def SDT_X86Call : SDTypeProfile<0, 1, [SDTCisVT<0, i32>]>;
-def SDTX86FpGet : SDTypeProfile<1, 0, [SDTCisVT<0, f64>]>;
-def SDTX86FpSet : SDTypeProfile<0, 1, [SDTCisFP<0>]>;
-
-def SDTX86Fld : SDTypeProfile<1, 2, [SDTCisVT<0, f64>,
- SDTCisPtrTy<1>, SDTCisVT<2, OtherVT>]>;
-def SDTX86Fst : SDTypeProfile<0, 3, [SDTCisFP<0>,
- SDTCisPtrTy<1>, SDTCisVT<2, OtherVT>]>;
-def SDTX86Fild : SDTypeProfile<1, 2, [SDTCisVT<0, f64>, SDTCisPtrTy<1>,
- SDTCisVT<2, OtherVT>]>;
-def SDTX86FpToIMem: SDTypeProfile<0, 2, [SDTCisFP<0>, SDTCisPtrTy<1>]>;
-
def SDTX86RepStr : SDTypeProfile<0, 1, [SDTCisVT<0, OtherVT>]>;
def SDTX86RdTsc : SDTypeProfile<0, 0, []>;
+def SDTX86Wrapper : SDTypeProfile<1, 1, [SDTCisSameAs<0, 1>, SDTCisPtrTy<0>]>;
+
def X86shld : SDNode<"X86ISD::SHLD", SDTIntShiftDOp>;
def X86shrd : SDNode<"X86ISD::SHRD", SDTIntShiftDOp>;
-def X86fand : SDNode<"X86ISD::FAND", SDTFPBinOp,
- [SDNPCommutative, SDNPAssociative]>;
-def X86fxor : SDNode<"X86ISD::FXOR", SDTFPBinOp,
- [SDNPCommutative, SDNPAssociative]>;
-
def X86cmp : SDNode<"X86ISD::CMP" , SDTX86CmpTest,
[SDNPOutFlag]>;
def X86test : SDNode<"X86ISD::TEST", SDTX86CmpTest,
def X86call : SDNode<"X86ISD::CALL", SDT_X86Call,
[SDNPHasChain, SDNPOutFlag, SDNPOptInFlag]>;
-def X86fpget : SDNode<"X86ISD::FP_GET_RESULT", SDTX86FpGet,
- [SDNPHasChain, SDNPInFlag, SDNPOutFlag]>;
-def X86fpset : SDNode<"X86ISD::FP_SET_RESULT", SDTX86FpSet,
- [SDNPHasChain, SDNPOutFlag]>;
-
-def X86fld : SDNode<"X86ISD::FLD", SDTX86Fld,
- [SDNPHasChain]>;
-def X86fst : SDNode<"X86ISD::FST", SDTX86Fst,
- [SDNPHasChain, SDNPInFlag]>;
-def X86fild : SDNode<"X86ISD::FILD", SDTX86Fild,
- [SDNPHasChain]>;
-def X86fildflag: SDNode<"X86ISD::FILD_FLAG",SDTX86Fild,
- [SDNPHasChain, SDNPOutFlag]>;
-def X86fp_to_i16mem : SDNode<"X86ISD::FP_TO_INT16_IN_MEM", SDTX86FpToIMem,
- [SDNPHasChain]>;
-def X86fp_to_i32mem : SDNode<"X86ISD::FP_TO_INT32_IN_MEM", SDTX86FpToIMem,
- [SDNPHasChain]>;
-def X86fp_to_i64mem : SDNode<"X86ISD::FP_TO_INT64_IN_MEM", SDTX86FpToIMem,
- [SDNPHasChain]>;
-
def X86rep_stos: SDNode<"X86ISD::REP_STOS", SDTX86RepStr,
- [SDNPHasChain, SDNPInFlag]>;
+ [SDNPHasChain, SDNPInFlag, SDNPOutFlag]>;
def X86rep_movs: SDNode<"X86ISD::REP_MOVS", SDTX86RepStr,
- [SDNPHasChain, SDNPInFlag]>;
+ [SDNPHasChain, SDNPInFlag, SDNPOutFlag]>;
def X86rdtsc : SDNode<"X86ISD::RDTSC_DAG",SDTX86RdTsc,
[SDNPHasChain, SDNPOutFlag]>;
-def X86loadp : SDNode<"X86ISD::LOAD_PACK", SDTLoad,
- [SDNPHasChain]>;
+def X86Wrapper : SDNode<"X86ISD::Wrapper", SDTX86Wrapper>;
//===----------------------------------------------------------------------===//
// X86 Operand Definitions.
def i16mem : X86MemOperand<"printi16mem">;
def i32mem : X86MemOperand<"printi32mem">;
def i64mem : X86MemOperand<"printi64mem">;
+def i128mem : X86MemOperand<"printi128mem">;
def f32mem : X86MemOperand<"printf32mem">;
def f64mem : X86MemOperand<"printf64mem">;
def f128mem : X86MemOperand<"printf128mem">;
// Define X86 specific addressing mode.
def addr : ComplexPattern<i32, 4, "SelectAddr", []>;
def leaaddr : ComplexPattern<i32, 4, "SelectLEAAddr",
- [add, frameindex, constpool]>;
+ [add, mul, shl, frameindex]>;
//===----------------------------------------------------------------------===//
// X86 Instruction Format Definitions.
//===----------------------------------------------------------------------===//
// X86 Instruction Predicate Definitions.
+def HasMMX : Predicate<"Subtarget->hasMMX()">;
def HasSSE1 : Predicate<"Subtarget->hasSSE1()">;
def HasSSE2 : Predicate<"Subtarget->hasSSE2()">;
def HasSSE3 : Predicate<"Subtarget->hasSSE3()">;
return (unsigned)N->getValue() == (unsigned char)N->getValue();
}]>;
-def fp32imm0 : PatLeaf<(f32 fpimm), [{
- return N->isExactlyValue(+0.0);
-}]>;
-
-def fp64imm0 : PatLeaf<(f64 fpimm), [{
- return N->isExactlyValue(+0.0);
-}]>;
-
-def fp64immneg0 : PatLeaf<(f64 fpimm), [{
- return N->isExactlyValue(-0.0);
-}]>;
-
-def fp64imm1 : PatLeaf<(f64 fpimm), [{
- return N->isExactlyValue(+1.0);
-}]>;
-
-def fp64immneg1 : PatLeaf<(f64 fpimm), [{
- return N->isExactlyValue(-1.0);
-}]>;
-
// Helper fragments for loads.
def loadi8 : PatFrag<(ops node:$ptr), (i8 (load node:$ptr))>;
def loadi16 : PatFrag<(ops node:$ptr), (i16 (load node:$ptr))>;
def loadi32 : PatFrag<(ops node:$ptr), (i32 (load node:$ptr))>;
+
def loadf32 : PatFrag<(ops node:$ptr), (f32 (load node:$ptr))>;
def loadf64 : PatFrag<(ops node:$ptr), (f64 (load node:$ptr))>;
def zextloadi32i16 : PatFrag<(ops node:$ptr), (i32 (zextload node:$ptr, i16))>;
def extloadi8i1 : PatFrag<(ops node:$ptr), (i8 (extload node:$ptr, i1))>;
-def extloadf64f32 : PatFrag<(ops node:$ptr), (f64 (extload node:$ptr, f32))>;
-
-def X86loadpf32 : PatFrag<(ops node:$ptr), (f32 (X86loadp node:$ptr))>;
-def X86loadpf64 : PatFrag<(ops node:$ptr), (f64 (X86loadp node:$ptr))>;
//===----------------------------------------------------------------------===//
// Instruction templates...
def IMPLICIT_DEF_R32 : I<0, Pseudo, (ops R32:$dst),
"#IMPLICIT_DEF $dst",
[(set R32:$dst, (undef))]>;
-def IMPLICIT_DEF_FR32 : I<0, Pseudo, (ops FR32:$dst),
- "#IMPLICIT_DEF $dst",
- [(set FR32:$dst, (undef))]>, Requires<[HasSSE2]>;
-def IMPLICIT_DEF_FR64 : I<0, Pseudo, (ops FR64:$dst),
- "#IMPLICIT_DEF $dst",
- [(set FR64:$dst, (undef))]>, Requires<[HasSSE2]>;
-
-
-// CMOV* - Used to implement the SSE SELECT DAG operation. Expanded by the
-// scheduler into a branch sequence.
-let usesCustomDAGSchedInserter = 1 in { // Expanded by the scheduler.
- def CMOV_FR32 : I<0, Pseudo,
- (ops FR32:$dst, FR32:$t, FR32:$f, i8imm:$cond),
- "#CMOV_FR32 PSEUDO!",
- [(set FR32:$dst, (X86cmov FR32:$t, FR32:$f, imm:$cond))]>;
- def CMOV_FR64 : I<0, Pseudo,
- (ops FR64:$dst, FR64:$t, FR64:$f, i8imm:$cond),
- "#CMOV_FR64 PSEUDO!",
- [(set FR64:$dst, (X86cmov FR64:$t, FR64:$f, imm:$cond))]>;
-}
-
-let usesCustomDAGSchedInserter = 1 in { // Expanded by the scheduler.
- def FP_TO_INT16_IN_MEM : I<0, Pseudo,
- (ops i16mem:$dst, RFP:$src),
- "#FP_TO_INT16_IN_MEM PSEUDO!",
- [(X86fp_to_i16mem RFP:$src, addr:$dst)]>;
- def FP_TO_INT32_IN_MEM : I<0, Pseudo,
- (ops i32mem:$dst, RFP:$src),
- "#FP_TO_INT32_IN_MEM PSEUDO!",
- [(X86fp_to_i32mem RFP:$src, addr:$dst)]>;
- def FP_TO_INT64_IN_MEM : I<0, Pseudo,
- (ops i64mem:$dst, RFP:$src),
- "#FP_TO_INT64_IN_MEM PSEUDO!",
- [(X86fp_to_i64mem RFP:$src, addr:$dst)]>;
-}
-
-
-let isTerminator = 1 in
- let Defs = [FP0, FP1, FP2, FP3, FP4, FP5, FP6] in
- def FP_REG_KILL : I<0, Pseudo, (ops), "#FP_REG_KILL", []>;
-
// Nop
def NOOP : I<0x90, RawFrm, (ops), "nop", []>;
//
def IN8rr : I<0xEC, RawFrm, (ops),
"in{b} {%dx, %al|%AL, %DX}",
- [(set AL, (readport DX))]>, Imp<[DX], [AL]>;
+ []>, Imp<[DX], [AL]>;
def IN16rr : I<0xED, RawFrm, (ops),
"in{w} {%dx, %ax|%AX, %DX}",
- [(set AX, (readport DX))]>, Imp<[DX], [AX]>, OpSize;
+ []>, Imp<[DX], [AX]>, OpSize;
def IN32rr : I<0xED, RawFrm, (ops),
"in{l} {%dx, %eax|%EAX, %DX}",
- [(set EAX, (readport DX))]>, Imp<[DX],[EAX]>;
+ []>, Imp<[DX],[EAX]>;
def IN8ri : Ii8<0xE4, RawFrm, (ops i16i8imm:$port),
"in{b} {$port, %al|%AL, $port}",
- [(set AL, (readport i16immZExt8:$port))]>,
+ []>,
Imp<[], [AL]>;
def IN16ri : Ii8<0xE5, RawFrm, (ops i16i8imm:$port),
"in{w} {$port, %ax|%AX, $port}",
- [(set AX, (readport i16immZExt8:$port))]>,
+ []>,
Imp<[], [AX]>, OpSize;
def IN32ri : Ii8<0xE5, RawFrm, (ops i16i8imm:$port),
"in{l} {$port, %eax|%EAX, $port}",
- [(set EAX, (readport i16immZExt8:$port))]>,
+ []>,
Imp<[],[EAX]>;
def OUT8rr : I<0xEE, RawFrm, (ops),
"out{b} {%al, %dx|%DX, %AL}",
- [(writeport AL, DX)]>, Imp<[DX, AL], []>;
+ []>, Imp<[DX, AL], []>;
def OUT16rr : I<0xEF, RawFrm, (ops),
"out{w} {%ax, %dx|%DX, %AX}",
- [(writeport AX, DX)]>, Imp<[DX, AX], []>, OpSize;
+ []>, Imp<[DX, AX], []>, OpSize;
def OUT32rr : I<0xEF, RawFrm, (ops),
"out{l} {%eax, %dx|%DX, %EAX}",
- [(writeport EAX, DX)]>, Imp<[DX, EAX], []>;
+ []>, Imp<[DX, EAX], []>;
def OUT8ir : Ii8<0xE6, RawFrm, (ops i16i8imm:$port),
"out{b} {%al, $port|$port, %AL}",
- [(writeport AL, i16immZExt8:$port)]>,
+ []>,
Imp<[AL], []>;
def OUT16ir : Ii8<0xE7, RawFrm, (ops i16i8imm:$port),
"out{w} {%ax, $port|$port, %AX}",
- [(writeport AX, i16immZExt8:$port)]>,
+ []>,
Imp<[AX], []>, OpSize;
def OUT32ir : Ii8<0xE7, RawFrm, (ops i16i8imm:$port),
"out{l} {%eax, $port|$port, %EAX}",
- [(writeport EAX, i16immZExt8:$port)]>,
+ []>,
Imp<[EAX], []>;
//===----------------------------------------------------------------------===//
"mov{l} {$src, $dst|$dst, $src}",
[(store R32:$src, addr:$dst)]>;
-// Pseudo-instructions that map movr0 to xor.
-// FIXME: remove when we can teach regalloc that xor reg, reg is ok.
-def MOV8r0 : I<0x30, MRMInitReg, (ops R8 :$dst),
- "xor{b} $dst, $dst",
- [(set R8:$dst, 0)]>;
-def MOV16r0 : I<0x31, MRMInitReg, (ops R16:$dst),
- "xor{w} $dst, $dst",
- [(set R16:$dst, 0)]>, OpSize;
-def MOV32r0 : I<0x31, MRMInitReg, (ops R32:$dst),
- "xor{l} $dst, $dst",
- [(set R32:$dst, 0)]>;
-
//===----------------------------------------------------------------------===//
// Fixed-Register Multiplication and Division Instructions...
//
// Double shift instructions (generalizations of rotate)
-
def SHLD32rrCL : I<0xA5, MRMDestReg, (ops R32:$dst, R32:$src1, R32:$src2),
"shld{l} {%cl, $src2, $dst|$dst, $src2, %CL}",
[(set R32:$dst, (X86shld R32:$src1, R32:$src2, CL))]>,
[(set R32:$dst, (add R32:$src1, imm:$src2))]>;
}
-// FIXME: move ADD16ri8 above ADD16ri to optimize for space.
def ADD16ri8 : Ii8<0x83, MRM0r, (ops R16:$dst, R16:$src1, i16i8imm:$src2),
"add{w} {$src2, $dst|$dst, $src2}",
[(set R16:$dst, (add R16:$src1, i16immSExt8:$src2))]>,
def SBB8mi : Ii32<0x80, MRM3m, (ops i8mem:$dst, i8imm:$src2),
"sbb{b} {$src2, $dst|$dst, $src2}",
[(store (sube (loadi8 addr:$dst), imm:$src2), addr:$dst)]>;
- def SBB16mi : Ii32<0x81, MRM3m, (ops i16mem:$dst, i16imm:$src2),
- "sbb{w} {$src2, $dst|$dst, $src2}",
- [(store (sube (loadi16 addr:$dst), imm:$src2), addr:$dst)]>,
- OpSize;
def SBB32mi : Ii32<0x81, MRM3m, (ops i32mem:$dst, i32imm:$src2),
"sbb{l} {$src2, $dst|$dst, $src2}",
[(store (sube (loadi32 addr:$dst), imm:$src2), addr:$dst)]>;
- def SBB16mi8 : Ii8<0x83, MRM3m, (ops i16mem:$dst, i16i8imm :$src2),
- "sbb{w} {$src2, $dst|$dst, $src2}",
- [(store (sube (load addr:$dst), i16immSExt8:$src2), addr:$dst)]>,
- OpSize;
def SBB32mi8 : Ii8<0x83, MRM3m, (ops i32mem:$dst, i32i8imm :$src2),
"sbb{l} {$src2, $dst|$dst, $src2}",
[(store (sube (load addr:$dst), i32immSExt8:$src2), addr:$dst)]>;
}
-def SBB8ri : Ii8<0x80, MRM3r, (ops R8:$dst, R8:$src1, i8imm:$src2),
- "sbb{b} {$src2, $dst|$dst, $src2}",
- [(set R8:$dst, (sube R8:$src1, imm:$src2))]>;
-def SBB16ri : Ii16<0x81, MRM3r, (ops R16:$dst, R16:$src1, i16imm:$src2),
- "sbb{w} {$src2, $dst|$dst, $src2}",
- [(set R16:$dst, (sube R16:$src1, imm:$src2))]>, OpSize;
-
def SBB32rm : I<0x1B, MRMSrcMem, (ops R32:$dst, R32:$src1, i32mem:$src2),
"sbb{l} {$src2, $dst|$dst, $src2}",
[(set R32:$dst, (sube R32:$src1, (load addr:$src2)))]>;
def SBB32ri : Ii32<0x81, MRM3r, (ops R32:$dst, R32:$src1, i32imm:$src2),
"sbb{l} {$src2, $dst|$dst, $src2}",
[(set R32:$dst, (sube R32:$src1, imm:$src2))]>;
-
-def SBB16ri8 : Ii8<0x83, MRM3r, (ops R16:$dst, R16:$src1, i16i8imm:$src2),
- "sbb{w} {$src2, $dst|$dst, $src2}",
- [(set R16:$dst, (sube R16:$src1, i16immSExt8:$src2))]>,
- OpSize;
def SBB32ri8 : Ii8<0x83, MRM3r, (ops R32:$dst, R32:$src1, i32i8imm:$src2),
"sbb{l} {$src2, $dst|$dst, $src2}",
[(set R32:$dst, (sube R32:$src1, i32immSExt8:$src2))]>;
[(set R32:$dst, (zextloadi32i16 addr:$src))]>, TB;
//===----------------------------------------------------------------------===//
-// XMM Floating point support (requires SSE / SSE2)
-//===----------------------------------------------------------------------===//
-
-def MOVSSrr : I<0x10, MRMSrcReg, (ops FR32:$dst, FR32:$src),
- "movss {$src, $dst|$dst, $src}", []>,
- Requires<[HasSSE1]>, XS;
-def MOVSDrr : I<0x10, MRMSrcReg, (ops FR64:$dst, FR64:$src),
- "movsd {$src, $dst|$dst, $src}", []>,
- Requires<[HasSSE2]>, XD;
-
-def MOVSSrm : I<0x10, MRMSrcMem, (ops FR32:$dst, f32mem:$src),
- "movss {$src, $dst|$dst, $src}",
- [(set FR32:$dst, (loadf32 addr:$src))]>,
- Requires<[HasSSE1]>, XS;
-def MOVSSmr : I<0x11, MRMDestMem, (ops f32mem:$dst, FR32:$src),
- "movss {$src, $dst|$dst, $src}",
- [(store FR32:$src, addr:$dst)]>,
- Requires<[HasSSE1]>, XS;
-def MOVSDrm : I<0x10, MRMSrcMem, (ops FR64:$dst, f64mem:$src),
- "movsd {$src, $dst|$dst, $src}",
- [(set FR64:$dst, (loadf64 addr:$src))]>,
- Requires<[HasSSE2]>, XD;
-def MOVSDmr : I<0x11, MRMDestMem, (ops f64mem:$dst, FR64:$src),
- "movsd {$src, $dst|$dst, $src}",
- [(store FR64:$src, addr:$dst)]>,
- Requires<[HasSSE2]>, XD;
-
-def CVTTSS2SIrr: I<0x2C, MRMSrcReg, (ops R32:$dst, FR32:$src),
- "cvttss2si {$src, $dst|$dst, $src}",
- [(set R32:$dst, (fp_to_sint FR32:$src))]>,
- Requires<[HasSSE1]>, XS;
-def CVTTSS2SIrm: I<0x2C, MRMSrcMem, (ops R32:$dst, f32mem:$src),
- "cvttss2si {$src, $dst|$dst, $src}",
- [(set R32:$dst, (fp_to_sint (loadf32 addr:$src)))]>,
- Requires<[HasSSE1]>, XS;
-def CVTTSD2SIrr: I<0x2C, MRMSrcReg, (ops R32:$dst, FR64:$src),
- "cvttsd2si {$src, $dst|$dst, $src}",
- [(set R32:$dst, (fp_to_sint FR64:$src))]>,
- Requires<[HasSSE2]>, XD;
-def CVTTSD2SIrm: I<0x2C, MRMSrcMem, (ops R32:$dst, f64mem:$src),
- "cvttsd2si {$src, $dst|$dst, $src}",
- [(set R32:$dst, (fp_to_sint (loadf64 addr:$src)))]>,
- Requires<[HasSSE2]>, XD;
-def CVTSS2SDrr: I<0x5A, MRMSrcReg, (ops FR64:$dst, FR32:$src),
- "cvtss2sd {$src, $dst|$dst, $src}",
- [(set FR64:$dst, (fextend FR32:$src))]>,
- Requires<[HasSSE2]>, XS;
-def CVTSS2SDrm: I<0x5A, MRMSrcMem, (ops FR64:$dst, f32mem:$src),
- "cvtss2sd {$src, $dst|$dst, $src}",
- [(set FR64:$dst, (fextend (loadf32 addr:$src)))]>,
- Requires<[HasSSE2]>, XS;
-def CVTSD2SSrr: I<0x5A, MRMSrcReg, (ops FR32:$dst, FR64:$src),
- "cvtsd2ss {$src, $dst|$dst, $src}",
- [(set FR32:$dst, (fround FR64:$src))]>,
- Requires<[HasSSE2]>, XD;
-def CVTSD2SSrm: I<0x5A, MRMSrcMem, (ops FR32:$dst, f64mem:$src),
- "cvtsd2ss {$src, $dst|$dst, $src}",
- [(set FR32:$dst, (fround (loadf64 addr:$src)))]>,
- Requires<[HasSSE2]>, XD;
-def CVTSI2SSrr: I<0x2A, MRMSrcReg, (ops FR32:$dst, R32:$src),
- "cvtsi2ss {$src, $dst|$dst, $src}",
- [(set FR32:$dst, (sint_to_fp R32:$src))]>,
- Requires<[HasSSE2]>, XS;
-def CVTSI2SSrm: I<0x2A, MRMSrcMem, (ops FR32:$dst, i32mem:$src),
- "cvtsi2ss {$src, $dst|$dst, $src}",
- [(set FR32:$dst, (sint_to_fp (loadi32 addr:$src)))]>,
- Requires<[HasSSE2]>, XS;
-def CVTSI2SDrr: I<0x2A, MRMSrcReg, (ops FR64:$dst, R32:$src),
- "cvtsi2sd {$src, $dst|$dst, $src}",
- [(set FR64:$dst, (sint_to_fp R32:$src))]>,
- Requires<[HasSSE2]>, XD;
-def CVTSI2SDrm: I<0x2A, MRMSrcMem, (ops FR64:$dst, i32mem:$src),
- "cvtsi2sd {$src, $dst|$dst, $src}",
- [(set FR64:$dst, (sint_to_fp (loadi32 addr:$src)))]>,
- Requires<[HasSSE2]>, XD;
-
-def SQRTSSrr : I<0x51, MRMSrcReg, (ops FR32:$dst, FR32:$src),
- "sqrtss {$src, $dst|$dst, $src}",
- [(set FR32:$dst, (fsqrt FR32:$src))]>,
- Requires<[HasSSE1]>, XS;
-def SQRTSSrm : I<0x51, MRMSrcMem, (ops FR32:$dst, f32mem:$src),
- "sqrtss {$src, $dst|$dst, $src}",
- [(set FR32:$dst, (fsqrt (loadf32 addr:$src)))]>,
- Requires<[HasSSE1]>, XS;
-def SQRTSDrr : I<0x51, MRMSrcReg, (ops FR64:$dst, FR64:$src),
- "sqrtsd {$src, $dst|$dst, $src}",
- [(set FR64:$dst, (fsqrt FR64:$src))]>,
- Requires<[HasSSE2]>, XD;
-def SQRTSDrm : I<0x51, MRMSrcMem, (ops FR64:$dst, f64mem:$src),
- "sqrtsd {$src, $dst|$dst, $src}",
- [(set FR64:$dst, (fsqrt (loadf64 addr:$src)))]>,
- Requires<[HasSSE2]>, XD;
-
-def UCOMISSrr: I<0x2E, MRMSrcReg, (ops FR32:$src1, FR32:$src2),
- "ucomiss {$src2, $src1|$src1, $src2}",
- [(X86cmp FR32:$src1, FR32:$src2)]>,
- Requires<[HasSSE1]>, TB;
-def UCOMISSrm: I<0x2E, MRMSrcMem, (ops FR32:$src1, f32mem:$src2),
- "ucomiss {$src2, $src1|$src1, $src2}",
- [(X86cmp FR32:$src1, (loadf32 addr:$src2))]>,
- Requires<[HasSSE1]>, TB;
-def UCOMISDrr: I<0x2E, MRMSrcReg, (ops FR64:$src1, FR64:$src2),
- "ucomisd {$src2, $src1|$src1, $src2}",
- [(X86cmp FR64:$src1, FR64:$src2)]>,
- Requires<[HasSSE2]>, TB, OpSize;
-def UCOMISDrm: I<0x2E, MRMSrcMem, (ops FR64:$src1, f64mem:$src2),
- "ucomisd {$src2, $src1|$src1, $src2}",
- [(X86cmp FR64:$src1, (loadf64 addr:$src2))]>,
- Requires<[HasSSE2]>, TB, OpSize;
-
-// Pseudo-instructions that map fld0 to pxor for sse.
-// FIXME: remove when we can teach regalloc that xor reg, reg is ok.
-def FLD0SS : I<0xEF, MRMInitReg, (ops FR32:$dst),
- "pxor $dst, $dst", [(set FR32:$dst, fp32imm0)]>,
- Requires<[HasSSE1]>, TB, OpSize;
-def FLD0SD : I<0xEF, MRMInitReg, (ops FR64:$dst),
- "pxor $dst, $dst", [(set FR64:$dst, fp64imm0)]>,
- Requires<[HasSSE2]>, TB, OpSize;
-
-let isTwoAddress = 1 in {
-// SSE Scalar Arithmetic
-let isCommutable = 1 in {
-def ADDSSrr : I<0x58, MRMSrcReg, (ops FR32:$dst, FR32:$src1, FR32:$src2),
- "addss {$src2, $dst|$dst, $src2}",
- [(set FR32:$dst, (fadd FR32:$src1, FR32:$src2))]>,
- Requires<[HasSSE1]>, XS;
-def ADDSDrr : I<0x58, MRMSrcReg, (ops FR64:$dst, FR64:$src1, FR64:$src2),
- "addsd {$src2, $dst|$dst, $src2}",
- [(set FR64:$dst, (fadd FR64:$src1, FR64:$src2))]>,
- Requires<[HasSSE2]>, XD;
-def MULSSrr : I<0x59, MRMSrcReg, (ops FR32:$dst, FR32:$src1, FR32:$src2),
- "mulss {$src2, $dst|$dst, $src2}",
- [(set FR32:$dst, (fmul FR32:$src1, FR32:$src2))]>,
- Requires<[HasSSE1]>, XS;
-def MULSDrr : I<0x59, MRMSrcReg, (ops FR64:$dst, FR64:$src1, FR64:$src2),
- "mulsd {$src2, $dst|$dst, $src2}",
- [(set FR64:$dst, (fmul FR64:$src1, FR64:$src2))]>,
- Requires<[HasSSE2]>, XD;
-}
-
-def ADDSSrm : I<0x58, MRMSrcMem, (ops FR32:$dst, FR32:$src1, f32mem:$src2),
- "addss {$src2, $dst|$dst, $src2}",
- [(set FR32:$dst, (fadd FR32:$src1, (loadf32 addr:$src2)))]>,
- Requires<[HasSSE1]>, XS;
-def ADDSDrm : I<0x58, MRMSrcMem, (ops FR64:$dst, FR64:$src1, f64mem:$src2),
- "addsd {$src2, $dst|$dst, $src2}",
- [(set FR64:$dst, (fadd FR64:$src1, (loadf64 addr:$src2)))]>,
- Requires<[HasSSE2]>, XD;
-def MULSSrm : I<0x59, MRMSrcMem, (ops FR32:$dst, FR32:$src1, f32mem:$src2),
- "mulss {$src2, $dst|$dst, $src2}",
- [(set FR32:$dst, (fmul FR32:$src1, (loadf32 addr:$src2)))]>,
- Requires<[HasSSE1]>, XS;
-def MULSDrm : I<0x59, MRMSrcMem, (ops FR64:$dst, FR64:$src1, f64mem:$src2),
- "mulsd {$src2, $dst|$dst, $src2}",
- [(set FR64:$dst, (fmul FR64:$src1, (loadf64 addr:$src2)))]>,
- Requires<[HasSSE2]>, XD;
-
-def DIVSSrr : I<0x5E, MRMSrcReg, (ops FR32:$dst, FR32:$src1, FR32:$src2),
- "divss {$src2, $dst|$dst, $src2}",
- [(set FR32:$dst, (fdiv FR32:$src1, FR32:$src2))]>,
- Requires<[HasSSE1]>, XS;
-def DIVSSrm : I<0x5E, MRMSrcMem, (ops FR32:$dst, FR32:$src1, f32mem:$src2),
- "divss {$src2, $dst|$dst, $src2}",
- [(set FR32:$dst, (fdiv FR32:$src1, (loadf32 addr:$src2)))]>,
- Requires<[HasSSE1]>, XS;
-def DIVSDrr : I<0x5E, MRMSrcReg, (ops FR64:$dst, FR64:$src1, FR64:$src2),
- "divsd {$src2, $dst|$dst, $src2}",
- [(set FR64:$dst, (fdiv FR64:$src1, FR64:$src2))]>,
- Requires<[HasSSE2]>, XD;
-def DIVSDrm : I<0x5E, MRMSrcMem, (ops FR64:$dst, FR64:$src1, f64mem:$src2),
- "divsd {$src2, $dst|$dst, $src2}",
- [(set FR64:$dst, (fdiv FR64:$src1, (loadf64 addr:$src2)))]>,
- Requires<[HasSSE2]>, XD;
-
-def SUBSSrr : I<0x5C, MRMSrcReg, (ops FR32:$dst, FR32:$src1, FR32:$src2),
- "subss {$src2, $dst|$dst, $src2}",
- [(set FR32:$dst, (fsub FR32:$src1, FR32:$src2))]>,
- Requires<[HasSSE1]>, XS;
-def SUBSSrm : I<0x5C, MRMSrcMem, (ops FR32:$dst, FR32:$src1, f32mem:$src2),
- "subss {$src2, $dst|$dst, $src2}",
- [(set FR32:$dst, (fsub FR32:$src1, (loadf32 addr:$src2)))]>,
- Requires<[HasSSE1]>, XS;
-def SUBSDrr : I<0x5C, MRMSrcReg, (ops FR64:$dst, FR64:$src1, FR64:$src2),
- "subsd {$src2, $dst|$dst, $src2}",
- [(set FR64:$dst, (fsub FR64:$src1, FR64:$src2))]>,
- Requires<[HasSSE2]>, XD;
-def SUBSDrm : I<0x5C, MRMSrcMem, (ops FR64:$dst, FR64:$src1, f64mem:$src2),
- "subsd {$src2, $dst|$dst, $src2}",
- [(set FR64:$dst, (fsub FR64:$src1, (loadf64 addr:$src2)))]>,
- Requires<[HasSSE2]>, XD;
-
-// SSE compare
-def CMPSSrr : I<0xC2, MRMSrcReg,
- (ops FR32:$dst, FR32:$src1, FR32:$src, SSECC:$cc),
- "cmp${cc}ss {$src, $dst|$dst, $src}", []>,
- Requires<[HasSSE1]>, XS;
-def CMPSSrm : I<0xC2, MRMSrcMem,
- (ops FR32:$dst, FR32:$src1, f32mem:$src, SSECC:$cc),
- "cmp${cc}ss {$src, $dst|$dst, $src}", []>,
- Requires<[HasSSE1]>, XS;
-def CMPSDrr : I<0xC2, MRMSrcReg,
- (ops FR64:$dst, FR64:$src1, FR64:$src, SSECC:$cc),
- "cmp${cc}sd {$src, $dst|$dst, $src}", []>,
- Requires<[HasSSE1]>, XD;
-def CMPSDrm : I<0xC2, MRMSrcMem,
- (ops FR64:$dst, FR64:$src1, f64mem:$src, SSECC:$cc),
- "cmp${cc}sd {$src, $dst|$dst, $src}", []>,
- Requires<[HasSSE2]>, XD;
-
-// SSE Logical - these all operate on packed values
-let isCommutable = 1 in {
-def ANDPSrr : I<0x54, MRMSrcReg, (ops FR32:$dst, FR32:$src1, FR32:$src2),
- "andps {$src2, $dst|$dst, $src2}",
- [(set FR32:$dst, (X86fand FR32:$src1, FR32:$src2))]>,
- Requires<[HasSSE1]>, TB;
-def ANDPDrr : I<0x54, MRMSrcReg, (ops FR64:$dst, FR64:$src1, FR64:$src2),
- "andpd {$src2, $dst|$dst, $src2}",
- [(set FR64:$dst, (X86fand FR64:$src1, FR64:$src2))]>,
- Requires<[HasSSE2]>, TB, OpSize;
-def ORPSrr : I<0x56, MRMSrcReg, (ops FR32:$dst, FR32:$src1, FR32:$src2),
- "orps {$src2, $dst|$dst, $src2}", []>,
- Requires<[HasSSE1]>, TB;
-def ORPDrr : I<0x56, MRMSrcReg, (ops FR64:$dst, FR64:$src1, FR64:$src2),
- "orpd {$src2, $dst|$dst, $src2}", []>,
- Requires<[HasSSE2]>, TB, OpSize;
-def XORPSrr : I<0x57, MRMSrcReg, (ops FR32:$dst, FR32:$src1, FR32:$src2),
- "xorps {$src2, $dst|$dst, $src2}",
- [(set FR32:$dst, (X86fxor FR32:$src1, FR32:$src2))]>,
- Requires<[HasSSE1]>, TB;
-def XORPDrr : I<0x57, MRMSrcReg, (ops FR64:$dst, FR64:$src1, FR64:$src2),
- "xorpd {$src2, $dst|$dst, $src2}",
- [(set FR64:$dst, (X86fxor FR64:$src1, FR64:$src2))]>,
- Requires<[HasSSE2]>, TB, OpSize;
-}
-def ANDPSrm : I<0x54, MRMSrcMem, (ops FR32:$dst, FR32:$src1, f128mem:$src2),
- "andps {$src2, $dst|$dst, $src2}",
- [(set FR32:$dst, (X86fand FR32:$src1,
- (X86loadpf32 addr:$src2)))]>,
- Requires<[HasSSE1]>, TB;
-def ANDPDrm : I<0x54, MRMSrcMem, (ops FR64:$dst, FR64:$src1, f128mem:$src2),
- "andpd {$src2, $dst|$dst, $src2}",
- [(set FR64:$dst, (X86fand FR64:$src1,
- (X86loadpf64 addr:$src2)))]>,
- Requires<[HasSSE2]>, TB, OpSize;
-def ORPSrm : I<0x56, MRMSrcMem, (ops FR32:$dst, FR32:$src1, f128mem:$src2),
- "orps {$src2, $dst|$dst, $src2}", []>,
- Requires<[HasSSE1]>, TB;
-def ORPDrm : I<0x56, MRMSrcMem, (ops FR64:$dst, FR64:$src1, f128mem:$src2),
- "orpd {$src2, $dst|$dst, $src2}", []>,
- Requires<[HasSSE2]>, TB, OpSize;
-def XORPSrm : I<0x57, MRMSrcMem, (ops FR32:$dst, FR32:$src1, f128mem:$src2),
- "xorps {$src2, $dst|$dst, $src2}",
- [(set FR32:$dst, (X86fxor FR32:$src1,
- (X86loadpf32 addr:$src2)))]>,
- Requires<[HasSSE1]>, TB;
-def XORPDrm : I<0x57, MRMSrcMem, (ops FR64:$dst, FR64:$src1, f128mem:$src2),
- "xorpd {$src2, $dst|$dst, $src2}",
- [(set FR64:$dst, (X86fxor FR64:$src1,
- (X86loadpf64 addr:$src2)))]>,
- Requires<[HasSSE2]>, TB, OpSize;
-
-def ANDNPSrr : I<0x55, MRMSrcReg, (ops FR32:$dst, FR32:$src1, FR32:$src2),
- "andnps {$src2, $dst|$dst, $src2}", []>,
- Requires<[HasSSE1]>, TB;
-def ANDNPSrm : I<0x55, MRMSrcMem, (ops FR32:$dst, FR32:$src1, f128mem:$src2),
- "andnps {$src2, $dst|$dst, $src2}", []>,
- Requires<[HasSSE1]>, TB;
-def ANDNPDrr : I<0x55, MRMSrcReg, (ops FR64:$dst, FR64:$src1, FR64:$src2),
- "andnpd {$src2, $dst|$dst, $src2}", []>,
- Requires<[HasSSE2]>, TB, OpSize;
-def ANDNPDrm : I<0x55, MRMSrcMem, (ops FR64:$dst, FR64:$src1, f128mem:$src2),
- "andnpd {$src2, $dst|$dst, $src2}", []>,
- Requires<[HasSSE2]>, TB, OpSize;
-}
-
-//===----------------------------------------------------------------------===//
-// Floating Point Stack Support
+// Miscellaneous Instructions
//===----------------------------------------------------------------------===//
-// Floating point support. All FP Stack operations are represented with two
-// instructions here. The first instruction, generated by the instruction
-// selector, uses "RFP" registers: a traditional register file to reference
-// floating point values. These instructions are all psuedo instructions and
-// use the "Fp" prefix. The second instruction is defined with FPI, which is
-// the actual instruction emitted by the assembler. The FP stackifier pass
-// converts one to the other after register allocation occurs.
-//
-// Note that the FpI instruction should have instruction selection info (e.g.
-// a pattern) and the FPI instruction should have emission info (e.g. opcode
-// encoding and asm printing info).
-
-// FPI - Floating Point Instruction template.
-class FPI<bits<8> o, Format F, dag ops, string asm> : I<o, F, ops, asm, []> {}
-
-// FpI_ - Floating Point Psuedo Instruction template. Not Predicated.
-class FpI_<dag ops, FPFormat fp, list<dag> pattern>
- : X86Inst<0, Pseudo, NoImm, ops, ""> {
- let FPForm = fp; let FPFormBits = FPForm.Value;
- let Pattern = pattern;
-}
-
-// Random Pseudo Instructions.
-def FpGETRESULT : FpI_<(ops RFP:$dst), SpecialFP,
- [(set RFP:$dst, X86fpget)]>; // FPR = ST(0)
-
-let noResults = 1 in
- def FpSETRESULT : FpI_<(ops RFP:$src), SpecialFP,
- [(X86fpset RFP:$src)]>, Imp<[], [ST0]>; // ST(0) = FPR
-
-// FpI - Floating Point Psuedo Instruction template. Predicated on FPStack.
-class FpI<dag ops, FPFormat fp, list<dag> pattern> :
- FpI_<ops, fp, pattern>, Requires<[FPStack]>;
-
-
-def FpMOV : FpI<(ops RFP:$dst, RFP:$src), SpecialFP, []>; // f1 = fmov f2
-
-// Arithmetic
-// Add, Sub, Mul, Div.
-def FpADD : FpI<(ops RFP:$dst, RFP:$src1, RFP:$src2), TwoArgFP,
- [(set RFP:$dst, (fadd RFP:$src1, RFP:$src2))]>;
-def FpSUB : FpI<(ops RFP:$dst, RFP:$src1, RFP:$src2), TwoArgFP,
- [(set RFP:$dst, (fsub RFP:$src1, RFP:$src2))]>;
-def FpMUL : FpI<(ops RFP:$dst, RFP:$src1, RFP:$src2), TwoArgFP,
- [(set RFP:$dst, (fmul RFP:$src1, RFP:$src2))]>;
-def FpDIV : FpI<(ops RFP:$dst, RFP:$src1, RFP:$src2), TwoArgFP,
- [(set RFP:$dst, (fdiv RFP:$src1, RFP:$src2))]>;
-
-class FPST0rInst<bits<8> o, string asm>
- : FPI<o, AddRegFrm, (ops RST:$op), asm>, D8;
-class FPrST0Inst<bits<8> o, string asm>
- : FPI<o, AddRegFrm, (ops RST:$op), asm>, DC;
-class FPrST0PInst<bits<8> o, string asm>
- : FPI<o, AddRegFrm, (ops RST:$op), asm>, DE;
-
-// Binary Ops with a memory source.
-def FpADD32m : FpI<(ops RFP:$dst, RFP:$src1, f32mem:$src2), OneArgFPRW,
- [(set RFP:$dst, (fadd RFP:$src1,
- (extloadf64f32 addr:$src2)))]>;
- // ST(0) = ST(0) + [mem32]
-def FpADD64m : FpI<(ops RFP:$dst, RFP:$src1, f64mem:$src2), OneArgFPRW,
- [(set RFP:$dst, (fadd RFP:$src1, (loadf64 addr:$src2)))]>;
- // ST(0) = ST(0) + [mem64]
-def FpMUL32m : FpI<(ops RFP:$dst, RFP:$src1, f32mem:$src2), OneArgFPRW,
- [(set RFP:$dst, (fmul RFP:$src1,
- (extloadf64f32 addr:$src2)))]>;
- // ST(0) = ST(0) * [mem32]
-def FpMUL64m : FpI<(ops RFP:$dst, RFP:$src1, f64mem:$src2), OneArgFPRW,
- [(set RFP:$dst, (fmul RFP:$src1, (loadf64 addr:$src2)))]>;
- // ST(0) = ST(0) * [mem64]
-def FpSUB32m : FpI<(ops RFP:$dst, RFP:$src1, f32mem:$src2), OneArgFPRW,
- [(set RFP:$dst, (fsub RFP:$src1,
- (extloadf64f32 addr:$src2)))]>;
- // ST(0) = ST(0) - [mem32]
-def FpSUB64m : FpI<(ops RFP:$dst, RFP:$src1, f64mem:$src2), OneArgFPRW,
- [(set RFP:$dst, (fsub RFP:$src1, (loadf64 addr:$src2)))]>;
- // ST(0) = ST(0) - [mem64]
-def FpSUBR32m : FpI<(ops RFP:$dst, RFP:$src1, f32mem:$src2), OneArgFPRW,
- [(set RFP:$dst, (fsub (extloadf64f32 addr:$src2),
- RFP:$src1))]>;
- // ST(0) = [mem32] - ST(0)
-def FpSUBR64m : FpI<(ops RFP:$dst, RFP:$src1, f64mem:$src2), OneArgFPRW,
- [(set RFP:$dst, (fsub (loadf64 addr:$src2), RFP:$src1))]>;
- // ST(0) = [mem64] - ST(0)
-def FpDIV32m : FpI<(ops RFP:$dst, RFP:$src1, f32mem:$src2), OneArgFPRW,
- [(set RFP:$dst, (fdiv RFP:$src1,
- (extloadf64f32 addr:$src2)))]>;
- // ST(0) = ST(0) / [mem32]
-def FpDIV64m : FpI<(ops RFP:$dst, RFP:$src1, f64mem:$src2), OneArgFPRW,
- [(set RFP:$dst, (fdiv RFP:$src1, (loadf64 addr:$src2)))]>;
- // ST(0) = ST(0) / [mem64]
-def FpDIVR32m : FpI<(ops RFP:$dst, RFP:$src1, f32mem:$src2), OneArgFPRW,
- [(set RFP:$dst, (fdiv (extloadf64f32 addr:$src2),
- RFP:$src1))]>;
- // ST(0) = [mem32] / ST(0)
-def FpDIVR64m : FpI<(ops RFP:$dst, RFP:$src1, f64mem:$src2), OneArgFPRW,
- [(set RFP:$dst, (fdiv (loadf64 addr:$src2), RFP:$src1))]>;
- // ST(0) = [mem64] / ST(0)
-
-
-def FADD32m : FPI<0xD8, MRM0m, (ops f32mem:$src), "fadd{s} $src">;
-def FADD64m : FPI<0xDC, MRM0m, (ops f64mem:$src), "fadd{l} $src">;
-def FMUL32m : FPI<0xD8, MRM1m, (ops f32mem:$src), "fmul{s} $src">;
-def FMUL64m : FPI<0xDC, MRM1m, (ops f64mem:$src), "fmul{l} $src">;
-def FSUB32m : FPI<0xD8, MRM4m, (ops f32mem:$src), "fsub{s} $src">;
-def FSUB64m : FPI<0xDC, MRM4m, (ops f64mem:$src), "fsub{l} $src">;
-def FSUBR32m : FPI<0xD8, MRM5m, (ops f32mem:$src), "fsubr{s} $src">;
-def FSUBR64m : FPI<0xDC, MRM5m, (ops f64mem:$src), "fsubr{l} $src">;
-def FDIV32m : FPI<0xD8, MRM6m, (ops f32mem:$src), "fdiv{s} $src">;
-def FDIV64m : FPI<0xDC, MRM6m, (ops f64mem:$src), "fdiv{l} $src">;
-def FDIVR32m : FPI<0xD8, MRM7m, (ops f32mem:$src), "fdivr{s} $src">;
-def FDIVR64m : FPI<0xDC, MRM7m, (ops f64mem:$src), "fdivr{l} $src">;
-
-def FpIADD16m : FpI<(ops RFP:$dst, RFP:$src1, i16mem:$src2), OneArgFPRW,
- [(set RFP:$dst, (fadd RFP:$src1,
- (X86fild addr:$src2, i16)))]>;
- // ST(0) = ST(0) + [mem16int]
-def FpIADD32m : FpI<(ops RFP:$dst, RFP:$src1, i32mem:$src2), OneArgFPRW,
- [(set RFP:$dst, (fadd RFP:$src1,
- (X86fild addr:$src2, i32)))]>;
- // ST(0) = ST(0) + [mem32int]
-def FpIMUL16m : FpI<(ops RFP:$dst, RFP:$src1, i16mem:$src2), OneArgFPRW,
- [(set RFP:$dst, (fmul RFP:$src1,
- (X86fild addr:$src2, i16)))]>;
- // ST(0) = ST(0) * [mem16int]
-def FpIMUL32m : FpI<(ops RFP:$dst, RFP:$src1, i32mem:$src2), OneArgFPRW,
- [(set RFP:$dst, (fmul RFP:$src1,
- (X86fild addr:$src2, i32)))]>;
- // ST(0) = ST(0) * [mem32int]
-def FpISUB16m : FpI<(ops RFP:$dst, RFP:$src1, i16mem:$src2), OneArgFPRW,
- [(set RFP:$dst, (fsub RFP:$src1,
- (X86fild addr:$src2, i16)))]>;
- // ST(0) = ST(0) - [mem16int]
-def FpISUB32m : FpI<(ops RFP:$dst, RFP:$src1, i32mem:$src2), OneArgFPRW,
- [(set RFP:$dst, (fsub RFP:$src1,
- (X86fild addr:$src2, i32)))]>;
- // ST(0) = ST(0) - [mem32int]
-def FpISUBR16m : FpI<(ops RFP:$dst, RFP:$src1, i16mem:$src2), OneArgFPRW,
- [(set RFP:$dst, (fsub (X86fild addr:$src2, i16),
- RFP:$src1))]>;
- // ST(0) = [mem16int] - ST(0)
-def FpISUBR32m : FpI<(ops RFP:$dst, RFP:$src1, i32mem:$src2), OneArgFPRW,
- [(set RFP:$dst, (fsub (X86fild addr:$src2, i32),
- RFP:$src1))]>;
- // ST(0) = [mem32int] - ST(0)
-def FpIDIV16m : FpI<(ops RFP:$dst, RFP:$src1, i16mem:$src2), OneArgFPRW,
- [(set RFP:$dst, (fdiv RFP:$src1,
- (X86fild addr:$src2, i16)))]>;
- // ST(0) = ST(0) / [mem16int]
-def FpIDIV32m : FpI<(ops RFP:$dst, RFP:$src1, i32mem:$src2), OneArgFPRW,
- [(set RFP:$dst, (fdiv RFP:$src1,
- (X86fild addr:$src2, i32)))]>;
- // ST(0) = ST(0) / [mem32int]
-def FpIDIVR16m : FpI<(ops RFP:$dst, RFP:$src1, i16mem:$src2), OneArgFPRW,
- [(set RFP:$dst, (fdiv (X86fild addr:$src2, i16),
- RFP:$src1))]>;
- // ST(0) = [mem16int] / ST(0)
-def FpIDIVR32m : FpI<(ops RFP:$dst, RFP:$src1, i32mem:$src2), OneArgFPRW,
- [(set RFP:$dst, (fdiv (X86fild addr:$src2, i32),
- RFP:$src1))]>;
- // ST(0) = [mem32int] / ST(0)
-
-def FIADD16m : FPI<0xDE, MRM0m, (ops i16mem:$src), "fiadd{s} $src">;
-def FIADD32m : FPI<0xDA, MRM0m, (ops i32mem:$src), "fiadd{l} $src">;
-def FIMUL16m : FPI<0xDE, MRM1m, (ops i16mem:$src), "fimul{s} $src">;
-def FIMUL32m : FPI<0xDA, MRM1m, (ops i32mem:$src), "fimul{l} $src">;
-def FISUB16m : FPI<0xDE, MRM4m, (ops i16mem:$src), "fisub{s} $src">;
-def FISUB32m : FPI<0xDA, MRM4m, (ops i32mem:$src), "fisub{l} $src">;
-def FISUBR16m : FPI<0xDE, MRM5m, (ops i16mem:$src), "fisubr{s} $src">;
-def FISUBR32m : FPI<0xDA, MRM5m, (ops i32mem:$src), "fisubr{l} $src">;
-def FIDIV16m : FPI<0xDE, MRM6m, (ops i16mem:$src), "fidiv{s} $src">;
-def FIDIV32m : FPI<0xDA, MRM6m, (ops i32mem:$src), "fidiv{l} $src">;
-def FIDIVR16m : FPI<0xDE, MRM7m, (ops i16mem:$src), "fidivr{s} $src">;
-def FIDIVR32m : FPI<0xDA, MRM7m, (ops i32mem:$src), "fidivr{l} $src">;
-
-// NOTE: GAS and apparently all other AT&T style assemblers have a broken notion
-// of some of the 'reverse' forms of the fsub and fdiv instructions. As such,
-// we have to put some 'r's in and take them out of weird places.
-def FADDST0r : FPST0rInst <0xC0, "fadd $op">;
-def FADDrST0 : FPrST0Inst <0xC0, "fadd {%st(0), $op|$op, %ST(0)}">;
-def FADDPrST0 : FPrST0PInst<0xC0, "faddp $op">;
-def FSUBRST0r : FPST0rInst <0xE8, "fsubr $op">;
-def FSUBrST0 : FPrST0Inst <0xE8, "fsub{r} {%st(0), $op|$op, %ST(0)}">;
-def FSUBPrST0 : FPrST0PInst<0xE8, "fsub{r}p $op">;
-def FSUBST0r : FPST0rInst <0xE0, "fsub $op">;
-def FSUBRrST0 : FPrST0Inst <0xE0, "fsub{|r} {%st(0), $op|$op, %ST(0)}">;
-def FSUBRPrST0 : FPrST0PInst<0xE0, "fsub{|r}p $op">;
-def FMULST0r : FPST0rInst <0xC8, "fmul $op">;
-def FMULrST0 : FPrST0Inst <0xC8, "fmul {%st(0), $op|$op, %ST(0)}">;
-def FMULPrST0 : FPrST0PInst<0xC8, "fmulp $op">;
-def FDIVRST0r : FPST0rInst <0xF8, "fdivr $op">;
-def FDIVrST0 : FPrST0Inst <0xF8, "fdiv{r} {%st(0), $op|$op, %ST(0)}">;
-def FDIVPrST0 : FPrST0PInst<0xF8, "fdiv{r}p $op">;
-def FDIVST0r : FPST0rInst <0xF0, "fdiv $op">;
-def FDIVRrST0 : FPrST0Inst <0xF0, "fdiv{|r} {%st(0), $op|$op, %ST(0)}">;
-def FDIVRPrST0 : FPrST0PInst<0xF0, "fdiv{|r}p $op">;
-
-
-// Unary operations.
-def FpCHS : FpI<(ops RFP:$dst, RFP:$src), OneArgFPRW,
- [(set RFP:$dst, (fneg RFP:$src))]>;
-def FpABS : FpI<(ops RFP:$dst, RFP:$src), OneArgFPRW,
- [(set RFP:$dst, (fabs RFP:$src))]>;
-def FpSQRT : FpI<(ops RFP:$dst, RFP:$src), OneArgFPRW,
- [(set RFP:$dst, (fsqrt RFP:$src))]>;
-def FpSIN : FpI<(ops RFP:$dst, RFP:$src), OneArgFPRW,
- [(set RFP:$dst, (fsin RFP:$src))]>;
-def FpCOS : FpI<(ops RFP:$dst, RFP:$src), OneArgFPRW,
- [(set RFP:$dst, (fcos RFP:$src))]>;
-def FpTST : FpI<(ops RFP:$src), OneArgFP,
- []>;
-
-def FCHS : FPI<0xE0, RawFrm, (ops), "fchs">, D9;
-def FABS : FPI<0xE1, RawFrm, (ops), "fabs">, D9;
-def FSQRT : FPI<0xFA, RawFrm, (ops), "fsqrt">, D9;
-def FSIN : FPI<0xFE, RawFrm, (ops), "fsin">, D9;
-def FCOS : FPI<0xFF, RawFrm, (ops), "fcos">, D9;
-def FTST : FPI<0xE4, RawFrm, (ops), "ftst">, D9;
-
-
-// Floating point cmovs.
-let isTwoAddress = 1 in {
- def FpCMOVB : FpI<(ops RFP:$dst, RFP:$src1, RFP:$src2), CondMovFP,
- [(set RFP:$dst, (X86cmov RFP:$src1, RFP:$src2,
- X86_COND_B))]>;
- def FpCMOVBE : FpI<(ops RFP:$dst, RFP:$src1, RFP:$src2), CondMovFP,
- [(set RFP:$dst, (X86cmov RFP:$src1, RFP:$src2,
- X86_COND_BE))]>;
- def FpCMOVE : FpI<(ops RFP:$dst, RFP:$src1, RFP:$src2), CondMovFP,
- [(set RFP:$dst, (X86cmov RFP:$src1, RFP:$src2,
- X86_COND_E))]>;
- def FpCMOVP : FpI<(ops RFP:$dst, RFP:$src1, RFP:$src2), CondMovFP,
- [(set RFP:$dst, (X86cmov RFP:$src1, RFP:$src2,
- X86_COND_P))]>;
- def FpCMOVNB : FpI<(ops RFP:$dst, RFP:$src1, RFP:$src2), CondMovFP,
- [(set RFP:$dst, (X86cmov RFP:$src1, RFP:$src2,
- X86_COND_AE))]>;
- def FpCMOVNBE: FpI<(ops RFP:$dst, RFP:$src1, RFP:$src2), CondMovFP,
- [(set RFP:$dst, (X86cmov RFP:$src1, RFP:$src2,
- X86_COND_A))]>;
- def FpCMOVNE : FpI<(ops RFP:$dst, RFP:$src1, RFP:$src2), CondMovFP,
- [(set RFP:$dst, (X86cmov RFP:$src1, RFP:$src2,
- X86_COND_NE))]>;
- def FpCMOVNP : FpI<(ops RFP:$dst, RFP:$src1, RFP:$src2), CondMovFP,
- [(set RFP:$dst, (X86cmov RFP:$src1, RFP:$src2,
- X86_COND_NP))]>;
-}
-
-def FCMOVB : FPI<0xC0, AddRegFrm, (ops RST:$op),
- "fcmovb {$op, %st(0)|%ST(0), $op}">, DA;
-def FCMOVBE : FPI<0xD0, AddRegFrm, (ops RST:$op),
- "fcmovbe {$op, %st(0)|%ST(0), $op}">, DA;
-def FCMOVE : FPI<0xC8, AddRegFrm, (ops RST:$op),
- "fcmove {$op, %st(0)|%ST(0), $op}">, DA;
-def FCMOVP : FPI<0xD8, AddRegFrm, (ops RST:$op),
- "fcmovu {$op, %st(0)|%ST(0), $op}">, DA;
-def FCMOVNB : FPI<0xC0, AddRegFrm, (ops RST:$op),
- "fcmovnb {$op, %st(0)|%ST(0), $op}">, DB;
-def FCMOVNBE : FPI<0xD0, AddRegFrm, (ops RST:$op),
- "fcmovnbe {$op, %st(0)|%ST(0), $op}">, DB;
-def FCMOVNE : FPI<0xC8, AddRegFrm, (ops RST:$op),
- "fcmovne {$op, %st(0)|%ST(0), $op}">, DB;
-def FCMOVNP : FPI<0xD8, AddRegFrm, (ops RST:$op),
- "fcmovnu {$op, %st(0)|%ST(0), $op}">, DB;
-
-// Floating point loads & stores.
-def FpLD32m : FpI<(ops RFP:$dst, f32mem:$src), ZeroArgFP,
- [(set RFP:$dst, (extloadf64f32 addr:$src))]>;
-def FpLD64m : FpI<(ops RFP:$dst, f64mem:$src), ZeroArgFP,
- [(set RFP:$dst, (loadf64 addr:$src))]>;
-def FpILD16m : FpI<(ops RFP:$dst, i16mem:$src), ZeroArgFP,
- [(set RFP:$dst, (X86fild addr:$src, i16))]>;
-def FpILD32m : FpI<(ops RFP:$dst, i32mem:$src), ZeroArgFP,
- [(set RFP:$dst, (X86fild addr:$src, i32))]>;
-def FpILD64m : FpI<(ops RFP:$dst, i64mem:$src), ZeroArgFP,
- [(set RFP:$dst, (X86fild addr:$src, i64))]>;
-
-def FpST32m : FpI<(ops f32mem:$op, RFP:$src), OneArgFP,
- [(truncstore RFP:$src, addr:$op, f32)]>;
-def FpST64m : FpI<(ops f64mem:$op, RFP:$src), OneArgFP,
- [(store RFP:$src, addr:$op)]>;
-
-def FpSTP32m : FpI<(ops f32mem:$op, RFP:$src), OneArgFP, []>;
-def FpSTP64m : FpI<(ops f64mem:$op, RFP:$src), OneArgFP, []>;
-def FpIST16m : FpI<(ops i16mem:$op, RFP:$src), OneArgFP, []>;
-def FpIST32m : FpI<(ops i32mem:$op, RFP:$src), OneArgFP, []>;
-def FpIST64m : FpI<(ops i64mem:$op, RFP:$src), OneArgFP, []>;
-
-def FLD32m : FPI<0xD9, MRM0m, (ops f32mem:$src), "fld{s} $src">;
-def FLD64m : FPI<0xDD, MRM0m, (ops f64mem:$src), "fld{l} $src">;
-def FILD16m : FPI<0xDF, MRM0m, (ops i16mem:$src), "fild{s} $src">;
-def FILD32m : FPI<0xDB, MRM0m, (ops i32mem:$src), "fild{l} $src">;
-def FILD64m : FPI<0xDF, MRM5m, (ops i64mem:$src), "fild{ll} $src">;
-def FST32m : FPI<0xD9, MRM2m, (ops f32mem:$dst), "fst{s} $dst">;
-def FST64m : FPI<0xDD, MRM2m, (ops f64mem:$dst), "fst{l} $dst">;
-def FSTP32m : FPI<0xD9, MRM3m, (ops f32mem:$dst), "fstp{s} $dst">;
-def FSTP64m : FPI<0xDD, MRM3m, (ops f64mem:$dst), "fstp{l} $dst">;
-def FIST16m : FPI<0xDF, MRM2m, (ops i16mem:$dst), "fist{s} $dst">;
-def FIST32m : FPI<0xDB, MRM2m, (ops i32mem:$dst), "fist{l} $dst">;
-def FISTP16m : FPI<0xDF, MRM3m, (ops i16mem:$dst), "fistp{s} $dst">;
-def FISTP32m : FPI<0xDB, MRM3m, (ops i32mem:$dst), "fistp{l} $dst">;
-def FISTP64m : FPI<0xDF, MRM7m, (ops i64mem:$dst), "fistp{ll} $dst">;
-
-// FP Stack manipulation instructions.
-def FLDrr : FPI<0xC0, AddRegFrm, (ops RST:$op), "fld $op">, D9;
-def FSTrr : FPI<0xD0, AddRegFrm, (ops RST:$op), "fst $op">, DD;
-def FSTPrr : FPI<0xD8, AddRegFrm, (ops RST:$op), "fstp $op">, DD;
-def FXCH : FPI<0xC8, AddRegFrm, (ops RST:$op), "fxch $op">, D9;
-
-// Floating point constant loads.
-def FpLD0 : FpI<(ops RFP:$dst), ZeroArgFP,
- [(set RFP:$dst, fp64imm0)]>;
-def FpLD1 : FpI<(ops RFP:$dst), ZeroArgFP,
- [(set RFP:$dst, fp64imm1)]>;
-
-def FLD0 : FPI<0xEE, RawFrm, (ops), "fldz">, D9;
-def FLD1 : FPI<0xE8, RawFrm, (ops), "fld1">, D9;
-
-
-// Floating point compares.
-def FpUCOMr : FpI<(ops RFP:$lhs, RFP:$rhs), CompareFP,
- []>; // FPSW = cmp ST(0) with ST(i)
-def FpUCOMIr : FpI<(ops RFP:$lhs, RFP:$rhs), CompareFP,
- [(X86cmp RFP:$lhs, RFP:$rhs)]>; // CC = cmp ST(0) with ST(i)
-
-def FUCOMr : FPI<0xE0, AddRegFrm, // FPSW = cmp ST(0) with ST(i)
- (ops RST:$reg),
- "fucom $reg">, DD, Imp<[ST0],[]>;
-def FUCOMPr : FPI<0xE8, AddRegFrm, // FPSW = cmp ST(0) with ST(i), pop
- (ops RST:$reg),
- "fucomp $reg">, DD, Imp<[ST0],[]>;
-def FUCOMPPr : FPI<0xE9, RawFrm, // cmp ST(0) with ST(1), pop, pop
- (ops),
- "fucompp">, DA, Imp<[ST0],[]>;
-
-def FUCOMIr : FPI<0xE8, AddRegFrm, // CC = cmp ST(0) with ST(i)
- (ops RST:$reg),
- "fucomi {$reg, %st(0)|%ST(0), $reg}">, DB, Imp<[ST0],[]>;
-def FUCOMIPr : FPI<0xE8, AddRegFrm, // CC = cmp ST(0) with ST(i), pop
- (ops RST:$reg),
- "fucomip {$reg, %st(0)|%ST(0), $reg}">, DF, Imp<[ST0],[]>;
-
-
-// Floating point flag ops.
-def FNSTSW8r : I<0xE0, RawFrm, // AX = fp flags
- (ops), "fnstsw", []>, DF, Imp<[],[AX]>;
-
-def FNSTCW16m : I<0xD9, MRM7m, // [mem16] = X87 control world
- (ops i16mem:$dst), "fnstcw $dst", []>;
-def FLDCW16m : I<0xD9, MRM5m, // X87 control world = [mem16]
- (ops i16mem:$dst), "fldcw $dst", []>;
-
+def RDTSC : I<0x31, RawFrm, (ops), "rdtsc", [(X86rdtsc)]>,
+ TB, Imp<[],[EAX,EDX]>;
//===----------------------------------------------------------------------===//
-// XMM Packed Floating point support (requires SSE / SSE2)
+// Alias Instructions
//===----------------------------------------------------------------------===//
-def MOVAPSrr : I<0x28, MRMSrcReg, (ops V4F4:$dst, V4F4:$src),
- "movaps {$src, $dst|$dst, $src}", []>,
- Requires<[HasSSE1]>, TB;
-def MOVAPDrr : I<0x28, MRMSrcReg, (ops V2F8:$dst, V2F8:$src),
- "movapd {$src, $dst|$dst, $src}", []>,
- Requires<[HasSSE2]>, TB, OpSize;
-
-def MOVAPSrm : I<0x28, MRMSrcMem, (ops V4F4:$dst, f128mem:$src),
- "movaps {$src, $dst|$dst, $src}", []>,
- Requires<[HasSSE1]>, TB;
-def MOVAPSmr : I<0x29, MRMDestMem, (ops f128mem:$dst, V4F4:$src),
- "movaps {$src, $dst|$dst, $src}",[]>,
- Requires<[HasSSE1]>, TB;
-def MOVAPDrm : I<0x28, MRMSrcMem, (ops V2F8:$dst, f128mem:$src),
- "movapd {$src, $dst|$dst, $src}", []>,
- Requires<[HasSSE1]>, TB, OpSize;
-def MOVAPDmr : I<0x29, MRMDestMem, (ops f128mem:$dst, V2F8:$src),
- "movapd {$src, $dst|$dst, $src}",[]>,
- Requires<[HasSSE2]>, TB, OpSize;
-
-// Alias instructions to do FR32 / FR64 reg-to-reg copy using movaps / movapd.
-// Upper bits are disregarded.
-def FsMOVAPSrr : I<0x28, MRMSrcReg, (ops V4F4:$dst, V4F4:$src),
- "movaps {$src, $dst|$dst, $src}", []>,
- Requires<[HasSSE1]>, TB;
-def FsMOVAPDrr : I<0x28, MRMSrcReg, (ops V2F8:$dst, V2F8:$src),
- "movapd {$src, $dst|$dst, $src}", []>,
- Requires<[HasSSE2]>, TB, OpSize;
-
-// Alias instructions to load FR32 / FR64 from f128mem using movaps / movapd.
-// Upper bits are disregarded.
-def FsMOVAPSrm : I<0x28, MRMSrcMem, (ops FR32:$dst, f128mem:$src),
- "movaps {$src, $dst|$dst, $src}",
- [(set FR32:$dst, (X86loadpf32 addr:$src))]>,
- Requires<[HasSSE1]>, TB;
-def FsMOVAPDrm : I<0x28, MRMSrcMem, (ops FR64:$dst, f128mem:$src),
- "movapd {$src, $dst|$dst, $src}",
- [(set FR64:$dst, (X86loadpf64 addr:$src))]>,
- Requires<[HasSSE2]>, TB, OpSize;
-
+// Alias instructions that map movr0 to xor.
+// FIXME: remove when we can teach regalloc that xor reg, reg is ok.
+def MOV8r0 : I<0x30, MRMInitReg, (ops R8 :$dst),
+ "xor{b} $dst, $dst",
+ [(set R8:$dst, 0)]>;
+def MOV16r0 : I<0x31, MRMInitReg, (ops R16:$dst),
+ "xor{w} $dst, $dst",
+ [(set R16:$dst, 0)]>, OpSize;
+def MOV32r0 : I<0x31, MRMInitReg, (ops R32:$dst),
+ "xor{l} $dst, $dst",
+ [(set R32:$dst, 0)]>;
//===----------------------------------------------------------------------===//
-// Miscellaneous Instructions
-//===----------------------------------------------------------------------===//
+// DWARF Pseudo Instructions
+//
-def RDTSC : I<0x31, RawFrm, (ops), "rdtsc", [(X86rdtsc)]>,
- TB, Imp<[],[EAX,EDX]>;
+def DWARF_LOC : I<0, Pseudo, (ops i32imm:$line, i32imm:$col, i32imm:$file),
+ "; .loc $file, $line, $col",
+ [(dwarf_loc (i32 imm:$line), (i32 imm:$col),
+ (i32 imm:$file))]>;
+def DWARF_LABEL : I<0, Pseudo, (ops i32imm:$id),
+ "\nLdebug_loc${id:debug}:",
+ [(dwarf_label (i32 imm:$id))]>;
//===----------------------------------------------------------------------===//
// Non-Instruction Patterns
//===----------------------------------------------------------------------===//
-// GlobalAddress and ExternalSymbol
-def : Pat<(i32 globaladdr:$dst), (MOV32ri tglobaladdr:$dst)>;
-def : Pat<(i32 externalsym:$dst), (MOV32ri texternalsym:$dst)>;
+// ConstantPool GlobalAddress, ExternalSymbol
+def : Pat<(i32 (X86Wrapper tconstpool :$dst)), (MOV32ri tconstpool :$dst)>;
+def : Pat<(i32 (X86Wrapper tglobaladdr :$dst)), (MOV32ri tglobaladdr :$dst)>;
+def : Pat<(i32 (X86Wrapper texternalsym:$dst)), (MOV32ri texternalsym:$dst)>;
+
+def : Pat<(add R32:$src1, (X86Wrapper tconstpool:$src2)),
+ (ADD32ri R32:$src1, tconstpool:$src2)>;
+def : Pat<(add R32:$src1, (X86Wrapper tglobaladdr :$src2)),
+ (ADD32ri R32:$src1, tglobaladdr:$src2)>;
+def : Pat<(add R32:$src1, (X86Wrapper texternalsym:$src2)),
+ (ADD32ri R32:$src1, texternalsym:$src2)>;
+
+def : Pat<(store (X86Wrapper tconstpool:$src), addr:$dst),
+ (MOV32mi addr:$dst, tconstpool:$src)>;
+def : Pat<(store (X86Wrapper tglobaladdr:$src), addr:$dst),
+ (MOV32mi addr:$dst, tglobaladdr:$src)>;
+def : Pat<(store (X86Wrapper texternalsym:$src), addr:$dst),
+ (MOV32mi addr:$dst, texternalsym:$src)>;
// Calls
def : Pat<(X86call tglobaladdr:$dst),
def : Pat<(i32 (anyext R8 :$src)), (MOVZX32rr8 R8 :$src)>;
def : Pat<(i32 (anyext R16:$src)), (MOVZX32rr16 R16:$src)>;
-// Required for RET of f32 / f64 values.
-def : Pat<(X86fld addr:$src, f32), (FpLD32m addr:$src)>;
-def : Pat<(X86fld addr:$src, f64), (FpLD64m addr:$src)>;
-
-// Required for CALL which return f32 / f64 values.
-def : Pat<(X86fst RFP:$src, addr:$op, f32), (FpST32m addr:$op, RFP:$src)>;
-def : Pat<(X86fst RFP:$src, addr:$op, f64), (FpST64m addr:$op, RFP:$src)>;
-
-// Floating point constant -0.0 and -1.0
-def : Pat<(f64 fp64immneg0), (FpCHS (FpLD0))>, Requires<[FPStack]>;
-def : Pat<(f64 fp64immneg1), (FpCHS (FpLD1))>, Requires<[FPStack]>;
-
-// Used to conv. i64 to f64 since there isn't a SSE version.
-def : Pat<(X86fildflag addr:$src, i64), (FpILD64m addr:$src)>;
-
//===----------------------------------------------------------------------===//
// Some peepholes
//===----------------------------------------------------------------------===//
def : Pat<(store (or (shl (loadi16 addr:$dst), CL:$amt),
(srl R16:$src2, (sub 16, CL:$amt))), addr:$dst),
(SHLD16mrCL addr:$dst, R16:$src2)>;
+
+
+//===----------------------------------------------------------------------===//
+// Floating Point Stack Support
+//===----------------------------------------------------------------------===//
+
+include "X86InstrFPStack.td"
+
+//===----------------------------------------------------------------------===//
+// MMX and XMM Packed Integer support (requires MMX, SSE, and SSE2)
+//===----------------------------------------------------------------------===//
+
+include "X86InstrMMX.td"
+
+//===----------------------------------------------------------------------===//
+// XMM Floating point support (requires SSE / SSE2)
+//===----------------------------------------------------------------------===//
+
+include "X86InstrSSE.td"