// MMX Multiclasses
//===----------------------------------------------------------------------===//
+let Sched = WriteVecALU in {
def MMX_INTALU_ITINS : OpndItins<
IIC_MMX_ALU_RR, IIC_MMX_ALU_RM
>;
def MMX_PHADDSUBD : OpndItins<
IIC_MMX_PHADDSUBD_RR, IIC_MMX_PHADDSUBD_RM
>;
+}
+
+let Sched = WriteVecLogic in
+def MMX_INTALU_ITINS_VECLOGICSCHED : OpndItins<
+ IIC_MMX_ALU_RR, IIC_MMX_ALU_RM
+>;
+let Sched = WriteVecIMul in
def MMX_PMUL_ITINS : OpndItins<
IIC_MMX_PMUL, IIC_MMX_PMUL
>;
+let Sched = WriteVecIMul in {
def MMX_PSADBW_ITINS : OpndItins<
IIC_MMX_PSADBW, IIC_MMX_PSADBW
>;
def MMX_MISC_FUNC_ITINS : OpndItins<
IIC_MMX_MISC_FUNC_MEM, IIC_MMX_MISC_FUNC_REG
>;
+}
def MMX_SHIFT_ITINS : ShiftOpndItins<
IIC_MMX_SHIFT_RR, IIC_MMX_SHIFT_RM, IIC_MMX_SHIFT_RI
>;
+let Sched = WriteShuffle in {
def MMX_UNPCK_H_ITINS : OpndItins<
IIC_MMX_UNPCK_H_RR, IIC_MMX_UNPCK_H_RM
>;
def MMX_PSHUF_ITINS : OpndItins<
IIC_MMX_PSHUF, IIC_MMX_PSHUF
>;
+} // Sched
+let Sched = WriteCvtF2I in {
def MMX_CVT_PD_ITINS : OpndItins<
IIC_MMX_CVT_PD_RR, IIC_MMX_CVT_PD_RM
>;
def MMX_CVT_PS_ITINS : OpndItins<
IIC_MMX_CVT_PS_RR, IIC_MMX_CVT_PS_RM
>;
+}
let Constraints = "$src1 = $dst" in {
// MMXI_binop_rm_int - Simple MMX binary operator based on intrinsic.
def irr : MMXI<opc, MRMSrcReg, (outs VR64:$dst),
(ins VR64:$src1, VR64:$src2),
!strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"),
- [(set VR64:$dst, (IntId VR64:$src1, VR64:$src2))], itins.rr> {
+ [(set VR64:$dst, (IntId VR64:$src1, VR64:$src2))], itins.rr>,
+ Sched<[itins.Sched]> {
let isCommutable = Commutable;
}
def irm : MMXI<opc, MRMSrcMem, (outs VR64:$dst),
!strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"),
[(set VR64:$dst, (IntId VR64:$src1,
(bitconvert (load_mmx addr:$src2))))],
- itins.rm>;
+ itins.rm>, Sched<[itins.Sched.Folded, ReadAfterLd]>;
}
multiclass MMXI_binop_rmi_int<bits<8> opc, bits<8> opc2, Format ImmForm,
def rr : MMXI<opc, MRMSrcReg, (outs VR64:$dst),
(ins VR64:$src1, VR64:$src2),
!strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"),
- [(set VR64:$dst, (IntId VR64:$src1, VR64:$src2))], itins.rr>;
+ [(set VR64:$dst, (IntId VR64:$src1, VR64:$src2))], itins.rr>,
+ Sched<[WriteVecShift]>;
def rm : MMXI<opc, MRMSrcMem, (outs VR64:$dst),
(ins VR64:$src1, i64mem:$src2),
!strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"),
[(set VR64:$dst, (IntId VR64:$src1,
(bitconvert (load_mmx addr:$src2))))],
- itins.rm>;
+ itins.rm>, Sched<[WriteVecShiftLd, ReadAfterLd]>;
def ri : MMXIi8<opc2, ImmForm, (outs VR64:$dst),
- (ins VR64:$src1, i32i8imm:$src2),
+ (ins VR64:$src1, i32u8imm:$src2),
!strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"),
- [(set VR64:$dst, (IntId2 VR64:$src1, (i32 imm:$src2)))], itins.ri>;
+ [(set VR64:$dst, (IntId2 VR64:$src1, imm:$src2))], itins.ri>,
+ Sched<[WriteVecShift]>;
}
}
Intrinsic IntId64, OpndItins itins> {
def rr64 : MMXSS38I<opc, MRMSrcReg, (outs VR64:$dst), (ins VR64:$src),
!strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"),
- [(set VR64:$dst, (IntId64 VR64:$src))], itins.rr>;
+ [(set VR64:$dst, (IntId64 VR64:$src))], itins.rr>,
+ Sched<[itins.Sched]>;
def rm64 : MMXSS38I<opc, MRMSrcMem, (outs VR64:$dst), (ins i64mem:$src),
!strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"),
[(set VR64:$dst,
(IntId64 (bitconvert (memopmmx addr:$src))))],
- itins.rm>;
+ itins.rm>, Sched<[itins.Sched.Folded]>;
}
/// Binary MMX instructions requiring SSSE3.
def rr64 : MMXSS38I<opc, MRMSrcReg, (outs VR64:$dst),
(ins VR64:$src1, VR64:$src2),
!strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"),
- [(set VR64:$dst, (IntId64 VR64:$src1, VR64:$src2))], itins.rr>;
+ [(set VR64:$dst, (IntId64 VR64:$src1, VR64:$src2))], itins.rr>,
+ Sched<[itins.Sched]>;
def rm64 : MMXSS38I<opc, MRMSrcMem, (outs VR64:$dst),
(ins VR64:$src1, i64mem:$src2),
!strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"),
[(set VR64:$dst,
(IntId64 VR64:$src1,
- (bitconvert (memopmmx addr:$src2))))], itins.rm>;
+ (bitconvert (memopmmx addr:$src2))))], itins.rm>,
+ Sched<[itins.Sched.Folded, ReadAfterLd]>;
}
}
/// PALIGN MMX instructions (require SSSE3).
multiclass ssse3_palign_mm<string asm, Intrinsic IntId> {
def R64irr : MMXSS3AI<0x0F, MRMSrcReg, (outs VR64:$dst),
- (ins VR64:$src1, VR64:$src2, i8imm:$src3),
- !strconcat(asm, "\t{$src3, $src2, $dst|$dst, $src2, $src3}"),
- [(set VR64:$dst, (IntId VR64:$src1, VR64:$src2, (i8 imm:$src3)))]>;
+ (ins VR64:$src1, VR64:$src2, u8imm:$src3),
+ !strconcat(asm, "\t{$src3, $src2, $dst|$dst, $src2, $src3}"),
+ [(set VR64:$dst, (IntId VR64:$src1, VR64:$src2, (i8 imm:$src3)))]>,
+ Sched<[WriteShuffle]>;
def R64irm : MMXSS3AI<0x0F, MRMSrcMem, (outs VR64:$dst),
- (ins VR64:$src1, i64mem:$src2, i8imm:$src3),
+ (ins VR64:$src1, i64mem:$src2, u8imm:$src3),
!strconcat(asm, "\t{$src3, $src2, $dst|$dst, $src2, $src3}"),
[(set VR64:$dst, (IntId VR64:$src1,
- (bitconvert (load_mmx addr:$src2)), (i8 imm:$src3)))]>;
+ (bitconvert (load_mmx addr:$src2)), (i8 imm:$src3)))]>,
+ Sched<[WriteShuffleLd, ReadAfterLd]>;
}
multiclass sse12_cvt_pint<bits<8> opc, RegisterClass SrcRC, RegisterClass DstRC,
Intrinsic Int, X86MemOperand x86memop, PatFrag ld_frag,
string asm, OpndItins itins, Domain d> {
def irr : MMXPI<opc, MRMSrcReg, (outs DstRC:$dst), (ins SrcRC:$src), asm,
- [(set DstRC:$dst, (Int SrcRC:$src))], itins.rr, d>;
+ [(set DstRC:$dst, (Int SrcRC:$src))], itins.rr, d>,
+ Sched<[itins.Sched]>;
def irm : MMXPI<opc, MRMSrcMem, (outs DstRC:$dst), (ins x86memop:$src), asm,
- [(set DstRC:$dst, (Int (ld_frag addr:$src)))], itins.rm, d>;
+ [(set DstRC:$dst, (Int (ld_frag addr:$src)))], itins.rm, d>,
+ Sched<[itins.Sched.Folded]>;
}
multiclass sse12_cvt_pint_3addr<bits<8> opc, RegisterClass SrcRC,
RegisterClass DstRC, Intrinsic Int, X86MemOperand x86memop,
PatFrag ld_frag, string asm, Domain d> {
- def irr : PI<opc, MRMSrcReg, (outs DstRC:$dst),(ins DstRC:$src1, SrcRC:$src2),
- asm, [(set DstRC:$dst, (Int DstRC:$src1, SrcRC:$src2))],
- IIC_DEFAULT, d>;
- def irm : PI<opc, MRMSrcMem, (outs DstRC:$dst),
- (ins DstRC:$src1, x86memop:$src2), asm,
- [(set DstRC:$dst, (Int DstRC:$src1, (ld_frag addr:$src2)))],
- IIC_DEFAULT, d>;
+ def irr : MMXPI<opc, MRMSrcReg, (outs DstRC:$dst),
+ (ins DstRC:$src1, SrcRC:$src2), asm,
+ [(set DstRC:$dst, (Int DstRC:$src1, SrcRC:$src2))],
+ NoItinerary, d>, Sched<[WriteCvtI2F]>;
+ def irm : MMXPI<opc, MRMSrcMem, (outs DstRC:$dst),
+ (ins DstRC:$src1, x86memop:$src2), asm,
+ [(set DstRC:$dst, (Int DstRC:$src1, (ld_frag addr:$src2)))],
+ NoItinerary, d>, Sched<[WriteCvtI2FLd]>;
}
//===----------------------------------------------------------------------===//
//===----------------------------------------------------------------------===//
def MMX_EMMS : MMXI<0x77, RawFrm, (outs), (ins), "emms",
- [(int_x86_mmx_emms)]>;
+ [(int_x86_mmx_emms)], IIC_MMX_EMMS>;
//===----------------------------------------------------------------------===//
// MMX Scalar Instructions
// Data Transfer Instructions
def MMX_MOVD64rr : MMXI<0x6E, MRMSrcReg, (outs VR64:$dst), (ins GR32:$src),
"movd\t{$src, $dst|$dst, $src}",
- [(set VR64:$dst,
+ [(set VR64:$dst,
(x86mmx (scalar_to_vector GR32:$src)))],
- IIC_MMX_MOV_MM_RM>;
-let canFoldAsLoad = 1 in
+ IIC_MMX_MOV_MM_RM>, Sched<[WriteMove]>;
def MMX_MOVD64rm : MMXI<0x6E, MRMSrcMem, (outs VR64:$dst), (ins i32mem:$src),
"movd\t{$src, $dst|$dst, $src}",
[(set VR64:$dst,
(x86mmx (scalar_to_vector (loadi32 addr:$src))))],
- IIC_MMX_MOV_MM_RM>;
+ IIC_MMX_MOV_MM_RM>, Sched<[WriteLoad]>;
+
+let Predicates = [HasMMX] in {
+ let AddedComplexity = 15 in
+ def : Pat<(x86mmx (MMX_X86movw2d GR32:$src)),
+ (MMX_MOVD64rr GR32:$src)>;
+ let AddedComplexity = 20 in
+ def : Pat<(x86mmx (MMX_X86movw2d (loadi32 addr:$src))),
+ (MMX_MOVD64rm addr:$src)>;
+}
+
let mayStore = 1 in
def MMX_MOVD64mr : MMXI<0x7E, MRMDestMem, (outs), (ins i32mem:$dst, VR64:$src),
- "movd\t{$src, $dst|$dst, $src}", [], IIC_MMX_MOV_MM_RM>;
+ "movd\t{$src, $dst|$dst, $src}", [], IIC_MMX_MOV_MM_RM>,
+ Sched<[WriteStore]>;
-// Low word of MMX to GPR.
-def MMX_X86movd2w : SDNode<"X86ISD::MMX_MOVD2W", SDTypeProfile<1, 1,
- [SDTCisVT<0, i32>, SDTCisVT<1, x86mmx>]>>;
def MMX_MOVD64grr : MMXI<0x7E, MRMDestReg, (outs GR32:$dst), (ins VR64:$src),
"movd\t{$src, $dst|$dst, $src}",
[(set GR32:$dst,
- (MMX_X86movd2w (x86mmx VR64:$src)))], IIC_MMX_MOV_REG_MM>;
+ (MMX_X86movd2w (x86mmx VR64:$src)))],
+ IIC_MMX_MOV_REG_MM>, Sched<[WriteMove]>;
-let neverHasSideEffects = 1 in
+let isBitcast = 1 in
def MMX_MOVD64to64rr : MMXRI<0x6E, MRMSrcReg, (outs VR64:$dst), (ins GR64:$src),
"movd\t{$src, $dst|$dst, $src}",
- [], IIC_MMX_MOV_MM_RM>;
+ [(set VR64:$dst, (bitconvert GR64:$src))],
+ IIC_MMX_MOV_MM_RM>, Sched<[WriteMove]>;
+
+let isCodeGenOnly = 1, ForceDisassemble = 1, hasSideEffects = 0, mayLoad = 1 in
+def MMX_MOVD64to64rm : MMXRI<0x6E, MRMSrcMem, (outs VR64:$dst),
+ (ins i64mem:$src), "movd\t{$src, $dst|$dst, $src}",
+ [], IIC_MMX_MOVQ_RM>, Sched<[WriteLoad]>;
// These are 64 bit moves, but since the OS X assembler doesn't
// recognize a register-register movq, we write them as
// movd.
+let SchedRW = [WriteMove], isBitcast = 1 in {
def MMX_MOVD64from64rr : MMXRI<0x7E, MRMDestReg,
(outs GR64:$dst), (ins VR64:$src),
- "movd\t{$src, $dst|$dst, $src}",
+ "movd\t{$src, $dst|$dst, $src}",
[(set GR64:$dst,
(bitconvert VR64:$src))], IIC_MMX_MOV_REG_MM>;
-def MMX_MOVD64rrv164 : MMXRI<0x6E, MRMSrcReg, (outs VR64:$dst), (ins GR64:$src),
- "movd\t{$src, $dst|$dst, $src}",
- [(set VR64:$dst,
- (bitconvert GR64:$src))], IIC_MMX_MOV_MM_RM>;
-let neverHasSideEffects = 1 in
+let hasSideEffects = 0 in
def MMX_MOVQ64rr : MMXI<0x6F, MRMSrcReg, (outs VR64:$dst), (ins VR64:$src),
"movq\t{$src, $dst|$dst, $src}", [],
IIC_MMX_MOVQ_RR>;
+let isCodeGenOnly = 1, ForceDisassemble = 1, hasSideEffects = 0 in {
+def MMX_MOVQ64rr_REV : MMXI<0x7F, MRMDestReg, (outs VR64:$dst), (ins VR64:$src),
+ "movq\t{$src, $dst|$dst, $src}", [],
+ IIC_MMX_MOVQ_RR>;
+}
+} // SchedRW
+
+let isCodeGenOnly = 1, ForceDisassemble = 1, hasSideEffects = 0, mayStore = 1 in
+def MMX_MOVD64from64rm : MMXRI<0x7E, MRMDestMem,
+ (outs i64mem:$dst), (ins VR64:$src),
+ "movd\t{$src, $dst|$dst, $src}",
+ [], IIC_MMX_MOV_REG_MM>, Sched<[WriteStore]>;
+
+let SchedRW = [WriteLoad] in {
let canFoldAsLoad = 1 in
def MMX_MOVQ64rm : MMXI<0x6F, MRMSrcMem, (outs VR64:$dst), (ins i64mem:$src),
"movq\t{$src, $dst|$dst, $src}",
[(set VR64:$dst, (load_mmx addr:$src))],
IIC_MMX_MOVQ_RM>;
+} // SchedRW
+let SchedRW = [WriteStore] in
def MMX_MOVQ64mr : MMXI<0x7F, MRMDestMem, (outs), (ins i64mem:$dst, VR64:$src),
"movq\t{$src, $dst|$dst, $src}",
[(store (x86mmx VR64:$src), addr:$dst)],
IIC_MMX_MOVQ_RM>;
+let SchedRW = [WriteMove] in {
def MMX_MOVDQ2Qrr : MMXSDIi8<0xD6, MRMSrcReg, (outs VR64:$dst),
(ins VR128:$src), "movdq2q\t{$src, $dst|$dst, $src}",
[(set VR64:$dst,
(i64 (bitconvert (x86mmx VR64:$src))))))],
IIC_MMX_MOVQ_RR>;
-let neverHasSideEffects = 1 in
+let isCodeGenOnly = 1, hasSideEffects = 1 in {
def MMX_MOVQ2FR64rr: MMXS2SIi8<0xD6, MRMSrcReg, (outs FR64:$dst),
(ins VR64:$src), "movq2dq\t{$src, $dst|$dst, $src}",
[], IIC_MMX_MOVQ_RR>;
def MMX_MOVFR642Qrr: MMXSDIi8<0xD6, MRMSrcReg, (outs VR64:$dst),
(ins FR64:$src), "movdq2q\t{$src, $dst|$dst, $src}",
[], IIC_MMX_MOVQ_RR>;
+}
+} // SchedRW
def MMX_MOVNTQmr : MMXI<0xE7, MRMDestMem, (outs), (ins i64mem:$dst, VR64:$src),
"movntq\t{$src, $dst|$dst, $src}",
[(int_x86_mmx_movnt_dq addr:$dst, VR64:$src)],
- IIC_MMX_MOVQ_RM>;
-
-let AddedComplexity = 15 in
-// movd to MMX register zero-extends
-def MMX_MOVZDI2PDIrr : MMXI<0x6E, MRMSrcReg, (outs VR64:$dst), (ins GR32:$src),
- "movd\t{$src, $dst|$dst, $src}",
- [(set VR64:$dst,
- (x86mmx (X86vzmovl (x86mmx (scalar_to_vector GR32:$src)))))],
- IIC_MMX_MOV_MM_RM>;
-let AddedComplexity = 20 in
-def MMX_MOVZDI2PDIrm : MMXI<0x6E, MRMSrcMem, (outs VR64:$dst),
- (ins i32mem:$src),
- "movd\t{$src, $dst|$dst, $src}",
- [(set VR64:$dst,
- (x86mmx (X86vzmovl (x86mmx
- (scalar_to_vector (loadi32 addr:$src))))))],
- IIC_MMX_MOV_MM_RM>;
+ IIC_MMX_MOVQ_RM>, Sched<[WriteStore]>;
+
+let Predicates = [HasMMX] in {
+ let AddedComplexity = 15 in
+ // movd to MMX register zero-extends
+ def : Pat<(x86mmx (X86vzmovl (x86mmx (scalar_to_vector GR32:$src)))),
+ (MMX_MOVD64rr GR32:$src)>;
+ let AddedComplexity = 20 in
+ def : Pat<(x86mmx (X86vzmovl (x86mmx (scalar_to_vector (loadi32 addr:$src))))),
+ (MMX_MOVD64rm addr:$src)>;
+}
// Arithmetic Instructions
defm MMX_PABSB : SS3I_unop_rm_int_mm<0x1C, "pabsb", int_x86_ssse3_pabs_b,
defm MMX_PSUBB : MMXI_binop_rm_int<0xF8, "psubb", int_x86_mmx_psub_b,
MMX_INTALU_ITINS>;
defm MMX_PSUBW : MMXI_binop_rm_int<0xF9, "psubw", int_x86_mmx_psub_w,
- MMX_INTALU_ITINS, 1>;
+ MMX_INTALU_ITINS>;
defm MMX_PSUBD : MMXI_binop_rm_int<0xFA, "psubd", int_x86_mmx_psub_d,
- MMX_INTALU_ITINS, 1>;
+ MMX_INTALU_ITINS>;
defm MMX_PSUBQ : MMXI_binop_rm_int<0xFB, "psubq", int_x86_mmx_psub_q,
- MMX_INTALUQ_ITINS, 1>;
+ MMX_INTALUQ_ITINS>;
defm MMX_PSUBSB : MMXI_binop_rm_int<0xE8, "psubsb" , int_x86_mmx_psubs_b,
- MMX_INTALU_ITINS, 1>;
+ MMX_INTALU_ITINS>;
defm MMX_PSUBSW : MMXI_binop_rm_int<0xE9, "psubsw" , int_x86_mmx_psubs_w,
- MMX_INTALU_ITINS, 1>;
+ MMX_INTALU_ITINS>;
defm MMX_PSUBUSB : MMXI_binop_rm_int<0xD8, "psubusb", int_x86_mmx_psubus_b,
- MMX_INTALU_ITINS, 1>;
+ MMX_INTALU_ITINS>;
defm MMX_PSUBUSW : MMXI_binop_rm_int<0xD9, "psubusw", int_x86_mmx_psubus_w,
- MMX_INTALU_ITINS, 1>;
+ MMX_INTALU_ITINS>;
defm MMX_PHSUBW : SS3I_binop_rm_int_mm<0x05, "phsubw", int_x86_ssse3_phsub_w,
MMX_PHADDSUBW>;
// Logical Instructions
defm MMX_PAND : MMXI_binop_rm_int<0xDB, "pand", int_x86_mmx_pand,
- MMX_INTALU_ITINS, 1>;
+ MMX_INTALU_ITINS_VECLOGICSCHED, 1>;
defm MMX_POR : MMXI_binop_rm_int<0xEB, "por" , int_x86_mmx_por,
- MMX_INTALU_ITINS, 1>;
+ MMX_INTALU_ITINS_VECLOGICSCHED, 1>;
defm MMX_PXOR : MMXI_binop_rm_int<0xEF, "pxor", int_x86_mmx_pxor,
- MMX_INTALU_ITINS, 1>;
+ MMX_INTALU_ITINS_VECLOGICSCHED, 1>;
defm MMX_PANDN : MMXI_binop_rm_int<0xDF, "pandn", int_x86_mmx_pandn,
- MMX_INTALU_ITINS>;
+ MMX_INTALU_ITINS_VECLOGICSCHED>;
// Shift Instructions
defm MMX_PSRLW : MMXI_binop_rmi_int<0xD1, 0x71, MRM2r, "psrlw",
int_x86_mmx_psrl_q, int_x86_mmx_psrli_q,
MMX_SHIFT_ITINS>;
+def : Pat<(int_x86_mmx_psrl_w VR64:$src1, (load_mvmmx addr:$src2)),
+ (MMX_PSRLWrm VR64:$src1, addr:$src2)>;
+def : Pat<(int_x86_mmx_psrl_d VR64:$src1, (load_mvmmx addr:$src2)),
+ (MMX_PSRLDrm VR64:$src1, addr:$src2)>;
+def : Pat<(int_x86_mmx_psrl_q VR64:$src1, (load_mvmmx addr:$src2)),
+ (MMX_PSRLQrm VR64:$src1, addr:$src2)>;
+
defm MMX_PSLLW : MMXI_binop_rmi_int<0xF1, 0x71, MRM6r, "psllw",
int_x86_mmx_psll_w, int_x86_mmx_pslli_w,
MMX_SHIFT_ITINS>;
int_x86_mmx_psll_q, int_x86_mmx_pslli_q,
MMX_SHIFT_ITINS>;
+def : Pat<(int_x86_mmx_psll_w VR64:$src1, (load_mvmmx addr:$src2)),
+ (MMX_PSLLWrm VR64:$src1, addr:$src2)>;
+def : Pat<(int_x86_mmx_psll_d VR64:$src1, (load_mvmmx addr:$src2)),
+ (MMX_PSLLDrm VR64:$src1, addr:$src2)>;
+def : Pat<(int_x86_mmx_psll_q VR64:$src1, (load_mvmmx addr:$src2)),
+ (MMX_PSLLQrm VR64:$src1, addr:$src2)>;
+
defm MMX_PSRAW : MMXI_binop_rmi_int<0xE1, 0x71, MRM4r, "psraw",
int_x86_mmx_psra_w, int_x86_mmx_psrai_w,
MMX_SHIFT_ITINS>;
int_x86_mmx_psra_d, int_x86_mmx_psrai_d,
MMX_SHIFT_ITINS>;
+def : Pat<(int_x86_mmx_psra_w VR64:$src1, (load_mvmmx addr:$src2)),
+ (MMX_PSRAWrm VR64:$src1, addr:$src2)>;
+def : Pat<(int_x86_mmx_psra_d VR64:$src1, (load_mvmmx addr:$src2)),
+ (MMX_PSRADrm VR64:$src1, addr:$src2)>;
+
// Comparison Instructions
defm MMX_PCMPEQB : MMXI_binop_rm_int<0x74, "pcmpeqb", int_x86_mmx_pcmpeq_b,
MMX_INTALU_ITINS>;
MMX_INTALU_ITINS>;
// -- Unpack Instructions
-defm MMX_PUNPCKHBW : MMXI_binop_rm_int<0x68, "punpckhbw",
+defm MMX_PUNPCKHBW : MMXI_binop_rm_int<0x68, "punpckhbw",
int_x86_mmx_punpckhbw,
MMX_UNPCK_H_ITINS>;
-defm MMX_PUNPCKHWD : MMXI_binop_rm_int<0x69, "punpckhwd",
+defm MMX_PUNPCKHWD : MMXI_binop_rm_int<0x69, "punpckhwd",
int_x86_mmx_punpckhwd,
MMX_UNPCK_H_ITINS>;
-defm MMX_PUNPCKHDQ : MMXI_binop_rm_int<0x6A, "punpckhdq",
+defm MMX_PUNPCKHDQ : MMXI_binop_rm_int<0x6A, "punpckhdq",
int_x86_mmx_punpckhdq,
MMX_UNPCK_H_ITINS>;
-defm MMX_PUNPCKLBW : MMXI_binop_rm_int<0x60, "punpcklbw",
+defm MMX_PUNPCKLBW : MMXI_binop_rm_int<0x60, "punpcklbw",
int_x86_mmx_punpcklbw,
MMX_UNPCK_L_ITINS>;
-defm MMX_PUNPCKLWD : MMXI_binop_rm_int<0x61, "punpcklwd",
+defm MMX_PUNPCKLWD : MMXI_binop_rm_int<0x61, "punpcklwd",
int_x86_mmx_punpcklwd,
MMX_UNPCK_L_ITINS>;
defm MMX_PUNPCKLDQ : MMXI_binop_rm_int<0x62, "punpckldq",
MMX_PSHUF_ITINS>;
def MMX_PSHUFWri : MMXIi8<0x70, MRMSrcReg,
- (outs VR64:$dst), (ins VR64:$src1, i8imm:$src2),
+ (outs VR64:$dst), (ins VR64:$src1, u8imm:$src2),
"pshufw\t{$src2, $src1, $dst|$dst, $src1, $src2}",
[(set VR64:$dst,
(int_x86_sse_pshuf_w VR64:$src1, imm:$src2))],
- IIC_MMX_PSHUF>;
+ IIC_MMX_PSHUF>, Sched<[WriteShuffle]>;
def MMX_PSHUFWmi : MMXIi8<0x70, MRMSrcMem,
- (outs VR64:$dst), (ins i64mem:$src1, i8imm:$src2),
+ (outs VR64:$dst), (ins i64mem:$src1, u8imm:$src2),
"pshufw\t{$src2, $src1, $dst|$dst, $src1, $src2}",
[(set VR64:$dst,
(int_x86_sse_pshuf_w (load_mmx addr:$src1),
imm:$src2))],
- IIC_MMX_PSHUF>;
+ IIC_MMX_PSHUF>, Sched<[WriteShuffleLd]>;
// -- Conversion Instructions
defm MMX_CVTPS2PI : sse12_cvt_pint<0x2D, VR128, VR64, int_x86_sse_cvtps2pi,
f64mem, load, "cvtps2pi\t{$src, $dst|$dst, $src}",
- MMX_CVT_PS_ITINS, SSEPackedSingle>, TB;
+ MMX_CVT_PS_ITINS, SSEPackedSingle>, PS;
defm MMX_CVTPD2PI : sse12_cvt_pint<0x2D, VR128, VR64, int_x86_sse_cvtpd2pi,
f128mem, memop, "cvtpd2pi\t{$src, $dst|$dst, $src}",
- MMX_CVT_PD_ITINS, SSEPackedDouble>, TB, OpSize;
+ MMX_CVT_PD_ITINS, SSEPackedDouble>, PD;
defm MMX_CVTTPS2PI : sse12_cvt_pint<0x2C, VR128, VR64, int_x86_sse_cvttps2pi,
f64mem, load, "cvttps2pi\t{$src, $dst|$dst, $src}",
- MMX_CVT_PS_ITINS, SSEPackedSingle>, TB;
+ MMX_CVT_PS_ITINS, SSEPackedSingle>, PS;
defm MMX_CVTTPD2PI : sse12_cvt_pint<0x2C, VR128, VR64, int_x86_sse_cvttpd2pi,
f128mem, memop, "cvttpd2pi\t{$src, $dst|$dst, $src}",
- MMX_CVT_PD_ITINS, SSEPackedDouble>, TB, OpSize;
+ MMX_CVT_PD_ITINS, SSEPackedDouble>, PD;
defm MMX_CVTPI2PD : sse12_cvt_pint<0x2A, VR64, VR128, int_x86_sse_cvtpi2pd,
i64mem, load, "cvtpi2pd\t{$src, $dst|$dst, $src}",
- MMX_CVT_PD_ITINS, SSEPackedDouble>, TB, OpSize;
+ MMX_CVT_PD_ITINS, SSEPackedDouble>, PD;
let Constraints = "$src1 = $dst" in {
defm MMX_CVTPI2PS : sse12_cvt_pint_3addr<0x2A, VR64, VR128,
int_x86_sse_cvtpi2ps,
i64mem, load, "cvtpi2ps\t{$src2, $dst|$dst, $src2}",
- SSEPackedSingle>, TB;
+ SSEPackedSingle>, PS;
}
// Extract / Insert
def MMX_PEXTRWirri: MMXIi8<0xC5, MRMSrcReg,
- (outs GR32:$dst), (ins VR64:$src1, i32i8imm:$src2),
- "pextrw\t{$src2, $src1, $dst|$dst, $src1, $src2}",
- [(set GR32:$dst, (int_x86_mmx_pextr_w VR64:$src1,
- (iPTR imm:$src2)))],
- IIC_MMX_PEXTR>;
+ (outs GR32orGR64:$dst), (ins VR64:$src1, i32u8imm:$src2),
+ "pextrw\t{$src2, $src1, $dst|$dst, $src1, $src2}",
+ [(set GR32orGR64:$dst, (int_x86_mmx_pextr_w VR64:$src1,
+ imm:$src2))],
+ IIC_MMX_PEXTR>, Sched<[WriteShuffle]>;
let Constraints = "$src1 = $dst" in {
def MMX_PINSRWirri : MMXIi8<0xC4, MRMSrcReg,
- (outs VR64:$dst),
- (ins VR64:$src1, GR32:$src2, i32i8imm:$src3),
+ (outs VR64:$dst),
+ (ins VR64:$src1, GR32orGR64:$src2, i32u8imm:$src3),
"pinsrw\t{$src3, $src2, $dst|$dst, $src2, $src3}",
[(set VR64:$dst, (int_x86_mmx_pinsr_w VR64:$src1,
- GR32:$src2, (iPTR imm:$src3)))],
- IIC_MMX_PINSRW>;
+ GR32orGR64:$src2, imm:$src3))],
+ IIC_MMX_PINSRW>, Sched<[WriteShuffle]>;
def MMX_PINSRWirmi : MMXIi8<0xC4, MRMSrcMem,
(outs VR64:$dst),
- (ins VR64:$src1, i16mem:$src2, i32i8imm:$src3),
+ (ins VR64:$src1, i16mem:$src2, i32u8imm:$src3),
"pinsrw\t{$src3, $src2, $dst|$dst, $src2, $src3}",
[(set VR64:$dst, (int_x86_mmx_pinsr_w VR64:$src1,
(i32 (anyext (loadi16 addr:$src2))),
- (iPTR imm:$src3)))],
- IIC_MMX_PINSRW>;
+ imm:$src3))],
+ IIC_MMX_PINSRW>, Sched<[WriteShuffleLd, ReadAfterLd]>;
}
// Mask creation
-def MMX_PMOVMSKBrr : MMXI<0xD7, MRMSrcReg, (outs GR32:$dst), (ins VR64:$src),
+def MMX_PMOVMSKBrr : MMXI<0xD7, MRMSrcReg, (outs GR32orGR64:$dst),
+ (ins VR64:$src),
"pmovmskb\t{$src, $dst|$dst, $src}",
- [(set GR32:$dst,
+ [(set GR32orGR64:$dst,
(int_x86_mmx_pmovmskb VR64:$src))]>;
(x86mmx (MMX_MOVQ64rm addr:$src))>;
// Misc.
+let SchedRW = [WriteShuffle] in {
let Uses = [EDI] in
-def MMX_MASKMOVQ : MMXI<0xF7, MRMSrcReg, (outs), (ins VR64:$src, VR64:$mask),
- "maskmovq\t{$mask, $src|$src, $mask}",
- [(int_x86_mmx_maskmovq VR64:$src, VR64:$mask, EDI)],
- IIC_MMX_MASKMOV>;
+def MMX_MASKMOVQ : MMXI32<0xF7, MRMSrcReg, (outs), (ins VR64:$src, VR64:$mask),
+ "maskmovq\t{$mask, $src|$src, $mask}",
+ [(int_x86_mmx_maskmovq VR64:$src, VR64:$mask, EDI)],
+ IIC_MMX_MASKMOV>;
let Uses = [RDI] in
def MMX_MASKMOVQ64: MMXI64<0xF7, MRMSrcReg, (outs), (ins VR64:$src, VR64:$mask),
"maskmovq\t{$mask, $src|$src, $mask}",
[(int_x86_mmx_maskmovq VR64:$src, VR64:$mask, RDI)],
IIC_MMX_MASKMOV>;
+}
// 64-bit bit convert.
let Predicates = [HasSSE2] in {
-def : Pat<(x86mmx (bitconvert (i64 GR64:$src))),
- (MMX_MOVD64to64rr GR64:$src)>;
-def : Pat<(i64 (bitconvert (x86mmx VR64:$src))),
- (MMX_MOVD64from64rr VR64:$src)>;
def : Pat<(f64 (bitconvert (x86mmx VR64:$src))),
(MMX_MOVQ2FR64rr VR64:$src)>;
def : Pat<(x86mmx (bitconvert (f64 FR64:$src))),
projects / oota-llvm.git / blobdiff