def Mem16 : MemType<2>;
def Mem32 : MemType<3>;
def Mem64 : MemType<4>;
-def Mem80 : MemType<4>;
+def Mem80 : MemType<5>;
def Mem128 : MemType<6>;
// FPFormat - This specifies what form this FP instruction has. This is used by
def OneArgFP : FPFormat<2>;
def OneArgFPRW : FPFormat<3>;
def TwoArgFP : FPFormat<4>;
-def SpecialFP : FPFormat<5>;
+def CondMovFP : FPFormat<5>;
+def SpecialFP : FPFormat<6>;
class X86Inst<string nam, bits<8> opcod, Format f, MemType m, ImmType i> : Instruction {
// 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?
+ bit printImplicitUsesBefore = 0; // Should we print implicit uses before this inst?
+ bit printImplicitUsesAfter = 0; // Should we print implicit uses after this inst?
bits<4> Prefix = 0; // Which prefix byte does this inst have?
FPFormat FPForm; // What flavor of FP instruction is this?
class I<string n, bits<8> o, Format f> : X86Inst<n, o, f, NoMem, NoImm>;
-class IM<string n, bits<8> o, Format f, MemType m> : X86Inst<n, o, f, m, NoImm>;
-class IM8 <string n, bits<8> o, Format f> : IM<n, o, f, Mem8 >;
-class IM16<string n, bits<8> o, Format f> : IM<n, o, f, Mem16>;
-class IM32<string n, bits<8> o, Format f> : IM<n, o, f, Mem32>;
+class Im<string n, bits<8> o, Format f, MemType m> : X86Inst<n, o, f, m, NoImm>;
+class Im8 <string n, bits<8> o, Format f> : Im<n, o, f, Mem8 >;
+class Im16<string n, bits<8> o, Format f> : Im<n, o, f, Mem16>;
+class Im32<string n, bits<8> o, Format f> : Im<n, o, f, Mem32>;
-class II<string n, bits<8> o, Format f, ImmType i> : X86Inst<n, o, f, NoMem, i>;
-class II8 <string n, bits<8> o, Format f> : II<n, o, f, Imm8 >;
-class II16<string n, bits<8> o, Format f> : II<n, o, f, Imm16>;
-class II32<string n, bits<8> o, Format f> : II<n, o, f, Imm32>;
+class Ii<string n, bits<8> o, Format f, ImmType i> : X86Inst<n, o, f, NoMem, i>;
+class Ii8 <string n, bits<8> o, Format f> : Ii<n, o, f, Imm8 >;
+class Ii16<string n, bits<8> o, Format f> : Ii<n, o, f, Imm16>;
+class Ii32<string n, bits<8> o, Format f> : Ii<n, o, f, Imm32>;
-class I8MI <string n, bits<8> o, Format f> : X86Inst<n, o, f, Mem8 , Imm8 >;
-class I16MI<string n, bits<8> o, Format f> : X86Inst<n, o, f, Mem16, Imm16>;
-class I32MI<string n, bits<8> o, Format f> : X86Inst<n, o, f, Mem32, Imm32>;
+class Im8i8 <string n, bits<8> o, Format f> : X86Inst<n, o, f, Mem8 , Imm8 >;
+class Im16i16<string n, bits<8> o, Format f> : X86Inst<n, o, f, Mem16, Imm16>;
+class Im32i32<string n, bits<8> o, Format f> : X86Inst<n, o, f, Mem32, Imm32>;
-class IM16I8<string n, bits<8> o, Format f> : X86Inst<n, o, f, Mem16, Imm8>;
-class IM32I8<string n, bits<8> o, Format f> : X86Inst<n, o, f, Mem32, Imm8>;
+class Im16i8<string n, bits<8> o, Format f> : X86Inst<n, o, f, Mem16, Imm8>;
+class Im32i8<string n, bits<8> o, Format f> : X86Inst<n, o, f, Mem32, Imm8>;
// Helper for shift instructions
-class UsesCL { list<Register> Uses = [CL]; bit printImplicitUses = 1; }
+class UsesCL { list<Register> Uses = [CL]; bit printImplicitUsesAfter = 1; }
//===----------------------------------------------------------------------===//
// Instruction list...
// All calls clobber the non-callee saved registers...
let Defs = [EAX, ECX, EDX, FP0, FP1, FP2, FP3, FP4, FP5, FP6] in {
def CALLpcrel32 : I <"call", 0xE8, RawFrm>;
- def CALLr32 : I <"call", 0xFF, MRM2r>;
- def CALLm32 : IM32<"call", 0xFF, MRM2m>;
+ def CALL32r : I <"call", 0xFF, MRM2r>;
+ def CALL32m : Im32<"call", 0xFF, MRM2m>;
}
// Miscellaneous Instructions...
//
def LEAVE : I<"leave", 0xC9, RawFrm>, Imp<[EBP,ESP],[EBP,ESP]>;
-def POPr32 : I<"pop", 0x58, AddRegFrm>, Imp<[ESP],[ESP]>;
+def POP32r : I<"pop", 0x58, AddRegFrm>, Imp<[ESP],[ESP]>;
let isTwoAddress = 1 in // R32 = bswap R32
- def BSWAPr32 : I<"bswap", 0xC8, AddRegFrm>, TB;
+ def BSWAP32r : I<"bswap", 0xC8, AddRegFrm>, TB;
-def XCHGrr8 : I <"xchg", 0x86, MRMDestReg>; // xchg R8, R8
-def XCHGrr16 : I <"xchg", 0x87, MRMDestReg>, OpSize; // xchg R16, R16
-def XCHGrr32 : I <"xchg", 0x87, MRMDestReg>; // xchg R32, R32
-def XCHGmr8 : IM8 <"xchg", 0x86, MRMDestMem>; // xchg [mem8], R8
-def XCHGmr16 : IM16<"xchg", 0x87, MRMDestMem>, OpSize; // xchg [mem16], R16
-def XCHGmr32 : IM32<"xchg", 0x87, MRMDestMem>; // xchg [mem32], R32
-def XCHGrm8 : IM8 <"xchg", 0x86, MRMSrcMem >; // xchg R8, [mem8]
-def XCHGrm16 : IM16<"xchg", 0x87, MRMSrcMem >, OpSize; // xchg R16, [mem16]
-def XCHGrm32 : IM32<"xchg", 0x87, MRMSrcMem >; // xchg R32, [mem32]
+def XCHG8rr : I <"xchg", 0x86, MRMDestReg>; // xchg R8, R8
+def XCHG16rr : I <"xchg", 0x87, MRMDestReg>, OpSize; // xchg R16, R16
+def XCHG32rr : I <"xchg", 0x87, MRMDestReg>; // xchg R32, R32
+def XCHG8mr : Im8 <"xchg", 0x86, MRMDestMem>; // xchg [mem8], R8
+def XCHG16mr : Im16<"xchg", 0x87, MRMDestMem>, OpSize; // xchg [mem16], R16
+def XCHG32mr : Im32<"xchg", 0x87, MRMDestMem>; // xchg [mem32], R32
+def XCHG8rm : Im8 <"xchg", 0x86, MRMSrcMem >; // xchg R8, [mem8]
+def XCHG16rm : Im16<"xchg", 0x87, MRMSrcMem >, OpSize; // xchg R16, [mem16]
+def XCHG32rm : Im32<"xchg", 0x87, MRMSrcMem >; // xchg R32, [mem32]
-def LEAr16 : IM32<"lea", 0x8D, MRMSrcMem>, OpSize; // R16 = lea [mem]
-def LEAr32 : IM32<"lea", 0x8D, MRMSrcMem>; // R32 = lea [mem]
+def LEA16r : Im32<"lea", 0x8D, MRMSrcMem>, OpSize; // R16 = lea [mem]
+def LEA32r : Im32<"lea", 0x8D, MRMSrcMem>; // R32 = lea [mem]
def REP_MOVSB : I<"rep movsb", 0xA4, RawFrm>, REP,
//===----------------------------------------------------------------------===//
// Move Instructions...
//
-def MOVrr8 : I <"mov", 0x88, MRMDestReg>, Pattern<(set R8 , R8 )>;
-def MOVrr16 : I <"mov", 0x89, MRMDestReg>, OpSize, Pattern<(set R16, R16)>;
-def MOVrr32 : I <"mov", 0x89, MRMDestReg>, Pattern<(set R32, R32)>;
-def MOVri8 : II8 <"mov", 0xB0, AddRegFrm >, Pattern<(set R8 , imm )>;
-def MOVri16 : II16 <"mov", 0xB8, AddRegFrm >, OpSize, Pattern<(set R16, imm)>;
-def MOVri32 : II32 <"mov", 0xB8, AddRegFrm >, Pattern<(set R32, imm)>;
-def MOVmi8 : I8MI <"mov", 0xC6, MRM0m >; // [mem8] = imm8
-def MOVmi16 : I16MI<"mov", 0xC7, MRM0m >, OpSize; // [mem16] = imm16
-def MOVmi32 : I32MI<"mov", 0xC7, MRM0m >; // [mem32] = imm32
-
-def MOVrm8 : IM8 <"mov", 0x8A, MRMSrcMem>; // R8 = [mem8]
-def MOVrm16 : IM16 <"mov", 0x8B, MRMSrcMem>, OpSize, // R16 = [mem16]
+def MOV8rr : I <"mov", 0x88, MRMDestReg>, Pattern<(set R8 , R8 )>;
+def MOV16rr : I <"mov", 0x89, MRMDestReg>, OpSize, Pattern<(set R16, R16)>;
+def MOV32rr : I <"mov", 0x89, MRMDestReg>, Pattern<(set R32, R32)>;
+def MOV8ri : Ii8 <"mov", 0xB0, AddRegFrm >, Pattern<(set R8 , imm )>;
+def MOV16ri : Ii16 <"mov", 0xB8, AddRegFrm >, OpSize, Pattern<(set R16, imm)>;
+def MOV32ri : Ii32 <"mov", 0xB8, AddRegFrm >, Pattern<(set R32, imm)>;
+def MOV8mi : Im8i8 <"mov", 0xC6, MRM0m >; // [mem8] = imm8
+def MOV16mi : Im16i16<"mov", 0xC7, MRM0m >, OpSize; // [mem16] = imm16
+def MOV32mi : Im32i32<"mov", 0xC7, MRM0m >; // [mem32] = imm32
+
+def MOV8rm : Im8 <"mov", 0x8A, MRMSrcMem>; // R8 = [mem8]
+def MOV16rm : Im16 <"mov", 0x8B, MRMSrcMem>, OpSize, // R16 = [mem16]
Pattern<(set R16, (load (plus R32, (plus (times imm, R32), imm))))>;
-def MOVrm32 : IM32 <"mov", 0x8B, MRMSrcMem>, // R32 = [mem32]
+def MOV32rm : Im32 <"mov", 0x8B, MRMSrcMem>, // R32 = [mem32]
Pattern<(set R32, (load (plus R32, (plus (times imm, R32), imm))))>;
-def MOVmr8 : IM8 <"mov", 0x88, MRMDestMem>; // [mem8] = R8
-def MOVmr16 : IM16 <"mov", 0x89, MRMDestMem>, OpSize; // [mem16] = R16
-def MOVmr32 : IM32 <"mov", 0x89, MRMDestMem>; // [mem32] = R32
+def MOV8mr : Im8 <"mov", 0x88, MRMDestMem>; // [mem8] = R8
+def MOV16mr : Im16 <"mov", 0x89, MRMDestMem>, OpSize; // [mem16] = R16
+def MOV32mr : Im32 <"mov", 0x89, MRMDestMem>; // [mem32] = R32
//===----------------------------------------------------------------------===//
// Fixed-Register Multiplication and Division Instructions...
//
// Extra precision multiplication
-def MULr8 : I <"mul", 0xF6, MRM4r>, Imp<[AL],[AX]>; // AL,AH = AL*R8
-def MULr16 : I <"mul", 0xF7, MRM4r>, Imp<[AX],[AX,DX]>, OpSize; // AX,DX = AX*R16
-def MULr32 : I <"mul", 0xF7, MRM4r>, Imp<[EAX],[EAX,EDX]>; // EAX,EDX = EAX*R32
-def MULm8 : IM8 <"mul", 0xF6, MRM4m>, Imp<[AL],[AX]>; // AL,AH = AL*[mem8]
-def MULm16 : IM16<"mul", 0xF7, MRM4m>, Imp<[AX],[AX,DX]>, OpSize; // AX,DX = AX*[mem16]
-def MULm32 : IM32<"mul", 0xF7, MRM4m>, Imp<[EAX],[EAX,EDX]>; // EAX,EDX = EAX*[mem32]
+def MUL8r : I <"mul", 0xF6, MRM4r>, Imp<[AL],[AX]>; // AL,AH = AL*R8
+def MUL16r : I <"mul", 0xF7, MRM4r>, Imp<[AX],[AX,DX]>, OpSize; // AX,DX = AX*R16
+def MUL32r : I <"mul", 0xF7, MRM4r>, Imp<[EAX],[EAX,EDX]>; // EAX,EDX = EAX*R32
+def MUL8m : Im8 <"mul", 0xF6, MRM4m>, Imp<[AL],[AX]>; // AL,AH = AL*[mem8]
+def MUL16m : Im16<"mul", 0xF7, MRM4m>, Imp<[AX],[AX,DX]>, OpSize; // AX,DX = AX*[mem16]
+def MUL32m : Im32<"mul", 0xF7, MRM4m>, Imp<[EAX],[EAX,EDX]>; // EAX,EDX = EAX*[mem32]
// unsigned division/remainder
-def DIVr8 : I <"div", 0xF6, MRM6r>, Imp<[AX],[AX]>; // AX/r8 = AL,AH
-def DIVr16 : I <"div", 0xF7, MRM6r>, Imp<[AX,DX],[AX,DX]>, OpSize; // DX:AX/r16 = AX,DX
-def DIVr32 : I <"div", 0xF7, MRM6r>, Imp<[EAX,EDX],[EAX,EDX]>; // EDX:EAX/r32 = EAX,EDX
-def DIVm8 : IM8 <"div", 0xF6, MRM6m>, Imp<[AX],[AX]>; // AX/[mem8] = AL,AH
-def DIVm16 : IM16<"div", 0xF7, MRM6m>, Imp<[AX,DX],[AX,DX]>, OpSize; // DX:AX/[mem16] = AX,DX
-def DIVm32 : IM32<"div", 0xF7, MRM6m>, Imp<[EAX,EDX],[EAX,EDX]>; // EDX:EAX/[mem32] = EAX,EDX
+def DIV8r : I <"div", 0xF6, MRM6r>, Imp<[AX],[AX]>; // AX/r8 = AL,AH
+def DIV16r : I <"div", 0xF7, MRM6r>, Imp<[AX,DX],[AX,DX]>, OpSize; // DX:AX/r16 = AX,DX
+def DIV32r : I <"div", 0xF7, MRM6r>, Imp<[EAX,EDX],[EAX,EDX]>; // EDX:EAX/r32 = EAX,EDX
+def DIV8m : Im8 <"div", 0xF6, MRM6m>, Imp<[AX],[AX]>; // AX/[mem8] = AL,AH
+def DIV16m : Im16<"div", 0xF7, MRM6m>, Imp<[AX,DX],[AX,DX]>, OpSize; // DX:AX/[mem16] = AX,DX
+def DIV32m : Im32<"div", 0xF7, MRM6m>, Imp<[EAX,EDX],[EAX,EDX]>; // EDX:EAX/[mem32] = EAX,EDX
// signed division/remainder
-def IDIVr8 : I <"idiv",0xF6, MRM7r>, Imp<[AX],[AX]>; // AX/r8 = AL,AH
-def IDIVr16: I <"idiv",0xF7, MRM7r>, Imp<[AX,DX],[AX,DX]>, OpSize; // DX:AX/r16 = AX,DX
-def IDIVr32: I <"idiv",0xF7, MRM7r>, Imp<[EAX,EDX],[EAX,EDX]>; // EDX:EAX/r32 = EAX,EDX
-def IDIVm8 : IM8 <"idiv",0xF6, MRM7m>, Imp<[AX],[AX]>; // AX/[mem8] = AL,AH
-def IDIVm16: IM16<"idiv",0xF7, MRM7m>, Imp<[AX,DX],[AX,DX]>, OpSize; // DX:AX/[mem16] = AX,DX
-def IDIVm32: IM32<"idiv",0xF7, MRM7m>, Imp<[EAX,EDX],[EAX,EDX]>; // EDX:EAX/[mem32] = EAX,EDX
+def IDIV8r : I <"idiv",0xF6, MRM7r>, Imp<[AX],[AX]>; // AX/r8 = AL,AH
+def IDIV16r: I <"idiv",0xF7, MRM7r>, Imp<[AX,DX],[AX,DX]>, OpSize; // DX:AX/r16 = AX,DX
+def IDIV32r: I <"idiv",0xF7, MRM7r>, Imp<[EAX,EDX],[EAX,EDX]>; // EDX:EAX/r32 = EAX,EDX
+def IDIV8m : Im8 <"idiv",0xF6, MRM7m>, Imp<[AX],[AX]>; // AX/[mem8] = AL,AH
+def IDIV16m: Im16<"idiv",0xF7, MRM7m>, Imp<[AX,DX],[AX,DX]>, OpSize; // DX:AX/[mem16] = AX,DX
+def IDIV32m: Im32<"idiv",0xF7, MRM7m>, Imp<[EAX,EDX],[EAX,EDX]>; // EDX:EAX/[mem32] = EAX,EDX
// Sign-extenders for division
def CBW : I<"cbw", 0x98, RawFrm >, Imp<[AL],[AH]>; // AX = signext(AL)
//
let isTwoAddress = 1 in {
-// Conditional moves. These are modelled as X = cmovXX Y, Z. Eventually
-// register allocated to cmovXX XY, Z
-def CMOVErr16 : I<"cmove", 0x44, MRMSrcReg>, TB, OpSize; // if ==, R16 = R16
-def CMOVNErr32: I<"cmovne",0x45, MRMSrcReg>, TB; // if !=, R32 = R32
-def CMOVSrr32 : I<"cmovs", 0x48, MRMSrcReg>, TB; // if signed, R32 = R32
+// Conditional moves
+def CMOVB16rr : I <"cmovb", 0x42, MRMSrcReg>, TB, OpSize; // if <u, R16 = R16
+def CMOVB16rm : Im16<"cmovb", 0x42, MRMSrcMem>, TB, OpSize; // if <u, R16 = [mem16]
+def CMOVB32rr : I <"cmovb", 0x42, MRMSrcReg>, TB; // if <u, R32 = R32
+def CMOVB32rm : Im32<"cmovb", 0x42, MRMSrcMem>, TB; // if <u, R32 = [mem32]
+
+def CMOVAE16rr: I <"cmovae", 0x43, MRMSrcReg>, TB, OpSize; // if >=u, R16 = R16
+def CMOVAE16rm: Im16<"cmovae", 0x43, MRMSrcMem>, TB, OpSize; // if >=u, R16 = [mem16]
+def CMOVAE32rr: I <"cmovae", 0x43, MRMSrcReg>, TB; // if >=u, R32 = R32
+def CMOVAE32rm: Im32<"cmovae", 0x43, MRMSrcMem>, TB; // if >=u, R32 = [mem32]
+
+def CMOVE16rr : I <"cmove", 0x44, MRMSrcReg>, TB, OpSize; // if ==, R16 = R16
+def CMOVE16rm : Im16<"cmove", 0x44, MRMSrcMem>, TB, OpSize; // if ==, R16 = [mem16]
+def CMOVE32rr : I <"cmove", 0x44, MRMSrcReg>, TB; // if ==, R32 = R32
+def CMOVE32rm : Im32<"cmove", 0x44, MRMSrcMem>, TB; // if ==, R32 = [mem32]
+
+def CMOVNE16rr: I <"cmovne",0x45, MRMSrcReg>, TB, OpSize; // if !=, R16 = R16
+def CMOVNE16rm: Im16<"cmovne",0x45, MRMSrcMem>, TB, OpSize; // if !=, R16 = [mem16]
+def CMOVNE32rr: I <"cmovne",0x45, MRMSrcReg>, TB; // if !=, R32 = R32
+def CMOVNE32rm: Im32<"cmovne",0x45, MRMSrcMem>, TB; // if !=, R32 = [mem32]
+
+def CMOVBE16rr: I <"cmovbe",0x46, MRMSrcReg>, TB, OpSize; // if <=u, R16 = R16
+def CMOVBE16rm: Im16<"cmovbe",0x46, MRMSrcMem>, TB, OpSize; // if <=u, R16 = [mem16]
+def CMOVBE32rr: I <"cmovbe",0x46, MRMSrcReg>, TB; // if <=u, R32 = R32
+def CMOVBE32rm: Im32<"cmovbe",0x46, MRMSrcMem>, TB; // if <=u, R32 = [mem32]
+
+def CMOVA16rr : I <"cmova", 0x47, MRMSrcReg>, TB, OpSize; // if >u, R16 = R16
+def CMOVA16rm : Im16<"cmova", 0x47, MRMSrcMem>, TB, OpSize; // if >u, R16 = [mem16]
+def CMOVA32rr : I <"cmova", 0x47, MRMSrcReg>, TB; // if >u, R32 = R32
+def CMOVA32rm : Im32<"cmova", 0x47, MRMSrcMem>, TB; // if >u, R32 = [mem32]
+
+def CMOVS16rr : I <"cmovs", 0x48, MRMSrcReg>, TB, OpSize; // if signed, R16 = R16
+def CMOVS16rm : Im16<"cmovs", 0x48, MRMSrcMem>, TB, OpSize; // if signed, R16 = [mem16]
+def CMOVS32rr : I <"cmovs", 0x48, MRMSrcReg>, TB; // if signed, R32 = R32
+def CMOVS32rm : Im32<"cmovs", 0x48, MRMSrcMem>, TB; // if signed, R32 = [mem32]
+
+def CMOVNS16rr: I <"cmovns",0x49, MRMSrcReg>, TB, OpSize; // if !signed, R16 = R16
+def CMOVNS16rm: Im16<"cmovns",0x49, MRMSrcMem>, TB, OpSize; // if !signed, R16 = [mem16]
+def CMOVNS32rr: I <"cmovns",0x49, MRMSrcReg>, TB; // if !signed, R32 = R32
+def CMOVNS32rm: Im32<"cmovns",0x49, MRMSrcMem>, TB; // if !signed, R32 = [mem32]
+
+def CMOVL16rr : I <"cmovl", 0x4C, MRMSrcReg>, TB, OpSize; // if <s, R16 = R16
+def CMOVL16rm : Im16<"cmovl", 0x4C, MRMSrcMem>, TB, OpSize; // if <s, R16 = [mem16]
+def CMOVL32rr : I <"cmovl", 0x4C, MRMSrcReg>, TB; // if <s, R32 = R32
+def CMOVL32rm : Im32<"cmovl", 0x4C, MRMSrcMem>, TB; // if <s, R32 = [mem32]
+
+def CMOVGE16rr: I <"cmovge",0x4D, MRMSrcReg>, TB, OpSize; // if >=s, R16 = R16
+def CMOVGE16rm: Im16<"cmovge",0x4D, MRMSrcMem>, TB, OpSize; // if >=s, R16 = [mem16]
+def CMOVGE32rr: I <"cmovge",0x4D, MRMSrcReg>, TB; // if >=s, R32 = R32
+def CMOVGE32rm: Im32<"cmovge",0x4D, MRMSrcMem>, TB; // if >=s, R32 = [mem32]
+
+def CMOVLE16rr: I <"cmovle",0x4E, MRMSrcReg>, TB, OpSize; // if <=s, R16 = R16
+def CMOVLE16rm: Im16<"cmovle",0x4E, MRMSrcMem>, TB, OpSize; // if <=s, R16 = [mem16]
+def CMOVLE32rr: I <"cmovle",0x4E, MRMSrcReg>, TB; // if <=s, R32 = R32
+def CMOVLE32rm: Im32<"cmovle",0x4E, MRMSrcMem>, TB; // if <=s, R32 = [mem32]
+
+def CMOVG16rr : I <"cmovg", 0x4F, MRMSrcReg>, TB, OpSize; // if >s, R16 = R16
+def CMOVG16rm : Im16<"cmovg", 0x4F, MRMSrcMem>, TB, OpSize; // if >s, R16 = [mem16]
+def CMOVG32rr : I <"cmovg", 0x4F, MRMSrcReg>, TB; // if >s, R32 = R32
+def CMOVG32rm : Im32<"cmovg", 0x4F, MRMSrcMem>, TB; // if >s, R32 = [mem32]
// unary instructions
-def NEGr8 : I <"neg", 0xF6, MRM3r>; // R8 = -R8 = 0-R8
-def NEGr16 : I <"neg", 0xF7, MRM3r>, OpSize; // R16 = -R16 = 0-R16
-def NEGr32 : I <"neg", 0xF7, MRM3r>; // R32 = -R32 = 0-R32
-def NEGm8 : IM8 <"neg", 0xF6, MRM3m>; // [mem8] = -[mem8] = 0-[mem8]
-def NEGm16 : IM16<"neg", 0xF7, MRM3m>, OpSize; // [mem16] = -[mem16] = 0-[mem16]
-def NEGm32 : IM32<"neg", 0xF7, MRM3m>; // [mem32] = -[mem32] = 0-[mem32]
-
-def NOTr8 : I <"not", 0xF6, MRM2r>; // R8 = ~R8 = R8^-1
-def NOTr16 : I <"not", 0xF7, MRM2r>, OpSize; // R16 = ~R16 = R16^-1
-def NOTr32 : I <"not", 0xF7, MRM2r>; // R32 = ~R32 = R32^-1
-def NOTm8 : IM8 <"not", 0xF6, MRM2m>; // [mem8] = ~[mem8] = [mem8^-1]
-def NOTm16 : IM16<"not", 0xF7, MRM2m>, OpSize; // [mem16] = ~[mem16] = [mem16^-1]
-def NOTm32 : IM32<"not", 0xF7, MRM2m>; // [mem32] = ~[mem32] = [mem32^-1]
-
-def INCr8 : I <"inc", 0xFE, MRM0r>; // ++R8
-def INCr16 : I <"inc", 0xFF, MRM0r>, OpSize; // ++R16
-def INCr32 : I <"inc", 0xFF, MRM0r>; // ++R32
-def INCm8 : IM8 <"inc", 0xFE, MRM0m>; // ++R8
-def INCm16 : IM16<"inc", 0xFF, MRM0m>, OpSize; // ++R16
-def INCm32 : IM32<"inc", 0xFF, MRM0m>; // ++R32
-
-def DECr8 : I <"dec", 0xFE, MRM1r>; // --R8
-def DECr16 : I <"dec", 0xFF, MRM1r>, OpSize; // --R16
-def DECr32 : I <"dec", 0xFF, MRM1r>; // --R32
-def DECm8 : IM8 <"dec", 0xFE, MRM1m>; // --[mem8]
-def DECm16 : IM16<"dec", 0xFF, MRM1m>, OpSize; // --[mem16]
-def DECm32 : IM32<"dec", 0xFF, MRM1m>; // --[mem32]
+def NEG8r : I <"neg", 0xF6, MRM3r>; // R8 = -R8 = 0-R8
+def NEG16r : I <"neg", 0xF7, MRM3r>, OpSize; // R16 = -R16 = 0-R16
+def NEG32r : I <"neg", 0xF7, MRM3r>; // R32 = -R32 = 0-R32
+def NEG8m : Im8 <"neg", 0xF6, MRM3m>; // [mem8] = -[mem8] = 0-[mem8]
+def NEG16m : Im16<"neg", 0xF7, MRM3m>, OpSize; // [mem16] = -[mem16] = 0-[mem16]
+def NEG32m : Im32<"neg", 0xF7, MRM3m>; // [mem32] = -[mem32] = 0-[mem32]
+
+def NOT8r : I <"not", 0xF6, MRM2r>; // R8 = ~R8 = R8^-1
+def NOT16r : I <"not", 0xF7, MRM2r>, OpSize; // R16 = ~R16 = R16^-1
+def NOT32r : I <"not", 0xF7, MRM2r>; // R32 = ~R32 = R32^-1
+def NOT8m : Im8 <"not", 0xF6, MRM2m>; // [mem8] = ~[mem8] = [mem8^-1]
+def NOT16m : Im16<"not", 0xF7, MRM2m>, OpSize; // [mem16] = ~[mem16] = [mem16^-1]
+def NOT32m : Im32<"not", 0xF7, MRM2m>; // [mem32] = ~[mem32] = [mem32^-1]
+
+def INC8r : I <"inc", 0xFE, MRM0r>; // ++R8
+def INC16r : I <"inc", 0xFF, MRM0r>, OpSize; // ++R16
+def INC32r : I <"inc", 0xFF, MRM0r>; // ++R32
+def INC8m : Im8 <"inc", 0xFE, MRM0m>; // ++R8
+def INC16m : Im16<"inc", 0xFF, MRM0m>, OpSize; // ++R16
+def INC32m : Im32<"inc", 0xFF, MRM0m>; // ++R32
+
+def DEC8r : I <"dec", 0xFE, MRM1r>; // --R8
+def DEC16r : I <"dec", 0xFF, MRM1r>, OpSize; // --R16
+def DEC32r : I <"dec", 0xFF, MRM1r>; // --R32
+def DEC8m : Im8 <"dec", 0xFE, MRM1m>; // --[mem8]
+def DEC16m : Im16<"dec", 0xFF, MRM1m>, OpSize; // --[mem16]
+def DEC32m : Im32<"dec", 0xFF, MRM1m>; // --[mem32]
// Logical operators...
-def ANDrr8 : I <"and", 0x20, MRMDestReg>, Pattern<(set R8 , (and R8 , R8 ))>;
-def ANDrr16 : I <"and", 0x21, MRMDestReg>, OpSize, Pattern<(set R16, (and R16, R16))>;
-def ANDrr32 : I <"and", 0x21, MRMDestReg>, Pattern<(set R32, (and R32, R32))>;
-def ANDmr8 : IM8 <"and", 0x20, MRMDestMem>; // [mem8] &= R8
-def ANDmr16 : IM16 <"and", 0x21, MRMDestMem>, OpSize; // [mem16] &= R16
-def ANDmr32 : IM32 <"and", 0x21, MRMDestMem>; // [mem32] &= R32
-def ANDrm8 : IM8 <"and", 0x22, MRMSrcMem >; // R8 &= [mem8]
-def ANDrm16 : IM16 <"and", 0x23, MRMSrcMem >, OpSize; // R16 &= [mem16]
-def ANDrm32 : IM32 <"and", 0x23, MRMSrcMem >; // R32 &= [mem32]
-
-def ANDri8 : II8 <"and", 0x80, MRM4r >, Pattern<(set R8 , (and R8 , imm))>;
-def ANDri16 : II16 <"and", 0x81, MRM4r >, OpSize, Pattern<(set R16, (and R16, imm))>;
-def ANDri32 : II32 <"and", 0x81, MRM4r >, Pattern<(set R32, (and R32, imm))>;
-def ANDmi8 : I8MI <"and", 0x80, MRM4m >; // [mem8] &= imm8
-def ANDmi16 : I16MI <"and", 0x81, MRM4m >, OpSize; // [mem16] &= imm16
-def ANDmi32 : I32MI <"and", 0x81, MRM4m >; // [mem32] &= imm32
-
-def ANDri16b : II8 <"and", 0x83, MRM4r >, OpSize; // R16 &= imm8
-def ANDri32b : II8 <"and", 0x83, MRM4r >; // R32 &= imm8
-def ANDmi16b : IM16I8<"and", 0x83, MRM4m >, OpSize; // [mem16] &= imm8
-def ANDmi32b : IM32I8<"and", 0x83, MRM4m >; // [mem32] &= imm8
-
-
-def ORrr8 : I <"or" , 0x08, MRMDestReg>, Pattern<(set R8 , (or R8 , R8 ))>;
-def ORrr16 : I <"or" , 0x09, MRMDestReg>, OpSize, Pattern<(set R16, (or R16, R16))>;
-def ORrr32 : I <"or" , 0x09, MRMDestReg>, Pattern<(set R32, (or R32, R32))>;
-def ORmr8 : IM8 <"or" , 0x08, MRMDestMem>; // [mem8] |= R8
-def ORmr16 : IM16 <"or" , 0x09, MRMDestMem>, OpSize; // [mem16] |= R16
-def ORmr32 : IM32 <"or" , 0x09, MRMDestMem>; // [mem32] |= R32
-def ORrm8 : IM8 <"or" , 0x0A, MRMSrcMem >; // R8 |= [mem8]
-def ORrm16 : IM16 <"or" , 0x0B, MRMSrcMem >, OpSize; // R16 |= [mem16]
-def ORrm32 : IM32 <"or" , 0x0B, MRMSrcMem >; // R32 |= [mem32]
-
-def ORri8 : II8 <"or" , 0x80, MRM1r >, Pattern<(set R8 , (or R8 , imm))>;
-def ORri16 : II16 <"or" , 0x81, MRM1r >, OpSize, Pattern<(set R16, (or R16, imm))>;
-def ORri32 : II32 <"or" , 0x81, MRM1r >, Pattern<(set R32, (or R32, imm))>;
-def ORmi8 : I8MI <"or" , 0x80, MRM1m >; // [mem8] |= imm8
-def ORmi16 : I16MI <"or" , 0x81, MRM1m >, OpSize; // [mem16] |= imm16
-def ORmi32 : I32MI <"or" , 0x81, MRM1m >; // [mem32] |= imm32
-
-def ORri16b : II8 <"or" , 0x83, MRM1r >, OpSize; // R16 |= imm8
-def ORri32b : II8 <"or" , 0x83, MRM1r >; // R32 |= imm8
-def ORmi16b : IM16I8<"or" , 0x83, MRM1m >, OpSize; // [mem16] |= imm8
-def ORmi32b : IM32I8<"or" , 0x83, MRM1m >; // [mem32] |= imm8
-
-
-def XORrr8 : I <"xor", 0x30, MRMDestReg>, Pattern<(set R8 , (xor R8 , R8 ))>;
-def XORrr16 : I <"xor", 0x31, MRMDestReg>, OpSize, Pattern<(set R16, (xor R16, R16))>;
-def XORrr32 : I <"xor", 0x31, MRMDestReg>, Pattern<(set R32, (xor R32, R32))>;
-def XORmr8 : IM8 <"xor", 0x30, MRMDestMem>; // [mem8] ^= R8
-def XORmr16 : IM16 <"xor", 0x31, MRMDestMem>, OpSize; // [mem16] ^= R16
-def XORmr32 : IM32 <"xor", 0x31, MRMDestMem>; // [mem32] ^= R32
-def XORrm8 : IM8 <"xor", 0x32, MRMSrcMem >; // R8 ^= [mem8]
-def XORrm16 : IM16 <"xor", 0x33, MRMSrcMem >, OpSize; // R16 ^= [mem16]
-def XORrm32 : IM32 <"xor", 0x33, MRMSrcMem >; // R32 ^= [mem32]
-
-def XORri8 : II8 <"xor", 0x80, MRM6r >, Pattern<(set R8 , (xor R8 , imm))>;
-def XORri16 : II16 <"xor", 0x81, MRM6r >, OpSize, Pattern<(set R16, (xor R16, imm))>;
-def XORri32 : II32 <"xor", 0x81, MRM6r >, Pattern<(set R32, (xor R32, imm))>;
-def XORmi8 : I8MI <"xor", 0x80, MRM6m >; // [mem8] ^= R8
-def XORmi16 : I16MI <"xor", 0x81, MRM6m >, OpSize; // [mem16] ^= R16
-def XORmi32 : I32MI <"xor", 0x81, MRM6m >; // [mem32] ^= R32
-
-def XORri16b : II8 <"xor", 0x83, MRM6r >, OpSize; // R16 ^= imm8
-def XORri32b : II8 <"xor", 0x83, MRM6r >; // R32 ^= imm8
-def XORmi16b : IM16I8<"xor", 0x83, MRM6m >, OpSize; // [mem16] ^= imm8
-def XORmi32b : IM32I8<"xor", 0x83, MRM6m >; // [mem32] ^= imm8
+def AND8rr : I <"and", 0x20, MRMDestReg>, Pattern<(set R8 , (and R8 , R8 ))>;
+def AND16rr : I <"and", 0x21, MRMDestReg>, OpSize, Pattern<(set R16, (and R16, R16))>;
+def AND32rr : I <"and", 0x21, MRMDestReg>, Pattern<(set R32, (and R32, R32))>;
+def AND8mr : Im8 <"and", 0x20, MRMDestMem>; // [mem8] &= R8
+def AND16mr : Im16 <"and", 0x21, MRMDestMem>, OpSize; // [mem16] &= R16
+def AND32mr : Im32 <"and", 0x21, MRMDestMem>; // [mem32] &= R32
+def AND8rm : Im8 <"and", 0x22, MRMSrcMem >; // R8 &= [mem8]
+def AND16rm : Im16 <"and", 0x23, MRMSrcMem >, OpSize; // R16 &= [mem16]
+def AND32rm : Im32 <"and", 0x23, MRMSrcMem >; // R32 &= [mem32]
+
+def AND8ri : Ii8 <"and", 0x80, MRM4r >, Pattern<(set R8 , (and R8 , imm))>;
+def AND16ri : Ii16 <"and", 0x81, MRM4r >, OpSize, Pattern<(set R16, (and R16, imm))>;
+def AND32ri : Ii32 <"and", 0x81, MRM4r >, Pattern<(set R32, (and R32, imm))>;
+def AND8mi : Im8i8 <"and", 0x80, MRM4m >; // [mem8] &= imm8
+def AND16mi : Im16i16<"and", 0x81, MRM4m >, OpSize; // [mem16] &= imm16
+def AND32mi : Im32i32<"and", 0x81, MRM4m >; // [mem32] &= imm32
+
+def AND16ri8 : Ii8 <"and", 0x83, MRM4r >, OpSize; // R16 &= imm8
+def AND32ri8 : Ii8 <"and", 0x83, MRM4r >; // R32 &= imm8
+def AND16mi8 : Im16i8<"and", 0x83, MRM4m >, OpSize; // [mem16] &= imm8
+def AND32mi8 : Im32i8<"and", 0x83, MRM4m >; // [mem32] &= imm8
+
+
+def OR8rr : I <"or" , 0x08, MRMDestReg>, Pattern<(set R8 , (or R8 , R8 ))>;
+def OR16rr : I <"or" , 0x09, MRMDestReg>, OpSize, Pattern<(set R16, (or R16, R16))>;
+def OR32rr : I <"or" , 0x09, MRMDestReg>, Pattern<(set R32, (or R32, R32))>;
+def OR8mr : Im8 <"or" , 0x08, MRMDestMem>; // [mem8] |= R8
+def OR16mr : Im16 <"or" , 0x09, MRMDestMem>, OpSize; // [mem16] |= R16
+def OR32mr : Im32 <"or" , 0x09, MRMDestMem>; // [mem32] |= R32
+def OR8rm : Im8 <"or" , 0x0A, MRMSrcMem >; // R8 |= [mem8]
+def OR16rm : Im16 <"or" , 0x0B, MRMSrcMem >, OpSize; // R16 |= [mem16]
+def OR32rm : Im32 <"or" , 0x0B, MRMSrcMem >; // R32 |= [mem32]
+
+def OR8ri : Ii8 <"or" , 0x80, MRM1r >, Pattern<(set R8 , (or R8 , imm))>;
+def OR16ri : Ii16 <"or" , 0x81, MRM1r >, OpSize, Pattern<(set R16, (or R16, imm))>;
+def OR32ri : Ii32 <"or" , 0x81, MRM1r >, Pattern<(set R32, (or R32, imm))>;
+def OR8mi : Im8i8 <"or" , 0x80, MRM1m >; // [mem8] |= imm8
+def OR16mi : Im16i16<"or" , 0x81, MRM1m >, OpSize; // [mem16] |= imm16
+def OR32mi : Im32i32<"or" , 0x81, MRM1m >; // [mem32] |= imm32
+
+def OR16ri8 : Ii8 <"or" , 0x83, MRM1r >, OpSize; // R16 |= imm8
+def OR32ri8 : Ii8 <"or" , 0x83, MRM1r >; // R32 |= imm8
+def OR16mi8 : Im16i8<"or" , 0x83, MRM1m >, OpSize; // [mem16] |= imm8
+def OR32mi8 : Im32i8<"or" , 0x83, MRM1m >; // [mem32] |= imm8
+
+
+def XOR8rr : I <"xor", 0x30, MRMDestReg>, Pattern<(set R8 , (xor R8 , R8 ))>;
+def XOR16rr : I <"xor", 0x31, MRMDestReg>, OpSize, Pattern<(set R16, (xor R16, R16))>;
+def XOR32rr : I <"xor", 0x31, MRMDestReg>, Pattern<(set R32, (xor R32, R32))>;
+def XOR8mr : Im8 <"xor", 0x30, MRMDestMem>; // [mem8] ^= R8
+def XOR16mr : Im16 <"xor", 0x31, MRMDestMem>, OpSize; // [mem16] ^= R16
+def XOR32mr : Im32 <"xor", 0x31, MRMDestMem>; // [mem32] ^= R32
+def XOR8rm : Im8 <"xor", 0x32, MRMSrcMem >; // R8 ^= [mem8]
+def XOR16rm : Im16 <"xor", 0x33, MRMSrcMem >, OpSize; // R16 ^= [mem16]
+def XOR32rm : Im32 <"xor", 0x33, MRMSrcMem >; // R32 ^= [mem32]
+
+def XOR8ri : Ii8 <"xor", 0x80, MRM6r >, Pattern<(set R8 , (xor R8 , imm))>;
+def XOR16ri : Ii16 <"xor", 0x81, MRM6r >, OpSize, Pattern<(set R16, (xor R16, imm))>;
+def XOR32ri : Ii32 <"xor", 0x81, MRM6r >, Pattern<(set R32, (xor R32, imm))>;
+def XOR8mi : Im8i8 <"xor", 0x80, MRM6m >; // [mem8] ^= R8
+def XOR16mi : Im16i16<"xor", 0x81, MRM6m >, OpSize; // [mem16] ^= R16
+def XOR32mi : Im32i32<"xor", 0x81, MRM6m >; // [mem32] ^= R32
+
+def XOR16ri8 : Ii8 <"xor", 0x83, MRM6r >, OpSize; // R16 ^= imm8
+def XOR32ri8 : Ii8 <"xor", 0x83, MRM6r >; // R32 ^= imm8
+def XOR16mi8 : Im16i8<"xor", 0x83, MRM6m >, OpSize; // [mem16] ^= imm8
+def XOR32mi8 : Im32i8<"xor", 0x83, MRM6m >; // [mem32] ^= imm8
// Shift instructions
-def SHLrCL8 : I <"shl", 0xD2, MRM4r > , UsesCL; // R8 <<= cl
-def SHLrCL16 : I <"shl", 0xD3, MRM4r >, OpSize, UsesCL; // R16 <<= cl
-def SHLrCL32 : I <"shl", 0xD3, MRM4r > , UsesCL; // R32 <<= cl
-def SHLmCL8 : IM8 <"shl", 0xD2, MRM4m > , UsesCL; // [mem8] <<= cl
-def SHLmCL16 : IM16 <"shl", 0xD3, MRM4m >, OpSize, UsesCL; // [mem16] <<= cl
-def SHLmCL32 : IM32 <"shl", 0xD3, MRM4m > , UsesCL; // [mem32] <<= cl
-
-def SHLri8 : II8 <"shl", 0xC0, MRM4r >; // R8 <<= imm8
-def SHLri16 : II8 <"shl", 0xC1, MRM4r >, OpSize; // R16 <<= imm8
-def SHLri32 : II8 <"shl", 0xC1, MRM4r >; // R32 <<= imm8
-def SHLmi8 : I8MI <"shl", 0xC0, MRM4m >; // [mem8] <<= imm8
-def SHLmi16 : IM16I8<"shl", 0xC1, MRM4m >, OpSize; // [mem16] <<= imm8
-def SHLmi32 : IM32I8<"shl", 0xC1, MRM4m >; // [mem32] <<= imm8
-
-def SHRrCL8 : I <"shr", 0xD2, MRM5r > , UsesCL; // R8 >>= cl
-def SHRrCL16 : I <"shr", 0xD3, MRM5r >, OpSize, UsesCL; // R16 >>= cl
-def SHRrCL32 : I <"shr", 0xD3, MRM5r > , UsesCL; // R32 >>= cl
-def SHRmCL8 : IM8 <"shr", 0xD2, MRM5m > , UsesCL; // [mem8] >>= cl
-def SHRmCL16 : IM16 <"shr", 0xD3, MRM5m >, OpSize, UsesCL; // [mem16] >>= cl
-def SHRmCL32 : IM32 <"shr", 0xD3, MRM5m > , UsesCL; // [mem32] >>= cl
-
-def SHRri8 : II8 <"shr", 0xC0, MRM5r >; // R8 >>= imm8
-def SHRri16 : II8 <"shr", 0xC1, MRM5r >, OpSize; // R16 >>= imm8
-def SHRri32 : II8 <"shr", 0xC1, MRM5r >; // R32 >>= imm8
-def SHRmi8 : I8MI <"shr", 0xC0, MRM5m >; // [mem8] >>= imm8
-def SHRmi16 : IM16I8<"shr", 0xC1, MRM5m >, OpSize; // [mem16] >>= imm8
-def SHRmi32 : IM32I8<"shr", 0xC1, MRM5m >; // [mem32] >>= imm8
-
-def SARrCL8 : I <"sar", 0xD2, MRM7r > , UsesCL; // R8 >>>= cl
-def SARrCL16 : I <"sar", 0xD3, MRM7r >, OpSize, UsesCL; // R16 >>>= cl
-def SARrCL32 : I <"sar", 0xD3, MRM7r > , UsesCL; // R32 >>>= cl
-def SARmCL8 : IM8 <"sar", 0xD2, MRM7m > , UsesCL; // [mem8] >>>= cl
-def SARmCL16 : IM16 <"sar", 0xD3, MRM7m >, OpSize, UsesCL; // [mem16] >>>= cl
-def SARmCL32 : IM32 <"sar", 0xD3, MRM7m > , UsesCL; // [mem32] >>>= cl
-
-def SARri8 : II8 <"sar", 0xC0, MRM7r >; // R8 >>>= imm8
-def SARri16 : II8 <"sar", 0xC1, MRM7r >, OpSize; // R16 >>>= imm8
-def SARri32 : II8 <"sar", 0xC1, MRM7r >; // R32 >>>= imm8
-def SARmi8 : I8MI <"sar", 0xC0, MRM7m >; // [mem8] >>>= imm8
-def SARmi16 : IM16I8<"sar", 0xC1, MRM7m >, OpSize; // [mem16] >>>= imm8
-def SARmi32 : IM32I8<"sar", 0xC1, MRM7m >; // [mem32] >>>= imm8
-
-def SHLDrrCL32 : I <"shld", 0xA5, MRMDestReg>, TB, UsesCL; // R32 <<= R32,R32 cl
-def SHLDmrCL32 : I <"shld", 0xA5, MRMDestMem>, TB, UsesCL; // [mem32] <<= [mem32],R32 cl
-def SHLDrri32 : II8 <"shld", 0xA4, MRMDestReg>, TB; // R32 <<= R32,R32 imm8
-def SHLDmri32 : II8 <"shld", 0xA4, MRMDestMem>, TB; // [mem32] <<= [mem32],R32 imm8
-
-def SHRDrrCL32 : I <"shrd", 0xAD, MRMDestReg>, TB, UsesCL; // R32 >>= R32,R32 cl
-def SHRDmrCL32 : I <"shrd", 0xAD, MRMDestMem>, TB, UsesCL; // [mem32] >>= [mem32],R32 cl
-def SHRDrri32 : II8 <"shrd", 0xAC, MRMDestReg>, TB; // R32 >>= R32,R32 imm8
-def SHRDmri32 : II8 <"shrd", 0xAC, MRMDestMem>, TB; // [mem32] >>= [mem32],R32 imm8
+// FIXME: provide shorter instructions when imm8 == 1
+def SHL8rCL : I <"shl", 0xD2, MRM4r > , UsesCL; // R8 <<= cl
+def SHL16rCL : I <"shl", 0xD3, MRM4r >, OpSize, UsesCL; // R16 <<= cl
+def SHL32rCL : I <"shl", 0xD3, MRM4r > , UsesCL; // R32 <<= cl
+def SHL8mCL : Im8 <"shl", 0xD2, MRM4m > , UsesCL; // [mem8] <<= cl
+def SHL16mCL : Im16 <"shl", 0xD3, MRM4m >, OpSize, UsesCL; // [mem16] <<= cl
+def SHL32mCL : Im32 <"shl", 0xD3, MRM4m > , UsesCL; // [mem32] <<= cl
+
+def SHL8ri : Ii8 <"shl", 0xC0, MRM4r >; // R8 <<= imm8
+def SHL16ri : Ii8 <"shl", 0xC1, MRM4r >, OpSize; // R16 <<= imm8
+def SHL32ri : Ii8 <"shl", 0xC1, MRM4r >; // R32 <<= imm8
+def SHL8mi : Im8i8 <"shl", 0xC0, MRM4m >; // [mem8] <<= imm8
+def SHL16mi : Im16i8<"shl", 0xC1, MRM4m >, OpSize; // [mem16] <<= imm8
+def SHL32mi : Im32i8<"shl", 0xC1, MRM4m >; // [mem32] <<= imm8
+
+def SHR8rCL : I <"shr", 0xD2, MRM5r > , UsesCL; // R8 >>= cl
+def SHR16rCL : I <"shr", 0xD3, MRM5r >, OpSize, UsesCL; // R16 >>= cl
+def SHR32rCL : I <"shr", 0xD3, MRM5r > , UsesCL; // R32 >>= cl
+def SHR8mCL : Im8 <"shr", 0xD2, MRM5m > , UsesCL; // [mem8] >>= cl
+def SHR16mCL : Im16 <"shr", 0xD3, MRM5m >, OpSize, UsesCL; // [mem16] >>= cl
+def SHR32mCL : Im32 <"shr", 0xD3, MRM5m > , UsesCL; // [mem32] >>= cl
+
+def SHR8ri : Ii8 <"shr", 0xC0, MRM5r >; // R8 >>= imm8
+def SHR16ri : Ii8 <"shr", 0xC1, MRM5r >, OpSize; // R16 >>= imm8
+def SHR32ri : Ii8 <"shr", 0xC1, MRM5r >; // R32 >>= imm8
+def SHR8mi : Im8i8 <"shr", 0xC0, MRM5m >; // [mem8] >>= imm8
+def SHR16mi : Im16i8<"shr", 0xC1, MRM5m >, OpSize; // [mem16] >>= imm8
+def SHR32mi : Im32i8<"shr", 0xC1, MRM5m >; // [mem32] >>= imm8
+
+def SAR8rCL : I <"sar", 0xD2, MRM7r > , UsesCL; // R8 >>>= cl
+def SAR16rCL : I <"sar", 0xD3, MRM7r >, OpSize, UsesCL; // R16 >>>= cl
+def SAR32rCL : I <"sar", 0xD3, MRM7r > , UsesCL; // R32 >>>= cl
+def SAR8mCL : Im8 <"sar", 0xD2, MRM7m > , UsesCL; // [mem8] >>>= cl
+def SAR16mCL : Im16 <"sar", 0xD3, MRM7m >, OpSize, UsesCL; // [mem16] >>>= cl
+def SAR32mCL : Im32 <"sar", 0xD3, MRM7m > , UsesCL; // [mem32] >>>= cl
+
+def SAR8ri : Ii8 <"sar", 0xC0, MRM7r >; // R8 >>>= imm8
+def SAR16ri : Ii8 <"sar", 0xC1, MRM7r >, OpSize; // R16 >>>= imm8
+def SAR32ri : Ii8 <"sar", 0xC1, MRM7r >; // R32 >>>= imm8
+def SAR8mi : Im8i8 <"sar", 0xC0, MRM7m >; // [mem8] >>>= imm8
+def SAR16mi : Im16i8<"sar", 0xC1, MRM7m >, OpSize; // [mem16] >>>= imm8
+def SAR32mi : Im32i8<"sar", 0xC1, MRM7m >; // [mem32] >>>= imm8
+
+def SHLD32rrCL : I <"shld", 0xA5, MRMDestReg>, TB, UsesCL; // R32 <<= R32,R32 cl
+def SHLD32mrCL : Im32 <"shld", 0xA5, MRMDestMem>, TB, UsesCL; // [mem32] <<= [mem32],R32 cl
+def SHLD32rri8 : Ii8 <"shld", 0xA4, MRMDestReg>, TB; // R32 <<= R32,R32 imm8
+def SHLD32mri8 : Im32i8<"shld", 0xA4, MRMDestMem>, TB; // [mem32] <<= [mem32],R32 imm8
+
+def SHRD32rrCL : I <"shrd", 0xAD, MRMDestReg>, TB, UsesCL; // R32 >>= R32,R32 cl
+def SHRD32mrCL : Im32 <"shrd", 0xAD, MRMDestMem>, TB, UsesCL; // [mem32] >>= [mem32],R32 cl
+def SHRD32rri8 : Ii8 <"shrd", 0xAC, MRMDestReg>, TB; // R32 >>= R32,R32 imm8
+def SHRD32mri8 : Im32i8<"shrd", 0xAC, MRMDestMem>, TB; // [mem32] >>= [mem32],R32 imm8
// Arithmetic...
-def ADDrr8 : I <"add", 0x00, MRMDestReg>, Pattern<(set R8 , (plus R8 , R8 ))>;
-def ADDrr16 : I <"add", 0x01, MRMDestReg>, OpSize, Pattern<(set R16, (plus R16, R16))>;
-def ADDrr32 : I <"add", 0x01, MRMDestReg>, Pattern<(set R32, (plus R32, R32))>;
-def ADDmr8 : IM8 <"add", 0x00, MRMDestMem>; // [mem8] += R8
-def ADDmr16 : IM16 <"add", 0x01, MRMDestMem>, OpSize; // [mem16] += R16
-def ADDmr32 : IM32 <"add", 0x01, MRMDestMem>; // [mem32] += R32
-def ADDrm8 : IM8 <"add", 0x02, MRMSrcMem >; // R8 += [mem8]
-def ADDrm16 : IM16 <"add", 0x03, MRMSrcMem >, OpSize; // R16 += [mem16]
-def ADDrm32 : IM32 <"add", 0x03, MRMSrcMem >; // R32 += [mem32]
-
-def ADDri8 : II8 <"add", 0x80, MRM0r >, Pattern<(set R8 , (plus R8 , imm))>;
-def ADDri16 : II16 <"add", 0x81, MRM0r >, OpSize, Pattern<(set R16, (plus R16, imm))>;
-def ADDri32 : II32 <"add", 0x81, MRM0r >, Pattern<(set R32, (plus R32, imm))>;
-def ADDmi8 : I8MI <"add", 0x80, MRM0m >; // [mem8] += I8
-def ADDmi16 : I16MI <"add", 0x81, MRM0m >, OpSize; // [mem16] += I16
-def ADDmi32 : I32MI <"add", 0x81, MRM0m >; // [mem32] += I32
-
-def ADDri16b : II8 <"add", 0x83, MRM0r >, OpSize; // ADDri with sign extended 8 bit imm
-def ADDri32b : II8 <"add", 0x83, MRM0r >;
-def ADDmi16b : IM16I8<"add", 0x83, MRM0m >, OpSize; // [mem16] += I8
-def ADDmi32b : IM32I8<"add", 0x83, MRM0m >; // [mem32] += I8
-
-def ADCrr32 : I <"adc", 0x11, MRMDestReg>; // R32 += R32+Carry
-def ADCrm32 : I <"adc", 0x11, MRMSrcMem >; // R32 += [mem32]+Carry
-def ADCmr32 : I <"adc", 0x13, MRMDestMem>; // [mem32] += R32+Carry
-
-
-def SUBrr8 : I <"sub", 0x28, MRMDestReg>, Pattern<(set R8 , (minus R8 , R8 ))>;
-def SUBrr16 : I <"sub", 0x29, MRMDestReg>, OpSize, Pattern<(set R16, (minus R16, R16))>;
-def SUBrr32 : I <"sub", 0x29, MRMDestReg>, Pattern<(set R32, (minus R32, R32))>;
-def SUBmr8 : IM8 <"sub", 0x28, MRMDestMem>; // [mem8] -= R8
-def SUBmr16 : IM16 <"sub", 0x29, MRMDestMem>, OpSize; // [mem16] -= R16
-def SUBmr32 : IM32 <"sub", 0x29, MRMDestMem>; // [mem32] -= R32
-def SUBrm8 : IM8 <"sub", 0x2A, MRMSrcMem >; // R8 -= [mem8]
-def SUBrm16 : IM16 <"sub", 0x2B, MRMSrcMem >, OpSize; // R16 -= [mem16]
-def SUBrm32 : IM32 <"sub", 0x2B, MRMSrcMem >; // R32 -= [mem32]
-
-def SUBri8 : II8 <"sub", 0x80, MRM5r >, Pattern<(set R8 , (minus R8 , imm))>;
-def SUBri16 : II16 <"sub", 0x81, MRM5r >, OpSize, Pattern<(set R16, (minus R16, imm))>;
-def SUBri32 : II32 <"sub", 0x81, MRM5r >, Pattern<(set R32, (minus R32, imm))>;
-def SUBmi8 : I8MI <"sub", 0x80, MRM5m >; // [mem8] -= I8
-def SUBmi16 : I16MI <"sub", 0x81, MRM5m >, OpSize; // [mem16] -= I16
-def SUBmi32 : I32MI <"sub", 0x81, MRM5m >; // [mem32] -= I32
-
-def SUBri16b : II8 <"sub", 0x83, MRM5r >, OpSize;
-def SUBri32b : II8 <"sub", 0x83, MRM5r >;
-def SUBmi16b : IM16I8<"sub", 0x83, MRM5m >, OpSize; // [mem16] -= I8
-def SUBmi32b : IM32I8<"sub", 0x83, MRM5m >; // [mem32] -= I8
-
-def SBBrr32 : I <"sbb", 0x19, MRMDestReg>; // R32 -= R32+Borrow
-def SBBrm32 : IM32 <"sbb", 0x19, MRMSrcMem >; // R32 -= [mem32]+Borrow
-def SBBmr32 : IM32 <"sbb", 0x1B, MRMDestMem>; // [mem32] -= R32+Borrow
-
-def IMULrr16 : I <"imul", 0xAF, MRMSrcReg>, TB, OpSize, Pattern<(set R16, (times R16, R16))>;
-def IMULrr32 : I <"imul", 0xAF, MRMSrcReg>, TB , Pattern<(set R32, (times R32, R32))>;
-def IMULrm16 : IM16 <"imul", 0xAF, MRMSrcMem>, TB, OpSize;
-def IMULrm32 : IM32 <"imul", 0xAF, MRMSrcMem>, TB ;
+def ADD8rr : I <"add", 0x00, MRMDestReg>, Pattern<(set R8 , (plus R8 , R8 ))>;
+def ADD16rr : I <"add", 0x01, MRMDestReg>, OpSize, Pattern<(set R16, (plus R16, R16))>;
+def ADD32rr : I <"add", 0x01, MRMDestReg>, Pattern<(set R32, (plus R32, R32))>;
+def ADD8mr : Im8 <"add", 0x00, MRMDestMem>; // [mem8] += R8
+def ADD16mr : Im16 <"add", 0x01, MRMDestMem>, OpSize; // [mem16] += R16
+def ADD32mr : Im32 <"add", 0x01, MRMDestMem>; // [mem32] += R32
+def ADD8rm : Im8 <"add", 0x02, MRMSrcMem >; // R8 += [mem8]
+def ADD16rm : Im16 <"add", 0x03, MRMSrcMem >, OpSize; // R16 += [mem16]
+def ADD32rm : Im32 <"add", 0x03, MRMSrcMem >; // R32 += [mem32]
+
+def ADD8ri : Ii8 <"add", 0x80, MRM0r >, Pattern<(set R8 , (plus R8 , imm))>;
+def ADD16ri : Ii16 <"add", 0x81, MRM0r >, OpSize, Pattern<(set R16, (plus R16, imm))>;
+def ADD32ri : Ii32 <"add", 0x81, MRM0r >, Pattern<(set R32, (plus R32, imm))>;
+def ADD8mi : Im8i8 <"add", 0x80, MRM0m >; // [mem8] += I8
+def ADD16mi : Im16i16<"add", 0x81, MRM0m >, OpSize; // [mem16] += I16
+def ADD32mi : Im32i32<"add", 0x81, MRM0m >; // [mem32] += I32
+
+def ADD16ri8 : Ii8 <"add", 0x83, MRM0r >, OpSize; // ADDri with sign extended 8 bit imm
+def ADD32ri8 : Ii8 <"add", 0x83, MRM0r >;
+def ADD16mi8 : Im16i8<"add", 0x83, MRM0m >, OpSize; // [mem16] += I8
+def ADD32mi8 : Im32i8<"add", 0x83, MRM0m >; // [mem32] += I8
+
+def ADC32rr : I <"adc", 0x11, MRMDestReg>; // R32 += R32+Carry
+def ADC32rm : Im32 <"adc", 0x11, MRMSrcMem >; // R32 += [mem32]+Carry
+def ADC32mr : Im32 <"adc", 0x13, MRMDestMem>; // [mem32] += R32+Carry
+def ADC32ri : Ii32 <"adc", 0x81, MRM2r >; // R32 += I32+Carry
+def ADC32ri8 : Ii8 <"adc", 0x83, MRM2r >; // R32 += I8+Carry
+def ADC32mi : Im32i32<"adc", 0x81, MRM2m >; // [mem32] += I32+Carry
+def ADC32mi8 : Im32i8 <"adc", 0x83, MRM2m >; // [mem32] += I8+Carry
+
+def SUB8rr : I <"sub", 0x28, MRMDestReg>, Pattern<(set R8 , (minus R8 , R8 ))>;
+def SUB16rr : I <"sub", 0x29, MRMDestReg>, OpSize, Pattern<(set R16, (minus R16, R16))>;
+def SUB32rr : I <"sub", 0x29, MRMDestReg>, Pattern<(set R32, (minus R32, R32))>;
+def SUB8mr : Im8 <"sub", 0x28, MRMDestMem>; // [mem8] -= R8
+def SUB16mr : Im16 <"sub", 0x29, MRMDestMem>, OpSize; // [mem16] -= R16
+def SUB32mr : Im32 <"sub", 0x29, MRMDestMem>; // [mem32] -= R32
+def SUB8rm : Im8 <"sub", 0x2A, MRMSrcMem >; // R8 -= [mem8]
+def SUB16rm : Im16 <"sub", 0x2B, MRMSrcMem >, OpSize; // R16 -= [mem16]
+def SUB32rm : Im32 <"sub", 0x2B, MRMSrcMem >; // R32 -= [mem32]
+
+def SUB8ri : Ii8 <"sub", 0x80, MRM5r >, Pattern<(set R8 , (minus R8 , imm))>;
+def SUB16ri : Ii16 <"sub", 0x81, MRM5r >, OpSize, Pattern<(set R16, (minus R16, imm))>;
+def SUB32ri : Ii32 <"sub", 0x81, MRM5r >, Pattern<(set R32, (minus R32, imm))>;
+def SUB8mi : Im8i8 <"sub", 0x80, MRM5m >; // [mem8] -= I8
+def SUB16mi : Im16i16<"sub", 0x81, MRM5m >, OpSize; // [mem16] -= I16
+def SUB32mi : Im32i32<"sub", 0x81, MRM5m >; // [mem32] -= I32
+
+def SUB16ri8 : Ii8 <"sub", 0x83, MRM5r >, OpSize;
+def SUB32ri8 : Ii8 <"sub", 0x83, MRM5r >;
+def SUB16mi8 : Im16i8<"sub", 0x83, MRM5m >, OpSize; // [mem16] -= I8
+def SUB32mi8 : Im32i8<"sub", 0x83, MRM5m >; // [mem32] -= I8
+
+def SBB32rr : I <"sbb", 0x19, MRMDestReg>; // R32 -= R32+Borrow
+def SBB32rm : Im32 <"sbb", 0x19, MRMSrcMem >; // R32 -= [mem32]+Borrow
+def SBB32mr : Im32 <"sbb", 0x1B, MRMDestMem>; // [mem32] -= R32+Borrow
+def SBB32ri : Ii32 <"adc", 0x81, MRM3r >; // R32 -= I32+Borrow
+def SBB32ri8 : Ii8 <"adc", 0x83, MRM3r >; // R32 -= I8+Borrow
+def SBB32mi : Im32i32<"adc", 0x81, MRM3m >; // [mem32] -= I32+Borrow
+def SBB32mi8 : Im32i8 <"adc", 0x83, MRM3m >; // [mem32] -= I8+Borrow
+
+def IMUL16rr : I <"imul", 0xAF, MRMSrcReg>, TB, OpSize, Pattern<(set R16, (times R16, R16))>;
+def IMUL32rr : I <"imul", 0xAF, MRMSrcReg>, TB , Pattern<(set R32, (times R32, R32))>;
+def IMUL16rm : Im16 <"imul", 0xAF, MRMSrcMem>, TB, OpSize;
+def IMUL32rm : Im32 <"imul", 0xAF, MRMSrcMem>, TB ;
} // end Two Address instructions
// These are suprisingly enough not two address instructions!
-def IMULrri16 : II16 <"imul", 0x69, MRMSrcReg>, OpSize; // R16 = R16*I16
-def IMULrri32 : II32 <"imul", 0x69, MRMSrcReg>; // R32 = R32*I32
-def IMULrri16b : II8 <"imul", 0x6B, MRMSrcReg>, OpSize; // R16 = R16*I8
-def IMULrri32b : II8 <"imul", 0x6B, MRMSrcReg>; // R32 = R32*I8
-def IMULrmi16 : I16MI <"imul", 0x69, MRMSrcMem>, OpSize; // R16 = [mem16]*I16
-def IMULrmi32 : I32MI <"imul", 0x69, MRMSrcMem>; // R32 = [mem32]*I32
-def IMULrmi16b : IM16I8<"imul", 0x6B, MRMSrcMem>, OpSize; // R16 = [mem16]*I8
-def IMULrmi32b : IM32I8<"imul", 0x6B, MRMSrcMem>; // R32 = [mem32]*I8
+def IMUL16rri : Ii16 <"imul", 0x69, MRMSrcReg>, OpSize; // R16 = R16*I16
+def IMUL32rri : Ii32 <"imul", 0x69, MRMSrcReg>; // R32 = R32*I32
+def IMUL16rri8 : Ii8 <"imul", 0x6B, MRMSrcReg>, OpSize; // R16 = R16*I8
+def IMUL32rri8 : Ii8 <"imul", 0x6B, MRMSrcReg>; // R32 = R32*I8
+def IMUL16rmi : Im16i16<"imul",0x69, MRMSrcMem>, OpSize; // R16 = [mem16]*I16
+def IMUL32rmi : Im32i32<"imul",0x69, MRMSrcMem>; // R32 = [mem32]*I32
+def IMUL16rmi8 : Im16i8<"imul", 0x6B, MRMSrcMem>, OpSize; // R16 = [mem16]*I8
+def IMUL32rmi8 : Im32i8<"imul", 0x6B, MRMSrcMem>; // R32 = [mem32]*I8
//===----------------------------------------------------------------------===//
// Test instructions are just like AND, except they don't generate a result.
-def TESTrr8 : I <"test", 0x84, MRMDestReg>; // flags = R8 & R8
-def TESTrr16 : I <"test", 0x85, MRMDestReg>, OpSize; // flags = R16 & R16
-def TESTrr32 : I <"test", 0x85, MRMDestReg>; // flags = R32 & R32
-def TESTmr8 : IM8 <"test", 0x84, MRMDestMem>; // flags = [mem8] & R8
-def TESTmr16 : IM16 <"test", 0x85, MRMDestMem>, OpSize; // flags = [mem16] & R16
-def TESTmr32 : IM32 <"test", 0x85, MRMDestMem>; // flags = [mem32] & R32
-def TESTrm8 : IM8 <"test", 0x84, MRMSrcMem >; // flags = R8 & [mem8]
-def TESTrm16 : IM16 <"test", 0x85, MRMSrcMem >, OpSize; // flags = R16 & [mem16]
-def TESTrm32 : IM32 <"test", 0x85, MRMSrcMem >; // flags = R32 & [mem32]
+def TEST8rr : I <"test", 0x84, MRMDestReg>; // flags = R8 & R8
+def TEST16rr : I <"test", 0x85, MRMDestReg>, OpSize; // flags = R16 & R16
+def TEST32rr : I <"test", 0x85, MRMDestReg>; // flags = R32 & R32
+def TEST8mr : Im8 <"test", 0x84, MRMDestMem>; // flags = [mem8] & R8
+def TEST16mr : Im16 <"test", 0x85, MRMDestMem>, OpSize; // flags = [mem16] & R16
+def TEST32mr : Im32 <"test", 0x85, MRMDestMem>; // flags = [mem32] & R32
+def TEST8rm : Im8 <"test", 0x84, MRMSrcMem >; // flags = R8 & [mem8]
+def TEST16rm : Im16 <"test", 0x85, MRMSrcMem >, OpSize; // flags = R16 & [mem16]
+def TEST32rm : Im32 <"test", 0x85, MRMSrcMem >; // flags = R32 & [mem32]
-def TESTri8 : II8 <"test", 0xF6, MRM0r >; // flags = R8 & imm8
-def TESTri16 : II16 <"test", 0xF7, MRM0r >, OpSize; // flags = R16 & imm16
-def TESTri32 : II32 <"test", 0xF7, MRM0r >; // flags = R32 & imm32
-def TESTmi8 : I8MI <"test", 0xF6, MRM0m >; // flags = [mem8] & imm8
-def TESTmi16 : I16MI<"test", 0xF7, MRM0m >, OpSize; // flags = [mem16] & imm16
-def TESTmi32 : I32MI<"test", 0xF7, MRM0m >; // flags = [mem32] & imm32
+def TEST8ri : Ii8 <"test", 0xF6, MRM0r >; // flags = R8 & imm8
+def TEST16ri : Ii16 <"test", 0xF7, MRM0r >, OpSize; // flags = R16 & imm16
+def TEST32ri : Ii32 <"test", 0xF7, MRM0r >; // flags = R32 & imm32
+def TEST8mi : Im8i8 <"test", 0xF6, MRM0m >; // flags = [mem8] & imm8
+def TEST16mi : Im16i16<"test", 0xF7, MRM0m >, OpSize; // flags = [mem16] & imm16
+def TEST32mi : Im32i32<"test", 0xF7, MRM0m >; // flags = [mem32] & imm32
// Condition code ops, incl. set if equal/not equal/...
def SAHF : I <"sahf" , 0x9E, RawFrm>, Imp<[AH],[]>; // flags = AH
+def LAHF : I <"lahf" , 0x9F, RawFrm>, Imp<[],[AH]>; // AH = flags
def SETBr : I <"setb" , 0x92, MRM0r>, TB; // R8 = < unsign
-def SETBm : IM8<"setb" , 0x92, MRM0m>, TB; // [mem8] = < unsign
+def SETBm : Im8<"setb" , 0x92, MRM0m>, TB; // [mem8] = < unsign
def SETAEr : I <"setae", 0x93, MRM0r>, TB; // R8 = >= unsign
-def SETAEm : IM8<"setae", 0x93, MRM0m>, TB; // [mem8] = >= unsign
+def SETAEm : Im8<"setae", 0x93, MRM0m>, TB; // [mem8] = >= unsign
def SETEr : I <"sete" , 0x94, MRM0r>, TB; // R8 = ==
-def SETEm : IM8<"sete" , 0x94, MRM0m>, TB; // [mem8] = ==
+def SETEm : Im8<"sete" , 0x94, MRM0m>, TB; // [mem8] = ==
def SETNEr : I <"setne", 0x95, MRM0r>, TB; // R8 = !=
-def SETNEm : IM8<"setne", 0x95, MRM0m>, TB; // [mem8] = !=
+def SETNEm : Im8<"setne", 0x95, MRM0m>, TB; // [mem8] = !=
def SETBEr : I <"setbe", 0x96, MRM0r>, TB; // R8 = <= unsign
-def SETBEm : IM8<"setbe", 0x96, MRM0m>, TB; // [mem8] = <= unsign
+def SETBEm : Im8<"setbe", 0x96, MRM0m>, TB; // [mem8] = <= unsign
def SETAr : I <"seta" , 0x97, MRM0r>, TB; // R8 = > signed
-def SETAm : IM8<"seta" , 0x97, MRM0m>, TB; // [mem8] = > signed
+def SETAm : Im8<"seta" , 0x97, MRM0m>, TB; // [mem8] = > signed
def SETSr : I <"sets" , 0x98, MRM0r>, TB; // R8 = <sign bit>
-def SETSm : IM8<"sets" , 0x98, MRM0m>, TB; // [mem8] = <sign bit>
+def SETSm : Im8<"sets" , 0x98, MRM0m>, TB; // [mem8] = <sign bit>
def SETNSr : I <"setns", 0x99, MRM0r>, TB; // R8 = !<sign bit>
-def SETNSm : IM8<"setns", 0x99, MRM0m>, TB; // [mem8] = !<sign bit>
+def SETNSm : Im8<"setns", 0x99, MRM0m>, TB; // [mem8] = !<sign bit>
def SETLr : I <"setl" , 0x9C, MRM0r>, TB; // R8 = < signed
-def SETLm : IM8<"setl" , 0x9C, MRM0m>, TB; // [mem8] = < signed
+def SETLm : Im8<"setl" , 0x9C, MRM0m>, TB; // [mem8] = < signed
def SETGEr : I <"setge", 0x9D, MRM0r>, TB; // R8 = >= signed
-def SETGEm : IM8<"setge", 0x9D, MRM0m>, TB; // [mem8] = >= signed
+def SETGEm : Im8<"setge", 0x9D, MRM0m>, TB; // [mem8] = >= signed
def SETLEr : I <"setle", 0x9E, MRM0r>, TB; // R8 = <= signed
-def SETLEm : IM8<"setle", 0x9E, MRM0m>, TB; // [mem8] = <= signed
+def SETLEm : Im8<"setle", 0x9E, MRM0m>, TB; // [mem8] = <= signed
def SETGr : I <"setg" , 0x9F, MRM0r>, TB; // R8 = < signed
-def SETGm : IM8<"setg" , 0x9F, MRM0m>, TB; // [mem8] = < signed
+def SETGm : Im8<"setg" , 0x9F, MRM0m>, TB; // [mem8] = < signed
// Integer comparisons
-def CMPrr8 : I <"cmp", 0x38, MRMDestReg>; // compare R8, R8
-def CMPrr16 : I <"cmp", 0x39, MRMDestReg>, OpSize; // compare R16, R16
-def CMPrr32 : I <"cmp", 0x39, MRMDestReg>, // compare R32, R32
+def CMP8rr : I <"cmp", 0x38, MRMDestReg>; // compare R8, R8
+def CMP16rr : I <"cmp", 0x39, MRMDestReg>, OpSize; // compare R16, R16
+def CMP32rr : I <"cmp", 0x39, MRMDestReg>, // compare R32, R32
Pattern<(isVoid (unspec2 R32, R32))>;
-def CMPmr8 : IM8 <"cmp", 0x38, MRMDestMem>; // compare [mem8], R8
-def CMPmr16 : IM16 <"cmp", 0x39, MRMDestMem>, OpSize; // compare [mem16], R16
-def CMPmr32 : IM32 <"cmp", 0x39, MRMDestMem>; // compare [mem32], R32
-def CMPrm8 : IM8 <"cmp", 0x3A, MRMSrcMem >; // compare R8, [mem8]
-def CMPrm16 : IM16 <"cmp", 0x3B, MRMSrcMem >, OpSize; // compare R16, [mem16]
-def CMPrm32 : IM32 <"cmp", 0x3B, MRMSrcMem >; // compare R32, [mem32]
-def CMPri8 : II8 <"cmp", 0x80, MRM7r >; // compare R8, imm8
-def CMPri16 : II16 <"cmp", 0x81, MRM7r >, OpSize; // compare R16, imm16
-def CMPri32 : II32 <"cmp", 0x81, MRM7r >; // compare R32, imm32
-def CMPmi8 : I8MI <"cmp", 0x80, MRM7m >; // compare [mem8], imm8
-def CMPmi16 : I16MI<"cmp", 0x81, MRM7m >, OpSize; // compare [mem16], imm16
-def CMPmi32 : I32MI<"cmp", 0x81, MRM7m >; // compare [mem32], imm32
+def CMP8mr : Im8 <"cmp", 0x38, MRMDestMem>; // compare [mem8], R8
+def CMP16mr : Im16 <"cmp", 0x39, MRMDestMem>, OpSize; // compare [mem16], R16
+def CMP32mr : Im32 <"cmp", 0x39, MRMDestMem>; // compare [mem32], R32
+def CMP8rm : Im8 <"cmp", 0x3A, MRMSrcMem >; // compare R8, [mem8]
+def CMP16rm : Im16 <"cmp", 0x3B, MRMSrcMem >, OpSize; // compare R16, [mem16]
+def CMP32rm : Im32 <"cmp", 0x3B, MRMSrcMem >; // compare R32, [mem32]
+def CMP8ri : Ii8 <"cmp", 0x80, MRM7r >; // compare R8, imm8
+def CMP16ri : Ii16 <"cmp", 0x81, MRM7r >, OpSize; // compare R16, imm16
+def CMP32ri : Ii32 <"cmp", 0x81, MRM7r >; // compare R32, imm32
+def CMP8mi : Im8i8 <"cmp", 0x80, MRM7m >; // compare [mem8], imm8
+def CMP16mi : Im16i16<"cmp", 0x81, MRM7m >, OpSize; // compare [mem16], imm16
+def CMP32mi : Im32i32<"cmp", 0x81, MRM7m >; // compare [mem32], imm32
// Sign/Zero extenders
-def MOVSXr16r8 : I <"movsx", 0xBE, MRMSrcReg>, TB, OpSize; // R16 = signext(R8)
-def MOVSXr32r8 : I <"movsx", 0xBE, MRMSrcReg>, TB; // R32 = signext(R8)
-def MOVSXr32r16: I <"movsx", 0xBF, MRMSrcReg>, TB; // R32 = signext(R16)
-def MOVSXr16m8 : IM8 <"movsx", 0xBE, MRMSrcMem>, TB, OpSize; // R16 = signext([mem8])
-def MOVSXr32m8 : IM8 <"movsx", 0xBE, MRMSrcMem>, TB; // R32 = signext([mem8])
-def MOVSXr32m16: IM16<"movsx", 0xBF, MRMSrcMem>, TB; // R32 = signext([mem16])
+def MOVSX16rr8 : I <"movsx", 0xBE, MRMSrcReg>, TB, OpSize; // R16 = signext(R8)
+def MOVSX32rr8 : I <"movsx", 0xBE, MRMSrcReg>, TB; // R32 = signext(R8)
+def MOVSX32rr16: I <"movsx", 0xBF, MRMSrcReg>, TB; // R32 = signext(R16)
+def MOVSX16rm8 : Im8 <"movsx", 0xBE, MRMSrcMem>, TB, OpSize; // R16 = signext([mem8])
+def MOVSX32rm8 : Im8 <"movsx", 0xBE, MRMSrcMem>, TB; // R32 = signext([mem8])
+def MOVSX32rm16: Im16<"movsx", 0xBF, MRMSrcMem>, TB; // R32 = signext([mem16])
-def MOVZXr16r8 : I <"movzx", 0xB6, MRMSrcReg>, TB, OpSize; // R16 = zeroext(R8)
-def MOVZXr32r8 : I <"movzx", 0xB6, MRMSrcReg>, TB; // R32 = zeroext(R8)
-def MOVZXr32r16: I <"movzx", 0xB7, MRMSrcReg>, TB; // R32 = zeroext(R16)
-def MOVZXr16m8 : IM8 <"movzx", 0xB6, MRMSrcMem>, TB, OpSize; // R16 = zeroext([mem8])
-def MOVZXr32m8 : IM8 <"movzx", 0xB6, MRMSrcMem>, TB; // R32 = zeroext([mem8])
-def MOVZXr32m16: IM16<"movzx", 0xB7, MRMSrcMem>, TB; // R32 = zeroext([mem16])
+def MOVZX16rr8 : I <"movzx", 0xB6, MRMSrcReg>, TB, OpSize; // R16 = zeroext(R8)
+def MOVZX32rr8 : I <"movzx", 0xB6, MRMSrcReg>, TB; // R32 = zeroext(R8)
+def MOVZX32rr16: I <"movzx", 0xB7, MRMSrcReg>, TB; // R32 = zeroext(R16)
+def MOVZX16rm8 : Im8 <"movzx", 0xB6, MRMSrcMem>, TB, OpSize; // R16 = zeroext([mem8])
+def MOVZX32rm8 : Im8 <"movzx", 0xB6, MRMSrcMem>, TB; // R32 = zeroext([mem8])
+def MOVZX32rm16: Im16<"movzx", 0xB7, MRMSrcMem>, TB; // R32 = zeroext([mem16])
//===----------------------------------------------------------------------===//
class FPIM<string n, bits<8> o, Format F, FPFormat fp, MemType m> : FPInst<n, o, F, fp, m, NoImm>;
-class FPIM16<string n, bits<8> o, Format F, FPFormat fp> : FPIM<n, o, F, fp, Mem16>;
-class FPIM32<string n, bits<8> o, Format F, FPFormat fp> : FPIM<n, o, F, fp, Mem32>;
-class FPIM64<string n, bits<8> o, Format F, FPFormat fp> : FPIM<n, o, F, fp, Mem64>;
-class FPIM80<string n, bits<8> o, Format F, FPFormat fp> : FPIM<n, o, F, fp, Mem80>;
+class FPI16m<string n, bits<8> o, Format F, FPFormat fp> : FPIM<n, o, F, fp, Mem16>;
+class FPI32m<string n, bits<8> o, Format F, FPFormat fp> : FPIM<n, o, F, fp, Mem32>;
+class FPI64m<string n, bits<8> o, Format F, FPFormat fp> : FPIM<n, o, F, fp, Mem64>;
+class FPI80m<string n, bits<8> o, Format F, FPFormat fp> : FPIM<n, o, F, fp, Mem80>;
// Pseudo instructions for floating point. We use these pseudo instructions
// because they can be expanded by the fp spackifier into one of many different
def FpGETRESULT : FPI<"FGETRESULT",0, Pseudo, SpecialFP>; // FPR = ST(0)
def FpSETRESULT : FPI<"FSETRESULT",0, Pseudo, SpecialFP>; // ST(0) = FPR
+
+// Floating point cmovs...
+let isTwoAddress = 1, Uses = [ST0], Defs = [ST0], printImplicitUsesBefore = 1 in {
+ def FCMOVB : FPI <"fcmovb" , 0xC0, AddRegFrm, CondMovFP>, DA; // fcmovb ST(i) -> ST(0)
+ def FCMOVBE : FPI <"fcmovbe", 0xD0, AddRegFrm, CondMovFP>, DA; // fcmovbe ST(i) -> ST(0)
+ def FCMOVE : FPI <"fcmove" , 0xC8, AddRegFrm, CondMovFP>, DA; // fcmove ST(i) -> ST(0)
+ def FCMOVAE : FPI <"fcmovae", 0xC0, AddRegFrm, CondMovFP>, DB; // fcmovae ST(i) -> ST(0)
+ def FCMOVA : FPI <"fcmova" , 0xD0, AddRegFrm, CondMovFP>, DB; // fcmova ST(i) -> ST(0)
+ def FCMOVNE : FPI <"fcmovne", 0xC8, AddRegFrm, CondMovFP>, DB; // fcmovne ST(i) -> ST(0)
+}
+
// Floating point loads & stores...
def FLDrr : FPI <"fld" , 0xC0, AddRegFrm, NotFP>, D9; // push(ST(i))
-def FLDr32 : FPIM32 <"fld" , 0xD9, MRM0m , ZeroArgFP>; // load float
-def FLDr64 : FPIM64 <"fld" , 0xDD, MRM0m , ZeroArgFP>; // load double
-def FLDr80 : FPIM80 <"fld" , 0xDB, MRM5m , ZeroArgFP>; // load extended
-def FILDr16 : FPIM16 <"fild" , 0xDF, MRM0m , ZeroArgFP>; // load signed short
-def FILDr32 : FPIM32 <"fild" , 0xDB, MRM0m , ZeroArgFP>; // load signed int
-def FILDr64 : FPIM64 <"fild" , 0xDF, MRM5m , ZeroArgFP>; // load signed long
+def FLD32m : FPI32m <"fld" , 0xD9, MRM0m , ZeroArgFP>; // load float
+def FLD64m : FPI64m <"fld" , 0xDD, MRM0m , ZeroArgFP>; // load double
+def FLD80m : FPI80m <"fld" , 0xDB, MRM5m , ZeroArgFP>; // load extended
+def FILD16m : FPI16m <"fild" , 0xDF, MRM0m , ZeroArgFP>; // load signed short
+def FILD32m : FPI32m <"fild" , 0xDB, MRM0m , ZeroArgFP>; // load signed int
+def FILD64m : FPI64m <"fild" , 0xDF, MRM5m , ZeroArgFP>; // load signed long
def FSTrr : FPI <"fst" , 0xD0, AddRegFrm, NotFP >, DD; // ST(i) = ST(0)
def FSTPrr : FPI <"fstp", 0xD8, AddRegFrm, NotFP >, DD; // ST(i) = ST(0), pop
-def FSTr32 : FPIM32 <"fst" , 0xD9, MRM2m , OneArgFP>; // store float
-def FSTr64 : FPIM64 <"fst" , 0xDD, MRM2m , OneArgFP>; // store double
-def FSTPr32 : FPIM32 <"fstp", 0xD9, MRM3m , OneArgFP>; // store float, pop
-def FSTPr64 : FPIM64 <"fstp", 0xDD, MRM3m , OneArgFP>; // store double, pop
-def FSTPr80 : FPIM80 <"fstp", 0xDB, MRM7m , OneArgFP>; // store extended, pop
-
-def FISTr16 : FPIM16 <"fist", 0xDF, MRM2m , OneArgFP>; // store signed short
-def FISTr32 : FPIM32 <"fist", 0xDB, MRM2m , OneArgFP>; // store signed int
-def FISTPr16 : FPIM16 <"fistp", 0xDF, MRM3m , NotFP >; // store signed short, pop
-def FISTPr32 : FPIM32 <"fistp", 0xDB, MRM3m , NotFP >; // store signed int, pop
-def FISTPr64 : FPIM64 <"fistpll", 0xDF, MRM7m , OneArgFP>; // store signed long, pop
+def FST32m : FPI32m <"fst" , 0xD9, MRM2m , OneArgFP>; // store float
+def FST64m : FPI64m <"fst" , 0xDD, MRM2m , OneArgFP>; // store double
+def FSTP32m : FPI32m <"fstp", 0xD9, MRM3m , OneArgFP>; // store float, pop
+def FSTP64m : FPI64m <"fstp", 0xDD, MRM3m , OneArgFP>; // store double, pop
+def FSTP80m : FPI80m <"fstp", 0xDB, MRM7m , OneArgFP>; // store extended, pop
+
+def FIST16m : FPI16m <"fist", 0xDF, MRM2m , OneArgFP>; // store signed short
+def FIST32m : FPI32m <"fist", 0xDB, MRM2m , OneArgFP>; // store signed int
+def FISTP16m : FPI16m <"fistp", 0xDF, MRM3m , NotFP >; // store signed short, pop
+def FISTP32m : FPI32m <"fistp", 0xDB, MRM3m , NotFP >; // store signed int, pop
+def FISTP64m : FPI64m <"fistpll", 0xDF, MRM7m , OneArgFP>; // store signed long, pop
def FXCH : FPI <"fxch", 0xC8, AddRegFrm, NotFP>, D9; // fxch ST(i), ST(0)
list<Register> Defs = [ST0];
}
class FPrST0Inst<string n, bits<8> o> : I<n, o, AddRegFrm>, DC {
- bit printImplicitUses = 1;
+ bit printImplicitUsesAfter = 1;
list<Register> Uses = [ST0];
}
class FPrST0PInst<string n, bits<8> o> : I<n, o, AddRegFrm>, DE {
def FUCOMPPr : I<"fucompp", 0xE9, RawFrm >, DA, Imp<[ST0],[]>; // compare ST(0) with ST(1), pop, pop
// Floating point flag ops
-def FNSTSWr8 : I <"fnstsw" , 0xE0, RawFrm>, DF, Imp<[],[AX]>; // AX = fp flags
-def FNSTCWm16 : IM16<"fnstcw" , 0xD9, MRM7m >; // [mem16] = X87 control world
-def FLDCWm16 : IM16<"fldcw" , 0xD9, MRM5m >; // X87 control world = [mem16]
+def FNSTSW8r : I <"fnstsw" , 0xE0, RawFrm>, DF, Imp<[],[AX]>; // AX = fp flags
+def FNSTCW16m : Im16<"fnstcw" , 0xD9, MRM7m >; // [mem16] = X87 control world
+def FLDCW16m : Im16<"fldcw" , 0xD9, MRM5m >; // X87 control world = [mem16]
//===----------------------------------------------------------------------===//
//
def RET_R32 : Expander<(ret R32:$reg),
- [(MOVrr32 EAX, R32:$reg),
+ [(MOV32rr 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:$fi, 1, 0/*NoReg*/, 0)]>;
+ [(MOV16rm R16:$dest, frameidx:$fi, 1, 0/*NoReg*/, 0)]>;
def LOAD_FI32 : Expander<(set R32:$dest, (load frameidx:$fi)),
- [(MOVrm32 R32:$dest, frameidx:$fi, 1, 0/*NoReg*/, 0)]>;
+ [(MOV32rm R32:$dest, frameidx:$fi, 1, 0/*NoReg*/, 0)]>;
def LOAD_R16 : Expander<(set R16:$dest, (load R32:$src)),
- [(MOVrm16 R16:$dest, R32:$src, 1, 0/*NoReg*/, 0)]>;
+ [(MOV16rm R16:$dest, R32:$src, 1, 0/*NoReg*/, 0)]>;
def LOAD_R32 : Expander<(set R32:$dest, (load R32:$src)),
- [(MOVrm32 R32:$dest, R32:$src, 1, 0/*NoReg*/, 0)]>;
+ [(MOV32rm R32:$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:$a2),
+ [(CMP32rr R32:$a1, R32:$a2),
(JE basicblock:$d1),
(JMP basicblock:$d2)]>;