAction(ISD::FP_TO_UINT, MVT::v8i16, Custom);
    setOperationAction(ISD::FP_TO_SINT, MVT::v4i16, Custom);
    setOperationAction(ISD::FP_TO_SINT, MVT::v8i16, Custom);
 
    setOperationAction(ISD::FP_ROUND,   MVT::v2f32, Expand);
    setOperationAction(ISD::FP_EXTEND,  MVT::v2f64, Expand);
 
    // NEON does not have single instruction CTPOP for vectors with element
    // types wider than 8-bits.  However, custom lowering can leverage the
    // v8i8/v16i8 vcnt instruction.
    setOperationAction(ISD::CTPOP,      MVT::v2i32, Custom);
    setOperationAction(ISD::CTPOP,      MVT::v4i32, Custom);
    setOperationAction(ISD::CTPOP,      MVT::v4i16, Custom);
    setOperationAction(ISD::CTPOP,      MVT::v8i16, Custom);
    setOperationAction(ISD::CTPOP,      MVT::v1i64, Custom);
    setOperationAction(ISD::CTPOP,      MVT::v2i64, Custom);
 
    setOperationAction(ISD::CTLZ,       MVT::v1i64, Expand);
    setOperationAction(ISD::CTLZ,       MVT::v2i64, Expand);
 
    // NEON does not have single instruction CTTZ for vectors.
    setOperationAction(ISD::CTTZ, MVT::v8i8, Custom);
    setOperationAction(ISD::CTTZ, MVT::v4i16, Custom);
    setOperationAction(ISD::CTTZ, MVT::v2i32, Custom);
    setOperationAction(ISD::CTTZ, MVT::v1i64, Custom);
 
    setOperationAction(ISD::CTTZ, MVT::v16i8, Custom);
    setOperationAction(ISD::CTTZ, MVT::v8i16, Custom);
    setOperationAction(ISD::CTTZ, MVT::v4i32, Custom);
    setOperationAction(ISD::CTTZ, MVT::v2i64, Custom);
 
    setOperationAction(ISD::CTTZ_ZERO_UNDEF, MVT::v8i8, Custom);
    setOperationAction(ISD::CTTZ_ZERO_UNDEF, MVT::v4i16, Custom);
    setOperationAction(ISD::CTTZ_ZERO_UNDEF, MVT::v2i32, Custom);
    setOperationAction(ISD::CTTZ_ZERO_UNDEF, MVT::v1i64, Custom);
 
    setOperationAction(ISD::CTTZ_ZERO_UNDEF, MVT::v16i8, Custom);
    setOperationAction(ISD::CTTZ_ZERO_UNDEF, MVT::v8i16, Custom);
    setOperationAction(ISD::CTTZ_ZERO_UNDEF, MVT::v4i32, Custom);
    setOperationAction(ISD::CTTZ_ZERO_UNDEF, MVT::v2i64, Custom);
 
    for (MVT VT : MVT::fixedlen_vector_valuetypes()) {
      setOperationAction(ISD::MULHS, VT, Expand);
      setOperationAction(ISD::MULHU, VT, Expand);
    }
 
    // NEON only has FMA instructions as of VFP4.
    if (!Subtarget->hasVFP4Base()) {
      setOperationAction(ISD::FMA, MVT::v2f32, Expand);
      setOperationAction(ISD::FMA, MVT::v4f32, Expand);
    }
 
    setTargetDAGCombine({ISD::SHL, ISD::SRL, ISD::SRA, ISD::FP_TO_SINT,
                         ISD::FP_TO_UINT, ISD::FDIV, ISD::LOAD});
 
    // It is legal to extload from v4i8 to v4i16 or v4i32.
    for (MVT Ty : {MVT::v8i8, MVT::v4i8, MVT::v2i8, MVT::v4i16, MVT::v2i16,
                   MVT::v2i32}) {
      for (MVT VT : MVT::integer_fixedlen_vector_valuetypes()) {
        setLoadExtAction(ISD::EXTLOAD, VT, Ty, Legal);
        setLoadExtAction(ISD::ZEXTLOAD, VT, Ty, Legal);
        setLoadExtAction(ISD::SEXTLOAD, VT, Ty, Legal);
      }
    }
  }
 
  if (Subtarget->hasNEON() || Subtarget->hasMVEIntegerOps()) {
    setTargetDAGCombine(
        {ISD::BUILD_VECTOR, ISD::VECTOR_SHUFFLE, ISD::INSERT_SUBVECTOR,
         ISD::INSERT_VECTOR_ELT, ISD::EXTRACT_VECTOR_ELT,
         ISD::SIGN_EXTEND_INREG, ISD::STORE, ISD::SIGN_EXTEND, ISD::ZERO_EXTEND,
         ISD::ANY_EXTEND, ISD::INTRINSIC_WO_CHAIN, ISD::INTRINSIC_W_CHAIN,
         ISD::INTRINSIC_VOID, ISD::VECREDUCE_ADD, ISD::ADD, ISD::BITCAST});
  }
  if (Subtarget->hasMVEIntegerOps()) {
    setTargetDAGCombine({ISD::SMIN, ISD::UMIN, ISD::SMAX, ISD::UMAX,
                         ISD::FP_EXTEND, ISD::SELECT, ISD::SELECT_CC,
                         ISD::SETCC});
  }
  if (Subtarget->hasMVEFloatOps()) {
    setTargetDAGCombine(ISD::FADD);
  }
 
  if (!Subtarget->hasFP64()) {
    // When targeting a floating-point unit with only single-precision
    // operations, f64 is legal for the few double-precision instructions which
    // are present However, no double-precision operations other than moves,
    // loads and stores are provided by the hardware.
    setOperationAction(ISD::FADD,       MVT::f64, Expand);
    setOperationAction(ISD::FSUB,       MVT::f64, Expand);
    setOperationAction(ISD::FMUL,       MVT::f64, Expand);
    setOperationAction(ISD::FMA,        MVT::f64, Expand);
    setOperationAction(ISD::FDIV,       MVT::f64, Expand);
    setOperationAction(ISD::FREM,       MVT::f64, Expand);
    setOperationAction(ISD::FCOPYSIGN,  MVT::f64, Expand);
    setOperationAction(ISD::FGETSIGN,   MVT::f64, Expand);
    setOperationAction(ISD::FNEG,       MVT::f64, Expand);
    setOperationAction(ISD::FABS,       MVT::f64, Expand);
    setOperationAction(ISD::FSQRT,      MVT::f64, Expand);
    setOperationAction(ISD::FSIN,       MVT::f64, Expand);
    setOperationAction(ISD::FCOS,       MVT::f64, Expand);
    setOperationAction(ISD::FPOW,       MVT::f64, Expand);
    setOperationAction(ISD::FLOG,       MVT::f64, Expand);
    setOperationAction(ISD::FLOG2,      MVT::f64, Expand);
    setOperationAction(ISD::FLOG10,     MVT::f64, Expand);
    setOperationAction(ISD::FEXP,       MVT::f64, Expand);
    setOperationAction(ISD::FEXP2,      MVT::f64, Expand);
    setOperationAction(ISD::FCEIL,      MVT::f64, Expand);
    setOperationAction(ISD::FTRUNC,     MVT::f64, Expand);
    setOperationAction(ISD::FRINT,      MVT::f64, Expand);
    setOperationAction(ISD::FNEARBYINT, MVT::f64, Expand);
    setOperationAction(ISD::FFLOOR,     MVT::f64, Expand);
    setOperationAction(ISD::SINT_TO_FP, MVT::i32, Custom);
    setOperationAction(ISD::UINT_TO_FP, MVT::i32, Custom);
    setOperationAction(ISD::FP_TO_SINT, MVT::i32, Custom);
    setOperationAction(ISD::FP_TO_UINT, MVT::i32, Custom);
    setOperationAction(ISD::FP_TO_SINT, MVT::f64, Custom);
    setOperationAction(ISD::FP_TO_UINT, MVT::f64, Custom);
    setOperationAction(ISD::FP_ROUND,   MVT::f32, Custom);
    setOperationAction(ISD::STRICT_FP_TO_SINT, MVT::i32, Custom);
    setOperationAction(ISD::STRICT_FP_TO_UINT, MVT::i32, Custom);
    setOperationAction(ISD::STRICT_FP_TO_SINT, MVT::f64, Custom);
    setOperationAction(ISD::STRICT_FP_TO_UINT, MVT::f64, Custom);
    setOperationAction(ISD::STRICT_FP_ROUND,   MVT::f32, Custom);
  }
 
  if (!Subtarget->hasFP64() || !Subtarget->hasFPARMv8Base()) {
    setOperationAction(ISD::FP_EXTEND,  MVT::f64, Custom);
    setOperationAction(ISD::STRICT_FP_EXTEND, MVT::f64, Custom);
    if (Subtarget->hasFullFP16()) {
      setOperationAction(ISD::FP_ROUND,  MVT::f16, Custom);
      setOperationAction(ISD::STRICT_FP_ROUND, MVT::f16, Custom);
    }
  }
 
  if (!Subtarget->hasFP16()) {
    setOperationAction(ISD::FP_EXTEND,  MVT::f32, Custom);
    setOperationAction(ISD::STRICT_FP_EXTEND, MVT::f32, Custom);
  }
 
  computeRegisterProperties(Subtarget->getRegisterInfo());
 
  // ARM does not have floating-point extending loads.
  for (MVT VT : MVT::fp_valuetypes()) {
    setLoadExtAction(ISD::EXTLOAD, VT, MVT::f32, Expand);
    setLoadExtAction(ISD::EXTLOAD, VT, MVT::f16, Expand);
  }
 
  // ... or truncating stores
  setTruncStoreAction(MVT::f64, MVT::f32, Expand);
  setTruncStoreAction(MVT::f32, MVT::f16, Expand);
  setTruncStoreAction(MVT::f64, MVT::f16, Expand);
 
  // ARM does not have i1 sign extending load.
  for (MVT VT : MVT::integer_valuetypes())
    setLoadExtAction(ISD::SEXTLOAD, VT, MVT::i1, Promote);
 
  // ARM supports all 4 flavors of integer indexed load / store.
  if (!Subtarget->isThumb1Only()) {
    for (unsigned im = (unsigned)ISD::PRE_INC;
         im != (unsigned)ISD::LAST_INDEXED_MODE; ++im) {
      setIndexedLoadAction(im,  MVT::i1,  Legal);
      setIndexedLoadAction(im,  MVT::i8,  Legal);
      setIndexedLoadAction(im,  MVT::i16, Legal);
      setIndexedLoadAction(im,  MVT::i32, Legal);
      setIndexedStoreAction(im, MVT::i1,  Legal);
      setIndexedStoreAction(im, MVT::i8,  Legal);
      setIndexedStoreAction(im, MVT::i16, Legal);
      setIndexedStoreAction(im, MVT::i32, Legal);
    }
  } else {
    // Thumb-1 has limited post-inc load/store support - LDM r0!, {r1}.
    setIndexedLoadAction(ISD::POST_INC, MVT::i32,  Legal);
    setIndexedStoreAction(ISD::POST_INC, MVT::i32,  Legal);
  }
 
  setOperationAction(ISD::SADDO, MVT::i32, Custom);
  setOperationAction(ISD::UADDO, MVT::i32, Custom);
  setOperationAction(ISD::SSUBO, MVT::i32, Custom);
  setOperationAction(ISD::USUBO, MVT::i32, Custom);
 
  setOperationAction(ISD::ADDCARRY, MVT::i32, Custom);
  setOperationAction(ISD::SUBCARRY, MVT::i32, Custom);
  if (Subtarget->hasDSP()) {
    setOperationAction(ISD::SADDSAT, MVT::i8, Custom);
    setOperationAction(ISD::SSUBSAT, MVT::i8, Custom);
    setOperationAction(ISD::SADDSAT, MVT::i16, Custom);
    setOperationAction(ISD::SSUBSAT, MVT::i16, Custom);
    setOperationAction(ISD::UADDSAT, MVT::i8, Custom);
    setOperationAction(ISD::USUBSAT, MVT::i8, Custom);
    setOperationAction(ISD::UADDSAT, MVT::i16, Custom);
    setOperationAction(ISD::USUBSAT, MVT::i16, Custom);
  }
  if (Subtarget->hasBaseDSP()) {
    setOperationAction(ISD::SADDSAT, MVT::i32, Legal);
    setOperationAction(ISD::SSUBSAT, MVT::i32, Legal);
  }
 
  // i64 operation support.
  setOperationAction(ISD::MUL,     MVT::i64, Expand);
  setOperationAction(ISD::MULHU,   MVT::i32, Expand);
  if (Subtarget->isThumb1Only()) {
    setOperationAction(ISD::UMUL_LOHI, MVT::i32, Expand);
    setOperationAction(ISD::SMUL_LOHI, MVT::i32, Expand);
  }
  if (Subtarget->isThumb1Only() || !Subtarget->hasV6Ops()
      || (Subtarget->isThumb2() && !Subtarget->hasDSP()))
    setOperationAction(ISD::MULHS, MVT::i32, Expand);
 
  setOperationAction(ISD::SHL_PARTS, MVT::i32, Custom);
  setOperationAction(ISD::SRA_PARTS, MVT::i32, Custom);
  setOperationAction(ISD::SRL_PARTS, MVT::i32, Custom);
  setOperationAction(ISD::SRL,       MVT::i64, Custom);
  setOperationAction(ISD::SRA,       MVT::i64, Custom);
  setOperationAction(ISD::INTRINSIC_VOID, MVT::Other, Custom);
  setOperationAction(ISD::INTRINSIC_WO_CHAIN, MVT::i64, Custom);
  setOperationAction(ISD::LOAD, MVT::i64, Custom);
  setOperationAction(ISD::STORE, MVT::i64, Custom);
 
  // MVE lowers 64 bit shifts to lsll and lsrl
  // assuming that ISD::SRL and SRA of i64 are already marked custom
  if (Subtarget->hasMVEIntegerOps())
    setOperationAction(ISD::SHL, MVT::i64, Custom);
 
  // Expand to __aeabi_l{lsl,lsr,asr} calls for Thumb1.
  if (Subtarget->isThumb1Only()) {
    setOperationAction(ISD::SHL_PARTS, MVT::i32, Expand);
    setOperationAction(ISD::SRA_PARTS, MVT::i32, Expand);
    setOperationAction(ISD::SRL_PARTS, MVT::i32, Expand);
  }
 
  if (!Subtarget->isThumb1Only() && Subtarget->hasV6T2Ops())
    setOperationAction(ISD::BITREVERSE, MVT::i32, Legal);
 
  // ARM does not have ROTL.
  setOperationAction(ISD::ROTL, MVT::i32, Expand);
  for (MVT VT : MVT::fixedlen_vector_valuetypes()) {
    setOperationAction(ISD::ROTL, VT, Expand);
    setOperationAction(ISD::ROTR, VT, Expand);
  }
  setOperationAction(ISD::CTTZ,  MVT::i32, Custom);
  setOperationAction(ISD::CTPOP, MVT::i32, Expand);
  if (!Subtarget->hasV5TOps() || Subtarget->isThumb1Only()) {
    setOperationAction(ISD::CTLZ, MVT::i32, Expand);
    setOperationAction(ISD::CTLZ_ZERO_UNDEF, MVT::i32, LibCall);
  }
 
  // @llvm.readcyclecounter requires the Performance Monitors extension.
  // Default to the 0 expansion on unsupported platforms.
  // FIXME: Technically there are older ARM CPUs that have
  // implementation-specific ways of obtaining this information.
  if (Subtarget->hasPerfMon())
    setOperationAction(ISD::READCYCLECOUNTER, MVT::i64, Custom);
 
  // Only ARMv6 has BSWAP.
  if (!Subtarget->hasV6Ops())
    setOperationAction(ISD::BSWAP, MVT::i32, Expand);
 
  bool hasDivide = Subtarget->isThumb() ? Subtarget->hasDivideInThumbMode()
                                        : Subtarget->hasDivideInARMMode();
  if (!hasDivide) {
    // These are expanded into libcalls if the cpu doesn't have HW divider.
    setOperationAction(ISD::SDIV,  MVT::i32, LibCall);
    setOperationAction(ISD::UDIV,  MVT::i32, LibCall);
  }
 
  if (Subtarget->isTargetWindows() && !Subtarget->hasDivideInThumbMode()) {
    setOperationAction(ISD::SDIV, MVT::i32, Custom);
    setOperationAction(ISD::UDIV, MVT::i32, Custom);
 
    setOperationAction(ISD::SDIV, MVT::i64, Custom);
    setOperationAction(ISD::UDIV, MVT::i64, Custom);
  }
 
  setOperationAction(ISD::SREM,  MVT::i32, Expand);
  setOperationAction(ISD::UREM,  MVT::i32, Expand);
 
  // Register based DivRem for AEABI (RTABI 4.2)
  if (Subtarget->isTargetAEABI() || Subtarget->isTargetAndroid() ||
      Subtarget->isTargetGNUAEABI() || Subtarget->isTargetMuslAEABI() ||
      Subtarget->isTargetWindows()) {
    setOperationAction(ISD::SREM, MVT::i64, Custom);
    setOperationAction(ISD::UREM, MVT::i64, Custom);
    HasStandaloneRem = false;
 
    if (Subtarget->isTargetWindows()) {
      const struct {
        const RTLIB::Libcall Op;
        const char * const Name;
        const CallingConv::ID CC;
      } LibraryCalls[] = {
        { RTLIB::SDIVREM_I8, "__rt_sdiv", CallingConv::ARM_AAPCS },
        { RTLIB::SDIVREM_I16, "__rt_sdiv", CallingConv::ARM_AAPCS },
        { RTLIB::SDIVREM_I32, "__rt_sdiv", CallingConv::ARM_AAPCS },
        { RTLIB::SDIVREM_I64, "__rt_sdiv64", CallingConv::ARM_AAPCS },
 
        { RTLIB::UDIVREM_I8, "__rt_udiv", CallingConv::ARM_AAPCS },
        { RTLIB::UDIVREM_I16, "__rt_udiv", CallingConv::ARM_AAPCS },
        { RTLIB::UDIVREM_I32, "__rt_udiv", CallingConv::ARM_AAPCS },
        { RTLIB::UDIVREM_I64, "__rt_udiv64", CallingConv::ARM_AAPCS },
      };
 
      for (const auto &LC : LibraryCalls) {
        setLibcallName(LC.Op, LC.Name);
        setLibcallCallingConv(LC.Op, LC.CC);
      }
    } else {
      const struct {
        const RTLIB::Libcall Op;
        const char * const Name;
        const CallingConv::ID CC;
      } LibraryCalls[] = {
        { RTLIB::SDIVREM_I8, "__aeabi_idivmod", CallingConv::ARM_AAPCS },
        { RTLIB::SDIVREM_I16, "__aeabi_idivmod", CallingConv::ARM_AAPCS },
        { RTLIB::SDIVREM_I32, "__aeabi_idivmod", CallingConv::ARM_AAPCS },
        { RTLIB::SDIVREM_I64, "__aeabi_ldivmod", CallingConv::ARM_AAPCS },
 
        { RTLIB::UDIVREM_I8, "__aeabi_uidivmod", CallingConv::ARM_AAPCS },
        { RTLIB::UDIVREM_I16, "__aeabi_uidivmod", CallingConv::ARM_AAPCS },
        { RTLIB::UDIVREM_I32, "__aeabi_uidivmod", CallingConv::ARM_AAPCS },
        { RTLIB::UDIVREM_I64, "__aeabi_uldivmod", CallingConv::ARM_AAPCS },
      };
 
      for (const auto &LC : LibraryCalls) {
        setLibcallName(LC.Op, LC.Name);
        setLibcallCallingConv(LC.Op, LC.CC);
      }
    }
 
    setOperationAction(ISD::SDIVREM, MVT::i32, Custom);
    setOperationAction(ISD::UDIVREM, MVT::i32, Custom);
    setOperationAction(ISD::SDIVREM, MVT::i64, Custom);
    setOperationAction(ISD::UDIVREM, MVT::i64, Custom);
  } else {
    setOperationAction(ISD::SDIVREM, MVT::i32, Expand);
    setOperationAction(ISD::UDIVREM, MVT::i32, Expand);
  }
 
  if (Subtarget->getTargetTriple().isOSMSVCRT()) {
    // MSVCRT doesn't have powi; fall back to pow
    setLibcallName(RTLIB::POWI_F32, nullptr);
    setLibcallName(RTLIB::POWI_F64, nullptr);
  }
 
  setOperationAction(ISD::GlobalAddress, MVT::i32,   Custom);
  setOperationAction(ISD::ConstantPool,  MVT::i32,   Custom);
  setOperationAction(ISD::GlobalTLSAddress, MVT::i32, Custom);
  setOperationAction(ISD::BlockAddress, MVT::i32, Custom);
 
  setOperationAction(ISD::TRAP, MVT::Other, Legal);
  setOperationAction(ISD::DEBUGTRAP, MVT::Other, Legal);
 
  // Use the default implementation.
  setOperationAction(ISD::VASTART,            MVT::Other, Custom);
  setOperationAction(ISD::VAARG,              MVT::Other, Expand);
  setOperationAction(ISD::VACOPY,             MVT::Other, Expand);
  setOperationAction(ISD::VAEND,              MVT::Other, Expand);
  setOperationAction(ISD::STACKSAVE,          MVT::Other, Expand);
  setOperationAction(ISD::STACKRESTORE,       MVT::Other, Expand);
 
  if (Subtarget->isTargetWindows())
    setOperationAction(ISD::DYNAMIC_STACKALLOC, MVT::i32, Custom);
  else
    setOperationAction(ISD::DYNAMIC_STACKALLOC, MVT::i32, Expand);
 
  // ARMv6 Thumb1 (except for CPUs that support dmb / dsb) and earlier use
  // the default expansion.
  InsertFencesForAtomic = false;
  if (Subtarget->hasAnyDataBarrier() &&
      (!Subtarget->isThumb() || Subtarget->hasV8MBaselineOps())) {
    // ATOMIC_FENCE needs custom lowering; the others should have been expanded
    // to ldrex/strex loops already.
    setOperationAction(ISD::ATOMIC_FENCE,     MVT::Other, Custom);
    if (!Subtarget->isThumb() || !Subtarget->isMClass())
      setOperationAction(ISD::ATOMIC_CMP_SWAP,  MVT::i64, Custom);
 
    // On v8, we have particularly efficient implementations of atomic fences
    // if they can be combined with nearby atomic loads and stores.
    if (!Subtarget->hasAcquireRelease() ||
        getTargetMachine().getOptLevel() == 0) {
      // Automatically insert fences (dmb ish) around ATOMIC_SWAP etc.
      InsertFencesForAtomic = true;
    }
  } else {
    // If there's anything we can use as a barrier, go through custom lowering
    // for ATOMIC_FENCE.
    // If target has DMB in thumb, Fences can be inserted.
    if (Subtarget->hasDataBarrier())
      InsertFencesForAtomic = true;
 
    setOperationAction(ISD::ATOMIC_FENCE,   MVT::Other,
                       Subtarget->hasAnyDataBarrier() ? Custom : Expand);
 
    // Set them all for expansion, which will force libcalls.
    setOperationAction(ISD::ATOMIC_CMP_SWAP,  MVT::i32, Expand);
    setOperationAction(ISD::ATOMIC_SWAP,      MVT::i32, Expand);
    setOperationAction(ISD::ATOMIC_LOAD_ADD,  MVT::i32, Expand);
    setOperationAction(ISD::ATOMIC_LOAD_SUB,  MVT::i32, Expand);
    setOperationAction(ISD::ATOMIC_LOAD_AND,  MVT::i32, Expand);
    setOperationAction(ISD::ATOMIC_LOAD_OR,   MVT::i32, Expand);
    setOperationAction(ISD::ATOMIC_LOAD_XOR,  MVT::i32, Expand);
    setOperationAction(ISD::ATOMIC_LOAD_NAND, MVT::i32, Expand);
    setOperationAction(ISD::ATOMIC_LOAD_MIN, MVT::i32, Expand);
    setOperationAction(ISD::ATOMIC_LOAD_MAX, MVT::i32, Expand);
    setOperationAction(ISD::ATOMIC_LOAD_UMIN, MVT::i32, Expand);
    setOperationAction(ISD::ATOMIC_LOAD_UMAX, MVT::i32, Expand);
    // Mark ATOMIC_LOAD and ATOMIC_STORE custom so we can handle the
    // Unordered/Monotonic case.
    if (!InsertFencesForAtomic) {
      setOperationAction(ISD::ATOMIC_LOAD, MVT::i32, Custom);
      setOperationAction(ISD::ATOMIC_STORE, MVT::i32, Custom);
    }
  }
 
  // Compute supported atomic widths.
  if (Subtarget->isTargetLinux() ||
      (!Subtarget->isMClass() && Subtarget->hasV6Ops())) {
    // For targets where __sync_* routines are reliably available, we use them
    // if necessary.
    //
    // ARM Linux always supports 64-bit atomics through kernel-assisted atomic
    // routines (kernel 3.1 or later). FIXME: Not with compiler-rt?
    //
    // ARMv6 targets have native instructions in ARM mode. For Thumb mode,
    // such targets should provide __sync_* routines, which use the ARM mode
    // instructions. (ARMv6 doesn't have dmb, but it has an equivalent
    // encoding; see ARMISD::MEMBARRIER_MCR.)
    setMaxAtomicSizeInBitsSupported(64);
  } else if ((Subtarget->isMClass() && Subtarget->hasV8MBaselineOps()) ||
             Subtarget->hasForced32BitAtomics()) {
    // Cortex-M (besides Cortex-M0) have 32-bit atomics.
    setMaxAtomicSizeInBitsSupported(32);
  } else {
    // We can't assume anything about other targets; just use libatomic
    // routines.
    setMaxAtomicSizeInBitsSupported(0);
  }
 
  setMaxDivRemBitWidthSupported(64);
 
  setOperationAction(ISD::PREFETCH,         MVT::Other, Custom);
 
  // Requires SXTB/SXTH, available on v6 and up in both ARM and Thumb modes.
  if (!Subtarget->hasV6Ops()) {
    setOperationAction(ISD::SIGN_EXTEND_INREG, MVT::i16, Expand);
    setOperationAction(ISD::SIGN_EXTEND_INREG, MVT::i8,  Expand);
  }
  setOperationAction(ISD::SIGN_EXTEND_INREG, MVT::i1, Expand);
 
  if (!Subtarget->useSoftFloat() && Subtarget->hasFPRegs() &&
      !Subtarget->isThumb1Only()) {
    // Turn f64->i64 into VMOVRRD, i64 -> f64 to VMOVDRR
    // iff target supports vfp2.
    setOperationAction(ISD::BITCAST, MVT::i64, Custom);
    setOperationAction(ISD::FLT_ROUNDS_, MVT::i32, Custom);
    setOperationAction(ISD::SET_ROUNDING, MVT::Other, Custom);
  }
 
  // We want to custom lower some of our intrinsics.
  setOperationAction(ISD::INTRINSIC_WO_CHAIN, MVT::Other, Custom);
  setOperationAction(ISD::EH_SJLJ_SETJMP, MVT::i32, Custom);
  setOperationAction(ISD::EH_SJLJ_LONGJMP, MVT::Other, Custom);
  setOperationAction(ISD::EH_SJLJ_SETUP_DISPATCH, MVT::Other, Custom);
  if (Subtarget->useSjLjEH())
    setLibcallName(RTLIB::UNWIND_RESUME, "_Unwind_SjLj_Resume");
 
  setOperationAction(ISD::SETCC,     MVT::i32, Expand);
  setOperationAction(ISD::SETCC,     MVT::f32, Expand);
  setOperationAction(ISD::SETCC,     MVT::f64, Expand);
  setOperationAction(ISD::SELECT,    MVT::i32, Custom);
  setOperationAction(ISD::SELECT,    MVT::f32, Custom);
  setOperationAction(ISD::SELECT,    MVT::f64, Custom);
  setOperationAction(ISD::SELECT_CC, MVT::i32, Custom);
  setOperationAction(ISD::SELECT_CC, MVT::f32, Custom);
  setOperationAction(ISD::SELECT_CC, MVT::f64, Custom);
  if (Subtarget->hasFullFP16()) {
    setOperationAction(ISD::SETCC,     MVT::f16, Expand);
    setOperationAction(ISD::SELECT,    MVT::f16, Custom);
    setOperationAction(ISD::SELECT_CC, MVT::f16, Custom);
  }
 
  setOperationAction(ISD::SETCCCARRY, MVT::i32, Custom);
 
  setOperationAction(ISD::BRCOND,    MVT::Other, Custom);
  setOperationAction(ISD::BR_CC,     MVT::i32,   Custom);
  if (Subtarget->hasFullFP16())
      setOperationAction(ISD::BR_CC, MVT::f16,   Custom);
  setOperationAction(ISD::BR_CC,     MVT::f32,   Custom);
  setOperationAction(ISD::BR_CC,     MVT::f64,   Custom);
  setOperationAction(ISD::BR_JT,     MVT::Other, Custom);
 
  // We don't support sin/cos/fmod/copysign/pow
  setOperationAction(ISD::FSIN,      MVT::f64, Expand);
  setOperationAction(ISD::FSIN,      MVT::f32, Expand);
  setOperationAction(ISD::FCOS,      MVT::f32, Expand);
  setOperationAction(ISD::FCOS,      MVT::f64, Expand);
  setOperationAction(ISD::FSINCOS,   MVT::f64, Expand);
  setOperationAction(ISD::FSINCOS,   MVT::f32, Expand);
  setOperationAction(ISD::FREM,      MVT::f64, Expand);
  setOperationAction(ISD::FREM,      MVT::f32, Expand);
  if (!Subtarget->useSoftFloat() && Subtarget->hasVFP2Base() &&
      !Subtarget->isThumb1Only()) {
    setOperationAction(ISD::FCOPYSIGN, MVT::f64, Custom);
    setOperationAction(ISD::FCOPYSIGN, MVT::f32, Custom);
  }
  setOperationAction(ISD::FPOW,      MVT::f64, Expand);
  setOperationAction(ISD::FPOW,      MVT::f32, Expand);
 
  if (!Subtarget->hasVFP4Base()) {
    setOperationAction(ISD::FMA, MVT::f64, Expand);
    setOperationAction(ISD::FMA, MVT::f32, Expand);
  }
 
  // Various VFP goodness
  if (!Subtarget->useSoftFloat() && !Subtarget->isThumb1Only()) {
    // FP-ARMv8 adds f64 <-> f16 conversion. Before that it should be expanded.
    if (!Subtarget->hasFPARMv8Base() || !Subtarget->hasFP64()) {
      setOperationAction(ISD::FP16_TO_FP, MVT::f64, Expand);
      setOperationAction(ISD::FP_TO_FP16, MVT::f64, Expand);
    }
 
    // fp16 is a special v7 extension that adds f16 <-> f32 conversions.
    if (!Subtarget->hasFP16()) {
      setOperationAction(ISD::FP16_TO_FP, MVT::f32, Expand);
      setOperationAction(ISD::FP_TO_FP16, MVT::f32, Expand);
    }
 
    // Strict floating-point comparisons need custom lowering.
    setOperationAction(ISD::STRICT_FSETCC,  MVT::f16, Custom);
    setOperationAction(ISD::STRICT_FSETCCS, MVT::f16, Custom);
    setOperationAction(ISD::STRICT_FSETCC,  MVT::f32, Custom);
    setOperationAction(ISD::STRICT_FSETCCS, MVT::f32, Custom);
    setOperationAction(ISD::STRICT_FSETCC,  MVT::f64, Custom);
    setOperationAction(ISD::STRICT_FSETCCS, MVT::f64, Custom);
  }
 
  // Use __sincos_stret if available.
  if (getLibcallName(RTLIB::SINCOS_STRET_F32) != nullptr &&
      getLibcallName(RTLIB::SINCOS_STRET_F64) != nullptr) {
    setOperationAction(ISD::FSINCOS, MVT::f64, Custom);
    setOperationAction(ISD::FSINCOS, MVT::f32, Custom);
  }
 
  // FP-ARMv8 implements a lot of rounding-like FP operations.
  if (Subtarget->hasFPARMv8Base()) {
    setOperationAction(ISD::FFLOOR, MVT::f32, Legal);
    setOperationAction(ISD::FCEIL, MVT::f32, Legal);
    setOperationAction(ISD::FROUND, MVT::f32, Legal);
    setOperationAction(ISD::FTRUNC, MVT::f32, Legal);
    setOperationAction(ISD::FNEARBYINT, MVT::f32, Legal);
    setOperationAction(ISD::FRINT, MVT::f32, Legal);
    setOperationAction(ISD::FMINNUM, MVT::f32, Legal);
    setOperationAction(ISD::FMAXNUM, MVT::f32, Legal);
    if (Subtarget->hasNEON()) {
      setOperationAction(ISD::FMINNUM, MVT::v2f32, Legal);
      setOperationAction(ISD::FMAXNUM, MVT::v2f32, Legal);
      setOperationAction(ISD::FMINNUM, MVT::v4f32, Legal);
      setOperationAction(ISD::FMAXNUM, MVT::v4f32, Legal);
    }
 
    if (Subtarget->hasFP64()) {
      setOperationAction(ISD::FFLOOR, MVT::f64, Legal);
      setOperationAction(ISD::FCEIL, MVT::f64, Legal);
      setOperationAction(ISD::FROUND, MVT::f64, Legal);
      setOperationAction(ISD::FTRUNC, MVT::f64, Legal);
      setOperationAction(ISD::FNEARBYINT, MVT::f64, Legal);
      setOperationAction(ISD::FRINT, MVT::f64, Legal);
      setOperationAction(ISD::FMINNUM, MVT::f64, Legal);
      setOperationAction(ISD::FMAXNUM, MVT::f64, Legal);
    }
  }
 
  // FP16 often need to be promoted to call lib functions
  if (Subtarget->hasFullFP16()) {
    setOperationAction(ISD::FREM, MVT::f16, Promote);
    setOperationAction(ISD::FCOPYSIGN, MVT::f16, Expand);
    setOperationAction(ISD::FSIN, MVT::f16, Promote);
    setOperationAction(ISD::FCOS, MVT::f16, Promote);
    setOperationAction(ISD::FSINCOS, MVT::f16, Promote);
    setOperationAction(ISD::FPOWI, MVT::f16, Promote);
    setOperationAction(ISD::FPOW, MVT::f16, Promote);
    setOperationAction(ISD::FEXP, MVT::f16, Promote);
    setOperationAction(ISD::FEXP2, MVT::f16, Promote);
    setOperationAction(ISD::FLOG, MVT::f16, Promote);
    setOperationAction(ISD::FLOG10, MVT::f16, Promote);
    setOperationAction(ISD::FLOG2, MVT::f16, Promote);
 
    setOperationAction(ISD::FROUND, MVT::f16, Legal);
  }
 
  if (Subtarget->hasNEON()) {
    // vmin and vmax aren't available in a scalar form, so we can use
    // a NEON instruction with an undef lane instead.  This has a performance
    // penalty on some cores, so we don't do this unless we have been
    // asked to by the core tuning model.
    if (Subtarget->useNEONForSinglePrecisionFP()) {
      setOperationAction(ISD::FMINIMUM, MVT::f32, Legal);
      setOperationAction(ISD::FMAXIMUM, MVT::f32, Legal);
      setOperationAction(ISD::FMINIMUM, MVT::f16, Legal);
      setOperationAction(ISD::FMAXIMUM, MVT::f16, Legal);
    }
    setOperationAction(ISD::FMINIMUM, MVT::v2f32, Legal);
    setOperationAction(ISD::FMAXIMUM, MVT::v2f32, Legal);
    setOperationAction(ISD::FMINIMUM, MVT::v4f32, Legal);
    setOperationAction(ISD::FMAXIMUM, MVT::v4f32, Legal);
 
    if (Subtarget->hasFullFP16()) {
      setOperationAction(ISD::FMINNUM, MVT::v4f16, Legal);
      setOperationAction(ISD::FMAXNUM, MVT::v4f16, Legal);
      setOperationAction(ISD::FMINNUM, MVT::v8f16, Legal);
      setOperationAction(ISD::FMAXNUM, MVT::v8f16, Legal);
 
      setOperationAction(ISD::FMINIMUM, MVT::v4f16, Legal);
      setOperationAction(ISD::FMAXIMUM, MVT::v4f16, Legal);
      setOperationAction(ISD::FMINIMUM, MVT::v8f16, Legal);
      setOperationAction(ISD::FMAXIMUM, MVT::v8f16, Legal);
    }
  }
 
  // We have target-specific dag combine patterns for the following nodes:
  // ARMISD::VMOVRRD  - No need to call setTargetDAGCombine
  setTargetDAGCombine(
      {ISD::ADD, ISD::SUB, ISD::MUL, ISD::AND, ISD::OR, ISD::XOR});
 
  if (Subtarget->hasMVEIntegerOps())
    setTargetDAGCombine(ISD::VSELECT);
 
  if (Subtarget->hasV6Ops())
    setTargetDAGCombine(ISD::SRL);
  if (Subtarget->isThumb1Only())
    setTargetDAGCombine(ISD::SHL);
  // Attempt to lower smin/smax to ssat/usat
  if ((!Subtarget->isThumb() && Subtarget->hasV6Ops()) ||
      Subtarget->isThumb2()) {
    setTargetDAGCombine({ISD::SMIN, ISD::SMAX});
  }
 
  setStackPointerRegisterToSaveRestore(ARM::SP);
 
  if (Subtarget->useSoftFloat() || Subtarget->isThumb1Only() ||
      !Subtarget->hasVFP2Base() || Subtarget->hasMinSize())
    setSchedulingPreference(Sched::RegPressure);
  else
    setSchedulingPreference(Sched::Hybrid);
 
  //// temporary - rewrite interface to use type
  MaxStoresPerMemset = 8;
  MaxStoresPerMemsetOptSize = 4;
  MaxStoresPerMemcpy = 4; // For @llvm.memcpy -> sequence of stores
  MaxStoresPerMemcpyOptSize = 2;
  MaxStoresPerMemmove = 4; // For @llvm.memmove -> sequence of stores
  MaxStoresPerMemmoveOptSize = 2;
 
  // On ARM arguments smaller than 4 bytes are extended, so all arguments
  // are at least 4 bytes aligned.
  setMinStackArgumentAlignment(Align(4));
 
  // Prefer likely predicted branches to selects on out-of-order cores.
  PredictableSelectIsExpensive = Subtarget->getSchedModel().isOutOfOrder();
 
  setPrefLoopAlignment(Align(1ULL << Subtarget->getPrefLoopLogAlignment()));
 
  setMinFunctionAlignment(Subtarget->isThumb() ? Align(2) : Align(4));
 
  if (Subtarget->isThumb() || Subtarget->isThumb2())
    setTargetDAGCombine(ISD::ABS);
}
 
bool ARMTargetLowering::useSoftFloat() const {
  return Subtarget->useSoftFloat();
}
 
// FIXME: It might make sense to define the representative register class as the
// nearest super-register that has a non-null superset. For example, DPR_VFP2 is
// a super-register of SPR, and DPR is a superset if DPR_VFP2. Consequently,
// SPR's representative would be DPR_VFP2. This should work well if register
// pressure tracking were modified such that a register use would increment the
// pressure of the register class's representative and all of it's super
// classes' representatives transitively. We have not implemented this because
// of the difficulty prior to coalescing of modeling operand register classes
// due to the common occurrence of cross class copies and subregister insertions
// and extractions.
std::pair<const TargetRegisterClass *, uint8_t>
ARMTargetLowering::findRepresentativeClass(const TargetRegisterInfo *TRI,
                                           MVT VT) const {
  const TargetRegisterClass *RRC = nullptr;
  uint8_t Cost = 1;
  switch (VT.SimpleTy) {
  default:
    return TargetLowering::findRepresentativeClass(TRI, VT);
  // Use DPR as representative register class for all floating point
  // and vector types. Since there are 32 SPR registers and 32 DPR registers so
  // the cost is 1 for both f32 and f64.
  case MVT::f32: case MVT::f64: case MVT::v8i8: case MVT::v4i16:
  case MVT::v2i32: case MVT::v1i64: case MVT::v2f32:
    RRC = &ARM::DPRRegClass;
    // When NEON is used for SP, only half of the register file is available
    // because operations that define both SP and DP results will be constrained
    // to the VFP2 class (D0-D15). We currently model this constraint prior to
    // coalescing by double-counting the SP regs. See the FIXME above.
    if (Subtarget->useNEONForSinglePrecisionFP())
      Cost = 2;
    break;
  case MVT::v16i8: case MVT::v8i16: case MVT::v4i32: case MVT::v2i64:
  case MVT::v4f32: case MVT::v2f64:
    RRC = &ARM::DPRRegClass;
    Cost = 2;
    break;
  case MVT::v4i64:
    RRC = &ARM::DPRRegClass;
    Cost = 4;
    break;
  case MVT::v8i64:
    RRC = &ARM::DPRRegClass;
    Cost = 8;
    break;
  }
  return std::make_pair(RRC, Cost);
}
 
const char *ARMTargetLowering::getTargetNodeName(unsigned Opcode) const {
#define MAKE_CASE(V)                                                           \
  case V:                                                                      \
    return #V;
  switch ((ARMISD::NodeType)Opcode) {
  case ARMISD::FIRST_NUMBER:
    break;
    MAKE_CASE(ARMISD::Wrapper)
    MAKE_CASE(ARMISD::WrapperPIC)
    MAKE_CASE(ARMISD::WrapperJT)
    MAKE_CASE(ARMISD::COPY_STRUCT_BYVAL)
    MAKE_CASE(ARMISD::CALL)
    MAKE_CASE(ARMISD::CALL_PRED)
    MAKE_CASE(ARMISD::CALL_NOLINK)
    MAKE_CASE(ARMISD::tSECALL)
    MAKE_CASE(ARMISD::t2CALL_BTI)
    MAKE_CASE(ARMISD::BRCOND)
    MAKE_CASE(ARMISD::BR_JT)
    MAKE_CASE(ARMISD::BR2_JT)
    MAKE_CASE(ARMISD::RET_FLAG)
    MAKE_CASE(ARMISD::SERET_FLAG)
    MAKE_CASE(ARMISD::INTRET_FLAG)
    MAKE_CASE(ARMISD::PIC_ADD)
    MAKE_CASE(ARMISD::CMP)
    MAKE_CASE(ARMISD::CMN)
    MAKE_CASE(ARMISD::CMPZ)
    MAKE_CASE(ARMISD::CMPFP)
    MAKE_CASE(ARMISD::CMPFPE)
    MAKE_CASE(ARMISD::CMPFPw0)
    MAKE_CASE(ARMISD::CMPFPEw0)
    MAKE_CASE(ARMISD::BCC_i64)
    MAKE_CASE(ARMISD::FMSTAT)
    MAKE_CASE(ARMISD::CMOV)
    MAKE_CASE(ARMISD::SUBS)
    MAKE_CASE(ARMISD::SSAT)
    MAKE_CASE(ARMISD::USAT)
    MAKE_CASE(ARMISD::ASRL)
    MAKE_CASE(ARMISD::LSRL)
    MAKE_CASE(ARMISD::LSLL)
    MAKE_CASE(ARMISD::SRL_FLAG)
    MAKE_CASE(ARMISD::SRA_FLAG)
    MAKE_CASE(ARMISD::RRX)
    MAKE_CASE(ARMISD::ADDC)
    MAKE_CASE(ARMISD::ADDE)
    MAKE_CASE(ARMISD::SUBC)
    MAKE_CASE(ARMISD::SUBE)
    MAKE_CASE(ARMISD::LSLS)
    MAKE_CASE(ARMISD::VMOVRRD)
    MAKE_CASE(ARMISD::VMOVDRR)
    MAKE_CASE(ARMISD::VMOVhr)
    MAKE_CASE(ARMISD::VMOVrh)
    MAKE_CASE(ARMISD::VMOVSR)
    MAKE_CASE(ARMISD::EH_SJLJ_SETJMP)
    MAKE_CASE(ARMISD::EH_SJLJ_LONGJMP)
    MAKE_CASE(ARMISD::EH_SJLJ_SETUP_DISPATCH)
    MAKE_CASE(ARMISD::TC_RETURN)
    MAKE_CASE(ARMISD::THREAD_POINTER)
    MAKE_CASE(ARMISD::DYN_ALLOC)
    MAKE_CASE(ARMISD::MEMBARRIER_MCR)
    MAKE_CASE(ARMISD::PRELOAD)
    MAKE_CASE(ARMISD::LDRD)
    MAKE_CASE(ARMISD::STRD)
    MAKE_CASE(ARMISD::WIN__CHKSTK)
    MAKE_CASE(ARMISD::WIN__DBZCHK)
    MAKE_CASE(ARMISD::PREDICATE_CAST)
    MAKE_CASE(ARMISD::VECTOR_REG_CAST)
    MAKE_CASE(ARMISD::MVESEXT)
    MAKE_CASE(ARMISD::MVEZEXT)
    MAKE_CASE(ARMISD::MVETRUNC)
    MAKE_CASE(ARMISD::VCMP)
    MAKE_CASE(ARMISD::VCMPZ)
    MAKE_CASE(ARMISD::VTST)
    MAKE_CASE(ARMISD::VSHLs)
    MAKE_CASE(ARMISD::VSHLu)
    MAKE_CASE(ARMISD::VSHLIMM)
    MAKE_CASE(ARMISD::VSHRsIMM)
    MAKE_CASE(ARMISD::VSHRuIMM)
    MAKE_CASE(ARMISD::VRSHRsIMM)
    MAKE_CASE(ARMISD::VRSHRuIMM)
    MAKE_CASE(ARMISD::VRSHRNIMM)
    MAKE_CASE(ARMISD::VQSHLsIMM)
    MAKE_CASE(ARMISD::VQSHLuIMM)
    MAKE_CASE(ARMISD::VQSHLsuIMM)
    MAKE_CASE(ARMISD::VQSHRNsIMM)
    MAKE_CASE(ARMISD::VQSHRNuIMM)
    MAKE_CASE(ARMISD::VQSHRNsuIMM)
    MAKE_CASE(ARMISD::VQRSHRNsIMM)
    MAKE_CASE(ARMISD::VQRSHRNuIMM)
    MAKE_CASE(ARMISD::VQRSHRNsuIMM)
    MAKE_CASE(ARMISD::VSLIIMM)
    MAKE_CASE(ARMISD::VSRIIMM)
    MAKE_CASE(ARMISD::VGETLANEu)
    MAKE_CASE(ARMISD::VGETLANEs)
    MAKE_CASE(ARMISD::VMOVIMM)
    MAKE_CASE(ARMISD::VMVNIMM)
    MAKE_CASE(ARMISD::VMOVFPIMM)
    MAKE_CASE(ARMISD::VDUP)
    MAKE_CASE(ARMISD::VDUPLANE)
    MAKE_CASE(ARMISD::VEXT)
    MAKE_CASE(ARMISD::VREV64)
    MAKE_CASE(ARMISD::VREV32)
    MAKE_CASE(ARMISD::VREV16)
    MAKE_CASE(ARMISD::VZIP)
    MAKE_CASE(ARMISD::VUZP)
    MAKE_CASE(ARMISD::VTRN)
    MAKE_CASE(ARMISD::VTBL1)
    MAKE_CASE(ARMISD::VTBL2)
    MAKE_CASE(ARMISD::VMOVN)
    MAKE_CASE(ARMISD::VQMOVNs)
    MAKE_CASE(ARMISD::VQMOVNu)
    MAKE_CASE(ARMISD::VCVTN)
    MAKE_CASE(ARMISD::VCVTL)
    MAKE_CASE(ARMISD::VIDUP)
    MAKE_CASE(ARMISD::VMULLs)
    MAKE_CASE(ARMISD::VMULLu)
    MAKE_CASE(ARMISD::VQDMULH)
    MAKE_CASE(ARMISD::VADDVs)
    MAKE_CASE(ARMISD::VADDVu)
    MAKE_CASE(ARMISD::VADDVps)
    MAKE_CASE(ARMISD::VADDVpu)
    MAKE_CASE(ARMISD::VADDLVs)
    MAKE_CASE(ARMISD::VADDLVu)
    MAKE_CASE(ARMISD::VADDLVAs)
    MAKE_CASE(ARMISD::VADDLVAu)
    MAKE_CASE(ARMISD::VADDLVps)
    MAKE_CASE(ARMISD::VADDLVpu)
    MAKE_CASE(ARMISD::VADDLVAps)
    MAKE_CASE(ARMISD::VADDLVApu)
    MAKE_CASE(ARMISD::VMLAVs)
    MAKE_CASE(ARMISD::VMLAVu)
    MAKE_CASE(ARMISD::VMLAVps)
    MAKE_CASE(ARMISD::VMLAVpu)
    MAKE_CASE(ARMISD::VMLALVs)
    MAKE_CASE(ARMISD::VMLALVu)
    MAKE_CASE(ARMISD::VMLALVps)
    MAKE_CASE(ARMISD::VMLALVpu)
    MAKE_CASE(ARMISD::VMLALVAs)
    MAKE_CASE(ARMISD::VMLALVAu)
    MAKE_CASE(ARMISD::VMLALVAps)
    MAKE_CASE(ARMISD::VMLALVApu)
    MAKE_CASE(ARMISD::VMINVu)
    MAKE_CASE(ARMISD::VMINVs)
    MAKE_CASE(ARMISD::VMAXVu)
    MAKE_CASE(ARMISD::VMAXVs)
    MAKE_CASE(ARMISD::UMAAL)
    MAKE_CASE(ARMISD::UMLAL)
    MAKE_CASE(ARMISD::SMLAL)
    MAKE_CASE(ARMISD::SMLALBB)
    MAKE_CASE(ARMISD::SMLALBT)
    MAKE_CASE(ARMISD::SMLALTB)
    MAKE_CASE(ARMISD::SMLALTT)
    MAKE_CASE(ARMISD::SMULWB)
    MAKE_CASE(ARMISD::SMULWT)
    MAKE_CASE(ARMISD::SMLALD)
    MAKE_CASE(ARMISD::SMLALDX)
    MAKE_CASE(ARMISD::SMLSLD)
    MAKE_CASE(ARMISD::SMLSLDX)
    MAKE_CASE(ARMISD::SMMLAR)
    MAKE_CASE(ARMISD::SMMLSR)
    MAKE_CASE(ARMISD::QADD16b)
    MAKE_CASE(ARMISD::QSUB16b)
    MAKE_CASE(ARMISD::QADD8b)
    MAKE_CASE(ARMISD::QSUB8b)
    MAKE_CASE(ARMISD::UQADD16b)
    MAKE_CASE(ARMISD::UQSUB16b)
    MAKE_CASE(ARMISD::UQADD8b)
    MAKE_CASE(ARMISD::UQSUB8b)
    MAKE_CASE(ARMISD::BUILD_VECTOR)
    MAKE_CASE(ARMISD::BFI)
    MAKE_CASE(ARMISD::VORRIMM)
    MAKE_CASE(ARMISD::VBICIMM)
    MAKE_CASE(ARMISD::VBSP)
    MAKE_CASE(ARMISD::MEMCPY)
    MAKE_CASE(ARMISD::VLD1DUP)
    MAKE_CASE(ARMISD::VLD2DUP)
    MAKE_CASE(ARMISD::VLD3DUP)
    MAKE_CASE(ARMISD::VLD4DUP)
    MAKE_CASE(ARMISD::VLD1_UPD)
    MAKE_CASE(ARMISD::VLD2_UPD)
    MAKE_CASE(ARMISD::VLD3_UPD)
    MAKE_CASE(ARMISD::VLD4_UPD)
    MAKE_CASE(ARMISD::VLD1x2_UPD)
    MAKE_CASE(ARMISD::VLD1x3_UPD)
    MAKE_CASE(ARMISD::VLD1x4_UPD)
    MAKE_CASE(ARMISD::VLD2LN_UPD)
    MAKE_CASE(ARMISD::VLD3LN_UPD)
    MAKE_CASE(ARMISD::VLD4LN_UPD)
    MAKE_CASE(ARMISD::VLD1DUP_UPD)
    MAKE_CASE(ARMISD::VLD2DUP_UPD)
    MAKE_CASE(ARMISD::VLD3DUP_UPD)
    MAKE_CASE(ARMISD::VLD4DUP_UPD)
    MAKE_CASE(ARMISD::VST1_UPD)
    MAKE_CASE(ARMISD::VST2_UPD)
    MAKE_CASE(ARMISD::VST3_UPD)
    MAKE_CASE(ARMISD::VST4_UPD)
    MAKE_CASE(ARMISD::VST1x2_UPD)
    MAKE_CASE(ARMISD::VST1x3_UPD)
    MAKE_CASE(ARMISD::VST1x4_UPD)
    MAKE_CASE(ARMISD::VST2LN_UPD)
    MAKE_CASE(ARMISD::VST3LN_UPD)
    MAKE_CASE(ARMISD::VST4LN_UPD)
    MAKE_CASE(ARMISD::WLS)
    MAKE_CASE(ARMISD::WLSSETUP)
    MAKE_CASE(ARMISD::LE)
    MAKE_CASE(ARMISD::LOOP_DEC)
    MAKE_CASE(ARMISD::CSINV)
    MAKE_CASE(ARMISD::CSNEG)
    MAKE_CASE(ARMISD::CSINC)
    MAKE_CASE(ARMISD::MEMCPYLOOP)
    MAKE_CASE(ARMISD::MEMSETLOOP)
#undef MAKE_CASE
  }
  return nullptr;
}
 
EVT ARMTargetLowering::getSetCCResultType(const DataLayout &DL, LLVMContext &,
                                          EVT VT) const {
  if (!VT.isVector())
    return getPointerTy(DL);
 
  // MVE has a predicate register.
  if ((Subtarget->hasMVEIntegerOps() &&
       (VT == MVT::v2i64 || VT == MVT::v4i32 || VT == MVT::v8i16 ||
        VT == MVT::v16i8)) ||
      (Subtarget->hasMVEFloatOps() &&
       (VT == MVT::v2f64 || VT == MVT::v4f32 || VT == MVT::v8f16)))
    return MVT::getVectorVT(MVT::i1, VT.getVectorElementCount());
  return VT.changeVectorElementTypeToInteger();
}
 
/// getRegClassFor - Return the register class that should be used for the
/// specified value type.
const TargetRegisterClass *
ARMTargetLowering::getRegClassFor(MVT VT, bool isDivergent) const {
  (void)isDivergent;
  // Map v4i64 to QQ registers but do not make the type legal. Similarly map
  // v8i64 to QQQQ registers. v4i64 and v8i64 are only used for REG_SEQUENCE to
  // load / store 4 to 8 consecutive NEON D registers, or 2 to 4 consecutive
  // MVE Q registers.
  if (Subtarget->hasNEON()) {
    if (VT == MVT::v4i64)
      return &ARM::QQPRRegClass;
    if (VT == MVT::v8i64)
      return &ARM::QQQQPRRegClass;
  }
  if (Subtarget->hasMVEIntegerOps()) {
    if (VT == MVT::v4i64)
      return &ARM::MQQPRRegClass;
    if (VT == MVT::v8i64)
      return &ARM::MQQQQPRRegClass;
  }
  return TargetLowering::getRegClassFor(VT);
}
 
// memcpy, and other memory intrinsics, typically tries to use LDM/STM if the
// source/dest is aligned and the copy size is large enough. We therefore want
// to align such objects passed to memory intrinsics.
bool ARMTargetLowering::shouldAlignPointerArgs(CallInst *CI, unsigned &MinSize,
                                               Align &PrefAlign) const {
  if (!isa<MemIntrinsic>(CI))
    return false;
  MinSize = 8;
  // On ARM11 onwards (excluding M class) 8-byte aligned LDM is typically 1
  // cycle faster than 4-byte aligned LDM.
  PrefAlign =
      (Subtarget->hasV6Ops() && !Subtarget->isMClass() ? Align(8) : Align(4));
  return true;
}
 
// Create a fast isel object.
FastISel *
ARMTargetLowering::createFastISel(FunctionLoweringInfo &funcInfo,
                                  const TargetLibraryInfo *libInfo) const {
  return ARM::createFastISel(funcInfo, libInfo);
}
 
Sched::Preference ARMTargetLowering::getSchedulingPreference(SDNode *N) const {
  unsigned NumVals = N->getNumValues();
  if (!NumVals)
    return Sched::RegPressure;
 
  for (unsigned i = 0; i != NumVals; ++i) {
    EVT VT = N->getValueType(i);
    if (VT == MVT::Glue || VT == MVT::Other)
      continue;
    if (VT.isFloatingPoint() || VT.isVector())
      return Sched::ILP;
  }
 
  if (!N->isMachineOpcode())
    return Sched::RegPressure;
 
  // Load are scheduled for latency even if there instruction itinerary
  // is not available.
  const TargetInstrInfo *TII = Subtarget->getInstrInfo();
  const MCInstrDesc &MCID = TII->get(N->getMachineOpcode());
 
  if (MCID.getNumDefs() == 0)
    return Sched::RegPressure;
  if (!Itins->isEmpty() &&
      Itins->getOperandCycle(MCID.getSchedClass(), 0) > 2)
    return Sched::ILP;
 
  return Sched::RegPressure;
}
 
//===----------------------------------------------------------------------===//
// Lowering Code
//===----------------------------------------------------------------------===//
 
static bool isSRL16(const SDValue &Op) {
  if (Op.getOpcode() != ISD::SRL)
    return false;
  if (auto Const = dyn_cast<ConstantSDNode>(Op.getOperand(1)))
    return Const->getZExtValue() == 16;
  return false;
}
 
static bool isSRA16(const SDValue &Op) {
  if (Op.getOpcode() != ISD::SRA)
    return false;
  if (auto Const = dyn_cast<ConstantSDNode>(Op.getOperand(1)))
    return Const->getZExtValue() == 16;
  return false;
}
 
static bool isSHL16(const SDValue &Op) {
  if (Op.getOpcode() != ISD::SHL)
    return false;
  if (auto Const = dyn_cast<ConstantSDNode>(Op.getOperand(1)))
    return Const->getZExtValue() == 16;
  return false;
}
 
// Check for a signed 16-bit value. We special case SRA because it makes it
// more simple when also looking for SRAs that aren't sign extending a
// smaller value. Without the check, we'd need to take extra care with
// checking order for some operations.
static bool isS16(const SDValue &Op, SelectionDAG &DAG) {
  if (isSRA16(Op))
    return isSHL16(Op.getOperand(0));
  return DAG.ComputeNumSignBits(Op) == 17;
}
 
/// IntCCToARMCC - Convert a DAG integer condition code to an ARM CC
static ARMCC::CondCodes IntCCToARMCC(ISD::CondCode CC) {
  switch (CC) {
  default: llvm_unreachable("Unknown condition code!");
  case ISD::SETNE:  return ARMCC::NE;
  case ISD::SETEQ:  return ARMCC::EQ;
  case ISD::SETGT:  return ARMCC::GT;
  case ISD::SETGE:  return ARMCC::GE;
  case ISD::SETLT:  return ARMCC::LT;
  case ISD::SETLE:  return ARMCC::LE;
  case ISD::SETUGT: return ARMCC::HI;
  case ISD::SETUGE: return ARMCC::HS;
  case ISD::SETULT: return ARMCC::LO;
  case ISD::SETULE: return ARMCC::LS;
  }
}
 
/// FPCCToARMCC - Convert a DAG fp condition code to an ARM CC.
static void FPCCToARMCC(ISD::CondCode CC, ARMCC::CondCodes &CondCode,
                        ARMCC::CondCodes &CondCode2) {
  CondCode2 = ARMCC::AL;
  switch (CC) {
  default: llvm_unreachable("Unknown FP condition!");
  case ISD::SETEQ:
  case ISD::SETOEQ: CondCode = ARMCC::EQ; break;
  case ISD::SETGT:
  case ISD::SETOGT: CondCode = ARMCC::GT; break;
  case ISD::SETGE:
  case ISD::SETOGE: CondCode = ARMCC::GE; break;
  case ISD::SETOLT: CondCode = ARMCC::MI; break;
  case ISD::SETOLE: CondCode = ARMCC::LS; break;
  case ISD::SETONE: CondCode = ARMCC::MI; CondCode2 = ARMCC::GT; break;
  case ISD::SETO:   CondCode = ARMCC::VC; break;
  case ISD::SETUO:  CondCode = ARMCC::VS; break;
  case ISD::SETUEQ: CondCode = ARMCC::EQ; CondCode2 = ARMCC::VS; break;
  case ISD::SETUGT: CondCode = ARMCC::HI; break;
  case ISD::SETUGE: CondCode = ARMCC::PL; break;
  case ISD::SETLT:
  case ISD::SETULT: CondCode = ARMCC::LT; break;
  case ISD::SETLE:
  case ISD::SETULE: CondCode = ARMCC::LE; break;
  case ISD::SETNE:
  case ISD::SETUNE: CondCode = ARMCC::NE; break;
  }
}
 
//===----------------------------------------------------------------------===//
//                      Calling Convention Implementation
//===----------------------------------------------------------------------===//
 
/// getEffectiveCallingConv - Get the effective calling convention, taking into
/// account presence of floating point hardware and calling convention
/// limitations, such as support for variadic functions.
CallingConv::ID
ARMTargetLowering::getEffectiveCallingConv(CallingConv::ID CC,
                                           bool isVarArg) const {
  switch (CC) {
  default:
    report_fatal_error("Unsupported calling convention");
  case CallingConv::ARM_AAPCS:
  case CallingConv::ARM_APCS:
  case CallingConv::GHC:
  case CallingConv::CFGuard_Check:
    return CC;
  case CallingConv::PreserveMost:
    return CallingConv::PreserveMost;
  case CallingConv::ARM_AAPCS_VFP:
  case CallingConv::Swift:
  case CallingConv::SwiftTail:
    return isVarArg ? CallingConv::ARM_AAPCS : CallingConv::ARM_AAPCS_VFP;
  case CallingConv::C:
  case CallingConv::Tail:
    if (!Subtarget->isAAPCS_ABI())
      return CallingConv::ARM_APCS;
    else if (Subtarget->hasVFP2Base() && !Subtarget->isThumb1Only() &&
             getTargetMachine().Options.FloatABIType == FloatABI::Hard &&
             !isVarArg)
      return CallingConv::ARM_AAPCS_VFP;
    else
      return CallingConv::ARM_AAPCS;
  case CallingConv::Fast:
  case CallingConv::CXX_FAST_TLS:
    if (!Subtarget->isAAPCS_ABI()) {
      if (Subtarget->hasVFP2Base() && !Subtarget->isThumb1Only() && !isVarArg)
        return CallingConv::Fast;
      return CallingConv::ARM_APCS;
    } else if (Subtarget->hasVFP2Base() &&
               !Subtarget->isThumb1Only() && !isVarArg)
      return CallingConv::ARM_AAPCS_VFP;
    else
      return CallingConv::ARM_AAPCS;
  }
}
 
CCAssignFn *ARMTargetLowering::CCAssignFnForCall(CallingConv::ID CC,
                                                 bool isVarArg) const {
  return CCAssignFnForNode(CC, false, isVarArg);
}
 
CCAssignFn *ARMTargetLowering::CCAssignFnForReturn(CallingConv::ID CC,
                                                   bool isVarArg) const {
  return CCAssignFnForNode(CC, true, isVarArg);
}
 
/// CCAssignFnForNode - Selects the correct CCAssignFn for the given
/// CallingConvention.
CCAssignFn *ARMTargetLowering::CCAssignFnForNode(CallingConv::ID CC,
                                                 bool Return,
                                                 bool isVarArg) const {
  switch (getEffectiveCallingConv(CC, isVarArg)) {
  default:
    report_fatal_error("Unsupported calling convention");
  case CallingConv::ARM_APCS:
    return (Return ? RetCC_ARM_APCS : CC_ARM_APCS);
  case CallingConv::ARM_AAPCS:
    return (Return ? RetCC_ARM_AAPCS : CC_ARM_AAPCS);
  case CallingConv::ARM_AAPCS_VFP:
    return (Return ? RetCC_ARM_AAPCS_VFP : CC_ARM_AAPCS_VFP);
  case CallingConv::Fast:
    return (Return ? RetFastCC_ARM_APCS : FastCC_ARM_APCS);
  case CallingConv::GHC:
    return (Return ? RetCC_ARM_APCS : CC_ARM_APCS_GHC);
  case CallingConv::PreserveMost:
    return (Return ? RetCC_ARM_AAPCS : CC_ARM_AAPCS);
  case CallingConv::CFGuard_Check:
    return (Return ? RetCC_ARM_AAPCS : CC_ARM_Win32_CFGuard_Check);
  }
}
 
SDValue ARMTargetLowering::MoveToHPR(const SDLoc &dl, SelectionDAG &DAG,
                                     MVT LocVT, MVT ValVT, SDValue Val) const {
  Val = DAG.getNode(ISD::BITCAST, dl, MVT::getIntegerVT(LocVT.getSizeInBits()),
                    Val);
  if (Subtarget->hasFullFP16()) {
    Val = DAG.getNode(ARMISD::VMOVhr, dl, ValVT, Val);
  } else {
    Val = DAG.getNode(ISD::TRUNCATE, dl,
                      MVT::getIntegerVT(ValVT.getSizeInBits()), Val);
    Val = DAG.getNode(ISD::BITCAST, dl, ValVT, Val);
  }
  return Val;
}
 
SDValue ARMTargetLowering::MoveFromHPR(const SDLoc &dl, SelectionDAG &DAG,
                                       MVT LocVT, MVT ValVT,
                                       SDValue Val) const {
  if (Subtarget->hasFullFP16()) {
    Val = DAG.getNode(ARMISD::VMOVrh, dl,
                      MVT::getIntegerVT(LocVT.getSizeInBits()), Val);
  } else {
    Val = DAG.getNode(ISD::BITCAST, dl,
                      MVT::getIntegerVT(ValVT.getSizeInBits()), Val);
    Val = DAG.getNode(ISD::ZERO_EXTEND, dl,
                      MVT::getIntegerVT(LocVT.getSizeInBits()), Val);
  }
  return DAG.getNode(ISD::BITCAST, dl, LocVT, Val);
}
 
/// LowerCallResult - Lower the result values of a call into the
/// appropriate copies out of appropriate physical registers.
SDValue ARMTargetLowering::LowerCallResult(
    SDValue Chain, SDValue InFlag, CallingConv::ID CallConv, bool isVarArg,
    const SmallVectorImpl<ISD::InputArg> &Ins, const SDLoc &dl,
    SelectionDAG &DAG, SmallVectorImpl<SDValue> &InVals, bool isThisReturn,
    SDValue ThisVal) const {
  // Assign locations to each value returned by this call.
  SmallVector<CCValAssign, 16> RVLocs;
  CCState CCInfo(CallConv, isVarArg, DAG.getMachineFunction(), RVLocs,
                 *DAG.getContext());
  CCInfo.AnalyzeCallResult(Ins, CCAssignFnForReturn(CallConv, isVarArg));
 
  // Copy all of the result registers out of their specified physreg.
  for (unsigned i = 0; i != RVLocs.size(); ++i) {
    CCValAssign VA = RVLocs[i];
 
    // Pass 'this' value directly from the argument to return value, to avoid
    // reg unit interference
    if (i == 0 && isThisReturn) {
      assert(!VA.needsCustom() && VA.getLocVT() == MVT::i32 &&
             "unexpected return calling convention register assignment");
      InVals.push_back(ThisVal);
      continue;
    }
 
    SDValue Val;
    if (VA.needsCustom() &&
        (VA.getLocVT() == MVT::f64 || VA.getLocVT() == MVT::v2f64)) {
      // Handle f64 or half of a v2f64.
      SDValue Lo = DAG.getCopyFromReg(Chain, dl, VA.getLocReg(), MVT::i32,
                                      InFlag);
      Chain = Lo.getValue(1);
      InFlag = Lo.getValue(2);
      VA = RVLocs[++i]; // skip ahead to next loc
      SDValue Hi = DAG.getCopyFromReg(Chain, dl, VA.getLocReg(), MVT::i32,
                                      InFlag);
      Chain = Hi.getValue(1);
      InFlag = Hi.getValue(2);
      if (!Subtarget->isLittle())
        std::swap (Lo, Hi);
      Val = DAG.getNode(ARMISD::VMOVDRR, dl, MVT::f64, Lo, Hi);
 
      if (VA.getLocVT() == MVT::v2f64) {
        SDValue Vec = DAG.getNode(ISD::UNDEF, dl, MVT::v2f64);
        Vec = DAG.getNode(ISD::INSERT_VECTOR_ELT, dl, MVT::v2f64, Vec, Val,
                          DAG.getConstant(0, dl, MVT::i32));
 
        VA = RVLocs[++i]; // skip ahead to next loc
        Lo = DAG.getCopyFromReg(Chain, dl, VA.getLocReg(), MVT::i32, InFlag);
        Chain = Lo.getValue(1);
        InFlag = Lo.getValue(2);
        VA = RVLocs[++i]; // skip ahead to next loc
        Hi = DAG.getCopyFromReg(Chain, dl, VA.getLocReg(), MVT::i32, InFlag);
        Chain = Hi.getValue(1);
        InFlag = Hi.getValue(2);
        if (!Subtarget->isLittle())
          std::swap (Lo, Hi);
        Val = DAG.getNode(ARMISD::VMOVDRR, dl, MVT::f64, Lo, Hi);
        Val = DAG.getNode(ISD::INSERT_VECTOR_ELT, dl, MVT::v2f64, Vec, Val,
                          DAG.getConstant(1, dl, MVT::i32));
      }
    } else {
      Val = DAG.getCopyFromReg(Chain, dl, VA.getLocReg(), VA.getLocVT(),
                               InFlag);
      Chain = Val.getValue(1);
      InFlag = Val.getValue(2);
    }
 
    switch (VA.getLocInfo()) {
    default: llvm_unreachable("Unknown loc info!");
    case CCValAssign::Full: break;
    case CCValAssign::BCvt:
      Val = DAG.getNode(ISD::BITCAST, dl, VA.getValVT(), Val);
      break;
    }
 
    // f16 arguments have their size extended to 4 bytes and passed as if they
    // had been copied to the LSBs of a 32-bit register.
    // For that, it's passed extended to i32 (soft ABI) or to f32 (hard ABI)
    if (VA.needsCustom() &&
        (VA.getValVT() == MVT::f16 || VA.getValVT() == MVT::bf16))
      Val = MoveToHPR(dl, DAG, VA.getLocVT(), VA.getValVT(), Val);
 
    InVals.push_back(Val);
  }
 
  return Chain;
}
 
std::pair<SDValue, MachinePointerInfo> ARMTargetLowering::computeAddrForCallArg(
    const SDLoc &dl, SelectionDAG &DAG, const CCValAssign &VA, SDValue StackPtr,
    bool IsTailCall, int SPDiff) const {
  SDValue DstAddr;
  MachinePointerInfo DstInfo;
  int32_t Offset = VA.getLocMemOffset();
  MachineFunction &MF = DAG.getMachineFunction();
 
  if (IsTailCall) {
        Offset += SPDiff;
        auto PtrVT = getPointerTy(DAG.getDataLayout());
        int Size = VA.getLocVT().getFixedSizeInBits() / 8;
        int FI = MF.getFrameInfo().CreateFixedObject(Size, Offset, true);
        DstAddr = DAG.getFrameIndex(FI, PtrVT);
        DstInfo =
            MachinePointerInfo::getFixedStack(DAG.getMachineFunction(), FI);
  } else {
        SDValue PtrOff = DAG.getIntPtrConstant(Offset, dl);
        DstAddr = DAG.getNode(ISD::ADD, dl, getPointerTy(DAG.getDataLayout()),
                              StackPtr, PtrOff);
        DstInfo =
            MachinePointerInfo::getStack(DAG.getMachineFunction(), Offset);
  }
 
  return std::make_pair(DstAddr, DstInfo);
}
 
void ARMTargetLowering::PassF64ArgInRegs(const SDLoc &dl, SelectionDAG &DAG,
                                         SDValue Chain, SDValue &Arg,
                                         RegsToPassVector &RegsToPass,
                                         CCValAssign &VA, CCValAssign &NextVA,
                                         SDValue &StackPtr,
                                         SmallVectorImpl<SDValue> &MemOpChains,
                                         bool IsTailCall,
                                         int SPDiff) const {
  SDValue fmrrd = DAG.getNode(ARMISD::VMOVRRD, dl,
                              DAG.getVTList(MVT::i32, MVT::i32), Arg);
  unsigned id = Subtarget->isLittle() ? 0 : 1;
  RegsToPass.push_back(std::make_pair(VA.getLocReg(), fmrrd.getValue(id)));
 
  if (NextVA.isRegLoc())
    RegsToPass.push_back(std::make_pair(NextVA.getLocReg(), fmrrd.getValue(1-id)));
  else {
    assert(NextVA.isMemLoc());
    if (!StackPtr.getNode())
      StackPtr = DAG.getCopyFromReg(Chain, dl, ARM::SP,
                                    getPointerTy(DAG.getDataLayout()));
 
    SDValue DstAddr;
    MachinePointerInfo DstInfo;
    std::tie(DstAddr, DstInfo) =
        computeAddrForCallArg(dl, DAG, NextVA, StackPtr, IsTailCall, SPDiff);
    MemOpChains.push_back(
        DAG.getStore(Chain, dl, fmrrd.getValue(1 - id), DstAddr, DstInfo));
  }
}
 
static bool canGuaranteeTCO(CallingConv::ID CC, bool GuaranteeTailCalls) {
  return (CC == CallingConv::Fast && GuaranteeTailCalls) ||
         CC == CallingConv::Tail || CC == CallingConv::SwiftTail;
}
 
/// LowerCall - Lowering a call into a callseq_start <-
/// ARMISD:CALL <- callseq_end chain. Also add input and output parameter
/// nodes.
SDValue
ARMTargetLowering::LowerCall(TargetLowering::CallLoweringInfo &CLI,
                             SmallVectorImpl<SDValue> &InVals) const {
  SelectionDAG &DAG                     = CLI.DAG;
  SDLoc &dl                             = CLI.DL;
  SmallVectorImpl<ISD::OutputArg> &Outs = CLI.Outs;
  SmallVectorImpl<SDValue> &OutVals     = CLI.OutVals;
  SmallVectorImpl<ISD::InputArg> &Ins   = CLI.Ins;
  SDValue Chain                         = CLI.Chain;
  SDValue Callee                        = CLI.Callee;
  bool &isTailCall                      = CLI.IsTailCall;
  CallingConv::ID CallConv              = CLI.CallConv;
  bool doesNotRet                       = CLI.DoesNotReturn;
  bool isVarArg                         = CLI.IsVarArg;
 
  MachineFunction &MF = DAG.getMachineFunction();
  ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
  MachineFunction::CallSiteInfo CSInfo;
  bool isStructRet = (Outs.empty()) ? false : Outs[0].Flags.isSRet();
  bool isThisReturn = false;
  bool isCmseNSCall   = false;
  bool isSibCall = false;
  bool PreferIndirect = false;
  bool GuardWithBTI = false;
 
  // Lower 'returns_twice' calls to a pseudo-instruction.
  if (CLI.CB && CLI.CB->getAttributes().hasFnAttr(Attribute::ReturnsTwice) &&
      !Subtarget->noBTIAtReturnTwice())
    GuardWithBTI = AFI->branchTargetEnforcement();
 
  // Determine whether this is a non-secure function call.
  if (CLI.CB && CLI.CB->getAttributes().hasFnAttr("cmse_nonsecure_call"))
    isCmseNSCall = true;
 
  // Disable tail calls if they're not supported.
  if (!Subtarget->supportsTailCall())
    isTailCall = false;
 
  // For both the non-secure calls and the returns from a CMSE entry function,
  // the function needs to do some extra work afte r the call, or before the
  // return, respectively, thus it cannot end with atail call
  if (isCmseNSCall || AFI->isCmseNSEntryFunction())
    isTailCall = false;
 
  if (isa<GlobalAddressSDNode>(Callee)) {
    // If we're optimizing for minimum size and the function is called three or
    // more times in this block, we can improve codesize by calling indirectly
    // as BLXr has a 16-bit encoding.
    auto *GV = cast<GlobalAddressSDNode>(Callee)->getGlobal();
    if (CLI.CB) {
      auto *BB = CLI.CB->getParent();
      PreferIndirect = Subtarget->isThumb() && Subtarget->hasMinSize() &&
                       count_if(GV->users(), [&BB](const User *U) {
                         return isa<Instruction>(U) &&
                                cast<Instruction>(U)->getParent() == BB;
                       }) > 2;
    }
  }
  if (isTailCall) {
    // Check if it's really possible to do a tail call.
    isTailCall = IsEligibleForTailCallOptimization(
        Callee, CallConv, isVarArg, isStructRet,
        MF.getFunction().hasStructRetAttr(), Outs, OutVals, Ins, DAG,
        PreferIndirect);
 
    if (isTailCall && !getTargetMachine().Options.GuaranteedTailCallOpt &&
        CallConv != CallingConv::Tail && CallConv != CallingConv::SwiftTail)
      isSibCall = true;
 
    // We don't support GuaranteedTailCallOpt for ARM, only automatically
    // detected sibcalls.
    if (isTailCall)
      ++NumTailCalls;
  }
 
  if (!isTailCall && CLI.CB && CLI.CB->isMustTailCall())
    report_fatal_error("failed to perform tail call elimination on a call "
                       "site marked musttail");
  // Analyze operands of the call, assigning locations to each operand.
  SmallVector<CCValAssign, 16> ArgLocs;
  CCState CCInfo(CallConv, isVarArg, DAG.getMachineFunction(), ArgLocs,
                 *DAG.getContext());
  CCInfo.AnalyzeCallOperands(Outs, CCAssignFnForCall(CallConv, isVarArg));
 
  // Get a count of how many bytes are to be pushed on the stack.
  unsigned NumBytes = CCInfo.getNextStackOffset();
 
  // SPDiff is the byte offset of the call's argument area from the callee's.
  // Stores to callee stack arguments will be placed in FixedStackSlots offset
  // by this amount for a tail call. In a sibling call it must be 0 because the
  // caller will deallocate the entire stack and the callee still expects its
  // arguments to begin at SP+0. Completely unused for non-tail calls.
  int SPDiff = 0;
 
  if (isTailCall && !isSibCall) {
    auto FuncInfo = MF.getInfo<ARMFunctionInfo>();
    unsigned NumReusableBytes = FuncInfo->getArgumentStackSize();
 
    // Since callee will pop argument stack as a tail call, we must keep the
    // popped size 16-byte aligned.
    Align StackAlign = DAG.getDataLayout().getStackAlignment();
    NumBytes = alignTo(NumBytes, StackAlign);
 
    // SPDiff will be negative if this tail call requires more space than we
    // would automatically have in our incoming argument space. Positive if we
    // can actually shrink the stack.
    SPDiff = NumReusableBytes - NumBytes;
 
    // If this call requires more stack than we have available from
    // LowerFormalArguments, tell FrameLowering to reserve space for it.
    if (SPDiff < 0 && AFI->getArgRegsSaveSize() < (unsigned)-SPDiff)
      AFI->setArgRegsSaveSize(-SPDiff);
  }
 
  if (isSibCall) {
    // For sibling tail calls, memory operands are available in our caller's stack.
    NumBytes = 0;
  } else {
    // Adjust the stack pointer for the new arguments...
    // These operations are automatically eliminated by the prolog/epilog pass
    Chain = DAG.getCALLSEQ_START(Chain, isTailCall ? 0 : NumBytes, 0, dl);
  }
 
  SDValue StackPtr =
      DAG.getCopyFromReg(Chain, dl, ARM::SP, getPointerTy(DAG.getDataLayout()));
 
  RegsToPassVector RegsToPass;
  SmallVector<SDValue, 8> MemOpChains;
 
  // During a tail call, stores to the argument area must happen after all of
  // the function's incoming arguments have been loaded because they may alias.
  // This is done by folding in a TokenFactor from LowerFormalArguments, but
  // there's no point in doing so repeatedly so this tracks whether that's
  // happened yet.
  bool AfterFormalArgLoads = false;
 
  // Walk the register/memloc assignments, inserting copies/loads.  In the case
  // of tail call optimization, arguments are handled later.
  for (unsigned i = 0, realArgIdx = 0, e = ArgLocs.size();
       i != e;
       ++i, ++realArgIdx) {
    CCValAssign &VA = ArgLocs[i];
    SDValue Arg = OutVals[realArgIdx];
    ISD::ArgFlagsTy Flags = Outs[realArgIdx].Flags;
    bool isByVal = Flags.isByVal();
 
    // Promote the value if needed.
    switch (VA.getLocInfo()) {
    default: llvm_unreachable("Unknown loc info!");
    case CCValAssign::Full: break;
    case CCValAssign::SExt:
      Arg = DAG.getNode(ISD::SIGN_EXTEND, dl, VA.getLocVT(), Arg);
      break;
    case CCValAssign::ZExt:
      Arg = DAG.getNode(ISD::ZERO_EXTEND, dl, VA.getLocVT(), Arg);
      break;
    case CCValAssign::AExt:
      Arg = DAG.getNode(ISD::ANY_EXTEND, dl, VA.getLocVT(), Arg);
      break;
    case CCValAssign::BCvt:
      Arg = DAG.getNode(ISD::BITCAST, dl, VA.getLocVT(), Arg);
      break;
    }
 
    if (isTailCall && VA.isMemLoc() && !AfterFormalArgLoads) {
      Chain = DAG.getStackArgumentTokenFactor(Chain);
      AfterFormalArgLoads = true;
    }
 
    // f16 arguments have their size extended to 4 bytes and passed as if they
    // had been copied to the LSBs of a 32-bit register.
    // For that, it's passed extended to i32 (soft ABI) or to f32 (hard ABI)
    if (VA.needsCustom() &&
        (VA.getValVT() == MVT::f16 || VA.getValVT() == MVT::bf16)) {
      Arg = MoveFromHPR(dl, DAG, VA.getLocVT(), VA.getValVT(), Arg);
    } else {
      // f16 arguments could have been extended prior to argument lowering.
      // Mask them arguments if this is a CMSE nonsecure call.
      auto ArgVT = Outs[realArgIdx].ArgVT;
      if (isCmseNSCall && (ArgVT == MVT::f16)) {
        auto LocBits = VA.getLocVT().getSizeInBits();
        auto MaskValue = APInt::getLowBitsSet(LocBits, ArgVT.getSizeInBits());
        SDValue Mask =
            DAG.getConstant(MaskValue, dl, MVT::getIntegerVT(LocBits));
        Arg = DAG.getNode(ISD::BITCAST, dl, MVT::getIntegerVT(LocBits), Arg);
        Arg = DAG.getNode(ISD::AND, dl, MVT::getIntegerVT(LocBits), Arg, Mask);
        Arg = DAG.getNode(ISD::BITCAST, dl, VA.getLocVT(), Arg);
      }
    }
 
    // f64 and v2f64 might be passed in i32 pairs and must be split into pieces
    if (VA.needsCustom() && VA.getLocVT() == MVT::v2f64) {
      SDValue Op0 = DAG.getNode(ISD::EXTRACT_VECTOR_ELT, dl, MVT::f64, Arg,
                                DAG.getConstant(0, dl, MVT::i32));
      SDValue Op1 = DAG.getNode(ISD::EXTRACT_VECTOR_ELT, dl, MVT::f64, Arg,
                                DAG.getConstant(1, dl, MVT::i32));
 
      PassF64ArgInRegs(dl, DAG, Chain, Op0, RegsToPass, VA, ArgLocs[++i],
                       StackPtr, MemOpChains, isTailCall, SPDiff);
 
      VA = ArgLocs[++i]; // skip ahead to next loc
      if (VA.isRegLoc()) {
        PassF64ArgInRegs(dl, DAG, Chain, Op1, RegsToPass, VA, ArgLocs[++i],
                         StackPtr, MemOpChains, isTailCall, SPDiff);
      } else {
        assert(VA.isMemLoc());
        SDValue DstAddr;
        MachinePointerInfo DstInfo;
        std::tie(DstAddr, DstInfo) =
            computeAddrForCallArg(dl, DAG, VA, StackPtr, isTailCall, SPDiff);
        MemOpChains.push_back(DAG.getStore(Chain, dl, Op1, DstAddr, DstInfo));
      }
    } else if (VA.needsCustom() && VA.getLocVT() == MVT::f64) {
      PassF64ArgInRegs(dl, DAG, Chain, Arg, RegsToPass, VA, ArgLocs[++i],
                       StackPtr, MemOpChains, isTailCall, SPDiff);
    } else if (VA.isRegLoc()) {
      if (realArgIdx == 0 && Flags.isReturned() && !Flags.isSwiftSelf() &&
          Outs[0].VT == MVT::i32) {
        assert(VA.getLocVT() == MVT::i32 &&
               "unexpected calling convention register assignment");
        assert(!Ins.empty() && Ins[0].VT == MVT::i32 &&
               "unexpected use of 'returned'");
        isThisReturn = true;
      }
      const TargetOptions &Options = DAG.getTarget().Options;
      if (Options.EmitCallSiteInfo)
        CSInfo.emplace_back(VA.getLocReg(), i);
      RegsToPass.push_back(std::make_pair(VA.getLocReg(), Arg));
    } else if (isByVal) {
      assert(VA.isMemLoc());
      unsigned offset = 0;
 
      // True if this byval aggregate will be split between registers
      // and memory.
      unsigned ByValArgsCount = CCInfo.getInRegsParamsCount();
      unsigned CurByValIdx = CCInfo.getInRegsParamsProcessed();
 
      if (CurByValIdx < ByValArgsCount) {
 
        unsigned RegBegin, RegEnd;
        CCInfo.getInRegsParamInfo(CurByValIdx, RegBegin, RegEnd);
 
        EVT PtrVT =
            DAG.getTargetLoweringInfo().getPointerTy(DAG.getDataLayout());
        unsigned int i, j;
        for (i = 0, j = RegBegin; j < RegEnd; i++, j++) {
          SDValue Const = DAG.getConstant(4*i, dl, MVT::i32);
          SDValue AddArg = DAG.getNode(ISD::ADD, dl, PtrVT, Arg, Const);
          SDValue Load =
              DAG.getLoad(PtrVT, dl, Chain, AddArg, MachinePointerInfo(),
                          DAG.InferPtrAlign(AddArg));
          MemOpChains.push_back(Load.getValue(1));
          RegsToPass.push_back(std::make_pair(j, Load));
        }
 
        // If parameter size outsides register area, "offset" value
        // helps us to calculate stack slot for remained part properly.
        offset = RegEnd - RegBegin;
 
        CCInfo.nextInRegsParam();
      }
 
      if (Flags.getByValSize() > 4*offset) {
        auto PtrVT = getPointerTy(DAG.getDataLayout());
        SDValue Dst;
        MachinePointerInfo DstInfo;
        std::tie(Dst, DstInfo) =
            computeAddrForCallArg(dl, DAG, VA, StackPtr, isTailCall, SPDiff);
        SDValue SrcOffset = DAG.getIntPtrConstant(4*offset, dl);
        SDValue Src = DAG.getNode(ISD::ADD, dl, PtrVT, Arg, SrcOffset);
        SDValue SizeNode = DAG.getConstant(Flags.getByValSize() - 4*offset, dl,
                                           MVT::i32);
        SDValue AlignNode =
            DAG.getConstant(Flags.getNonZeroByValAlign().value(), dl, MVT::i32);
 
        SDVTList VTs = DAG.getVTList(MVT::Other, MVT::Glue);
        SDValue Ops[] = { Chain, Dst, Src, SizeNode, AlignNode};
        MemOpChains.push_back(DAG.getNode(ARMISD::COPY_STRUCT_BYVAL, dl, VTs,
                                          Ops));
      }
    } else {
      assert(VA.isMemLoc());
      SDValue DstAddr;
      MachinePointerInfo DstInfo;
      std::tie(DstAddr, DstInfo) =
          computeAddrForCallArg(dl, DAG, VA, StackPtr, isTailCall, SPDiff);
 
      SDValue Store = DAG.getStore(Chain, dl, Arg, DstAddr, DstInfo);
      MemOpChains.push_back(Store);
    }
  }
 
  if (!MemOpChains.empty())
    Chain = DAG.getNode(ISD::TokenFactor, dl, MVT::Other, MemOpChains);
 
  // Build a sequence of copy-to-reg nodes chained together with token chain
  // and flag operands which copy the outgoing args into the appropriate regs.
  SDValue InFlag;
  for (unsigned i = 0, e = RegsToPass.size(); i != e; ++i) {
    Chain = DAG.getCopyToReg(Chain, dl, RegsToPass[i].first,
                             RegsToPass[i].second, InFlag);
    InFlag = Chain.getValue(1);
  }
 
  // If the callee is a GlobalAddress/ExternalSymbol node (quite common, every
  // direct call is) turn it into a TargetGlobalAddress/TargetExternalSymbol
  // node so that legalize doesn't hack it.
  bool isDirect = false;
 
  const TargetMachine &TM = getTargetMachine();
  const Module *Mod = MF.getFunction().getParent();
  const GlobalValue *GVal = nullptr;
  if (GlobalAddressSDNode *G = dyn_cast<GlobalAddressSDNode>(Callee))
    GVal = G->getGlobal();
  bool isStub =
      !TM.shouldAssumeDSOLocal(*Mod, GVal) && Subtarget->isTargetMachO();
 
  bool isARMFunc = !Subtarget->isThumb() || (isStub && !Subtarget->isMClass());
  bool isLocalARMFunc = false;
  auto PtrVt = getPointerTy(DAG.getDataLayout());
 
  if (Subtarget->genLongCalls()) {
    assert((!isPositionIndependent() || Subtarget->isTargetWindows()) &&
           "long-calls codegen is not position independent!");
    // Handle a global address or an external symbol. If it's not one of
    // those, the target's already in a register, so we don't need to do
    // anything extra.
    if (isa<GlobalAddressSDNode>(Callee)) {
      // When generating execute-only code we use movw movt pair.
      // Currently execute-only is only available for architectures that
      // support movw movt, so we are safe to assume that.
      if (Subtarget->genExecuteOnly()) {
        assert(Subtarget->useMovt() &&
               "long-calls with execute-only requires movt and movw!");
        ++NumMovwMovt;
        Callee = DAG.getNode(ARMISD::Wrapper, dl, PtrVt,
                             DAG.getTargetGlobalAddress(GVal, dl, PtrVt));
      } else {
        // Create a constant pool entry for the callee address
        unsigned ARMPCLabelIndex = AFI->createPICLabelUId();
        ARMConstantPoolValue *CPV = ARMConstantPoolConstant::Create(
            GVal, ARMPCLabelIndex, ARMCP::CPValue, 0);
 
        // Get the address of the callee into a register
        SDValue Addr = DAG.getTargetConstantPool(CPV, PtrVt, Align(4));
        Addr = DAG.getNode(ARMISD::Wrapper, dl, MVT::i32, Addr);
        Callee = DAG.getLoad(
            PtrVt, dl, DAG.getEntryNode(), Addr,
            MachinePointerInfo::getConstantPool(DAG.getMachineFunction()));
      }
    } else if (ExternalSymbolSDNode *S=dyn_cast<ExternalSymbolSDNode>(Callee)) {
      const char *Sym = S->getSymbol();
 
      // When generating execute-only code we use movw movt pair.
      // Currently execute-only is only available for architectures that
      // support movw movt, so we are safe to assume that.
      if (Subtarget->genExecuteOnly()) {
        assert(Subtarget->useMovt() &&
               "long-calls with execute-only requires movt and movw!");
        ++NumMovwMovt;
        Callee = DAG.getNode(ARMISD::Wrapper, dl, PtrVt,
                             DAG.getTargetGlobalAddress(GVal, dl, PtrVt));
      } else {
        // Create a constant pool entry for the callee address
        unsigned ARMPCLabelIndex = AFI->createPICLabelUId();
        ARMConstantPoolValue *CPV = ARMConstantPoolSymbol::Create(
            *DAG.getContext(), Sym, ARMPCLabelIndex, 0);
 
        // Get the address of the callee into a register
        SDValue Addr = DAG.getTargetConstantPool(CPV, PtrVt, Align(4));
        Addr = DAG.getNode(ARMISD::Wrapper, dl, MVT::i32, Addr);
        Callee = DAG.getLoad(
            PtrVt, dl, DAG.getEntryNode(), Addr,
            MachinePointerInfo::getConstantPool(DAG.getMachineFunction()));
      }
    }
  } else if (isa<GlobalAddressSDNode>(Callee)) {
    if (!PreferIndirect) {
      isDirect = true;
      bool isDef = GVal->isStrongDefinitionForLinker();
 
      // ARM call to a local ARM function is predicable.
      isLocalARMFunc = !Subtarget->isThumb() && (isDef || !ARMInterworking);
      // tBX takes a register source operand.
      if (isStub && Subtarget->isThumb1Only() && !Subtarget->hasV5TOps()) {
        assert(Subtarget->isTargetMachO() && "WrapperPIC use on non-MachO?");
        Callee = DAG.getNode(
            ARMISD::WrapperPIC, dl, PtrVt,
            DAG.getTargetGlobalAddress(GVal, dl, PtrVt, 0, ARMII::MO_NONLAZY));
        Callee = DAG.getLoad(
            PtrVt, dl, DAG.getEntryNode(), Callee,
            MachinePointerInfo::getGOT(DAG.getMachineFunction()), MaybeAlign(),
            MachineMemOperand::MODereferenceable |
                MachineMemOperand::MOInvariant);
      } else if (Subtarget->isTargetCOFF()) {
        assert(Subtarget->isTargetWindows() &&
               "Windows is the only supported COFF target");
        unsigned TargetFlags = ARMII::MO_NO_FLAG;
        if (GVal->hasDLLImportStorageClass())
          TargetFlags = ARMII::MO_DLLIMPORT;
        else if (!TM.shouldAssumeDSOLocal(*GVal->getParent(), GVal))
          TargetFlags = ARMII::MO_COFFSTUB;
        Callee = DAG.getTargetGlobalAddress(GVal, dl, PtrVt, /*offset=*/0,
                                            TargetFlags);
        if (TargetFlags & (ARMII::MO_DLLIMPORT | ARMII::MO_COFFSTUB))
          Callee =
              DAG.getLoad(PtrVt, dl, DAG.getEntryNode(),
                          DAG.getNode(ARMISD::Wrapper, dl, PtrVt, Callee),
                          MachinePointerInfo::getGOT(DAG.getMachineFunction()));
      } else {
        Callee = DAG.getTargetGlobalAddress(GVal, dl, PtrVt, 0, 0);
      }
    }
  } else if (ExternalSymbolSDNode *S = dyn_cast<ExternalSymbolSDNode>(Callee)) {
    isDirect = true;
    // tBX takes a register source operand.
    const char *Sym = S->getSymbol();
    if (isARMFunc && Subtarget->isThumb1Only() && !Subtarget->hasV5TOps()) {
      unsigned ARMPCLabelIndex = AFI->createPICLabelUId();
      ARMConstantPoolValue *CPV =
        ARMConstantPoolSymbol::Create(*DAG.getContext(), Sym,
                                      ARMPCLabelIndex, 4);
      SDValue CPAddr = DAG.getTargetConstantPool(CPV, PtrVt, Align(4));
      CPAddr = DAG.getNode(ARMISD::Wrapper, dl, MVT::i32, CPAddr);
      Callee = DAG.getLoad(
          PtrVt, dl, DAG.getEntryNode(), CPAddr,
          MachinePointerInfo::getConstantPool(DAG.getMachineFunction()));
      SDValue PICLabel = DAG.getConstant(ARMPCLabelIndex, dl, MVT::i32);
      Callee = DAG.getNode(ARMISD::PIC_ADD, dl, PtrVt, Callee, PICLabel);
    } else {
      Callee = DAG.getTargetExternalSymbol(Sym, PtrVt, 0);
    }
  }
 
  if (isCmseNSCall) {
    assert(!isARMFunc && !isDirect &&
           "Cannot handle call to ARM function or direct call");
    if (NumBytes > 0) {
      DiagnosticInfoUnsupported Diag(DAG.getMachineFunction().getFunction(),
                                     "call to non-secure function would "
                                     "require passing arguments on stack",
                                     dl.getDebugLoc());
      DAG.getContext()->diagnose(Diag);
    }
    if (isStructRet) {
      DiagnosticInfoUnsupported Diag(
          DAG.getMachineFunction().getFunction(),
          "call to non-secure function would return value through pointer",
          dl.getDebugLoc());
      DAG.getContext()->diagnose(Diag);
    }
  }
 
  // FIXME: handle tail calls differently.
  unsigned CallOpc;
  if (Subtarget->isThumb()) {
    if (GuardWithBTI)
      CallOpc = ARMISD::t2CALL_BTI;
    else if (isCmseNSCall)
      CallOpc = ARMISD::tSECALL;
    else if ((!isDirect || isARMFunc) && !Subtarget->hasV5TOps())
      CallOpc = ARMISD::CALL_NOLINK;
    else
      CallOpc = ARMISD::CALL;
  } else {
    if (!isDirect && !Subtarget->hasV5TOps())
      CallOpc = ARMISD::CALL_NOLINK;
    else if (doesNotRet && isDirect && Subtarget->hasRetAddrStack() &&
             // Emit regular call when code size is the priority
             !Subtarget->hasMinSize())
      // "mov lr, pc; b _foo" to avoid confusing the RSP
      CallOpc = ARMISD::CALL_NOLINK;
    else
      CallOpc = isLocalARMFunc ? ARMISD::CALL_PRED : ARMISD::CALL;
  }
 
  // We don't usually want to end the call-sequence here because we would tidy
  // the frame up *after* the call, however in the ABI-changing tail-call case
  // we've carefully laid out the parameters so that when sp is reset they'll be
  // in the correct location.
  if (isTailCall && !isSibCall) {
    Chain = DAG.getCALLSEQ_END(Chain, 0, 0, InFlag, dl);
    InFlag = Chain.getValue(1);
  }
 
  std::vector<SDValue> Ops;
  Ops.push_back(Chain);
  Ops.push_back(Callee);
 
  if (isTailCall) {
    Ops.push_back(DAG.getTargetConstant(SPDiff, dl, MVT::i32));
  }
 
  // Add argument registers to the end of the list so that they are known live
  // into the call.
  for (unsigned i = 0, e = RegsToPass.size(); i != e; ++i)
    Ops.push_back(DAG.getRegister(RegsToPass[i].first,
                                  RegsToPass[i].second.getValueType()));
 
  // Add a register mask operand representing the call-preserved registers.
  const uint32_t *Mask;
  const ARMBaseRegisterInfo *ARI = Subtarget->getRegisterInfo();
  if (isThisReturn) {
    // For 'this' returns, use the R0-preserving mask if applicable
    Mask = ARI->getThisReturnPreservedMask(MF, CallConv);
    if (!Mask) {
      // Set isThisReturn to false if the calling convention is not one that
      // allows 'returned' to be modeled in this way, so LowerCallResult does
      // not try to pass 'this' straight through
      isThisReturn = false;
      Mask = ARI->getCallPreservedMask(MF, CallConv);
    }
  } else
    Mask = ARI->getCallPreservedMask(MF, CallConv);
 
  assert(Mask && "Missing call preserved mask for calling convention");
  Ops.push_back(DAG.getRegisterMask(Mask));
 
  if (InFlag.getNode())
    Ops.push_back(InFlag);
 
  SDVTList NodeTys = DAG.getVTList(MVT::Other, MVT::Glue);
  if (isTailCall) {
    MF.getFrameInfo().setHasTailCall();
    SDValue Ret = DAG.getNode(ARMISD::TC_RETURN, dl, NodeTys, Ops);
    DAG.addCallSiteInfo(Ret.getNode(), std::move(CSInfo));
    return Ret;
  }
 
  // Returns a chain and a flag for retval copy to use.
  Chain = DAG.getNode(CallOpc, dl, NodeTys, Ops);
  DAG.addNoMergeSiteInfo(Chain.getNode(), CLI.NoMerge);
  InFlag = Chain.getValue(1);
  DAG.addCallSiteInfo(Chain.getNode(), std::move(CSInfo));
 
  // If we're guaranteeing tail-calls will be honoured, the callee must
  // pop its own argument stack on return. But this call is *not* a tail call so
  // we need to undo that after it returns to restore the status-quo.
  bool TailCallOpt = getTargetMachine().Options.GuaranteedTailCallOpt;
  uint64_t CalleePopBytes =
      canGuaranteeTCO(CallConv, TailCallOpt) ? alignTo(NumBytes, 16) : -1ULL;
 
  Chain = DAG.getCALLSEQ_END(Chain, NumBytes, CalleePopBytes, InFlag, dl);
  if (!Ins.empty())
    InFlag = Chain.getValue(1);
 
  // Handle result values, copying them out of physregs into vregs that we
  // return.
  return LowerCallResult(Chain, InFlag, CallConv, isVarArg, Ins, dl, DAG,
                         InVals, isThisReturn,
                         isThisReturn ? OutVals[0] : SDValue());
}
 
/// HandleByVal - Every parameter *after* a byval parameter is passed
/// on the stack.  Remember the next parameter register to allocate,
/// and then confiscate the rest of the parameter registers to insure
/// this.
void ARMTargetLowering::HandleByVal(CCState *State, unsigned &Size,
                                    Align Alignment) const {
  // Byval (as with any stack) slots are always at least 4 byte aligned.
  Alignment = std::max(Alignment, Align(4));
 
  unsigned Reg = State->AllocateReg(GPRArgRegs);
  if (!Reg)
    return;
 
  unsigned AlignInRegs = Alignment.value() / 4;
  unsigned Waste = (ARM::R4 - Reg) % AlignInRegs;
  for (unsigned i = 0; i < Waste; ++i)
    Reg = State->AllocateReg(GPRArgRegs);
 
  if (!Reg)
    return;
 
  unsigned Excess = 4 * (ARM::R4 - Reg);
 
  // Special case when NSAA != SP and parameter size greater than size of
  // all remained GPR regs. In that case we can't split parameter, we must
  // send it to stack. We also must set NCRN to R4, so waste all
  // remained registers.
  const unsigned NSAAOffset = State->getNextStackOffset();
  if (NSAAOffset != 0 && Size > Excess) {
    while (State->AllocateReg(GPRArgRegs))
      ;
    return;
  }
 
  // First register for byval parameter is the first register that wasn't
  // allocated before this method call, so it would be "reg".
  // If parameter is small enough to be saved in range [reg, r4), then
  // the end (first after last) register would be reg + param-size-in-regs,
  // else parameter would be splitted between registers and stack,
  // end register would be r4 in this case.
  unsigned ByValRegBegin = Reg;
  unsigned ByValRegEnd = std::min<unsigned>(Reg + Size / 4, ARM::R4);
  State->addInRegsParamInfo(ByValRegBegin, ByValRegEnd);
  // Note, first register is allocated in the beginning of function already,
  // allocate remained amount of registers we need.
  for (unsigned i = Reg + 1; i != ByValRegEnd; ++i)
    State->AllocateReg(GPRArgRegs);
  // A byval parameter that is split between registers and memory needs its
  // size truncated here.
  // In the case where the entire structure fits in registers, we set the
  // size in memory to zero.
  Size = std::max<int>(Size - Excess, 0);
}
 
/// MatchingStackOffset - Return true if the given stack call argument is
/// already available in the same position (relatively) of the caller's
/// incoming argument stack.
static
bool MatchingStackOffset(SDValue Arg, unsigned Offset, ISD::ArgFlagsTy Flags,
                         MachineFrameInfo &MFI, const MachineRegisterInfo *MRI,
                         const TargetInstrInfo *TII) {
  unsigned Bytes = Arg.getValueSizeInBits() / 8;
  int FI = std::numeric_limits<int>::max();
  if (Arg.getOpcode() == ISD::CopyFromReg) {
    Register VR = cast<RegisterSDNode>(Arg.getOperand(1))->getReg();
    if (!Register::isVirtualRegister(VR))
      return false;
    MachineInstr *Def = MRI->getVRegDef(VR);
    if (!Def)
      return false;
    if (!Flags.isByVal()) {
      if (!TII->isLoadFromStackSlot(*Def, FI))
        return false;
    } else {
      return false;
    }
  } else if (LoadSDNode *Ld = dyn_cast<LoadSDNode>(Arg)) {
    if (Flags.isByVal())
      // ByVal argument is passed in as a pointer but it's now being
      // dereferenced. e.g.
      // define @foo(%struct.X* %A) {
      //   tail call @bar(%struct.X* byval %A)
      // }
      return false;
    SDValue Ptr = Ld->getBasePtr();
    FrameIndexSDNode *FINode = dyn_cast<FrameIndexSDNode>(Ptr);
    if (!FINode)
      return false;
    FI = FINode->getIndex();
  } else
    return false;
 
  assert(FI != std::numeric_limits<int>::max());
  if (!MFI.isFixedObjectIndex(FI))
    return false;
  return Offset == MFI.getObjectOffset(FI) && Bytes == MFI.getObjectSize(FI);
}
 
/// IsEligibleForTailCallOptimization - Check whether the call is eligible
/// for tail call optimization. Targets which want to do tail call
/// optimization should implement this function.
bool ARMTargetLowering::IsEligibleForTailCallOptimization(
    SDValue Callee, CallingConv::ID CalleeCC, bool isVarArg,
    bool isCalleeStructRet, bool isCallerStructRet,
    const SmallVectorImpl<ISD::OutputArg> &Outs,
    const SmallVectorImpl<SDValue> &OutVals,
    const SmallVectorImpl<ISD::InputArg> &Ins, SelectionDAG &DAG,
    const bool isIndirect) const {
  MachineFunction &MF = DAG.getMachineFunction();
  const Function &CallerF = MF.getFunction();
  CallingConv::ID CallerCC = CallerF.getCallingConv();
 
  assert(Subtarget->supportsTailCall());
 
  // Indirect tail calls cannot be optimized for Thumb1 if the args
  // to the call take up r0-r3. The reason is that there are no legal registers
  // left to hold the pointer to the function to be called.
  // Similarly, if the function uses return address sign and authentication,
  // r12 is needed to hold the PAC and is not available to hold the callee
  // address.
  if (Outs.size() >= 4 &&
      (!isa<GlobalAddressSDNode>(Callee.getNode()) || isIndirect)) {
    if (Subtarget->isThumb1Only())
      return false;
    // Conservatively assume the function spills LR.
    if (MF.getInfo<ARMFunctionInfo>()->shouldSignReturnAddress(true))
      return false;
  }
 
  // Look for obvious safe cases to perform tail call optimization that do not
  // require ABI changes. This is what gcc calls sibcall.
 
  // Exception-handling functions need a special set of instructions to indicate
  // a return to the hardware. Tail-calling another function would probably
  // break this.
  if (CallerF.hasFnAttribute("interrupt"))
    return false;
 
  if (canGuaranteeTCO(CalleeCC, getTargetMachine().Options.GuaranteedTailCallOpt))
    return CalleeCC == CallerCC;
 
  // Also avoid sibcall optimization if either caller or callee uses struct
  // return semantics.
  if (isCalleeStructRet || isCallerStructRet)
    return false;
 
  // Externally-defined functions with weak linkage should not be
  // tail-called on ARM when the OS does not support dynamic
  // pre-emption of symbols, as the AAELF spec requires normal calls
  // to undefined weak functions to be replaced with a NOP or jump to the
  // next instruction. The behaviour of branch instructions in this
  // situation (as used for tail calls) is implementation-defined, so we
  // cannot rely on the linker replacing the tail call with a return.
  if (GlobalAddressSDNode *G = dyn_cast<GlobalAddressSDNode>(Callee)) {
    const GlobalValue *GV = G->getGlobal();
    const Triple &TT = getTargetMachine().getTargetTriple();
    if (GV->hasExternalWeakLinkage() &&
        (!TT.isOSWindows() || TT.isOSBinFormatELF() || TT.isOSBinFormatMachO()))
      return false;
  }
 
  // Check that the call results are passed in the same way.
  LLVMContext &C = *DAG.getContext();
  if (!CCState::resultsCompatible(
          getEffectiveCallingConv(CalleeCC, isVarArg),
          getEffectiveCallingConv(CallerCC, CallerF.isVarArg()), MF, C, Ins,
          CCAssignFnForReturn(CalleeCC, isVarArg),
          CCAssignFnForReturn(CallerCC, CallerF.isVarArg())))
    return false;
  // The callee has to preserve all registers the caller needs to preserve.
  const ARMBaseRegisterInfo *TRI = Subtarget->getRegisterInfo();
  const uint32_t *CallerPreserved = TRI->getCallPreservedMask(MF, CallerCC);
  if (CalleeCC != CallerCC) {
    const uint32_t *CalleePreserved = TRI->getCallPreservedMask(MF, CalleeCC);
    if (!TRI->regmaskSubsetEqual(CallerPreserved, CalleePreserved))
      return false;
  }
 
  // If Caller's vararg or byval argument has been split between registers and
  // stack, do not perform tail call, since part of the argument is in caller's
  // local frame.
  const ARMFunctionInfo *AFI_Caller = MF.getInfo<ARMFunctionInfo>();
  if (AFI_Caller->getArgRegsSaveSize())
    return false;
 
  // If the callee takes no arguments then go on to check the results of the
  // call.
  if (!Outs.empty()) {
    // Check if stack adjustment is needed. For now, do not do this if any
    // argument is passed on the stack.
    SmallVector<CCValAssign, 16> ArgLocs;
    CCState CCInfo(CalleeCC, isVarArg, MF, ArgLocs, C);
    CCInfo.AnalyzeCallOperands(Outs, CCAssignFnForCall(CalleeCC, isVarArg));
    if (CCInfo.getNextStackOffset()) {
      // Check if the arguments are already laid out in the right way as
      // the caller's fixed stack objects.
      MachineFrameInfo &MFI = MF.getFrameInfo();
      const MachineRegisterInfo *MRI = &MF.getRegInfo();
      const TargetInstrInfo *TII = Subtarget->getInstrInfo();
      for (unsigned i = 0, realArgIdx = 0, e = ArgLocs.size();
           i != e;
           ++i, ++realArgIdx) {
        CCValAssign &VA = ArgLocs[i];
        EVT RegVT = VA.getLocVT();
        SDValue Arg = OutVals[realArgIdx];
        ISD::ArgFlagsTy Flags = Outs[realArgIdx].Flags;
        if (VA.getLocInfo() == CCValAssign::Indirect)
          return false;
        if (VA.needsCustom() && (RegVT == MVT::f64 || RegVT == MVT::v2f64)) {
          // f64 and vector types are split into multiple registers or
          // register/stack-slot combinations.  The types will not match
          // the registers; give up on memory f64 refs until we figure
          // out what to do about this.
          if (!VA.isRegLoc())
            return false;
          if (!ArgLocs[++i].isRegLoc())
            return false;
          if (RegVT == MVT::v2f64) {
            if (!ArgLocs[++i].isRegLoc())
              return false;
            if (!ArgLocs[++i].isRegLoc())
              return false;
          }
        } else if (!VA.isRegLoc()) {
          if (!MatchingStackOffset(Arg, VA.getLocMemOffset(), Flags,
                                   MFI, MRI, TII))
            return false;
        }
      }
    }
 
    const MachineRegisterInfo &MRI = MF.getRegInfo();
    if (!parametersInCSRMatch(MRI, CallerPreserved, ArgLocs, OutVals))
      return false;
  }
 
  return true;
}
 
bool
ARMTargetLowering::CanLowerReturn(CallingConv::ID CallConv,
                                  MachineFunction &MF, bool isVarArg,
                                  const SmallVectorImpl<ISD::OutputArg> &Outs,
                                  LLVMContext &Context) const {
  SmallVector<CCValAssign, 16> RVLocs;
  CCState CCInfo(CallConv, isVarArg, MF, RVLocs, Context);
  return CCInfo.CheckReturn(Outs, CCAssignFnForReturn(CallConv, isVarArg));
}
 
static SDValue LowerInterruptReturn(SmallVectorImpl<SDValue> &RetOps,
                                    const SDLoc &DL, SelectionDAG &DAG) {
  const MachineFunction &MF = DAG.getMachineFunction();
  const Function &F = MF.getFunction();
 
  StringRef IntKind = F.getFnAttribute("interrupt").getValueAsString();
 
  // See ARM ARM v7 B1.8.3. On exception entry LR is set to a possibly offset
  // version of the "preferred return address". These offsets affect the return
  // instruction if this is a return from PL1 without hypervisor extensions.
  //    IRQ/FIQ: +4     "subs pc, lr, #4"
  //    SWI:     0      "subs pc, lr, #0"
  //    ABORT:   +4     "subs pc, lr, #4"
  //    UNDEF:   +4/+2  "subs pc, lr, #0"
  // UNDEF varies depending on where the exception came from ARM or Thumb
  // mode. Alongside GCC, we throw our hands up in disgust and pretend it's 0.
 
  int64_t LROffset;
  if (IntKind == "" || IntKind == "IRQ" || IntKind == "FIQ" ||
      IntKind == "ABORT")
    LROffset = 4;
  else if (IntKind == "SWI" || IntKind == "UNDEF")
    LROffset = 0;
  else
    report_fatal_error("Unsupported interrupt attribute. If present, value "
                       "must be one of: IRQ, FIQ, SWI, ABORT or UNDEF");
 
  RetOps.insert(RetOps.begin() + 1,
                DAG.getConstant(LROffset, DL, MVT::i32, false));
 
  return DAG.getNode(ARMISD::INTRET_FLAG, DL, MVT::Other, RetOps);
}
 
SDValue
ARMTargetLowering::LowerReturn(SDValue Chain, CallingConv::ID CallConv,
                               bool isVarArg,
                               const SmallVectorImpl<ISD::OutputArg> &Outs,
                               const SmallVectorImpl<SDValue> &OutVals,
                               const SDLoc &dl, SelectionDAG &DAG) const {
  // CCValAssign - represent the assignment of the return value to a location.
  SmallVector<CCValAssign, 16> RVLocs;
 
  // CCState - Info about the registers and stack slots.
  CCState CCInfo(CallConv, isVarArg, DAG.getMachineFunction(), RVLocs,
                 *DAG.getContext());
 
  // Analyze outgoing return values.
  CCInfo.AnalyzeReturn(Outs, CCAssignFnForReturn(CallConv, isVarArg));
 
  SDValue Flag;
  SmallVector<SDValue, 4> RetOps;
  RetOps.push_back(Chain); // Operand #0 = Chain (updated below)
  bool isLittleEndian = Subtarget->isLittle();
 
  MachineFunction &MF = DAG.getMachineFunction();
  ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
  AFI->setReturnRegsCount(RVLocs.size());
 
 // Report error if cmse entry function returns structure through first ptr arg.
  if (AFI->isCmseNSEntryFunction() && MF.getFunction().hasStructRetAttr()) {
    // Note: using an empty SDLoc(), as the first line of the function is a
    // better place to report than the last line.
    DiagnosticInfoUnsupported Diag(
        DAG.getMachineFunction().getFunction(),
        "secure entry function would return value through pointer",
        SDLoc().getDebugLoc());
    DAG.getContext()->diagnose(Diag);
  }
 
  // Copy the result values into the output registers.
  for (unsigned i = 0, realRVLocIdx = 0;
       i != RVLocs.size();
       ++i, ++realRVLocIdx) {
    CCValAssign &VA = RVLocs[i];
    assert(VA.isRegLoc() && "Can only return in registers!");
 
    SDValue Arg = OutVals[realRVLocIdx];
    bool ReturnF16 = false;
 
    if (Subtarget->hasFullFP16() && Subtarget->isTargetHardFloat()) {
      // Half-precision return values can be returned like this:
      //
      // t11 f16 = fadd ...
      // t12: i16 = bitcast t11
      //   t13: i32 = zero_extend t12
      // t14: f32 = bitcast t13  <~~~~~~~ Arg
      //
      // to avoid code generation for bitcasts, we simply set Arg to the node
      // that produces the f16 value, t11 in this case.
      //
      if (Arg.getValueType() == MVT::f32 && Arg.getOpcode() == ISD::BITCAST) {
        SDValue ZE = Arg.getOperand(0);
        if (ZE.getOpcode() == ISD::ZERO_EXTEND && ZE.getValueType() == MVT::i32) {
          SDValue BC = ZE.getOperand(0);
          if (BC.getOpcode() == ISD::BITCAST && BC.getValueType() == MVT::i16) {
            Arg = BC.getOperand(0);
            ReturnF16 = true;
          }
        }
      }
    }
 
    switch (VA.getLocInfo()) {
    default: llvm_unreachable("Unknown loc info!");
    case CCValAssign::Full: break;
    case CCValAssign::BCvt:
      if (!ReturnF16)
        Arg = DAG.getNode(ISD::BITCAST, dl, VA.getLocVT(), Arg);
      break;
    }
 
    // Mask f16 arguments if this is a CMSE nonsecure entry.
    auto RetVT = Outs[realRVLocIdx].ArgVT;
    if (AFI->isCmseNSEntryFunction() && (RetVT == MVT::f16)) {
      if (VA.needsCustom() && VA.getValVT() == MVT::f16) {
        Arg = MoveFromHPR(dl, DAG, VA.getLocVT(), VA.getValVT(), Arg);
      } else {
        auto LocBits = VA.getLocVT().getSizeInBits();
        auto MaskValue = APInt::getLowBitsSet(LocBits, RetVT.getSizeInBits());
        SDValue Mask =
            DAG.getConstant(MaskValue, dl, MVT::getIntegerVT(LocBits));
        Arg = DAG.getNode(ISD::BITCAST, dl, MVT::getIntegerVT(LocBits), Arg);
        Arg = DAG.getNode(ISD::AND, dl, MVT::getIntegerVT(LocBits), Arg, Mask);
        Arg = DAG.getNode(ISD::BITCAST, dl, VA.getLocVT(), Arg);
      }
    }
 
    if (VA.needsCustom() &&
        (VA.getLocVT() == MVT::v2f64 || VA.getLocVT() == MVT::f64)) {
      if (VA.getLocVT() == MVT::v2f64) {
        // Extract the first half and return it in two registers.
        SDValue Half = DAG.getNode(ISD::EXTRACT_VECTOR_ELT, dl, MVT::f64, Arg,
                                   DAG.getConstant(0, dl, MVT::i32));
        SDValue HalfGPRs = DAG.getNode(ARMISD::VMOVRRD, dl,
                                       DAG.getVTList(MVT::i32, MVT::i32), Half);
 
        Chain =
            DAG.getCopyToReg(Chain, dl, VA.getLocReg(),
                             HalfGPRs.getValue(isLittleEndian ? 0 : 1), Flag);
        Flag = Chain.getValue(1);
        RetOps.push_back(DAG.getRegister(VA.getLocReg(), VA.getLocVT()));
        VA = RVLocs[++i]; // skip ahead to next loc
        Chain =
            DAG.getCopyToReg(Chain, dl, VA.getLocReg(),
                             HalfGPRs.getValue(isLittleEndian ? 1 : 0), Flag);
        Flag = Chain.getValue(1);
        RetOps.push_back(DAG.getRegister(VA.getLocReg(), VA.getLocVT()));
        VA = RVLocs[++i]; // skip ahead to next loc
 
        // Extract the 2nd half and fall through to handle it as an f64 value.
        Arg = DAG.getNode(ISD::EXTRACT_VECTOR_ELT, dl, MVT::f64, Arg,
                          DAG.getConstant(1, dl, MVT::i32));
      }
      // Legalize ret f64 -> ret 2 x i32.  We always have fmrrd if f64 is
      // available.
      SDValue fmrrd = DAG.getNode(ARMISD::VMOVRRD, dl,
                                  DAG.getVTList(MVT::i32, MVT::i32), Arg);
      Chain = DAG.getCopyToReg(Chain, dl, VA.getLocReg(),
                               fmrrd.getValue(isLittleEndian ? 0 : 1), Flag);
      Flag = Chain.getValue(1);
      RetOps.push_back(DAG.getRegister(VA.getLocReg(), VA.getLocVT()));
      VA = RVLocs[++i]; // skip ahead to next loc
      Chain = DAG.getCopyToReg(Chain, dl, VA.getLocReg(),
                               fmrrd.getValue(isLittleEndian ? 1 : 0), Flag);
    } else
      Chain = DAG.getCopyToReg(Chain, dl, VA.getLocReg(), Arg, Flag);
 
    // Guarantee that all emitted copies are
    // stuck together, avoiding something bad.
    Flag = Chain.getValue(1);
    RetOps.push_back(DAG.getRegister(
        VA.getLocReg(), ReturnF16 ? Arg.getValueType() : VA.getLocVT()));
  }
  const ARMBaseRegisterInfo *TRI = Subtarget->getRegisterInfo();
  const MCPhysReg *I =
      TRI->getCalleeSavedRegsViaCopy(&DAG.getMachineFunction());
  if (I) {
    for (; *I; ++I) {
      if (ARM::GPRRegClass.contains(*I))
        RetOps.push_back(DAG.getRegister(*I, MVT::i32));
      else if (ARM::DPRRegClass.contains(*I))
        RetOps.push_back(DAG.getRegister(*I, MVT::getFloatingPointVT(64)));
      else
        llvm_unreachable("Unexpected register class in CSRsViaCopy!");
    }
  }
 
  // Update chain and glue.
  RetOps[0] = Chain;
  if (Flag.getNode())
    RetOps.push_back(Flag);
 
  // CPUs which aren't M-class use a special sequence to return from
  // exceptions (roughly, any instruction setting pc and cpsr simultaneously,
  // though we use "subs pc, lr, #N").
  //
  // M-class CPUs actually use a normal return sequence with a special
  // (hardware-provided) value in LR, so the normal code path works.
  if (DAG.getMachineFunction().getFunction().hasFnAttribute("interrupt") &&
      !Subtarget->isMClass()) {
    if (Subtarget->isThumb1Only())
      report_fatal_error("interrupt attribute is not supported in Thumb1");
    return LowerInterruptReturn(RetOps, dl, DAG);
  }
 
  ARMISD::NodeType RetNode = AFI->isCmseNSEntryFunction() ? ARMISD::SERET_FLAG :
                                                            ARMISD::RET_FLAG;
  return DAG.getNode(RetNode, dl, MVT::Other, RetOps);
}
 
bool ARMTargetLowering::isUsedByReturnOnly(SDNode *N, SDValue &Chain) const {
  if (N->getNumValues() != 1)
    return false;
  if (!N->hasNUsesOfValue(1, 0))
    return false;
 
  SDValue TCChain = Chain;
  SDNode *Copy = *N->use_begin();
  if (Copy->getOpcode() == ISD::CopyToReg) {
    // If the copy has a glue operand, we conservatively assume it isn't safe to
    // perform a tail call.
    if (Copy->getOperand(Copy->getNumOperands()-1).getValueType() == MVT::Glue)
      return false;
    TCChain = Copy->getOperand(0);
  } else if (Copy->getOpcode() == ARMISD::VMOVRRD) {
    SDNode *VMov = Copy;
    // f64 returned in a pair of GPRs.
    SmallPtrSet<SDNode*, 2> Copies;
    for (SDNode *U : VMov->uses()) {
      if (U->getOpcode() != ISD::CopyToReg)
        return false;
      Copies.insert(U);
    }
    if (Copies.size() > 2)
      return false;
 
    for (SDNode *U : VMov->uses()) {
      SDValue UseChain = U->getOperand(0);
      if (Copies.count(UseChain.getNode()))
        // Second CopyToReg
        Copy = U;
      else {
        // We are at the top of this chain.
        // If the copy has a glue operand, we conservatively assume it
        // isn't safe to perform a tail call.
        if (U->getOperand(U->getNumOperands() - 1).getValueType() == MVT::Glue)
          return false;
        // First CopyToReg
        TCChain = UseChain;
      }
    }
  } else if (Copy->getOpcode() == ISD::BITCAST) {
    // f32 returned in a single GPR.
    if (!Copy->hasOneUse())
      return false;
    Copy = *Copy->use_begin();
    if (Copy->getOpcode() != ISD::CopyToReg || !Copy->hasNUsesOfValue(1, 0))
      return false;
    // If the copy has a glue operand, we conservatively assume it isn't safe to
    // perform a tail call.
    if (Copy->getOperand(Copy->getNumOperands()-1).getValueType() == MVT::Glue)
      return false;
    TCChain = Copy->getOperand(0);
  } else {
    return false;
  }
 
  bool HasRet = false;
  for (const SDNode *U : Copy->uses()) {
    if (U->getOpcode() != ARMISD::RET_FLAG &&
        U->getOpcode() != ARMISD::INTRET_FLAG)
      return false;
    HasRet = true;
  }
 
  if (!HasRet)
    return false;
 
  Chain = TCChain;
  return true;
}
 
bool ARMTargetLowering::mayBeEmittedAsTailCall(const CallInst *CI) const {
  if (!Subtarget->supportsTailCall())
    return false;
 
  if (!CI->isTailCall())
    return false;
 
  return true;
}
 
// Trying to write a 64 bit value so need to split into two 32 bit values first,
// and pass the lower and high parts through.
static SDValue LowerWRITE_REGISTER(SDValue Op, SelectionDAG &DAG) {
  SDLoc DL(Op);
  SDValue WriteValue = Op->getOperand(2);
 
  // This function is only supposed to be called for i64 type argument.
  assert(WriteValue.getValueType() == MVT::i64
          && "LowerWRITE_REGISTER called for non-i64 type argument.");
 
  SDValue Lo = DAG.getNode(ISD::EXTRACT_ELEMENT, DL, MVT::i32, WriteValue,
                           DAG.getConstant(0, DL, MVT::i32));
  SDValue Hi = DAG.getNode(ISD::EXTRACT_ELEMENT, DL, MVT::i32, WriteValue,
                           DAG.getConstant(1, DL, MVT::i32));
  SDValue Ops[] = { Op->getOperand(0), Op->getOperand(1), Lo, Hi };
  return DAG.getNode(ISD::WRITE_REGISTER, DL, MVT::Other, Ops);
}
 
// ConstantPool, JumpTable, GlobalAddress, and ExternalSymbol are lowered as
// their target counterpart wrapped in the ARMISD::Wrapper node. Suppose N is
// one of the above mentioned nodes. It has to be wrapped because otherwise
// Select(N) returns N. So the raw TargetGlobalAddress nodes, etc. can only
// be used to form addressing mode. These wrapped nodes will be selected
// into MOVi.
SDValue ARMTargetLowering::LowerConstantPool(SDValue Op,
                                             SelectionDAG &DAG) const {
  EVT PtrVT = Op.getValueType();
  // FIXME there is no actual debug info here
  SDLoc dl(Op);
  ConstantPoolSDNode *CP = cast<ConstantPoolSDNode>(Op);
  SDValue Res;
 
  // When generating execute-only code Constant Pools must be promoted to the
  // global data section. It's a bit ugly that we can't share them across basic
  // blocks, but this way we guarantee that execute-only behaves correct with
  // position-independent addressing modes.
  if (Subtarget->genExecuteOnly()) {
    auto AFI = DAG.getMachineFunction().getInfo<ARMFunctionInfo>();
    auto T = const_cast<Type*>(CP->getType());
    auto C = const_cast<Constant*>(CP->getConstVal());
    auto M = const_cast<Module*>(DAG.getMachineFunction().
                                 getFunction().getParent());
    auto GV = new GlobalVariable(
                    *M, T, /*isConstant=*/true, GlobalVariable::InternalLinkage, C,
                    Twine(DAG.getDataLayout().getPrivateGlobalPrefix()) + "CP" +
                    Twine(DAG.getMachineFunction().getFunctionNumber()) + "_" +
                    Twine(AFI->createPICLabelUId())
                  );
    SDValue GA = DAG.getTargetGlobalAddress(dyn_cast<GlobalValue>(GV),
                                            dl, PtrVT);
    return LowerGlobalAddress(GA, DAG);
  }
 
  // The 16-bit ADR instruction can only encode offsets that are multiples of 4,
  // so we need to align to at least 4 bytes when we don't have 32-bit ADR.
  Align CPAlign = CP->getAlign();
  if (Subtarget->isThumb1Only())
    CPAlign = std::max(CPAlign, Align(4));
  if (CP->isMachineConstantPoolEntry())
    Res =
        DAG.getTargetConstantPool(CP->getMachineCPVal(), PtrVT, CPAlign);
  else
    Res = DAG.getTargetConstantPool(CP->getConstVal(), PtrVT, CPAlign);
  return DAG.getNode(ARMISD::Wrapper, dl, MVT::i32, Res);
}
 
unsigned ARMTargetLowering::getJumpTableEncoding() const {
  return MachineJumpTableInfo::EK_Inline;
}
 
SDValue ARMTargetLowering::LowerBlockAddress(SDValue Op,
                                             SelectionDAG &DAG) const {
  MachineFunction &MF = DAG.getMachineFunction();
  ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
  unsigned ARMPCLabelIndex = 0;
  SDLoc DL(Op);
  EVT PtrVT = getPointerTy(DAG.getDataLayout());
  const BlockAddress *BA = cast<BlockAddressSDNode>(Op)->getBlockAddress();
  SDValue CPAddr;
  bool IsPositionIndependent = isPositionIndependent() || Subtarget->isROPI();
  if (!IsPositionIndependent) {
    CPAddr = DAG.getTargetConstantPool(BA, PtrVT, Align(4));
  } else {
    unsigned PCAdj = Subtarget->isThumb() ? 4 : 8;
    ARMPCLabelIndex = AFI->createPICLabelUId();
    ARMConstantPoolValue *CPV =
      ARMConstantPoolConstant::Create(BA, ARMPCLabelIndex,
                                      ARMCP::CPBlockAddress, PCAdj);
    CPAddr = DAG.getTargetConstantPool(CPV, PtrVT, Align(4));
  }
  CPAddr = DAG.getNode(ARMISD::Wrapper, DL, PtrVT, CPAddr);
  SDValue Result = DAG.getLoad(
      PtrVT, DL, DAG.getEntryNode(), CPAddr,
      MachinePointerInfo::getConstantPool(DAG.getMachineFunction()));
  if (!IsPositionIndependent)
    return Result;
  SDValue PICLabel = DAG.getConstant(ARMPCLabelIndex, DL, MVT::i32);
  return DAG.getNode(ARMISD::PIC_ADD, DL, PtrVT, Result, PICLabel);
}
 
/// Convert a TLS address reference into the correct sequence of loads
/// and calls to compute the variable's address for Darwin, and return an
/// SDValue containing the final node.
 
/// Darwin only has one TLS scheme which must be capable of dealing with the
/// fully general situation, in the worst case. This means:
///     + "extern __thread" declaration.
///     + Defined in a possibly unknown dynamic library.
///
/// The general system is that each __thread variable has a [3 x i32] descriptor
/// which contains information used by the runtime to calculate the address. The
/// only part of this the compiler needs to know about is the first word, which
/// contains a function pointer that must be called with the address of the
/// entire descriptor in "r0".
///
/// Since this descriptor may be in a different unit, in general access must
/// proceed along the usual ARM rules. A common sequence to produce is:
///
///     movw rT1, :lower16:_var$non_lazy_ptr
///     movt rT1, :upper16:_var$non_lazy_ptr
///     ldr r0, [rT1]
///     ldr rT2, [r0]
///     blx rT2
///     [...address now in r0...]
SDValue
ARMTargetLowering::LowerGlobalTLSAddressDarwin(SDValue Op,
                                               SelectionDAG &DAG) const {
  assert(Subtarget->isTargetDarwin() &&
         "This function expects a Darwin target");
  SDLoc DL(Op);
 
  // First step is to get the address of the actua global symbol. This is where
  // the TLS descriptor lives.
  SDValue DescAddr = LowerGlobalAddressDarwin(Op, DAG);
 
  // The first entry in the descriptor is a function pointer that we must call
  // to obtain the address of the variable.
  SDValue Chain = DAG.getEntryNode();
  SDValue FuncTLVGet = DAG.getLoad(
      MVT::i32, DL, Chain, DescAddr,
      MachinePointerInfo::getGOT(DAG.getMachineFunction()), Align(4),
      MachineMemOperand::MONonTemporal | MachineMemOperand::MODereferenceable |
          MachineMemOperand::MOInvariant);
  Chain = FuncTLVGet.getValue(1);
 
  MachineFunction &F = DAG.getMachineFunction();
  MachineFrameInfo &MFI = F.getFrameInfo();
  MFI.setAdjustsStack(true);
 
  // TLS calls preserve all registers except those that absolutely must be
  // trashed: R0 (it takes an argument), LR (it's a call) and CPSR (let's not be
  // silly).
  auto TRI =
      getTargetMachine().getSubtargetImpl(F.getFunction())->getRegisterInfo();
  auto ARI = static_cast<const ARMRegisterInfo *>(TRI);
  const uint32_t *Mask = ARI->getTLSCallPreservedMask(DAG.getMachineFunction());
 
  // Finally, we can make the call. This is just a degenerate version of a
  // normal AArch64 call node: r0 takes the address of the descriptor, and
  // returns the address of the variable in this thread.
  Chain = DAG.getCopyToReg(Chain, DL, ARM::R0, DescAddr, SDValue());
  Chain =
      DAG.getNode(ARMISD::CALL, DL, DAG.getVTList(MVT::Other, MVT::Glue),
                  Chain, FuncTLVGet, DAG.getRegister(ARM::R0, MVT::i32),
                  DAG.getRegisterMask(Mask), Chain.getValue(1));
  return DAG.getCopyFromReg(Chain, DL, ARM::R0, MVT::i32, Chain.getValue(1));
}
 
SDValue
ARMTargetLowering::LowerGlobalTLSAddressWindows(SDValue Op,
                                                SelectionDAG &DAG) const {
  assert(Subtarget->isTargetWindows() && "Windows specific TLS lowering");
 
  SDValue Chain = DAG.getEntryNode();
  EVT PtrVT = getPointerTy(DAG.getDataLayout());
  SDLoc DL(Op);
 
  // Load the current TEB (thread environment block)
  SDValue Ops[] = {Chain,
                   DAG.getTargetConstant(Intrinsic::arm_mrc, DL, MVT::i32),
                   DAG.getTargetConstant(15, DL, MVT::i32),
                   DAG.getTargetConstant(0, DL, MVT::i32),
                   DAG.getTargetConstant(13, DL, MVT::i32),
                   DAG.getTargetConstant(0, DL, MVT::i32),
                   DAG.getTargetConstant(2, DL, MVT::i32)};
  SDValue CurrentTEB = DAG.getNode(ISD::INTRINSIC_W_CHAIN, DL,
                                   DAG.getVTList(MVT::i32, MVT::Other), Ops);
 
  SDValue TEB = CurrentTEB.getValue(0);
  Chain = CurrentTEB.getValue(1);
 
  // Load the ThreadLocalStoragePointer from the TEB
  // A pointer to the TLS array is located at offset 0x2c from the TEB.
  SDValue TLSArray =
      DAG.getNode(ISD::ADD, DL, PtrVT, TEB, DAG.getIntPtrConstant(0x2c, DL));
  TLSArray = DAG.getLoad(PtrVT, DL, Chain, TLSArray, MachinePointerInfo());
 
  // The pointer to the thread's TLS data area is at the TLS Index scaled by 4
  // offset into the TLSArray.
 
  // Load the TLS index from the C runtime
  SDValue TLSIndex =
      DAG.getTargetExternalSymbol("_tls_index", PtrVT, ARMII::MO_NO_FLAG);
  TLSIndex = DAG.getNode(ARMISD::Wrapper, DL, PtrVT, TLSIndex);
  TLSIndex = DAG.getLoad(PtrVT, DL, Chain, TLSIndex, MachinePointerInfo());
 
  SDValue Slot = DAG.getNode(ISD::SHL, DL, PtrVT, TLSIndex,
                              DAG.getConstant(2, DL, MVT::i32));
  SDValue TLS = DAG.getLoad(PtrVT, DL, Chain,
                            DAG.getNode(ISD::ADD, DL, PtrVT, TLSArray, Slot),
                            MachinePointerInfo());
 
  // Get the offset of the start of the .tls section (section base)
  const auto *GA = cast<GlobalAddressSDNode>(Op);
  auto *CPV = ARMConstantPoolConstant::Create(GA->getGlobal(), ARMCP::SECREL);
  SDValue Offset = DAG.getLoad(
      PtrVT, DL, Chain,
      DAG.getNode(ARMISD::Wrapper, DL, MVT::i32,
                  DAG.getTargetConstantPool(CPV, PtrVT, Align(4))),
      MachinePointerInfo::getConstantPool(DAG.getMachineFunction()));
 
  return DAG.getNode(ISD::ADD, DL, PtrVT, TLS, Offset);
}
 
// Lower ISD::GlobalTLSAddress using the "general dynamic" model
SDValue
ARMTargetLowering::LowerToTLSGeneralDynamicModel(GlobalAddressSDNode *GA,
                                                 SelectionDAG &DAG) const {
  SDLoc dl(GA);
  EVT PtrVT = getPointerTy(DAG.getDataLayout());
  unsigned char PCAdj = Subtarget->isThumb() ? 4 : 8;
  MachineFunction &MF = DAG.getMachineFunction();
  ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
  unsigned ARMPCLabelIndex = AFI->createPICLabelUId();
  ARMConstantPoolValue *CPV =
    ARMConstantPoolConstant::Create(GA->getGlobal(), ARMPCLabelIndex,
                                    ARMCP::CPValue, PCAdj, ARMCP::TLSGD, true);
  SDValue Argument = DAG.getTargetConstantPool(CPV, PtrVT, Align(4));
  Argument = DAG.getNode(ARMISD::Wrapper, dl, MVT::i32, Argument);
  Argument = DAG.getLoad(
      PtrVT, dl, DAG.getEntryNode(), Argument,
      MachinePointerInfo::getConstantPool(DAG.getMachineFunction()));
  SDValue Chain = Argument.getValue(1);
 
  SDValue PICLabel = DAG.getConstant(ARMPCLabelIndex, dl, MVT::i32);
  Argument = DAG.getNode(ARMISD::PIC_ADD, dl, PtrVT, Argument, PICLabel);
 
  // call __tls_get_addr.
  ArgListTy Args;
  ArgListEntry Entry;
  Entry.Node = Argument;
  Entry.Ty = (Type *) Type::getInt32Ty(*DAG.getContext());
  Args.push_back(Entry);
 
  // FIXME: is there useful debug info available here?
  TargetLowering::CallLoweringInfo CLI(DAG);
  CLI.setDebugLoc(dl).setChain(Chain).setLibCallee(
      CallingConv::C, Type::getInt32Ty(*DAG.getContext()),
      DAG.getExternalSymbol("__tls_get_addr", PtrVT), std::move(Args));
 
  std::pair<SDValue, SDValue> CallResult = LowerCallTo(CLI);
  return CallResult.first;
}
 
// Lower ISD::GlobalTLSAddress using the "initial exec" or
// "local exec" model.
SDValue
ARMTargetLowering::LowerToTLSExecModels(GlobalAddressSDNode *GA,
                                        SelectionDAG &DAG,
                                        TLSModel::Model model) const {
  const GlobalValue *GV = GA->getGlobal();
  SDLoc dl(GA);
  SDValue Offset;
  SDValue Chain = DAG.getEntryNode();
  EVT PtrVT = getPointerTy(DAG.getDataLayout());
  // Get the Thread Pointer
  SDValue ThreadPointer = DAG.getNode(ARMISD::THREAD_POINTER, dl, PtrVT);
 
  if (model == TLSModel::InitialExec) {
    MachineFunction &MF = DAG.getMachineFunction();
    ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
    unsigned ARMPCLabelIndex = AFI->createPICLabelUId();
    // Initial exec model.
    unsigned char PCAdj = Subtarget->isThumb() ? 4 : 8;
    ARMConstantPoolValue *CPV =
      ARMConstantPoolConstant::Create(GA->getGlobal(), ARMPCLabelIndex,
                                      ARMCP::CPValue, PCAdj, ARMCP::GOTTPOFF,
                                      true);
    Offset = DAG.getTargetConstantPool(CPV, PtrVT, Align(4));
    Offset = DAG.getNode(ARMISD::Wrapper, dl, MVT::i32, Offset);
    Offset = DAG.getLoad(
        PtrVT, dl, Chain, Offset,
        MachinePointerInfo::getConstantPool(DAG.getMachineFunction()));
    Chain = Offset.getValue(1);
 
    SDValue PICLabel = DAG.getConstant(ARMPCLabelIndex, dl, MVT::i32);
    Offset = DAG.getNode(ARMISD::PIC_ADD, dl, PtrVT, Offset, PICLabel);
 
    Offset = DAG.getLoad(
        PtrVT, dl, Chain, Offset,
        MachinePointerInfo::getConstantPool(DAG.getMachineFunction()));
  } else {
    // local exec model
    assert(model == TLSModel::LocalExec);
    ARMConstantPoolValue *CPV =
      ARMConstantPoolConstant::Create(GV, ARMCP::TPOFF);
    Offset = DAG.getTargetConstantPool(CPV, PtrVT, Align(4));
    Offset = DAG.getNode(ARMISD::Wrapper, dl, MVT::i32, Offset);
    Offset = DAG.getLoad(
        PtrVT, dl, Chain, Offset,
        MachinePointerInfo::getConstantPool(DAG.getMachineFunction()));
  }
 
  // The address of the thread local variable is the add of the thread
  // pointer with the offset of the variable.
  return DAG.getNode(ISD::ADD, dl, PtrVT, ThreadPointer, Offset);
}
 
SDValue
ARMTargetLowering::LowerGlobalTLSAddress(SDValue Op, SelectionDAG &DAG) const {
  GlobalAddressSDNode *GA = cast<GlobalAddressSDNode>(Op);
  if (DAG.getTarget().useEmulatedTLS())
    return LowerToTLSEmulatedModel(GA, DAG);
 
  if (Subtarget->isTargetDarwin())
    return LowerGlobalTLSAddressDarwin(Op, DAG);
 
  if (Subtarget->isTargetWindows())
    return LowerGlobalTLSAddressWindows(Op, DAG);
 
  // TODO: implement the "local dynamic" model
  assert(Subtarget->isTargetELF() && "Only ELF implemented here");
  TLSModel::Model model = getTargetMachine().getTLSModel(GA->getGlobal());
 
  switch (model) {
    case TLSModel::GeneralDynamic:
    case TLSModel::LocalDynamic:
      return LowerToTLSGeneralDynamicModel(GA, DAG);
    case TLSModel::InitialExec:
    case TLSModel::LocalExec:
      return LowerToTLSExecModels(GA, DAG, model);
  }
  llvm_unreachable("bogus TLS model");
}
 
/// Return true if all users of V are within function F, looking through
/// ConstantExprs.
static bool allUsersAreInFunction(const Value *V, const Function *F) {
  SmallVector<const User*,4> Worklist(V->users());
  while (!Worklist.empty()) {
    auto *U = Worklist.pop_back_val();
    if (isa<ConstantExpr>(U)) {
      append_range(Worklist, U->users());
      continue;
    }
 
    auto *I = dyn_cast<Instruction>(U);
    if (!I || I->getParent()->getParent() != F)
      return false;
  }
  return true;
}
 
static SDValue promoteToConstantPool(const ARMTargetLowering *TLI,
                                     const GlobalValue *GV, SelectionDAG &DAG,
                                     EVT PtrVT, const SDLoc &dl) {
  // If we're creating a pool entry for a constant global with unnamed address,
  // and the global is small enough, we can emit it inline into the constant pool
  // to save ourselves an indirection.
  //
  // This is a win if the constant is only used in one function (so it doesn't
  // need to be duplicated) or duplicating the constant wouldn't increase code
  // size (implying the constant is no larger than 4 bytes).
  const Function &F = DAG.getMachineFunction().getFunction();
 
  // We rely on this decision to inline being idemopotent and unrelated to the
  // use-site. We know that if we inline a variable at one use site, we'll
  // inline it elsewhere too (and reuse the constant pool entry). Fast-isel