>;
}
+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 = WriteVecALU in {
+let Sched = WriteVecIMul in {
def MMX_PSADBW_ITINS : OpndItins<
IIC_MMX_PSADBW, IIC_MMX_PSADBW
>;
(bitconvert (load_mmx addr:$src2))))],
itins.rm>, Sched<[WriteVecShiftLd, ReadAfterLd]>;
def ri : MMXIi8<opc2, ImmForm, (outs VR64:$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:$src2))], itins.ri>,
Sched<[WriteVecShift]>;
}
}
/// PALIGN MMX instructions (require SSSE3).
multiclass ssse3_palign_mm<string asm, Intrinsic IntId> {
def R64irr : MMXSS3AI<0x0F, MRMSrcReg, (outs VR64:$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:$src3)))]>,
+ Sched<[WriteShuffleLd, ReadAfterLd]>;
}
multiclass sse12_cvt_pint<bits<8> opc, RegisterClass SrcRC, RegisterClass DstRC,
def irr : MMXPI<opc, MRMSrcReg, (outs DstRC:$dst),
(ins DstRC:$src1, SrcRC:$src2), asm,
[(set DstRC:$dst, (Int DstRC:$src1, SrcRC:$src2))],
- NoItinerary, d>;
+ NoItinerary, d>, Sched<[WriteCvtI2F]>;
def irm : MMXPI<opc, MRMSrcMem, (outs DstRC:$dst),
(ins DstRC:$src1, x86memop:$src2), asm,
[(set DstRC:$dst, (Int DstRC:$src1, (ld_frag addr:$src2)))],
- NoItinerary, d>;
+ NoItinerary, d>, Sched<[WriteCvtI2FLd]>;
}
//===----------------------------------------------------------------------===//
// Data Transfer Instructions
def MMX_MOVD64rr : MMXI<0x6E, MRMSrcReg, (outs VR64:$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>, Sched<[WriteMove]>;
-let canFoldAsLoad = 1 in
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:$src))))],
IIC_MMX_MOV_MM_RM>, Sched<[WriteLoad]>;
+
+let Predicates = [HasMMX] in {
+ let AddedComplexity = 15 in
+ def : Pat<(x86mmx (MMX_X86movw2d GR32:$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, $src}", [], IIC_MMX_MOV_MM_RM>,
Sched<[WriteStore]>;
-// Low word of MMX to GPR.
-def MMX_X86movd2w : SDNode<"X86ISD::MMX_MOVD2W", SDTypeProfile<1, 1,
- [SDTCisVT<0, i32>, SDTCisVT<1, x86mmx>]>>;
def MMX_MOVD64grr : MMXI<0x7E, MRMDestReg, (outs GR32:$dst), (ins VR64:$src),
"movd\t{$src, $dst|$dst, $src}",
[(set GR32:$dst,
(MMX_X86movd2w (x86mmx VR64:$src)))],
IIC_MMX_MOV_REG_MM>, Sched<[WriteMove]>;
+let isBitcast = 1 in
def MMX_MOVD64to64rr : MMXRI<0x6E, MRMSrcReg, (outs VR64:$dst), (ins GR64:$src),
"movd\t{$src, $dst|$dst, $src}",
[(set VR64:$dst, (bitconvert GR64:$src))],
IIC_MMX_MOV_MM_RM>, Sched<[WriteMove]>;
+let isCodeGenOnly = 1, ForceDisassemble = 1, hasSideEffects = 0, mayLoad = 1 in
+def MMX_MOVD64to64rm : MMXRI<0x6E, MRMSrcMem, (outs VR64:$dst),
+ (ins i64mem:$src), "movd\t{$src, $dst|$dst, $src}",
+ [], IIC_MMX_MOVQ_RM>, Sched<[WriteLoad]>;
+
// These are 64 bit moves, but since the OS X assembler doesn't
// recognize a register-register movq, we write them as
// movd.
-let SchedRW = [WriteMove] in {
+let SchedRW = [WriteMove], isBitcast = 1 in {
def MMX_MOVD64from64rr : MMXRI<0x7E, MRMDestReg,
(outs GR64:$dst), (ins VR64:$src),
- "movd\t{$src, $dst|$dst, $src}",
+ "movd\t{$src, $dst|$dst, $src}",
[(set GR64:$dst,
(bitconvert VR64:$src))], IIC_MMX_MOV_REG_MM>;
-let neverHasSideEffects = 1 in
+let hasSideEffects = 0 in
def MMX_MOVQ64rr : MMXI<0x6F, MRMSrcReg, (outs VR64:$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:$dst)],
IIC_MMX_MOVQ_RM>;
-} // SchedRW
let SchedRW = [WriteMove] in {
def MMX_MOVDQ2Qrr : MMXSDIi8<0xD6, MRMSrcReg, (outs VR64:$dst),
// 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>, 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:$src1),
// Extract / Insert
def MMX_PEXTRWirri: MMXIi8<0xC5, MRMSrcReg,
- (outs GR32orGR64:$dst), (ins VR64:$src1, i32i8imm:$src2),
+ (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:$src1,
- (iPTR imm:$src2)))],
+ imm:$src2))],
IIC_MMX_PEXTR>, Sched<[WriteShuffle]>;
let Constraints = "$src1 = $dst" in {
def MMX_PINSRWirri : MMXIi8<0xC4, MRMSrcReg,
- (outs VR64:$dst),
- (ins VR64:$src1, GR32orGR64:$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:$src1,
- GR32orGR64:$src2, (iPTR imm:$src3)))],
+ GR32orGR64:$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:$src3)))],
+ imm:$src3))],
IIC_MMX_PINSRW>, Sched<[WriteShuffleLd, ReadAfterLd]>;
}