xref: /freebsd-12.1/contrib/llvm/lib/Target/ARM/ARMFrameLowering.cpp (revision 2cab237b)

//===- ARMFrameLowering.cpp - ARM Frame Information -----------------------===//
//
//                     The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//
// This file contains the ARM implementation of TargetFrameLowering class.
//
//===----------------------------------------------------------------------===//

#include "ARMFrameLowering.h"
#include "ARMBaseInstrInfo.h"
#include "ARMBaseRegisterInfo.h"
#include "ARMConstantPoolValue.h"
#include "ARMMachineFunctionInfo.h"
#include "ARMSubtarget.h"
#include "MCTargetDesc/ARMAddressingModes.h"
#include "MCTargetDesc/ARMBaseInfo.h"
#include "Utils/ARMBaseInfo.h"
#include "llvm/ADT/BitVector.h"
#include "llvm/ADT/STLExtras.h"
#include "llvm/ADT/SmallPtrSet.h"
#include "llvm/ADT/SmallVector.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/MachineModuleInfo.h"
#include "llvm/CodeGen/MachineOperand.h"
#include "llvm/CodeGen/MachineRegisterInfo.h"
#include "llvm/CodeGen/RegisterScavenging.h"
#include "llvm/CodeGen/TargetInstrInfo.h"
#include "llvm/CodeGen/TargetOpcodes.h"
#include "llvm/CodeGen/TargetRegisterInfo.h"
#include "llvm/CodeGen/TargetSubtargetInfo.h"
#include "llvm/IR/Attributes.h"
#include "llvm/IR/CallingConv.h"
#include "llvm/IR/DebugLoc.h"
#include "llvm/IR/Function.h"
#include "llvm/MC/MCContext.h"
#include "llvm/MC/MCDwarf.h"
#include "llvm/MC/MCInstrDesc.h"
#include "llvm/MC/MCRegisterInfo.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/MathExtras.h"
#include "llvm/Support/raw_ostream.h"
#include "llvm/Target/TargetMachine.h"
#include "llvm/Target/TargetOptions.h"
#include <algorithm>
#include <cassert>
#include <cstddef>
#include <cstdint>
#include <iterator>
#include <utility>
#include <vector>

#define DEBUG_TYPE "arm-frame-lowering"

using namespace llvm;

static cl::opt<bool>
SpillAlignedNEONRegs("align-neon-spills", cl::Hidden, cl::init(true),
                     cl::desc("Align ARM NEON spills in prolog and epilog"));

static MachineBasicBlock::iterator
skipAlignedDPRCS2Spills(MachineBasicBlock::iterator MI,
                        unsigned NumAlignedDPRCS2Regs);

ARMFrameLowering::ARMFrameLowering(const ARMSubtarget &sti)
    : TargetFrameLowering(StackGrowsDown, sti.getStackAlignment(), 0, 4),
      STI(sti) {}

bool ARMFrameLowering::noFramePointerElim(const MachineFunction &MF) const {
  // iOS always has a FP for backtracking, force other targets to keep their FP
  // when doing FastISel. The emitted code is currently superior, and in cases
  // like test-suite's lencod FastISel isn't quite correct when FP is eliminated.
  return TargetFrameLowering::noFramePointerElim(MF) ||
         MF.getSubtarget<ARMSubtarget>().useFastISel();
}

/// hasFP - Return true if the specified function should have a dedicated frame
/// pointer register.  This is true if the function has variable sized allocas
/// or if frame pointer elimination is disabled.
bool ARMFrameLowering::hasFP(const MachineFunction &MF) const {
  const TargetRegisterInfo *RegInfo = MF.getSubtarget().getRegisterInfo();
  const MachineFrameInfo &MFI = MF.getFrameInfo();

  // ABI-required frame pointer.
  if (MF.getTarget().Options.DisableFramePointerElim(MF))
    return true;

  // Frame pointer required for use within this function.
  return (RegInfo->needsStackRealignment(MF) ||
          MFI.hasVarSizedObjects() ||
          MFI.isFrameAddressTaken());
}

/// hasReservedCallFrame - Under normal circumstances, when a frame pointer is
/// not required, we reserve argument space for call sites in the function
/// immediately on entry to the current function.  This eliminates the need for
/// add/sub sp brackets around call sites.  Returns true if the call frame is
/// included as part of the stack frame.
bool ARMFrameLowering::hasReservedCallFrame(const MachineFunction &MF) const {
  const MachineFrameInfo &MFI = MF.getFrameInfo();
  unsigned CFSize = MFI.getMaxCallFrameSize();
  // It's not always a good idea to include the call frame as part of the
  // stack frame. ARM (especially Thumb) has small immediate offset to
  // address the stack frame. So a large call frame can cause poor codegen
  // and may even makes it impossible to scavenge a register.
  if (CFSize >= ((1 << 12) - 1) / 2)  // Half of imm12
    return false;

  return !MFI.hasVarSizedObjects();
}

/// canSimplifyCallFramePseudos - If there is a reserved call frame, the
/// call frame pseudos can be simplified.  Unlike most targets, having a FP
/// is not sufficient here since we still may reference some objects via SP
/// even when FP is available in Thumb2 mode.
bool
ARMFrameLowering::canSimplifyCallFramePseudos(const MachineFunction &MF) const {
  return hasReservedCallFrame(MF) || MF.getFrameInfo().hasVarSizedObjects();
}

static bool isCSRestore(MachineInstr &MI, const ARMBaseInstrInfo &TII,
                        const MCPhysReg *CSRegs) {
  // Integer spill area is handled with "pop".
  if (isPopOpcode(MI.getOpcode())) {
    // The first two operands are predicates. The last two are
    // imp-def and imp-use of SP. Check everything in between.
    for (int i = 5, e = MI.getNumOperands(); i != e; ++i)
      if (!isCalleeSavedRegister(MI.getOperand(i).getReg(), CSRegs))
        return false;
    return true;
  }
  if ((MI.getOpcode() == ARM::LDR_POST_IMM ||
       MI.getOpcode() == ARM::LDR_POST_REG ||
       MI.getOpcode() == ARM::t2LDR_POST) &&
      isCalleeSavedRegister(MI.getOperand(0).getReg(), CSRegs) &&
      MI.getOperand(1).getReg() == ARM::SP)
    return true;

  return false;
}

static void emitRegPlusImmediate(
    bool isARM, MachineBasicBlock &MBB, MachineBasicBlock::iterator &MBBI,
    const DebugLoc &dl, const ARMBaseInstrInfo &TII, unsigned DestReg,
    unsigned SrcReg, int NumBytes, unsigned MIFlags = MachineInstr::NoFlags,
    ARMCC::CondCodes Pred = ARMCC::AL, unsigned PredReg = 0) {
  if (isARM)
    emitARMRegPlusImmediate(MBB, MBBI, dl, DestReg, SrcReg, NumBytes,
                            Pred, PredReg, TII, MIFlags);
  else
    emitT2RegPlusImmediate(MBB, MBBI, dl, DestReg, SrcReg, NumBytes,
                           Pred, PredReg, TII, MIFlags);
}

static void emitSPUpdate(bool isARM, MachineBasicBlock &MBB,
                         MachineBasicBlock::iterator &MBBI, const DebugLoc &dl,
                         const ARMBaseInstrInfo &TII, int NumBytes,
                         unsigned MIFlags = MachineInstr::NoFlags,
                         ARMCC::CondCodes Pred = ARMCC::AL,
                         unsigned PredReg = 0) {
  emitRegPlusImmediate(isARM, MBB, MBBI, dl, TII, ARM::SP, ARM::SP, NumBytes,
                       MIFlags, Pred, PredReg);
}

static int sizeOfSPAdjustment(const MachineInstr &MI) {
  int RegSize;
  switch (MI.getOpcode()) {
  case ARM::VSTMDDB_UPD:
    RegSize = 8;
    break;
  case ARM::STMDB_UPD:
  case ARM::t2STMDB_UPD:
    RegSize = 4;
    break;
  case ARM::t2STR_PRE:
  case ARM::STR_PRE_IMM:
    return 4;
  default:
    llvm_unreachable("Unknown push or pop like instruction");
  }

  int count = 0;
  // ARM and Thumb2 push/pop insts have explicit "sp, sp" operands (+
  // pred) so the list starts at 4.
  for (int i = MI.getNumOperands() - 1; i >= 4; --i)
    count += RegSize;
  return count;
}

static bool WindowsRequiresStackProbe(const MachineFunction &MF,
                                      size_t StackSizeInBytes) {
  const MachineFrameInfo &MFI = MF.getFrameInfo();
  const Function &F = MF.getFunction();
  unsigned StackProbeSize = (MFI.getStackProtectorIndex() > 0) ? 4080 : 4096;
  if (F.hasFnAttribute("stack-probe-size"))
    F.getFnAttribute("stack-probe-size")
        .getValueAsString()
        .getAsInteger(0, StackProbeSize);
  return StackSizeInBytes >= StackProbeSize;
}

namespace {

struct StackAdjustingInsts {
  struct InstInfo {
    MachineBasicBlock::iterator I;
    unsigned SPAdjust;
    bool BeforeFPSet;
  };

  SmallVector<InstInfo, 4> Insts;

  void addInst(MachineBasicBlock::iterator I, unsigned SPAdjust,
               bool BeforeFPSet = false) {
    InstInfo Info = {I, SPAdjust, BeforeFPSet};
    Insts.push_back(Info);
  }

  void addExtraBytes(const MachineBasicBlock::iterator I, unsigned ExtraBytes) {
    auto Info =
        llvm::find_if(Insts, [&](InstInfo &Info) { return Info.I == I; });
    assert(Info != Insts.end() && "invalid sp adjusting instruction");
    Info->SPAdjust += ExtraBytes;
  }

  void emitDefCFAOffsets(MachineBasicBlock &MBB, const DebugLoc &dl,
                         const ARMBaseInstrInfo &TII, bool HasFP) {
    MachineFunction &MF = *MBB.getParent();
    unsigned CFAOffset = 0;
    for (auto &Info : Insts) {
      if (HasFP && !Info.BeforeFPSet)
        return;

      CFAOffset -= Info.SPAdjust;
      unsigned CFIIndex = MF.addFrameInst(
          MCCFIInstruction::createDefCfaOffset(nullptr, CFAOffset));
      BuildMI(MBB, std::next(Info.I), dl,
              TII.get(TargetOpcode::CFI_INSTRUCTION))
              .addCFIIndex(CFIIndex)
              .setMIFlags(MachineInstr::FrameSetup);
    }
  }
};

} // end anonymous namespace

/// Emit an instruction sequence that will align the address in
/// register Reg by zero-ing out the lower bits.  For versions of the
/// architecture that support Neon, this must be done in a single
/// instruction, since skipAlignedDPRCS2Spills assumes it is done in a
/// single instruction. That function only gets called when optimizing
/// spilling of D registers on a core with the Neon instruction set
/// present.
static void emitAligningInstructions(MachineFunction &MF, ARMFunctionInfo *AFI,
                                     const TargetInstrInfo &TII,
                                     MachineBasicBlock &MBB,
                                     MachineBasicBlock::iterator MBBI,
                                     const DebugLoc &DL, const unsigned Reg,
                                     const unsigned Alignment,
                                     const bool MustBeSingleInstruction) {
  const ARMSubtarget &AST =
      static_cast<const ARMSubtarget &>(MF.getSubtarget());
  const bool CanUseBFC = AST.hasV6T2Ops() || AST.hasV7Ops();
  const unsigned AlignMask = Alignment - 1;
  const unsigned NrBitsToZero = countTrailingZeros(Alignment);
  assert(!AFI->isThumb1OnlyFunction() && "Thumb1 not supported");
  if (!AFI->isThumbFunction()) {
    // if the BFC instruction is available, use that to zero the lower
    // bits:
    //   bfc Reg, #0, log2(Alignment)
    // otherwise use BIC, if the mask to zero the required number of bits
    // can be encoded in the bic immediate field
    //   bic Reg, Reg, Alignment-1
    // otherwise, emit
    //   lsr Reg, Reg, log2(Alignment)
    //   lsl Reg, Reg, log2(Alignment)
    if (CanUseBFC) {
      BuildMI(MBB, MBBI, DL, TII.get(ARM::BFC), Reg)
          .addReg(Reg, RegState::Kill)
          .addImm(~AlignMask)
          .add(predOps(ARMCC::AL));
    } else if (AlignMask <= 255) {
      BuildMI(MBB, MBBI, DL, TII.get(ARM::BICri), Reg)
          .addReg(Reg, RegState::Kill)
          .addImm(AlignMask)
          .add(predOps(ARMCC::AL))
          .add(condCodeOp());
    } else {
      assert(!MustBeSingleInstruction &&
             "Shouldn't call emitAligningInstructions demanding a single "
             "instruction to be emitted for large stack alignment for a target "
             "without BFC.");
      BuildMI(MBB, MBBI, DL, TII.get(ARM::MOVsi), Reg)
          .addReg(Reg, RegState::Kill)
          .addImm(ARM_AM::getSORegOpc(ARM_AM::lsr, NrBitsToZero))
          .add(predOps(ARMCC::AL))
          .add(condCodeOp());
      BuildMI(MBB, MBBI, DL, TII.get(ARM::MOVsi), Reg)
          .addReg(Reg, RegState::Kill)
          .addImm(ARM_AM::getSORegOpc(ARM_AM::lsl, NrBitsToZero))
          .add(predOps(ARMCC::AL))
          .add(condCodeOp());
    }
  } else {
    // Since this is only reached for Thumb-2 targets, the BFC instruction
    // should always be available.
    assert(CanUseBFC);
    BuildMI(MBB, MBBI, DL, TII.get(ARM::t2BFC), Reg)
        .addReg(Reg, RegState::Kill)
        .addImm(~AlignMask)
        .add(predOps(ARMCC::AL));
  }
}

/// We need the offset of the frame pointer relative to other MachineFrameInfo
/// offsets which are encoded relative to SP at function begin.
/// See also emitPrologue() for how the FP is set up.
/// Unfortunately we cannot determine this value in determineCalleeSaves() yet
/// as assignCalleeSavedSpillSlots() hasn't run at this point. Instead we use
/// this to produce a conservative estimate that we check in an assert() later.
static int getMaxFPOffset(const Function &F, const ARMFunctionInfo &AFI) {
  // This is a conservative estimation: Assume the frame pointer being r7 and
  // pc("r15") up to r8 getting spilled before (= 8 registers).
  return -AFI.getArgRegsSaveSize() - (8 * 4);
}

void ARMFrameLowering::emitPrologue(MachineFunction &MF,
                                    MachineBasicBlock &MBB) const {
  MachineBasicBlock::iterator MBBI = MBB.begin();
  MachineFrameInfo  &MFI = MF.getFrameInfo();
  ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
  MachineModuleInfo &MMI = MF.getMMI();
  MCContext &Context = MMI.getContext();
  const TargetMachine &TM = MF.getTarget();
  const MCRegisterInfo *MRI = Context.getRegisterInfo();
  const ARMBaseRegisterInfo *RegInfo = STI.getRegisterInfo();
  const ARMBaseInstrInfo &TII = *STI.getInstrInfo();
  assert(!AFI->isThumb1OnlyFunction() &&
         "This emitPrologue does not support Thumb1!");
  bool isARM = !AFI->isThumbFunction();
  unsigned Align = STI.getFrameLowering()->getStackAlignment();
  unsigned ArgRegsSaveSize = AFI->getArgRegsSaveSize();
  unsigned NumBytes = MFI.getStackSize();
  const std::vector<CalleeSavedInfo> &CSI = MFI.getCalleeSavedInfo();

  // Debug location must be unknown since the first debug location is used
  // to determine the end of the prologue.
  DebugLoc dl;

  unsigned FramePtr = RegInfo->getFrameRegister(MF);

  // Determine the sizes of each callee-save spill areas and record which frame
  // belongs to which callee-save spill areas.
  unsigned GPRCS1Size = 0, GPRCS2Size = 0, DPRCSSize = 0;
  int FramePtrSpillFI = 0;
  int D8SpillFI = 0;

  // All calls are tail calls in GHC calling conv, and functions have no
  // prologue/epilogue.
  if (MF.getFunction().getCallingConv() == CallingConv::GHC)
    return;

  StackAdjustingInsts DefCFAOffsetCandidates;
  bool HasFP = hasFP(MF);

  // Allocate the vararg register save area.
  if (ArgRegsSaveSize) {
    emitSPUpdate(isARM, MBB, MBBI, dl, TII, -ArgRegsSaveSize,
                 MachineInstr::FrameSetup);
    DefCFAOffsetCandidates.addInst(std::prev(MBBI), ArgRegsSaveSize, true);
  }

  if (!AFI->hasStackFrame() &&
      (!STI.isTargetWindows() || !WindowsRequiresStackProbe(MF, NumBytes))) {
    if (NumBytes - ArgRegsSaveSize != 0) {
      emitSPUpdate(isARM, MBB, MBBI, dl, TII, -(NumBytes - ArgRegsSaveSize),
                   MachineInstr::FrameSetup);
      DefCFAOffsetCandidates.addInst(std::prev(MBBI),
                                     NumBytes - ArgRegsSaveSize, true);
    }
    DefCFAOffsetCandidates.emitDefCFAOffsets(MBB, dl, TII, HasFP);
    return;
  }

  // Determine spill area sizes.
  for (unsigned i = 0, e = CSI.size(); i != e; ++i) {
    unsigned Reg = CSI[i].getReg();
    int FI = CSI[i].getFrameIdx();
    switch (Reg) {
    case ARM::R8:
    case ARM::R9:
    case ARM::R10:
    case ARM::R11:
    case ARM::R12:
      if (STI.splitFramePushPop(MF)) {
        GPRCS2Size += 4;
        break;
      }
      LLVM_FALLTHROUGH;
    case ARM::R0:
    case ARM::R1:
    case ARM::R2:
    case ARM::R3:
    case ARM::R4:
    case ARM::R5:
    case ARM::R6:
    case ARM::R7:
    case ARM::LR:
      if (Reg == FramePtr)
        FramePtrSpillFI = FI;
      GPRCS1Size += 4;
      break;
    default:
      // This is a DPR. Exclude the aligned DPRCS2 spills.
      if (Reg == ARM::D8)
        D8SpillFI = FI;
      if (Reg < ARM::D8 || Reg >= ARM::D8 + AFI->getNumAlignedDPRCS2Regs())
        DPRCSSize += 8;
    }
  }

  // Move past area 1.
  MachineBasicBlock::iterator LastPush = MBB.end(), GPRCS1Push, GPRCS2Push;
  if (GPRCS1Size > 0) {
    GPRCS1Push = LastPush = MBBI++;
    DefCFAOffsetCandidates.addInst(LastPush, GPRCS1Size, true);
  }

  // Determine starting offsets of spill areas.
  unsigned GPRCS1Offset = NumBytes - ArgRegsSaveSize - GPRCS1Size;
  unsigned GPRCS2Offset = GPRCS1Offset - GPRCS2Size;
  unsigned DPRAlign = DPRCSSize ? std::min(8U, Align) : 4U;
  unsigned DPRGapSize = (GPRCS1Size + GPRCS2Size + ArgRegsSaveSize) % DPRAlign;
  unsigned DPRCSOffset = GPRCS2Offset - DPRGapSize - DPRCSSize;
  int FramePtrOffsetInPush = 0;
  if (HasFP) {
    int FPOffset = MFI.getObjectOffset(FramePtrSpillFI);
    assert(getMaxFPOffset(MF.getFunction(), *AFI) <= FPOffset &&
           "Max FP estimation is wrong");
    FramePtrOffsetInPush = FPOffset + ArgRegsSaveSize;
    AFI->setFramePtrSpillOffset(MFI.getObjectOffset(FramePtrSpillFI) +
                                NumBytes);
  }
  AFI->setGPRCalleeSavedArea1Offset(GPRCS1Offset);
  AFI->setGPRCalleeSavedArea2Offset(GPRCS2Offset);
  AFI->setDPRCalleeSavedAreaOffset(DPRCSOffset);

  // Move past area 2.
  if (GPRCS2Size > 0) {
    GPRCS2Push = LastPush = MBBI++;
    DefCFAOffsetCandidates.addInst(LastPush, GPRCS2Size);
  }

  // Prolog/epilog inserter assumes we correctly align DPRs on the stack, so our
  // .cfi_offset operations will reflect that.
  if (DPRGapSize) {
    assert(DPRGapSize == 4 && "unexpected alignment requirements for DPRs");
    if (LastPush != MBB.end() &&
        tryFoldSPUpdateIntoPushPop(STI, MF, &*LastPush, DPRGapSize))
      DefCFAOffsetCandidates.addExtraBytes(LastPush, DPRGapSize);
    else {
      emitSPUpdate(isARM, MBB, MBBI, dl, TII, -DPRGapSize,
                   MachineInstr::FrameSetup);
      DefCFAOffsetCandidates.addInst(std::prev(MBBI), DPRGapSize);
    }
  }

  // Move past area 3.
  if (DPRCSSize > 0) {
    // Since vpush register list cannot have gaps, there may be multiple vpush
    // instructions in the prologue.
    while (MBBI != MBB.end() && MBBI->getOpcode() == ARM::VSTMDDB_UPD) {
      DefCFAOffsetCandidates.addInst(MBBI, sizeOfSPAdjustment(*MBBI));
      LastPush = MBBI++;
    }
  }

  // Move past the aligned DPRCS2 area.
  if (AFI->getNumAlignedDPRCS2Regs() > 0) {
    MBBI = skipAlignedDPRCS2Spills(MBBI, AFI->getNumAlignedDPRCS2Regs());
    // The code inserted by emitAlignedDPRCS2Spills realigns the stack, and
    // leaves the stack pointer pointing to the DPRCS2 area.
    //
    // Adjust NumBytes to represent the stack slots below the DPRCS2 area.
    NumBytes += MFI.getObjectOffset(D8SpillFI);
  } else
    NumBytes = DPRCSOffset;

  if (STI.isTargetWindows() && WindowsRequiresStackProbe(MF, NumBytes)) {
    uint32_t NumWords = NumBytes >> 2;

    if (NumWords < 65536)
      BuildMI(MBB, MBBI, dl, TII.get(ARM::t2MOVi16), ARM::R4)
          .addImm(NumWords)
          .setMIFlags(MachineInstr::FrameSetup)
          .add(predOps(ARMCC::AL));
    else
      BuildMI(MBB, MBBI, dl, TII.get(ARM::t2MOVi32imm), ARM::R4)
        .addImm(NumWords)
        .setMIFlags(MachineInstr::FrameSetup);

    switch (TM.getCodeModel()) {
    case CodeModel::Small:
    case CodeModel::Medium:
    case CodeModel::Kernel:
      BuildMI(MBB, MBBI, dl, TII.get(ARM::tBL))
          .add(predOps(ARMCC::AL))
          .addExternalSymbol("__chkstk")
          .addReg(ARM::R4, RegState::Implicit)
          .setMIFlags(MachineInstr::FrameSetup);
      break;
    case CodeModel::Large:
      BuildMI(MBB, MBBI, dl, TII.get(ARM::t2MOVi32imm), ARM::R12)
        .addExternalSymbol("__chkstk")
        .setMIFlags(MachineInstr::FrameSetup);

      BuildMI(MBB, MBBI, dl, TII.get(ARM::tBLXr))
          .add(predOps(ARMCC::AL))
          .addReg(ARM::R12, RegState::Kill)
          .addReg(ARM::R4, RegState::Implicit)
          .setMIFlags(MachineInstr::FrameSetup);
      break;
    }

    BuildMI(MBB, MBBI, dl, TII.get(ARM::t2SUBrr), ARM::SP)
        .addReg(ARM::SP, RegState::Kill)
        .addReg(ARM::R4, RegState::Kill)
        .setMIFlags(MachineInstr::FrameSetup)
        .add(predOps(ARMCC::AL))
        .add(condCodeOp());
    NumBytes = 0;
  }

  if (NumBytes) {
    // Adjust SP after all the callee-save spills.
    if (AFI->getNumAlignedDPRCS2Regs() == 0 &&
        tryFoldSPUpdateIntoPushPop(STI, MF, &*LastPush, NumBytes))
      DefCFAOffsetCandidates.addExtraBytes(LastPush, NumBytes);
    else {
      emitSPUpdate(isARM, MBB, MBBI, dl, TII, -NumBytes,
                   MachineInstr::FrameSetup);
      DefCFAOffsetCandidates.addInst(std::prev(MBBI), NumBytes);
    }

    if (HasFP && isARM)
      // Restore from fp only in ARM mode: e.g. sub sp, r7, #24
      // Note it's not safe to do this in Thumb2 mode because it would have
      // taken two instructions:
      // mov sp, r7
      // sub sp, #24
      // If an interrupt is taken between the two instructions, then sp is in
      // an inconsistent state (pointing to the middle of callee-saved area).
      // The interrupt handler can end up clobbering the registers.
      AFI->setShouldRestoreSPFromFP(true);
  }

  // Set FP to point to the stack slot that contains the previous FP.
  // For iOS, FP is R7, which has now been stored in spill area 1.
  // Otherwise, if this is not iOS, all the callee-saved registers go
  // into spill area 1, including the FP in R11.  In either case, it
  // is in area one and the adjustment needs to take place just after
  // that push.
  if (HasFP) {
    MachineBasicBlock::iterator AfterPush = std::next(GPRCS1Push);
    unsigned PushSize = sizeOfSPAdjustment(*GPRCS1Push);
    emitRegPlusImmediate(!AFI->isThumbFunction(), MBB, AfterPush,
                         dl, TII, FramePtr, ARM::SP,
                         PushSize + FramePtrOffsetInPush,
                         MachineInstr::FrameSetup);
    if (FramePtrOffsetInPush + PushSize != 0) {
      unsigned CFIIndex = MF.addFrameInst(MCCFIInstruction::createDefCfa(
          nullptr, MRI->getDwarfRegNum(FramePtr, true),
          -(ArgRegsSaveSize - FramePtrOffsetInPush)));
      BuildMI(MBB, AfterPush, dl, TII.get(TargetOpcode::CFI_INSTRUCTION))
          .addCFIIndex(CFIIndex)
          .setMIFlags(MachineInstr::FrameSetup);
    } else {
      unsigned CFIIndex =
          MF.addFrameInst(MCCFIInstruction::createDefCfaRegister(
              nullptr, MRI->getDwarfRegNum(FramePtr, true)));
      BuildMI(MBB, AfterPush, dl, TII.get(TargetOpcode::CFI_INSTRUCTION))
          .addCFIIndex(CFIIndex)
          .setMIFlags(MachineInstr::FrameSetup);
    }
  }

  // Now that the prologue's actual instructions are finalised, we can insert
  // the necessary DWARF cf instructions to describe the situation. Start by
  // recording where each register ended up:
  if (GPRCS1Size > 0) {
    MachineBasicBlock::iterator Pos = std::next(GPRCS1Push);
    int CFIIndex;
    for (const auto &Entry : CSI) {
      unsigned Reg = Entry.getReg();
      int FI = Entry.getFrameIdx();
      switch (Reg) {
      case ARM::R8:
      case ARM::R9:
      case ARM::R10:
      case ARM::R11:
      case ARM::R12:
        if (STI.splitFramePushPop(MF))
          break;
        LLVM_FALLTHROUGH;
      case ARM::R0:
      case ARM::R1:
      case ARM::R2:
      case ARM::R3:
      case ARM::R4:
      case ARM::R5:
      case ARM::R6:
      case ARM::R7:
      case ARM::LR:
        CFIIndex = MF.addFrameInst(MCCFIInstruction::createOffset(
            nullptr, MRI->getDwarfRegNum(Reg, true), MFI.getObjectOffset(FI)));
        BuildMI(MBB, Pos, dl, TII.get(TargetOpcode::CFI_INSTRUCTION))
            .addCFIIndex(CFIIndex)
            .setMIFlags(MachineInstr::FrameSetup);
        break;
      }
    }
  }

  if (GPRCS2Size > 0) {
    MachineBasicBlock::iterator Pos = std::next(GPRCS2Push);
    for (const auto &Entry : CSI) {
      unsigned Reg = Entry.getReg();
      int FI = Entry.getFrameIdx();
      switch (Reg) {
      case ARM::R8:
      case ARM::R9:
      case ARM::R10:
      case ARM::R11:
      case ARM::R12:
        if (STI.splitFramePushPop(MF)) {
          unsigned DwarfReg =  MRI->getDwarfRegNum(Reg, true);
          unsigned Offset = MFI.getObjectOffset(FI);
          unsigned CFIIndex = MF.addFrameInst(
              MCCFIInstruction::createOffset(nullptr, DwarfReg, Offset));
          BuildMI(MBB, Pos, dl, TII.get(TargetOpcode::CFI_INSTRUCTION))
              .addCFIIndex(CFIIndex)
              .setMIFlags(MachineInstr::FrameSetup);
        }
        break;
      }
    }
  }

  if (DPRCSSize > 0) {
    // Since vpush register list cannot have gaps, there may be multiple vpush
    // instructions in the prologue.
    MachineBasicBlock::iterator Pos = std::next(LastPush);
    for (const auto &Entry : CSI) {
      unsigned Reg = Entry.getReg();
      int FI = Entry.getFrameIdx();
      if ((Reg >= ARM::D0 && Reg <= ARM::D31) &&
          (Reg < ARM::D8 || Reg >= ARM::D8 + AFI->getNumAlignedDPRCS2Regs())) {
        unsigned DwarfReg = MRI->getDwarfRegNum(Reg, true);
        unsigned Offset = MFI.getObjectOffset(FI);
        unsigned CFIIndex = MF.addFrameInst(
            MCCFIInstruction::createOffset(nullptr, DwarfReg, Offset));
        BuildMI(MBB, Pos, dl, TII.get(TargetOpcode::CFI_INSTRUCTION))
            .addCFIIndex(CFIIndex)
            .setMIFlags(MachineInstr::FrameSetup);
      }
    }
  }

  // Now we can emit descriptions of where the canonical frame address was
  // throughout the process. If we have a frame pointer, it takes over the job
  // half-way through, so only the first few .cfi_def_cfa_offset instructions
  // actually get emitted.
  DefCFAOffsetCandidates.emitDefCFAOffsets(MBB, dl, TII, HasFP);

  if (STI.isTargetELF() && hasFP(MF))
    MFI.setOffsetAdjustment(MFI.getOffsetAdjustment() -
                            AFI->getFramePtrSpillOffset());

  AFI->setGPRCalleeSavedArea1Size(GPRCS1Size);
  AFI->setGPRCalleeSavedArea2Size(GPRCS2Size);
  AFI->setDPRCalleeSavedGapSize(DPRGapSize);
  AFI->setDPRCalleeSavedAreaSize(DPRCSSize);

  // If we need dynamic stack realignment, do it here. Be paranoid and make
  // sure if we also have VLAs, we have a base pointer for frame access.
  // If aligned NEON registers were spilled, the stack has already been
  // realigned.
  if (!AFI->getNumAlignedDPRCS2Regs() && RegInfo->needsStackRealignment(MF)) {
    unsigned MaxAlign = MFI.getMaxAlignment();
    assert(!AFI->isThumb1OnlyFunction());
    if (!AFI->isThumbFunction()) {
      emitAligningInstructions(MF, AFI, TII, MBB, MBBI, dl, ARM::SP, MaxAlign,
                               false);
    } else {
      // We cannot use sp as source/dest register here, thus we're using r4 to
      // perform the calculations. We're emitting the following sequence:
      // mov r4, sp
      // -- use emitAligningInstructions to produce best sequence to zero
      // -- out lower bits in r4
      // mov sp, r4
      // FIXME: It will be better just to find spare register here.
      BuildMI(MBB, MBBI, dl, TII.get(ARM::tMOVr), ARM::R4)
          .addReg(ARM::SP, RegState::Kill)
          .add(predOps(ARMCC::AL));
      emitAligningInstructions(MF, AFI, TII, MBB, MBBI, dl, ARM::R4, MaxAlign,
                               false);
      BuildMI(MBB, MBBI, dl, TII.get(ARM::tMOVr), ARM::SP)
          .addReg(ARM::R4, RegState::Kill)
          .add(predOps(ARMCC::AL));
    }

    AFI->setShouldRestoreSPFromFP(true);
  }

  // If we need a base pointer, set it up here. It's whatever the value
  // of the stack pointer is at this point. Any variable size objects
  // will be allocated after this, so we can still use the base pointer
  // to reference locals.
  // FIXME: Clarify FrameSetup flags here.
  if (RegInfo->hasBasePointer(MF)) {
    if (isARM)
      BuildMI(MBB, MBBI, dl, TII.get(ARM::MOVr), RegInfo->getBaseRegister())
          .addReg(ARM::SP)
          .add(predOps(ARMCC::AL))
          .add(condCodeOp());
    else
      BuildMI(MBB, MBBI, dl, TII.get(ARM::tMOVr), RegInfo->getBaseRegister())
          .addReg(ARM::SP)
          .add(predOps(ARMCC::AL));
  }

  // If the frame has variable sized objects then the epilogue must restore
  // the sp from fp. We can assume there's an FP here since hasFP already
  // checks for hasVarSizedObjects.
  if (MFI.hasVarSizedObjects())
    AFI->setShouldRestoreSPFromFP(true);
}

void ARMFrameLowering::emitEpilogue(MachineFunction &MF,
                                    MachineBasicBlock &MBB) const {
  MachineFrameInfo &MFI = MF.getFrameInfo();
  ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
  const TargetRegisterInfo *RegInfo = MF.getSubtarget().getRegisterInfo();
  const ARMBaseInstrInfo &TII =
      *static_cast<const ARMBaseInstrInfo *>(MF.getSubtarget().getInstrInfo());
  assert(!AFI->isThumb1OnlyFunction() &&
         "This emitEpilogue does not support Thumb1!");
  bool isARM = !AFI->isThumbFunction();

  unsigned ArgRegsSaveSize = AFI->getArgRegsSaveSize();
  int NumBytes = (int)MFI.getStackSize();
  unsigned FramePtr = RegInfo->getFrameRegister(MF);

  // All calls are tail calls in GHC calling conv, and functions have no
  // prologue/epilogue.
  if (MF.getFunction().getCallingConv() == CallingConv::GHC)
    return;

  // First put ourselves on the first (from top) terminator instructions.
  MachineBasicBlock::iterator MBBI = MBB.getFirstTerminator();
  DebugLoc dl = MBBI != MBB.end() ? MBBI->getDebugLoc() : DebugLoc();

  if (!AFI->hasStackFrame()) {
    if (NumBytes - ArgRegsSaveSize != 0)
      emitSPUpdate(isARM, MBB, MBBI, dl, TII, NumBytes - ArgRegsSaveSize);
  } else {
    // Unwind MBBI to point to first LDR / VLDRD.
    const MCPhysReg *CSRegs = RegInfo->getCalleeSavedRegs(&MF);
    if (MBBI != MBB.begin()) {
      do {
        --MBBI;
      } while (MBBI != MBB.begin() && isCSRestore(*MBBI, TII, CSRegs));
      if (!isCSRestore(*MBBI, TII, CSRegs))
        ++MBBI;
    }

    // Move SP to start of FP callee save spill area.
    NumBytes -= (ArgRegsSaveSize +
                 AFI->getGPRCalleeSavedArea1Size() +
                 AFI->getGPRCalleeSavedArea2Size() +
                 AFI->getDPRCalleeSavedGapSize() +
                 AFI->getDPRCalleeSavedAreaSize());

    // Reset SP based on frame pointer only if the stack frame extends beyond
    // frame pointer stack slot or target is ELF and the function has FP.
    if (AFI->shouldRestoreSPFromFP()) {
      NumBytes = AFI->getFramePtrSpillOffset() - NumBytes;
      if (NumBytes) {
        if (isARM)
          emitARMRegPlusImmediate(MBB, MBBI, dl, ARM::SP, FramePtr, -NumBytes,
                                  ARMCC::AL, 0, TII);
        else {
          // It's not possible to restore SP from FP in a single instruction.
          // For iOS, this looks like:
          // mov sp, r7
          // sub sp, #24
          // This is bad, if an interrupt is taken after the mov, sp is in an
          // inconsistent state.
          // Use the first callee-saved register as a scratch register.
          assert(!MFI.getPristineRegs(MF).test(ARM::R4) &&
                 "No scratch register to restore SP from FP!");
          emitT2RegPlusImmediate(MBB, MBBI, dl, ARM::R4, FramePtr, -NumBytes,
                                 ARMCC::AL, 0, TII);
          BuildMI(MBB, MBBI, dl, TII.get(ARM::tMOVr), ARM::SP)
              .addReg(ARM::R4)
              .add(predOps(ARMCC::AL));
        }
      } else {
        // Thumb2 or ARM.
        if (isARM)
          BuildMI(MBB, MBBI, dl, TII.get(ARM::MOVr), ARM::SP)
              .addReg(FramePtr)
              .add(predOps(ARMCC::AL))
              .add(condCodeOp());
        else
          BuildMI(MBB, MBBI, dl, TII.get(ARM::tMOVr), ARM::SP)
              .addReg(FramePtr)
              .add(predOps(ARMCC::AL));
      }
    } else if (NumBytes &&
               !tryFoldSPUpdateIntoPushPop(STI, MF, &*MBBI, NumBytes))
      emitSPUpdate(isARM, MBB, MBBI, dl, TII, NumBytes);

    // Increment past our save areas.
    if (MBBI != MBB.end() && AFI->getDPRCalleeSavedAreaSize()) {
      MBBI++;
      // Since vpop register list cannot have gaps, there may be multiple vpop
      // instructions in the epilogue.
      while (MBBI != MBB.end() && MBBI->getOpcode() == ARM::VLDMDIA_UPD)
        MBBI++;
    }
    if (AFI->getDPRCalleeSavedGapSize()) {
      assert(AFI->getDPRCalleeSavedGapSize() == 4 &&
             "unexpected DPR alignment gap");
      emitSPUpdate(isARM, MBB, MBBI, dl, TII, AFI->getDPRCalleeSavedGapSize());
    }

    if (AFI->getGPRCalleeSavedArea2Size()) MBBI++;
    if (AFI->getGPRCalleeSavedArea1Size()) MBBI++;
  }

  if (ArgRegsSaveSize)
    emitSPUpdate(isARM, MBB, MBBI, dl, TII, ArgRegsSaveSize);
}

/// getFrameIndexReference - Provide a base+offset reference to an FI slot for
/// debug info.  It's the same as what we use for resolving the code-gen
/// references for now.  FIXME: This can go wrong when references are
/// SP-relative and simple call frames aren't used.
int
ARMFrameLowering::getFrameIndexReference(const MachineFunction &MF, int FI,
                                         unsigned &FrameReg) const {
  return ResolveFrameIndexReference(MF, FI, FrameReg, 0);
}

int
ARMFrameLowering::ResolveFrameIndexReference(const MachineFunction &MF,
                                             int FI, unsigned &FrameReg,
                                             int SPAdj) const {
  const MachineFrameInfo &MFI = MF.getFrameInfo();
  const ARMBaseRegisterInfo *RegInfo = static_cast<const ARMBaseRegisterInfo *>(
      MF.getSubtarget().getRegisterInfo());
  const ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
  int Offset = MFI.getObjectOffset(FI) + MFI.getStackSize();
  int FPOffset = Offset - AFI->getFramePtrSpillOffset();
  bool isFixed = MFI.isFixedObjectIndex(FI);

  FrameReg = ARM::SP;
  Offset += SPAdj;

  // SP can move around if there are allocas.  We may also lose track of SP
  // when emergency spilling inside a non-reserved call frame setup.
  bool hasMovingSP = !hasReservedCallFrame(MF);

  // When dynamically realigning the stack, use the frame pointer for
  // parameters, and the stack/base pointer for locals.
  if (RegInfo->needsStackRealignment(MF)) {
    assert(hasFP(MF) && "dynamic stack realignment without a FP!");
    if (isFixed) {
      FrameReg = RegInfo->getFrameRegister(MF);
      Offset = FPOffset;
    } else if (hasMovingSP) {
      assert(RegInfo->hasBasePointer(MF) &&
             "VLAs and dynamic stack alignment, but missing base pointer!");
      FrameReg = RegInfo->getBaseRegister();
    }
    return Offset;
  }

  // If there is a frame pointer, use it when we can.
  if (hasFP(MF) && AFI->hasStackFrame()) {
    // Use frame pointer to reference fixed objects. Use it for locals if
    // there are VLAs (and thus the SP isn't reliable as a base).
    if (isFixed || (hasMovingSP && !RegInfo->hasBasePointer(MF))) {
      FrameReg = RegInfo->getFrameRegister(MF);
      return FPOffset;
    } else if (hasMovingSP) {
      assert(RegInfo->hasBasePointer(MF) && "missing base pointer!");
      if (AFI->isThumb2Function()) {
        // Try to use the frame pointer if we can, else use the base pointer
        // since it's available. This is handy for the emergency spill slot, in
        // particular.
        if (FPOffset >= -255 && FPOffset < 0) {
          FrameReg = RegInfo->getFrameRegister(MF);
          return FPOffset;
        }
      }
    } else if (AFI->isThumb2Function()) {
      // Use  add <rd>, sp, #<imm8>
      //      ldr <rd>, [sp, #<imm8>]
      // if at all possible to save space.
      if (Offset >= 0 && (Offset & 3) == 0 && Offset <= 1020)
        return Offset;
      // In Thumb2 mode, the negative offset is very limited. Try to avoid
      // out of range references. ldr <rt>,[<rn>, #-<imm8>]
      if (FPOffset >= -255 && FPOffset < 0) {
        FrameReg = RegInfo->getFrameRegister(MF);
        return FPOffset;
      }
    } else if (Offset > (FPOffset < 0 ? -FPOffset : FPOffset)) {
      // Otherwise, use SP or FP, whichever is closer to the stack slot.
      FrameReg = RegInfo->getFrameRegister(MF);
      return FPOffset;
    }
  }
  // Use the base pointer if we have one.
  if (RegInfo->hasBasePointer(MF))
    FrameReg = RegInfo->getBaseRegister();
  return Offset;
}

void ARMFrameLowering::emitPushInst(MachineBasicBlock &MBB,
                                    MachineBasicBlock::iterator MI,
                                    const std::vector<CalleeSavedInfo> &CSI,
                                    unsigned StmOpc, unsigned StrOpc,
                                    bool NoGap,
                                    bool(*Func)(unsigned, bool),
                                    unsigned NumAlignedDPRCS2Regs,
                                    unsigned MIFlags) const {
  MachineFunction &MF = *MBB.getParent();
  const TargetInstrInfo &TII = *MF.getSubtarget().getInstrInfo();
  const TargetRegisterInfo &TRI = *STI.getRegisterInfo();

  DebugLoc DL;

  using RegAndKill = std::pair<unsigned, bool>;

  SmallVector<RegAndKill, 4> Regs;
  unsigned i = CSI.size();
  while (i != 0) {
    unsigned LastReg = 0;
    for (; i != 0; --i) {
      unsigned Reg = CSI[i-1].getReg();
      if (!(Func)(Reg, STI.splitFramePushPop(MF))) continue;

      // D-registers in the aligned area DPRCS2 are NOT spilled here.
      if (Reg >= ARM::D8 && Reg < ARM::D8 + NumAlignedDPRCS2Regs)
        continue;

      const MachineRegisterInfo &MRI = MF.getRegInfo();
      bool isLiveIn = MRI.isLiveIn(Reg);
      if (!isLiveIn && !MRI.isReserved(Reg))
        MBB.addLiveIn(Reg);
      // If NoGap is true, push consecutive registers and then leave the rest
      // for other instructions. e.g.
      // vpush {d8, d10, d11} -> vpush {d8}, vpush {d10, d11}
      if (NoGap && LastReg && LastReg != Reg-1)
        break;
      LastReg = Reg;
      // Do not set a kill flag on values that are also marked as live-in. This
      // happens with the @llvm-returnaddress intrinsic and with arguments
      // passed in callee saved registers.
      // Omitting the kill flags is conservatively correct even if the live-in
      // is not used after all.
      Regs.push_back(std::make_pair(Reg, /*isKill=*/!isLiveIn));
    }

    if (Regs.empty())
      continue;

    std::sort(Regs.begin(), Regs.end(), [&](const RegAndKill &LHS,
                                            const RegAndKill &RHS) {
      return TRI.getEncodingValue(LHS.first) < TRI.getEncodingValue(RHS.first);
    });

    if (Regs.size() > 1 || StrOpc== 0) {
      MachineInstrBuilder MIB = BuildMI(MBB, MI, DL, TII.get(StmOpc), ARM::SP)
                                    .addReg(ARM::SP)
                                    .setMIFlags(MIFlags)
                                    .add(predOps(ARMCC::AL));
      for (unsigned i = 0, e = Regs.size(); i < e; ++i)
        MIB.addReg(Regs[i].first, getKillRegState(Regs[i].second));
    } else if (Regs.size() == 1) {
      BuildMI(MBB, MI, DL, TII.get(StrOpc), ARM::SP)
          .addReg(Regs[0].first, getKillRegState(Regs[0].second))
          .addReg(ARM::SP)
          .setMIFlags(MIFlags)
          .addImm(-4)
          .add(predOps(ARMCC::AL));
    }
    Regs.clear();

    // Put any subsequent vpush instructions before this one: they will refer to
    // higher register numbers so need to be pushed first in order to preserve
    // monotonicity.
    if (MI != MBB.begin())
      --MI;
  }
}

void ARMFrameLowering::emitPopInst(MachineBasicBlock &MBB,
                                   MachineBasicBlock::iterator MI,
                                   std::vector<CalleeSavedInfo> &CSI,
                                   unsigned LdmOpc, unsigned LdrOpc,
                                   bool isVarArg, bool NoGap,
                                   bool(*Func)(unsigned, bool),
                                   unsigned NumAlignedDPRCS2Regs) const {
  MachineFunction &MF = *MBB.getParent();
  const TargetInstrInfo &TII = *MF.getSubtarget().getInstrInfo();
  const TargetRegisterInfo &TRI = *STI.getRegisterInfo();
  ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
  DebugLoc DL;
  bool isTailCall = false;
  bool isInterrupt = false;
  bool isTrap = false;
  if (MBB.end() != MI) {
    DL = MI->getDebugLoc();
    unsigned RetOpcode = MI->getOpcode();
    isTailCall = (RetOpcode == ARM::TCRETURNdi || RetOpcode == ARM::TCRETURNri);
    isInterrupt =
        RetOpcode == ARM::SUBS_PC_LR || RetOpcode == ARM::t2SUBS_PC_LR;
    isTrap =
        RetOpcode == ARM::TRAP || RetOpcode == ARM::TRAPNaCl ||
        RetOpcode == ARM::tTRAP;
  }

  SmallVector<unsigned, 4> Regs;
  unsigned i = CSI.size();
  while (i != 0) {
    unsigned LastReg = 0;
    bool DeleteRet = false;
    for (; i != 0; --i) {
      CalleeSavedInfo &Info = CSI[i-1];
      unsigned Reg = Info.getReg();
      if (!(Func)(Reg, STI.splitFramePushPop(MF))) continue;

      // The aligned reloads from area DPRCS2 are not inserted here.
      if (Reg >= ARM::D8 && Reg < ARM::D8 + NumAlignedDPRCS2Regs)
        continue;

      if (Reg == ARM::LR && !isTailCall && !isVarArg && !isInterrupt &&
          !isTrap && STI.hasV5TOps()) {
        if (MBB.succ_empty()) {
          Reg = ARM::PC;
          DeleteRet = true;
          LdmOpc = AFI->isThumbFunction() ? ARM::t2LDMIA_RET : ARM::LDMIA_RET;
          // We 'restore' LR into PC so it is not live out of the return block:
          // Clear Restored bit.
          Info.setRestored(false);
        } else
          LdmOpc = AFI->isThumbFunction() ? ARM::t2LDMIA_UPD : ARM::LDMIA_UPD;
        // Fold the return instruction into the LDM.
      }

      // If NoGap is true, pop consecutive registers and then leave the rest
      // for other instructions. e.g.
      // vpop {d8, d10, d11} -> vpop {d8}, vpop {d10, d11}
      if (NoGap && LastReg && LastReg != Reg-1)
        break;

      LastReg = Reg;
      Regs.push_back(Reg);
    }

    if (Regs.empty())
      continue;

    std::sort(Regs.begin(), Regs.end(), [&](unsigned LHS, unsigned RHS) {
      return TRI.getEncodingValue(LHS) < TRI.getEncodingValue(RHS);
    });

    if (Regs.size() > 1 || LdrOpc == 0) {
      MachineInstrBuilder MIB = BuildMI(MBB, MI, DL, TII.get(LdmOpc), ARM::SP)
                                    .addReg(ARM::SP)
                                    .add(predOps(ARMCC::AL));
      for (unsigned i = 0, e = Regs.size(); i < e; ++i)
        MIB.addReg(Regs[i], getDefRegState(true));
      if (DeleteRet) {
        if (MI != MBB.end()) {
          MIB.copyImplicitOps(*MI);
          MI->eraseFromParent();
        }
      }
      MI = MIB;
    } else if (Regs.size() == 1) {
      // If we adjusted the reg to PC from LR above, switch it back here. We
      // only do that for LDM.
      if (Regs[0] == ARM::PC)
        Regs[0] = ARM::LR;
      MachineInstrBuilder MIB =
        BuildMI(MBB, MI, DL, TII.get(LdrOpc), Regs[0])
          .addReg(ARM::SP, RegState::Define)
          .addReg(ARM::SP);
      // ARM mode needs an extra reg0 here due to addrmode2. Will go away once
      // that refactoring is complete (eventually).
      if (LdrOpc == ARM::LDR_POST_REG || LdrOpc == ARM::LDR_POST_IMM) {
        MIB.addReg(0);
        MIB.addImm(ARM_AM::getAM2Opc(ARM_AM::add, 4, ARM_AM::no_shift));
      } else
        MIB.addImm(4);
      MIB.add(predOps(ARMCC::AL));
    }
    Regs.clear();

    // Put any subsequent vpop instructions after this one: they will refer to
    // higher register numbers so need to be popped afterwards.
    if (MI != MBB.end())
      ++MI;
  }
}

/// Emit aligned spill instructions for NumAlignedDPRCS2Regs D-registers
/// starting from d8.  Also insert stack realignment code and leave the stack
/// pointer pointing to the d8 spill slot.
static void emitAlignedDPRCS2Spills(MachineBasicBlock &MBB,
                                    MachineBasicBlock::iterator MI,
                                    unsigned NumAlignedDPRCS2Regs,
                                    const std::vector<CalleeSavedInfo> &CSI,
                                    const TargetRegisterInfo *TRI) {
  MachineFunction &MF = *MBB.getParent();
  ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
  DebugLoc DL = MI != MBB.end() ? MI->getDebugLoc() : DebugLoc();
  const TargetInstrInfo &TII = *MF.getSubtarget().getInstrInfo();
  MachineFrameInfo &MFI = MF.getFrameInfo();

  // Mark the D-register spill slots as properly aligned.  Since MFI computes
  // stack slot layout backwards, this can actually mean that the d-reg stack
  // slot offsets can be wrong. The offset for d8 will always be correct.
  for (unsigned i = 0, e = CSI.size(); i != e; ++i) {
    unsigned DNum = CSI[i].getReg() - ARM::D8;
    if (DNum > NumAlignedDPRCS2Regs - 1)
      continue;
    int FI = CSI[i].getFrameIdx();
    // The even-numbered registers will be 16-byte aligned, the odd-numbered
    // registers will be 8-byte aligned.
    MFI.setObjectAlignment(FI, DNum % 2 ? 8 : 16);

    // The stack slot for D8 needs to be maximally aligned because this is
    // actually the point where we align the stack pointer.  MachineFrameInfo
    // computes all offsets relative to the incoming stack pointer which is a
    // bit weird when realigning the stack.  Any extra padding for this
    // over-alignment is not realized because the code inserted below adjusts
    // the stack pointer by numregs * 8 before aligning the stack pointer.
    if (DNum == 0)
      MFI.setObjectAlignment(FI, MFI.getMaxAlignment());
  }

  // Move the stack pointer to the d8 spill slot, and align it at the same
  // time. Leave the stack slot address in the scratch register r4.
  //
  //   sub r4, sp, #numregs * 8
  //   bic r4, r4, #align - 1
  //   mov sp, r4
  //
  bool isThumb = AFI->isThumbFunction();
  assert(!AFI->isThumb1OnlyFunction() && "Can't realign stack for thumb1");
  AFI->setShouldRestoreSPFromFP(true);

  // sub r4, sp, #numregs * 8
  // The immediate is <= 64, so it doesn't need any special encoding.
  unsigned Opc = isThumb ? ARM::t2SUBri : ARM::SUBri;
  BuildMI(MBB, MI, DL, TII.get(Opc), ARM::R4)
      .addReg(ARM::SP)
      .addImm(8 * NumAlignedDPRCS2Regs)
      .add(predOps(ARMCC::AL))
      .add(condCodeOp());

  unsigned MaxAlign = MF.getFrameInfo().getMaxAlignment();
  // We must set parameter MustBeSingleInstruction to true, since
  // skipAlignedDPRCS2Spills expects exactly 3 instructions to perform
  // stack alignment.  Luckily, this can always be done since all ARM
  // architecture versions that support Neon also support the BFC
  // instruction.
  emitAligningInstructions(MF, AFI, TII, MBB, MI, DL, ARM::R4, MaxAlign, true);

  // mov sp, r4
  // The stack pointer must be adjusted before spilling anything, otherwise
  // the stack slots could be clobbered by an interrupt handler.
  // Leave r4 live, it is used below.
  Opc = isThumb ? ARM::tMOVr : ARM::MOVr;
  MachineInstrBuilder MIB = BuildMI(MBB, MI, DL, TII.get(Opc), ARM::SP)
                                .addReg(ARM::R4)
                                .add(predOps(ARMCC::AL));
  if (!isThumb)
    MIB.add(condCodeOp());

  // Now spill NumAlignedDPRCS2Regs registers starting from d8.
  // r4 holds the stack slot address.
  unsigned NextReg = ARM::D8;

  // 16-byte aligned vst1.64 with 4 d-regs and address writeback.
  // The writeback is only needed when emitting two vst1.64 instructions.
  if (NumAlignedDPRCS2Regs >= 6) {
    unsigned SupReg = TRI->getMatchingSuperReg(NextReg, ARM::dsub_0,
                                               &ARM::QQPRRegClass);
    MBB.addLiveIn(SupReg);
    BuildMI(MBB, MI, DL, TII.get(ARM::VST1d64Qwb_fixed), ARM::R4)
        .addReg(ARM::R4, RegState::Kill)
        .addImm(16)
        .addReg(NextReg)
        .addReg(SupReg, RegState::ImplicitKill)
        .add(predOps(ARMCC::AL));
    NextReg += 4;
    NumAlignedDPRCS2Regs -= 4;
  }

  // We won't modify r4 beyond this point.  It currently points to the next
  // register to be spilled.
  unsigned R4BaseReg = NextReg;

  // 16-byte aligned vst1.64 with 4 d-regs, no writeback.
  if (NumAlignedDPRCS2Regs >= 4) {
    unsigned SupReg = TRI->getMatchingSuperReg(NextReg, ARM::dsub_0,
                                               &ARM::QQPRRegClass);
    MBB.addLiveIn(SupReg);
    BuildMI(MBB, MI, DL, TII.get(ARM::VST1d64Q))
        .addReg(ARM::R4)
        .addImm(16)
        .addReg(NextReg)
        .addReg(SupReg, RegState::ImplicitKill)
        .add(predOps(ARMCC::AL));
    NextReg += 4;
    NumAlignedDPRCS2Regs -= 4;
  }

  // 16-byte aligned vst1.64 with 2 d-regs.
  if (NumAlignedDPRCS2Regs >= 2) {
    unsigned SupReg = TRI->getMatchingSuperReg(NextReg, ARM::dsub_0,
                                               &ARM::QPRRegClass);
    MBB.addLiveIn(SupReg);
    BuildMI(MBB, MI, DL, TII.get(ARM::VST1q64))
        .addReg(ARM::R4)
        .addImm(16)
        .addReg(SupReg)
        .add(predOps(ARMCC::AL));
    NextReg += 2;
    NumAlignedDPRCS2Regs -= 2;
  }

  // Finally, use a vanilla vstr.64 for the odd last register.
  if (NumAlignedDPRCS2Regs) {
    MBB.addLiveIn(NextReg);
    // vstr.64 uses addrmode5 which has an offset scale of 4.
    BuildMI(MBB, MI, DL, TII.get(ARM::VSTRD))
        .addReg(NextReg)
        .addReg(ARM::R4)
        .addImm((NextReg - R4BaseReg) * 2)
        .add(predOps(ARMCC::AL));
  }

  // The last spill instruction inserted should kill the scratch register r4.
  std::prev(MI)->addRegisterKilled(ARM::R4, TRI);
}

/// Skip past the code inserted by emitAlignedDPRCS2Spills, and return an
/// iterator to the following instruction.
static MachineBasicBlock::iterator
skipAlignedDPRCS2Spills(MachineBasicBlock::iterator MI,
                        unsigned NumAlignedDPRCS2Regs) {
  //   sub r4, sp, #numregs * 8
  //   bic r4, r4, #align - 1
  //   mov sp, r4
  ++MI; ++MI; ++MI;
  assert(MI->mayStore() && "Expecting spill instruction");

  // These switches all fall through.
  switch(NumAlignedDPRCS2Regs) {
  case 7:
    ++MI;
    assert(MI->mayStore() && "Expecting spill instruction");
    LLVM_FALLTHROUGH;
  default:
    ++MI;
    assert(MI->mayStore() && "Expecting spill instruction");
    LLVM_FALLTHROUGH;
  case 1:
  case 2:
  case 4:
    assert(MI->killsRegister(ARM::R4) && "Missed kill flag");
    ++MI;
  }
  return MI;
}

/// Emit aligned reload instructions for NumAlignedDPRCS2Regs D-registers
/// starting from d8.  These instructions are assumed to execute while the
/// stack is still aligned, unlike the code inserted by emitPopInst.
static void emitAlignedDPRCS2Restores(MachineBasicBlock &MBB,
                                      MachineBasicBlock::iterator MI,
                                      unsigned NumAlignedDPRCS2Regs,
                                      const std::vector<CalleeSavedInfo> &CSI,
                                      const TargetRegisterInfo *TRI) {
  MachineFunction &MF = *MBB.getParent();
  ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
  DebugLoc DL = MI != MBB.end() ? MI->getDebugLoc() : DebugLoc();
  const TargetInstrInfo &TII = *MF.getSubtarget().getInstrInfo();

  // Find the frame index assigned to d8.
  int D8SpillFI = 0;
  for (unsigned i = 0, e = CSI.size(); i != e; ++i)
    if (CSI[i].getReg() == ARM::D8) {
      D8SpillFI = CSI[i].getFrameIdx();
      break;
    }

  // Materialize the address of the d8 spill slot into the scratch register r4.
  // This can be fairly complicated if the stack frame is large, so just use
  // the normal frame index elimination mechanism to do it.  This code runs as
  // the initial part of the epilog where the stack and base pointers haven't
  // been changed yet.
  bool isThumb = AFI->isThumbFunction();
  assert(!AFI->isThumb1OnlyFunction() && "Can't realign stack for thumb1");

  unsigned Opc = isThumb ? ARM::t2ADDri : ARM::ADDri;
  BuildMI(MBB, MI, DL, TII.get(Opc), ARM::R4)
      .addFrameIndex(D8SpillFI)
      .addImm(0)
      .add(predOps(ARMCC::AL))
      .add(condCodeOp());

  // Now restore NumAlignedDPRCS2Regs registers starting from d8.
  unsigned NextReg = ARM::D8;

  // 16-byte aligned vld1.64 with 4 d-regs and writeback.
  if (NumAlignedDPRCS2Regs >= 6) {
    unsigned SupReg = TRI->getMatchingSuperReg(NextReg, ARM::dsub_0,
                                               &ARM::QQPRRegClass);
    BuildMI(MBB, MI, DL, TII.get(ARM::VLD1d64Qwb_fixed), NextReg)
        .addReg(ARM::R4, RegState::Define)
        .addReg(ARM::R4, RegState::Kill)
        .addImm(16)
        .addReg(SupReg, RegState::ImplicitDefine)
        .add(predOps(ARMCC::AL));
    NextReg += 4;
    NumAlignedDPRCS2Regs -= 4;
  }

  // We won't modify r4 beyond this point.  It currently points to the next
  // register to be spilled.
  unsigned R4BaseReg = NextReg;

  // 16-byte aligned vld1.64 with 4 d-regs, no writeback.
  if (NumAlignedDPRCS2Regs >= 4) {
    unsigned SupReg = TRI->getMatchingSuperReg(NextReg, ARM::dsub_0,
                                               &ARM::QQPRRegClass);
    BuildMI(MBB, MI, DL, TII.get(ARM::VLD1d64Q), NextReg)
        .addReg(ARM::R4)
        .addImm(16)
        .addReg(SupReg, RegState::ImplicitDefine)
        .add(predOps(ARMCC::AL));
    NextReg += 4;
    NumAlignedDPRCS2Regs -= 4;
  }

  // 16-byte aligned vld1.64 with 2 d-regs.
  if (NumAlignedDPRCS2Regs >= 2) {
    unsigned SupReg = TRI->getMatchingSuperReg(NextReg, ARM::dsub_0,
                                               &ARM::QPRRegClass);
    BuildMI(MBB, MI, DL, TII.get(ARM::VLD1q64), SupReg)
        .addReg(ARM::R4)
        .addImm(16)
        .add(predOps(ARMCC::AL));
    NextReg += 2;
    NumAlignedDPRCS2Regs -= 2;
  }

  // Finally, use a vanilla vldr.64 for the remaining odd register.
  if (NumAlignedDPRCS2Regs)
    BuildMI(MBB, MI, DL, TII.get(ARM::VLDRD), NextReg)
        .addReg(ARM::R4)
        .addImm(2 * (NextReg - R4BaseReg))
        .add(predOps(ARMCC::AL));

  // Last store kills r4.
  std::prev(MI)->addRegisterKilled(ARM::R4, TRI);
}

bool ARMFrameLowering::spillCalleeSavedRegisters(MachineBasicBlock &MBB,
                                        MachineBasicBlock::iterator MI,
                                        const std::vector<CalleeSavedInfo> &CSI,
                                        const TargetRegisterInfo *TRI) const {
  if (CSI.empty())
    return false;

  MachineFunction &MF = *MBB.getParent();
  ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();

  unsigned PushOpc = AFI->isThumbFunction() ? ARM::t2STMDB_UPD : ARM::STMDB_UPD;
  unsigned PushOneOpc = AFI->isThumbFunction() ?
    ARM::t2STR_PRE : ARM::STR_PRE_IMM;
  unsigned FltOpc = ARM::VSTMDDB_UPD;
  unsigned NumAlignedDPRCS2Regs = AFI->getNumAlignedDPRCS2Regs();
  emitPushInst(MBB, MI, CSI, PushOpc, PushOneOpc, false, &isARMArea1Register, 0,
               MachineInstr::FrameSetup);
  emitPushInst(MBB, MI, CSI, PushOpc, PushOneOpc, false, &isARMArea2Register, 0,
               MachineInstr::FrameSetup);
  emitPushInst(MBB, MI, CSI, FltOpc, 0, true, &isARMArea3Register,
               NumAlignedDPRCS2Regs, MachineInstr::FrameSetup);

  // The code above does not insert spill code for the aligned DPRCS2 registers.
  // The stack realignment code will be inserted between the push instructions
  // and these spills.
  if (NumAlignedDPRCS2Regs)
    emitAlignedDPRCS2Spills(MBB, MI, NumAlignedDPRCS2Regs, CSI, TRI);

  return true;
}

bool ARMFrameLowering::restoreCalleeSavedRegisters(MachineBasicBlock &MBB,
                                        MachineBasicBlock::iterator MI,
                                        std::vector<CalleeSavedInfo> &CSI,
                                        const TargetRegisterInfo *TRI) const {
  if (CSI.empty())
    return false;

  MachineFunction &MF = *MBB.getParent();
  ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
  bool isVarArg = AFI->getArgRegsSaveSize() > 0;
  unsigned NumAlignedDPRCS2Regs = AFI->getNumAlignedDPRCS2Regs();

  // The emitPopInst calls below do not insert reloads for the aligned DPRCS2
  // registers. Do that here instead.
  if (NumAlignedDPRCS2Regs)
    emitAlignedDPRCS2Restores(MBB, MI, NumAlignedDPRCS2Regs, CSI, TRI);

  unsigned PopOpc = AFI->isThumbFunction() ? ARM::t2LDMIA_UPD : ARM::LDMIA_UPD;
  unsigned LdrOpc = AFI->isThumbFunction() ? ARM::t2LDR_POST :ARM::LDR_POST_IMM;
  unsigned FltOpc = ARM::VLDMDIA_UPD;
  emitPopInst(MBB, MI, CSI, FltOpc, 0, isVarArg, true, &isARMArea3Register,
              NumAlignedDPRCS2Regs);
  emitPopInst(MBB, MI, CSI, PopOpc, LdrOpc, isVarArg, false,
              &isARMArea2Register, 0);
  emitPopInst(MBB, MI, CSI, PopOpc, LdrOpc, isVarArg, false,
              &isARMArea1Register, 0);

  return true;
}

// FIXME: Make generic?
static unsigned GetFunctionSizeInBytes(const MachineFunction &MF,
                                       const ARMBaseInstrInfo &TII) {
  unsigned FnSize = 0;
  for (auto &MBB : MF) {
    for (auto &MI : MBB)
      FnSize += TII.getInstSizeInBytes(MI);
  }
  return FnSize;
}

/// estimateRSStackSizeLimit - Look at each instruction that references stack
/// frames and return the stack size limit beyond which some of these
/// instructions will require a scratch register during their expansion later.
// FIXME: Move to TII?
static unsigned estimateRSStackSizeLimit(MachineFunction &MF,
                                         const TargetFrameLowering *TFI) {
  const ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
  unsigned Limit = (1 << 12) - 1;
  for (auto &MBB : MF) {
    for (auto &MI : MBB) {
      for (unsigned i = 0, e = MI.getNumOperands(); i != e; ++i) {
        if (!MI.getOperand(i).isFI())
          continue;

        // When using ADDri to get the address of a stack object, 255 is the
        // largest offset guaranteed to fit in the immediate offset.
        if (MI.getOpcode() == ARM::ADDri) {
          Limit = std::min(Limit, (1U << 8) - 1);
          break;
        }

        // Otherwise check the addressing mode.
        switch (MI.getDesc().TSFlags & ARMII::AddrModeMask) {
        case ARMII::AddrMode3:
        case ARMII::AddrModeT2_i8:
          Limit = std::min(Limit, (1U << 8) - 1);
          break;
        case ARMII::AddrMode5:
        case ARMII::AddrModeT2_i8s4:
          Limit = std::min(Limit, ((1U << 8) - 1) * 4);
          break;
        case ARMII::AddrModeT2_i12:
          // i12 supports only positive offset so these will be converted to
          // i8 opcodes. See llvm::rewriteT2FrameIndex.
          if (TFI->hasFP(MF) && AFI->hasStackFrame())
            Limit = std::min(Limit, (1U << 8) - 1);
          break;
        case ARMII::AddrMode4:
        case ARMII::AddrMode6:
          // Addressing modes 4 & 6 (load/store) instructions can't encode an
          // immediate offset for stack references.
          return 0;
        default:
          break;
        }
        break; // At most one FI per instruction
      }
    }
  }

  return Limit;
}

// In functions that realign the stack, it can be an advantage to spill the
// callee-saved vector registers after realigning the stack. The vst1 and vld1
// instructions take alignment hints that can improve performance.
static void
checkNumAlignedDPRCS2Regs(MachineFunction &MF, BitVector &SavedRegs) {
  MF.getInfo<ARMFunctionInfo>()->setNumAlignedDPRCS2Regs(0);
  if (!SpillAlignedNEONRegs)
    return;

  // Naked functions don't spill callee-saved registers.
  if (MF.getFunction().hasFnAttribute(Attribute::Naked))
    return;

  // We are planning to use NEON instructions vst1 / vld1.
  if (!static_cast<const ARMSubtarget &>(MF.getSubtarget()).hasNEON())
    return;

  // Don't bother if the default stack alignment is sufficiently high.
  if (MF.getSubtarget().getFrameLowering()->getStackAlignment() >= 8)
    return;

  // Aligned spills require stack realignment.
  if (!static_cast<const ARMBaseRegisterInfo *>(
           MF.getSubtarget().getRegisterInfo())->canRealignStack(MF))
    return;

  // We always spill contiguous d-registers starting from d8. Count how many
  // needs spilling.  The register allocator will almost always use the
  // callee-saved registers in order, but it can happen that there are holes in
  // the range.  Registers above the hole will be spilled to the standard DPRCS
  // area.
  unsigned NumSpills = 0;
  for (; NumSpills < 8; ++NumSpills)
    if (!SavedRegs.test(ARM::D8 + NumSpills))
      break;

  // Don't do this for just one d-register. It's not worth it.
  if (NumSpills < 2)
    return;

  // Spill the first NumSpills D-registers after realigning the stack.
  MF.getInfo<ARMFunctionInfo>()->setNumAlignedDPRCS2Regs(NumSpills);

  // A scratch register is required for the vst1 / vld1 instructions.
  SavedRegs.set(ARM::R4);
}

void ARMFrameLowering::determineCalleeSaves(MachineFunction &MF,
                                            BitVector &SavedRegs,
                                            RegScavenger *RS) const {
  TargetFrameLowering::determineCalleeSaves(MF, SavedRegs, RS);
  // This tells PEI to spill the FP as if it is any other callee-save register
  // to take advantage the eliminateFrameIndex machinery. This also ensures it
  // is spilled in the order specified by getCalleeSavedRegs() to make it easier
  // to combine multiple loads / stores.
  bool CanEliminateFrame = true;
  bool CS1Spilled = false;
  bool LRSpilled = false;
  unsigned NumGPRSpills = 0;
  unsigned NumFPRSpills = 0;
  SmallVector<unsigned, 4> UnspilledCS1GPRs;
  SmallVector<unsigned, 4> UnspilledCS2GPRs;
  const ARMBaseRegisterInfo *RegInfo = static_cast<const ARMBaseRegisterInfo *>(
      MF.getSubtarget().getRegisterInfo());
  const ARMBaseInstrInfo &TII =
      *static_cast<const ARMBaseInstrInfo *>(MF.getSubtarget().getInstrInfo());
  ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
  MachineFrameInfo &MFI = MF.getFrameInfo();
  MachineRegisterInfo &MRI = MF.getRegInfo();
  const TargetRegisterInfo *TRI = MF.getSubtarget().getRegisterInfo();
  (void)TRI;  // Silence unused warning in non-assert builds.
  unsigned FramePtr = RegInfo->getFrameRegister(MF);

  // Spill R4 if Thumb2 function requires stack realignment - it will be used as
  // scratch register. Also spill R4 if Thumb2 function has varsized objects,
  // since it's not always possible to restore sp from fp in a single
  // instruction.
  // FIXME: It will be better just to find spare register here.
  if (AFI->isThumb2Function() &&
      (MFI.hasVarSizedObjects() || RegInfo->needsStackRealignment(MF)))
    SavedRegs.set(ARM::R4);

  if (AFI->isThumb1OnlyFunction()) {
    // Spill LR if Thumb1 function uses variable length argument lists.
    if (AFI->getArgRegsSaveSize() > 0)
      SavedRegs.set(ARM::LR);

    // Spill R4 if Thumb1 epilogue has to restore SP from FP or the function
    // requires stack alignment.  We don't know for sure what the stack size
    // will be, but for this, an estimate is good enough. If there anything
    // changes it, it'll be a spill, which implies we've used all the registers
    // and so R4 is already used, so not marking it here will be OK.
    // FIXME: It will be better just to find spare register here.
    if (MFI.hasVarSizedObjects() || RegInfo->needsStackRealignment(MF) ||
        MFI.estimateStackSize(MF) > 508)
      SavedRegs.set(ARM::R4);
  }

  // See if we can spill vector registers to aligned stack.
  checkNumAlignedDPRCS2Regs(MF, SavedRegs);

  // Spill the BasePtr if it's used.
  if (RegInfo->hasBasePointer(MF))
    SavedRegs.set(RegInfo->getBaseRegister());

  // Don't spill FP if the frame can be eliminated. This is determined
  // by scanning the callee-save registers to see if any is modified.
  const MCPhysReg *CSRegs = RegInfo->getCalleeSavedRegs(&MF);
  for (unsigned i = 0; CSRegs[i]; ++i) {
    unsigned Reg = CSRegs[i];
    bool Spilled = false;
    if (SavedRegs.test(Reg)) {
      Spilled = true;
      CanEliminateFrame = false;
    }

    if (!ARM::GPRRegClass.contains(Reg)) {
      if (Spilled) {
        if (ARM::SPRRegClass.contains(Reg))
          NumFPRSpills++;
        else if (ARM::DPRRegClass.contains(Reg))
          NumFPRSpills += 2;
        else if (ARM::QPRRegClass.contains(Reg))
          NumFPRSpills += 4;
      }
      continue;
    }

    if (Spilled) {
      NumGPRSpills++;

      if (!STI.splitFramePushPop(MF)) {
        if (Reg == ARM::LR)
          LRSpilled = true;
        CS1Spilled = true;
        continue;
      }

      // Keep track if LR and any of R4, R5, R6, and R7 is spilled.
      switch (Reg) {
      case ARM::LR:
        LRSpilled = true;
        LLVM_FALLTHROUGH;
      case ARM::R0: case ARM::R1:
      case ARM::R2: case ARM::R3:
      case ARM::R4: case ARM::R5:
      case ARM::R6: case ARM::R7:
        CS1Spilled = true;
        break;
      default:
        break;
      }
    } else {
      if (!STI.splitFramePushPop(MF)) {
        UnspilledCS1GPRs.push_back(Reg);
        continue;
      }

      switch (Reg) {
      case ARM::R0: case ARM::R1:
      case ARM::R2: case ARM::R3:
      case ARM::R4: case ARM::R5:
      case ARM::R6: case ARM::R7:
      case ARM::LR:
        UnspilledCS1GPRs.push_back(Reg);
        break;
      default:
        UnspilledCS2GPRs.push_back(Reg);
        break;
      }
    }
  }

  bool ForceLRSpill = false;
  if (!LRSpilled && AFI->isThumb1OnlyFunction()) {
    unsigned FnSize = GetFunctionSizeInBytes(MF, TII);
    // Force LR to be spilled if the Thumb function size is > 2048. This enables
    // use of BL to implement far jump. If it turns out that it's not needed
    // then the branch fix up path will undo it.
    if (FnSize >= (1 << 11)) {
      CanEliminateFrame = false;
      ForceLRSpill = true;
    }
  }

  // If any of the stack slot references may be out of range of an immediate
  // offset, make sure a register (or a spill slot) is available for the
  // register scavenger. Note that if we're indexing off the frame pointer, the
  // effective stack size is 4 bytes larger since the FP points to the stack
  // slot of the previous FP. Also, if we have variable sized objects in the
  // function, stack slot references will often be negative, and some of
  // our instructions are positive-offset only, so conservatively consider
  // that case to want a spill slot (or register) as well. Similarly, if
  // the function adjusts the stack pointer during execution and the
  // adjustments aren't already part of our stack size estimate, our offset
  // calculations may be off, so be conservative.
  // FIXME: We could add logic to be more precise about negative offsets
  //        and which instructions will need a scratch register for them. Is it
  //        worth the effort and added fragility?
  unsigned EstimatedStackSize =
      MFI.estimateStackSize(MF) + 4 * (NumGPRSpills + NumFPRSpills);

  // Determine biggest (positive) SP offset in MachineFrameInfo.
  int MaxFixedOffset = 0;
  for (int I = MFI.getObjectIndexBegin(); I < 0; ++I) {
    int MaxObjectOffset = MFI.getObjectOffset(I) + MFI.getObjectSize(I);
    MaxFixedOffset = std::max(MaxFixedOffset, MaxObjectOffset);
  }

  bool HasFP = hasFP(MF);
  if (HasFP) {
    if (AFI->hasStackFrame())
      EstimatedStackSize += 4;
  } else {
    // If FP is not used, SP will be used to access arguments, so count the
    // size of arguments into the estimation.
    EstimatedStackSize += MaxFixedOffset;
  }
  EstimatedStackSize += 16; // For possible paddings.

  unsigned EstimatedRSStackSizeLimit = estimateRSStackSizeLimit(MF, this);
  int MaxFPOffset = getMaxFPOffset(MF.getFunction(), *AFI);
  bool BigFrameOffsets = EstimatedStackSize >= EstimatedRSStackSizeLimit ||
    MFI.hasVarSizedObjects() ||
    (MFI.adjustsStack() && !canSimplifyCallFramePseudos(MF)) ||
    // For large argument stacks fp relative addressed may overflow.
    (HasFP && (MaxFixedOffset - MaxFPOffset) >= (int)EstimatedRSStackSizeLimit);
  if (BigFrameOffsets ||
      !CanEliminateFrame || RegInfo->cannotEliminateFrame(MF)) {
    AFI->setHasStackFrame(true);

    if (HasFP) {
      SavedRegs.set(FramePtr);
      // If the frame pointer is required by the ABI, also spill LR so that we
      // emit a complete frame record.
      if (MF.getTarget().Options.DisableFramePointerElim(MF) && !LRSpilled) {
        SavedRegs.set(ARM::LR);
        LRSpilled = true;
        NumGPRSpills++;
        auto LRPos = llvm::find(UnspilledCS1GPRs, ARM::LR);
        if (LRPos != UnspilledCS1GPRs.end())
          UnspilledCS1GPRs.erase(LRPos);
      }
      auto FPPos = llvm::find(UnspilledCS1GPRs, FramePtr);
      if (FPPos != UnspilledCS1GPRs.end())
        UnspilledCS1GPRs.erase(FPPos);
      NumGPRSpills++;
      if (FramePtr == ARM::R7)
        CS1Spilled = true;
    }

    // This is true when we inserted a spill for an unused register that can now
    // be used for register scavenging.
    bool ExtraCSSpill = false;

    if (AFI->isThumb1OnlyFunction()) {
      // For Thumb1-only targets, we need some low registers when we save and
      // restore the high registers (which aren't allocatable, but could be
      // used by inline assembly) because the push/pop instructions can not
      // access high registers. If necessary, we might need to push more low
      // registers to ensure that there is at least one free that can be used
      // for the saving & restoring, and preferably we should ensure that as
      // many as are needed are available so that fewer push/pop instructions
      // are required.

      // Low registers which are not currently pushed, but could be (r4-r7).
      SmallVector<unsigned, 4> AvailableRegs;

      // Unused argument registers (r0-r3) can be clobbered in the prologue for
      // free.
      int EntryRegDeficit = 0;
      for (unsigned Reg : {ARM::R0, ARM::R1, ARM::R2, ARM::R3}) {
        if (!MF.getRegInfo().isLiveIn(Reg)) {
          --EntryRegDeficit;
          DEBUG(dbgs() << printReg(Reg, TRI)
                       << " is unused argument register, EntryRegDeficit = "
                       << EntryRegDeficit << "\n");
        }
      }

      // Unused return registers can be clobbered in the epilogue for free.
      int ExitRegDeficit = AFI->getReturnRegsCount() - 4;
      DEBUG(dbgs() << AFI->getReturnRegsCount()
                   << " return regs used, ExitRegDeficit = " << ExitRegDeficit
                   << "\n");

      int RegDeficit = std::max(EntryRegDeficit, ExitRegDeficit);
      DEBUG(dbgs() << "RegDeficit = " << RegDeficit << "\n");

      // r4-r6 can be used in the prologue if they are pushed by the first push
      // instruction.
      for (unsigned Reg : {ARM::R4, ARM::R5, ARM::R6}) {
        if (SavedRegs.test(Reg)) {
          --RegDeficit;
          DEBUG(dbgs() << printReg(Reg, TRI)
                       << " is saved low register, RegDeficit = " << RegDeficit
                       << "\n");
        } else {
          AvailableRegs.push_back(Reg);
          DEBUG(dbgs()
                << printReg(Reg, TRI)
                << " is non-saved low register, adding to AvailableRegs\n");
        }
      }

      // r7 can be used if it is not being used as the frame pointer.
      if (!HasFP) {
        if (SavedRegs.test(ARM::R7)) {
          --RegDeficit;
          DEBUG(dbgs() << "%r7 is saved low register, RegDeficit = "
                       << RegDeficit << "\n");
        } else {
          AvailableRegs.push_back(ARM::R7);
          DEBUG(dbgs()
                << "%r7 is non-saved low register, adding to AvailableRegs\n");
        }
      }

      // Each of r8-r11 needs to be copied to a low register, then pushed.
      for (unsigned Reg : {ARM::R8, ARM::R9, ARM::R10, ARM::R11}) {
        if (SavedRegs.test(Reg)) {
          ++RegDeficit;
          DEBUG(dbgs() << printReg(Reg, TRI)
                       << " is saved high register, RegDeficit = " << RegDeficit
                       << "\n");
        }
      }

      // LR can only be used by PUSH, not POP, and can't be used at all if the
      // llvm.returnaddress intrinsic is used. This is only worth doing if we
      // are more limited at function entry than exit.
      if ((EntryRegDeficit > ExitRegDeficit) &&
          !(MF.getRegInfo().isLiveIn(ARM::LR) &&
            MF.getFrameInfo().isReturnAddressTaken())) {
        if (SavedRegs.test(ARM::LR)) {
          --RegDeficit;
          DEBUG(dbgs() << "%lr is saved register, RegDeficit = " << RegDeficit
                       << "\n");
        } else {
          AvailableRegs.push_back(ARM::LR);
          DEBUG(dbgs() << "%lr is not saved, adding to AvailableRegs\n");
        }
      }

      // If there are more high registers that need pushing than low registers
      // available, push some more low registers so that we can use fewer push
      // instructions. This might not reduce RegDeficit all the way to zero,
      // because we can only guarantee that r4-r6 are available, but r8-r11 may
      // need saving.
      DEBUG(dbgs() << "Final RegDeficit = " << RegDeficit << "\n");
      for (; RegDeficit > 0 && !AvailableRegs.empty(); --RegDeficit) {
        unsigned Reg = AvailableRegs.pop_back_val();
        DEBUG(dbgs() << "Spilling " << printReg(Reg, TRI)
                     << " to make up reg deficit\n");
        SavedRegs.set(Reg);
        NumGPRSpills++;
        CS1Spilled = true;
        assert(!MRI.isReserved(Reg) && "Should not be reserved");
        if (!MRI.isPhysRegUsed(Reg))
          ExtraCSSpill = true;
        UnspilledCS1GPRs.erase(llvm::find(UnspilledCS1GPRs, Reg));
        if (Reg == ARM::LR)
          LRSpilled = true;
      }
      DEBUG(dbgs() << "After adding spills, RegDeficit = " << RegDeficit << "\n");
    }

    // If LR is not spilled, but at least one of R4, R5, R6, and R7 is spilled.
    // Spill LR as well so we can fold BX_RET to the registers restore (LDM).
    if (!LRSpilled && CS1Spilled) {
      SavedRegs.set(ARM::LR);
      NumGPRSpills++;
      SmallVectorImpl<unsigned>::iterator LRPos;
      LRPos = llvm::find(UnspilledCS1GPRs, (unsigned)ARM::LR);
      if (LRPos != UnspilledCS1GPRs.end())
        UnspilledCS1GPRs.erase(LRPos);

      ForceLRSpill = false;
      if (!MRI.isReserved(ARM::LR) && !MRI.isPhysRegUsed(ARM::LR))
        ExtraCSSpill = true;
    }

    // If stack and double are 8-byte aligned and we are spilling an odd number
    // of GPRs, spill one extra callee save GPR so we won't have to pad between
    // the integer and double callee save areas.
    DEBUG(dbgs() << "NumGPRSpills = " << NumGPRSpills << "\n");
    unsigned TargetAlign = getStackAlignment();
    if (TargetAlign >= 8 && (NumGPRSpills & 1)) {
      if (CS1Spilled && !UnspilledCS1GPRs.empty()) {
        for (unsigned i = 0, e = UnspilledCS1GPRs.size(); i != e; ++i) {
          unsigned Reg = UnspilledCS1GPRs[i];
          // Don't spill high register if the function is thumb.  In the case of
          // Windows on ARM, accept R11 (frame pointer)
          if (!AFI->isThumbFunction() ||
              (STI.isTargetWindows() && Reg == ARM::R11) ||
              isARMLowRegister(Reg) || Reg == ARM::LR) {
            SavedRegs.set(Reg);
            DEBUG(dbgs() << "Spilling " << printReg(Reg, TRI)
                         << " to make up alignment\n");
            if (!MRI.isReserved(Reg) && !MRI.isPhysRegUsed(Reg))
              ExtraCSSpill = true;
            break;
          }
        }
      } else if (!UnspilledCS2GPRs.empty() && !AFI->isThumb1OnlyFunction()) {
        unsigned Reg = UnspilledCS2GPRs.front();
        SavedRegs.set(Reg);
        DEBUG(dbgs() << "Spilling " << printReg(Reg, TRI)
                     << " to make up alignment\n");
        if (!MRI.isReserved(Reg) && !MRI.isPhysRegUsed(Reg))
          ExtraCSSpill = true;
      }
    }

    // Estimate if we might need to scavenge a register at some point in order
    // to materialize a stack offset. If so, either spill one additional
    // callee-saved register or reserve a special spill slot to facilitate
    // register scavenging. Thumb1 needs a spill slot for stack pointer
    // adjustments also, even when the frame itself is small.
    if (BigFrameOffsets && !ExtraCSSpill) {
      // If any non-reserved CS register isn't spilled, just spill one or two
      // extra. That should take care of it!
      unsigned NumExtras = TargetAlign / 4;
      SmallVector<unsigned, 2> Extras;
      while (NumExtras && !UnspilledCS1GPRs.empty()) {
        unsigned Reg = UnspilledCS1GPRs.back();
        UnspilledCS1GPRs.pop_back();
        if (!MRI.isReserved(Reg) &&
            (!AFI->isThumb1OnlyFunction() || isARMLowRegister(Reg) ||
             Reg == ARM::LR)) {
          Extras.push_back(Reg);
          NumExtras--;
        }
      }
      // For non-Thumb1 functions, also check for hi-reg CS registers
      if (!AFI->isThumb1OnlyFunction()) {
        while (NumExtras && !UnspilledCS2GPRs.empty()) {
          unsigned Reg = UnspilledCS2GPRs.back();
          UnspilledCS2GPRs.pop_back();
          if (!MRI.isReserved(Reg)) {
            Extras.push_back(Reg);
            NumExtras--;
          }
        }
      }
      if (NumExtras == 0) {
        for (unsigned Reg : Extras) {
          SavedRegs.set(Reg);
          if (!MRI.isPhysRegUsed(Reg))
            ExtraCSSpill = true;
        }
      }
      if (!ExtraCSSpill && !AFI->isThumb1OnlyFunction()) {
        // note: Thumb1 functions spill to R12, not the stack.  Reserve a slot
        // closest to SP or frame pointer.
        assert(RS && "Register scavenging not provided");
        const TargetRegisterClass &RC = ARM::GPRRegClass;
        unsigned Size = TRI->getSpillSize(RC);
        unsigned Align = TRI->getSpillAlignment(RC);
        RS->addScavengingFrameIndex(MFI.CreateStackObject(Size, Align, false));
      }
    }
  }

  if (ForceLRSpill) {
    SavedRegs.set(ARM::LR);
    AFI->setLRIsSpilledForFarJump(true);
  }
}

MachineBasicBlock::iterator ARMFrameLowering::eliminateCallFramePseudoInstr(
    MachineFunction &MF, MachineBasicBlock &MBB,
    MachineBasicBlock::iterator I) const {
  const ARMBaseInstrInfo &TII =
      *static_cast<const ARMBaseInstrInfo *>(MF.getSubtarget().getInstrInfo());
  if (!hasReservedCallFrame(MF)) {
    // If we have alloca, convert as follows:
    // ADJCALLSTACKDOWN -> sub, sp, sp, amount
    // ADJCALLSTACKUP   -> add, sp, sp, amount
    MachineInstr &Old = *I;
    DebugLoc dl = Old.getDebugLoc();
    unsigned Amount = TII.getFrameSize(Old);
    if (Amount != 0) {
      // We need to keep the stack aligned properly.  To do this, we round the
      // amount of space needed for the outgoing arguments up to the next
      // alignment boundary.
      Amount = alignSPAdjust(Amount);

      ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
      assert(!AFI->isThumb1OnlyFunction() &&
             "This eliminateCallFramePseudoInstr does not support Thumb1!");
      bool isARM = !AFI->isThumbFunction();

      // Replace the pseudo instruction with a new instruction...
      unsigned Opc = Old.getOpcode();
      int PIdx = Old.findFirstPredOperandIdx();
      ARMCC::CondCodes Pred =
          (PIdx == -1) ? ARMCC::AL
                       : (ARMCC::CondCodes)Old.getOperand(PIdx).getImm();
      unsigned PredReg = TII.getFramePred(Old);
      if (Opc == ARM::ADJCALLSTACKDOWN || Opc == ARM::tADJCALLSTACKDOWN) {
        emitSPUpdate(isARM, MBB, I, dl, TII, -Amount, MachineInstr::NoFlags,
                     Pred, PredReg);
      } else {
        assert(Opc == ARM::ADJCALLSTACKUP || Opc == ARM::tADJCALLSTACKUP);
        emitSPUpdate(isARM, MBB, I, dl, TII, Amount, MachineInstr::NoFlags,
                     Pred, PredReg);
      }
    }
  }
  return MBB.erase(I);
}

/// Get the minimum constant for ARM that is greater than or equal to the
/// argument. In ARM, constants can have any value that can be produced by
/// rotating an 8-bit value to the right by an even number of bits within a
/// 32-bit word.
static uint32_t alignToARMConstant(uint32_t Value) {
  unsigned Shifted = 0;

  if (Value == 0)
      return 0;

  while (!(Value & 0xC0000000)) {
      Value = Value << 2;
      Shifted += 2;
  }

  bool Carry = (Value & 0x00FFFFFF);
  Value = ((Value & 0xFF000000) >> 24) + Carry;

  if (Value & 0x0000100)
      Value = Value & 0x000001FC;

  if (Shifted > 24)
      Value = Value >> (Shifted - 24);
  else
      Value = Value << (24 - Shifted);

  return Value;
}

// The stack limit in the TCB is set to this many bytes above the actual
// stack limit.
static const uint64_t kSplitStackAvailable = 256;

// Adjust the function prologue to enable split stacks. This currently only
// supports android and linux.
//
// The ABI of the segmented stack prologue is a little arbitrarily chosen, but
// must be well defined in order to allow for consistent implementations of the
// __morestack helper function. The ABI is also not a normal ABI in that it
// doesn't follow the normal calling conventions because this allows the
// prologue of each function to be optimized further.
//
// Currently, the ABI looks like (when calling __morestack)
//
//  * r4 holds the minimum stack size requested for this function call
//  * r5 holds the stack size of the arguments to the function
//  * the beginning of the function is 3 instructions after the call to
//    __morestack
//
// Implementations of __morestack should use r4 to allocate a new stack, r5 to
// place the arguments on to the new stack, and the 3-instruction knowledge to
// jump directly to the body of the function when working on the new stack.
//
// An old (and possibly no longer compatible) implementation of __morestack for
// ARM can be found at [1].
//
// [1] - https://github.com/mozilla/rust/blob/86efd9/src/rt/arch/arm/morestack.S
void ARMFrameLowering::adjustForSegmentedStacks(
    MachineFunction &MF, MachineBasicBlock &PrologueMBB) const {
  unsigned Opcode;
  unsigned CFIIndex;
  const ARMSubtarget *ST = &MF.getSubtarget<ARMSubtarget>();
  bool Thumb = ST->isThumb();

  // Sadly, this currently doesn't support varargs, platforms other than
  // android/linux. Note that thumb1/thumb2 are support for android/linux.
  if (MF.getFunction().isVarArg())
    report_fatal_error("Segmented stacks do not support vararg functions.");
  if (!ST->isTargetAndroid() && !ST->isTargetLinux())
    report_fatal_error("Segmented stacks not supported on this platform.");

  MachineFrameInfo &MFI = MF.getFrameInfo();
  MachineModuleInfo &MMI = MF.getMMI();
  MCContext &Context = MMI.getContext();
  const MCRegisterInfo *MRI = Context.getRegisterInfo();
  const ARMBaseInstrInfo &TII =
      *static_cast<const ARMBaseInstrInfo *>(MF.getSubtarget().getInstrInfo());
  ARMFunctionInfo *ARMFI = MF.getInfo<ARMFunctionInfo>();
  DebugLoc DL;

  uint64_t StackSize = MFI.getStackSize();

  // Do not generate a prologue for functions with a stack of size zero
  if (StackSize == 0)
    return;

  // Use R4 and R5 as scratch registers.
  // We save R4 and R5 before use and restore them before leaving the function.
  unsigned ScratchReg0 = ARM::R4;
  unsigned ScratchReg1 = ARM::R5;
  uint64_t AlignedStackSize;

  MachineBasicBlock *PrevStackMBB = MF.CreateMachineBasicBlock();
  MachineBasicBlock *PostStackMBB = MF.CreateMachineBasicBlock();
  MachineBasicBlock *AllocMBB = MF.CreateMachineBasicBlock();
  MachineBasicBlock *GetMBB = MF.CreateMachineBasicBlock();
  MachineBasicBlock *McrMBB = MF.CreateMachineBasicBlock();

  // Grab everything that reaches PrologueMBB to update there liveness as well.
  SmallPtrSet<MachineBasicBlock *, 8> BeforePrologueRegion;
  SmallVector<MachineBasicBlock *, 2> WalkList;
  WalkList.push_back(&PrologueMBB);

  do {
    MachineBasicBlock *CurMBB = WalkList.pop_back_val();
    for (MachineBasicBlock *PredBB : CurMBB->predecessors()) {
      if (BeforePrologueRegion.insert(PredBB).second)
        WalkList.push_back(PredBB);
    }
  } while (!WalkList.empty());

  // The order in that list is important.
  // The blocks will all be inserted before PrologueMBB using that order.
  // Therefore the block that should appear first in the CFG should appear
  // first in the list.
  MachineBasicBlock *AddedBlocks[] = {PrevStackMBB, McrMBB, GetMBB, AllocMBB,
                                      PostStackMBB};

  for (MachineBasicBlock *B : AddedBlocks)
    BeforePrologueRegion.insert(B);

  for (const auto &LI : PrologueMBB.liveins()) {
    for (MachineBasicBlock *PredBB : BeforePrologueRegion)
      PredBB->addLiveIn(LI);
  }

  // Remove the newly added blocks from the list, since we know
  // we do not have to do the following updates for them.
  for (MachineBasicBlock *B : AddedBlocks) {
    BeforePrologueRegion.erase(B);
    MF.insert(PrologueMBB.getIterator(), B);
  }

  for (MachineBasicBlock *MBB : BeforePrologueRegion) {
    // Make sure the LiveIns are still sorted and unique.
    MBB->sortUniqueLiveIns();
    // Replace the edges to PrologueMBB by edges to the sequences
    // we are about to add.
    MBB->ReplaceUsesOfBlockWith(&PrologueMBB, AddedBlocks[0]);
  }

  // The required stack size that is aligned to ARM constant criterion.
  AlignedStackSize = alignToARMConstant(StackSize);

  // When the frame size is less than 256 we just compare the stack
  // boundary directly to the value of the stack pointer, per gcc.
  bool CompareStackPointer = AlignedStackSize < kSplitStackAvailable;

  // We will use two of the callee save registers as scratch registers so we
  // need to save those registers onto the stack.
  // We will use SR0 to hold stack limit and SR1 to hold the stack size
  // requested and arguments for __morestack().
  // SR0: Scratch Register #0
  // SR1: Scratch Register #1
  // push {SR0, SR1}
  if (Thumb) {
    BuildMI(PrevStackMBB, DL, TII.get(ARM::tPUSH))
        .add(predOps(ARMCC::AL))
        .addReg(ScratchReg0)
        .addReg(ScratchReg1);
  } else {
    BuildMI(PrevStackMBB, DL, TII.get(ARM::STMDB_UPD))
        .addReg(ARM::SP, RegState::Define)
        .addReg(ARM::SP)
        .add(predOps(ARMCC::AL))
        .addReg(ScratchReg0)
        .addReg(ScratchReg1);
  }

  // Emit the relevant DWARF information about the change in stack pointer as
  // well as where to find both r4 and r5 (the callee-save registers)
  CFIIndex =
      MF.addFrameInst(MCCFIInstruction::createDefCfaOffset(nullptr, -8));
  BuildMI(PrevStackMBB, DL, TII.get(TargetOpcode::CFI_INSTRUCTION))
      .addCFIIndex(CFIIndex);
  CFIIndex = MF.addFrameInst(MCCFIInstruction::createOffset(
      nullptr, MRI->getDwarfRegNum(ScratchReg1, true), -4));
  BuildMI(PrevStackMBB, DL, TII.get(TargetOpcode::CFI_INSTRUCTION))
      .addCFIIndex(CFIIndex);
  CFIIndex = MF.addFrameInst(MCCFIInstruction::createOffset(
      nullptr, MRI->getDwarfRegNum(ScratchReg0, true), -8));
  BuildMI(PrevStackMBB, DL, TII.get(TargetOpcode::CFI_INSTRUCTION))
      .addCFIIndex(CFIIndex);

  // mov SR1, sp
  if (Thumb) {
    BuildMI(McrMBB, DL, TII.get(ARM::tMOVr), ScratchReg1)
        .addReg(ARM::SP)
        .add(predOps(ARMCC::AL));
  } else if (CompareStackPointer) {
    BuildMI(McrMBB, DL, TII.get(ARM::MOVr), ScratchReg1)
        .addReg(ARM::SP)
        .add(predOps(ARMCC::AL))
        .add(condCodeOp());
  }

  // sub SR1, sp, #StackSize
  if (!CompareStackPointer && Thumb) {
    BuildMI(McrMBB, DL, TII.get(ARM::tSUBi8), ScratchReg1)
        .add(condCodeOp())
        .addReg(ScratchReg1)
        .addImm(AlignedStackSize)
        .add(predOps(ARMCC::AL));
  } else if (!CompareStackPointer) {
    BuildMI(McrMBB, DL, TII.get(ARM::SUBri), ScratchReg1)
        .addReg(ARM::SP)
        .addImm(AlignedStackSize)
        .add(predOps(ARMCC::AL))
        .add(condCodeOp());
  }

  if (Thumb && ST->isThumb1Only()) {
    unsigned PCLabelId = ARMFI->createPICLabelUId();
    ARMConstantPoolValue *NewCPV = ARMConstantPoolSymbol::Create(
        MF.getFunction().getContext(), "__STACK_LIMIT", PCLabelId, 0);
    MachineConstantPool *MCP = MF.getConstantPool();
    unsigned CPI = MCP->getConstantPoolIndex(NewCPV, 4);

    // ldr SR0, [pc, offset(STACK_LIMIT)]
    BuildMI(GetMBB, DL, TII.get(ARM::tLDRpci), ScratchReg0)
        .addConstantPoolIndex(CPI)
        .add(predOps(ARMCC::AL));

    // ldr SR0, [SR0]
    BuildMI(GetMBB, DL, TII.get(ARM::tLDRi), ScratchReg0)
        .addReg(ScratchReg0)
        .addImm(0)
        .add(predOps(ARMCC::AL));
  } else {
    // Get TLS base address from the coprocessor
    // mrc p15, #0, SR0, c13, c0, #3
    BuildMI(McrMBB, DL, TII.get(ARM::MRC), ScratchReg0)
        .addImm(15)
        .addImm(0)
        .addImm(13)
        .addImm(0)
        .addImm(3)
        .add(predOps(ARMCC::AL));

    // Use the last tls slot on android and a private field of the TCP on linux.
    assert(ST->isTargetAndroid() || ST->isTargetLinux());
    unsigned TlsOffset = ST->isTargetAndroid() ? 63 : 1;

    // Get the stack limit from the right offset
    // ldr SR0, [sr0, #4 * TlsOffset]
    BuildMI(GetMBB, DL, TII.get(ARM::LDRi12), ScratchReg0)
        .addReg(ScratchReg0)
        .addImm(4 * TlsOffset)
        .add(predOps(ARMCC::AL));
  }

  // Compare stack limit with stack size requested.
  // cmp SR0, SR1
  Opcode = Thumb ? ARM::tCMPr : ARM::CMPrr;
  BuildMI(GetMBB, DL, TII.get(Opcode))
      .addReg(ScratchReg0)
      .addReg(ScratchReg1)
      .add(predOps(ARMCC::AL));

  // This jump is taken if StackLimit < SP - stack required.
  Opcode = Thumb ? ARM::tBcc : ARM::Bcc;
  BuildMI(GetMBB, DL, TII.get(Opcode)).addMBB(PostStackMBB)
       .addImm(ARMCC::LO)
       .addReg(ARM::CPSR);


  // Calling __morestack(StackSize, Size of stack arguments).
  // __morestack knows that the stack size requested is in SR0(r4)
  // and amount size of stack arguments is in SR1(r5).

  // Pass first argument for the __morestack by Scratch Register #0.
  //   The amount size of stack required
  if (Thumb) {
    BuildMI(AllocMBB, DL, TII.get(ARM::tMOVi8), ScratchReg0)
        .add(condCodeOp())
        .addImm(AlignedStackSize)
        .add(predOps(ARMCC::AL));
  } else {
    BuildMI(AllocMBB, DL, TII.get(ARM::MOVi), ScratchReg0)
        .addImm(AlignedStackSize)
        .add(predOps(ARMCC::AL))
        .add(condCodeOp());
  }
  // Pass second argument for the __morestack by Scratch Register #1.
  //   The amount size of stack consumed to save function arguments.
  if (Thumb) {
    BuildMI(AllocMBB, DL, TII.get(ARM::tMOVi8), ScratchReg1)
        .add(condCodeOp())
        .addImm(alignToARMConstant(ARMFI->getArgumentStackSize()))
        .add(predOps(ARMCC::AL));
  } else {
    BuildMI(AllocMBB, DL, TII.get(ARM::MOVi), ScratchReg1)
        .addImm(alignToARMConstant(ARMFI->getArgumentStackSize()))
        .add(predOps(ARMCC::AL))
        .add(condCodeOp());
  }

  // push {lr} - Save return address of this function.
  if (Thumb) {
    BuildMI(AllocMBB, DL, TII.get(ARM::tPUSH))
        .add(predOps(ARMCC::AL))
        .addReg(ARM::LR);
  } else {
    BuildMI(AllocMBB, DL, TII.get(ARM::STMDB_UPD))
        .addReg(ARM::SP, RegState::Define)
        .addReg(ARM::SP)
        .add(predOps(ARMCC::AL))
        .addReg(ARM::LR);
  }

  // Emit the DWARF info about the change in stack as well as where to find the
  // previous link register
  CFIIndex =
      MF.addFrameInst(MCCFIInstruction::createDefCfaOffset(nullptr, -12));
  BuildMI(AllocMBB, DL, TII.get(TargetOpcode::CFI_INSTRUCTION))
      .addCFIIndex(CFIIndex);
  CFIIndex = MF.addFrameInst(MCCFIInstruction::createOffset(
        nullptr, MRI->getDwarfRegNum(ARM::LR, true), -12));
  BuildMI(AllocMBB, DL, TII.get(TargetOpcode::CFI_INSTRUCTION))
      .addCFIIndex(CFIIndex);

  // Call __morestack().
  if (Thumb) {
    BuildMI(AllocMBB, DL, TII.get(ARM::tBL))
        .add(predOps(ARMCC::AL))
        .addExternalSymbol("__morestack");
  } else {
    BuildMI(AllocMBB, DL, TII.get(ARM::BL))
        .addExternalSymbol("__morestack");
  }

  // pop {lr} - Restore return address of this original function.
  if (Thumb) {
    if (ST->isThumb1Only()) {
      BuildMI(AllocMBB, DL, TII.get(ARM::tPOP))
          .add(predOps(ARMCC::AL))
          .addReg(ScratchReg0);
      BuildMI(AllocMBB, DL, TII.get(ARM::tMOVr), ARM::LR)
          .addReg(ScratchReg0)
          .add(predOps(ARMCC::AL));
    } else {
      BuildMI(AllocMBB, DL, TII.get(ARM::t2LDR_POST))
          .addReg(ARM::LR, RegState::Define)
          .addReg(ARM::SP, RegState::Define)
          .addReg(ARM::SP)
          .addImm(4)
          .add(predOps(ARMCC::AL));
    }
  } else {
    BuildMI(AllocMBB, DL, TII.get(ARM::LDMIA_UPD))
        .addReg(ARM::SP, RegState::Define)
        .addReg(ARM::SP)
        .add(predOps(ARMCC::AL))
        .addReg(ARM::LR);
  }

  // Restore SR0 and SR1 in case of __morestack() was called.
  // __morestack() will skip PostStackMBB block so we need to restore
  // scratch registers from here.
  // pop {SR0, SR1}
  if (Thumb) {
    BuildMI(AllocMBB, DL, TII.get(ARM::tPOP))
        .add(predOps(ARMCC::AL))
        .addReg(ScratchReg0)
        .addReg(ScratchReg1);
  } else {
    BuildMI(AllocMBB, DL, TII.get(ARM::LDMIA_UPD))
        .addReg(ARM::SP, RegState::Define)
        .addReg(ARM::SP)
        .add(predOps(ARMCC::AL))
        .addReg(ScratchReg0)
        .addReg(ScratchReg1);
  }

  // Update the CFA offset now that we've popped
  CFIIndex = MF.addFrameInst(MCCFIInstruction::createDefCfaOffset(nullptr, 0));
  BuildMI(AllocMBB, DL, TII.get(TargetOpcode::CFI_INSTRUCTION))
      .addCFIIndex(CFIIndex);

  // Return from this function.
  BuildMI(AllocMBB, DL, TII.get(ST->getReturnOpcode())).add(predOps(ARMCC::AL));

  // Restore SR0 and SR1 in case of __morestack() was not called.
  // pop {SR0, SR1}
  if (Thumb) {
    BuildMI(PostStackMBB, DL, TII.get(ARM::tPOP))
        .add(predOps(ARMCC::AL))
        .addReg(ScratchReg0)
        .addReg(ScratchReg1);
  } else {
    BuildMI(PostStackMBB, DL, TII.get(ARM::LDMIA_UPD))
        .addReg(ARM::SP, RegState::Define)
        .addReg(ARM::SP)
        .add(predOps(ARMCC::AL))
        .addReg(ScratchReg0)
        .addReg(ScratchReg1);
  }

  // Update the CFA offset now that we've popped
  CFIIndex = MF.addFrameInst(MCCFIInstruction::createDefCfaOffset(nullptr, 0));
  BuildMI(PostStackMBB, DL, TII.get(TargetOpcode::CFI_INSTRUCTION))
      .addCFIIndex(CFIIndex);

  // Tell debuggers that r4 and r5 are now the same as they were in the
  // previous function, that they're the "Same Value".
  CFIIndex = MF.addFrameInst(MCCFIInstruction::createSameValue(
      nullptr, MRI->getDwarfRegNum(ScratchReg0, true)));
  BuildMI(PostStackMBB, DL, TII.get(TargetOpcode::CFI_INSTRUCTION))
      .addCFIIndex(CFIIndex);
  CFIIndex = MF.addFrameInst(MCCFIInstruction::createSameValue(
      nullptr, MRI->getDwarfRegNum(ScratchReg1, true)));
  BuildMI(PostStackMBB, DL, TII.get(TargetOpcode::CFI_INSTRUCTION))
      .addCFIIndex(CFIIndex);

  // Organizing MBB lists
  PostStackMBB->addSuccessor(&PrologueMBB);

  AllocMBB->addSuccessor(PostStackMBB);

  GetMBB->addSuccessor(PostStackMBB);
  GetMBB->addSuccessor(AllocMBB);

  McrMBB->addSuccessor(GetMBB);

  PrevStackMBB->addSuccessor(McrMBB);

#ifdef EXPENSIVE_CHECKS
  MF.verify();
#endif
}