C++ MachineBasicBlock::begin方法代码示例

void AArch64FrameLowering::emitPrologue(MachineFunction &MF,
                                        MachineBasicBlock &MBB) const {
  MachineBasicBlock::iterator MBBI = MBB.begin();
  const MachineFrameInfo *MFI = MF.getFrameInfo();
  const Function *Fn = MF.getFunction();
  const AArch64Subtarget &Subtarget = MF.getSubtarget<AArch64Subtarget>();
  const AArch64RegisterInfo *RegInfo = Subtarget.getRegisterInfo();
  const TargetInstrInfo *TII = Subtarget.getInstrInfo();
  MachineModuleInfo &MMI = MF.getMMI();
  AArch64FunctionInfo *AFI = MF.getInfo<AArch64FunctionInfo>();
  bool needsFrameMoves = MMI.hasDebugInfo() || Fn->needsUnwindTableEntry();
  bool HasFP = hasFP(MF);

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

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

  int NumBytes = (int)MFI->getStackSize();
  if (!AFI->hasStackFrame()) {
    assert(!HasFP && "unexpected function without stack frame but with FP");

    // All of the stack allocation is for locals.
    AFI->setLocalStackSize(NumBytes);

    // Label used to tie together the PROLOG_LABEL and the MachineMoves.
    MCSymbol *FrameLabel = MMI.getContext().createTempSymbol();

    // REDZONE: If the stack size is less than 128 bytes, we don't need
    // to actually allocate.
    if (NumBytes && !canUseRedZone(MF)) {
      emitFrameOffset(MBB, MBBI, DL, AArch64::SP, AArch64::SP, -NumBytes, TII,
                      MachineInstr::FrameSetup);

      // Encode the stack size of the leaf function.
      unsigned CFIIndex = MMI.addFrameInst(
          MCCFIInstruction::createDefCfaOffset(FrameLabel, -NumBytes));
      BuildMI(MBB, MBBI, DL, TII->get(TargetOpcode::CFI_INSTRUCTION))
          .addCFIIndex(CFIIndex)
          .setMIFlags(MachineInstr::FrameSetup);
    } else if (NumBytes) {
      ++NumRedZoneFunctions;
    }

    return;
  }

  // Only set up FP if we actually need to.
  int FPOffset = 0;
  if (HasFP)
    // Frame pointer is fp = sp - 16.
    FPOffset = AFI->getCalleeSavedStackSize() - 16;

  // Move past the saves of the callee-saved registers.
  MachineBasicBlock::iterator End = MBB.end();
  while (MBBI != End && MBBI->getFlag(MachineInstr::FrameSetup))
    ++MBBI;
  NumBytes -= AFI->getCalleeSavedStackSize();
  assert(NumBytes >= 0 && "Negative stack allocation size!?");
  if (HasFP) {
    // Issue    sub fp, sp, FPOffset or
    //          mov fp,sp          when FPOffset is zero.
    // Note: All stores of callee-saved registers are marked as "FrameSetup".
    // This code marks the instruction(s) that set the FP also.
    emitFrameOffset(MBB, MBBI, DL, AArch64::FP, AArch64::SP, FPOffset, TII,
                    MachineInstr::FrameSetup);
  }

  // All of the remaining stack allocations are for locals.
  AFI->setLocalStackSize(NumBytes);

  // Allocate space for the rest of the frame.

  const unsigned Alignment = MFI->getMaxAlignment();
  const bool NeedsRealignment = RegInfo->needsStackRealignment(MF);
  unsigned scratchSPReg = AArch64::SP;
  if (NumBytes && NeedsRealignment) {
    scratchSPReg = findScratchNonCalleeSaveRegister(&MBB);
    assert(scratchSPReg != AArch64::NoRegister);
  }

  // If we're a leaf function, try using the red zone.
  if (NumBytes && !canUseRedZone(MF))
    // FIXME: in the case of dynamic re-alignment, NumBytes doesn't have
    // the correct value here, as NumBytes also includes padding bytes,
    // which shouldn't be counted here.
    emitFrameOffset(MBB, MBBI, DL, scratchSPReg, AArch64::SP, -NumBytes, TII,
                    MachineInstr::FrameSetup);

  if (NumBytes && NeedsRealignment) {
    const unsigned NrBitsToZero = countTrailingZeros(Alignment);
    assert(NrBitsToZero > 1);
    assert(scratchSPReg != AArch64::SP);

    // SUB X9, SP, NumBytes
    //   -- X9 is temporary register, so shouldn't contain any live data here,
//.........这里部分代码省略.........

/// LowerAtomicPHINode - Lower the PHI node at the top of the specified block,
/// under the assuption that it needs to be lowered in a way that supports
/// atomic execution of PHIs.  This lowering method is always correct all of the
/// time.
///
void PHIElimination::LowerAtomicPHINode(
                                      MachineBasicBlock &MBB,
                                      MachineBasicBlock::iterator AfterPHIsIt) {
  ++NumAtomic;
  // Unlink the PHI node from the basic block, but don't delete the PHI yet.
  MachineInstr *MPhi = MBB.remove(MBB.begin());

  unsigned NumSrcs = (MPhi->getNumOperands() - 1) / 2;
  unsigned DestReg = MPhi->getOperand(0).getReg();
  assert(MPhi->getOperand(0).getSubReg() == 0 && "Can't handle sub-reg PHIs");
  bool isDead = MPhi->getOperand(0).isDead();

  // Create a new register for the incoming PHI arguments.
  MachineFunction &MF = *MBB.getParent();
  unsigned IncomingReg = 0;
  bool reusedIncoming = false;  // Is IncomingReg reused from an earlier PHI?

  // Insert a register to register copy at the top of the current block (but
  // after any remaining phi nodes) which copies the new incoming register
  // into the phi node destination.
  const TargetInstrInfo *TII = MF.getTarget().getInstrInfo();
  if (isSourceDefinedByImplicitDef(MPhi, MRI))
    // If all sources of a PHI node are implicit_def, just emit an
    // implicit_def instead of a copy.
    BuildMI(MBB, AfterPHIsIt, MPhi->getDebugLoc(),
            TII->get(TargetOpcode::IMPLICIT_DEF), DestReg);
  else {
    // Can we reuse an earlier PHI node? This only happens for critical edges,
    // typically those created by tail duplication.
    unsigned &entry = LoweredPHIs[MPhi];
    if (entry) {
      // An identical PHI node was already lowered. Reuse the incoming register.
      IncomingReg = entry;
      reusedIncoming = true;
      ++NumReused;
      DEBUG(dbgs() << "Reusing " << PrintReg(IncomingReg) << " for " << *MPhi);
    } else {
      const TargetRegisterClass *RC = MF.getRegInfo().getRegClass(DestReg);
      entry = IncomingReg = MF.getRegInfo().createVirtualRegister(RC);
    }
    BuildMI(MBB, AfterPHIsIt, MPhi->getDebugLoc(),
            TII->get(TargetOpcode::COPY), DestReg)
      .addReg(IncomingReg);
  }

  // Update live variable information if there is any.
  LiveVariables *LV = getAnalysisIfAvailable<LiveVariables>();
  if (LV) {
    MachineInstr *PHICopy = prior(AfterPHIsIt);

    if (IncomingReg) {
      LiveVariables::VarInfo &VI = LV->getVarInfo(IncomingReg);

      // Increment use count of the newly created virtual register.
      VI.NumUses++;
      LV->setPHIJoin(IncomingReg);

      // When we are reusing the incoming register, it may already have been
      // killed in this block. The old kill will also have been inserted at
      // AfterPHIsIt, so it appears before the current PHICopy.
      if (reusedIncoming)
        if (MachineInstr *OldKill = VI.findKill(&MBB)) {
          DEBUG(dbgs() << "Remove old kill from " << *OldKill);
          LV->removeVirtualRegisterKilled(IncomingReg, OldKill);
          DEBUG(MBB.dump());
        }

      // Add information to LiveVariables to know that the incoming value is
      // killed.  Note that because the value is defined in several places (once
      // each for each incoming block), the "def" block and instruction fields
      // for the VarInfo is not filled in.
      LV->addVirtualRegisterKilled(IncomingReg, PHICopy);
    }

    // Since we are going to be deleting the PHI node, if it is the last use of
    // any registers, or if the value itself is dead, we need to move this
    // information over to the new copy we just inserted.
    LV->removeVirtualRegistersKilled(MPhi);

    // If the result is dead, update LV.
    if (isDead) {
      LV->addVirtualRegisterDead(DestReg, PHICopy);
      LV->removeVirtualRegisterDead(DestReg, MPhi);
    }
  }

  // Adjust the VRegPHIUseCount map to account for the removal of this PHI node.
  for (unsigned i = 1; i != MPhi->getNumOperands(); i += 2)
    --VRegPHIUseCount[BBVRegPair(MPhi->getOperand(i+1).getMBB()->getNumber(),
                                 MPhi->getOperand(i).getReg())];

  // Now loop over all of the incoming arguments, changing them to copy into the
  // IncomingReg register in the corresponding predecessor basic block.
  SmallPtrSet<MachineBasicBlock*, 8> MBBsInsertedInto;
  for (int i = NumSrcs - 1; i >= 0; --i) {
//.........这里部分代码省略.........

/// insertCSRSpillsAndRestores - Insert spill and restore code for
/// callee saved registers used in the function, handling shrink wrapping.
///
void PEI::insertCSRSpillsAndRestores(MachineFunction &Fn) {
  // Get callee saved register information.
  MachineFrameInfo *MFI = Fn.getFrameInfo();
  const std::vector<CalleeSavedInfo> &CSI = MFI->getCalleeSavedInfo();

  MFI->setCalleeSavedInfoValid(true);

  // Early exit if no callee saved registers are modified!
  if (CSI.empty())
    return;

  const TargetInstrInfo &TII = *Fn.getTarget().getInstrInfo();
  const TargetFrameLowering *TFI = Fn.getTarget().getFrameLowering();
  const TargetRegisterInfo *TRI = Fn.getTarget().getRegisterInfo();
  MachineBasicBlock::iterator I;

  if (!ShrinkWrapThisFunction) {
    // Spill using target interface.
    I = EntryBlock->begin();
    if (!TFI->spillCalleeSavedRegisters(*EntryBlock, I, CSI, TRI)) {
      for (unsigned i = 0, e = CSI.size(); i != e; ++i) {
        // Add the callee-saved register as live-in.
        // It's killed at the spill.
        EntryBlock->addLiveIn(CSI[i].getReg());

        // Insert the spill to the stack frame.
        unsigned Reg = CSI[i].getReg();
        const TargetRegisterClass *RC = TRI->getMinimalPhysRegClass(Reg);
        TII.storeRegToStackSlot(*EntryBlock, I, Reg, true,
                                CSI[i].getFrameIdx(), RC, TRI);
      }
    }

    // Restore using target interface.
    for (unsigned ri = 0, re = ReturnBlocks.size(); ri != re; ++ri) {
      MachineBasicBlock* MBB = ReturnBlocks[ri];
      I = MBB->end(); --I;

      // Skip over all terminator instructions, which are part of the return
      // sequence.
      MachineBasicBlock::iterator I2 = I;
      while (I2 != MBB->begin() && (--I2)->isTerminator())
        I = I2;

      bool AtStart = I == MBB->begin();
      MachineBasicBlock::iterator BeforeI = I;
      if (!AtStart)
        --BeforeI;

      // Restore all registers immediately before the return and any
      // terminators that precede it.
      if (!TFI->restoreCalleeSavedRegisters(*MBB, I, CSI, TRI)) {
        for (unsigned i = 0, e = CSI.size(); i != e; ++i) {
          unsigned Reg = CSI[i].getReg();
          const TargetRegisterClass *RC = TRI->getMinimalPhysRegClass(Reg);
          TII.loadRegFromStackSlot(*MBB, I, Reg,
                                   CSI[i].getFrameIdx(),
                                   RC, TRI);
          assert(I != MBB->begin() &&
                 "loadRegFromStackSlot didn't insert any code!");
          // Insert in reverse order.  loadRegFromStackSlot can insert
          // multiple instructions.
          if (AtStart)
            I = MBB->begin();
          else {
            I = BeforeI;
            ++I;
          }
        }
      }
    }
    return;
  }

  // Insert spills.
  std::vector<CalleeSavedInfo> blockCSI;
  for (CSRegBlockMap::iterator BI = CSRSave.begin(),
         BE = CSRSave.end(); BI != BE; ++BI) {
    MachineBasicBlock* MBB = BI->first;
    CSRegSet save = BI->second;

    if (save.empty())
      continue;

    blockCSI.clear();
    for (CSRegSet::iterator RI = save.begin(),
           RE = save.end(); RI != RE; ++RI) {
      blockCSI.push_back(CSI[*RI]);
    }
    assert(blockCSI.size() > 0 &&
           "Could not collect callee saved register info");

    I = MBB->begin();

    // When shrink wrapping, use stack slot stores/loads.
    for (unsigned i = 0, e = blockCSI.size(); i != e; ++i) {
      // Add the callee-saved register as live-in.
//.........这里部分代码省略.........

// Branch analysis.
// Note: If the condition register is set to CTR or CTR8 then this is a
// BDNZ (imm == 1) or BDZ (imm == 0) branch.
bool PPCInstrInfo::AnalyzeBranch(MachineBasicBlock &MBB,MachineBasicBlock *&TBB,
                                 MachineBasicBlock *&FBB,
                                 SmallVectorImpl<MachineOperand> &Cond,
                                 bool AllowModify) const {
  bool isPPC64 = TM.getSubtargetImpl()->isPPC64();

  // If the block has no terminators, it just falls into the block after it.
  MachineBasicBlock::iterator I = MBB.end();
  if (I == MBB.begin())
    return false;
  --I;
  while (I->isDebugValue()) {
    if (I == MBB.begin())
      return false;
    --I;
  }
  if (!isUnpredicatedTerminator(I))
    return false;

  // Get the last instruction in the block.
  MachineInstr *LastInst = I;

  // If there is only one terminator instruction, process it.
  if (I == MBB.begin() || !isUnpredicatedTerminator(--I)) {
    if (LastInst->getOpcode() == PPC::B) {
      if (!LastInst->getOperand(0).isMBB())
        return true;
      TBB = LastInst->getOperand(0).getMBB();
      return false;
    } else if (LastInst->getOpcode() == PPC::BCC) {
      if (!LastInst->getOperand(2).isMBB())
        return true;
      // Block ends with fall-through condbranch.
      TBB = LastInst->getOperand(2).getMBB();
      Cond.push_back(LastInst->getOperand(0));
      Cond.push_back(LastInst->getOperand(1));
      return false;
    } else if (LastInst->getOpcode() == PPC::BDNZ8 ||
               LastInst->getOpcode() == PPC::BDNZ) {
      if (!LastInst->getOperand(0).isMBB())
        return true;
      if (DisableCTRLoopAnal)
        return true;
      TBB = LastInst->getOperand(0).getMBB();
      Cond.push_back(MachineOperand::CreateImm(1));
      Cond.push_back(MachineOperand::CreateReg(isPPC64 ? PPC::CTR8 : PPC::CTR,
                                               true));
      return false;
    } else if (LastInst->getOpcode() == PPC::BDZ8 ||
               LastInst->getOpcode() == PPC::BDZ) {
      if (!LastInst->getOperand(0).isMBB())
        return true;
      if (DisableCTRLoopAnal)
        return true;
      TBB = LastInst->getOperand(0).getMBB();
      Cond.push_back(MachineOperand::CreateImm(0));
      Cond.push_back(MachineOperand::CreateReg(isPPC64 ? PPC::CTR8 : PPC::CTR,
                                               true));
      return false;
    }

    // Otherwise, don't know what this is.
    return true;
  }

  // Get the instruction before it if it's a terminator.
  MachineInstr *SecondLastInst = I;

  // If there are three terminators, we don't know what sort of block this is.
  if (SecondLastInst && I != MBB.begin() &&
      isUnpredicatedTerminator(--I))
    return true;

  // If the block ends with PPC::B and PPC:BCC, handle it.
  if (SecondLastInst->getOpcode() == PPC::BCC &&
      LastInst->getOpcode() == PPC::B) {
    if (!SecondLastInst->getOperand(2).isMBB() ||
        !LastInst->getOperand(0).isMBB())
      return true;
    TBB =  SecondLastInst->getOperand(2).getMBB();
    Cond.push_back(SecondLastInst->getOperand(0));
    Cond.push_back(SecondLastInst->getOperand(1));
    FBB = LastInst->getOperand(0).getMBB();
    return false;
  } else if ((SecondLastInst->getOpcode() == PPC::BDNZ8 ||
              SecondLastInst->getOpcode() == PPC::BDNZ) &&
      LastInst->getOpcode() == PPC::B) {
    if (!SecondLastInst->getOperand(0).isMBB() ||
        !LastInst->getOperand(0).isMBB())
      return true;
    if (DisableCTRLoopAnal)
      return true;
    TBB = SecondLastInst->getOperand(0).getMBB();
    Cond.push_back(MachineOperand::CreateImm(1));
    Cond.push_back(MachineOperand::CreateReg(isPPC64 ? PPC::CTR8 : PPC::CTR,
                                             true));
    FBB = LastInst->getOperand(0).getMBB();
//.........这里部分代码省略.........

bool ARMExpandPseudo::ExpandMBB(MachineBasicBlock &MBB) {
  bool Modified = false;

  MachineBasicBlock::iterator MBBI = MBB.begin(), E = MBB.end();
  while (MBBI != E) {
    MachineInstr &MI = *MBBI;
    MachineBasicBlock::iterator NMBBI = llvm::next(MBBI);

    unsigned Opcode = MI.getOpcode();
    switch (Opcode) {
    default: break;
    case ARM::tLDRpci_pic: 
    case ARM::t2LDRpci_pic: {
      unsigned NewLdOpc = (Opcode == ARM::tLDRpci_pic)
        ? ARM::tLDRpci : ARM::t2LDRpci;
      unsigned DstReg = MI.getOperand(0).getReg();
      if (!MI.getOperand(0).isDead()) {
        MachineInstr *NewMI =
          AddDefaultPred(BuildMI(MBB, MBBI, MI.getDebugLoc(),
                                 TII->get(NewLdOpc), DstReg)
                         .addOperand(MI.getOperand(1)));
        NewMI->setMemRefs(MI.memoperands_begin(), MI.memoperands_end());
        BuildMI(MBB, MBBI, MI.getDebugLoc(), TII->get(ARM::tPICADD))
          .addReg(DstReg, getDefRegState(true))
          .addReg(DstReg)
          .addOperand(MI.getOperand(2));
      }
      MI.eraseFromParent();
      Modified = true;
      break;
    }
    case ARM::t2MOVi32imm: {
      unsigned DstReg = MI.getOperand(0).getReg();
      if (!MI.getOperand(0).isDead()) {
        const MachineOperand &MO = MI.getOperand(1);
        MachineInstrBuilder LO16, HI16;

        LO16 = BuildMI(MBB, MBBI, MI.getDebugLoc(), TII->get(ARM::t2MOVi16),
                       DstReg);
        HI16 = BuildMI(MBB, MBBI, MI.getDebugLoc(), TII->get(ARM::t2MOVTi16))
          .addReg(DstReg, getDefRegState(true)).addReg(DstReg);

        if (MO.isImm()) {
          unsigned Imm = MO.getImm();
          unsigned Lo16 = Imm & 0xffff;
          unsigned Hi16 = (Imm >> 16) & 0xffff;
          LO16 = LO16.addImm(Lo16);
          HI16 = HI16.addImm(Hi16);
        } else {
          GlobalValue *GV = MO.getGlobal();
          unsigned TF = MO.getTargetFlags();
          LO16 = LO16.addGlobalAddress(GV, MO.getOffset(), TF | ARMII::MO_LO16);
          HI16 = HI16.addGlobalAddress(GV, MO.getOffset(), TF | ARMII::MO_HI16);
          // FIXME: What's about memoperands?
        }
        AddDefaultPred(LO16);
        AddDefaultPred(HI16);
      }
      MI.eraseFromParent();
      Modified = true;
    }
    // FIXME: expand t2MOVi32imm
    }

bool Thumb2ITBlockPass::InsertITInstructions(MachineBasicBlock &MBB) {
  bool Modified = false;

  SmallSet<unsigned, 4> Defs;
  SmallSet<unsigned, 4> Uses;
  MachineBasicBlock::iterator MBBI = MBB.begin(), E = MBB.end();
  while (MBBI != E) {
    MachineInstr *MI = &*MBBI;
    DebugLoc dl = MI->getDebugLoc();
    unsigned PredReg = 0;
    ARMCC::CondCodes CC = llvm::getITInstrPredicate(MI, PredReg);
    if (CC == ARMCC::AL) {
      ++MBBI;
      continue;
    }

    Defs.clear();
    Uses.clear();
    TrackDefUses(MI, Defs, Uses, TRI);

    // Insert an IT instruction.
    MachineInstrBuilder MIB = BuildMI(MBB, MBBI, dl, TII->get(ARM::t2IT))
      .addImm(CC);

    // Add implicit use of ITSTATE to IT block instructions.
    MI->addOperand(MachineOperand::CreateReg(ARM::ITSTATE, false/*ifDef*/,
                                             true/*isImp*/, false/*isKill*/));

    MachineInstr *LastITMI = MI;
    MachineBasicBlock::iterator InsertPos = MIB;
    ++MBBI;

    // Form IT block.
    ARMCC::CondCodes OCC = ARMCC::getOppositeCondition(CC);
    unsigned Mask = 0, Pos = 3;
    // Branches, including tricky ones like LDM_RET, need to end an IT
    // block so check the instruction we just put in the block.
    for (; MBBI != E && Pos &&
           (!MI->getDesc().isBranch() && !MI->getDesc().isReturn()) ; ++MBBI) {
      if (MBBI->isDebugValue())
        continue;

      MachineInstr *NMI = &*MBBI;
      MI = NMI;

      unsigned NPredReg = 0;
      ARMCC::CondCodes NCC = llvm::getITInstrPredicate(NMI, NPredReg);
      if (NCC == CC || NCC == OCC) {
        Mask |= (NCC & 1) << Pos;
        // Add implicit use of ITSTATE.
        NMI->addOperand(MachineOperand::CreateReg(ARM::ITSTATE, false/*ifDef*/,
                                               true/*isImp*/, false/*isKill*/));
        LastITMI = NMI;
      } else {
        if (NCC == ARMCC::AL &&
            MoveCopyOutOfITBlock(NMI, CC, OCC, Defs, Uses)) {
          --MBBI;
          MBB.remove(NMI);
          MBB.insert(InsertPos, NMI);
          ++NumMovedInsts;
          continue;
        }
        break;
      }
      TrackDefUses(NMI, Defs, Uses, TRI);
      --Pos;
    }

    // Finalize IT mask.
    Mask |= (1 << Pos);
    // Tag along (firstcond[0] << 4) with the mask.
    Mask |= (CC & 1) << 4;
    MIB.addImm(Mask);

    // Last instruction in IT block kills ITSTATE.
    LastITMI->findRegisterUseOperand(ARM::ITSTATE)->setIsKill();

    Modified = true;
    ++NumITs;
  }

  return Modified;
}

void MCS51FrameLowering::emitEpilogue(MachineFunction &MF,
                                       MachineBasicBlock &MBB) const {
  const MachineFrameInfo *MFI = MF.getFrameInfo();
  MCS51MachineFunctionInfo *MCS51FI = MF.getInfo<MCS51MachineFunctionInfo>();
  const MCS51InstrInfo &TII =
    *static_cast<const MCS51InstrInfo*>(MF.getTarget().getInstrInfo());

  MachineBasicBlock::iterator MBBI = MBB.getLastNonDebugInstr();
  unsigned RetOpcode = MBBI->getOpcode();
  DebugLoc DL = MBBI->getDebugLoc();

  switch (RetOpcode) {
  case MCS51::RET:
  case MCS51::RETI: break;  // These are ok
  default:
    llvm_unreachable("Can only insert epilog into returning blocks");
  }

  // Get the number of bytes to allocate from the FrameInfo
  uint64_t StackSize = MFI->getStackSize();
  unsigned CSSize = MCS51FI->getCalleeSavedFrameSize();
  uint64_t NumBytes = 0;

  if (hasFP(MF)) {
    // Calculate required stack adjustment
    uint64_t FrameSize = StackSize - 2;
    NumBytes = FrameSize - CSSize;

    // pop FPW.
    BuildMI(MBB, MBBI, DL, TII.get(MCS51::POP16r), MCS51::FPW);
  } else
    NumBytes = StackSize - CSSize;

  // Skip the callee-saved pop instructions.
  while (MBBI != MBB.begin()) {
    MachineBasicBlock::iterator PI = prior(MBBI);
    unsigned Opc = PI->getOpcode();
    if (Opc != MCS51::POP16r && !PI->isTerminator())
      break;
    --MBBI;
  }

  DL = MBBI->getDebugLoc();

  // If there is an ADD16ri or SUB16ri of SPW immediately before this
  // instruction, merge the two instructions.
  //if (NumBytes || MFI->hasVarSizedObjects())
  //  mergeSPUpdatesUp(MBB, MBBI, StackPtr, &NumBytes);

  if (MFI->hasVarSizedObjects()) {
    BuildMI(MBB, MBBI, DL,
            TII.get(MCS51::MOV16rr), MCS51::SPW).addReg(MCS51::FPW);
    if (CSSize) {
      MachineInstr *MI =
        BuildMI(MBB, MBBI, DL,
                TII.get(MCS51::SUB16ri), MCS51::SPW)
        .addReg(MCS51::SPW).addImm(CSSize);
      // The SRW implicit def is dead.
      MI->getOperand(3).setIsDead();
    }
  } else {
    // adjust stack pointer back: SPW += numbytes
    if (NumBytes) {
      MachineInstr *MI =
        BuildMI(MBB, MBBI, DL, TII.get(MCS51::ADD16ri), MCS51::SPW)
        .addReg(MCS51::SPW).addImm(NumBytes);
      // The SRW implicit def is dead.
      MI->getOperand(3).setIsDead();
    }
  }
}

bool PPCBSel::runOnMachineFunction(MachineFunction &Fn) {
  const PPCInstrInfo *TII =
                static_cast<const PPCInstrInfo*>(Fn.getTarget().getInstrInfo());
  // Give the blocks of the function a dense, in-order, numbering.
  Fn.RenumberBlocks();
  BlockSizes.resize(Fn.getNumBlockIDs());

  // Measure each MBB and compute a size for the entire function.
  unsigned FuncSize = 0;
  for (MachineFunction::iterator MFI = Fn.begin(), E = Fn.end(); MFI != E;
       ++MFI) {
    MachineBasicBlock *MBB = MFI;

    unsigned BlockSize = 0;
    for (MachineBasicBlock::iterator MBBI = MBB->begin(), EE = MBB->end();
         MBBI != EE; ++MBBI)
      BlockSize += TII->GetInstSizeInBytes(MBBI);
    
    BlockSizes[MBB->getNumber()] = BlockSize;
    FuncSize += BlockSize;
  }
  
  // If the entire function is smaller than the displacement of a branch field,
  // we know we don't need to shrink any branches in this function.  This is a
  // common case.
  if (FuncSize < (1 << 15)) {
    BlockSizes.clear();
    return false;
  }
  
  // For each conditional branch, if the offset to its destination is larger
  // than the offset field allows, transform it into a long branch sequence
  // like this:
  //   short branch:
  //     bCC MBB
  //   long branch:
  //     b!CC $PC+8
  //     b MBB
  //
  bool MadeChange = true;
  bool EverMadeChange = false;
  while (MadeChange) {
    // Iteratively expand branches until we reach a fixed point.
    MadeChange = false;
  
    for (MachineFunction::iterator MFI = Fn.begin(), E = Fn.end(); MFI != E;
         ++MFI) {
      MachineBasicBlock &MBB = *MFI;
      unsigned MBBStartOffset = 0;
      for (MachineBasicBlock::iterator I = MBB.begin(), E = MBB.end();
           I != E; ++I) {
        MachineBasicBlock *Dest = nullptr;
        if (I->getOpcode() == PPC::BCC && !I->getOperand(2).isImm())
          Dest = I->getOperand(2).getMBB();
        else if ((I->getOpcode() == PPC::BC || I->getOpcode() == PPC::BCn) &&
                 !I->getOperand(1).isImm())
          Dest = I->getOperand(1).getMBB();
        else if ((I->getOpcode() == PPC::BDNZ8 || I->getOpcode() == PPC::BDNZ ||
                  I->getOpcode() == PPC::BDZ8  || I->getOpcode() == PPC::BDZ) &&
                 !I->getOperand(0).isImm())
          Dest = I->getOperand(0).getMBB();

        if (!Dest) {
          MBBStartOffset += TII->GetInstSizeInBytes(I);
          continue;
        }
        
        // Determine the offset from the current branch to the destination
        // block.
        int BranchSize;
        if (Dest->getNumber() <= MBB.getNumber()) {
          // If this is a backwards branch, the delta is the offset from the
          // start of this block to this branch, plus the sizes of all blocks
          // from this block to the dest.
          BranchSize = MBBStartOffset;
          
          for (unsigned i = Dest->getNumber(), e = MBB.getNumber(); i != e; ++i)
            BranchSize += BlockSizes[i];
        } else {
          // Otherwise, add the size of the blocks between this block and the
          // dest to the number of bytes left in this block.
          BranchSize = -MBBStartOffset;

          for (unsigned i = MBB.getNumber(), e = Dest->getNumber(); i != e; ++i)
            BranchSize += BlockSizes[i];
        }

        // If this branch is in range, ignore it.
        if (isInt<16>(BranchSize)) {
          MBBStartOffset += 4;
          continue;
        }

        // Otherwise, we have to expand it to a long branch.
        MachineInstr *OldBranch = I;
        DebugLoc dl = OldBranch->getDebugLoc();
 
        if (I->getOpcode() == PPC::BCC) {
          // The BCC operands are:
          // 0. PPC branch predicate
//.........这里部分代码省略.........

/// insertCSRSpillsAndRestores - Insert spill and restore code for
/// callee saved registers used in the function.
///
void PEI::insertCSRSpillsAndRestores(MachineFunction &Fn) {
  // Get callee saved register information.
  MachineFrameInfo *MFI = Fn.getFrameInfo();
  const std::vector<CalleeSavedInfo> &CSI = MFI->getCalleeSavedInfo();

  MFI->setCalleeSavedInfoValid(true);

  // Early exit if no callee saved registers are modified!
  if (CSI.empty())
    return;

  const TargetInstrInfo &TII = *Fn.getSubtarget().getInstrInfo();
  const TargetFrameLowering *TFI = Fn.getSubtarget().getFrameLowering();
  const TargetRegisterInfo *TRI = Fn.getSubtarget().getRegisterInfo();
  MachineBasicBlock::iterator I;

  // Spill using target interface.
  I = EntryBlock->begin();
  if (!TFI->spillCalleeSavedRegisters(*EntryBlock, I, CSI, TRI)) {
    for (unsigned i = 0, e = CSI.size(); i != e; ++i) {
      // Add the callee-saved register as live-in.
      // It's killed at the spill.
      EntryBlock->addLiveIn(CSI[i].getReg());

      // Insert the spill to the stack frame.
      unsigned Reg = CSI[i].getReg();
      const TargetRegisterClass *RC = TRI->getMinimalPhysRegClass(Reg);
      TII.storeRegToStackSlot(*EntryBlock, I, Reg, true, CSI[i].getFrameIdx(),
                              RC, TRI);
    }
  }

  // Restore using target interface.
  for (unsigned ri = 0, re = ReturnBlocks.size(); ri != re; ++ri) {
    MachineBasicBlock *MBB = ReturnBlocks[ri];
    I = MBB->end();
    --I;

    // Skip over all terminator instructions, which are part of the return
    // sequence.
    MachineBasicBlock::iterator I2 = I;
    while (I2 != MBB->begin() && (--I2)->isTerminator())
      I = I2;

    bool AtStart = I == MBB->begin();
    MachineBasicBlock::iterator BeforeI = I;
    if (!AtStart)
      --BeforeI;

    // Restore all registers immediately before the return and any
    // terminators that precede it.
    if (!TFI->restoreCalleeSavedRegisters(*MBB, I, CSI, TRI)) {
      for (unsigned i = 0, e = CSI.size(); i != e; ++i) {
        unsigned Reg = CSI[i].getReg();
        const TargetRegisterClass *RC = TRI->getMinimalPhysRegClass(Reg);
        TII.loadRegFromStackSlot(*MBB, I, Reg, CSI[i].getFrameIdx(), RC, TRI);
        assert(I != MBB->begin() &&
               "loadRegFromStackSlot didn't insert any code!");
        // Insert in reverse order.  loadRegFromStackSlot can insert
        // multiple instructions.
        if (AtStart)
          I = MBB->begin();
        else {
          I = BeforeI;
          ++I;
        }
      }
    }
  }
}

/// Walk the specified loop in the CFG (defined by all blocks dominated by the
/// specified header block, and that are in the current loop) in depth first
/// order w.r.t the DominatorTree. This allows us to visit definitions before
/// uses, allowing us to hoist a loop body in one pass without iteration.
///
void MachineLICM::HoistOutOfLoop(MachineDomTreeNode *HeaderN) {
  MachineBasicBlock *Preheader = getCurPreheader();
  if (!Preheader)
    return;

  SmallVector<MachineDomTreeNode*, 32> Scopes;
  SmallVector<MachineDomTreeNode*, 8> WorkList;
  DenseMap<MachineDomTreeNode*, MachineDomTreeNode*> ParentMap;
  DenseMap<MachineDomTreeNode*, unsigned> OpenChildren;

  // Perform a DFS walk to determine the order of visit.
  WorkList.push_back(HeaderN);
  while (!WorkList.empty()) {
    MachineDomTreeNode *Node = WorkList.pop_back_val();
    assert(Node && "Null dominator tree node?");
    MachineBasicBlock *BB = Node->getBlock();

    // If the header of the loop containing this basic block is a landing pad,
    // then don't try to hoist instructions out of this loop.
    const MachineLoop *ML = MLI->getLoopFor(BB);
    if (ML && ML->getHeader()->isEHPad())
      continue;

    // If this subregion is not in the top level loop at all, exit.
    if (!CurLoop->contains(BB))
      continue;

    Scopes.push_back(Node);
    const std::vector<MachineDomTreeNode*> &Children = Node->getChildren();
    unsigned NumChildren = Children.size();

    // Don't hoist things out of a large switch statement.  This often causes
    // code to be hoisted that wasn't going to be executed, and increases
    // register pressure in a situation where it's likely to matter.
    if (BB->succ_size() >= 25)
      NumChildren = 0;

    OpenChildren[Node] = NumChildren;
    // Add children in reverse order as then the next popped worklist node is
    // the first child of this node.  This means we ultimately traverse the
    // DOM tree in exactly the same order as if we'd recursed.
    for (int i = (int)NumChildren-1; i >= 0; --i) {
      MachineDomTreeNode *Child = Children[i];
      ParentMap[Child] = Node;
      WorkList.push_back(Child);
    }
  }

  if (Scopes.size() == 0)
    return;

  // Compute registers which are livein into the loop headers.
  RegSeen.clear();
  BackTrace.clear();
  InitRegPressure(Preheader);

  // Now perform LICM.
  for (MachineDomTreeNode *Node : Scopes) {
    MachineBasicBlock *MBB = Node->getBlock();

    EnterScope(MBB);

    // Process the block
    SpeculationState = SpeculateUnknown;
    for (MachineBasicBlock::iterator
         MII = MBB->begin(), E = MBB->end(); MII != E; ) {
      MachineBasicBlock::iterator NextMII = MII; ++NextMII;
      MachineInstr *MI = &*MII;
      if (!Hoist(MI, Preheader))
        UpdateRegPressure(MI);
      MII = NextMII;
    }

    // If it's a leaf node, it's done. Traverse upwards to pop ancestors.
    ExitScopeIfDone(Node, OpenChildren, ParentMap);
  }
}

bool SystemZInstrInfo::AnalyzeBranch(MachineBasicBlock &MBB,
                                     MachineBasicBlock *&TBB,
                                     MachineBasicBlock *&FBB,
                                     SmallVectorImpl<MachineOperand> &Cond,
                                     bool AllowModify) const {
  // Most of the code and comments here are boilerplate.

  // Start from the bottom of the block and work up, examining the
  // terminator instructions.
  MachineBasicBlock::iterator I = MBB.end();
  while (I != MBB.begin()) {
    --I;
    if (I->isDebugValue())
      continue;

    // Working from the bottom, when we see a non-terminator instruction, we're
    // done.
    if (!isUnpredicatedTerminator(I))
      break;

    // A terminator that isn't a branch can't easily be handled by this
    // analysis.
    unsigned ThisCond;
    const MachineOperand *ThisTarget;
    if (!isBranch(I, ThisCond, ThisTarget))
      return true;

    // Can't handle indirect branches.
    if (!ThisTarget->isMBB())
      return true;

    if (ThisCond == SystemZ::CCMASK_ANY) {
      // Handle unconditional branches.
      if (!AllowModify) {
        TBB = ThisTarget->getMBB();
        continue;
      }

      // If the block has any instructions after a JMP, delete them.
      while (llvm::next(I) != MBB.end())
        llvm::next(I)->eraseFromParent();

      Cond.clear();
      FBB = 0;

      // Delete the JMP if it's equivalent to a fall-through.
      if (MBB.isLayoutSuccessor(ThisTarget->getMBB())) {
        TBB = 0;
        I->eraseFromParent();
        I = MBB.end();
        continue;
      }

      // TBB is used to indicate the unconditinal destination.
      TBB = ThisTarget->getMBB();
      continue;
    }

    // Working from the bottom, handle the first conditional branch.
    if (Cond.empty()) {
      // FIXME: add X86-style branch swap
      FBB = TBB;
      TBB = ThisTarget->getMBB();
      Cond.push_back(MachineOperand::CreateImm(ThisCond));
      continue;
    }

    // Handle subsequent conditional branches.
    assert(Cond.size() == 1);
    assert(TBB);

    // Only handle the case where all conditional branches branch to the same
    // destination.
    if (TBB != ThisTarget->getMBB())
      return true;

    // If the conditions are the same, we can leave them alone.
    unsigned OldCond = Cond[0].getImm();
    if (OldCond == ThisCond)
      continue;

    // FIXME: Try combining conditions like X86 does.  Should be easy on Z!
  }

  return false;
}

void AArch64FrameLowering::emitEpilogue(MachineFunction &MF,
                                        MachineBasicBlock &MBB) const {
  MachineBasicBlock::iterator MBBI = MBB.getLastNonDebugInstr();
  MachineFrameInfo *MFI = MF.getFrameInfo();
  const AArch64Subtarget &Subtarget = MF.getSubtarget<AArch64Subtarget>();
  const TargetInstrInfo *TII = Subtarget.getInstrInfo();
  DebugLoc DL;
  bool IsTailCallReturn = false;
  if (MBB.end() != MBBI) {
    DL = MBBI->getDebugLoc();
    unsigned RetOpcode = MBBI->getOpcode();
    IsTailCallReturn = RetOpcode == AArch64::TCRETURNdi ||
      RetOpcode == AArch64::TCRETURNri;
  }
  int NumBytes = MFI->getStackSize();
  const AArch64FunctionInfo *AFI = MF.getInfo<AArch64FunctionInfo>();

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

  // Initial and residual are named for consistency with the prologue. Note that
  // in the epilogue, the residual adjustment is executed first.
  uint64_t ArgumentPopSize = 0;
  if (IsTailCallReturn) {
    MachineOperand &StackAdjust = MBBI->getOperand(1);

    // For a tail-call in a callee-pops-arguments environment, some or all of
    // the stack may actually be in use for the call's arguments, this is
    // calculated during LowerCall and consumed here...
    ArgumentPopSize = StackAdjust.getImm();
  } else {
    // ... otherwise the amount to pop is *all* of the argument space,
    // conveniently stored in the MachineFunctionInfo by
    // LowerFormalArguments. This will, of course, be zero for the C calling
    // convention.
    ArgumentPopSize = AFI->getArgumentStackToRestore();
  }

  // The stack frame should be like below,
  //
  //      ----------------------                     ---
  //      |                    |                      |
  //      | BytesInStackArgArea|              CalleeArgStackSize
  //      | (NumReusableBytes) |                (of tail call)
  //      |                    |                     ---
  //      |                    |                      |
  //      ---------------------|        ---           |
  //      |                    |         |            |
  //      |   CalleeSavedReg   |         |            |
  //      | (CalleeSavedStackSize)|      |            |
  //      |                    |         |            |
  //      ---------------------|         |         NumBytes
  //      |                    |     StackSize  (StackAdjustUp)
  //      |   LocalStackSize   |         |            |
  //      | (covering callee   |         |            |
  //      |       args)        |         |            |
  //      |                    |         |            |
  //      ----------------------        ---          ---
  //
  // So NumBytes = StackSize + BytesInStackArgArea - CalleeArgStackSize
  //             = StackSize + ArgumentPopSize
  //
  // AArch64TargetLowering::LowerCall figures out ArgumentPopSize and keeps
  // it as the 2nd argument of AArch64ISD::TC_RETURN.

  // Move past the restores of the callee-saved registers.
  MachineBasicBlock::iterator LastPopI = MBB.getFirstTerminator();
  MachineBasicBlock::iterator Begin = MBB.begin();
  while (LastPopI != Begin) {
    --LastPopI;
    if (!LastPopI->getFlag(MachineInstr::FrameDestroy)) {
      ++LastPopI;
      break;
    }
  }
  NumBytes -= AFI->getCalleeSavedStackSize();
  assert(NumBytes >= 0 && "Negative stack allocation size!?");

  if (!hasFP(MF)) {
    bool RedZone = canUseRedZone(MF);
    // If this was a redzone leaf function, we don't need to restore the
    // stack pointer (but we may need to pop stack args for fastcc).
    if (RedZone && ArgumentPopSize == 0)
      return;

    bool NoCalleeSaveRestore = AFI->getCalleeSavedStackSize() == 0;
    int StackRestoreBytes = RedZone ? 0 : NumBytes;
    if (NoCalleeSaveRestore)
      StackRestoreBytes += ArgumentPopSize;
    emitFrameOffset(MBB, LastPopI, DL, AArch64::SP, AArch64::SP,
                    StackRestoreBytes, TII, MachineInstr::FrameDestroy);
    // If we were able to combine the local stack pop with the argument pop,
    // then we're done.
    if (NoCalleeSaveRestore || ArgumentPopSize == 0)
      return;
    NumBytes = 0;
  }

//.........这里部分代码省略.........

//.........这里部分代码省略.........
  unsigned Scratch = I->getOperand(3).getReg();

  unsigned Opcode = 0;
  unsigned OR = 0;
  unsigned AND = 0;
  unsigned NOR = 0;
  bool IsNand = false;
  switch (I->getOpcode()) {
  case Mips::ATOMIC_LOAD_ADD_I32_POSTRA:
    Opcode = Mips::ADDu;
    break;
  case Mips::ATOMIC_LOAD_SUB_I32_POSTRA:
    Opcode = Mips::SUBu;
    break;
  case Mips::ATOMIC_LOAD_AND_I32_POSTRA:
    Opcode = Mips::AND;
    break;
  case Mips::ATOMIC_LOAD_OR_I32_POSTRA:
    Opcode = Mips::OR;
    break;
  case Mips::ATOMIC_LOAD_XOR_I32_POSTRA:
    Opcode = Mips::XOR;
    break;
  case Mips::ATOMIC_LOAD_NAND_I32_POSTRA:
    IsNand = true;
    AND = Mips::AND;
    NOR = Mips::NOR;
    break;
  case Mips::ATOMIC_SWAP_I32_POSTRA:
    OR = Mips::OR;
    break;
  case Mips::ATOMIC_LOAD_ADD_I64_POSTRA:
    Opcode = Mips::DADDu;
    break;
  case Mips::ATOMIC_LOAD_SUB_I64_POSTRA:
    Opcode = Mips::DSUBu;
    break;
  case Mips::ATOMIC_LOAD_AND_I64_POSTRA:
    Opcode = Mips::AND64;
    break;
  case Mips::ATOMIC_LOAD_OR_I64_POSTRA:
    Opcode = Mips::OR64;
    break;
  case Mips::ATOMIC_LOAD_XOR_I64_POSTRA:
    Opcode = Mips::XOR64;
    break;
  case Mips::ATOMIC_LOAD_NAND_I64_POSTRA:
    IsNand = true;
    AND = Mips::AND64;
    NOR = Mips::NOR64;
    break;
  case Mips::ATOMIC_SWAP_I64_POSTRA:
    OR = Mips::OR64;
    break;
  default:
    llvm_unreachable("Unknown pseudo atomic!");
  }

  const BasicBlock *LLVM_BB = BB.getBasicBlock();
  MachineBasicBlock *loopMBB = MF->CreateMachineBasicBlock(LLVM_BB);
  MachineBasicBlock *exitMBB = MF->CreateMachineBasicBlock(LLVM_BB);
  MachineFunction::iterator It = ++BB.getIterator();
  MF->insert(It, loopMBB);
  MF->insert(It, exitMBB);

  exitMBB->splice(exitMBB->begin(), &BB, std::next(I), BB.end());
  exitMBB->transferSuccessorsAndUpdatePHIs(&BB);

  BB.addSuccessor(loopMBB, BranchProbability::getOne());
  loopMBB->addSuccessor(exitMBB);
  loopMBB->addSuccessor(loopMBB);
  loopMBB->normalizeSuccProbs();

  BuildMI(loopMBB, DL, TII->get(LL), OldVal).addReg(Ptr).addImm(0);
  assert((OldVal != Ptr) && "Clobbered the wrong ptr reg!");
  assert((OldVal != Incr) && "Clobbered the wrong reg!");
  if (Opcode) {
    BuildMI(loopMBB, DL, TII->get(Opcode), Scratch).addReg(OldVal).addReg(Incr);
  } else if (IsNand) {
    assert(AND && NOR &&
           "Unknown nand instruction for atomic pseudo expansion");
    BuildMI(loopMBB, DL, TII->get(AND), Scratch).addReg(OldVal).addReg(Incr);
    BuildMI(loopMBB, DL, TII->get(NOR), Scratch).addReg(ZERO).addReg(Scratch);
  } else {
    assert(OR && "Unknown instruction for atomic pseudo expansion!");
    BuildMI(loopMBB, DL, TII->get(OR), Scratch).addReg(Incr).addReg(ZERO);
  }

  BuildMI(loopMBB, DL, TII->get(SC), Scratch).addReg(Scratch).addReg(Ptr).addImm(0);
  BuildMI(loopMBB, DL, TII->get(BEQ)).addReg(Scratch).addReg(ZERO).addMBB(loopMBB);

  NMBBI = BB.end();
  I->eraseFromParent();

  LivePhysRegs LiveRegs;
  computeAndAddLiveIns(LiveRegs, *loopMBB);
  computeAndAddLiveIns(LiveRegs, *exitMBB);

  return true;
}

//.........这里部分代码省略.........
  // "zombie" define of that preg. E.g., EFLAGS. (<rdar://problem/8030636>)
  for (unsigned I = 0, E = MI->getNumOperands(); I != E; ++I) {
    const MachineOperand &MO = MI->getOperand(I);
    if (!MO.isReg()) continue;
    unsigned Reg = MO.getReg();
    if (Reg == 0 || !TargetRegisterInfo::isPhysicalRegister(Reg)) continue;
    if (SuccToSinkTo->isLiveIn(Reg))
      return false;
  }

  DEBUG(dbgs() << "Sink instr " << *MI << "\tinto block " << *SuccToSinkTo);

  // If the block has multiple predecessors, this is a critical edge.
  // Decide if we can sink along it or need to break the edge.
  if (SuccToSinkTo->pred_size() > 1) {
    // We cannot sink a load across a critical edge - there may be stores in
    // other code paths.
    bool TryBreak = false;
    bool store = true;
    if (!MI->isSafeToMove(AA, store)) {
      DEBUG(dbgs() << " *** NOTE: Won't sink load along critical edge.\n");
      TryBreak = true;
    }

    // We don't want to sink across a critical edge if we don't dominate the
    // successor. We could be introducing calculations to new code paths.
    if (!TryBreak && !DT->dominates(ParentBlock, SuccToSinkTo)) {
      DEBUG(dbgs() << " *** NOTE: Critical edge found\n");
      TryBreak = true;
    }

    // Don't sink instructions into a loop.
    if (!TryBreak && LI->isLoopHeader(SuccToSinkTo)) {
      DEBUG(dbgs() << " *** NOTE: Loop header found\n");
      TryBreak = true;
    }

    // Otherwise we are OK with sinking along a critical edge.
    if (!TryBreak)
      DEBUG(dbgs() << "Sinking along critical edge.\n");
    else {
      // Mark this edge as to be split.
      // If the edge can actually be split, the next iteration of the main loop
      // will sink MI in the newly created block.
      bool Status =
        PostponeSplitCriticalEdge(MI, ParentBlock, SuccToSinkTo, BreakPHIEdge);
      if (!Status)
        DEBUG(dbgs() << " *** PUNTING: Not legal or profitable to "
              "break critical edge\n");
      // The instruction will not be sunk this time.
      return false;
    }
  }

  if (BreakPHIEdge) {
    // BreakPHIEdge is true if all the uses are in the successor MBB being
    // sunken into and they are all PHI nodes. In this case, machine-sink must
    // break the critical edge first.
    bool Status = PostponeSplitCriticalEdge(MI, ParentBlock,
                                            SuccToSinkTo, BreakPHIEdge);
    if (!Status)
      DEBUG(dbgs() << " *** PUNTING: Not legal or profitable to "
            "break critical edge\n");
    // The instruction will not be sunk this time.
    return false;
  }

  // Determine where to insert into. Skip phi nodes.
  MachineBasicBlock::iterator InsertPos = SuccToSinkTo->begin();
  while (InsertPos != SuccToSinkTo->end() && InsertPos->isPHI())
    ++InsertPos;

  // collect matching debug values.
  SmallVector<MachineInstr *, 2> DbgValuesToSink;
  collectDebugValues(MI, DbgValuesToSink);

  // Move the instruction.
  SuccToSinkTo->splice(InsertPos, ParentBlock, MI,
                       ++MachineBasicBlock::iterator(MI));

  // Move debug values.
  for (SmallVectorImpl<MachineInstr *>::iterator DBI = DbgValuesToSink.begin(),
         DBE = DbgValuesToSink.end(); DBI != DBE; ++DBI) {
    MachineInstr *DbgMI = *DBI;
    SuccToSinkTo->splice(InsertPos, ParentBlock,  DbgMI,
                         ++MachineBasicBlock::iterator(DbgMI));
  }

  // Conservatively, clear any kill flags, since it's possible that they are no
  // longer correct.
  // Note that we have to clear the kill flags for any register this instruction
  // uses as we may sink over another instruction which currently kills the
  // used registers.
  for (MachineOperand &MO : MI->operands()) {
    if (MO.isReg() && MO.isUse())
      RegsToClearKillFlags.set(MO.getReg()); // Remember to clear kill flags.
  }

  return true;
}

bool MipsExpandPseudo::expandAtomicCmpSwapSubword(
    MachineBasicBlock &BB, MachineBasicBlock::iterator I,
    MachineBasicBlock::iterator &NMBBI) {

  MachineFunction *MF = BB.getParent();

  const bool ArePtrs64bit = STI->getABI().ArePtrs64bit();
  DebugLoc DL = I->getDebugLoc();
  unsigned LL, SC;

  unsigned ZERO = Mips::ZERO;
  unsigned BNE = Mips::BNE;
  unsigned BEQ = Mips::BEQ;
  unsigned SEOp =
      I->getOpcode() == Mips::ATOMIC_CMP_SWAP_I8_POSTRA ? Mips::SEB : Mips::SEH;

  if (STI->inMicroMipsMode()) {
      LL = STI->hasMips32r6() ? Mips::LL_MMR6 : Mips::LL_MM;
      SC = STI->hasMips32r6() ? Mips::SC_MMR6 : Mips::SC_MM;
      BNE = STI->hasMips32r6() ? Mips::BNEC_MMR6 : Mips::BNE_MM;
      BEQ = STI->hasMips32r6() ? Mips::BEQC_MMR6 : Mips::BEQ_MM;
  } else {
    LL = STI->hasMips32r6() ? (ArePtrs64bit ? Mips::LL64_R6 : Mips::LL_R6)
                            : (ArePtrs64bit ? Mips::LL64 : Mips::LL);
    SC = STI->hasMips32r6() ? (ArePtrs64bit ? Mips::SC64_R6 : Mips::SC_R6)
                            : (ArePtrs64bit ? Mips::SC64 : Mips::SC);
  }

  unsigned Dest = I->getOperand(0).getReg();
  unsigned Ptr = I->getOperand(1).getReg();
  unsigned Mask = I->getOperand(2).getReg();
  unsigned ShiftCmpVal = I->getOperand(3).getReg();
  unsigned Mask2 = I->getOperand(4).getReg();
  unsigned ShiftNewVal = I->getOperand(5).getReg();
  unsigned ShiftAmnt = I->getOperand(6).getReg();
  unsigned Scratch = I->getOperand(7).getReg();
  unsigned Scratch2 = I->getOperand(8).getReg();

  // insert new blocks after the current block
  const BasicBlock *LLVM_BB = BB.getBasicBlock();
  MachineBasicBlock *loop1MBB = MF->CreateMachineBasicBlock(LLVM_BB);
  MachineBasicBlock *loop2MBB = MF->CreateMachineBasicBlock(LLVM_BB);
  MachineBasicBlock *sinkMBB = MF->CreateMachineBasicBlock(LLVM_BB);
  MachineBasicBlock *exitMBB = MF->CreateMachineBasicBlock(LLVM_BB);
  MachineFunction::iterator It = ++BB.getIterator();
  MF->insert(It, loop1MBB);
  MF->insert(It, loop2MBB);
  MF->insert(It, sinkMBB);
  MF->insert(It, exitMBB);

  // Transfer the remainder of BB and its successor edges to exitMBB.
  exitMBB->splice(exitMBB->begin(), &BB,
                  std::next(MachineBasicBlock::iterator(I)), BB.end());
  exitMBB->transferSuccessorsAndUpdatePHIs(&BB);

  //  thisMBB:
  //    ...
  //    fallthrough --> loop1MBB
  BB.addSuccessor(loop1MBB, BranchProbability::getOne());
  loop1MBB->addSuccessor(sinkMBB);
  loop1MBB->addSuccessor(loop2MBB);
  loop1MBB->normalizeSuccProbs();
  loop2MBB->addSuccessor(loop1MBB);
  loop2MBB->addSuccessor(sinkMBB);
  loop2MBB->normalizeSuccProbs();
  sinkMBB->addSuccessor(exitMBB, BranchProbability::getOne());

  // loop1MBB:
  //   ll dest, 0(ptr)
  //   and Mask', dest, Mask
  //   bne Mask', ShiftCmpVal, exitMBB
  BuildMI(loop1MBB, DL, TII->get(LL), Scratch).addReg(Ptr).addImm(0);
  BuildMI(loop1MBB, DL, TII->get(Mips::AND), Scratch2)
      .addReg(Scratch)
      .addReg(Mask);
  BuildMI(loop1MBB, DL, TII->get(BNE))
    .addReg(Scratch2).addReg(ShiftCmpVal).addMBB(sinkMBB);

  // loop2MBB:
  //   and dest, dest, mask2
  //   or dest, dest, ShiftNewVal
  //   sc dest, dest, 0(ptr)
  //   beq dest, $0, loop1MBB
  BuildMI(loop2MBB, DL, TII->get(Mips::AND), Scratch)
      .addReg(Scratch, RegState::Kill)
      .addReg(Mask2);
  BuildMI(loop2MBB, DL, TII->get(Mips::OR), Scratch)
      .addReg(Scratch, RegState::Kill)
      .addReg(ShiftNewVal);
  BuildMI(loop2MBB, DL, TII->get(SC), Scratch)
      .addReg(Scratch, RegState::Kill)
      .addReg(Ptr)
      .addImm(0);
  BuildMI(loop2MBB, DL, TII->get(BEQ))
      .addReg(Scratch, RegState::Kill)
      .addReg(ZERO)
      .addMBB(loop1MBB);

  //  sinkMBB:
  //    srl     srlres, Mask', shiftamt
//.........这里部分代码省略.........