lib/Target/X86/X86RegisterInfo.td

//===- X86RegisterInfo.td - Describe the X86 Register File --*- tablegen -*-==//
// 
//                     The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
// 
//===----------------------------------------------------------------------===//
//
// This file describes the X86 Register file, defining the registers themselves,
// aliases between the registers, and the register classes built out of the
// registers.
//
//===----------------------------------------------------------------------===//

//===----------------------------------------------------------------------===//
//  Register definitions...
//
let Namespace = "X86" in {

  // In the register alias definitions below, we define which registers alias
  // which others.  We only specify which registers the small registers alias,
  // because the register file generator is smart enough to figure out that
  // AL aliases AX if we tell it that AX aliased AL (for example).

  // Dwarf numbering is different for 32-bit and 64-bit, and there are 
  // variations by target as well. Currently the first entry is for X86-64, 
  // second - for EH on X86-32/Darwin and third is 'generic' one (X86-32/Linux
  // and debug information on X86-32/Darwin)

  // 8-bit registers
  // Low registers
  def AL : Register<"al">, DwarfRegNum<[0, 0, 0]>;
  def DL : Register<"dl">, DwarfRegNum<[1, 2, 2]>;
  def CL : Register<"cl">, DwarfRegNum<[2, 1, 1]>;
  def BL : Register<"bl">, DwarfRegNum<[3, 3, 3]>;

  // X86-64 only
  def SIL : Register<"sil">, DwarfRegNum<[4, 6, 6]>;
  def DIL : Register<"dil">, DwarfRegNum<[5, 7, 7]>;
  def BPL : Register<"bpl">, DwarfRegNum<[6, 4, 5]>;
  def SPL : Register<"spl">, DwarfRegNum<[7, 5, 4]>;
  def R8B  : Register<"r8b">,  DwarfRegNum<[8, -2, -2]>;
  def R9B  : Register<"r9b">,  DwarfRegNum<[9, -2, -2]>;
  def R10B : Register<"r10b">, DwarfRegNum<[10, -2, -2]>;
  def R11B : Register<"r11b">, DwarfRegNum<[11, -2, -2]>;
  def R12B : Register<"r12b">, DwarfRegNum<[12, -2, -2]>;
  def R13B : Register<"r13b">, DwarfRegNum<[13, -2, -2]>;
  def R14B : Register<"r14b">, DwarfRegNum<[14, -2, -2]>;
  def R15B : Register<"r15b">, DwarfRegNum<[15, -2, -2]>;

  // High registers X86-32 only
  def AH : Register<"ah">, DwarfRegNum<[0, 0, 0]>;
  def DH : Register<"dh">, DwarfRegNum<[1, 2, 2]>;
  def CH : Register<"ch">, DwarfRegNum<[2, 1, 1]>;
  def BH : Register<"bh">, DwarfRegNum<[3, 3, 3]>;

  // 16-bit registers
  def AX : RegisterWithSubRegs<"ax", [AH,AL]>, DwarfRegNum<[0, 0, 0]>;
  def DX : RegisterWithSubRegs<"dx", [DH,DL]>, DwarfRegNum<[1, 2, 2]>;
  def CX : RegisterWithSubRegs<"cx", [CH,CL]>, DwarfRegNum<[2, 1, 1]>;
  def BX : RegisterWithSubRegs<"bx", [BH,BL]>, DwarfRegNum<[3, 3, 3]>;
  def SI : RegisterWithSubRegs<"si", [SIL]>, DwarfRegNum<[4, 6, 6]>;
  def DI : RegisterWithSubRegs<"di", [DIL]>, DwarfRegNum<[5, 7, 7]>;
  def BP : RegisterWithSubRegs<"bp", [BPL]>, DwarfRegNum<[6, 4, 5]>;
  def SP : RegisterWithSubRegs<"sp", [SPL]>, DwarfRegNum<[7, 5, 4]>;
  def IP : Register<"ip">, DwarfRegNum<[16]>;
  
  // X86-64 only
  def R8W  : RegisterWithSubRegs<"r8w", [R8B]>, DwarfRegNum<[8, -2, -2]>;
  def R9W  : RegisterWithSubRegs<"r9w", [R9B]>, DwarfRegNum<[9, -2, -2]>;
  def R10W : RegisterWithSubRegs<"r10w", [R10B]>, DwarfRegNum<[10, -2, -2]>;
  def R11W : RegisterWithSubRegs<"r11w", [R11B]>, DwarfRegNum<[11, -2, -2]>;
  def R12W : RegisterWithSubRegs<"r12w", [R12B]>, DwarfRegNum<[12, -2, -2]>;
  def R13W : RegisterWithSubRegs<"r13w", [R13B]>, DwarfRegNum<[13, -2, -2]>;
  def R14W : RegisterWithSubRegs<"r14w", [R14B]>, DwarfRegNum<[14, -2, -2]>;
  def R15W : RegisterWithSubRegs<"r15w", [R15B]>, DwarfRegNum<[15, -2, -2]>;

  // 32-bit registers
  def EAX : RegisterWithSubRegs<"eax", [AX]>, DwarfRegNum<[0, 0, 0]>;
  def EDX : RegisterWithSubRegs<"edx", [DX]>, DwarfRegNum<[1, 2, 2]>;
  def ECX : RegisterWithSubRegs<"ecx", [CX]>, DwarfRegNum<[2, 1, 1]>;
  def EBX : RegisterWithSubRegs<"ebx", [BX]>, DwarfRegNum<[3, 3, 3]>;
  def ESI : RegisterWithSubRegs<"esi", [SI]>, DwarfRegNum<[4, 6, 6]>;
  def EDI : RegisterWithSubRegs<"edi", [DI]>, DwarfRegNum<[5, 7, 7]>;
  def EBP : RegisterWithSubRegs<"ebp", [BP]>, DwarfRegNum<[6, 4, 5]>;
  def ESP : RegisterWithSubRegs<"esp", [SP]>, DwarfRegNum<[7, 5, 4]>;
  def EIP : RegisterWithSubRegs<"eip", [IP]>, DwarfRegNum<[16, 8, 8]>;  
  
  // X86-64 only
  def R8D  : RegisterWithSubRegs<"r8d", [R8W]>, DwarfRegNum<[8, -2, -2]>;
  def R9D  : RegisterWithSubRegs<"r9d", [R9W]>, DwarfRegNum<[9, -2, -2]>;
  def R10D : RegisterWithSubRegs<"r10d", [R10W]>, DwarfRegNum<[10, -2, -2]>;
  def R11D : RegisterWithSubRegs<"r11d", [R11W]>, DwarfRegNum<[11, -2, -2]>;
  def R12D : RegisterWithSubRegs<"r12d", [R12W]>, DwarfRegNum<[12, -2, -2]>;
  def R13D : RegisterWithSubRegs<"r13d", [R13W]>, DwarfRegNum<[13, -2, -2]>;
  def R14D : RegisterWithSubRegs<"r14d", [R14W]>, DwarfRegNum<[14, -2, -2]>;
  def R15D : RegisterWithSubRegs<"r15d", [R15W]>, DwarfRegNum<[15, -2, -2]>;

  // 64-bit registers, X86-64 only
  def RAX : RegisterWithSubRegs<"rax", [EAX]>, DwarfRegNum<[0, -2, -2]>;
  def RDX : RegisterWithSubRegs<"rdx", [EDX]>, DwarfRegNum<[1, -2, -2]>;
  def RCX : RegisterWithSubRegs<"rcx", [ECX]>, DwarfRegNum<[2, -2, -2]>;
  def RBX : RegisterWithSubRegs<"rbx", [EBX]>, DwarfRegNum<[3, -2, -2]>;
  def RSI : RegisterWithSubRegs<"rsi", [ESI]>, DwarfRegNum<[4, -2, -2]>;
  def RDI : RegisterWithSubRegs<"rdi", [EDI]>, DwarfRegNum<[5, -2, -2]>;
  def RBP : RegisterWithSubRegs<"rbp", [EBP]>, DwarfRegNum<[6, -2, -2]>;
  def RSP : RegisterWithSubRegs<"rsp", [ESP]>, DwarfRegNum<[7, -2, -2]>;

  def R8  : RegisterWithSubRegs<"r8", [R8D]>, DwarfRegNum<[8, -2, -2]>;
  def R9  : RegisterWithSubRegs<"r9", [R9D]>, DwarfRegNum<[9, -2, -2]>;
  def R10 : RegisterWithSubRegs<"r10", [R10D]>, DwarfRegNum<[10, -2, -2]>;
  def R11 : RegisterWithSubRegs<"r11", [R11D]>, DwarfRegNum<[11, -2, -2]>;
  def R12 : RegisterWithSubRegs<"r12", [R12D]>, DwarfRegNum<[12, -2, -2]>;
  def R13 : RegisterWithSubRegs<"r13", [R13D]>, DwarfRegNum<[13, -2, -2]>;
  def R14 : RegisterWithSubRegs<"r14", [R14D]>, DwarfRegNum<[14, -2, -2]>;
  def R15 : RegisterWithSubRegs<"r15", [R15D]>, DwarfRegNum<[15, -2, -2]>;
  def RIP : RegisterWithSubRegs<"rip", [EIP]>,  DwarfRegNum<[16, -2, -2]>;

  // MMX Registers. These are actually aliased to ST0 .. ST7
  def MM0 : Register<"mm0">, DwarfRegNum<[41, 29, 29]>;
  def MM1 : Register<"mm1">, DwarfRegNum<[42, 30, 30]>;
  def MM2 : Register<"mm2">, DwarfRegNum<[43, 31, 31]>;
  def MM3 : Register<"mm3">, DwarfRegNum<[44, 32, 32]>;
  def MM4 : Register<"mm4">, DwarfRegNum<[45, 33, 33]>;
  def MM5 : Register<"mm5">, DwarfRegNum<[46, 34, 34]>;
  def MM6 : Register<"mm6">, DwarfRegNum<[47, 35, 35]>;
  def MM7 : Register<"mm7">, DwarfRegNum<[48, 36, 36]>;
  
  // Pseudo Floating Point registers
  def FP0 : Register<"fp0">;
  def FP1 : Register<"fp1">;
  def FP2 : Register<"fp2">;
  def FP3 : Register<"fp3">;
  def FP4 : Register<"fp4">;
  def FP5 : Register<"fp5">;
  def FP6 : Register<"fp6">; 

  // XMM Registers, used by the various SSE instruction set extensions
  def XMM0: Register<"xmm0">, DwarfRegNum<[17, 21, 21]>;
  def XMM1: Register<"xmm1">, DwarfRegNum<[18, 22, 22]>;
  def XMM2: Register<"xmm2">, DwarfRegNum<[19, 23, 23]>;
  def XMM3: Register<"xmm3">, DwarfRegNum<[20, 24, 24]>;
  def XMM4: Register<"xmm4">, DwarfRegNum<[21, 25, 25]>;
  def XMM5: Register<"xmm5">, DwarfRegNum<[22, 26, 26]>;
  def XMM6: Register<"xmm6">, DwarfRegNum<[23, 27, 27]>;
  def XMM7: Register<"xmm7">, DwarfRegNum<[24, 28, 28]>;

  // X86-64 only
  def XMM8:  Register<"xmm8">,  DwarfRegNum<[25, -2, -2]>;
  def XMM9:  Register<"xmm9">,  DwarfRegNum<[26, -2, -2]>;
  def XMM10: Register<"xmm10">, DwarfRegNum<[27, -2, -2]>;
  def XMM11: Register<"xmm11">, DwarfRegNum<[28, -2, -2]>;
  def XMM12: Register<"xmm12">, DwarfRegNum<[29, -2, -2]>;
  def XMM13: Register<"xmm13">, DwarfRegNum<[30, -2, -2]>;
  def XMM14: Register<"xmm14">, DwarfRegNum<[31, -2, -2]>;
  def XMM15: Register<"xmm15">, DwarfRegNum<[32, -2, -2]>;

  // Floating point stack registers
  def ST0 : Register<"st(0)">, DwarfRegNum<[33, 12, 11]>;
  def ST1 : Register<"st(1)">, DwarfRegNum<[34, 13, 12]>;
  def ST2 : Register<"st(2)">, DwarfRegNum<[35, 14, 13]>;
  def ST3 : Register<"st(3)">, DwarfRegNum<[36, 15, 14]>;
  def ST4 : Register<"st(4)">, DwarfRegNum<[37, 16, 15]>;
  def ST5 : Register<"st(5)">, DwarfRegNum<[38, 17, 16]>;
  def ST6 : Register<"st(6)">, DwarfRegNum<[39, 18, 17]>;
  def ST7 : Register<"st(7)">, DwarfRegNum<[40, 19, 18]>; 

  // Status flags register
  def EFLAGS : Register<"flags">;

  // Segment registers
  def CS : Register<"cs">;
  def DS : Register<"ds">;
  def SS : Register<"ss">;
  def ES : Register<"es">;
  def FS : Register<"fs">;
  def GS : Register<"gs">;
}


//===----------------------------------------------------------------------===//
// Subregister Set Definitions... now that we have all of the pieces, define the
// sub registers for each register.
//

def x86_subreg_8bit    : PatLeaf<(i32 1)>;
def x86_subreg_16bit   : PatLeaf<(i32 2)>;
def x86_subreg_32bit   : PatLeaf<(i32 3)>;

def : SubRegSet<1, [AX, CX, DX, BX, SP,  BP,  SI,  DI,  
                    R8W, R9W, R10W, R11W, R12W, R13W, R14W, R15W],
                   [AL, CL, DL, BL, SPL, BPL, SIL, DIL, 
                    R8B, R9B, R10B, R11B, R12B, R13B, R14B, R15B]>;

// It's unclear if this subreg set is safe, given that not all registers
// in the class have an 'H' subreg.
// def : SubRegSet<2, [AX, CX, DX, BX],
//                    [AH, CH, DH, BH]>;

def : SubRegSet<1, [EAX, ECX, EDX, EBX, ESP, EBP, ESI, EDI,  
                    R8D, R9D, R10D, R11D, R12D, R13D, R14D, R15D],
                   [AL, CL, DL, BL, SPL, BPL, SIL, DIL, 
                    R8B, R9B, R10B, R11B, R12B, R13B, R14B, R15B]>;

def : SubRegSet<2, [EAX, ECX, EDX, EBX, ESP, EBP, ESI, EDI,  
                    R8D, R9D, R10D, R11D, R12D, R13D, R14D, R15D],
                   [AX,  CX,  DX,  BX,  SP,  BP,  SI,  DI, 
                    R8W, R9W, R10W, R11W, R12W, R13W, R14W, R15W]>;


def : SubRegSet<1, [RAX, RCX, RDX, RBX, RSP, RBP, RSI, RDI,  
                    R8,  R9,  R10, R11, R12, R13, R14, R15],
                   [AL, CL, DL, BL, SPL, BPL, SIL, DIL, 
                    R8B, R9B, R10B, R11B, R12B, R13B, R14B, R15B]>;

def : SubRegSet<2, [RAX, RCX, RDX, RBX, RSP, RBP, RSI, RDI,  
                    R8,  R9,  R10, R11, R12, R13, R14, R15],
                   [AX,  CX,  DX,  BX,  SP,  BP,  SI,  DI, 
                    R8W, R9W, R10W, R11W, R12W, R13W, R14W, R15W]>;
                    
def : SubRegSet<3, [RAX, RCX, RDX, RBX, RSP, RBP, RSI, RDI,  
                    R8,  R9,  R10, R11, R12, R13, R14, R15],
                   [EAX, ECX, EDX, EBX, ESP, EBP, ESI, EDI, 
                    R8D, R9D, R10D, R11D, R12D, R13D, R14D, R15D]>;

//===----------------------------------------------------------------------===//
// Register Class Definitions... now that we have all of the pieces, define the
// top-level register classes.  The order specified in the register list is
// implicitly defined to be the register allocation order.
//

// List call-clobbered registers before callee-save registers. RBX, RBP, (and 
// R12, R13, R14, and R15 for X86-64) are callee-save registers.
// In 64-mode, there are 12 additional i8 registers, SIL, DIL, BPL, SPL, and
// R8B, ... R15B. 
// Allocate R12 and R13 last, as these require an extra byte when
// encoded in x86_64 instructions.
// FIXME: Allow AH, CH, DH, BH in 64-mode for non-REX instructions,
def GR8 : RegisterClass<"X86", [i8],  8,
                        [AL, CL, DL, BL, AH, CH, DH, BH, SIL, DIL, BPL, SPL,
                         R8B, R9B, R10B, R11B, R14B, R15B, R12B, R13B]> {
  let MethodProtos = [{
    iterator allocation_order_begin(const MachineFunction &MF) const;
    iterator allocation_order_end(const MachineFunction &MF) const;
  }];
  let MethodBodies = [{
    // Does the function dedicate RBP / EBP to being a frame ptr?
    // If so, don't allocate SPL or BPL.
    static const unsigned X86_GR8_AO_64_fp[] = {
      X86::AL,   X86::CL,   X86::DL,   X86::SIL, X86::DIL,
      X86::R8B,  X86::R9B,  X86::R10B, X86::R11B,
      X86::BL,   X86::R14B, X86::R15B, X86::R12B, X86::R13B
    };
    // If not, just don't allocate SPL.
    static const unsigned X86_GR8_AO_64[] = {
      X86::AL,   X86::CL,   X86::DL,   X86::SIL, X86::DIL,
      X86::R8B,  X86::R9B,  X86::R10B, X86::R11B,
      X86::BL,   X86::R14B, X86::R15B, X86::R12B, X86::R13B, X86::BPL
    };
    // In 32-mode, none of the 8-bit registers aliases EBP or ESP.
    static const unsigned X86_GR8_AO_32[] = {
      X86::AL, X86::CL, X86::DL, X86::AH, X86::CH, X86::DH, X86::BL, X86::BH
    };

    GR8Class::iterator
    GR8Class::allocation_order_begin(const MachineFunction &MF) const {
      const TargetMachine &TM = MF.getTarget();
      const TargetRegisterInfo *RI = TM.getRegisterInfo();
      const X86Subtarget &Subtarget = TM.getSubtarget<X86Subtarget>();
      if (!Subtarget.is64Bit())
        return X86_GR8_AO_32;
      else if (RI->hasFP(MF))
        return X86_GR8_AO_64_fp;
      else
        return X86_GR8_AO_64;
    }

    GR8Class::iterator
    GR8Class::allocation_order_end(const MachineFunction &MF) const {
      const TargetMachine &TM = MF.getTarget();
      const TargetRegisterInfo *RI = TM.getRegisterInfo();
      const X86Subtarget &Subtarget = TM.getSubtarget<X86Subtarget>();
      if (!Subtarget.is64Bit())
        return X86_GR8_AO_32 + (sizeof(X86_GR8_AO_32) / sizeof(unsigned));
      else if (RI->hasFP(MF))
        return X86_GR8_AO_64_fp + (sizeof(X86_GR8_AO_64_fp) / sizeof(unsigned));
      else
        return X86_GR8_AO_64 + (sizeof(X86_GR8_AO_64) / sizeof(unsigned));
    }
  }];
}


def GR16 : RegisterClass<"X86", [i16], 16,
                         [AX, CX, DX, SI, DI, BX, BP, SP,
                          R8W, R9W, R10W, R11W, R14W, R15W, R12W, R13W]> {
  let SubRegClassList = [GR8];
  let MethodProtos = [{
    iterator allocation_order_begin(const MachineFunction &MF) const;
    iterator allocation_order_end(const MachineFunction &MF) const;
  }];
  let MethodBodies = [{
    // Does the function dedicate RBP / EBP to being a frame ptr?
    // If so, don't allocate SP or BP.
    static const unsigned X86_GR16_AO_64_fp[] = {
      X86::AX,  X86::CX,   X86::DX,   X86::SI,   X86::DI,
      X86::R8W, X86::R9W,  X86::R10W, X86::R11W,
      X86::BX, X86::R14W, X86::R15W,  X86::R12W, X86::R13W
    };
    static const unsigned X86_GR16_AO_32_fp[] = {
      X86::AX, X86::CX, X86::DX, X86::SI, X86::DI, X86::BX
    };
    // If not, just don't allocate SPL.
    static const unsigned X86_GR16_AO_64[] = {
      X86::AX,  X86::CX,   X86::DX,   X86::SI,   X86::DI,
      X86::R8W, X86::R9W,  X86::R10W, X86::R11W,
      X86::BX, X86::R14W, X86::R15W,  X86::R12W, X86::R13W, X86::BP
    };
    static const unsigned X86_GR16_AO_32[] = {
      X86::AX, X86::CX, X86::DX, X86::SI, X86::DI, X86::BX, X86::BP
    };

    GR16Class::iterator
    GR16Class::allocation_order_begin(const MachineFunction &MF) const {
      const TargetMachine &TM = MF.getTarget();
      const TargetRegisterInfo *RI = TM.getRegisterInfo();
      const X86Subtarget &Subtarget = TM.getSubtarget<X86Subtarget>();
      if (Subtarget.is64Bit()) {
        if (RI->hasFP(MF))
          return X86_GR16_AO_64_fp;
        else
          return X86_GR16_AO_64;
      } else {
        if (RI->hasFP(MF))
          return X86_GR16_AO_32_fp;
        else
          return X86_GR16_AO_32;
      }
    }

    GR16Class::iterator
    GR16Class::allocation_order_end(const MachineFunction &MF) const {
      const TargetMachine &TM = MF.getTarget();
      const TargetRegisterInfo *RI = TM.getRegisterInfo();
      const X86Subtarget &Subtarget = TM.getSubtarget<X86Subtarget>();
      if (Subtarget.is64Bit()) {
        if (RI->hasFP(MF))
          return X86_GR16_AO_64_fp+(sizeof(X86_GR16_AO_64_fp)/sizeof(unsigned));
        else
          return X86_GR16_AO_64 + (sizeof(X86_GR16_AO_64) / sizeof(unsigned));
      } else {
        if (RI->hasFP(MF))
          return X86_GR16_AO_32_fp+(sizeof(X86_GR16_AO_32_fp)/sizeof(unsigned));
        else
          return X86_GR16_AO_32 + (sizeof(X86_GR16_AO_32) / sizeof(unsigned));
      }
    }
  }];
}


def GR32 : RegisterClass<"X86", [i32], 32, 
                         [EAX, ECX, EDX, ESI, EDI, EBX, EBP, ESP,
                          R8D, R9D, R10D, R11D, R14D, R15D, R12D, R13D]> {
  let SubRegClassList = [GR8, GR16];
  let MethodProtos = [{
    iterator allocation_order_begin(const MachineFunction &MF) const;
    iterator allocation_order_end(const MachineFunction &MF) const;
  }];
  let MethodBodies = [{
    // Does the function dedicate RBP / EBP to being a frame ptr?
    // If so, don't allocate ESP or EBP.
    static const unsigned X86_GR32_AO_64_fp[] = {
      X86::EAX, X86::ECX,  X86::EDX,  X86::ESI,  X86::EDI,
      X86::R8D, X86::R9D,  X86::R10D, X86::R11D,
      X86::EBX, X86::R14D, X86::R15D, X86::R12D, X86::R13D
    };
    static const unsigned X86_GR32_AO_32_fp[] = {
      X86::EAX, X86::ECX, X86::EDX, X86::ESI, X86::EDI, X86::EBX
    };
    // If not, just don't allocate SPL.
    static const unsigned X86_GR32_AO_64[] = {
      X86::EAX, X86::ECX,  X86::EDX,  X86::ESI,  X86::EDI,
      X86::R8D, X86::R9D,  X86::R10D, X86::R11D,
      X86::EBX, X86::R14D, X86::R15D, X86::R12D, X86::R13D, X86::EBP
    };
    static const unsigned X86_GR32_AO_32[] = {
      X86::EAX, X86::ECX, X86::EDX, X86::ESI, X86::EDI, X86::EBX, X86::EBP
    };

    GR32Class::iterator
    GR32Class::allocation_order_begin(const MachineFunction &MF) const {
      const TargetMachine &TM = MF.getTarget();
      const TargetRegisterInfo *RI = TM.getRegisterInfo();
      const X86Subtarget &Subtarget = TM.getSubtarget<X86Subtarget>();
      if (Subtarget.is64Bit()) {
        if (RI->hasFP(MF))
          return X86_GR32_AO_64_fp;
        else
          return X86_GR32_AO_64;
      } else {
        if (RI->hasFP(MF))
          return X86_GR32_AO_32_fp;
        else
          return X86_GR32_AO_32;
      }
    }

    GR32Class::iterator
    GR32Class::allocation_order_end(const MachineFunction &MF) const {
      const TargetMachine &TM = MF.getTarget();
      const TargetRegisterInfo *RI = TM.getRegisterInfo();
      const X86Subtarget &Subtarget = TM.getSubtarget<X86Subtarget>();
      if (Subtarget.is64Bit()) {
        if (RI->hasFP(MF))
          return X86_GR32_AO_64_fp+(sizeof(X86_GR32_AO_64_fp)/sizeof(unsigned));
        else
          return X86_GR32_AO_64 + (sizeof(X86_GR32_AO_64) / sizeof(unsigned));
      } else {
        if (RI->hasFP(MF))
          return X86_GR32_AO_32_fp+(sizeof(X86_GR32_AO_32_fp)/sizeof(unsigned));
        else
          return X86_GR32_AO_32 + (sizeof(X86_GR32_AO_32) / sizeof(unsigned));
      }
    }
  }];
}


def GR64 : RegisterClass<"X86", [i64], 64, 
                         [RAX, RCX, RDX, RSI, RDI, R8, R9, R10, R11,
                          RBX, R14, R15, R12, R13, RBP, RSP]> {
  let SubRegClassList = [GR8, GR16, GR32];
  let MethodProtos = [{
    iterator allocation_order_end(const MachineFunction &MF) const;
  }];
  let MethodBodies = [{
    GR64Class::iterator
    GR64Class::allocation_order_end(const MachineFunction &MF) const {
      const TargetMachine &TM = MF.getTarget();
      const TargetRegisterInfo *RI = TM.getRegisterInfo();
      const X86Subtarget &Subtarget = TM.getSubtarget<X86Subtarget>();
      if (!Subtarget.is64Bit())
        return begin();  // None of these are allocatable in 32-bit.
      if (RI->hasFP(MF)) // Does the function dedicate RBP to being a frame ptr?
        return end()-2;  // If so, don't allocate RSP or RBP
      else
        return end()-1;  // If not, just don't allocate RSP
    }
  }];
}


// GR16, GR32 subclasses which contain registers that have GR8 sub-registers.
// These should only be used for 32-bit mode.
def GR16_ : RegisterClass<"X86", [i16], 16, [AX, CX, DX, BX]> {
  let SubRegClassList = [GR8];
}
def GR32_ : RegisterClass<"X86", [i32], 32, [EAX, ECX, EDX, EBX]> {
  let SubRegClassList = [GR8, GR16];
}

// A class to support the 'A' assembler constraint: EAX then EDX.
def GRAD : RegisterClass<"X86", [i32], 32, [EAX, EDX]>;

// Scalar SSE2 floating point registers.
def FR32 : RegisterClass<"X86", [f32], 32,
                         [XMM0, XMM1, XMM2, XMM3, XMM4, XMM5, XMM6, XMM7,
                          XMM8, XMM9, XMM10, XMM11,
                          XMM12, XMM13, XMM14, XMM15]> {
  let MethodProtos = [{
    iterator allocation_order_end(const MachineFunction &MF) const;
  }];
  let MethodBodies = [{
    FR32Class::iterator
    FR32Class::allocation_order_end(const MachineFunction &MF) const {
      const TargetMachine &TM = MF.getTarget();
      const X86Subtarget &Subtarget = TM.getSubtarget<X86Subtarget>();
      if (!Subtarget.is64Bit())
        return end()-8; // Only XMM0 to XMM7 are available in 32-bit mode.
      else
        return end();
    }
  }];
}

def FR64 : RegisterClass<"X86", [f64], 64,
                         [XMM0, XMM1, XMM2, XMM3, XMM4, XMM5, XMM6, XMM7,
                          XMM8, XMM9, XMM10, XMM11,
                          XMM12, XMM13, XMM14, XMM15]> {
  let MethodProtos = [{
    iterator allocation_order_end(const MachineFunction &MF) const;
  }];
  let MethodBodies = [{
    FR64Class::iterator
    FR64Class::allocation_order_end(const MachineFunction &MF) const {
      const TargetMachine &TM = MF.getTarget();
      const X86Subtarget &Subtarget = TM.getSubtarget<X86Subtarget>();
      if (!Subtarget.is64Bit())
        return end()-8; // Only XMM0 to XMM7 are available in 32-bit mode.
      else
        return end();
    }
  }];
}


// FIXME: This sets up the floating point register files as though they are f64
// values, though they really are f80 values.  This will cause us to spill
// values as 64-bit quantities instead of 80-bit quantities, which is much much
// faster on common hardware.  In reality, this should be controlled by a
// command line option or something.

def RFP32 : RegisterClass<"X86",[f32], 32, [FP0, FP1, FP2, FP3, FP4, FP5, FP6]>;
def RFP64 : RegisterClass<"X86",[f64], 32, [FP0, FP1, FP2, FP3, FP4, FP5, FP6]>;
def RFP80 : RegisterClass<"X86",[f80], 32, [FP0, FP1, FP2, FP3, FP4, FP5, FP6]>;

// Floating point stack registers (these are not allocatable by the
// register allocator - the floating point stackifier is responsible
// for transforming FPn allocations to STn registers)
def RST : RegisterClass<"X86", [f80, f64, f32], 32,
                        [ST0, ST1, ST2, ST3, ST4, ST5, ST6, ST7]> {
    let MethodProtos = [{
    iterator allocation_order_end(const MachineFunction &MF) const;
  }];
  let MethodBodies = [{
    RSTClass::iterator
    RSTClass::allocation_order_end(const MachineFunction &MF) const {
      return begin();
    }
  }];
}

// Generic vector registers: VR64 and VR128.
def VR64  : RegisterClass<"X86", [v8i8, v4i16, v2i32, v1i64, v2f32], 64,
                          [MM0, MM1, MM2, MM3, MM4, MM5, MM6, MM7]>;
def VR128 : RegisterClass<"X86", [v16i8, v8i16, v4i32, v2i64, v4f32, v2f64],128,
                          [XMM0, XMM1, XMM2, XMM3, XMM4, XMM5, XMM6, XMM7,
                           XMM8, XMM9, XMM10, XMM11,
                           XMM12, XMM13, XMM14, XMM15]> {
  let MethodProtos = [{
    iterator allocation_order_end(const MachineFunction &MF) const;
  }];
  let MethodBodies = [{
    VR128Class::iterator
    VR128Class::allocation_order_end(const MachineFunction &MF) const {
      const TargetMachine &TM = MF.getTarget();
      const X86Subtarget &Subtarget = TM.getSubtarget<X86Subtarget>();
      if (!Subtarget.is64Bit())
        return end()-8; // Only XMM0 to XMM7 are available in 32-bit mode.
      else
        return end();
    }
  }];
}

// Status flags registers.
def CCR : RegisterClass<"X86", [i32], 32, [EFLAGS]> {
  let CopyCost = -1;  // Don't allow copying of status registers.
}