llvm/lib/Target/X86/X86InstrInfo.td
Alkis Evlogimenos fa5229691f Fix argument size for SHL, SHR, SAR, SHLD and SHRD families of
instructions.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@11923 91177308-0d34-0410-b5e6-96231b3b80d8
2004-02-27 19:46:30 +00:00

717 lines
39 KiB
C++

//===- X86InstrInfo.td - Describe the X86 Instruction Set -------*- C++ -*-===//
//
// The LLVM Compiler Infrastructure
//
// This file was developed by the LLVM research group and is distributed under
// the University of Illinois Open Source License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//
// This file describes the X86 instruction set, defining the instructions, and
// properties of the instructions which are needed for code generation, machine
// code emission, and analysis.
//
//===----------------------------------------------------------------------===//
// Format specifies the encoding used by the instruction. This is part of the
// ad-hoc solution used to emit machine instruction encodings by our machine
// code emitter.
class Format<bits<5> val> {
bits<5> Value = val;
}
def Pseudo : Format<0>; def RawFrm : Format<1>;
def AddRegFrm : Format<2>; def MRMDestReg : Format<3>;
def MRMDestMem : Format<4>; def MRMSrcReg : Format<5>;
def MRMSrcMem : Format<6>;
def MRM0r : Format<16>; def MRM1r : Format<17>; def MRM2r : Format<18>;
def MRM3r : Format<19>; def MRM4r : Format<20>; def MRM5r : Format<21>;
def MRM6r : Format<22>; def MRM7r : Format<23>;
def MRM0m : Format<24>; def MRM1m : Format<25>; def MRM2m : Format<26>;
def MRM3m : Format<27>; def MRM4m : Format<28>; def MRM5m : Format<29>;
def MRM6m : Format<30>; def MRM7m : Format<31>;
// ArgType - This specifies the argument type used by an instruction. This is
// part of the ad-hoc solution used to emit machine instruction encodings by our
// machine code emitter.
class ArgType<bits<3> val> {
bits<3> Value = val;
}
def NoArg : ArgType<0>;
def Arg8 : ArgType<1>;
def Arg16 : ArgType<2>;
def Arg32 : ArgType<3>;
def Arg64 : ArgType<4>; // 64 bit int argument for FILD64
def ArgF32 : ArgType<5>;
def ArgF64 : ArgType<6>;
def ArgF80 : ArgType<6>;
// FPFormat - This specifies what form this FP instruction has. This is used by
// the Floating-Point stackifier pass.
class FPFormat<bits<3> val> {
bits<3> Value = val;
}
def NotFP : FPFormat<0>;
def ZeroArgFP : FPFormat<1>;
def OneArgFP : FPFormat<2>;
def OneArgFPRW : FPFormat<3>;
def TwoArgFP : FPFormat<4>;
def SpecialFP : FPFormat<5>;
class X86Inst<string nam, bits<8> opcod, Format f, ArgType a> : Instruction {
let Namespace = "X86";
let Name = nam;
bits<8> Opcode = opcod;
Format Form = f;
bits<5> FormBits = Form.Value;
ArgType Type = a;
bits<3> TypeBits = Type.Value;
// Attributes specific to X86 instructions...
bit hasOpSizePrefix = 0; // Does this inst have a 0x66 prefix?
bit printImplicitUses = 0; // Should we print implicit uses of this inst?
bits<4> Prefix = 0; // Which prefix byte does this inst have?
FPFormat FPForm; // What flavor of FP instruction is this?
bits<3> FPFormBits = 0;
}
class Imp<list<Register> uses, list<Register> defs> {
list<Register> Uses = uses;
list<Register> Defs = defs;
}
class Pattern<dag P> {
dag Pattern = P;
}
// Prefix byte classes which are used to indicate to the ad-hoc machine code
// emitter that various prefix bytes are required.
class OpSize { bit hasOpSizePrefix = 1; }
class TB { bits<4> Prefix = 1; }
class REP { bits<4> Prefix = 2; }
class D8 { bits<4> Prefix = 3; }
class D9 { bits<4> Prefix = 4; }
class DA { bits<4> Prefix = 5; }
class DB { bits<4> Prefix = 6; }
class DC { bits<4> Prefix = 7; }
class DD { bits<4> Prefix = 8; }
class DE { bits<4> Prefix = 9; }
class DF { bits<4> Prefix = 10; }
//===----------------------------------------------------------------------===//
// Instruction list...
//
def PHI : X86Inst<"PHI", 0, Pseudo, NoArg>; // PHI node...
def NOOP : X86Inst<"nop", 0x90, RawFrm, NoArg>; // nop
def ADJCALLSTACKDOWN : X86Inst<"ADJCALLSTACKDOWN", 0, Pseudo, NoArg>;
def ADJCALLSTACKUP : X86Inst<"ADJCALLSTACKUP", 0, Pseudo, NoArg>;
def IMPLICIT_USE : X86Inst<"IMPLICIT_USE", 0, Pseudo, NoArg>;
def IMPLICIT_DEF : X86Inst<"IMPLICIT_DEF", 0, Pseudo, NoArg>;
let isTerminator = 1 in
let Defs = [FP0, FP1, FP2, FP3, FP4, FP5, FP6] in
def FP_REG_KILL : X86Inst<"FP_REG_KILL", 0, Pseudo, NoArg>;
//===----------------------------------------------------------------------===//
// Control Flow Instructions...
//
// Return instruction...
let isTerminator = 1, isReturn = 1 in
def RET : X86Inst<"ret", 0xC3, RawFrm, NoArg>, Pattern<(retvoid)>;
// All branches are RawFrm, Void, Branch, and Terminators
let isBranch = 1, isTerminator = 1 in
class IBr<string name, bits<8> opcode> : X86Inst<name, opcode, RawFrm, NoArg>;
def JMP : IBr<"jmp", 0xE9>, Pattern<(br basicblock)>;
def JB : IBr<"jb" , 0x82>, TB;
def JAE : IBr<"jae", 0x83>, TB;
def JE : IBr<"je" , 0x84>, TB, Pattern<(isVoid (unspec1 basicblock))>;
def JNE : IBr<"jne", 0x85>, TB;
def JBE : IBr<"jbe", 0x86>, TB;
def JA : IBr<"ja" , 0x87>, TB;
def JS : IBr<"js" , 0x88>, TB;
def JNS : IBr<"jns", 0x89>, TB;
def JL : IBr<"jl" , 0x8C>, TB;
def JGE : IBr<"jge", 0x8D>, TB;
def JLE : IBr<"jle", 0x8E>, TB;
def JG : IBr<"jg" , 0x8F>, TB;
//===----------------------------------------------------------------------===//
// Call Instructions...
//
let isCall = 1 in
// All calls clobber the non-callee saved registers...
let Defs = [EAX, ECX, EDX, FP0, FP1, FP2, FP3, FP4, FP5, FP6] in {
def CALLpcrel32 : X86Inst<"call", 0xE8, RawFrm, NoArg>;
def CALLr32 : X86Inst<"call", 0xFF, MRM2r , Arg32>;
def CALLm32 : X86Inst<"call", 0xFF, MRM2m , Arg32>;
}
//===----------------------------------------------------------------------===//
// Miscellaneous Instructions...
//
def LEAVE : X86Inst<"leave", 0xC9, RawFrm, NoArg>, Imp<[EBP,ESP],[EBP,ESP]>;
def POPr32 : X86Inst<"pop", 0x58, AddRegFrm, Arg32>, Imp<[ESP],[ESP]>;
let isTwoAddress = 1 in // R32 = bswap R32
def BSWAPr32 : X86Inst<"bswap", 0xC8, AddRegFrm, Arg32>, TB;
def XCHGrr8 : X86Inst<"xchg", 0x86, MRMDestReg, Arg8>; // xchg R8, R8
def XCHGrr16 : X86Inst<"xchg", 0x87, MRMDestReg, Arg16>, OpSize;// xchg R16, R16
def XCHGrr32 : X86Inst<"xchg", 0x87, MRMDestReg, Arg32>; // xchg R32, R32
def XCHGmr8 : X86Inst<"xchg", 0x86, MRMDestMem, Arg8>; // xchg [mem8], R8
def XCHGmr16 : X86Inst<"xchg", 0x87, MRMDestMem, Arg16>, OpSize;// xchg [mem16], R16
def XCHGmr32 : X86Inst<"xchg", 0x87, MRMDestMem, Arg32>; // xchg [mem32], R32
def XCHGrm8 : X86Inst<"xchg", 0x86, MRMSrcMem , Arg8>; // xchg R8, [mem8]
def XCHGrm16 : X86Inst<"xchg", 0x87, MRMSrcMem , Arg16>, OpSize;// xchg R16, [mem16]
def XCHGrm32 : X86Inst<"xchg", 0x87, MRMSrcMem , Arg32>; // xchg R32, [mem32]
def LEAr16 : X86Inst<"lea", 0x8D, MRMSrcMem, Arg16>, OpSize; // R16 = lea [mem]
def LEAr32 : X86Inst<"lea", 0x8D, MRMSrcMem, Arg32>; // R32 = lea [mem]
def REP_MOVSB : X86Inst<"rep movsb", 0xA4, RawFrm, NoArg>, REP,
Imp<[ECX,EDI,ESI], [ECX,EDI,ESI]>;
def REP_MOVSW : X86Inst<"rep movsw", 0xA5, RawFrm, NoArg>, REP, OpSize,
Imp<[ECX,EDI,ESI], [ECX,EDI,ESI]>;
def REP_MOVSD : X86Inst<"rep movsd", 0xA5, RawFrm, NoArg>, REP,
Imp<[ECX,EDI,ESI], [ECX,EDI,ESI]>;
def REP_STOSB : X86Inst<"rep stosb", 0xAA, RawFrm, NoArg>, REP,
Imp<[AL,ECX,EDI], [ECX,EDI]>;
def REP_STOSW : X86Inst<"rep stosw", 0xAB, RawFrm, NoArg>, REP, OpSize,
Imp<[AX,ECX,EDI], [ECX,EDI]>;
def REP_STOSD : X86Inst<"rep stosd", 0xAB, RawFrm, NoArg>, REP,
Imp<[EAX,ECX,EDI], [ECX,EDI]>;
//===----------------------------------------------------------------------===//
// Move Instructions...
//
def MOVrr8 : X86Inst<"mov", 0x88, MRMDestReg, Arg8>, Pattern<(set R8 , R8 )>;
def MOVrr16 : X86Inst<"mov", 0x89, MRMDestReg, Arg16>, OpSize, Pattern<(set R16, R16)>;
def MOVrr32 : X86Inst<"mov", 0x89, MRMDestReg, Arg32>, Pattern<(set R32, R32)>;
def MOVri8 : X86Inst<"mov", 0xB0, AddRegFrm , Arg8>, Pattern<(set R8 , imm )>;
def MOVri16 : X86Inst<"mov", 0xB8, AddRegFrm , Arg16>, OpSize, Pattern<(set R16, imm)>;
def MOVri32 : X86Inst<"mov", 0xB8, AddRegFrm , Arg32>, Pattern<(set R32, imm)>;
def MOVmi8 : X86Inst<"mov", 0xC6, MRM0m , Arg8>; // [mem] = imm8
def MOVmi16 : X86Inst<"mov", 0xC7, MRM0m , Arg16>, OpSize; // [mem] = imm16
def MOVmi32 : X86Inst<"mov", 0xC7, MRM0m , Arg32>; // [mem] = imm32
def MOVrm8 : X86Inst<"mov", 0x8A, MRMSrcMem , Arg8>; // R8 = [mem]
def MOVrm16 : X86Inst<"mov", 0x8B, MRMSrcMem , Arg16>, OpSize, // R16 = [mem]
Pattern<(set R16, (load (plus R32, (plus (times imm, R32), imm))))>;
def MOVrm32 : X86Inst<"mov", 0x8B, MRMSrcMem , Arg32>, // R32 = [mem]
Pattern<(set R32, (load (plus R32, (plus (times imm, R32), imm))))>;
def MOVmr8 : X86Inst<"mov", 0x88, MRMDestMem, Arg8>; // [mem] = R8
def MOVmr16 : X86Inst<"mov", 0x89, MRMDestMem, Arg16>, OpSize; // [mem] = R16
def MOVmr32 : X86Inst<"mov", 0x89, MRMDestMem, Arg32>; // [mem] = R32
//===----------------------------------------------------------------------===//
// Fixed-Register Multiplication and Division Instructions...
//
// Extra precision multiplication
def MULr8 : X86Inst<"mul", 0xF6, MRM4r , Arg8 >, Imp<[AL],[AX]>; // AL,AH = AL*R8
def MULr16 : X86Inst<"mul", 0xF7, MRM4r , Arg16>, Imp<[AX],[AX,DX]>, OpSize; // AX,DX = AX*R16
def MULr32 : X86Inst<"mul", 0xF7, MRM4r , Arg32>, Imp<[EAX],[EAX,EDX]>; // EAX,EDX = EAX*R32
def MULm8 : X86Inst<"mul", 0xF6, MRM4m , Arg8 >, Imp<[AL],[AX]>; // AL,AH = AL*[mem8]
def MULm16 : X86Inst<"mul", 0xF7, MRM4m , Arg16>, Imp<[AX],[AX,DX]>, OpSize; // AX,DX = AX*[mem16]
def MULm32 : X86Inst<"mul", 0xF7, MRM4m , Arg32>, Imp<[EAX],[EAX,EDX]>; // EAX,EDX = EAX*[mem32]
// unsigned division/remainder
def DIVr8 : X86Inst<"div", 0xF6, MRM6r , Arg8 >, Imp<[AX],[AX]>; // AX/r8 = AL,AH
def DIVr16 : X86Inst<"div", 0xF7, MRM6r , Arg16>, Imp<[AX,DX],[AX,DX]>, OpSize; // DX:AX/r16 = AX,DX
def DIVr32 : X86Inst<"div", 0xF7, MRM6r , Arg32>, Imp<[EAX,EDX],[EAX,EDX]>; // EDX:EAX/r32 = EAX,EDX
def DIVm8 : X86Inst<"div", 0xF6, MRM6m , Arg8 >, Imp<[AX],[AX]>; // AX/[mem8] = AL,AH
def DIVm16 : X86Inst<"div", 0xF7, MRM6m , Arg16>, Imp<[AX,DX],[AX,DX]>, OpSize; // DX:AX/[mem16] = AX,DX
def DIVm32 : X86Inst<"div", 0xF7, MRM6m , Arg32>, Imp<[EAX,EDX],[EAX,EDX]>; // EDX:EAX/[mem32] = EAX,EDX
// signed division/remainder
def IDIVr8 : X86Inst<"idiv",0xF6, MRM7r , Arg8 >, Imp<[AX],[AX]>; // AX/r8 = AL,AH
def IDIVr16: X86Inst<"idiv",0xF7, MRM7r , Arg16>, Imp<[AX,DX],[AX,DX]>, OpSize; // DX:AX/r16 = AX,DX
def IDIVr32: X86Inst<"idiv",0xF7, MRM7r , Arg32>, Imp<[EAX,EDX],[EAX,EDX]>; // EDX:EAX/r32 = EAX,EDX
def IDIVm8 : X86Inst<"idiv",0xF6, MRM7m , Arg8 >, Imp<[AX],[AX]>; // AX/[mem8] = AL,AH
def IDIVm16: X86Inst<"idiv",0xF7, MRM7m , Arg16>, Imp<[AX,DX],[AX,DX]>, OpSize; // DX:AX/[mem16] = AX,DX
def IDIVm32: X86Inst<"idiv",0xF7, MRM7m , Arg32>, Imp<[EAX,EDX],[EAX,EDX]>; // EDX:EAX/[mem32] = EAX,EDX
// Sign-extenders for division
def CBW : X86Inst<"cbw", 0x98, RawFrm, Arg8 >, Imp<[AL],[AH]>; // AX = signext(AL)
def CWD : X86Inst<"cwd", 0x99, RawFrm, Arg8 >, Imp<[AX],[DX]>; // DX:AX = signext(AX)
def CDQ : X86Inst<"cdq", 0x99, RawFrm, Arg8 >, Imp<[EAX],[EDX]>; // EDX:EAX = signext(EAX)
//===----------------------------------------------------------------------===//
// Two address Instructions...
//
let isTwoAddress = 1 in { // Define some helper classes to make defs shorter.
class I2A8 <string n, bits<8> o, Format F> : X86Inst<n, o, F, Arg8>;
class I2A16<string n, bits<8> o, Format F> : X86Inst<n, o, F, Arg16>;
class I2A32<string n, bits<8> o, Format F> : X86Inst<n, o, F, Arg32>;
}
// unary instructions
def NEGr8 : I2A8 <"neg", 0xF6, MRM3r >; // R8 = -R8 = 0-R8
def NEGr16 : I2A16<"neg", 0xF7, MRM3r >, OpSize; // R16 = -R16 = 0-R16
def NEGr32 : I2A32<"neg", 0xF7, MRM3r >; // R32 = -R32 = 0-R32
def NEGm8 : I2A8 <"neg", 0xF6, MRM3m >; // [mem8] = -[mem8] = 0-[mem8]
def NEGm16 : I2A16<"neg", 0xF7, MRM3m >, OpSize; // [mem16] = -[mem16] = 0-[mem16]
def NEGm32 : I2A32<"neg", 0xF7, MRM3m >; // [mem32] = -[mem32] = 0-[mem32]
def NOTr8 : I2A8 <"not", 0xF6, MRM2r >; // R8 = ~R8 = R8^-1
def NOTr16 : I2A16<"not", 0xF7, MRM2r >, OpSize; // R16 = ~R16 = R16^-1
def NOTr32 : I2A32<"not", 0xF7, MRM2r >; // R32 = ~R32 = R32^-1
def NOTm8 : I2A8 <"not", 0xF6, MRM2m >; // [mem8] = ~[mem8] = [mem8^-1]
def NOTm16 : I2A16<"not", 0xF7, MRM2m >, OpSize; // [mem16] = ~[mem16] = [mem16^-1]
def NOTm32 : I2A32<"not", 0xF7, MRM2m >; // [mem32] = ~[mem32] = [mem32^-1]
def INCr8 : I2A8 <"inc", 0xFE, MRM0r >; // ++R8
def INCr16 : I2A16<"inc", 0xFF, MRM0r >, OpSize; // ++R16
def INCr32 : I2A32<"inc", 0xFF, MRM0r >; // ++R32
def INCm8 : I2A8 <"inc", 0xFE, MRM0m >; // ++R8
def INCm16 : I2A16<"inc", 0xFF, MRM0m >, OpSize; // ++R16
def INCm32 : I2A32<"inc", 0xFF, MRM0m >; // ++R32
def DECr8 : I2A8 <"dec", 0xFE, MRM1r >; // --R8
def DECr16 : I2A16<"dec", 0xFF, MRM1r >, OpSize; // --R16
def DECr32 : I2A32<"dec", 0xFF, MRM1r >; // --R32
def DECm8 : I2A8 <"dec", 0xFE, MRM1m >; // --[mem8]
def DECm16 : I2A16<"dec", 0xFF, MRM1m >, OpSize; // --[mem16]
def DECm32 : I2A32<"dec", 0xFF, MRM1m >; // --[mem32]
// Arithmetic...
def ADDrr8 : I2A8 <"add", 0x00, MRMDestReg>, Pattern<(set R8 , (plus R8 , R8 ))>;
def ADDrr16 : I2A16<"add", 0x01, MRMDestReg>, OpSize, Pattern<(set R16, (plus R16, R16))>;
def ADDrr32 : I2A32<"add", 0x01, MRMDestReg>, Pattern<(set R32, (plus R32, R32))>;
def ADDmr8 : I2A8 <"add", 0x00, MRMDestMem>; // [mem8] += R8
def ADDmr16 : I2A16<"add", 0x01, MRMDestMem>, OpSize; // [mem16] += R16
def ADDmr32 : I2A32<"add", 0x01, MRMDestMem>; // [mem32] += R32
def ADDrm8 : I2A8 <"add", 0x02, MRMSrcMem >; // R8 += [mem8]
def ADDrm16 : I2A16<"add", 0x03, MRMSrcMem >, OpSize; // R16 += [mem16]
def ADDrm32 : I2A32<"add", 0x03, MRMSrcMem >; // R32 += [mem32]
def ADDri8 : I2A8 <"add", 0x80, MRM0r >, Pattern<(set R8 , (plus R8 , imm))>;
def ADDri16 : I2A16<"add", 0x81, MRM0r >, OpSize, Pattern<(set R16, (plus R16, imm))>;
def ADDri32 : I2A32<"add", 0x81, MRM0r >, Pattern<(set R32, (plus R32, imm))>;
def ADDmi8 : I2A8 <"add", 0x80, MRM0m >; // [mem8] += I8
def ADDmi16 : I2A16<"add", 0x81, MRM0m >, OpSize; // [mem16] += I16
def ADDmi32 : I2A32<"add", 0x81, MRM0m >; // [mem32] += I8
def ADDri16b : I2A8 <"add", 0x83, MRM0r >, OpSize; // ADDri with sign extended 8 bit imm
def ADDri32b : I2A8 <"add", 0x83, MRM0r >;
def ADDmi16b : I2A8 <"add", 0x83, MRM0m >, OpSize; // [mem16] += I8
def ADDmi32b : I2A8 <"add", 0x83, MRM0m >; // [mem32] += I32
def ADCrr32 : I2A32<"adc", 0x11, MRMDestReg>; // R32 += R32+Carry
def ADCrm32 : I2A32<"adc", 0x11, MRMSrcMem >; // R32 += [mem32]+Carry
def ADCmr32 : I2A32<"adc", 0x13, MRMDestMem>; // [mem32] += R32+Carry
def SUBrr8 : I2A8 <"sub", 0x28, MRMDestReg>, Pattern<(set R8 , (minus R8 , R8 ))>;
def SUBrr16 : I2A16<"sub", 0x29, MRMDestReg>, OpSize, Pattern<(set R16, (minus R16, R16))>;
def SUBrr32 : I2A32<"sub", 0x29, MRMDestReg>, Pattern<(set R32, (minus R32, R32))>;
def SUBmr8 : I2A8 <"sub", 0x28, MRMDestMem>; // [mem8] -= R8
def SUBmr16 : I2A16<"sub", 0x29, MRMDestMem>, OpSize; // [mem16] -= R16
def SUBmr32 : I2A32<"sub", 0x29, MRMDestMem>; // [mem32] -= R32
def SUBrm8 : I2A8 <"sub", 0x2A, MRMSrcMem >; // R8 -= [mem8]
def SUBrm16 : I2A16<"sub", 0x2B, MRMSrcMem >, OpSize; // R16 -= [mem16]
def SUBrm32 : I2A32<"sub", 0x2B, MRMSrcMem >; // R32 -= [mem32]
def SUBri8 : I2A8 <"sub", 0x80, MRM5r >, Pattern<(set R8 , (minus R8 , imm))>;
def SUBri16 : I2A16<"sub", 0x81, MRM5r >, OpSize, Pattern<(set R16, (minus R16, imm))>;
def SUBri32 : I2A32<"sub", 0x81, MRM5r >, Pattern<(set R32, (minus R32, imm))>;
def SUBmi8 : I2A8 <"sub", 0x80, MRM5m >; // [mem8] -= I8
def SUBmi16 : I2A16<"sub", 0x81, MRM5m >, OpSize; // [mem16] -= I16
def SUBmi32 : I2A32<"sub", 0x81, MRM5m >; // [mem32] -= I8
def SUBri16b : I2A8 <"sub", 0x83, MRM5r >, OpSize;
def SUBri32b : I2A8 <"sub", 0x83, MRM5r >;
def SUBmi16b : I2A8 <"sub", 0x83, MRM5m >, OpSize; // [mem16] -= I8
def SUBmi32b : I2A8 <"sub", 0x83, MRM5m >; // [mem32] -= I32
def SBBrr32 : I2A32<"sbb", 0x19, MRMDestReg>; // R32 -= R32+Borrow
def SBBrm32 : I2A32<"sbb", 0x19, MRMSrcMem >; // R32 -= [mem32]+Borrow
def SBBmr32 : I2A32<"sbb", 0x1B, MRMDestMem>; // [mem32] -= R32+Borrow
def IMULrr16 : I2A16<"imul", 0xAF, MRMSrcReg>, TB, OpSize, Pattern<(set R16, (times R16, R16))>;
def IMULrr32 : I2A32<"imul", 0xAF, MRMSrcReg>, TB , Pattern<(set R32, (times R32, R32))>;
def IMULrm16 : I2A16<"imul", 0xAF, MRMSrcMem>, TB, OpSize;
def IMULrm32 : I2A32<"imul", 0xAF, MRMSrcMem>, TB ;
// These are suprisingly enough not two address instructions!
def IMULrri16 : X86Inst<"imul", 0x69, MRMSrcReg, Arg16>, OpSize; // R16 = R16*I16
def IMULrri32 : X86Inst<"imul", 0x69, MRMSrcReg, Arg32>; // R32 = R32*I32
def IMULrri16b : X86Inst<"imul", 0x6B, MRMSrcReg, Arg8 >, OpSize; // R16 = R16*I8
def IMULrri32b : X86Inst<"imul", 0x6B, MRMSrcReg, Arg8 >; // R32 = R32*I8
def IMULrmi16 : X86Inst<"imul", 0x69, MRMSrcMem, Arg16>, OpSize; // R16 = [mem16]*I16
def IMULrmi32 : X86Inst<"imul", 0x69, MRMSrcMem, Arg32>; // R32 = [mem32]*I32
def IMULrmi16b : X86Inst<"imul", 0x6B, MRMSrcMem, Arg8 >, OpSize; // R16 = [mem16]*I8
def IMULrmi32b : X86Inst<"imul", 0x6B, MRMSrcMem, Arg8 >; // R32 = [mem32]*I8
// Logical operators...
def ANDrr8 : I2A8 <"and", 0x20, MRMDestReg>, Pattern<(set R8 , (and R8 , R8 ))>;
def ANDrr16 : I2A16<"and", 0x21, MRMDestReg>, OpSize, Pattern<(set R16, (and R16, R16))>;
def ANDrr32 : I2A32<"and", 0x21, MRMDestReg>, Pattern<(set R32, (and R32, R32))>;
def ANDmr8 : I2A8 <"and", 0x20, MRMDestMem>; // [mem8] &= R8
def ANDmr16 : I2A16<"and", 0x21, MRMDestMem>, OpSize; // [mem16] &= R16
def ANDmr32 : I2A32<"and", 0x21, MRMDestMem>; // [mem32] &= R32
def ANDrm8 : I2A8 <"and", 0x22, MRMSrcMem >; // R8 &= [mem8]
def ANDrm16 : I2A16<"and", 0x23, MRMSrcMem >, OpSize; // R16 &= [mem16]
def ANDrm32 : I2A32<"and", 0x23, MRMSrcMem >; // R32 &= [mem32]
def ANDri8 : I2A8 <"and", 0x80, MRM4r >, Pattern<(set R8 , (and R8 , imm))>;
def ANDri16 : I2A16<"and", 0x81, MRM4r >, OpSize, Pattern<(set R16, (and R16, imm))>;
def ANDri32 : I2A32<"and", 0x81, MRM4r >, Pattern<(set R32, (and R32, imm))>;
def ANDmi8 : I2A8 <"and", 0x80, MRM4m >; // [mem8] &= imm8
def ANDmi16 : I2A16<"and", 0x81, MRM4m >, OpSize; // [mem16] &= imm16
def ANDmi32 : I2A32<"and", 0x81, MRM4m >; // [mem32] &= imm32
def ANDri16b : I2A8 <"and", 0x83, MRM4r >, OpSize; // R16 &= imm8
def ANDri32b : I2A8 <"and", 0x83, MRM4r >; // R32 &= imm8
def ANDmi16b : I2A8 <"and", 0x83, MRM4m >, OpSize; // [mem16] &= imm8
def ANDmi32b : I2A8 <"and", 0x83, MRM4m >; // [mem32] &= imm8
def ORrr8 : I2A8 <"or" , 0x08, MRMDestReg>, Pattern<(set R8 , (or R8 , R8 ))>;
def ORrr16 : I2A16<"or" , 0x09, MRMDestReg>, OpSize, Pattern<(set R16, (or R16, R16))>;
def ORrr32 : I2A32<"or" , 0x09, MRMDestReg>, Pattern<(set R32, (or R32, R32))>;
def ORmr8 : I2A8 <"or" , 0x08, MRMDestMem>; // [mem8] |= R8
def ORmr16 : I2A16<"or" , 0x09, MRMDestMem>, OpSize; // [mem16] |= R16
def ORmr32 : I2A32<"or" , 0x09, MRMDestMem>; // [mem32] |= R32
def ORrm8 : I2A8 <"or" , 0x0A, MRMSrcMem >; // R8 |= [mem8]
def ORrm16 : I2A16<"or" , 0x0B, MRMSrcMem >, OpSize; // R16 |= [mem16]
def ORrm32 : I2A32<"or" , 0x0B, MRMSrcMem >; // R32 |= [mem32]
def ORri8 : I2A8 <"or" , 0x80, MRM1r >, Pattern<(set R8 , (or R8 , imm))>;
def ORri16 : I2A16<"or" , 0x81, MRM1r >, OpSize, Pattern<(set R16, (or R16, imm))>;
def ORri32 : I2A32<"or" , 0x81, MRM1r >, Pattern<(set R32, (or R32, imm))>;
def ORmi8 : I2A8 <"or" , 0x80, MRM1m >; // [mem8] |= imm8
def ORmi16 : I2A16<"or" , 0x81, MRM1m >, OpSize; // [mem16] |= imm16
def ORmi32 : I2A32<"or" , 0x81, MRM1m >; // [mem32] |= imm32
def ORri16b : I2A8 <"or" , 0x83, MRM1r >, OpSize; // R16 |= imm8
def ORri32b : I2A8 <"or" , 0x83, MRM1r >; // R32 |= imm8
def ORmi16b : I2A8 <"or" , 0x83, MRM1m >, OpSize; // [mem16] |= imm8
def ORmi32b : I2A8 <"or" , 0x83, MRM1m >; // [mem32] |= imm8
def XORrr8 : I2A8 <"xor", 0x30, MRMDestReg>, Pattern<(set R8 , (xor R8 , R8 ))>;
def XORrr16 : I2A16<"xor", 0x31, MRMDestReg>, OpSize, Pattern<(set R16, (xor R16, R16))>;
def XORrr32 : I2A32<"xor", 0x31, MRMDestReg>, Pattern<(set R32, (xor R32, R32))>;
def XORmr8 : I2A8 <"xor", 0x30, MRMDestMem>; // [mem8] ^= R8
def XORmr16 : I2A16<"xor", 0x31, MRMDestMem>, OpSize; // [mem16] ^= R16
def XORmr32 : I2A32<"xor", 0x31, MRMDestMem>; // [mem32] ^= R32
def XORrm8 : I2A8 <"xor", 0x32, MRMSrcMem >; // R8 ^= [mem8]
def XORrm16 : I2A16<"xor", 0x33, MRMSrcMem >, OpSize; // R16 ^= [mem16]
def XORrm32 : I2A32<"xor", 0x33, MRMSrcMem >; // R32 ^= [mem32]
def XORri8 : I2A8 <"xor", 0x80, MRM6r >, Pattern<(set R8 , (xor R8 , imm))>;
def XORri16 : I2A16<"xor", 0x81, MRM6r >, OpSize, Pattern<(set R16, (xor R16, imm))>;
def XORri32 : I2A32<"xor", 0x81, MRM6r >, Pattern<(set R32, (xor R32, imm))>;
def XORmi8 : I2A8 <"xor", 0x80, MRM6m >; // [mem8] ^= R8
def XORmi16 : I2A16<"xor", 0x81, MRM6m >, OpSize; // [mem16] ^= R16
def XORmi32 : I2A32<"xor", 0x81, MRM6m >; // [mem32] ^= R32
def XORri16b : I2A8 <"xor", 0x83, MRM6r >, OpSize; // R16 ^= imm8
def XORri32b : I2A8 <"xor", 0x83, MRM6r >; // R32 ^= imm8
def XORmi16b : I2A8 <"xor", 0x83, MRM6m >, OpSize; // [mem16] ^= imm8
def XORmi32b : I2A8 <"xor", 0x83, MRM6m >; // [mem32] ^= imm8
// Test instructions are just like AND, except they don't generate a result.
def TESTrr8 : X86Inst<"test", 0x84, MRMDestReg, Arg8 >; // flags = R8 & R8
def TESTrr16 : X86Inst<"test", 0x85, MRMDestReg, Arg16>, OpSize; // flags = R16 & R16
def TESTrr32 : X86Inst<"test", 0x85, MRMDestReg, Arg32>; // flags = R32 & R32
def TESTmr8 : X86Inst<"test", 0x84, MRMDestMem, Arg8 >; // flags = [mem8] & R8
def TESTmr16 : X86Inst<"test", 0x85, MRMDestMem, Arg16>, OpSize; // flags = [mem16] & R16
def TESTmr32 : X86Inst<"test", 0x85, MRMDestMem, Arg32>; // flags = [mem32] & R32
def TESTrm8 : X86Inst<"test", 0x84, MRMSrcMem , Arg8 >; // flags = R8 & [mem8]
def TESTrm16 : X86Inst<"test", 0x85, MRMSrcMem , Arg16>, OpSize; // flags = R16 & [mem16]
def TESTrm32 : X86Inst<"test", 0x85, MRMSrcMem , Arg32>; // flags = R32 & [mem32]
def TESTri8 : X86Inst<"test", 0xF6, MRM0r , Arg8 >; // flags = R8 & imm8
def TESTri16 : X86Inst<"test", 0xF7, MRM0r , Arg16>, OpSize; // flags = R16 & imm16
def TESTri32 : X86Inst<"test", 0xF7, MRM0r , Arg32>; // flags = R32 & imm32
def TESTmi8 : X86Inst<"test", 0xF6, MRM0m , Arg8 >; // flags = [mem8] & imm8
def TESTmi16 : X86Inst<"test", 0xF7, MRM0m , Arg16>, OpSize; // flags = [mem16] & imm16
def TESTmi32 : X86Inst<"test", 0xF7, MRM0m , Arg32>; // flags = [mem32] & imm32
// Shift instructions
class UsesCL { list<Register> Uses = [CL]; bit printImplicitUses = 1; }
def SHLrCL8 : I2A8 <"shl", 0xD2, MRM4r > , UsesCL; // R8 <<= cl
def SHLrCL16 : I2A16<"shl", 0xD3, MRM4r >, OpSize, UsesCL; // R16 <<= cl
def SHLrCL32 : I2A32<"shl", 0xD3, MRM4r > , UsesCL; // R32 <<= cl
def SHLmCL8 : I2A8 <"shl", 0xD2, MRM4m > , UsesCL; // [mem8] <<= cl
def SHLmCL16 : I2A16<"shl", 0xD3, MRM4m >, OpSize, UsesCL; // [mem16] <<= cl
def SHLmCL32 : I2A32<"shl", 0xD3, MRM4m > , UsesCL; // [mem32] <<= cl
def SHLri8 : I2A8 <"shl", 0xC0, MRM4r >; // R8 <<= imm8
def SHLri16 : I2A16<"shl", 0xC1, MRM4r >, OpSize; // R16 <<= imm16
def SHLri32 : I2A32<"shl", 0xC1, MRM4r >; // R32 <<= imm32
def SHLmi8 : I2A8 <"shl", 0xC0, MRM4m >; // [mem8] <<= imm8
def SHLmi16 : I2A16<"shl", 0xC1, MRM4m >, OpSize; // [mem16] <<= imm16
def SHLmi32 : I2A32<"shl", 0xC1, MRM4m >; // [mem32] <<= imm32
def SHRrCL8 : I2A8 <"shr", 0xD2, MRM5r > , UsesCL; // R8 >>= cl
def SHRrCL16 : I2A16<"shr", 0xD3, MRM5r >, OpSize, UsesCL; // R16 >>= cl
def SHRrCL32 : I2A32<"shr", 0xD3, MRM5r > , UsesCL; // R32 >>= cl
def SHRmCL8 : I2A8 <"shr", 0xD2, MRM5m > , UsesCL; // [mem8] >>= cl
def SHRmCL16 : I2A16<"shr", 0xD3, MRM5m >, OpSize, UsesCL; // [mem16] >>= cl
def SHRmCL32 : I2A32<"shr", 0xD3, MRM5m > , UsesCL; // [mem32] >>= cl
def SHRri8 : I2A8 <"shr", 0xC0, MRM5r >; // R8 >>= imm8
def SHRri16 : I2A16<"shr", 0xC1, MRM5r >, OpSize; // R16 >>= imm16
def SHRri32 : I2A32<"shr", 0xC1, MRM5r >; // R32 >>= imm32
def SHRmi8 : I2A8 <"shr", 0xC0, MRM5m >; // [mem8] >>= imm8
def SHRmi16 : I2A16<"shr", 0xC1, MRM5m >, OpSize; // [mem16] >>= imm16
def SHRmi32 : I2A32<"shr", 0xC1, MRM5m >; // [mem32] >>= imm32
def SARrCL8 : I2A8 <"sar", 0xD2, MRM7r > , UsesCL; // R8 >>>= cl
def SARrCL16 : I2A16<"sar", 0xD3, MRM7r >, OpSize, UsesCL; // R16 >>>= cl
def SARrCL32 : I2A32<"sar", 0xD3, MRM7r > , UsesCL; // R32 >>>= cl
def SARmCL8 : I2A8 <"sar", 0xD2, MRM7m > , UsesCL; // [mem8] >>>= cl
def SARmCL16 : I2A16<"sar", 0xD3, MRM7m >, OpSize, UsesCL; // [mem16] >>>= cl
def SARmCL32 : I2A32<"sar", 0xD3, MRM7m > , UsesCL; // [mem32] >>>= cl
def SARri8 : I2A8 <"sar", 0xC0, MRM7r >; // R8 >>>= imm8
def SARri16 : I2A16<"sar", 0xC1, MRM7r >, OpSize; // R16 >>>= imm16
def SARri32 : I2A32<"sar", 0xC1, MRM7r >; // R32 >>>= imm32
def SARmi8 : I2A8 <"sar", 0xC0, MRM7m >; // [mem8] >>>= imm8
def SARmi16 : I2A16<"sar", 0xC1, MRM7m >, OpSize; // [mem16] >>>= imm16
def SARmi32 : I2A32<"sar", 0xC1, MRM7m >; // [mem32] >>>= imm32
def SHLDrrCL32 : I2A32<"shld", 0xA5, MRMDestReg>, TB, UsesCL; // R32 <<= R32,R32 cl
def SHLDmrCL32 : I2A32<"shld", 0xA5, MRMDestMem>, TB, UsesCL; // [mem32] <<= [mem32],R32 cl
def SHLDrri32 : I2A32<"shld", 0xA4, MRMDestReg>, TB; // R32 <<= R32,R32 imm8
def SHLDmri32 : I2A32<"shld", 0xA4, MRMDestMem>, TB; // [mem32] <<= [mem32],R32 imm8
def SHRDrrCL32 : I2A32<"shrd", 0xAD, MRMDestReg>, TB, UsesCL; // R32 >>= R32,R32 cl
def SHRDmrCL32 : I2A32<"shrd", 0xAD, MRMDestMem>, TB, UsesCL; // [mem32] >>= [mem32],R32 cl
def SHRDrri32 : I2A32<"shrd", 0xAC, MRMDestReg>, TB; // R32 >>= R32,R32 imm8
def SHRDmri32 : I2A32<"shrd", 0xAC, MRMDestMem>, TB; // [mem32] >>= [mem32],R32 imm8
// Condition code ops, incl. set if equal/not equal/...
def SAHF : X86Inst<"sahf" , 0x9E, RawFrm, Arg8>, Imp<[AH],[]>; // flags = AH
def SETBr : X86Inst<"setb" , 0x92, MRM0r , Arg8>, TB; // R8 = < unsign
def SETBm : X86Inst<"setb" , 0x92, MRM0m , Arg8>, TB; // [mem8] = < unsign
def SETAEr : X86Inst<"setae", 0x93, MRM0r , Arg8>, TB; // R8 = >= unsign
def SETAEm : X86Inst<"setae", 0x93, MRM0m , Arg8>, TB; // [mem8] = >= unsign
def SETEr : X86Inst<"sete" , 0x94, MRM0r , Arg8>, TB; // R8 = ==
def SETEm : X86Inst<"sete" , 0x94, MRM0m , Arg8>, TB; // [mem8] = ==
def SETNEr : X86Inst<"setne", 0x95, MRM0r , Arg8>, TB; // R8 = !=
def SETNEm : X86Inst<"setne", 0x95, MRM0m , Arg8>, TB; // [mem8] = !=
def SETBEr : X86Inst<"setbe", 0x96, MRM0r , Arg8>, TB; // R8 = <= unsign
def SETBEm : X86Inst<"setbe", 0x96, MRM0m , Arg8>, TB; // [mem8] = <= unsign
def SETAr : X86Inst<"seta" , 0x97, MRM0r , Arg8>, TB; // R8 = > signed
def SETAm : X86Inst<"seta" , 0x97, MRM0m , Arg8>, TB; // [mem8] = > signed
def SETSr : X86Inst<"sets" , 0x98, MRM0r , Arg8>, TB; // R8 = <sign bit>
def SETSm : X86Inst<"sets" , 0x98, MRM0m , Arg8>, TB; // [mem8] = <sign bit>
def SETNSr : X86Inst<"setns", 0x99, MRM0r , Arg8>, TB; // R8 = !<sign bit>
def SETNSm : X86Inst<"setns", 0x99, MRM0m , Arg8>, TB; // [mem8] = !<sign bit>
def SETLr : X86Inst<"setl" , 0x9C, MRM0r , Arg8>, TB; // R8 = < signed
def SETLm : X86Inst<"setl" , 0x9C, MRM0m , Arg8>, TB; // [mem8] = < signed
def SETGEr : X86Inst<"setge", 0x9D, MRM0r , Arg8>, TB; // R8 = >= signed
def SETGEm : X86Inst<"setge", 0x9D, MRM0m , Arg8>, TB; // [mem8] = >= signed
def SETLEr : X86Inst<"setle", 0x9E, MRM0r , Arg8>, TB; // R8 = <= signed
def SETLEm : X86Inst<"setle", 0x9E, MRM0m , Arg8>, TB; // [mem8] = <= signed
def SETGr : X86Inst<"setg" , 0x9F, MRM0r , Arg8>, TB; // R8 = < signed
def SETGm : X86Inst<"setg" , 0x9F, MRM0m , Arg8>, TB; // [mem8] = < signed
// Conditional moves. These are modelled as X = cmovXX Y, Z. Eventually
// register allocated to cmovXX XY, Z
def CMOVErr16 : I2A16<"cmove", 0x44, MRMSrcReg>, TB, OpSize; // if ==, R16 = R16
def CMOVNErr32: I2A32<"cmovne",0x45, MRMSrcReg>, TB; // if !=, R32 = R32
def CMOVSrr32 : I2A32<"cmovs", 0x48, MRMSrcReg>, TB; // if signed, R32 = R32
// Integer comparisons
def CMPrr8 : X86Inst<"cmp", 0x38, MRMDestReg, Arg8 >; // compare R8, R8
def CMPrr16 : X86Inst<"cmp", 0x39, MRMDestReg, Arg16>, OpSize; // compare R16, R16
def CMPrr32 : X86Inst<"cmp", 0x39, MRMDestReg, Arg32>, // compare R32, R32
Pattern<(isVoid (unspec2 R32, R32))>;
def CMPmr8 : X86Inst<"cmp", 0x38, MRMDestMem, Arg8 >; // compare [mem8], R8
def CMPmr16 : X86Inst<"cmp", 0x39, MRMDestMem, Arg16>, OpSize; // compare [mem16], R16
def CMPmr32 : X86Inst<"cmp", 0x39, MRMDestMem, Arg32>; // compare [mem32], R32
def CMPrm8 : X86Inst<"cmp", 0x3A, MRMSrcMem , Arg8 >; // compare R8, [mem8]
def CMPrm16 : X86Inst<"cmp", 0x3B, MRMSrcMem , Arg16>, OpSize; // compare R16, [mem16]
def CMPrm32 : X86Inst<"cmp", 0x3B, MRMSrcMem , Arg32>; // compare R32, [mem32]
def CMPri8 : X86Inst<"cmp", 0x80, MRM7r , Arg8 >; // compare R8, imm8
def CMPri16 : X86Inst<"cmp", 0x81, MRM7r , Arg16>, OpSize; // compare R16, imm16
def CMPri32 : X86Inst<"cmp", 0x81, MRM7r , Arg32>; // compare R32, imm32
def CMPmi8 : X86Inst<"cmp", 0x80, MRM7m , Arg8 >; // compare [mem8], imm8
def CMPmi16 : X86Inst<"cmp", 0x81, MRM7m , Arg16>, OpSize; // compare [mem16], imm16
def CMPmi32 : X86Inst<"cmp", 0x81, MRM7m , Arg32>; // compare [mem32], imm32
// Sign/Zero extenders
def MOVSXr16r8 : X86Inst<"movsx", 0xBE, MRMSrcReg, Arg8>, TB, OpSize; // R16 = signext(R8)
def MOVSXr32r8 : X86Inst<"movsx", 0xBE, MRMSrcReg, Arg8>, TB; // R32 = signext(R8)
def MOVSXr32r16: X86Inst<"movsx", 0xBF, MRMSrcReg, Arg16>,TB; // R32 = signext(R16)
def MOVSXr16m8 : X86Inst<"movsx", 0xBE, MRMSrcMem, Arg8>, TB, OpSize; // R16 = signext([mem8])
def MOVSXr32m8 : X86Inst<"movsx", 0xBE, MRMSrcMem, Arg8>, TB; // R32 = signext([mem8])
def MOVSXr32m16: X86Inst<"movsx", 0xBF, MRMSrcMem, Arg16>,TB; // R32 = signext([mem16])
def MOVZXr16r8 : X86Inst<"movzx", 0xB6, MRMSrcReg, Arg8>, TB, OpSize; // R16 = zeroext(R8)
def MOVZXr32r8 : X86Inst<"movzx", 0xB6, MRMSrcReg, Arg8>, TB; // R32 = zeroext(R8)
def MOVZXr32r16: X86Inst<"movzx", 0xB7, MRMSrcReg, Arg16>,TB; // R32 = zeroext(R16)
def MOVZXr16m8 : X86Inst<"movzx", 0xB6, MRMSrcMem, Arg8>, TB, OpSize; // R16 = zeroext([mem8])
def MOVZXr32m8 : X86Inst<"movzx", 0xB6, MRMSrcMem, Arg8>, TB; // R32 = zeroext([mem8])
def MOVZXr32m16: X86Inst<"movzx", 0xB7, MRMSrcMem, Arg16>,TB; // R32 = zeroext([mem16])
//===----------------------------------------------------------------------===//
// Floating point support
//===----------------------------------------------------------------------===//
// FIXME: These need to indicate mod/ref sets for FP regs... & FP 'TOP'
// Floating point pseudo instructions...
class FPInst<string n, bits<8> o, Format F, ArgType t, FPFormat fp>
: X86Inst<n, o, F, t> { let FPForm = fp; let FPFormBits = FPForm.Value; }
// Pseudo instructions for floating point. We use these pseudo instructions
// because they can be expanded by the fp spackifier into one of many different
// forms of instructions for doing these operations. Until the stackifier runs,
// we prefer to be abstract.
def FpMOV : FPInst<"FMOV", 0, Pseudo, ArgF80, SpecialFP>; // f1 = fmov f2
def FpADD : FPInst<"FADD", 0, Pseudo, ArgF80, TwoArgFP>; // f1 = fadd f2, f3
def FpSUB : FPInst<"FSUB", 0, Pseudo, ArgF80, TwoArgFP>; // f1 = fsub f2, f3
def FpMUL : FPInst<"FMUL", 0, Pseudo, ArgF80, TwoArgFP>; // f1 = fmul f2, f3
def FpDIV : FPInst<"FDIV", 0, Pseudo, ArgF80, TwoArgFP>; // f1 = fdiv f2, f3
def FpUCOM : FPInst<"FUCOM", 0, Pseudo, ArgF80, TwoArgFP>; // FPSW = fucom f1, f2
def FpGETRESULT : FPInst<"FGETRESULT",0, Pseudo, ArgF80, SpecialFP>; // FPR = ST(0)
def FpSETRESULT : FPInst<"FSETRESULT",0, Pseudo, ArgF80, SpecialFP>; // ST(0) = FPR
// Floating point loads & stores...
def FLDrr : FPInst<"fld" , 0xC0, AddRegFrm, ArgF80, NotFP>, D9; // push(ST(i))
def FLDr32 : FPInst<"fld" , 0xD9, MRM0m , ArgF32, ZeroArgFP>; // load float
def FLDr64 : FPInst<"fld" , 0xDD, MRM0m , ArgF64, ZeroArgFP>; // load double
def FLDr80 : FPInst<"fld" , 0xDB, MRM5m , ArgF80, ZeroArgFP>; // load extended
def FILDr16 : FPInst<"fild" , 0xDF, MRM0m , Arg16 , ZeroArgFP>; // load signed short
def FILDr32 : FPInst<"fild" , 0xDB, MRM0m , Arg32 , ZeroArgFP>; // load signed int
def FILDr64 : FPInst<"fild" , 0xDF, MRM5m , Arg64 , ZeroArgFP>; // load signed long
def FSTr32 : FPInst<"fst" , 0xD9, MRM2m , ArgF32, OneArgFP>; // store float
def FSTr64 : FPInst<"fst" , 0xDD, MRM2m , ArgF64, OneArgFP>; // store double
def FSTPr32 : FPInst<"fstp", 0xD9, MRM3m , ArgF32, OneArgFP>; // store float, pop
def FSTPr64 : FPInst<"fstp", 0xDD, MRM3m , ArgF64, OneArgFP>; // store double, pop
def FSTPr80 : FPInst<"fstp", 0xDB, MRM7m , ArgF80, OneArgFP>; // store extended, pop
def FSTrr : FPInst<"fst" , 0xD0, AddRegFrm, ArgF80, NotFP >, DD; // ST(i) = ST(0)
def FSTPrr : FPInst<"fstp", 0xD8, AddRegFrm, ArgF80, NotFP >, DD; // ST(i) = ST(0), pop
def FISTr16 : FPInst<"fist", 0xDF, MRM2m , Arg16 , OneArgFP>; // store signed short
def FISTr32 : FPInst<"fist", 0xDB, MRM2m , Arg32 , OneArgFP>; // store signed int
def FISTPr16 : FPInst<"fistp", 0xDF, MRM3m , Arg16 , NotFP >; // store signed short, pop
def FISTPr32 : FPInst<"fistp", 0xDB, MRM3m , Arg32 , NotFP >; // store signed int, pop
def FISTPr64 : FPInst<"fistpll", 0xDF, MRM7m , Arg64 , OneArgFP>; // store signed long, pop
def FXCH : FPInst<"fxch", 0xC8, AddRegFrm, ArgF80, NotFP>, D9; // fxch ST(i), ST(0)
// Floating point constant loads...
def FLD0 : FPInst<"fldz", 0xEE, RawFrm, ArgF80, ZeroArgFP>, D9;
def FLD1 : FPInst<"fld1", 0xE8, RawFrm, ArgF80, ZeroArgFP>, D9;
// Unary operations...
def FCHS : FPInst<"fchs", 0xE0, RawFrm, ArgF80, OneArgFPRW>, D9; // f1 = fchs f2
def FTST : FPInst<"ftst", 0xE4, RawFrm, ArgF80, OneArgFP>, D9; // ftst ST(0)
// Binary arithmetic operations...
class FPST0rInst<string n, bits<8> o>
: X86Inst<n, o, AddRegFrm, ArgF80>, D8 {
list<Register> Uses = [ST0];
list<Register> Defs = [ST0];
}
class FPrST0Inst<string n, bits<8> o>
: X86Inst<n, o, AddRegFrm, ArgF80>, DC {
bit printImplicitUses = 1;
list<Register> Uses = [ST0];
}
class FPrST0PInst<string n, bits<8> o>
: X86Inst<n, o, AddRegFrm, ArgF80>, DE {
list<Register> Uses = [ST0];
}
def FADDST0r : FPST0rInst <"fadd", 0xC0>;
def FADDrST0 : FPrST0Inst <"fadd", 0xC0>;
def FADDPrST0 : FPrST0PInst<"faddp", 0xC0>;
def FSUBRST0r : FPST0rInst <"fsubr", 0xE8>;
def FSUBrST0 : FPrST0Inst <"fsub", 0xE8>;
def FSUBPrST0 : FPrST0PInst<"fsubp", 0xE8>;
def FSUBST0r : FPST0rInst <"fsub", 0xE0>;
def FSUBRrST0 : FPrST0Inst <"fsubr", 0xE0>;
def FSUBRPrST0 : FPrST0PInst<"fsubrp", 0xE0>;
def FMULST0r : FPST0rInst <"fmul", 0xC8>;
def FMULrST0 : FPrST0Inst <"fmul", 0xC8>;
def FMULPrST0 : FPrST0PInst<"fmulp", 0xC8>;
def FDIVRST0r : FPST0rInst <"fdivr", 0xF8>;
def FDIVrST0 : FPrST0Inst <"fdiv", 0xF8>;
def FDIVPrST0 : FPrST0PInst<"fdivp", 0xF8>;
def FDIVST0r : FPST0rInst <"fdiv", 0xF0>; // ST(0) = ST(0) / ST(i)
def FDIVRrST0 : FPrST0Inst <"fdivr", 0xF0>; // ST(i) = ST(0) / ST(i)
def FDIVRPrST0 : FPrST0PInst<"fdivrp", 0xF0>; // ST(i) = ST(0) / ST(i), pop
// Floating point compares
def FUCOMr : X86Inst<"fucom" , 0xE0, AddRegFrm, ArgF80>, DD, Imp<[ST0],[]>; // FPSW = compare ST(0) with ST(i)
def FUCOMPr : X86Inst<"fucomp" , 0xE8, AddRegFrm, ArgF80>, DD, Imp<[ST0],[]>; // FPSW = compare ST(0) with ST(i), pop
def FUCOMPPr : X86Inst<"fucompp", 0xE9, RawFrm , ArgF80>, DA, Imp<[ST0],[]>; // compare ST(0) with ST(1), pop, pop
// Floating point flag ops
def FNSTSWr8 : X86Inst<"fnstsw" , 0xE0, RawFrm , ArgF80>, DF, Imp<[],[AX]>; // AX = fp flags
def FNSTCWm16 : X86Inst<"fnstcw" , 0xD9, MRM7m , Arg16 >; // [mem16] = X87 control world
def FLDCWm16 : X86Inst<"fldcw" , 0xD9, MRM5m , Arg16 >; // X87 control world = [mem16]
//===----------------------------------------------------------------------===//
// Instruction Expanders
//
def RET_R32 : Expander<(ret R32:$reg),
[(MOVrr32 EAX, R32:$reg),
(RET)]>;
// FIXME: This should eventually just be implemented by defining a frameidx as a
// value address for a load.
def LOAD_FI16 : Expander<(set R16:$dest, (load frameidx:$fi)),
[(MOVrm16 R16:$dest, frameidx:$fi, 1, 0/*NoReg*/, 0)]>;
def LOAD_FI32 : Expander<(set R32:$dest, (load frameidx:$fi)),
[(MOVrm32 R32:$dest, frameidx:$fi, 1, 0/*NoReg*/, 0)]>;
def LOAD_R16 : Expander<(set R16:$dest, (load R32:$src)),
[(MOVrm16 R16:$dest, R32:$src, 1, 0/*NoReg*/, 0)]>;
def LOAD_R32 : Expander<(set R32:$dest, (load R32:$src)),
[(MOVrm32 R32:$dest, R32:$src, 1, 0/*NoReg*/, 0)]>;
def BR_EQ : Expander<(brcond (seteq R32:$a1, R32:$a2),
basicblock:$d1, basicblock:$d2),
[(CMPrr32 R32:$a1, R32:$a2),
(JE basicblock:$d1),
(JMP basicblock:$d2)]>;