X86InstrInfo.cpp

      unsigned leaInReg = RegInfo.createVirtualRegister(&X86::GR32RegClass);
      unsigned leaOutReg = RegInfo.createVirtualRegister(&X86::GR32RegClass);
            
      // Build and insert into an implicit UNDEF value. This is OK because
      // well be shifting and then extracting the lower 16-bits. 
      MachineInstr *Undef = BuildMI(get(X86::IMPLICIT_DEF), leaInReg);
      
      MachineInstr *Ins = 
       BuildMI(get(X86::INSERT_SUBREG),leaInReg)
                    .addReg(leaInReg).addReg(Src).addImm(X86::SUBREG_16BIT);
      
      NewMI = BuildMI(get(Opc), leaOutReg)
        .addReg(0).addImm(1 << ShAmt).addReg(leaInReg).addImm(0);
      
      MachineInstr *Ext =
        BuildMI(get(X86::EXTRACT_SUBREG), Dest)
         .addReg(leaOutReg).addImm(X86::SUBREG_16BIT);
      Ext->copyKillDeadInfo(MI);
      
      MFI->insert(MBBI, Undef);
      MFI->insert(MBBI, Ins);            // Insert the insert_subreg
      LV.instructionChanged(MI, NewMI);  // Update live variables
      LV.addVirtualRegisterKilled(leaInReg, NewMI);
      MFI->insert(MBBI, NewMI);          // Insert the new inst
      LV.addVirtualRegisterKilled(leaOutReg, Ext);
      MFI->insert(MBBI, Ext);            // Insert the extract_subreg      
      return Ext;
    } else {
      NewMI = BuildMI(get(X86::LEA16r), Dest)
        .addReg(0).addImm(1 << ShAmt).addReg(Src).addImm(0);
    }
    break;
  }
  default: {
    // The following opcodes also sets the condition code register(s). Only
    // convert them to equivalent lea if the condition code register def's
    // are dead!
    if (hasLiveCondCodeDef(MI))
      return 0;

    bool is64Bit = TM.getSubtarget<X86Subtarget>().is64Bit();
    switch (MIOpc) {
    default: return 0;
    case X86::INC64r:
    case X86::INC32r: {
      assert(MI->getNumOperands() >= 2 && "Unknown inc instruction!");
      unsigned Opc = MIOpc == X86::INC64r ? X86::LEA64r
        : (is64Bit ? X86::LEA64_32r : X86::LEA32r);
      NewMI = addRegOffset(BuildMI(get(Opc), Dest), Src, 1);
      break;
    }
    case X86::INC16r:
    case X86::INC64_16r:
      if (DisableLEA16) return 0;
      assert(MI->getNumOperands() >= 2 && "Unknown inc instruction!");
      NewMI = addRegOffset(BuildMI(get(X86::LEA16r), Dest), Src, 1);
      break;
    case X86::DEC64r:
    case X86::DEC32r: {
      assert(MI->getNumOperands() >= 2 && "Unknown dec instruction!");
      unsigned Opc = MIOpc == X86::DEC64r ? X86::LEA64r
        : (is64Bit ? X86::LEA64_32r : X86::LEA32r);
      NewMI = addRegOffset(BuildMI(get(Opc), Dest), Src, -1);
      break;
    }
    case X86::DEC16r:
    case X86::DEC64_16r:
      if (DisableLEA16) return 0;
      assert(MI->getNumOperands() >= 2 && "Unknown dec instruction!");
      NewMI = addRegOffset(BuildMI(get(X86::LEA16r), Dest), Src, -1);
      break;
    case X86::ADD64rr:
    case X86::ADD32rr: {
      assert(MI->getNumOperands() >= 3 && "Unknown add instruction!");
      unsigned Opc = MIOpc == X86::ADD64rr ? X86::LEA64r
        : (is64Bit ? X86::LEA64_32r : X86::LEA32r);
      NewMI = addRegReg(BuildMI(get(Opc), Dest), Src,
                        MI->getOperand(2).getReg());
      break;
    }
    case X86::ADD16rr:
      if (DisableLEA16) return 0;
      assert(MI->getNumOperands() >= 3 && "Unknown add instruction!");
      NewMI = addRegReg(BuildMI(get(X86::LEA16r), Dest), Src,
                        MI->getOperand(2).getReg());
      break;
    case X86::ADD64ri32:
    case X86::ADD64ri8:
      assert(MI->getNumOperands() >= 3 && "Unknown add instruction!");
      if (MI->getOperand(2).isImmediate())
        NewMI = addRegOffset(BuildMI(get(X86::LEA64r), Dest), Src,
                             MI->getOperand(2).getImm());
      break;
    case X86::ADD32ri:
    case X86::ADD32ri8:
      assert(MI->getNumOperands() >= 3 && "Unknown add instruction!");
      if (MI->getOperand(2).isImmediate()) {
        unsigned Opc = is64Bit ? X86::LEA64_32r : X86::LEA32r;
        NewMI = addRegOffset(BuildMI(get(Opc), Dest), Src,
                             MI->getOperand(2).getImm());
      }
      break;
    case X86::ADD16ri:
    case X86::ADD16ri8:
      if (DisableLEA16) return 0;
      assert(MI->getNumOperands() >= 3 && "Unknown add instruction!");
      if (MI->getOperand(2).isImmediate())
        NewMI = addRegOffset(BuildMI(get(X86::LEA16r), Dest), Src,
                             MI->getOperand(2).getImm());
      break;
    case X86::SHL16ri:
      if (DisableLEA16) return 0;
    case X86::SHL32ri:
    case X86::SHL64ri: {
      assert(MI->getNumOperands() >= 3 && MI->getOperand(2).isImmediate() &&
             "Unknown shl instruction!");
      unsigned ShAmt = MI->getOperand(2).getImm();
      if (ShAmt == 1 || ShAmt == 2 || ShAmt == 3) {
        X86AddressMode AM;
        AM.Scale = 1 << ShAmt;
        AM.IndexReg = Src;
        unsigned Opc = MIOpc == X86::SHL64ri ? X86::LEA64r
          : (MIOpc == X86::SHL32ri
             ? (is64Bit ? X86::LEA64_32r : X86::LEA32r) : X86::LEA16r);
        NewMI = addFullAddress(BuildMI(get(Opc), Dest), AM);
      }
      break;
    }
    }
  }
  }

  if (!NewMI) return 0;

  NewMI->copyKillDeadInfo(MI);
  LV.instructionChanged(MI, NewMI);  // Update live variables
  MFI->insert(MBBI, NewMI);          // Insert the new inst    
  return NewMI;
}

/// commuteInstruction - We have a few instructions that must be hacked on to
/// commute them.
///
MachineInstr *X86InstrInfo::commuteInstruction(MachineInstr *MI) const {
  switch (MI->getOpcode()) {
  case X86::SHRD16rri8: // A = SHRD16rri8 B, C, I -> A = SHLD16rri8 C, B, (16-I)
  case X86::SHLD16rri8: // A = SHLD16rri8 B, C, I -> A = SHRD16rri8 C, B, (16-I)
  case X86::SHRD32rri8: // A = SHRD32rri8 B, C, I -> A = SHLD32rri8 C, B, (32-I)
  case X86::SHLD32rri8: // A = SHLD32rri8 B, C, I -> A = SHRD32rri8 C, B, (32-I)
  case X86::SHRD64rri8: // A = SHRD64rri8 B, C, I -> A = SHLD64rri8 C, B, (64-I)
  case X86::SHLD64rri8:{// A = SHLD64rri8 B, C, I -> A = SHRD64rri8 C, B, (64-I)
    unsigned Opc;
    unsigned Size;
    switch (MI->getOpcode()) {
    default: assert(0 && "Unreachable!");
    case X86::SHRD16rri8: Size = 16; Opc = X86::SHLD16rri8; break;
    case X86::SHLD16rri8: Size = 16; Opc = X86::SHRD16rri8; break;
    case X86::SHRD32rri8: Size = 32; Opc = X86::SHLD32rri8; break;
    case X86::SHLD32rri8: Size = 32; Opc = X86::SHRD32rri8; break;
    case X86::SHRD64rri8: Size = 64; Opc = X86::SHLD64rri8; break;
    case X86::SHLD64rri8: Size = 64; Opc = X86::SHRD64rri8; break;
    }
    unsigned Amt = MI->getOperand(3).getImm();
    unsigned A = MI->getOperand(0).getReg();
    unsigned B = MI->getOperand(1).getReg();
    unsigned C = MI->getOperand(2).getReg();
    bool BisKill = MI->getOperand(1).isKill();
    bool CisKill = MI->getOperand(2).isKill();
    // If machine instrs are no longer in two-address forms, update
    // destination register as well.
    if (A == B) {
      // Must be two address instruction!
      assert(MI->getDesc().getOperandConstraint(0, TOI::TIED_TO) &&
             "Expecting a two-address instruction!");
      A = C;
      CisKill = false;
    }
    return BuildMI(get(Opc), A).addReg(C, false, false, CisKill)
      .addReg(B, false, false, BisKill).addImm(Size-Amt);
  }
  case X86::CMOVB16rr:
  case X86::CMOVB32rr:
  case X86::CMOVB64rr:
  case X86::CMOVAE16rr:
  case X86::CMOVAE32rr:
  case X86::CMOVAE64rr:
  case X86::CMOVE16rr:
  case X86::CMOVE32rr:
  case X86::CMOVE64rr:
  case X86::CMOVNE16rr:
  case X86::CMOVNE32rr:
  case X86::CMOVNE64rr:
  case X86::CMOVBE16rr:
  case X86::CMOVBE32rr:
  case X86::CMOVBE64rr:
  case X86::CMOVA16rr:
  case X86::CMOVA32rr:
  case X86::CMOVA64rr:
  case X86::CMOVL16rr:
  case X86::CMOVL32rr:
  case X86::CMOVL64rr:
  case X86::CMOVGE16rr:
  case X86::CMOVGE32rr:
  case X86::CMOVGE64rr:
  case X86::CMOVLE16rr:
  case X86::CMOVLE32rr:
  case X86::CMOVLE64rr:
  case X86::CMOVG16rr:
  case X86::CMOVG32rr:
  case X86::CMOVG64rr:
  case X86::CMOVS16rr:
  case X86::CMOVS32rr:
  case X86::CMOVS64rr:
  case X86::CMOVNS16rr:
  case X86::CMOVNS32rr:
  case X86::CMOVNS64rr:
  case X86::CMOVP16rr:
  case X86::CMOVP32rr:
  case X86::CMOVP64rr:
  case X86::CMOVNP16rr:
  case X86::CMOVNP32rr:
  case X86::CMOVNP64rr: {
    unsigned Opc = 0;
    switch (MI->getOpcode()) {
    default: break;
    case X86::CMOVB16rr:  Opc = X86::CMOVAE16rr; break;
    case X86::CMOVB32rr:  Opc = X86::CMOVAE32rr; break;
    case X86::CMOVB64rr:  Opc = X86::CMOVAE64rr; break;
    case X86::CMOVAE16rr: Opc = X86::CMOVB16rr; break;
    case X86::CMOVAE32rr: Opc = X86::CMOVB32rr; break;
    case X86::CMOVAE64rr: Opc = X86::CMOVB64rr; break;
    case X86::CMOVE16rr:  Opc = X86::CMOVNE16rr; break;
    case X86::CMOVE32rr:  Opc = X86::CMOVNE32rr; break;
    case X86::CMOVE64rr:  Opc = X86::CMOVNE64rr; break;
    case X86::CMOVNE16rr: Opc = X86::CMOVE16rr; break;
    case X86::CMOVNE32rr: Opc = X86::CMOVE32rr; break;
    case X86::CMOVNE64rr: Opc = X86::CMOVE64rr; break;
    case X86::CMOVBE16rr: Opc = X86::CMOVA16rr; break;
    case X86::CMOVBE32rr: Opc = X86::CMOVA32rr; break;
    case X86::CMOVBE64rr: Opc = X86::CMOVA64rr; break;
    case X86::CMOVA16rr:  Opc = X86::CMOVBE16rr; break;
    case X86::CMOVA32rr:  Opc = X86::CMOVBE32rr; break;
    case X86::CMOVA64rr:  Opc = X86::CMOVBE64rr; break;
    case X86::CMOVL16rr:  Opc = X86::CMOVGE16rr; break;
    case X86::CMOVL32rr:  Opc = X86::CMOVGE32rr; break;
    case X86::CMOVL64rr:  Opc = X86::CMOVGE64rr; break;
    case X86::CMOVGE16rr: Opc = X86::CMOVL16rr; break;
    case X86::CMOVGE32rr: Opc = X86::CMOVL32rr; break;
    case X86::CMOVGE64rr: Opc = X86::CMOVL64rr; break;
    case X86::CMOVLE16rr: Opc = X86::CMOVG16rr; break;
    case X86::CMOVLE32rr: Opc = X86::CMOVG32rr; break;
    case X86::CMOVLE64rr: Opc = X86::CMOVG64rr; break;
    case X86::CMOVG16rr:  Opc = X86::CMOVLE16rr; break;
    case X86::CMOVG32rr:  Opc = X86::CMOVLE32rr; break;
    case X86::CMOVG64rr:  Opc = X86::CMOVLE64rr; break;
    case X86::CMOVS16rr:  Opc = X86::CMOVNS16rr; break;
    case X86::CMOVS32rr:  Opc = X86::CMOVNS32rr; break;
    case X86::CMOVS64rr:  Opc = X86::CMOVNS32rr; break;
    case X86::CMOVNS16rr: Opc = X86::CMOVS16rr; break;
    case X86::CMOVNS32rr: Opc = X86::CMOVS32rr; break;
    case X86::CMOVNS64rr: Opc = X86::CMOVS64rr; break;
    case X86::CMOVP16rr:  Opc = X86::CMOVNP16rr; break;
    case X86::CMOVP32rr:  Opc = X86::CMOVNP32rr; break;
    case X86::CMOVP64rr:  Opc = X86::CMOVNP32rr; break;
    case X86::CMOVNP16rr: Opc = X86::CMOVP16rr; break;
    case X86::CMOVNP32rr: Opc = X86::CMOVP32rr; break;
    case X86::CMOVNP64rr: Opc = X86::CMOVP64rr; break;
    }

    MI->setDesc(get(Opc));
    // Fallthrough intended.
  }
  default:
    return TargetInstrInfoImpl::commuteInstruction(MI);
  }
}

static X86::CondCode GetCondFromBranchOpc(unsigned BrOpc) {
  switch (BrOpc) {
  default: return X86::COND_INVALID;
  case X86::JE:  return X86::COND_E;
  case X86::JNE: return X86::COND_NE;
  case X86::JL:  return X86::COND_L;
  case X86::JLE: return X86::COND_LE;
  case X86::JG:  return X86::COND_G;
  case X86::JGE: return X86::COND_GE;
  case X86::JB:  return X86::COND_B;
  case X86::JBE: return X86::COND_BE;
  case X86::JA:  return X86::COND_A;
  case X86::JAE: return X86::COND_AE;
  case X86::JS:  return X86::COND_S;
  case X86::JNS: return X86::COND_NS;
  case X86::JP:  return X86::COND_P;
  case X86::JNP: return X86::COND_NP;
  case X86::JO:  return X86::COND_O;
  case X86::JNO: return X86::COND_NO;
  }
}

unsigned X86::GetCondBranchFromCond(X86::CondCode CC) {
  switch (CC) {
  default: assert(0 && "Illegal condition code!");
  case X86::COND_E:  return X86::JE;
  case X86::COND_NE: return X86::JNE;
  case X86::COND_L:  return X86::JL;
  case X86::COND_LE: return X86::JLE;
  case X86::COND_G:  return X86::JG;
  case X86::COND_GE: return X86::JGE;
  case X86::COND_B:  return X86::JB;
  case X86::COND_BE: return X86::JBE;
  case X86::COND_A:  return X86::JA;
  case X86::COND_AE: return X86::JAE;
  case X86::COND_S:  return X86::JS;
  case X86::COND_NS: return X86::JNS;
  case X86::COND_P:  return X86::JP;
  case X86::COND_NP: return X86::JNP;
  case X86::COND_O:  return X86::JO;
  case X86::COND_NO: return X86::JNO;
  }
}

/// GetOppositeBranchCondition - Return the inverse of the specified condition,
/// e.g. turning COND_E to COND_NE.
X86::CondCode X86::GetOppositeBranchCondition(X86::CondCode CC) {
  switch (CC) {
  default: assert(0 && "Illegal condition code!");
  case X86::COND_E:  return X86::COND_NE;
  case X86::COND_NE: return X86::COND_E;
  case X86::COND_L:  return X86::COND_GE;
  case X86::COND_LE: return X86::COND_G;
  case X86::COND_G:  return X86::COND_LE;
  case X86::COND_GE: return X86::COND_L;
  case X86::COND_B:  return X86::COND_AE;
  case X86::COND_BE: return X86::COND_A;
  case X86::COND_A:  return X86::COND_BE;
  case X86::COND_AE: return X86::COND_B;
  case X86::COND_S:  return X86::COND_NS;
  case X86::COND_NS: return X86::COND_S;
  case X86::COND_P:  return X86::COND_NP;
  case X86::COND_NP: return X86::COND_P;
  case X86::COND_O:  return X86::COND_NO;
  case X86::COND_NO: return X86::COND_O;
  }
}

bool X86InstrInfo::isUnpredicatedTerminator(const MachineInstr *MI) const {
  const TargetInstrDesc &TID = MI->getDesc();
  if (!TID.isTerminator()) return false;
  
  // Conditional branch is a special case.
  if (TID.isBranch() && !TID.isBarrier())
    return true;
  if (!TID.isPredicable())
    return true;
  return !isPredicated(MI);
}

// For purposes of branch analysis do not count FP_REG_KILL as a terminator.
static bool isBrAnalysisUnpredicatedTerminator(const MachineInstr *MI,
                                               const X86InstrInfo &TII) {
  if (MI->getOpcode() == X86::FP_REG_KILL)
    return false;
  return TII.isUnpredicatedTerminator(MI);
}

bool X86InstrInfo::AnalyzeBranch(MachineBasicBlock &MBB, 
                                 MachineBasicBlock *&TBB,
                                 MachineBasicBlock *&FBB,
                                 std::vector<MachineOperand> &Cond) const {
  // If the block has no terminators, it just falls into the block after it.
  MachineBasicBlock::iterator I = MBB.end();
  if (I == MBB.begin() || !isBrAnalysisUnpredicatedTerminator(--I, *this))
    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() || !isBrAnalysisUnpredicatedTerminator(--I, *this)) {
    if (!LastInst->getDesc().isBranch())
      return true;
    
    // If the block ends with a branch there are 3 possibilities:
    // it's an unconditional, conditional, or indirect branch.
    
    if (LastInst->getOpcode() == X86::JMP) {
      TBB = LastInst->getOperand(0).getMBB();
      return false;
    }
    X86::CondCode BranchCode = GetCondFromBranchOpc(LastInst->getOpcode());
    if (BranchCode == X86::COND_INVALID)
      return true;  // Can't handle indirect branch.

    // Otherwise, block ends with fall-through condbranch.
    TBB = LastInst->getOperand(0).getMBB();
    Cond.push_back(MachineOperand::CreateImm(BranchCode));
    return false;
  }
  
  // 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() &&
      isBrAnalysisUnpredicatedTerminator(--I, *this))
    return true;

  // If the block ends with X86::JMP and a conditional branch, handle it.
  X86::CondCode BranchCode = GetCondFromBranchOpc(SecondLastInst->getOpcode());
  if (BranchCode != X86::COND_INVALID && LastInst->getOpcode() == X86::JMP) {
    TBB = SecondLastInst->getOperand(0).getMBB();
    Cond.push_back(MachineOperand::CreateImm(BranchCode));
    FBB = LastInst->getOperand(0).getMBB();
    return false;
  }

  // If the block ends with two X86::JMPs, handle it.  The second one is not
  // executed, so remove it.
  if (SecondLastInst->getOpcode() == X86::JMP && 
      LastInst->getOpcode() == X86::JMP) {
    TBB = SecondLastInst->getOperand(0).getMBB();
    I = LastInst;
    I->eraseFromParent();
    return false;
  }

  // Otherwise, can't handle this.
  return true;
}

unsigned X86InstrInfo::RemoveBranch(MachineBasicBlock &MBB) const {
  MachineBasicBlock::iterator I = MBB.end();
  if (I == MBB.begin()) return 0;
  --I;
  if (I->getOpcode() != X86::JMP && 
      GetCondFromBranchOpc(I->getOpcode()) == X86::COND_INVALID)
    return 0;
  
  // Remove the branch.
  I->eraseFromParent();
  
  I = MBB.end();
  
  if (I == MBB.begin()) return 1;
  --I;
  if (GetCondFromBranchOpc(I->getOpcode()) == X86::COND_INVALID)
    return 1;
  
  // Remove the branch.
  I->eraseFromParent();
  return 2;
}

static const MachineInstrBuilder &X86InstrAddOperand(MachineInstrBuilder &MIB,
                                                     MachineOperand &MO) {
  if (MO.isRegister())
    MIB = MIB.addReg(MO.getReg(), MO.isDef(), MO.isImplicit(),
                     false, false, MO.getSubReg());
  else if (MO.isImmediate())
    MIB = MIB.addImm(MO.getImm());
  else if (MO.isFrameIndex())
    MIB = MIB.addFrameIndex(MO.getIndex());
  else if (MO.isGlobalAddress())
    MIB = MIB.addGlobalAddress(MO.getGlobal(), MO.getOffset());
  else if (MO.isConstantPoolIndex())
    MIB = MIB.addConstantPoolIndex(MO.getIndex(), MO.getOffset());
  else if (MO.isJumpTableIndex())
    MIB = MIB.addJumpTableIndex(MO.getIndex());
  else if (MO.isExternalSymbol())
    MIB = MIB.addExternalSymbol(MO.getSymbolName());
  else
    assert(0 && "Unknown operand for X86InstrAddOperand!");

  return MIB;
}

unsigned
X86InstrInfo::InsertBranch(MachineBasicBlock &MBB, MachineBasicBlock *TBB,
                           MachineBasicBlock *FBB,
                           const std::vector<MachineOperand> &Cond) const {
  // Shouldn't be a fall through.
  assert(TBB && "InsertBranch must not be told to insert a fallthrough");
  assert((Cond.size() == 1 || Cond.size() == 0) &&
         "X86 branch conditions have one component!");

  if (FBB == 0) { // One way branch.
    if (Cond.empty()) {
      // Unconditional branch?
      BuildMI(&MBB, get(X86::JMP)).addMBB(TBB);
    } else {
      // Conditional branch.
      unsigned Opc = GetCondBranchFromCond((X86::CondCode)Cond[0].getImm());
      BuildMI(&MBB, get(Opc)).addMBB(TBB);
    }
    return 1;
  }
  
  // Two-way Conditional branch.
  unsigned Opc = GetCondBranchFromCond((X86::CondCode)Cond[0].getImm());
  BuildMI(&MBB, get(Opc)).addMBB(TBB);
  BuildMI(&MBB, get(X86::JMP)).addMBB(FBB);
  return 2;
}

void X86InstrInfo::copyRegToReg(MachineBasicBlock &MBB,
                                MachineBasicBlock::iterator MI,
                                unsigned DestReg, unsigned SrcReg,
                                const TargetRegisterClass *DestRC,
                                const TargetRegisterClass *SrcRC) const {
  if (DestRC == SrcRC) {
    unsigned Opc;
    if (DestRC == &X86::GR64RegClass) {
      Opc = X86::MOV64rr;
    } else if (DestRC == &X86::GR32RegClass) {
      Opc = X86::MOV32rr;
    } else if (DestRC == &X86::GR16RegClass) {
      Opc = X86::MOV16rr;
    } else if (DestRC == &X86::GR8RegClass) {
      Opc = X86::MOV8rr;
    } else if (DestRC == &X86::GR32_RegClass) {
      Opc = X86::MOV32_rr;
    } else if (DestRC == &X86::GR16_RegClass) {
      Opc = X86::MOV16_rr;
    } else if (DestRC == &X86::RFP32RegClass) {
      Opc = X86::MOV_Fp3232;
    } else if (DestRC == &X86::RFP64RegClass || DestRC == &X86::RSTRegClass) {
      Opc = X86::MOV_Fp6464;
    } else if (DestRC == &X86::RFP80RegClass) {
      Opc = X86::MOV_Fp8080;
    } else if (DestRC == &X86::FR32RegClass) {
      Opc = X86::FsMOVAPSrr;
    } else if (DestRC == &X86::FR64RegClass) {
      Opc = X86::FsMOVAPDrr;
    } else if (DestRC == &X86::VR128RegClass) {
      Opc = X86::MOVAPSrr;
    } else if (DestRC == &X86::VR64RegClass) {
      Opc = X86::MMX_MOVQ64rr;
    } else {
      assert(0 && "Unknown regclass");
      abort();
    }
    BuildMI(MBB, MI, get(Opc), DestReg).addReg(SrcReg);
    return;
  }
  
  // Moving EFLAGS to / from another register requires a push and a pop.
  if (SrcRC == &X86::CCRRegClass) {
    assert(SrcReg == X86::EFLAGS);
    if (DestRC == &X86::GR64RegClass) {
      BuildMI(MBB, MI, get(X86::PUSHFQ));
      BuildMI(MBB, MI, get(X86::POP64r), DestReg);
      return;
    } else if (DestRC == &X86::GR32RegClass) {
      BuildMI(MBB, MI, get(X86::PUSHFD));
      BuildMI(MBB, MI, get(X86::POP32r), DestReg);
      return;
    }
  } else if (DestRC == &X86::CCRRegClass) {
    assert(DestReg == X86::EFLAGS);
    if (SrcRC == &X86::GR64RegClass) {
      BuildMI(MBB, MI, get(X86::PUSH64r)).addReg(SrcReg);
      BuildMI(MBB, MI, get(X86::POPFQ));
      return;
    } else if (SrcRC == &X86::GR32RegClass) {
      BuildMI(MBB, MI, get(X86::PUSH32r)).addReg(SrcReg);
      BuildMI(MBB, MI, get(X86::POPFD));
      return;
    }
  }
  
  // Moving from ST(0) turns into FpGET_ST0_32 etc.
  if (SrcRC == &X86::RSTRegClass) {
    // Copying from ST(0)/ST(1).
    assert((SrcReg == X86::ST0 || SrcReg == X86::ST1) &&
           "Can only copy from ST(0)/ST(1) right now");
    bool isST0 = SrcReg == X86::ST0;
    unsigned Opc;
    if (DestRC == &X86::RFP32RegClass)
      Opc = isST0 ? X86::FpGET_ST0_32 : X86::FpGET_ST1_32;
    else if (DestRC == &X86::RFP64RegClass)
      Opc = isST0 ? X86::FpGET_ST0_64 : X86::FpGET_ST1_64;
    else {
      assert(DestRC == &X86::RFP80RegClass);
      Opc = isST0 ? X86::FpGET_ST0_80 : X86::FpGET_ST1_80;
    }
    BuildMI(MBB, MI, get(Opc), DestReg);
    return;
  }

  // Moving to ST(0) turns into FpSET_ST0_32 etc.
  if (DestRC == &X86::RSTRegClass) {
    // Copying to ST(0).  FIXME: handle ST(1) also
    assert(DestReg == X86::ST0 && "Can only copy to TOS right now");
    unsigned Opc;
    if (SrcRC == &X86::RFP32RegClass)
      Opc = X86::FpSET_ST0_32;
    else if (SrcRC == &X86::RFP64RegClass)
      Opc = X86::FpSET_ST0_64;
    else {
      assert(SrcRC == &X86::RFP80RegClass);
      Opc = X86::FpSET_ST0_80;
    }
    BuildMI(MBB, MI, get(Opc)).addReg(SrcReg);
    return;
  }
  
  assert(0 && "Not yet supported!");
  abort();
}

static unsigned getStoreRegOpcode(const TargetRegisterClass *RC,
                                  unsigned StackAlign) {
  unsigned Opc = 0;
  if (RC == &X86::GR64RegClass) {
    Opc = X86::MOV64mr;
  } else if (RC == &X86::GR32RegClass) {
    Opc = X86::MOV32mr;
  } else if (RC == &X86::GR16RegClass) {
    Opc = X86::MOV16mr;
  } else if (RC == &X86::GR8RegClass) {
    Opc = X86::MOV8mr;
  } else if (RC == &X86::GR32_RegClass) {
    Opc = X86::MOV32_mr;
  } else if (RC == &X86::GR16_RegClass) {
    Opc = X86::MOV16_mr;
  } else if (RC == &X86::RFP80RegClass) {
    Opc = X86::ST_FpP80m;   // pops
  } else if (RC == &X86::RFP64RegClass) {
    Opc = X86::ST_Fp64m;
  } else if (RC == &X86::RFP32RegClass) {
    Opc = X86::ST_Fp32m;
  } else if (RC == &X86::FR32RegClass) {
    Opc = X86::MOVSSmr;
  } else if (RC == &X86::FR64RegClass) {
    Opc = X86::MOVSDmr;
  } else if (RC == &X86::VR128RegClass) {
    // FIXME: Use movaps once we are capable of selectively
    // aligning functions that spill SSE registers on 16-byte boundaries.
    Opc = StackAlign >= 16 ? X86::MOVAPSmr : X86::MOVUPSmr;
  } else if (RC == &X86::VR64RegClass) {
    Opc = X86::MMX_MOVQ64mr;
  } else {
    assert(0 && "Unknown regclass");
    abort();
  }

  return Opc;
}

void X86InstrInfo::storeRegToStackSlot(MachineBasicBlock &MBB,
                                       MachineBasicBlock::iterator MI,
                                       unsigned SrcReg, bool isKill, int FrameIdx,
                                       const TargetRegisterClass *RC) const {
  unsigned Opc = getStoreRegOpcode(RC, RI.getStackAlignment());
  addFrameReference(BuildMI(MBB, MI, get(Opc)), FrameIdx)
    .addReg(SrcReg, false, false, isKill);
}

void X86InstrInfo::storeRegToAddr(MachineFunction &MF, unsigned SrcReg,
                                  bool isKill,
                                  SmallVectorImpl<MachineOperand> &Addr,
                                  const TargetRegisterClass *RC,
                                  SmallVectorImpl<MachineInstr*> &NewMIs) const {
  unsigned Opc = getStoreRegOpcode(RC, RI.getStackAlignment());
  MachineInstrBuilder MIB = BuildMI(get(Opc));
  for (unsigned i = 0, e = Addr.size(); i != e; ++i)
    MIB = X86InstrAddOperand(MIB, Addr[i]);
  MIB.addReg(SrcReg, false, false, isKill);
  NewMIs.push_back(MIB);
}

static unsigned getLoadRegOpcode(const TargetRegisterClass *RC,
                                 unsigned StackAlign) {
  unsigned Opc = 0;
  if (RC == &X86::GR64RegClass) {
    Opc = X86::MOV64rm;
  } else if (RC == &X86::GR32RegClass) {
    Opc = X86::MOV32rm;
  } else if (RC == &X86::GR16RegClass) {
    Opc = X86::MOV16rm;
  } else if (RC == &X86::GR8RegClass) {
    Opc = X86::MOV8rm;
  } else if (RC == &X86::GR32_RegClass) {
    Opc = X86::MOV32_rm;
  } else if (RC == &X86::GR16_RegClass) {
    Opc = X86::MOV16_rm;
  } else if (RC == &X86::RFP80RegClass) {
    Opc = X86::LD_Fp80m;
  } else if (RC == &X86::RFP64RegClass) {
    Opc = X86::LD_Fp64m;
  } else if (RC == &X86::RFP32RegClass) {
    Opc = X86::LD_Fp32m;
  } else if (RC == &X86::FR32RegClass) {
    Opc = X86::MOVSSrm;
  } else if (RC == &X86::FR64RegClass) {
    Opc = X86::MOVSDrm;
  } else if (RC == &X86::VR128RegClass) {
    // FIXME: Use movaps once we are capable of selectively
    // aligning functions that spill SSE registers on 16-byte boundaries.
    Opc = StackAlign >= 16 ? X86::MOVAPSrm : X86::MOVUPSrm;
  } else if (RC == &X86::VR64RegClass) {
    Opc = X86::MMX_MOVQ64rm;
  } else {
    assert(0 && "Unknown regclass");
    abort();
  }

  return Opc;
}

void X86InstrInfo::loadRegFromStackSlot(MachineBasicBlock &MBB,
                                           MachineBasicBlock::iterator MI,
                                           unsigned DestReg, int FrameIdx,
                                           const TargetRegisterClass *RC) const{
  unsigned Opc = getLoadRegOpcode(RC, RI.getStackAlignment());
  addFrameReference(BuildMI(MBB, MI, get(Opc), DestReg), FrameIdx);
}

void X86InstrInfo::loadRegFromAddr(MachineFunction &MF, unsigned DestReg,
                                      SmallVectorImpl<MachineOperand> &Addr,
                                      const TargetRegisterClass *RC,
                                 SmallVectorImpl<MachineInstr*> &NewMIs) const {
  unsigned Opc = getLoadRegOpcode(RC, RI.getStackAlignment());
  MachineInstrBuilder MIB = BuildMI(get(Opc), DestReg);
  for (unsigned i = 0, e = Addr.size(); i != e; ++i)
    MIB = X86InstrAddOperand(MIB, Addr[i]);
  NewMIs.push_back(MIB);
}

bool X86InstrInfo::spillCalleeSavedRegisters(MachineBasicBlock &MBB,
                                                MachineBasicBlock::iterator MI,
                                const std::vector<CalleeSavedInfo> &CSI) const {
  if (CSI.empty())
    return false;

  bool is64Bit = TM.getSubtarget<X86Subtarget>().is64Bit();
  unsigned SlotSize = is64Bit ? 8 : 4;

  MachineFunction &MF = *MBB.getParent();
  X86MachineFunctionInfo *X86FI = MF.getInfo<X86MachineFunctionInfo>();
  X86FI->setCalleeSavedFrameSize(CSI.size() * SlotSize);
  
  unsigned Opc = is64Bit ? X86::PUSH64r : X86::PUSH32r;
  for (unsigned i = CSI.size(); i != 0; --i) {
    unsigned Reg = CSI[i-1].getReg();
    // Add the callee-saved register as live-in. It's killed at the spill.
    MBB.addLiveIn(Reg);
    BuildMI(MBB, MI, get(Opc)).addReg(Reg);
  }
  return true;
}

bool X86InstrInfo::restoreCalleeSavedRegisters(MachineBasicBlock &MBB,
                                                 MachineBasicBlock::iterator MI,
                                const std::vector<CalleeSavedInfo> &CSI) const {
  if (CSI.empty())
    return false;
    
  bool is64Bit = TM.getSubtarget<X86Subtarget>().is64Bit();

  unsigned Opc = is64Bit ? X86::POP64r : X86::POP32r;
  for (unsigned i = 0, e = CSI.size(); i != e; ++i) {
    unsigned Reg = CSI[i].getReg();
    BuildMI(MBB, MI, get(Opc), Reg);
  }
  return true;
}

static MachineInstr *FuseTwoAddrInst(unsigned Opcode,
                                     SmallVector<MachineOperand,4> &MOs,
                                 MachineInstr *MI, const TargetInstrInfo &TII) {
  // Create the base instruction with the memory operand as the first part.
  MachineInstr *NewMI = new MachineInstr(TII.get(Opcode), true);
  MachineInstrBuilder MIB(NewMI);
  unsigned NumAddrOps = MOs.size();
  for (unsigned i = 0; i != NumAddrOps; ++i)
    MIB = X86InstrAddOperand(MIB, MOs[i]);
  if (NumAddrOps < 4)  // FrameIndex only
    MIB.addImm(1).addReg(0).addImm(0);
  
  // Loop over the rest of the ri operands, converting them over.
  unsigned NumOps = MI->getDesc().getNumOperands()-2;
  for (unsigned i = 0; i != NumOps; ++i) {
    MachineOperand &MO = MI->getOperand(i+2);
    MIB = X86InstrAddOperand(MIB, MO);
  }
  for (unsigned i = NumOps+2, e = MI->getNumOperands(); i != e; ++i) {
    MachineOperand &MO = MI->getOperand(i);
    MIB = X86InstrAddOperand(MIB, MO);
  }
  return MIB;
}

static MachineInstr *FuseInst(unsigned Opcode, unsigned OpNo,
                              SmallVector<MachineOperand,4> &MOs,
                              MachineInstr *MI, const TargetInstrInfo &TII) {
  MachineInstr *NewMI = new MachineInstr(TII.get(Opcode), true);
  MachineInstrBuilder MIB(NewMI);
  
  for (unsigned i = 0, e = MI->getNumOperands(); i != e; ++i) {
    MachineOperand &MO = MI->getOperand(i);
    if (i == OpNo) {
      assert(MO.isRegister() && "Expected to fold into reg operand!");
      unsigned NumAddrOps = MOs.size();
      for (unsigned i = 0; i != NumAddrOps; ++i)
        MIB = X86InstrAddOperand(MIB, MOs[i]);
      if (NumAddrOps < 4)  // FrameIndex only
        MIB.addImm(1).addReg(0).addImm(0);
    } else {
      MIB = X86InstrAddOperand(MIB, MO);
    }
  }
  return MIB;
}

static MachineInstr *MakeM0Inst(const TargetInstrInfo &TII, unsigned Opcode,
                                SmallVector<MachineOperand,4> &MOs,
                                MachineInstr *MI) {
  MachineInstrBuilder MIB = BuildMI(TII.get(Opcode));

  unsigned NumAddrOps = MOs.size();
  for (unsigned i = 0; i != NumAddrOps; ++i)
    MIB = X86InstrAddOperand(MIB, MOs[i]);
  if (NumAddrOps < 4)  // FrameIndex only
    MIB.addImm(1).addReg(0).addImm(0);
  return MIB.addImm(0);
}

MachineInstr*
X86InstrInfo::foldMemoryOperand(MachineInstr *MI, unsigned i,
                                SmallVector<MachineOperand,4> &MOs) const {
  const DenseMap<unsigned*, unsigned> *OpcodeTablePtr = NULL;
  bool isTwoAddrFold = false;
  unsigned NumOps = MI->getDesc().getNumOperands();
  bool isTwoAddr = NumOps > 1 &&
    MI->getDesc().getOperandConstraint(1, TOI::TIED_TO) != -1;

  MachineInstr *NewMI = NULL;
  // Folding a memory location into the two-address part of a two-address
  // instruction is different than folding it other places.  It requires
  // replacing the *two* registers with the memory location.
  if (isTwoAddr && NumOps >= 2 && i < 2 &&
      MI->getOperand(0).isRegister() && 
      MI->getOperand(1).isRegister() &&
      MI->getOperand(0).getReg() == MI->getOperand(1).getReg()) { 
    OpcodeTablePtr = &RegOp2MemOpTable2Addr;
    isTwoAddrFold = true;
  } else if (i == 0) { // If operand 0
    if (MI->getOpcode() == X86::MOV16r0)
      NewMI = MakeM0Inst(*this, X86::MOV16mi, MOs, MI);
    else if (MI->getOpcode() == X86::MOV32r0)
      NewMI = MakeM0Inst(*this, X86::MOV32mi, MOs, MI);
    else if (MI->getOpcode() == X86::MOV64r0)
      NewMI = MakeM0Inst(*this, X86::MOV64mi32, MOs, MI);
    else if (MI->getOpcode() == X86::MOV8r0)
      NewMI = MakeM0Inst(*this, X86::MOV8mi, MOs, MI);
    if (NewMI) {
      NewMI->copyKillDeadInfo(MI);
      return NewMI;
    }
    
    OpcodeTablePtr = &RegOp2MemOpTable0;
  } else if (i == 1) {
    OpcodeTablePtr = &RegOp2MemOpTable1;
  } else if (i == 2) {
    OpcodeTablePtr = &RegOp2MemOpTable2;
  }
  
  // If table selected...
  if (OpcodeTablePtr) {
    // Find the Opcode to fuse
    DenseMap<unsigned*, unsigned>::iterator I =
      OpcodeTablePtr->find((unsigned*)MI->getOpcode());
    if (I != OpcodeTablePtr->end()) {
      if (isTwoAddrFold)
        NewMI = FuseTwoAddrInst(I->second, MOs, MI, *this);
      else
        NewMI = FuseInst(I->second, i, MOs, MI, *this);
      NewMI->copyKillDeadInfo(MI);
      return NewMI;
    }
  }
  
  // No fusion 
  if (PrintFailedFusing)
    cerr << "We failed to fuse operand " << i << *MI;
  return NULL;
}


MachineInstr* X86InstrInfo::foldMemoryOperand(MachineFunction &MF,
                                              MachineInstr *MI,
                                              SmallVectorImpl<unsigned> &Ops,
                                              int FrameIndex) const {
  // Check switch flag 
  if (NoFusing) return NULL;

  const MachineFrameInfo *MFI = MF.getFrameInfo();
  unsigned Alignment = MFI->getObjectAlignment(FrameIndex);
  // FIXME: Move alignment requirement into tables?
  if (Alignment < 16) {
    switch (MI->getOpcode()) {
    default: break;
    // Not always safe to fold movsd into these instructions since their load
    // folding variants expects the address to be 16 byte aligned.
    case X86::FsANDNPDrr:
    case X86::FsANDNPSrr:
    case X86::FsANDPDrr:
    case X86::FsANDPSrr:
    case X86::FsORPDrr:
    case X86::FsORPSrr:
    case X86::FsXORPDrr:
    case X86::FsXORPSrr:
      return NULL;
    }
  }

  if (Ops.size() == 2 && Ops[0] == 0 && Ops[1] == 1) {
    unsigned NewOpc = 0;
    switch (MI->getOpcode()) {
    default: return NULL;
    case X86::TEST8rr:  NewOpc = X86::CMP8ri; break;
    case X86::TEST16rr: NewOpc = X86::CMP16ri; break;
    case X86::TEST32rr: NewOpc = X86::CMP32ri; break;
    case X86::TEST64rr: NewOpc = X86::CMP64ri32; break;
    }
    // Change to CMPXXri r, 0 first.
    MI->setDesc(get(NewOpc));
    MI->getOperand(1).ChangeToImmediate(0);
  } else if (Ops.size() != 1)
    return NULL;

  SmallVector<MachineOperand,4> MOs;
  MOs.push_back(MachineOperand::CreateFI(FrameIndex));
  return foldMemoryOperand(MI, Ops[0], MOs);
}

MachineInstr* X86InstrInfo::foldMemoryOperand(MachineFunction &MF,
                                              MachineInstr *MI,
                                              SmallVectorImpl<unsigned> &Ops,
                                              MachineInstr *LoadMI) const {
  // Check switch flag 
  if (NoFusing) return NULL;

  unsigned Alignment = 0;
  for (unsigned i = 0, e = LoadMI->getNumMemOperands(); i != e; ++i) {
    const MachineMemOperand &MRO = LoadMI->getMemOperand(i);
    unsigned Align = MRO.getAlignment();
    if (Align > Alignment)
      Alignment = Align;
  }

  // FIXME: Move alignment requirement into tables?
  if (Alignment < 16) {
    switch (MI->getOpcode()) {
    default: break;
    // Not always safe to fold movsd into these instructions since their load
    // folding variants expects the address to be 16 byte aligned.
    case X86::FsANDNPDrr:
    case X86::FsANDNPSrr:
    case X86::FsANDPDrr:
    case X86::FsANDPSrr:
    case X86::FsORPDrr:
    case X86::FsORPSrr:
    case X86::FsXORPDrr:
    case X86::FsXORPSrr:
      return NULL;
    }
  }

  if (Ops.size() == 2 && Ops[0] == 0 && Ops[1] == 1) {
    unsigned NewOpc = 0;
    switch (MI->getOpcode()) {
    default: return NULL;
    case X86::TEST8rr:  NewOpc = X86::CMP8ri; break;
    case X86::TEST16rr: NewOpc = X86::CMP16ri; break;
    case X86::TEST32rr: NewOpc = X86::CMP32ri; break;
    case X86::TEST64rr: NewOpc = X86::CMP64ri32; break;
    }
    // Change to CMPXXri r, 0 first.
    MI->setDesc(get(NewOpc));
    MI->getOperand(1).ChangeToImmediate(0);
  } else if (Ops.size() != 1)
    return NULL;

  SmallVector<MachineOperand,4> MOs;
  unsigned NumOps = LoadMI->getDesc().getNumOperands();
  for (unsigned i = NumOps - 4; i != NumOps; ++i)
    MOs.push_back(LoadMI->getOperand(i));
  return foldMemoryOperand(MI, Ops[0], MOs);
}


bool X86InstrInfo::canFoldMemoryOperand(MachineInstr *MI,
                                        SmallVectorImpl<unsigned> &Ops) const {
  // Check switch flag 
  if (NoFusing) return 0;

  if (Ops.size() == 2 && Ops[0] == 0 && Ops[1] == 1) {
    switch (MI->getOpcode()) {