IIC_SSE_MOVU_P_RR, IIC_SSE_MOVU_P_RM
>;
+def SSE_DPPD_ITINS : OpndItins<
+ IIC_SSE_DPPD_RR, IIC_SSE_DPPD_RM
+>;
+
+def SSE_DPPS_ITINS : OpndItins<
+ IIC_SSE_DPPS_RR, IIC_SSE_DPPD_RM
+>;
+
+def DEFAULT_ITINS : OpndItins<
+ IIC_ALU_NONMEM, IIC_ALU_MEM
+>;
+
+def SSE_EXTRACT_ITINS : OpndItins<
+ IIC_SSE_EXTRACTPS_RR, IIC_SSE_EXTRACTPS_RM
+>;
+
+def SSE_INSERT_ITINS : OpndItins<
+ IIC_SSE_INSERTPS_RR, IIC_SSE_INSERTPS_RM
+>;
+
+def SSE_MPSADBW_ITINS : OpndItins<
+ IIC_SSE_MPSADBW_RR, IIC_SSE_MPSADBW_RM
+>;
+
+def SSE_PMULLD_ITINS : OpndItins<
+ IIC_SSE_PMULLD_RR, IIC_SSE_PMULLD_RM
+>;
+
//===----------------------------------------------------------------------===//
// SSE 1 & 2 Instructions Classes
//===----------------------------------------------------------------------===//
// SSE 1 & 2 - Move FP Scalar Instructions
//
// Move Instructions. Register-to-register movss/movsd is not used for FR32/64
-// register copies because it's a partial register update; FsMOVAPSrr/FsMOVAPDrr
-// is used instead. Register-to-register movss/movsd is not modeled as an
-// INSERT_SUBREG because INSERT_SUBREG requires that the insert be implementable
-// in terms of a copy, and just mentioned, we don't use movss/movsd for copies.
+// register copies because it's a partial register update; Register-to-register
+// movss/movsd is not modeled as an INSERT_SUBREG because INSERT_SUBREG requires
+// that the insert be implementable in terms of a copy, and just mentioned, we
+// don't use movss/movsd for copies.
//===----------------------------------------------------------------------===//
multiclass sse12_move_rr<RegisterClass RC, SDNode OpNode, ValueType vt,
(MOVUPSmr addr:$dst, VR128:$src)>;
}
-// Alias instruction to do FR32 or FR64 reg-to-reg copy using movaps. Upper
-// bits are disregarded. FIXME: Set encoding to pseudo!
-let neverHasSideEffects = 1, SchedRW = [WriteMove] in {
-def FsVMOVAPSrr : VPSI<0x28, MRMSrcReg, (outs FR32:$dst), (ins FR32:$src),
- "movaps\t{$src, $dst|$dst, $src}", [],
- IIC_SSE_MOVA_P_RR>, VEX;
-def FsVMOVAPDrr : VPDI<0x28, MRMSrcReg, (outs FR64:$dst), (ins FR64:$src),
- "movapd\t{$src, $dst|$dst, $src}", [],
- IIC_SSE_MOVA_P_RR>, VEX;
-def FsMOVAPSrr : PSI<0x28, MRMSrcReg, (outs FR32:$dst), (ins FR32:$src),
- "movaps\t{$src, $dst|$dst, $src}", [],
- IIC_SSE_MOVA_P_RR>;
-def FsMOVAPDrr : PDI<0x28, MRMSrcReg, (outs FR64:$dst), (ins FR64:$src),
- "movapd\t{$src, $dst|$dst, $src}", [],
- IIC_SSE_MOVA_P_RR>;
-}
-
// Alias instruction to load FR32 or FR64 from f128mem using movaps. Upper
// bits are disregarded. FIXME: Set encoding to pseudo!
let canFoldAsLoad = 1, isReMaterializable = 1, SchedRW = [WriteLoad] in {
"movapd\t{$src, $dst|$dst, $src}",
[(set FR64:$dst, (alignedloadfsf64 addr:$src))],
IIC_SSE_MOVA_P_RM>, VEX;
+ def FsMOVAPSrm : PSI<0x28, MRMSrcMem, (outs FR32:$dst), (ins f128mem:$src),
+ "movaps\t{$src, $dst|$dst, $src}",
+ [(set FR32:$dst, (alignedloadfsf32 addr:$src))],
+ IIC_SSE_MOVA_P_RM>;
+ def FsMOVAPDrm : PDI<0x28, MRMSrcMem, (outs FR64:$dst), (ins f128mem:$src),
+ "movapd\t{$src, $dst|$dst, $src}",
+ [(set FR64:$dst, (alignedloadfsf64 addr:$src))],
+ IIC_SSE_MOVA_P_RM>;
}
-def FsMOVAPSrm : PSI<0x28, MRMSrcMem, (outs FR32:$dst), (ins f128mem:$src),
- "movaps\t{$src, $dst|$dst, $src}",
- [(set FR32:$dst, (alignedloadfsf32 addr:$src))],
- IIC_SSE_MOVA_P_RM>;
-def FsMOVAPDrm : PDI<0x28, MRMSrcMem, (outs FR64:$dst), (ins f128mem:$src),
- "movapd\t{$src, $dst|$dst, $src}",
- [(set FR64:$dst, (alignedloadfsf64 addr:$src))],
- IIC_SSE_MOVA_P_RM>;
}
//===----------------------------------------------------------------------===//
let Constraints = "$src1 = $dst" in {
def Int_CVTSD2SSrr: I<0x5A, MRMSrcReg,
(outs VR128:$dst), (ins VR128:$src1, VR128:$src2),
- "cvtsd2ss\t{$src2, $src1, $dst|$dst, $src1, $src2}",
+ "cvtsd2ss\t{$src2, $dst|$dst, $src2}",
[(set VR128:$dst,
(int_x86_sse2_cvtsd2ss VR128:$src1, VR128:$src2))],
IIC_SSE_CVT_Scalar_RR>, XD, Requires<[UseSSE2]>,
Sched<[WriteCvtF2F]>;
def Int_CVTSD2SSrm: I<0x5A, MRMSrcReg,
(outs VR128:$dst), (ins VR128:$src1, sdmem:$src2),
- "cvtsd2ss\t{$src2, $src1, $dst|$dst, $src1, $src2}",
+ "cvtsd2ss\t{$src2, $dst|$dst, $src2}",
[(set VR128:$dst, (int_x86_sse2_cvtsd2ss
VR128:$src1, sse_load_f64:$src2))],
IIC_SSE_CVT_Scalar_RM>, XD, Requires<[UseSSE2]>,
def VCVTPS2DQrm : VPDI<0x5B, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
"cvtps2dq\t{$src, $dst|$dst, $src}",
[(set VR128:$dst,
- (int_x86_sse2_cvtps2dq (memopv4f32 addr:$src)))],
+ (int_x86_sse2_cvtps2dq (loadv4f32 addr:$src)))],
IIC_SSE_CVT_PS_RM>, VEX, Sched<[WriteCvtF2ILd]>;
def VCVTPS2DQYrr : VPDI<0x5B, MRMSrcReg, (outs VR256:$dst), (ins VR256:$src),
"cvtps2dq\t{$src, $dst|$dst, $src}",
def VCVTPS2DQYrm : VPDI<0x5B, MRMSrcMem, (outs VR256:$dst), (ins f256mem:$src),
"cvtps2dq\t{$src, $dst|$dst, $src}",
[(set VR256:$dst,
- (int_x86_avx_cvt_ps2dq_256 (memopv8f32 addr:$src)))],
+ (int_x86_avx_cvt_ps2dq_256 (loadv8f32 addr:$src)))],
IIC_SSE_CVT_PS_RM>, VEX, VEX_L, Sched<[WriteCvtF2ILd]>;
def CVTPS2DQrr : PDI<0x5B, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
"cvtps2dq\t{$src, $dst|$dst, $src}",
def VCVTPD2DQXrm : SDI<0xE6, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
"vcvtpd2dqx\t{$src, $dst|$dst, $src}",
[(set VR128:$dst,
- (int_x86_sse2_cvtpd2dq (memopv2f64 addr:$src)))]>, VEX,
+ (int_x86_sse2_cvtpd2dq (loadv2f64 addr:$src)))]>, VEX,
Sched<[WriteCvtF2ILd]>;
// YMM only
def VCVTPD2DQYrm : SDI<0xE6, MRMSrcMem, (outs VR128:$dst), (ins f256mem:$src),
"vcvtpd2dq{y}\t{$src, $dst|$dst, $src}",
[(set VR128:$dst,
- (int_x86_avx_cvt_pd2dq_256 (memopv4f64 addr:$src)))]>,
+ (int_x86_avx_cvt_pd2dq_256 (loadv4f64 addr:$src)))]>,
VEX, VEX_L, Sched<[WriteCvtF2ILd]>;
def : InstAlias<"vcvtpd2dq\t{$src, $dst|$dst, $src}",
(VCVTPD2DQYrr VR128:$dst, VR256:$src)>;
def VCVTTPS2DQrm : VS2SI<0x5B, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
"cvttps2dq\t{$src, $dst|$dst, $src}",
[(set VR128:$dst, (int_x86_sse2_cvttps2dq
- (memopv4f32 addr:$src)))],
+ (loadv4f32 addr:$src)))],
IIC_SSE_CVT_PS_RM>, VEX, Sched<[WriteCvtF2ILd]>;
def VCVTTPS2DQYrr : VS2SI<0x5B, MRMSrcReg, (outs VR256:$dst), (ins VR256:$src),
"cvttps2dq\t{$src, $dst|$dst, $src}",
def VCVTTPS2DQYrm : VS2SI<0x5B, MRMSrcMem, (outs VR256:$dst), (ins f256mem:$src),
"cvttps2dq\t{$src, $dst|$dst, $src}",
[(set VR256:$dst, (int_x86_avx_cvtt_ps2dq_256
- (memopv8f32 addr:$src)))],
+ (loadv8f32 addr:$src)))],
IIC_SSE_CVT_PS_RM>, VEX, VEX_L,
Sched<[WriteCvtF2ILd]>;
let Predicates = [HasAVX] in {
def : Pat<(v4f32 (sint_to_fp (v4i32 VR128:$src))),
(VCVTDQ2PSrr VR128:$src)>;
- def : Pat<(v4f32 (sint_to_fp (bc_v4i32 (memopv2i64 addr:$src)))),
+ def : Pat<(v4f32 (sint_to_fp (bc_v4i32 (loadv2i64 addr:$src)))),
(VCVTDQ2PSrm addr:$src)>;
def : Pat<(int_x86_sse2_cvtdq2ps VR128:$src),
(VCVTDQ2PSrr VR128:$src)>;
- def : Pat<(int_x86_sse2_cvtdq2ps (bc_v4i32 (memopv2i64 addr:$src))),
+ def : Pat<(int_x86_sse2_cvtdq2ps (bc_v4i32 (loadv2i64 addr:$src))),
(VCVTDQ2PSrm addr:$src)>;
def : Pat<(v4i32 (fp_to_sint (v4f32 VR128:$src))),
(VCVTTPS2DQrr VR128:$src)>;
- def : Pat<(v4i32 (fp_to_sint (memopv4f32 addr:$src))),
+ def : Pat<(v4i32 (fp_to_sint (loadv4f32 addr:$src))),
(VCVTTPS2DQrm addr:$src)>;
def : Pat<(v8f32 (sint_to_fp (v8i32 VR256:$src))),
(VCVTDQ2PSYrr VR256:$src)>;
- def : Pat<(v8f32 (sint_to_fp (bc_v8i32 (memopv4i64 addr:$src)))),
+ def : Pat<(v8f32 (sint_to_fp (bc_v8i32 (loadv4i64 addr:$src)))),
(VCVTDQ2PSYrm addr:$src)>;
def : Pat<(v8i32 (fp_to_sint (v8f32 VR256:$src))),
(VCVTTPS2DQYrr VR256:$src)>;
- def : Pat<(v8i32 (fp_to_sint (memopv8f32 addr:$src))),
+ def : Pat<(v8i32 (fp_to_sint (loadv8f32 addr:$src))),
(VCVTTPS2DQYrm addr:$src)>;
}
def VCVTTPD2DQXrm : VPDI<0xE6, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
"cvttpd2dqx\t{$src, $dst|$dst, $src}",
[(set VR128:$dst, (int_x86_sse2_cvttpd2dq
- (memopv2f64 addr:$src)))],
+ (loadv2f64 addr:$src)))],
IIC_SSE_CVT_PD_RM>, VEX, Sched<[WriteCvtF2ILd]>;
// YMM only
def VCVTTPD2DQYrm : VPDI<0xE6, MRMSrcMem, (outs VR128:$dst), (ins f256mem:$src),
"cvttpd2dq{y}\t{$src, $dst|$dst, $src}",
[(set VR128:$dst,
- (int_x86_avx_cvtt_pd2dq_256 (memopv4f64 addr:$src)))],
+ (int_x86_avx_cvtt_pd2dq_256 (loadv4f64 addr:$src)))],
IIC_SSE_CVT_PD_RM>, VEX, VEX_L, Sched<[WriteCvtF2ILd]>;
def : InstAlias<"vcvttpd2dq\t{$src, $dst|$dst, $src}",
(VCVTTPD2DQYrr VR128:$dst, VR256:$src)>;
let Predicates = [HasAVX] in {
def : Pat<(v4i32 (fp_to_sint (v4f64 VR256:$src))),
(VCVTTPD2DQYrr VR256:$src)>;
- def : Pat<(v4i32 (fp_to_sint (memopv4f64 addr:$src))),
+ def : Pat<(v4i32 (fp_to_sint (loadv4f64 addr:$src))),
(VCVTTPD2DQYrm addr:$src)>;
} // Predicates = [HasAVX]
def VCVTPS2PDYrm : I<0x5A, MRMSrcMem, (outs VR256:$dst), (ins f128mem:$src),
"vcvtps2pd\t{$src, $dst|$dst, $src}",
[(set VR256:$dst,
- (int_x86_avx_cvt_ps2_pd_256 (memopv4f32 addr:$src)))],
+ (int_x86_avx_cvt_ps2_pd_256 (loadv4f32 addr:$src)))],
IIC_SSE_CVT_PD_RM>, TB, VEX, VEX_L, Sched<[WriteCvtF2FLd]>;
}
"vcvtdq2pd\t{$src, $dst|$dst, $src}",
[(set VR256:$dst,
(int_x86_avx_cvtdq2_pd_256
- (bitconvert (memopv2i64 addr:$src))))]>, VEX, VEX_L,
+ (bitconvert (loadv2i64 addr:$src))))]>, VEX, VEX_L,
Sched<[WriteCvtI2FLd]>;
def VCVTDQ2PDYrr : S2SI<0xE6, MRMSrcReg, (outs VR256:$dst), (ins VR128:$src),
"vcvtdq2pd\t{$src, $dst|$dst, $src}",
let Predicates = [HasAVX] in {
def : Pat<(v4f64 (sint_to_fp (v4i32 VR128:$src))),
(VCVTDQ2PDYrr VR128:$src)>;
- def : Pat<(v4f64 (sint_to_fp (bc_v4i32 (memopv2i64 addr:$src)))),
+ def : Pat<(v4f64 (sint_to_fp (bc_v4i32 (loadv2i64 addr:$src)))),
(VCVTDQ2PDYrm addr:$src)>;
} // Predicates = [HasAVX]
def VCVTPD2PSXrm : VPDI<0x5A, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
"cvtpd2psx\t{$src, $dst|$dst, $src}",
[(set VR128:$dst,
- (int_x86_sse2_cvtpd2ps (memopv2f64 addr:$src)))],
+ (int_x86_sse2_cvtpd2ps (loadv2f64 addr:$src)))],
IIC_SSE_CVT_PD_RM>, VEX, Sched<[WriteCvtF2FLd]>;
// YMM only
def VCVTPD2PSYrm : VPDI<0x5A, MRMSrcMem, (outs VR128:$dst), (ins f256mem:$src),
"cvtpd2ps{y}\t{$src, $dst|$dst, $src}",
[(set VR128:$dst,
- (int_x86_avx_cvt_pd2_ps_256 (memopv4f64 addr:$src)))],
+ (int_x86_avx_cvt_pd2_ps_256 (loadv4f64 addr:$src)))],
IIC_SSE_CVT_PD_RM>, VEX, VEX_L, Sched<[WriteCvtF2FLd]>;
def : InstAlias<"vcvtpd2ps\t{$src, $dst|$dst, $src}",
(VCVTPD2PSYrr VR128:$dst, VR256:$src)>;
let Predicates = [HasAVX] in {
def : Pat<(int_x86_avx_cvtdq2_ps_256 VR256:$src),
(VCVTDQ2PSYrr VR256:$src)>;
- def : Pat<(int_x86_avx_cvtdq2_ps_256 (bitconvert (memopv4i64 addr:$src))),
+ def : Pat<(int_x86_avx_cvtdq2_ps_256 (bitconvert (loadv4i64 addr:$src))),
(VCVTDQ2PSYrm addr:$src)>;
// Match fround and fextend for 128/256-bit conversions
def : Pat<(v4f32 (X86vfpround (v2f64 VR128:$src))),
(VCVTPD2PSrr VR128:$src)>;
- def : Pat<(v4f32 (X86vfpround (memopv2f64 addr:$src))),
+ def : Pat<(v4f32 (X86vfpround (loadv2f64 addr:$src))),
(VCVTPD2PSXrm addr:$src)>;
def : Pat<(v4f32 (fround (v4f64 VR256:$src))),
(VCVTPD2PSYrr VR256:$src)>;
defm CMPSD : sse12_cmp_scalar<FR64, f64mem, SSECC, X86cmpsd, f64, loadf64,
"cmp${cc}sd\t{$src2, $dst|$dst, $src2}",
"cmpsd\t{$cc, $src2, $dst|$dst, $src2, $cc}",
- SSE_ALU_F32S>, // same latency as 32 bit compare
+ SSE_ALU_F64S>,
XD;
}
SSE_ALU_F32S>, XS;
defm Int_CMPSD : sse12_cmp_scalar_int<f64mem, SSECC, int_x86_sse2_cmp_sd,
"cmp${cc}sd\t{$src, $dst|$dst, $src}",
- SSE_ALU_F32S>, // same latency as f32
+ SSE_ALU_F64S>,
XD;
}
// sse12_cmp_packed - sse 1 & 2 compare packed instructions
multiclass sse12_cmp_packed<RegisterClass RC, X86MemOperand x86memop,
Operand CC, Intrinsic Int, string asm,
- string asm_alt, Domain d> {
+ string asm_alt, Domain d,
+ OpndItins itins = SSE_ALU_F32P> {
def rri : PIi8<0xC2, MRMSrcReg,
(outs RC:$dst), (ins RC:$src1, RC:$src2, CC:$cc), asm,
[(set RC:$dst, (Int RC:$src1, RC:$src2, imm:$cc))],
- IIC_SSE_CMPP_RR, d>,
+ itins.rr, d>,
Sched<[WriteFAdd]>;
def rmi : PIi8<0xC2, MRMSrcMem,
(outs RC:$dst), (ins RC:$src1, x86memop:$src2, CC:$cc), asm,
[(set RC:$dst, (Int RC:$src1, (memop addr:$src2), imm:$cc))],
- IIC_SSE_CMPP_RM, d>,
+ itins.rm, d>,
Sched<[WriteFAddLd, ReadAfterLd]>;
// Accept explicit immediate argument form instead of comparison code.
let neverHasSideEffects = 1 in {
def rri_alt : PIi8<0xC2, MRMSrcReg,
(outs RC:$dst), (ins RC:$src1, RC:$src2, i8imm:$cc),
- asm_alt, [], IIC_SSE_CMPP_RR, d>, Sched<[WriteFAdd]>;
+ asm_alt, [], itins.rr, d>, Sched<[WriteFAdd]>;
def rmi_alt : PIi8<0xC2, MRMSrcMem,
(outs RC:$dst), (ins RC:$src1, x86memop:$src2, i8imm:$cc),
- asm_alt, [], IIC_SSE_CMPP_RM, d>,
+ asm_alt, [], itins.rm, d>,
Sched<[WriteFAddLd, ReadAfterLd]>;
}
}
defm CMPPS : sse12_cmp_packed<VR128, f128mem, SSECC, int_x86_sse_cmp_ps,
"cmp${cc}ps\t{$src2, $dst|$dst, $src2}",
"cmpps\t{$cc, $src2, $dst|$dst, $src2, $cc}",
- SSEPackedSingle>, TB;
+ SSEPackedSingle, SSE_ALU_F32P>, TB;
defm CMPPD : sse12_cmp_packed<VR128, f128mem, SSECC, int_x86_sse2_cmp_pd,
"cmp${cc}pd\t{$src2, $dst|$dst, $src2}",
"cmppd\t{$cc, $src2, $dst|$dst, $src2, $cc}",
- SSEPackedDouble>, TB, OpSize;
+ SSEPackedDouble, SSE_ALU_F64P>, TB, OpSize;
}
let Predicates = [HasAVX] in {
defm VSHUFPS : sse12_shuffle<VR128, f128mem, v4f32,
"shufps\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}",
- memopv4f32, SSEPackedSingle>, TB, VEX_4V;
+ loadv4f32, SSEPackedSingle>, TB, VEX_4V;
defm VSHUFPSY : sse12_shuffle<VR256, f256mem, v8f32,
"shufps\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}",
- memopv8f32, SSEPackedSingle>, TB, VEX_4V, VEX_L;
+ loadv8f32, SSEPackedSingle>, TB, VEX_4V, VEX_L;
defm VSHUFPD : sse12_shuffle<VR128, f128mem, v2f64,
- "shufpd\t{$src3, $src2, $src1, $dst|$dst, $src2, $src2, $src3}",
- memopv2f64, SSEPackedDouble>, TB, OpSize, VEX_4V;
+ "shufpd\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}",
+ loadv2f64, SSEPackedDouble>, TB, OpSize, VEX_4V;
defm VSHUFPDY : sse12_shuffle<VR256, f256mem, v4f64,
- "shufpd\t{$src3, $src2, $src1, $dst|$dst, $src2, $src2, $src3}",
- memopv4f64, SSEPackedDouble>, TB, OpSize, VEX_4V, VEX_L;
+ "shufpd\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}",
+ loadv4f64, SSEPackedDouble>, TB, OpSize, VEX_4V, VEX_L;
let Constraints = "$src1 = $dst" in {
defm SHUFPS : sse12_shuffle<VR128, f128mem, v4f32,
let Predicates = [HasAVX] in {
def : Pat<(v4i32 (X86Shufp VR128:$src1,
- (bc_v4i32 (memopv2i64 addr:$src2)), (i8 imm:$imm))),
+ (bc_v4i32 (loadv2i64 addr:$src2)), (i8 imm:$imm))),
(VSHUFPSrmi VR128:$src1, addr:$src2, imm:$imm)>;
def : Pat<(v4i32 (X86Shufp VR128:$src1, VR128:$src2, (i8 imm:$imm))),
(VSHUFPSrri VR128:$src1, VR128:$src2, imm:$imm)>;
def : Pat<(v2i64 (X86Shufp VR128:$src1,
- (memopv2i64 addr:$src2), (i8 imm:$imm))),
+ (loadv2i64 addr:$src2), (i8 imm:$imm))),
(VSHUFPDrmi VR128:$src1, addr:$src2, imm:$imm)>;
def : Pat<(v2i64 (X86Shufp VR128:$src1, VR128:$src2, (i8 imm:$imm))),
(VSHUFPDrri VR128:$src1, VR128:$src2, imm:$imm)>;
def : Pat<(v8i32 (X86Shufp VR256:$src1, VR256:$src2, (i8 imm:$imm))),
(VSHUFPSYrri VR256:$src1, VR256:$src2, imm:$imm)>;
def : Pat<(v8i32 (X86Shufp VR256:$src1,
- (bc_v8i32 (memopv4i64 addr:$src2)), (i8 imm:$imm))),
+ (bc_v8i32 (loadv4i64 addr:$src2)), (i8 imm:$imm))),
(VSHUFPSYrmi VR256:$src1, addr:$src2, imm:$imm)>;
def : Pat<(v4i64 (X86Shufp VR256:$src1, VR256:$src2, (i8 imm:$imm))),
(VSHUFPDYrri VR256:$src1, VR256:$src2, imm:$imm)>;
def : Pat<(v4i64 (X86Shufp VR256:$src1,
- (memopv4i64 addr:$src2), (i8 imm:$imm))),
+ (loadv4i64 addr:$src2), (i8 imm:$imm))),
(VSHUFPDYrmi VR256:$src1, addr:$src2, imm:$imm)>;
}
Sched<[WriteShuffleLd, ReadAfterLd]>;
}
-defm VUNPCKHPS: sse12_unpack_interleave<0x15, X86Unpckh, v4f32, memopv4f32,
+defm VUNPCKHPS: sse12_unpack_interleave<0x15, X86Unpckh, v4f32, loadv4f32,
VR128, f128mem, "unpckhps\t{$src2, $src1, $dst|$dst, $src1, $src2}",
SSEPackedSingle>, TB, VEX_4V;
-defm VUNPCKHPD: sse12_unpack_interleave<0x15, X86Unpckh, v2f64, memopv2f64,
+defm VUNPCKHPD: sse12_unpack_interleave<0x15, X86Unpckh, v2f64, loadv2f64,
VR128, f128mem, "unpckhpd\t{$src2, $src1, $dst|$dst, $src1, $src2}",
SSEPackedDouble>, TB, OpSize, VEX_4V;
-defm VUNPCKLPS: sse12_unpack_interleave<0x14, X86Unpckl, v4f32, memopv4f32,
+defm VUNPCKLPS: sse12_unpack_interleave<0x14, X86Unpckl, v4f32, loadv4f32,
VR128, f128mem, "unpcklps\t{$src2, $src1, $dst|$dst, $src1, $src2}",
SSEPackedSingle>, TB, VEX_4V;
-defm VUNPCKLPD: sse12_unpack_interleave<0x14, X86Unpckl, v2f64, memopv2f64,
+defm VUNPCKLPD: sse12_unpack_interleave<0x14, X86Unpckl, v2f64, loadv2f64,
VR128, f128mem, "unpcklpd\t{$src2, $src1, $dst|$dst, $src1, $src2}",
SSEPackedDouble>, TB, OpSize, VEX_4V;
-defm VUNPCKHPSY: sse12_unpack_interleave<0x15, X86Unpckh, v8f32, memopv8f32,
+defm VUNPCKHPSY: sse12_unpack_interleave<0x15, X86Unpckh, v8f32, loadv8f32,
VR256, f256mem, "unpckhps\t{$src2, $src1, $dst|$dst, $src1, $src2}",
SSEPackedSingle>, TB, VEX_4V, VEX_L;
-defm VUNPCKHPDY: sse12_unpack_interleave<0x15, X86Unpckh, v4f64, memopv4f64,
+defm VUNPCKHPDY: sse12_unpack_interleave<0x15, X86Unpckh, v4f64, loadv4f64,
VR256, f256mem, "unpckhpd\t{$src2, $src1, $dst|$dst, $src1, $src2}",
SSEPackedDouble>, TB, OpSize, VEX_4V, VEX_L;
-defm VUNPCKLPSY: sse12_unpack_interleave<0x14, X86Unpckl, v8f32, memopv8f32,
+defm VUNPCKLPSY: sse12_unpack_interleave<0x14, X86Unpckl, v8f32, loadv8f32,
VR256, f256mem, "unpcklps\t{$src2, $src1, $dst|$dst, $src1, $src2}",
SSEPackedSingle>, TB, VEX_4V, VEX_L;
-defm VUNPCKLPDY: sse12_unpack_interleave<0x14, X86Unpckl, v4f64, memopv4f64,
+defm VUNPCKLPDY: sse12_unpack_interleave<0x14, X86Unpckl, v4f64, loadv4f64,
VR256, f256mem, "unpcklpd\t{$src2, $src1, $dst|$dst, $src1, $src2}",
SSEPackedDouble>, TB, OpSize, VEX_4V, VEX_L;
} // Constraints = "$src1 = $dst"
let Predicates = [HasAVX1Only] in {
- def : Pat<(v8i32 (X86Unpckl VR256:$src1, (bc_v8i32 (memopv4i64 addr:$src2)))),
+ def : Pat<(v8i32 (X86Unpckl VR256:$src1, (bc_v8i32 (loadv4i64 addr:$src2)))),
(VUNPCKLPSYrm VR256:$src1, addr:$src2)>;
def : Pat<(v8i32 (X86Unpckl VR256:$src1, VR256:$src2)),
(VUNPCKLPSYrr VR256:$src1, VR256:$src2)>;
- def : Pat<(v8i32 (X86Unpckh VR256:$src1, (bc_v8i32 (memopv4i64 addr:$src2)))),
+ def : Pat<(v8i32 (X86Unpckh VR256:$src1, (bc_v8i32 (loadv4i64 addr:$src2)))),
(VUNPCKHPSYrm VR256:$src1, addr:$src2)>;
def : Pat<(v8i32 (X86Unpckh VR256:$src1, VR256:$src2)),
(VUNPCKHPSYrr VR256:$src1, VR256:$src2)>;
- def : Pat<(v4i64 (X86Unpckl VR256:$src1, (memopv4i64 addr:$src2))),
+ def : Pat<(v4i64 (X86Unpckl VR256:$src1, (loadv4i64 addr:$src2))),
(VUNPCKLPDYrm VR256:$src1, addr:$src2)>;
def : Pat<(v4i64 (X86Unpckl VR256:$src1, VR256:$src2)),
(VUNPCKLPDYrr VR256:$src1, VR256:$src2)>;
- def : Pat<(v4i64 (X86Unpckh VR256:$src1, (memopv4i64 addr:$src2))),
+ def : Pat<(v4i64 (X86Unpckh VR256:$src1, (loadv4i64 addr:$src2))),
(VUNPCKHPDYrm VR256:$src1, addr:$src2)>;
def : Pat<(v4i64 (X86Unpckh VR256:$src1, VR256:$src2)),
(VUNPCKHPDYrr VR256:$src1, VR256:$src2)>;
/// sse12_extr_sign_mask - sse 1 & 2 unpack and interleave
multiclass sse12_extr_sign_mask<RegisterClass RC, Intrinsic Int, string asm,
Domain d> {
- def rr32 : PI<0x50, MRMSrcReg, (outs GR32:$dst), (ins RC:$src),
- !strconcat(asm, "\t{$src, $dst|$dst, $src}"),
- [(set GR32:$dst, (Int RC:$src))], IIC_SSE_MOVMSK, d>,
- Sched<[WriteVecLogic]>;
- def rr64 : PI<0x50, MRMSrcReg, (outs GR64:$dst), (ins RC:$src),
- !strconcat(asm, "\t{$src, $dst|$dst, $src}"), [],
- IIC_SSE_MOVMSK, d>, REX_W, Sched<[WriteVecLogic]>;
+ def rr : PI<0x50, MRMSrcReg, (outs GR32orGR64:$dst), (ins RC:$src),
+ !strconcat(asm, "\t{$src, $dst|$dst, $src}"),
+ [(set GR32orGR64:$dst, (Int RC:$src))], IIC_SSE_MOVMSK, d>,
+ Sched<[WriteVecLogic]>;
}
let Predicates = [HasAVX] in {
OpSize, VEX, VEX_L;
def : Pat<(i32 (X86fgetsign FR32:$src)),
- (VMOVMSKPSrr32 (COPY_TO_REGCLASS FR32:$src, VR128))>;
+ (VMOVMSKPSrr (COPY_TO_REGCLASS FR32:$src, VR128))>;
def : Pat<(i64 (X86fgetsign FR32:$src)),
- (VMOVMSKPSrr64 (COPY_TO_REGCLASS FR32:$src, VR128))>;
+ (SUBREG_TO_REG (i64 0),
+ (VMOVMSKPSrr (COPY_TO_REGCLASS FR32:$src, VR128)), sub_32bit)>;
def : Pat<(i32 (X86fgetsign FR64:$src)),
- (VMOVMSKPDrr32 (COPY_TO_REGCLASS FR64:$src, VR128))>;
+ (VMOVMSKPDrr (COPY_TO_REGCLASS FR64:$src, VR128))>;
def : Pat<(i64 (X86fgetsign FR64:$src)),
- (VMOVMSKPDrr64 (COPY_TO_REGCLASS FR64:$src, VR128))>;
-
- // Assembler Only
- def VMOVMSKPSr64r : PI<0x50, MRMSrcReg, (outs GR64:$dst), (ins VR128:$src),
- "movmskps\t{$src, $dst|$dst, $src}", [], IIC_SSE_MOVMSK,
- SSEPackedSingle>, TB, VEX, Sched<[WriteVecLogic]>;
- def VMOVMSKPDr64r : PI<0x50, MRMSrcReg, (outs GR64:$dst), (ins VR128:$src),
- "movmskpd\t{$src, $dst|$dst, $src}", [], IIC_SSE_MOVMSK,
- SSEPackedDouble>, TB,
- OpSize, VEX, Sched<[WriteVecLogic]>;
- def VMOVMSKPSYr64r : PI<0x50, MRMSrcReg, (outs GR64:$dst), (ins VR256:$src),
- "movmskps\t{$src, $dst|$dst, $src}", [], IIC_SSE_MOVMSK,
- SSEPackedSingle>, TB, VEX, VEX_L, Sched<[WriteVecLogic]>;
- def VMOVMSKPDYr64r : PI<0x50, MRMSrcReg, (outs GR64:$dst), (ins VR256:$src),
- "movmskpd\t{$src, $dst|$dst, $src}", [], IIC_SSE_MOVMSK,
- SSEPackedDouble>, TB,
- OpSize, VEX, VEX_L, Sched<[WriteVecLogic]>;
+ (SUBREG_TO_REG (i64 0),
+ (VMOVMSKPDrr (COPY_TO_REGCLASS FR64:$src, VR128)), sub_32bit)>;
}
defm MOVMSKPS : sse12_extr_sign_mask<VR128, int_x86_sse_movmsk_ps, "movmskps",
SSEPackedDouble>, TB, OpSize;
def : Pat<(i32 (X86fgetsign FR32:$src)),
- (MOVMSKPSrr32 (COPY_TO_REGCLASS FR32:$src, VR128))>,
+ (MOVMSKPSrr (COPY_TO_REGCLASS FR32:$src, VR128))>,
Requires<[UseSSE1]>;
def : Pat<(i64 (X86fgetsign FR32:$src)),
- (MOVMSKPSrr64 (COPY_TO_REGCLASS FR32:$src, VR128))>,
+ (SUBREG_TO_REG (i64 0),
+ (MOVMSKPSrr (COPY_TO_REGCLASS FR32:$src, VR128)), sub_32bit)>,
Requires<[UseSSE1]>;
def : Pat<(i32 (X86fgetsign FR64:$src)),
- (MOVMSKPDrr32 (COPY_TO_REGCLASS FR64:$src, VR128))>,
+ (MOVMSKPDrr (COPY_TO_REGCLASS FR64:$src, VR128))>,
Requires<[UseSSE2]>;
def : Pat<(i64 (X86fgetsign FR64:$src)),
- (MOVMSKPDrr64 (COPY_TO_REGCLASS FR64:$src, VR128))>,
+ (SUBREG_TO_REG (i64 0),
+ (MOVMSKPDrr (COPY_TO_REGCLASS FR64:$src, VR128)), sub_32bit)>,
Requires<[UseSSE2]>;
//===---------------------------------------------------------------------===//
OpndItins itins, bit IsCommutable = 0> {
let Predicates = [HasAVX] in
defm V#NAME : PDI_binop_rm<opc, !strconcat("v", OpcodeStr), Opcode, OpVT128,
- VR128, memopv2i64, i128mem, itins, IsCommutable, 0>, VEX_4V;
+ VR128, loadv2i64, i128mem, itins, IsCommutable, 0>, VEX_4V;
let Constraints = "$src1 = $dst" in
defm NAME : PDI_binop_rm<opc, OpcodeStr, Opcode, OpVT128, VR128,
let Predicates = [HasAVX2] in
defm V#NAME#Y : PDI_binop_rm<opc, !strconcat("v", OpcodeStr), Opcode,
- OpVT256, VR256, memopv4i64, i256mem, itins,
+ OpVT256, VR256, loadv4i64, i256mem, itins,
IsCommutable, 0>, VEX_4V, VEX_L;
}
}
// Alias bitwise logical operations using SSE logical ops on packed FP values.
-defm FsAND : sse12_fp_alias_pack_logical<0x54, "and", X86fand,
- SSE_BIT_ITINS_P>;
-defm FsOR : sse12_fp_alias_pack_logical<0x56, "or", X86for,
- SSE_BIT_ITINS_P>;
-defm FsXOR : sse12_fp_alias_pack_logical<0x57, "xor", X86fxor,
- SSE_BIT_ITINS_P>;
-
-let isCommutable = 0 in
- defm FsANDN : sse12_fp_alias_pack_logical<0x55, "andn", X86fandn,
+let isCodeGenOnly = 1 in {
+ defm FsAND : sse12_fp_alias_pack_logical<0x54, "and", X86fand,
SSE_BIT_ITINS_P>;
+ defm FsOR : sse12_fp_alias_pack_logical<0x56, "or", X86for,
+ SSE_BIT_ITINS_P>;
+ defm FsXOR : sse12_fp_alias_pack_logical<0x57, "xor", X86fxor,
+ SSE_BIT_ITINS_P>;
+
+ let isCommutable = 0 in
+ defm FsANDN : sse12_fp_alias_pack_logical<0x55, "andn", X86fandn,
+ SSE_BIT_ITINS_P>;
+}
/// sse12_fp_packed_logical - SSE 1 & 2 packed FP logical ops
///
!strconcat(OpcodeStr, "ps"), f256mem,
[(set VR256:$dst, (v4i64 (OpNode VR256:$src1, VR256:$src2)))],
[(set VR256:$dst, (OpNode (bc_v4i64 (v8f32 VR256:$src1)),
- (memopv4i64 addr:$src2)))], 0>, TB, VEX_4V, VEX_L;
+ (loadv4i64 addr:$src2)))], 0>, TB, VEX_4V, VEX_L;
defm V#NAME#PDY : sse12_fp_packed_logical_rm<opc, VR256, SSEPackedDouble,
!strconcat(OpcodeStr, "pd"), f256mem,
[(set VR256:$dst, (OpNode (bc_v4i64 (v4f64 VR256:$src1)),
(bc_v4i64 (v4f64 VR256:$src2))))],
[(set VR256:$dst, (OpNode (bc_v4i64 (v4f64 VR256:$src1)),
- (memopv4i64 addr:$src2)))], 0>,
+ (loadv4i64 addr:$src2)))], 0>,
TB, OpSize, VEX_4V, VEX_L;
// In AVX no need to add a pattern for 128-bit logical rr ps, because they
defm V#NAME#PS : sse12_fp_packed_logical_rm<opc, VR128, SSEPackedSingle,
!strconcat(OpcodeStr, "ps"), f128mem, [],
[(set VR128:$dst, (OpNode (bc_v2i64 (v4f32 VR128:$src1)),
- (memopv2i64 addr:$src2)))], 0>, TB, VEX_4V;
+ (loadv2i64 addr:$src2)))], 0>, TB, VEX_4V;
defm V#NAME#PD : sse12_fp_packed_logical_rm<opc, VR128, SSEPackedDouble,
!strconcat(OpcodeStr, "pd"), f128mem,
[(set VR128:$dst, (OpNode (bc_v2i64 (v2f64 VR128:$src1)),
(bc_v2i64 (v2f64 VR128:$src2))))],
[(set VR128:$dst, (OpNode (bc_v2i64 (v2f64 VR128:$src1)),
- (memopv2i64 addr:$src2)))], 0>,
+ (loadv2i64 addr:$src2)))], 0>,
TB, OpSize, VEX_4V;
let Constraints = "$src1 = $dst" in {
multiclass basic_sse12_fp_binop_p<bits<8> opc, string OpcodeStr,
SDNode OpNode, SizeItins itins> {
defm V#NAME#PS : sse12_fp_packed<opc, !strconcat(OpcodeStr, "ps"), OpNode,
- VR128, v4f32, f128mem, memopv4f32,
+ VR128, v4f32, f128mem, loadv4f32,
SSEPackedSingle, itins.s, 0>, TB, VEX_4V;
defm V#NAME#PD : sse12_fp_packed<opc, !strconcat(OpcodeStr, "pd"), OpNode,
- VR128, v2f64, f128mem, memopv2f64,
+ VR128, v2f64, f128mem, loadv2f64,
SSEPackedDouble, itins.d, 0>, TB, OpSize, VEX_4V;
defm V#NAME#PSY : sse12_fp_packed<opc, !strconcat(OpcodeStr, "ps"),
- OpNode, VR256, v8f32, f256mem, memopv8f32,
+ OpNode, VR256, v8f32, f256mem, loadv8f32,
SSEPackedSingle, itins.s, 0>, TB, VEX_4V, VEX_L;
defm V#NAME#PDY : sse12_fp_packed<opc, !strconcat(OpcodeStr, "pd"),
- OpNode, VR256, v4f64, f256mem, memopv4f64,
+ OpNode, VR256, v4f64, f256mem, loadv4f64,
SSEPackedDouble, itins.d, 0>, TB, OpSize, VEX_4V, VEX_L;
let Constraints = "$src1 = $dst" in {
def V#NAME#PSm : PSI<opc, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
!strconcat("v", OpcodeStr,
"ps\t{$src, $dst|$dst, $src}"),
- [(set VR128:$dst, (OpNode (memopv4f32 addr:$src)))],
+ [(set VR128:$dst, (OpNode (loadv4f32 addr:$src)))],
itins.rm>, VEX, Sched<[itins.Sched.Folded]>;
def V#NAME#PSYr : PSI<opc, MRMSrcReg, (outs VR256:$dst), (ins VR256:$src),
!strconcat("v", OpcodeStr,
def V#NAME#PSYm : PSI<opc, MRMSrcMem, (outs VR256:$dst), (ins f256mem:$src),
!strconcat("v", OpcodeStr,
"ps\t{$src, $dst|$dst, $src}"),
- [(set VR256:$dst, (OpNode (memopv8f32 addr:$src)))],
+ [(set VR256:$dst, (OpNode (loadv8f32 addr:$src)))],
itins.rm>, VEX, VEX_L, Sched<[itins.Sched.Folded]>;
}
def V#NAME#PSm_Int : PSI<opc, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
!strconcat("v", OpcodeStr,
"ps\t{$src, $dst|$dst, $src}"),
- [(set VR128:$dst, (V4F32Int (memopv4f32 addr:$src)))],
+ [(set VR128:$dst, (V4F32Int (loadv4f32 addr:$src)))],
itins.rm>, VEX, Sched<[itins.Sched.Folded]>;
def V#NAME#PSYr_Int : PSI<opc, MRMSrcReg, (outs VR256:$dst), (ins VR256:$src),
!strconcat("v", OpcodeStr,
(ins f256mem:$src),
!strconcat("v", OpcodeStr,
"ps\t{$src, $dst|$dst, $src}"),
- [(set VR256:$dst, (V8F32Int (memopv8f32 addr:$src)))],
+ [(set VR256:$dst, (V8F32Int (loadv8f32 addr:$src)))],
itins.rm>, VEX, VEX_L, Sched<[itins.Sched.Folded]>;
}
def V#NAME#PDm : PDI<opc, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
!strconcat("v", OpcodeStr,
"pd\t{$src, $dst|$dst, $src}"),
- [(set VR128:$dst, (OpNode (memopv2f64 addr:$src)))],
+ [(set VR128:$dst, (OpNode (loadv2f64 addr:$src)))],
itins.rm>, VEX, Sched<[itins.Sched.Folded]>;
def V#NAME#PDYr : PDI<opc, MRMSrcReg, (outs VR256:$dst), (ins VR256:$src),
!strconcat("v", OpcodeStr,
def V#NAME#PDYm : PDI<opc, MRMSrcMem, (outs VR256:$dst), (ins f256mem:$src),
!strconcat("v", OpcodeStr,
"pd\t{$src, $dst|$dst, $src}"),
- [(set VR256:$dst, (OpNode (memopv4f64 addr:$src)))],
+ [(set VR256:$dst, (OpNode (loadv4f64 addr:$src)))],
itins.rm>, VEX, VEX_L, Sched<[itins.Sched.Folded]>;
}
def : Pat<(f32 (X86frcp (load addr:$src))),
(VRCPSSm (f32 (IMPLICIT_DEF)), addr:$src)>,
Requires<[HasAVX, OptForSize]>;
-
+}
+let Predicates = [UseAVX] in {
def : Pat<(int_x86_sse_sqrt_ss VR128:$src),
(COPY_TO_REGCLASS (VSQRTSSr (f32 (IMPLICIT_DEF)),
(COPY_TO_REGCLASS VR128:$src, FR32)),
VR128)>;
def : Pat<(int_x86_sse2_sqrt_sd sse_load_f64:$src),
(VSQRTSDm_Int (v2f64 (IMPLICIT_DEF)), sse_load_f64:$src)>;
+}
+let Predicates = [HasAVX] in {
def : Pat<(int_x86_sse_rsqrt_ss VR128:$src),
(COPY_TO_REGCLASS (VRSQRTSSr (f32 (IMPLICIT_DEF)),
(COPY_TO_REGCLASS VR128:$src, FR32)),
// Pause. This "instruction" is encoded as "rep; nop", so even though it
// was introduced with SSE2, it's backward compatible.
-def PAUSE : I<0x90, RawFrm, (outs), (ins), "pause", [], IIC_SSE_PAUSE>, REP;
+def PAUSE : I<0x90, RawFrm, (outs), (ins),
+ "pause", [(int_x86_sse2_pause)], IIC_SSE_PAUSE>,
+ REP, Requires<[HasSSE2]>;
// Load, store, and memory fence
def SFENCE : I<0xAE, MRM_F8, (outs), (ins),
bit IsCommutable = 0> {
let Predicates = [HasAVX] in
defm V#NAME : PDI_binop_rm_int<opc, !strconcat("v", OpcodeStr), IntId128,
- VR128, memopv2i64, i128mem, itins,
+ VR128, loadv2i64, i128mem, itins,
IsCommutable, 0>, VEX_4V;
let Constraints = "$src1 = $dst" in
let Predicates = [HasAVX2] in
defm V#NAME#Y : PDI_binop_rm_int<opc, !strconcat("v", OpcodeStr), IntId256,
- VR256, memopv4i64, i256mem, itins,
+ VR256, loadv4i64, i256mem, itins,
IsCommutable, 0>, VEX_4V, VEX_L;
}
(bc_frag (memopv2i64 addr:$src2)))))], itins.rm>,
Sched<[WriteVecShiftLd, ReadAfterLd]>;
def ri : PDIi8<opc2, ImmForm, (outs RC:$dst),
- (ins RC:$src1, i32i8imm:$src2),
+ (ins RC:$src1, i8imm:$src2),
!if(Is2Addr,
!strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"),
!strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}")),
- [(set RC:$dst, (DstVT (OpNode2 RC:$src1, (i32 imm:$src2))))], itins.ri>,
+ [(set RC:$dst, (DstVT (OpNode2 RC:$src1, (i8 imm:$src2))))], itins.ri>,
Sched<[WriteVecShift]>;
}
defm PAVGW : PDI_binop_all_int<0xE3, "pavgw", int_x86_sse2_pavg_w,
int_x86_avx2_pavg_w, SSE_INTALU_ITINS_P, 1>;
defm PSADBW : PDI_binop_all_int<0xF6, "psadbw", int_x86_sse2_psad_bw,
- int_x86_avx2_psad_bw, SSE_INTALU_ITINS_P, 1>;
+ int_x86_avx2_psad_bw, SSE_PMADD, 1>;
let Predicates = [HasAVX] in
defm VPMULUDQ : PDI_binop_rm2<0xF4, "vpmuludq", X86pmuludq, v2i64, v4i32, VR128,
- memopv2i64, i128mem, SSE_INTMUL_ITINS_P, 1, 0>,
+ loadv2i64, i128mem, SSE_INTMUL_ITINS_P, 1, 0>,
VEX_4V;
let Predicates = [HasAVX2] in
defm VPMULUDQY : PDI_binop_rm2<0xF4, "vpmuludq", X86pmuludq, v4i64, v8i32,
- VR256, memopv4i64, i256mem,
+ VR256, loadv4i64, i256mem,
SSE_INTMUL_ITINS_P, 1, 0>, VEX_4V, VEX_L;
let Constraints = "$src1 = $dst" in
defm PMULUDQ : PDI_binop_rm2<0xF4, "pmuludq", X86pmuludq, v2i64, v4i32, VR128,
(outs VR128:$dst), (ins VR128:$src1, i32i8imm:$src2),
"pslldq\t{$src2, $dst|$dst, $src2}",
[(set VR128:$dst,
- (int_x86_sse2_psll_dq_bs VR128:$src1, imm:$src2))]>;
+ (int_x86_sse2_psll_dq_bs VR128:$src1, imm:$src2))],
+ IIC_SSE_INTSHDQ_P_RI>;
def PSRLDQri : PDIi8<0x73, MRM3r,
(outs VR128:$dst), (ins VR128:$src1, i32i8imm:$src2),
"psrldq\t{$src2, $dst|$dst, $src2}",
[(set VR128:$dst,
- (int_x86_sse2_psrl_dq_bs VR128:$src1, imm:$src2))]>;
+ (int_x86_sse2_psrl_dq_bs VR128:$src1, imm:$src2))],
+ IIC_SSE_INTSHDQ_P_RI>;
// PSRADQri doesn't exist in SSE[1-3].
}
} // Constraints = "$src1 = $dst"
"\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
[(set VR128:$dst,
(vt128 (OpNode VR128:$src1, (i8 imm:$src2))))],
- IIC_SSE_PSHUF>, VEX, Sched<[WriteShuffle]>;
+ IIC_SSE_PSHUF_RI>, VEX, Sched<[WriteShuffle]>;
def V#NAME#mi : Ii8<0x70, MRMSrcMem, (outs VR128:$dst),
(ins i128mem:$src1, i8imm:$src2),
!strconcat("v", OpcodeStr,
"\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
[(set VR128:$dst,
- (vt128 (OpNode (bitconvert (memopv2i64 addr:$src1)),
- (i8 imm:$src2))))], IIC_SSE_PSHUF>, VEX,
+ (vt128 (OpNode (bitconvert (loadv2i64 addr:$src1)),
+ (i8 imm:$src2))))], IIC_SSE_PSHUF_MI>, VEX,
Sched<[WriteShuffleLd]>;
}
"\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
[(set VR256:$dst,
(vt256 (OpNode VR256:$src1, (i8 imm:$src2))))],
- IIC_SSE_PSHUF>, VEX, VEX_L, Sched<[WriteShuffle]>;
+ IIC_SSE_PSHUF_RI>, VEX, VEX_L, Sched<[WriteShuffle]>;
def V#NAME#Ymi : Ii8<0x70, MRMSrcMem, (outs VR256:$dst),
(ins i256mem:$src1, i8imm:$src2),
!strconcat("v", OpcodeStr,
"\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
[(set VR256:$dst,
- (vt256 (OpNode (bitconvert (memopv4i64 addr:$src1)),
- (i8 imm:$src2))))], IIC_SSE_PSHUF>, VEX, VEX_L,
+ (vt256 (OpNode (bitconvert (loadv4i64 addr:$src1)),
+ (i8 imm:$src2))))], IIC_SSE_PSHUF_MI>, VEX, VEX_L,
Sched<[WriteShuffleLd]>;
}
"\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
[(set VR128:$dst,
(vt128 (OpNode VR128:$src1, (i8 imm:$src2))))],
- IIC_SSE_PSHUF>, Sched<[WriteShuffle]>;
+ IIC_SSE_PSHUF_RI>, Sched<[WriteShuffle]>;
def mi : Ii8<0x70, MRMSrcMem,
(outs VR128:$dst), (ins i128mem:$src1, i8imm:$src2),
!strconcat(OpcodeStr,
"\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
[(set VR128:$dst,
(vt128 (OpNode (bitconvert (memopv2i64 addr:$src1)),
- (i8 imm:$src2))))], IIC_SSE_PSHUF>,
+ (i8 imm:$src2))))], IIC_SSE_PSHUF_MI>,
Sched<[WriteShuffleLd]>;
}
}
defm PSHUFLW : sse2_pshuffle<"pshuflw", v8i16, v16i16, X86PShuflw>, XD;
let Predicates = [HasAVX] in {
- def : Pat<(v4f32 (X86PShufd (memopv4f32 addr:$src1), (i8 imm:$imm))),
+ def : Pat<(v4f32 (X86PShufd (loadv4f32 addr:$src1), (i8 imm:$imm))),
(VPSHUFDmi addr:$src1, imm:$imm)>;
def : Pat<(v4f32 (X86PShufd VR128:$src1, (i8 imm:$imm))),
(VPSHUFDri VR128:$src1, imm:$imm)>;
multiclass sse2_pinsrw<bit Is2Addr = 1> {
def rri : Ii8<0xC4, MRMSrcReg,
(outs VR128:$dst), (ins VR128:$src1,
- GR32:$src2, i32i8imm:$src3),
+ GR32orGR64:$src2, i32i8imm:$src3),
!if(Is2Addr,
"pinsrw\t{$src3, $src2, $dst|$dst, $src2, $src3}",
"vpinsrw\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}"),
[(set VR128:$dst,
- (X86pinsrw VR128:$src1, GR32:$src2, imm:$src3))], IIC_SSE_PINSRW>,
- Sched<[WriteShuffle]>;
+ (X86pinsrw VR128:$src1, GR32orGR64:$src2, imm:$src3))],
+ IIC_SSE_PINSRW>, Sched<[WriteShuffle]>;
def rmi : Ii8<0xC4, MRMSrcMem,
(outs VR128:$dst), (ins VR128:$src1,
i16mem:$src2, i32i8imm:$src3),
// Extract
let Predicates = [HasAVX] in
def VPEXTRWri : Ii8<0xC5, MRMSrcReg,
- (outs GR32:$dst), (ins VR128:$src1, i32i8imm:$src2),
+ (outs GR32orGR64:$dst), (ins VR128:$src1, i32i8imm:$src2),
"vpextrw\t{$src2, $src1, $dst|$dst, $src1, $src2}",
- [(set GR32:$dst, (X86pextrw (v8i16 VR128:$src1),
- imm:$src2))]>, TB, OpSize, VEX,
+ [(set GR32orGR64:$dst, (X86pextrw (v8i16 VR128:$src1),
+ imm:$src2))]>, TB, OpSize, VEX,
Sched<[WriteShuffle]>;
def PEXTRWri : PDIi8<0xC5, MRMSrcReg,
- (outs GR32:$dst), (ins VR128:$src1, i32i8imm:$src2),
+ (outs GR32orGR64:$dst), (ins VR128:$src1, i32i8imm:$src2),
"pextrw\t{$src2, $src1, $dst|$dst, $src1, $src2}",
- [(set GR32:$dst, (X86pextrw (v8i16 VR128:$src1),
- imm:$src2))], IIC_SSE_PEXTRW>,
+ [(set GR32orGR64:$dst, (X86pextrw (v8i16 VR128:$src1),
+ imm:$src2))], IIC_SSE_PEXTRW>,
Sched<[WriteShuffleLd, ReadAfterLd]>;
// Insert
-let Predicates = [HasAVX] in {
- defm VPINSRW : sse2_pinsrw<0>, TB, OpSize, VEX_4V;
- def VPINSRWrr64i : Ii8<0xC4, MRMSrcReg, (outs VR128:$dst),
- (ins VR128:$src1, GR64:$src2, i32i8imm:$src3),
- "vpinsrw\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}",
- []>, TB, OpSize, VEX_4V, Sched<[WriteShuffle]>;
-}
+let Predicates = [HasAVX] in
+defm VPINSRW : sse2_pinsrw<0>, TB, OpSize, VEX_4V;
-let Constraints = "$src1 = $dst" in
- defm PINSRW : sse2_pinsrw, TB, OpSize, Requires<[UseSSE2]>;
+let Predicates = [UseSSE2], Constraints = "$src1 = $dst" in
+defm PINSRW : sse2_pinsrw, TB, OpSize;
} // ExeDomain = SSEPackedInt
let ExeDomain = SSEPackedInt, SchedRW = [WriteVecLogic] in {
-def VPMOVMSKBrr : VPDI<0xD7, MRMSrcReg, (outs GR32:$dst), (ins VR128:$src),
+def VPMOVMSKBrr : VPDI<0xD7, MRMSrcReg, (outs GR32orGR64:$dst),
+ (ins VR128:$src),
"pmovmskb\t{$src, $dst|$dst, $src}",
- [(set GR32:$dst, (int_x86_sse2_pmovmskb_128 VR128:$src))],
+ [(set GR32orGR64:$dst, (int_x86_sse2_pmovmskb_128 VR128:$src))],
IIC_SSE_MOVMSK>, VEX;
-def VPMOVMSKBr64r : VPDI<0xD7, MRMSrcReg, (outs GR64:$dst), (ins VR128:$src),
- "pmovmskb\t{$src, $dst|$dst, $src}", [], IIC_SSE_MOVMSK>, VEX;
let Predicates = [HasAVX2] in {
-def VPMOVMSKBYrr : VPDI<0xD7, MRMSrcReg, (outs GR32:$dst), (ins VR256:$src),
+def VPMOVMSKBYrr : VPDI<0xD7, MRMSrcReg, (outs GR32orGR64:$dst),
+ (ins VR256:$src),
"pmovmskb\t{$src, $dst|$dst, $src}",
- [(set GR32:$dst, (int_x86_avx2_pmovmskb VR256:$src))]>, VEX, VEX_L;
-def VPMOVMSKBYr64r : VPDI<0xD7, MRMSrcReg, (outs GR64:$dst), (ins VR256:$src),
- "pmovmskb\t{$src, $dst|$dst, $src}", []>, VEX, VEX_L;
+ [(set GR32orGR64:$dst, (int_x86_avx2_pmovmskb VR256:$src))]>,
+ VEX, VEX_L;
}
-def PMOVMSKBrr : PDI<0xD7, MRMSrcReg, (outs GR32:$dst), (ins VR128:$src),
+def PMOVMSKBrr : PDI<0xD7, MRMSrcReg, (outs GR32orGR64:$dst), (ins VR128:$src),
"pmovmskb\t{$src, $dst|$dst, $src}",
- [(set GR32:$dst, (int_x86_sse2_pmovmskb_128 VR128:$src))],
+ [(set GR32orGR64:$dst, (int_x86_sse2_pmovmskb_128 VR128:$src))],
IIC_SSE_MOVMSK>;
} // ExeDomain = SSEPackedInt
let ExeDomain = SSEPackedInt, SchedRW = [WriteStore] in {
-let Uses = [EDI] in
+let Uses = [EDI], Predicates = [HasAVX,In32BitMode] in
def VMASKMOVDQU : VPDI<0xF7, MRMSrcReg, (outs),
(ins VR128:$src, VR128:$mask),
"maskmovdqu\t{$mask, $src|$src, $mask}",
[(int_x86_sse2_maskmov_dqu VR128:$src, VR128:$mask, EDI)],
IIC_SSE_MASKMOV>, VEX;
-let Uses = [RDI] in
+let Uses = [RDI], Predicates = [HasAVX,In64BitMode] in
def VMASKMOVDQU64 : VPDI<0xF7, MRMSrcReg, (outs),
(ins VR128:$src, VR128:$mask),
"maskmovdqu\t{$mask, $src|$src, $mask}",
[(int_x86_sse2_maskmov_dqu VR128:$src, VR128:$mask, RDI)],
IIC_SSE_MASKMOV>, VEX;
-let Uses = [EDI] in
+let Uses = [EDI], Predicates = [UseSSE2,In32BitMode] in
def MASKMOVDQU : PDI<0xF7, MRMSrcReg, (outs), (ins VR128:$src, VR128:$mask),
"maskmovdqu\t{$mask, $src|$src, $mask}",
[(int_x86_sse2_maskmov_dqu VR128:$src, VR128:$mask, EDI)],
IIC_SSE_MASKMOV>;
-let Uses = [RDI] in
+let Uses = [RDI], Predicates = [UseSSE2,In64BitMode] in
def MASKMOVDQU64 : PDI<0xF7, MRMSrcReg, (outs), (ins VR128:$src, VR128:$mask),
"maskmovdqu\t{$mask, $src|$src, $mask}",
[(int_x86_sse2_maskmov_dqu VR128:$src, VR128:$mask, RDI)],
IIC_SSE_MOVDQ>,
VEX, Sched<[WriteLoad]>;
def VMOV64toPQIrr : VRS2I<0x6E, MRMSrcReg, (outs VR128:$dst), (ins GR64:$src),
- "mov{d|q}\t{$src, $dst|$dst, $src}",
+ "movq\t{$src, $dst|$dst, $src}",
[(set VR128:$dst,
(v2i64 (scalar_to_vector GR64:$src)))],
IIC_SSE_MOVDQ>, VEX, Sched<[WriteMove]>;
+let isCodeGenOnly = 1 in
def VMOV64toSDrr : VRS2I<0x6E, MRMSrcReg, (outs FR64:$dst), (ins GR64:$src),
- "mov{d|q}\t{$src, $dst|$dst, $src}",
+ "movq\t{$src, $dst|$dst, $src}",
[(set FR64:$dst, (bitconvert GR64:$src))],
IIC_SSE_MOVDQ>, VEX, Sched<[WriteMove]>;
[(set VR128:$dst,
(v2i64 (scalar_to_vector GR64:$src)))],
IIC_SSE_MOVDQ>, Sched<[WriteMove]>;
+let isCodeGenOnly = 1 in
def MOV64toSDrr : RS2I<0x6E, MRMSrcReg, (outs FR64:$dst), (ins GR64:$src),
"mov{d|q}\t{$src, $dst|$dst, $src}",
[(set FR64:$dst, (bitconvert GR64:$src))],
//===---------------------------------------------------------------------===//
// Move Int Doubleword to Single Scalar
//
-def VMOVDI2SSrr : VS2I<0x6E, MRMSrcReg, (outs FR32:$dst), (ins GR32:$src),
- "movd\t{$src, $dst|$dst, $src}",
- [(set FR32:$dst, (bitconvert GR32:$src))],
- IIC_SSE_MOVDQ>, VEX, Sched<[WriteMove]>;
-
-def VMOVDI2SSrm : VS2I<0x6E, MRMSrcMem, (outs FR32:$dst), (ins i32mem:$src),
- "movd\t{$src, $dst|$dst, $src}",
- [(set FR32:$dst, (bitconvert (loadi32 addr:$src)))],
- IIC_SSE_MOVDQ>,
- VEX, Sched<[WriteLoad]>;
-def MOVDI2SSrr : S2I<0x6E, MRMSrcReg, (outs FR32:$dst), (ins GR32:$src),
- "movd\t{$src, $dst|$dst, $src}",
- [(set FR32:$dst, (bitconvert GR32:$src))],
- IIC_SSE_MOVDQ>, Sched<[WriteMove]>;
-
-def MOVDI2SSrm : S2I<0x6E, MRMSrcMem, (outs FR32:$dst), (ins i32mem:$src),
- "movd\t{$src, $dst|$dst, $src}",
- [(set FR32:$dst, (bitconvert (loadi32 addr:$src)))],
- IIC_SSE_MOVDQ>, Sched<[WriteLoad]>;
+let isCodeGenOnly = 1 in {
+ def VMOVDI2SSrr : VS2I<0x6E, MRMSrcReg, (outs FR32:$dst), (ins GR32:$src),
+ "movd\t{$src, $dst|$dst, $src}",
+ [(set FR32:$dst, (bitconvert GR32:$src))],
+ IIC_SSE_MOVDQ>, VEX, Sched<[WriteMove]>;
+
+ def VMOVDI2SSrm : VS2I<0x6E, MRMSrcMem, (outs FR32:$dst), (ins i32mem:$src),
+ "movd\t{$src, $dst|$dst, $src}",
+ [(set FR32:$dst, (bitconvert (loadi32 addr:$src)))],
+ IIC_SSE_MOVDQ>,
+ VEX, Sched<[WriteLoad]>;
+ def MOVDI2SSrr : S2I<0x6E, MRMSrcReg, (outs FR32:$dst), (ins GR32:$src),
+ "movd\t{$src, $dst|$dst, $src}",
+ [(set FR32:$dst, (bitconvert GR32:$src))],
+ IIC_SSE_MOVDQ>, Sched<[WriteMove]>;
+
+ def MOVDI2SSrm : S2I<0x6E, MRMSrcMem, (outs FR32:$dst), (ins i32mem:$src),
+ "movd\t{$src, $dst|$dst, $src}",
+ [(set FR32:$dst, (bitconvert (loadi32 addr:$src)))],
+ IIC_SSE_MOVDQ>, Sched<[WriteLoad]>;
+}
//===---------------------------------------------------------------------===//
// Move Packed Doubleword Int to Packed Double Int
//
let SchedRW = [WriteMove] in {
def VMOVPQIto64rr : VRS2I<0x7E, MRMDestReg, (outs GR64:$dst), (ins VR128:$src),
- "mov{d|q}\t{$src, $dst|$dst, $src}",
+ "movq\t{$src, $dst|$dst, $src}",
[(set GR64:$dst, (vector_extract (v2i64 VR128:$src),
(iPTR 0)))],
IIC_SSE_MOVD_ToGP>,
//===---------------------------------------------------------------------===//
// Bitcast FR64 <-> GR64
//
-let Predicates = [UseAVX] in
-def VMOV64toSDrm : VS2SI<0x7E, MRMSrcMem, (outs FR64:$dst), (ins i64mem:$src),
- "vmovq\t{$src, $dst|$dst, $src}",
- [(set FR64:$dst, (bitconvert (loadi64 addr:$src)))]>,
- VEX, Sched<[WriteLoad]>;
-def VMOVSDto64rr : VRS2I<0x7E, MRMDestReg, (outs GR64:$dst), (ins FR64:$src),
+let isCodeGenOnly = 1 in {
+ let Predicates = [UseAVX] in
+ def VMOV64toSDrm : VS2SI<0x7E, MRMSrcMem, (outs FR64:$dst), (ins i64mem:$src),
+ "movq\t{$src, $dst|$dst, $src}",
+ [(set FR64:$dst, (bitconvert (loadi64 addr:$src)))]>,
+ VEX, Sched<[WriteLoad]>;
+ def VMOVSDto64rr : VRS2I<0x7E, MRMDestReg, (outs GR64:$dst), (ins FR64:$src),
+ "movq\t{$src, $dst|$dst, $src}",
+ [(set GR64:$dst, (bitconvert FR64:$src))],
+ IIC_SSE_MOVDQ>, VEX, Sched<[WriteMove]>;
+ def VMOVSDto64mr : VRS2I<0x7E, MRMDestMem, (outs), (ins i64mem:$dst, FR64:$src),
+ "movq\t{$src, $dst|$dst, $src}",
+ [(store (i64 (bitconvert FR64:$src)), addr:$dst)],
+ IIC_SSE_MOVDQ>, VEX, Sched<[WriteStore]>;
+
+ def MOV64toSDrm : S2SI<0x7E, MRMSrcMem, (outs FR64:$dst), (ins i64mem:$src),
+ "movq\t{$src, $dst|$dst, $src}",
+ [(set FR64:$dst, (bitconvert (loadi64 addr:$src)))],
+ IIC_SSE_MOVDQ>, Sched<[WriteLoad]>;
+ def MOVSDto64rr : RS2I<0x7E, MRMDestReg, (outs GR64:$dst), (ins FR64:$src),
"mov{d|q}\t{$src, $dst|$dst, $src}",
[(set GR64:$dst, (bitconvert FR64:$src))],
- IIC_SSE_MOVDQ>, VEX, Sched<[WriteMove]>;
-def VMOVSDto64mr : VRS2I<0x7E, MRMDestMem, (outs), (ins i64mem:$dst, FR64:$src),
+ IIC_SSE_MOVD_ToGP>, Sched<[WriteMove]>;
+ def MOVSDto64mr : RS2I<0x7E, MRMDestMem, (outs), (ins i64mem:$dst, FR64:$src),
"movq\t{$src, $dst|$dst, $src}",
[(store (i64 (bitconvert FR64:$src)), addr:$dst)],
- IIC_SSE_MOVDQ>, VEX, Sched<[WriteStore]>;
-
-def MOV64toSDrm : S2SI<0x7E, MRMSrcMem, (outs FR64:$dst), (ins i64mem:$src),
- "movq\t{$src, $dst|$dst, $src}",
- [(set FR64:$dst, (bitconvert (loadi64 addr:$src)))],
- IIC_SSE_MOVDQ>, Sched<[WriteLoad]>;
-def MOVSDto64rr : RS2I<0x7E, MRMDestReg, (outs GR64:$dst), (ins FR64:$src),
- "mov{d|q}\t{$src, $dst|$dst, $src}",
- [(set GR64:$dst, (bitconvert FR64:$src))],
- IIC_SSE_MOVD_ToGP>, Sched<[WriteMove]>;
-def MOVSDto64mr : RS2I<0x7E, MRMDestMem, (outs), (ins i64mem:$dst, FR64:$src),
- "movq\t{$src, $dst|$dst, $src}",
- [(store (i64 (bitconvert FR64:$src)), addr:$dst)],
- IIC_SSE_MOVDQ>, Sched<[WriteStore]>;
+ IIC_SSE_MOVDQ>, Sched<[WriteStore]>;
+}
//===---------------------------------------------------------------------===//
// Move Scalar Single to Double Int
//
-def VMOVSS2DIrr : VS2I<0x7E, MRMDestReg, (outs GR32:$dst), (ins FR32:$src),
- "movd\t{$src, $dst|$dst, $src}",
- [(set GR32:$dst, (bitconvert FR32:$src))],
- IIC_SSE_MOVD_ToGP>, VEX, Sched<[WriteMove]>;
-def VMOVSS2DImr : VS2I<0x7E, MRMDestMem, (outs), (ins i32mem:$dst, FR32:$src),
- "movd\t{$src, $dst|$dst, $src}",
- [(store (i32 (bitconvert FR32:$src)), addr:$dst)],
- IIC_SSE_MOVDQ>, VEX, Sched<[WriteStore]>;
-def MOVSS2DIrr : S2I<0x7E, MRMDestReg, (outs GR32:$dst), (ins FR32:$src),
- "movd\t{$src, $dst|$dst, $src}",
- [(set GR32:$dst, (bitconvert FR32:$src))],
- IIC_SSE_MOVD_ToGP>, Sched<[WriteMove]>;
-def MOVSS2DImr : S2I<0x7E, MRMDestMem, (outs), (ins i32mem:$dst, FR32:$src),
- "movd\t{$src, $dst|$dst, $src}",
- [(store (i32 (bitconvert FR32:$src)), addr:$dst)],
- IIC_SSE_MOVDQ>, Sched<[WriteStore]>;
+let isCodeGenOnly = 1 in {
+ def VMOVSS2DIrr : VS2I<0x7E, MRMDestReg, (outs GR32:$dst), (ins FR32:$src),
+ "movd\t{$src, $dst|$dst, $src}",
+ [(set GR32:$dst, (bitconvert FR32:$src))],
+ IIC_SSE_MOVD_ToGP>, VEX, Sched<[WriteMove]>;
+ def VMOVSS2DImr : VS2I<0x7E, MRMDestMem, (outs), (ins i32mem:$dst, FR32:$src),
+ "movd\t{$src, $dst|$dst, $src}",
+ [(store (i32 (bitconvert FR32:$src)), addr:$dst)],
+ IIC_SSE_MOVDQ>, VEX, Sched<[WriteStore]>;
+ def MOVSS2DIrr : S2I<0x7E, MRMDestReg, (outs GR32:$dst), (ins FR32:$src),
+ "movd\t{$src, $dst|$dst, $src}",
+ [(set GR32:$dst, (bitconvert FR32:$src))],
+ IIC_SSE_MOVD_ToGP>, Sched<[WriteMove]>;
+ def MOVSS2DImr : S2I<0x7E, MRMDestMem, (outs), (ins i32mem:$dst, FR32:$src),
+ "movd\t{$src, $dst|$dst, $src}",
+ [(store (i32 (bitconvert FR32:$src)), addr:$dst)],
+ IIC_SSE_MOVDQ>, Sched<[WriteStore]>;
+}
//===---------------------------------------------------------------------===//
// Patterns and instructions to describe movd/movq to XMM register zero-extends
//
-let SchedRW = [WriteMove] in {
+let isCodeGenOnly = 1, SchedRW = [WriteMove] in {
let AddedComplexity = 15 in {
-def VMOVZDI2PDIrr : VS2I<0x6E, MRMSrcReg, (outs VR128:$dst), (ins GR32:$src),
- "movd\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst, (v4i32 (X86vzmovl
- (v4i32 (scalar_to_vector GR32:$src)))))],
- IIC_SSE_MOVDQ>, VEX;
def VMOVZQI2PQIrr : VS2I<0x6E, MRMSrcReg, (outs VR128:$dst), (ins GR64:$src),
- "mov{d|q}\t{$src, $dst|$dst, $src}", // X86-64 only
+ "movq\t{$src, $dst|$dst, $src}", // X86-64 only
[(set VR128:$dst, (v2i64 (X86vzmovl
(v2i64 (scalar_to_vector GR64:$src)))))],
IIC_SSE_MOVDQ>,
VEX, VEX_W;
-}
-let AddedComplexity = 15 in {
-def MOVZDI2PDIrr : S2I<0x6E, MRMSrcReg, (outs VR128:$dst), (ins GR32:$src),
- "movd\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst, (v4i32 (X86vzmovl
- (v4i32 (scalar_to_vector GR32:$src)))))],
- IIC_SSE_MOVDQ>;
def MOVZQI2PQIrr : RS2I<0x6E, MRMSrcReg, (outs VR128:$dst), (ins GR64:$src),
"mov{d|q}\t{$src, $dst|$dst, $src}", // X86-64 only
[(set VR128:$dst, (v2i64 (X86vzmovl
(v2i64 (scalar_to_vector GR64:$src)))))],
IIC_SSE_MOVDQ>;
}
-} // SchedRW
-
-let AddedComplexity = 20, SchedRW = [WriteLoad] in {
-def VMOVZDI2PDIrm : VS2I<0x6E, MRMSrcMem, (outs VR128:$dst), (ins i32mem:$src),
- "movd\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst,
- (v4i32 (X86vzmovl (v4i32 (scalar_to_vector
- (loadi32 addr:$src))))))],
- IIC_SSE_MOVDQ>, VEX;
-def MOVZDI2PDIrm : S2I<0x6E, MRMSrcMem, (outs VR128:$dst), (ins i32mem:$src),
- "movd\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst,
- (v4i32 (X86vzmovl (v4i32 (scalar_to_vector
- (loadi32 addr:$src))))))],
- IIC_SSE_MOVDQ>;
-} // AddedComplexity, SchedRW
+} // isCodeGenOnly, SchedRW
let Predicates = [UseAVX] in {
+ let AddedComplexity = 15 in
+ def : Pat<(v4i32 (X86vzmovl (v4i32 (scalar_to_vector GR32:$src)))),
+ (VMOVDI2PDIrr GR32:$src)>;
+
// AVX 128-bit movd/movq instruction write zeros in the high 128-bit part.
let AddedComplexity = 20 in {
+ def : Pat<(v4i32 (X86vzmovl (v4i32 (scalar_to_vector (loadi32 addr:$src))))),
+ (VMOVDI2PDIrm addr:$src)>;
def : Pat<(v4i32 (X86vzmovl (bc_v4i32 (loadv4f32 addr:$src)))),
- (VMOVZDI2PDIrm addr:$src)>;
+ (VMOVDI2PDIrm addr:$src)>;
def : Pat<(v4i32 (X86vzmovl (bc_v4i32 (loadv2i64 addr:$src)))),
- (VMOVZDI2PDIrm addr:$src)>;
+ (VMOVDI2PDIrm addr:$src)>;
}
// Use regular 128-bit instructions to match 256-bit scalar_to_vec+zext.
def : Pat<(v8i32 (X86vzmovl (insert_subvector undef,
(v4i32 (scalar_to_vector GR32:$src)),(iPTR 0)))),
- (SUBREG_TO_REG (i32 0), (VMOVZDI2PDIrr GR32:$src), sub_xmm)>;
+ (SUBREG_TO_REG (i32 0), (VMOVDI2PDIrr GR32:$src), sub_xmm)>;
def : Pat<(v4i64 (X86vzmovl (insert_subvector undef,
(v2i64 (scalar_to_vector GR64:$src)),(iPTR 0)))),
(SUBREG_TO_REG (i64 0), (VMOVZQI2PQIrr GR64:$src), sub_xmm)>;
}
-let Predicates = [UseSSE2], AddedComplexity = 20 in {
- def : Pat<(v4i32 (X86vzmovl (bc_v4i32 (loadv4f32 addr:$src)))),
- (MOVZDI2PDIrm addr:$src)>;
- def : Pat<(v4i32 (X86vzmovl (bc_v4i32 (loadv2i64 addr:$src)))),
- (MOVZDI2PDIrm addr:$src)>;
+let Predicates = [UseSSE2] in {
+ let AddedComplexity = 15 in
+ def : Pat<(v4i32 (X86vzmovl (v4i32 (scalar_to_vector GR32:$src)))),
+ (MOVDI2PDIrr GR32:$src)>;
+
+ let AddedComplexity = 20 in {
+ def : Pat<(v4i32 (X86vzmovl (v4i32 (scalar_to_vector (loadi32 addr:$src))))),
+ (MOVDI2PDIrm addr:$src)>;
+ def : Pat<(v4i32 (X86vzmovl (bc_v4i32 (loadv4f32 addr:$src)))),
+ (MOVDI2PDIrm addr:$src)>;
+ def : Pat<(v4i32 (X86vzmovl (bc_v4i32 (loadv2i64 addr:$src)))),
+ (MOVDI2PDIrm addr:$src)>;
+ }
}
// These are the correct encodings of the instructions so that we know how to
// compatibility with Darwin's buggy assembler.
def : InstAlias<"movq\t{$src, $dst|$dst, $src}",
(MOV64toPQIrr VR128:$dst, GR64:$src), 0>;
-def : InstAlias<"movq\t{$src, $dst|$dst, $src}",
- (MOV64toSDrr FR64:$dst, GR64:$src), 0>;
def : InstAlias<"movq\t{$src, $dst|$dst, $src}",
(MOVPQIto64rr GR64:$dst, VR128:$src), 0>;
-def : InstAlias<"movq\t{$src, $dst|$dst, $src}",
- (MOVSDto64rr GR64:$dst, FR64:$src), 0>;
-def : InstAlias<"movq\t{$src, $dst|$dst, $src}",
- (VMOVZQI2PQIrr VR128:$dst, GR64:$src), 0>;
-def : InstAlias<"movq\t{$src, $dst|$dst, $src}",
- (MOVZQI2PQIrr VR128:$dst, GR64:$src), 0>;
+// Allow "vmovd" but print "vmovq" since we don't need compatibility for AVX.
+def : InstAlias<"vmovd\t{$src, $dst|$dst, $src}",
+ (VMOV64toPQIrr VR128:$dst, GR64:$src), 0>;
+def : InstAlias<"vmovd\t{$src, $dst|$dst, $src}",
+ (VMOVPQIto64rr GR64:$dst, VR128:$src), 0>;
//===---------------------------------------------------------------------===//
// SSE2 - Move Quadword
[(int_x86_sse2_storel_dq addr:$dst, VR128:$src)],
IIC_SSE_MOVDQ>, Sched<[WriteStore]>;
-let AddedComplexity = 20 in
+let isCodeGenOnly = 1, AddedComplexity = 20 in {
def VMOVZQI2PQIrm : I<0x7E, MRMSrcMem, (outs VR128:$dst), (ins i64mem:$src),
"vmovq\t{$src, $dst|$dst, $src}",
[(set VR128:$dst,
IIC_SSE_MOVDQ>,
XS, VEX, Requires<[UseAVX]>, Sched<[WriteLoad]>;
-let AddedComplexity = 20 in
def MOVZQI2PQIrm : I<0x7E, MRMSrcMem, (outs VR128:$dst), (ins i64mem:$src),
"movq\t{$src, $dst|$dst, $src}",
[(set VR128:$dst,
(loadi64 addr:$src))))))],
IIC_SSE_MOVDQ>,
XS, Requires<[UseSSE2]>, Sched<[WriteLoad]>;
+}
let Predicates = [UseAVX], AddedComplexity = 20 in {
- def : Pat<(v2i64 (X86vzmovl (loadv2i64 addr:$src))),
- (VMOVZQI2PQIrm addr:$src)>;
def : Pat<(v2i64 (X86vzmovl (bc_v2i64 (loadv4f32 addr:$src)))),
(VMOVZQI2PQIrm addr:$src)>;
def : Pat<(v2i64 (X86vzload addr:$src)),
}
let Predicates = [UseSSE2], AddedComplexity = 20 in {
- def : Pat<(v2i64 (X86vzmovl (loadv2i64 addr:$src))),
- (MOVZQI2PQIrm addr:$src)>;
def : Pat<(v2i64 (X86vzmovl (bc_v2i64 (loadv4f32 addr:$src)))),
(MOVZQI2PQIrm addr:$src)>;
def : Pat<(v2i64 (X86vzload addr:$src)), (MOVZQI2PQIrm addr:$src)>;
XS, Requires<[UseSSE2]>;
} // SchedRW
-let SchedRW = [WriteVecLogicLd] in {
+let isCodeGenOnly = 1, SchedRW = [WriteVecLogicLd] in {
let AddedComplexity = 20 in
def VMOVZPQILo2PQIrm : I<0x7E, MRMSrcMem, (outs VR128:$dst), (ins i128mem:$src),
"vmovq\t{$src, $dst|$dst, $src}",
IIC_SSE_MOVDQ>,
XS, Requires<[UseSSE2]>;
}
-} // SchedRW
+} // isCodeGenOnly, SchedRW
let AddedComplexity = 20 in {
let Predicates = [UseAVX] in {
- def : Pat<(v2i64 (X86vzmovl (loadv2i64 addr:$src))),
- (VMOVZPQILo2PQIrm addr:$src)>;
def : Pat<(v2f64 (X86vzmovl (v2f64 VR128:$src))),
(VMOVZPQILo2PQIrr VR128:$src)>;
}
let Predicates = [UseSSE2] in {
- def : Pat<(v2i64 (X86vzmovl (loadv2i64 addr:$src))),
- (MOVZPQILo2PQIrm addr:$src)>;
def : Pat<(v2f64 (X86vzmovl (v2f64 VR128:$src))),
(MOVZPQILo2PQIrr VR128:$src)>;
}
}
-// Instructions to match in the assembler
-let SchedRW = [WriteMove] in {
-def VMOVQs64rr : VS2I<0x6E, MRMSrcReg, (outs VR128:$dst), (ins GR64:$src),
- "movq\t{$src, $dst|$dst, $src}", [],
- IIC_SSE_MOVDQ>, VEX, VEX_W;
-def VMOVQd64rr : VS2I<0x7E, MRMDestReg, (outs GR64:$dst), (ins VR128:$src),
- "movq\t{$src, $dst|$dst, $src}", [],
- IIC_SSE_MOVDQ>, VEX, VEX_W;
-// Recognize "movd" with GR64 destination, but encode as a "movq"
-def VMOVQd64rr_alt : VS2I<0x7E, MRMDestReg, (outs GR64:$dst), (ins VR128:$src),
- "movq\t{$src, $dst|$dst, $src}", [],
- IIC_SSE_MOVDQ>, VEX, VEX_W;
-} // SchedRW
-
-// Instructions for the disassembler
-// xr = XMM register
-// xm = mem64
-
-let SchedRW = [WriteMove] in {
-let Predicates = [UseAVX] in
-def VMOVQxrxr: I<0x7E, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
- "vmovq\t{$src, $dst|$dst, $src}", []>, VEX, XS;
-def MOVQxrxr : I<0x7E, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
- "movq\t{$src, $dst|$dst, $src}", [], IIC_SSE_MOVQ_RR>, XS;
-} // SchedRW
-
//===---------------------------------------------------------------------===//
// SSE3 - Replicate Single FP - MOVSHDUP and MOVSLDUP
//===---------------------------------------------------------------------===//
let Predicates = [HasAVX] in {
defm VMOVSHDUP : sse3_replicate_sfp<0x16, X86Movshdup, "vmovshdup",
- v4f32, VR128, memopv4f32, f128mem>, VEX;
+ v4f32, VR128, loadv4f32, f128mem>, VEX;
defm VMOVSLDUP : sse3_replicate_sfp<0x12, X86Movsldup, "vmovsldup",
- v4f32, VR128, memopv4f32, f128mem>, VEX;
+ v4f32, VR128, loadv4f32, f128mem>, VEX;
defm VMOVSHDUPY : sse3_replicate_sfp<0x16, X86Movshdup, "vmovshdup",
- v8f32, VR256, memopv8f32, f256mem>, VEX, VEX_L;
+ v8f32, VR256, loadv8f32, f256mem>, VEX, VEX_L;
defm VMOVSLDUPY : sse3_replicate_sfp<0x12, X86Movsldup, "vmovsldup",
- v8f32, VR256, memopv8f32, f256mem>, VEX, VEX_L;
+ v8f32, VR256, loadv8f32, f256mem>, VEX, VEX_L;
}
defm MOVSHDUP : sse3_replicate_sfp<0x16, X86Movshdup, "movshdup", v4f32, VR128,
memopv4f32, f128mem>;
let Predicates = [HasAVX] in {
def : Pat<(v4i32 (X86Movshdup VR128:$src)),
(VMOVSHDUPrr VR128:$src)>;
- def : Pat<(v4i32 (X86Movshdup (bc_v4i32 (memopv2i64 addr:$src)))),
+ def : Pat<(v4i32 (X86Movshdup (bc_v4i32 (loadv2i64 addr:$src)))),
(VMOVSHDUPrm addr:$src)>;
def : Pat<(v4i32 (X86Movsldup VR128:$src)),
(VMOVSLDUPrr VR128:$src)>;
- def : Pat<(v4i32 (X86Movsldup (bc_v4i32 (memopv2i64 addr:$src)))),
+ def : Pat<(v4i32 (X86Movsldup (bc_v4i32 (loadv2i64 addr:$src)))),
(VMOVSLDUPrm addr:$src)>;
def : Pat<(v8i32 (X86Movshdup VR256:$src)),
(VMOVSHDUPYrr VR256:$src)>;
- def : Pat<(v8i32 (X86Movshdup (bc_v8i32 (memopv4i64 addr:$src)))),
+ def : Pat<(v8i32 (X86Movshdup (bc_v8i32 (loadv4i64 addr:$src)))),
(VMOVSHDUPYrm addr:$src)>;
def : Pat<(v8i32 (X86Movsldup VR256:$src)),
(VMOVSLDUPYrr VR256:$src)>;
- def : Pat<(v8i32 (X86Movsldup (bc_v8i32 (memopv4i64 addr:$src)))),
+ def : Pat<(v8i32 (X86Movsldup (bc_v8i32 (loadv4i64 addr:$src)))),
(VMOVSLDUPYrm addr:$src)>;
}
defm MOVDDUP : sse3_replicate_dfp<"movddup">;
let Predicates = [HasAVX] in {
- def : Pat<(X86Movddup (memopv2f64 addr:$src)),
+ def : Pat<(X86Movddup (loadv2f64 addr:$src)),
(VMOVDDUPrm addr:$src)>, Requires<[HasAVX]>;
- def : Pat<(X86Movddup (bc_v2f64 (memopv4f32 addr:$src))),
+ def : Pat<(X86Movddup (bc_v2f64 (loadv4f32 addr:$src))),
(VMOVDDUPrm addr:$src)>, Requires<[HasAVX]>;
- def : Pat<(X86Movddup (bc_v2f64 (memopv2i64 addr:$src))),
+ def : Pat<(X86Movddup (bc_v2f64 (loadv2i64 addr:$src))),
(VMOVDDUPrm addr:$src)>, Requires<[HasAVX]>;
def : Pat<(X86Movddup (bc_v2f64
(v2i64 (scalar_to_vector (loadi64 addr:$src))))),
(VMOVDDUPrm addr:$src)>, Requires<[HasAVX]>;
// 256-bit version
- def : Pat<(X86Movddup (memopv4f64 addr:$src)),
+ def : Pat<(X86Movddup (loadv4f64 addr:$src)),
(VMOVDDUPYrm addr:$src)>;
- def : Pat<(X86Movddup (memopv4i64 addr:$src)),
+ def : Pat<(X86Movddup (loadv4i64 addr:$src)),
(VMOVDDUPYrm addr:$src)>;
def : Pat<(X86Movddup (v4i64 (scalar_to_vector (loadi64 addr:$src)))),
(VMOVDDUPYrm addr:$src)>;
// Helper fragments to match sext vXi1 to vXiY.
def v16i1sextv16i8 : PatLeaf<(v16i8 (X86pcmpgt (bc_v16i8 (v4i32 immAllZerosV)),
VR128:$src))>;
-def v8i1sextv8i16 : PatLeaf<(v8i16 (X86vsrai VR128:$src, (i32 15)))>;
-def v4i1sextv4i32 : PatLeaf<(v4i32 (X86vsrai VR128:$src, (i32 31)))>;
+def v8i1sextv8i16 : PatLeaf<(v8i16 (X86vsrai VR128:$src, (i8 15)))>;
+def v4i1sextv4i32 : PatLeaf<(v4i32 (X86vsrai VR128:$src, (i8 31)))>;
def v32i1sextv32i8 : PatLeaf<(v32i8 (X86pcmpgt (bc_v32i8 (v8i32 immAllZerosV)),
VR256:$src))>;
-def v16i1sextv16i16: PatLeaf<(v16i16 (X86vsrai VR256:$src, (i32 15)))>;
-def v8i1sextv8i32 : PatLeaf<(v8i32 (X86vsrai VR256:$src, (i32 31)))>;
+def v16i1sextv16i16: PatLeaf<(v16i16 (X86vsrai VR256:$src, (i8 15)))>;
+def v8i1sextv8i32 : PatLeaf<(v8i32 (X86vsrai VR256:$src, (i8 31)))>;
let Predicates = [HasAVX] in {
defm VPABSB : SS3I_unop_rm_int<0x1C, "vpabsb",
!strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
[(set VR256:$dst,
(IntId256 VR256:$src1,
- (bitconvert (memopv4i64 addr:$src2))))]>, OpSize;
+ (bitconvert (loadv4i64 addr:$src2))))]>, OpSize;
}
let ImmT = NoImm, Predicates = [HasAVX] in {
let isCommutable = 0 in {
defm VPHADDW : SS3I_binop_rm<0x01, "vphaddw", X86hadd, v8i16, VR128,
- memopv2i64, i128mem,
+ loadv2i64, i128mem,
SSE_PHADDSUBW, 0>, VEX_4V;
defm VPHADDD : SS3I_binop_rm<0x02, "vphaddd", X86hadd, v4i32, VR128,
- memopv2i64, i128mem,
+ loadv2i64, i128mem,
SSE_PHADDSUBD, 0>, VEX_4V;
defm VPHSUBW : SS3I_binop_rm<0x05, "vphsubw", X86hsub, v8i16, VR128,
- memopv2i64, i128mem,
+ loadv2i64, i128mem,
SSE_PHADDSUBW, 0>, VEX_4V;
defm VPHSUBD : SS3I_binop_rm<0x06, "vphsubd", X86hsub, v4i32, VR128,
- memopv2i64, i128mem,
+ loadv2i64, i128mem,
SSE_PHADDSUBD, 0>, VEX_4V;
defm VPSIGNB : SS3I_binop_rm<0x08, "vpsignb", X86psign, v16i8, VR128,
- memopv2i64, i128mem,
+ loadv2i64, i128mem,
SSE_PSIGN, 0>, VEX_4V;
defm VPSIGNW : SS3I_binop_rm<0x09, "vpsignw", X86psign, v8i16, VR128,
- memopv2i64, i128mem,
+ loadv2i64, i128mem,
SSE_PSIGN, 0>, VEX_4V;
defm VPSIGND : SS3I_binop_rm<0x0A, "vpsignd", X86psign, v4i32, VR128,
- memopv2i64, i128mem,
+ loadv2i64, i128mem,
SSE_PSIGN, 0>, VEX_4V;
defm VPSHUFB : SS3I_binop_rm<0x00, "vpshufb", X86pshufb, v16i8, VR128,
- memopv2i64, i128mem,
+ loadv2i64, i128mem,
SSE_PSHUFB, 0>, VEX_4V;
defm VPHADDSW : SS3I_binop_rm_int<0x03, "vphaddsw",
int_x86_ssse3_phadd_sw_128,
let ImmT = NoImm, Predicates = [HasAVX2] in {
let isCommutable = 0 in {
defm VPHADDWY : SS3I_binop_rm<0x01, "vphaddw", X86hadd, v16i16, VR256,
- memopv4i64, i256mem,
+ loadv4i64, i256mem,
SSE_PHADDSUBW, 0>, VEX_4V, VEX_L;
defm VPHADDDY : SS3I_binop_rm<0x02, "vphaddd", X86hadd, v8i32, VR256,
- memopv4i64, i256mem,
+ loadv4i64, i256mem,
SSE_PHADDSUBW, 0>, VEX_4V, VEX_L;
defm VPHSUBWY : SS3I_binop_rm<0x05, "vphsubw", X86hsub, v16i16, VR256,
- memopv4i64, i256mem,
+ loadv4i64, i256mem,
SSE_PHADDSUBW, 0>, VEX_4V, VEX_L;
defm VPHSUBDY : SS3I_binop_rm<0x06, "vphsubd", X86hsub, v8i32, VR256,
- memopv4i64, i256mem,
+ loadv4i64, i256mem,
SSE_PHADDSUBW, 0>, VEX_4V, VEX_L;
defm VPSIGNBY : SS3I_binop_rm<0x08, "vpsignb", X86psign, v32i8, VR256,
- memopv4i64, i256mem,
+ loadv4i64, i256mem,
SSE_PHADDSUBW, 0>, VEX_4V, VEX_L;
defm VPSIGNWY : SS3I_binop_rm<0x09, "vpsignw", X86psign, v16i16, VR256,
- memopv4i64, i256mem,
+ loadv4i64, i256mem,
SSE_PHADDSUBW, 0>, VEX_4V, VEX_L;
defm VPSIGNDY : SS3I_binop_rm<0x0A, "vpsignd", X86psign, v8i32, VR256,
- memopv4i64, i256mem,
+ loadv4i64, i256mem,
SSE_PHADDSUBW, 0>, VEX_4V, VEX_L;
defm VPSHUFBY : SS3I_binop_rm<0x00, "vpshufb", X86pshufb, v32i8, VR256,
- memopv4i64, i256mem,
+ loadv4i64, i256mem,
SSE_PHADDSUBW, 0>, VEX_4V, VEX_L;
defm VPHADDSW : SS3I_binop_rm_int_y<0x03, "vphaddsw",
int_x86_avx2_phadd_sw>, VEX_4V, VEX_L;
!strconcat(asm, "\t{$src3, $src2, $dst|$dst, $src2, $src3}"),
!strconcat(asm,
"\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}")),
- [], IIC_SSE_PALIGNR>, OpSize, Sched<[WriteShuffle]>;
+ [], IIC_SSE_PALIGNRR>, OpSize, Sched<[WriteShuffle]>;
let mayLoad = 1 in
def R128rm : SS3AI<0x0F, MRMSrcMem, (outs VR128:$dst),
(ins VR128:$src1, i128mem:$src2, i8imm:$src3),
!strconcat(asm, "\t{$src3, $src2, $dst|$dst, $src2, $src3}"),
!strconcat(asm,
"\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}")),
- [], IIC_SSE_PALIGNR>, OpSize, Sched<[WriteShuffleLd, ReadAfterLd]>;
+ [], IIC_SSE_PALIGNRM>, OpSize, Sched<[WriteShuffleLd, ReadAfterLd]>;
}
}
// SSE4.1 - Packed Move with Sign/Zero Extend
//===----------------------------------------------------------------------===//
-multiclass SS41I_binop_rm_int8<bits<8> opc, string OpcodeStr, Intrinsic IntId> {
+multiclass SS41I_binop_rm_int8<bits<8> opc, string OpcodeStr, Intrinsic IntId,
+ OpndItins itins = DEFAULT_ITINS> {
def rr : SS48I<opc, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
!strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"),
- [(set VR128:$dst, (IntId VR128:$src))]>, OpSize;
+ [(set VR128:$dst, (IntId VR128:$src))], itins.rr>, OpSize;
def rm : SS48I<opc, MRMSrcMem, (outs VR128:$dst), (ins i64mem:$src),
!strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"),
[(set VR128:$dst,
- (IntId (bitconvert (v2i64 (scalar_to_vector (loadi64 addr:$src))))))]>,
- OpSize;
+ (IntId (bitconvert (v2i64 (scalar_to_vector (loadi64 addr:$src))))))],
+ itins.rm>, OpSize;
}
multiclass SS41I_binop_rm_int16_y<bits<8> opc, string OpcodeStr,
def Yrm : SS48I<opc, MRMSrcMem, (outs VR256:$dst), (ins i128mem:$src),
!strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"),
- [(set VR256:$dst, (IntId (load addr:$src)))]>, OpSize;
+ [(set VR256:$dst, (IntId (load addr:$src)))]>,
+ OpSize;
}
let Predicates = [HasAVX] in {
-defm VPMOVSXBW : SS41I_binop_rm_int8<0x20, "vpmovsxbw", int_x86_sse41_pmovsxbw>,
- VEX;
-defm VPMOVSXWD : SS41I_binop_rm_int8<0x23, "vpmovsxwd", int_x86_sse41_pmovsxwd>,
- VEX;
-defm VPMOVSXDQ : SS41I_binop_rm_int8<0x25, "vpmovsxdq", int_x86_sse41_pmovsxdq>,
- VEX;
-defm VPMOVZXBW : SS41I_binop_rm_int8<0x30, "vpmovzxbw", int_x86_sse41_pmovzxbw>,
- VEX;
-defm VPMOVZXWD : SS41I_binop_rm_int8<0x33, "vpmovzxwd", int_x86_sse41_pmovzxwd>,
- VEX;
-defm VPMOVZXDQ : SS41I_binop_rm_int8<0x35, "vpmovzxdq", int_x86_sse41_pmovzxdq>,
- VEX;
+defm VPMOVSXBW : SS41I_binop_rm_int8<0x20, "vpmovsxbw",
+ int_x86_sse41_pmovsxbw>, VEX;
+defm VPMOVSXWD : SS41I_binop_rm_int8<0x23, "vpmovsxwd",
+ int_x86_sse41_pmovsxwd>, VEX;
+defm VPMOVSXDQ : SS41I_binop_rm_int8<0x25, "vpmovsxdq",
+ int_x86_sse41_pmovsxdq>, VEX;
+defm VPMOVZXBW : SS41I_binop_rm_int8<0x30, "vpmovzxbw",
+ int_x86_sse41_pmovzxbw>, VEX;
+defm VPMOVZXWD : SS41I_binop_rm_int8<0x33, "vpmovzxwd",
+ int_x86_sse41_pmovzxwd>, VEX;
+defm VPMOVZXDQ : SS41I_binop_rm_int8<0x35, "vpmovzxdq",
+ int_x86_sse41_pmovzxdq>, VEX;
}
let Predicates = [HasAVX2] in {
int_x86_avx2_pmovzxdq>, VEX, VEX_L;
}
-defm PMOVSXBW : SS41I_binop_rm_int8<0x20, "pmovsxbw", int_x86_sse41_pmovsxbw>;
-defm PMOVSXWD : SS41I_binop_rm_int8<0x23, "pmovsxwd", int_x86_sse41_pmovsxwd>;
-defm PMOVSXDQ : SS41I_binop_rm_int8<0x25, "pmovsxdq", int_x86_sse41_pmovsxdq>;
-defm PMOVZXBW : SS41I_binop_rm_int8<0x30, "pmovzxbw", int_x86_sse41_pmovzxbw>;
-defm PMOVZXWD : SS41I_binop_rm_int8<0x33, "pmovzxwd", int_x86_sse41_pmovzxwd>;
-defm PMOVZXDQ : SS41I_binop_rm_int8<0x35, "pmovzxdq", int_x86_sse41_pmovzxdq>;
+defm PMOVSXBW : SS41I_binop_rm_int8<0x20, "pmovsxbw", int_x86_sse41_pmovsxbw, SSE_INTALU_ITINS_P>;
+defm PMOVSXWD : SS41I_binop_rm_int8<0x23, "pmovsxwd", int_x86_sse41_pmovsxwd, SSE_INTALU_ITINS_P>;
+defm PMOVSXDQ : SS41I_binop_rm_int8<0x25, "pmovsxdq", int_x86_sse41_pmovsxdq, SSE_INTALU_ITINS_P>;
+defm PMOVZXBW : SS41I_binop_rm_int8<0x30, "pmovzxbw", int_x86_sse41_pmovzxbw, SSE_INTALU_ITINS_P>;
+defm PMOVZXWD : SS41I_binop_rm_int8<0x33, "pmovzxwd", int_x86_sse41_pmovzxwd, SSE_INTALU_ITINS_P>;
+defm PMOVZXDQ : SS41I_binop_rm_int8<0x35, "pmovzxdq", int_x86_sse41_pmovzxdq, SSE_INTALU_ITINS_P>;
let Predicates = [HasAVX] in {
// Common patterns involving scalar load.
(VPMOVZXDQYrr VR128:$src)>;
def : Pat<(v8i32 (X86vzmovly (v8i16 VR128:$src))),
(VPMOVZXWDYrr VR128:$src)>;
+ def : Pat<(v16i16 (X86vzmovly (v16i8 VR128:$src))),
+ (VPMOVZXBWYrr VR128:$src)>;
}
def : Pat<(v4i64 (X86vsmovl (v4i32 VR128:$src))), (VPMOVSXDQYrr VR128:$src)>;
def : Pat<(v8i32 (X86vsmovl (v8i16 VR128:$src))), (VPMOVSXWDYrr VR128:$src)>;
+ def : Pat<(v16i16 (X86vsmovl (v16i8 VR128:$src))), (VPMOVSXBWYrr VR128:$src)>;
}
let Predicates = [HasAVX] in {
def : Pat<(v2i64 (X86vsmovl (v4i32 VR128:$src))), (VPMOVSXDQrr VR128:$src)>;
def : Pat<(v4i32 (X86vsmovl (v8i16 VR128:$src))), (VPMOVSXWDrr VR128:$src)>;
+ def : Pat<(v8i16 (X86vsmovl (v16i8 VR128:$src))), (VPMOVSXBWrr VR128:$src)>;
}
let Predicates = [UseSSE41] in {
def : Pat<(v2i64 (X86vsmovl (v4i32 VR128:$src))), (PMOVSXDQrr VR128:$src)>;
def : Pat<(v4i32 (X86vsmovl (v8i16 VR128:$src))), (PMOVSXWDrr VR128:$src)>;
+ def : Pat<(v8i16 (X86vsmovl (v16i8 VR128:$src))), (PMOVSXBWrr VR128:$src)>;
}
-multiclass SS41I_binop_rm_int4<bits<8> opc, string OpcodeStr, Intrinsic IntId> {
+multiclass SS41I_binop_rm_int4<bits<8> opc, string OpcodeStr, Intrinsic IntId,
+ OpndItins itins = DEFAULT_ITINS> {
def rr : SS48I<opc, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
!strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"),
- [(set VR128:$dst, (IntId VR128:$src))]>, OpSize;
+ [(set VR128:$dst, (IntId VR128:$src))], itins.rr>, OpSize;
def rm : SS48I<opc, MRMSrcMem, (outs VR128:$dst), (ins i32mem:$src),
!strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"),
[(set VR128:$dst,
- (IntId (bitconvert (v4i32 (scalar_to_vector (loadi32 addr:$src))))))]>,
+ (IntId (bitconvert (v4i32 (scalar_to_vector (loadi32 addr:$src))))))],
+ itins.rm>,
OpSize;
}
int_x86_avx2_pmovzxwq>, VEX, VEX_L;
}
-defm PMOVSXBD : SS41I_binop_rm_int4<0x21, "pmovsxbd", int_x86_sse41_pmovsxbd>;
-defm PMOVSXWQ : SS41I_binop_rm_int4<0x24, "pmovsxwq", int_x86_sse41_pmovsxwq>;
-defm PMOVZXBD : SS41I_binop_rm_int4<0x31, "pmovzxbd", int_x86_sse41_pmovzxbd>;
-defm PMOVZXWQ : SS41I_binop_rm_int4<0x34, "pmovzxwq", int_x86_sse41_pmovzxwq>;
+defm PMOVSXBD : SS41I_binop_rm_int4<0x21, "pmovsxbd", int_x86_sse41_pmovsxbd,
+ SSE_INTALU_ITINS_P>;
+defm PMOVSXWQ : SS41I_binop_rm_int4<0x24, "pmovsxwq", int_x86_sse41_pmovsxwq,
+ SSE_INTALU_ITINS_P>;
+defm PMOVZXBD : SS41I_binop_rm_int4<0x31, "pmovzxbd", int_x86_sse41_pmovzxbd,
+ SSE_INTALU_ITINS_P>;
+defm PMOVZXWQ : SS41I_binop_rm_int4<0x34, "pmovzxwq", int_x86_sse41_pmovzxwq,
+ SSE_INTALU_ITINS_P>;
let Predicates = [HasAVX] in {
// Common patterns involving scalar load
(PMOVZXWQrm addr:$src)>;
}
-multiclass SS41I_binop_rm_int2<bits<8> opc, string OpcodeStr, Intrinsic IntId> {
+multiclass SS41I_binop_rm_int2<bits<8> opc, string OpcodeStr, Intrinsic IntId,
+ OpndItins itins = DEFAULT_ITINS> {
def rr : SS48I<opc, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
!strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"),
[(set VR128:$dst, (IntId VR128:$src))]>, OpSize;
defm VPMOVZXBQ : SS41I_binop_rm_int4_y<0x32, "vpmovzxbq",
int_x86_avx2_pmovzxbq>, VEX, VEX_L;
}
-defm PMOVSXBQ : SS41I_binop_rm_int2<0x22, "pmovsxbq", int_x86_sse41_pmovsxbq>;
-defm PMOVZXBQ : SS41I_binop_rm_int2<0x32, "pmovzxbq", int_x86_sse41_pmovzxbq>;
+defm PMOVSXBQ : SS41I_binop_rm_int2<0x22, "pmovsxbq", int_x86_sse41_pmovsxbq,
+ SSE_INTALU_ITINS_P>;
+defm PMOVZXBQ : SS41I_binop_rm_int2<0x32, "pmovzxbq", int_x86_sse41_pmovzxbq,
+ SSE_INTALU_ITINS_P>;
let Predicates = [HasAVX2] in {
def : Pat<(v16i16 (X86vsext (v16i8 VR128:$src))), (VPMOVSXBWYrr VR128:$src)>;
/// SS41I_binop_ext8 - SSE 4.1 extract 8 bits to 32 bit reg or 8 bit mem
multiclass SS41I_extract8<bits<8> opc, string OpcodeStr> {
- def rr : SS4AIi8<opc, MRMDestReg, (outs GR32:$dst),
+ def rr : SS4AIi8<opc, MRMDestReg, (outs GR32orGR64:$dst),
(ins VR128:$src1, i32i8imm:$src2),
!strconcat(OpcodeStr,
- "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
- [(set GR32:$dst, (X86pextrb (v16i8 VR128:$src1), imm:$src2))]>,
+ "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
+ [(set GR32orGR64:$dst, (X86pextrb (v16i8 VR128:$src1),
+ imm:$src2))]>,
OpSize;
let neverHasSideEffects = 1, mayStore = 1 in
def mr : SS4AIi8<opc, MRMDestMem, (outs),
(ins i8mem:$dst, VR128:$src1, i32i8imm:$src2),
!strconcat(OpcodeStr,
- "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
+ "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
[]>, OpSize;
// FIXME:
// There's an AssertZext in the way of writing the store pattern
// (store (i8 (trunc (X86pextrb (v16i8 VR128:$src1), imm:$src2))), addr:$dst)
}
-let Predicates = [HasAVX] in {
+let Predicates = [HasAVX] in
defm VPEXTRB : SS41I_extract8<0x14, "vpextrb">, VEX;
- def VPEXTRBrr64 : SS4AIi8<0x14, MRMDestReg, (outs GR64:$dst),
- (ins VR128:$src1, i32i8imm:$src2),
- "vpextrb\t{$src2, $src1, $dst|$dst, $src1, $src2}", []>, OpSize, VEX;
-}
defm PEXTRB : SS41I_extract8<0x14, "pextrb">;
/// SS41I_extract16 - SSE 4.1 extract 16 bits to memory destination
multiclass SS41I_extract16<bits<8> opc, string OpcodeStr> {
+ let isCodeGenOnly = 1, hasSideEffects = 0 in
+ def rr_REV : SS4AIi8<opc, MRMDestReg, (outs GR32orGR64:$dst),
+ (ins VR128:$src1, i32i8imm:$src2),
+ !strconcat(OpcodeStr,
+ "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
+ []>, OpSize;
+
let neverHasSideEffects = 1, mayStore = 1 in
def mr : SS4AIi8<opc, MRMDestMem, (outs),
(ins i16mem:$dst, VR128:$src1, i32i8imm:$src2),
/// SS41I_extractf32 - SSE 4.1 extract 32 bits fp value to int reg or memory
/// destination
-multiclass SS41I_extractf32<bits<8> opc, string OpcodeStr> {
- def rr : SS4AIi8<opc, MRMDestReg, (outs GR32:$dst),
+multiclass SS41I_extractf32<bits<8> opc, string OpcodeStr,
+ OpndItins itins = DEFAULT_ITINS> {
+ def rr : SS4AIi8<opc, MRMDestReg, (outs GR32orGR64:$dst),
(ins VR128:$src1, i32i8imm:$src2),
!strconcat(OpcodeStr,
"\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
- [(set GR32:$dst,
- (extractelt (bc_v4i32 (v4f32 VR128:$src1)), imm:$src2))]>,
+ [(set GR32orGR64:$dst,
+ (extractelt (bc_v4i32 (v4f32 VR128:$src1)), imm:$src2))],
+ itins.rr>,
OpSize;
def mr : SS4AIi8<opc, MRMDestMem, (outs),
(ins f32mem:$dst, VR128:$src1, i32i8imm:$src2),
!strconcat(OpcodeStr,
"\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
[(store (extractelt (bc_v4i32 (v4f32 VR128:$src1)), imm:$src2),
- addr:$dst)]>, OpSize;
+ addr:$dst)], itins.rm>, OpSize;
}
let ExeDomain = SSEPackedSingle in {
- let Predicates = [UseAVX] in {
+ let Predicates = [UseAVX] in
defm VEXTRACTPS : SS41I_extractf32<0x17, "vextractps">, VEX;
- def VEXTRACTPSrr64 : SS4AIi8<0x17, MRMDestReg, (outs GR64:$dst),
- (ins VR128:$src1, i32i8imm:$src2),
- "vextractps\t{$src2, $src1, $dst|$dst, $src1, $src2}",
- []>, OpSize, VEX;
- }
- defm EXTRACTPS : SS41I_extractf32<0x17, "extractps">;
+ defm EXTRACTPS : SS41I_extractf32<0x17, "extractps", SSE_EXTRACT_ITINS>;
}
// Also match an EXTRACTPS store when the store is done as f32 instead of i32.
multiclass SS41I_insert8<bits<8> opc, string asm, bit Is2Addr = 1> {
def rr : SS4AIi8<opc, MRMSrcReg, (outs VR128:$dst),
- (ins VR128:$src1, GR32:$src2, i32i8imm:$src3),
+ (ins VR128:$src1, GR32orGR64:$src2, i32i8imm:$src3),
!if(Is2Addr,
!strconcat(asm, "\t{$src3, $src2, $dst|$dst, $src2, $src3}"),
!strconcat(asm,
"\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}")),
[(set VR128:$dst,
- (X86pinsrb VR128:$src1, GR32:$src2, imm:$src3))]>, OpSize;
+ (X86pinsrb VR128:$src1, GR32orGR64:$src2, imm:$src3))]>, OpSize;
def rm : SS4AIi8<opc, MRMSrcMem, (outs VR128:$dst),
(ins VR128:$src1, i8mem:$src2, i32i8imm:$src3),
!if(Is2Addr,
// are optimized inserts that won't zero arbitrary elements in the destination
// vector. The next one matches the intrinsic and could zero arbitrary elements
// in the target vector.
-multiclass SS41I_insertf32<bits<8> opc, string asm, bit Is2Addr = 1> {
+multiclass SS41I_insertf32<bits<8> opc, string asm, bit Is2Addr = 1,
+ OpndItins itins = DEFAULT_ITINS> {
def rr : SS4AIi8<opc, MRMSrcReg, (outs VR128:$dst),
(ins VR128:$src1, VR128:$src2, u32u8imm:$src3),
!if(Is2Addr,
!strconcat(asm,
"\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}")),
[(set VR128:$dst,
- (X86insrtps VR128:$src1, VR128:$src2, imm:$src3))]>,
+ (X86insrtps VR128:$src1, VR128:$src2, imm:$src3))], itins.rr>,
OpSize;
def rm : SS4AIi8<opc, MRMSrcMem, (outs VR128:$dst),
(ins VR128:$src1, f32mem:$src2, u32u8imm:$src3),
[(set VR128:$dst,
(X86insrtps VR128:$src1,
(v4f32 (scalar_to_vector (loadf32 addr:$src2))),
- imm:$src3))]>, OpSize;
+ imm:$src3))], itins.rm>, OpSize;
}
let ExeDomain = SSEPackedSingle in {
- let Predicates = [HasAVX] in
+ let Predicates = [UseAVX] in
defm VINSERTPS : SS41I_insertf32<0x21, "vinsertps", 0>, VEX_4V;
let Constraints = "$src1 = $dst" in
- defm INSERTPS : SS41I_insertf32<0x21, "insertps">;
+ defm INSERTPS : SS41I_insertf32<0x21, "insertps", 1, SSE_INSERT_ITINS>;
}
//===----------------------------------------------------------------------===//
(outs RC:$dst), (ins RC:$src1, i32i8imm:$src2),
!strconcat(OpcodeStr,
"ps\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
- [(set RC:$dst, (V4F32Int RC:$src1, imm:$src2))]>,
+ [(set RC:$dst, (V4F32Int RC:$src1, imm:$src2))],
+ IIC_SSE_ROUNDPS_REG>,
OpSize;
// Vector intrinsic operation, mem
!strconcat(OpcodeStr,
"ps\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
[(set RC:$dst,
- (V4F32Int (mem_frag32 addr:$src1),imm:$src2))]>,
+ (V4F32Int (mem_frag32 addr:$src1),imm:$src2))],
+ IIC_SSE_ROUNDPS_MEM>,
OpSize;
} // ExeDomain = SSEPackedSingle
(outs RC:$dst), (ins RC:$src1, i32i8imm:$src2),
!strconcat(OpcodeStr,
"pd\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
- [(set RC:$dst, (V2F64Int RC:$src1, imm:$src2))]>,
+ [(set RC:$dst, (V2F64Int RC:$src1, imm:$src2))],
+ IIC_SSE_ROUNDPS_REG>,
OpSize;
// Vector intrinsic operation, mem
!strconcat(OpcodeStr,
"pd\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
[(set RC:$dst,
- (V2F64Int (mem_frag64 addr:$src1),imm:$src2))]>,
+ (V2F64Int (mem_frag64 addr:$src1),imm:$src2))],
+ IIC_SSE_ROUNDPS_REG>,
OpSize;
} // ExeDomain = SSEPackedDouble
}
let Predicates = [HasAVX] in {
// Intrinsic form
defm VROUND : sse41_fp_unop_rm<0x08, 0x09, "vround", f128mem, VR128,
- memopv4f32, memopv2f64,
+ loadv4f32, loadv2f64,
int_x86_sse41_round_ps,
int_x86_sse41_round_pd>, VEX;
defm VROUNDY : sse41_fp_unop_rm<0x08, 0x09, "vround", f256mem, VR256,
- memopv8f32, memopv4f64,
+ loadv8f32, loadv4f64,
int_x86_avx_round_ps_256,
int_x86_avx_round_pd_256>, VEX, VEX_L;
defm VROUND : sse41_fp_binop_rm<0x0A, 0x0B, "vround",
OpSize, VEX;
def VPTESTrm : SS48I<0x17, MRMSrcMem, (outs), (ins VR128:$src1, f128mem:$src2),
"vptest\t{$src2, $src1|$src1, $src2}",
- [(set EFLAGS,(X86ptest VR128:$src1, (memopv2i64 addr:$src2)))]>,
+ [(set EFLAGS,(X86ptest VR128:$src1, (loadv2i64 addr:$src2)))]>,
OpSize, VEX;
def VPTESTYrr : SS48I<0x17, MRMSrcReg, (outs), (ins VR256:$src1, VR256:$src2),
OpSize, VEX, VEX_L;
def VPTESTYrm : SS48I<0x17, MRMSrcMem, (outs), (ins VR256:$src1, i256mem:$src2),
"vptest\t{$src2, $src1|$src1, $src2}",
- [(set EFLAGS,(X86ptest VR256:$src1, (memopv4i64 addr:$src2)))]>,
+ [(set EFLAGS,(X86ptest VR256:$src1, (loadv4i64 addr:$src2)))]>,
OpSize, VEX, VEX_L;
}
let Defs = [EFLAGS], Predicates = [HasAVX] in {
let ExeDomain = SSEPackedSingle in {
-defm VTESTPS : avx_bittest<0x0E, "vtestps", VR128, f128mem, memopv4f32, v4f32>;
-defm VTESTPSY : avx_bittest<0x0E, "vtestps", VR256, f256mem, memopv8f32, v8f32>,
+defm VTESTPS : avx_bittest<0x0E, "vtestps", VR128, f128mem, loadv4f32, v4f32>;
+defm VTESTPSY : avx_bittest<0x0E, "vtestps", VR256, f256mem, loadv8f32, v8f32>,
VEX_L;
}
let ExeDomain = SSEPackedDouble in {
-defm VTESTPD : avx_bittest<0x0F, "vtestpd", VR128, f128mem, memopv2f64, v2f64>;
-defm VTESTPDY : avx_bittest<0x0F, "vtestpd", VR256, f256mem, memopv4f64, v4f64>,
+defm VTESTPD : avx_bittest<0x0F, "vtestpd", VR128, f128mem, loadv2f64, v2f64>;
+defm VTESTPDY : avx_bittest<0x0F, "vtestpd", VR256, f256mem, loadv4f64, v4f64>,
VEX_L;
}
}
let Defs = [EFLAGS], Predicates = [HasPOPCNT] in {
def POPCNT16rr : I<0xB8, MRMSrcReg, (outs GR16:$dst), (ins GR16:$src),
"popcnt{w}\t{$src, $dst|$dst, $src}",
- [(set GR16:$dst, (ctpop GR16:$src)), (implicit EFLAGS)]>,
+ [(set GR16:$dst, (ctpop GR16:$src)), (implicit EFLAGS)],
+ IIC_SSE_POPCNT_RR>,
OpSize, XS;
def POPCNT16rm : I<0xB8, MRMSrcMem, (outs GR16:$dst), (ins i16mem:$src),
"popcnt{w}\t{$src, $dst|$dst, $src}",
[(set GR16:$dst, (ctpop (loadi16 addr:$src))),
- (implicit EFLAGS)]>, OpSize, XS;
+ (implicit EFLAGS)], IIC_SSE_POPCNT_RM>, OpSize, XS;
def POPCNT32rr : I<0xB8, MRMSrcReg, (outs GR32:$dst), (ins GR32:$src),
"popcnt{l}\t{$src, $dst|$dst, $src}",
- [(set GR32:$dst, (ctpop GR32:$src)), (implicit EFLAGS)]>,
+ [(set GR32:$dst, (ctpop GR32:$src)), (implicit EFLAGS)],
+ IIC_SSE_POPCNT_RR>,
XS;
def POPCNT32rm : I<0xB8, MRMSrcMem, (outs GR32:$dst), (ins i32mem:$src),
"popcnt{l}\t{$src, $dst|$dst, $src}",
[(set GR32:$dst, (ctpop (loadi32 addr:$src))),
- (implicit EFLAGS)]>, XS;
+ (implicit EFLAGS)], IIC_SSE_POPCNT_RM>, XS;
def POPCNT64rr : RI<0xB8, MRMSrcReg, (outs GR64:$dst), (ins GR64:$src),
"popcnt{q}\t{$src, $dst|$dst, $src}",
- [(set GR64:$dst, (ctpop GR64:$src)), (implicit EFLAGS)]>,
+ [(set GR64:$dst, (ctpop GR64:$src)), (implicit EFLAGS)],
+ IIC_SSE_POPCNT_RR>,
XS;
def POPCNT64rm : RI<0xB8, MRMSrcMem, (outs GR64:$dst), (ins i64mem:$src),
"popcnt{q}\t{$src, $dst|$dst, $src}",
[(set GR64:$dst, (ctpop (loadi64 addr:$src))),
- (implicit EFLAGS)]>, XS;
+ (implicit EFLAGS)], IIC_SSE_POPCNT_RM>, XS;
}
/// SS41I_binop_rm_int - Simple SSE 4.1 binary operator
multiclass SS41I_binop_rm_int<bits<8> opc, string OpcodeStr,
- Intrinsic IntId128, bit Is2Addr = 1> {
+ Intrinsic IntId128, bit Is2Addr = 1,
+ OpndItins itins = DEFAULT_ITINS> {
let isCommutable = 1 in
def rr : SS48I<opc, MRMSrcReg, (outs VR128:$dst),
(ins VR128:$src1, VR128:$src2),
!if(Is2Addr,
!strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"),
!strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}")),
- [(set VR128:$dst, (IntId128 VR128:$src1, VR128:$src2))]>, OpSize;
+ [(set VR128:$dst, (IntId128 VR128:$src1, VR128:$src2))],
+ itins.rr>, OpSize;
def rm : SS48I<opc, MRMSrcMem, (outs VR128:$dst),
(ins VR128:$src1, i128mem:$src2),
!if(Is2Addr,
!strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}")),
[(set VR128:$dst,
(IntId128 VR128:$src1,
- (bitconvert (memopv2i64 addr:$src2))))]>, OpSize;
+ (bitconvert (memopv2i64 addr:$src2))))],
+ itins.rm>, OpSize;
}
/// SS41I_binop_rm_int_y - Simple SSE 4.1 binary operator
!strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
[(set VR256:$dst,
(IntId256 VR256:$src1,
- (bitconvert (memopv4i64 addr:$src2))))]>, OpSize;
+ (bitconvert (loadv4i64 addr:$src2))))]>, OpSize;
}
/// SS48I_binop_rm - Simple SSE41 binary operator.
multiclass SS48I_binop_rm<bits<8> opc, string OpcodeStr, SDNode OpNode,
ValueType OpVT, RegisterClass RC, PatFrag memop_frag,
- X86MemOperand x86memop, bit Is2Addr = 1> {
+ X86MemOperand x86memop, bit Is2Addr = 1,
+ OpndItins itins = DEFAULT_ITINS> {
let isCommutable = 1 in
def rr : SS48I<opc, MRMSrcReg, (outs RC:$dst),
(ins RC:$src1, RC:$src2),
defm VPACKUSDW : SS41I_binop_rm_int<0x2B, "vpackusdw", int_x86_sse41_packusdw,
0>, VEX_4V;
defm VPMINSB : SS48I_binop_rm<0x38, "vpminsb", X86smin, v16i8, VR128,
- memopv2i64, i128mem, 0>, VEX_4V;
+ loadv2i64, i128mem, 0>, VEX_4V;
defm VPMINSD : SS48I_binop_rm<0x39, "vpminsd", X86smin, v4i32, VR128,
- memopv2i64, i128mem, 0>, VEX_4V;
+ loadv2i64, i128mem, 0>, VEX_4V;
defm VPMINUD : SS48I_binop_rm<0x3B, "vpminud", X86umin, v4i32, VR128,
- memopv2i64, i128mem, 0>, VEX_4V;
+ loadv2i64, i128mem, 0>, VEX_4V;
defm VPMINUW : SS48I_binop_rm<0x3A, "vpminuw", X86umin, v8i16, VR128,
- memopv2i64, i128mem, 0>, VEX_4V;
+ loadv2i64, i128mem, 0>, VEX_4V;
defm VPMAXSB : SS48I_binop_rm<0x3C, "vpmaxsb", X86smax, v16i8, VR128,
- memopv2i64, i128mem, 0>, VEX_4V;
+ loadv2i64, i128mem, 0>, VEX_4V;
defm VPMAXSD : SS48I_binop_rm<0x3D, "vpmaxsd", X86smax, v4i32, VR128,
- memopv2i64, i128mem, 0>, VEX_4V;
+ loadv2i64, i128mem, 0>, VEX_4V;
defm VPMAXUD : SS48I_binop_rm<0x3F, "vpmaxud", X86umax, v4i32, VR128,
- memopv2i64, i128mem, 0>, VEX_4V;
+ loadv2i64, i128mem, 0>, VEX_4V;
defm VPMAXUW : SS48I_binop_rm<0x3E, "vpmaxuw", X86umax, v8i16, VR128,
- memopv2i64, i128mem, 0>, VEX_4V;
+ loadv2i64, i128mem, 0>, VEX_4V;
defm VPMULDQ : SS41I_binop_rm_int<0x28, "vpmuldq", int_x86_sse41_pmuldq,
0>, VEX_4V;
}
defm VPACKUSDW : SS41I_binop_rm_int_y<0x2B, "vpackusdw",
int_x86_avx2_packusdw>, VEX_4V, VEX_L;
defm VPMINSBY : SS48I_binop_rm<0x38, "vpminsb", X86smin, v32i8, VR256,
- memopv4i64, i256mem, 0>, VEX_4V, VEX_L;
+ loadv4i64, i256mem, 0>, VEX_4V, VEX_L;
defm VPMINSDY : SS48I_binop_rm<0x39, "vpminsd", X86smin, v8i32, VR256,
- memopv4i64, i256mem, 0>, VEX_4V, VEX_L;
+ loadv4i64, i256mem, 0>, VEX_4V, VEX_L;
defm VPMINUDY : SS48I_binop_rm<0x3B, "vpminud", X86umin, v8i32, VR256,
- memopv4i64, i256mem, 0>, VEX_4V, VEX_L;
+ loadv4i64, i256mem, 0>, VEX_4V, VEX_L;
defm VPMINUWY : SS48I_binop_rm<0x3A, "vpminuw", X86umin, v16i16, VR256,
- memopv4i64, i256mem, 0>, VEX_4V, VEX_L;
+ loadv4i64, i256mem, 0>, VEX_4V, VEX_L;
defm VPMAXSBY : SS48I_binop_rm<0x3C, "vpmaxsb", X86smax, v32i8, VR256,
- memopv4i64, i256mem, 0>, VEX_4V, VEX_L;
+ loadv4i64, i256mem, 0>, VEX_4V, VEX_L;
defm VPMAXSDY : SS48I_binop_rm<0x3D, "vpmaxsd", X86smax, v8i32, VR256,
- memopv4i64, i256mem, 0>, VEX_4V, VEX_L;
+ loadv4i64, i256mem, 0>, VEX_4V, VEX_L;
defm VPMAXUDY : SS48I_binop_rm<0x3F, "vpmaxud", X86umax, v8i32, VR256,
- memopv4i64, i256mem, 0>, VEX_4V, VEX_L;
+ loadv4i64, i256mem, 0>, VEX_4V, VEX_L;
defm VPMAXUWY : SS48I_binop_rm<0x3E, "vpmaxuw", X86umax, v16i16, VR256,
- memopv4i64, i256mem, 0>, VEX_4V, VEX_L;
+ loadv4i64, i256mem, 0>, VEX_4V, VEX_L;
defm VPMULDQ : SS41I_binop_rm_int_y<0x28, "vpmuldq",
int_x86_avx2_pmul_dq>, VEX_4V, VEX_L;
}
let isCommutable = 0 in
defm PACKUSDW : SS41I_binop_rm_int<0x2B, "packusdw", int_x86_sse41_packusdw>;
defm PMINSB : SS48I_binop_rm<0x38, "pminsb", X86smin, v16i8, VR128,
- memopv2i64, i128mem>;
+ memopv2i64, i128mem, 1, SSE_INTALU_ITINS_P>;
defm PMINSD : SS48I_binop_rm<0x39, "pminsd", X86smin, v4i32, VR128,
- memopv2i64, i128mem>;
+ memopv2i64, i128mem, 1, SSE_INTALU_ITINS_P>;
defm PMINUD : SS48I_binop_rm<0x3B, "pminud", X86umin, v4i32, VR128,
- memopv2i64, i128mem>;
+ memopv2i64, i128mem, 1, SSE_INTALU_ITINS_P>;
defm PMINUW : SS48I_binop_rm<0x3A, "pminuw", X86umin, v8i16, VR128,
- memopv2i64, i128mem>;
+ memopv2i64, i128mem, 1, SSE_INTALU_ITINS_P>;
defm PMAXSB : SS48I_binop_rm<0x3C, "pmaxsb", X86smax, v16i8, VR128,
- memopv2i64, i128mem>;
+ memopv2i64, i128mem, 1, SSE_INTALU_ITINS_P>;
defm PMAXSD : SS48I_binop_rm<0x3D, "pmaxsd", X86smax, v4i32, VR128,
- memopv2i64, i128mem>;
+ memopv2i64, i128mem, 1, SSE_INTALU_ITINS_P>;
defm PMAXUD : SS48I_binop_rm<0x3F, "pmaxud", X86umax, v4i32, VR128,
- memopv2i64, i128mem>;
+ memopv2i64, i128mem, 1, SSE_INTALU_ITINS_P>;
defm PMAXUW : SS48I_binop_rm<0x3E, "pmaxuw", X86umax, v8i16, VR128,
- memopv2i64, i128mem>;
- defm PMULDQ : SS41I_binop_rm_int<0x28, "pmuldq", int_x86_sse41_pmuldq>;
+ memopv2i64, i128mem, 1, SSE_INTALU_ITINS_P>;
+ defm PMULDQ : SS41I_binop_rm_int<0x28, "pmuldq", int_x86_sse41_pmuldq,
+ 1, SSE_INTMUL_ITINS_P>;
}
let Predicates = [HasAVX] in {
let Constraints = "$src1 = $dst" in {
defm PMULLD : SS48I_binop_rm<0x40, "pmulld", mul, v4i32, VR128,
- memopv2i64, i128mem>;
+ memopv2i64, i128mem, 1, SSE_PMULLD_ITINS>;
defm PCMPEQQ : SS48I_binop_rm<0x29, "pcmpeqq", X86pcmpeq, v2i64, VR128,
- memopv2i64, i128mem>;
+ memopv2i64, i128mem, 1, SSE_INTALUQ_ITINS_P>;
}
/// SS41I_binop_rmi_int - SSE 4.1 binary operator with 8-bit immediate
multiclass SS41I_binop_rmi_int<bits<8> opc, string OpcodeStr,
Intrinsic IntId, RegisterClass RC, PatFrag memop_frag,
- X86MemOperand x86memop, bit Is2Addr = 1> {
+ X86MemOperand x86memop, bit Is2Addr = 1,
+ OpndItins itins = DEFAULT_ITINS> {
let isCommutable = 1 in
def rri : SS4AIi8<opc, MRMSrcReg, (outs RC:$dst),
(ins RC:$src1, RC:$src2, u32u8imm:$src3),
"\t{$src3, $src2, $dst|$dst, $src2, $src3}"),
!strconcat(OpcodeStr,
"\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}")),
- [(set RC:$dst, (IntId RC:$src1, RC:$src2, imm:$src3))]>,
+ [(set RC:$dst, (IntId RC:$src1, RC:$src2, imm:$src3))], itins.rr>,
OpSize;
def rmi : SS4AIi8<opc, MRMSrcMem, (outs RC:$dst),
(ins RC:$src1, x86memop:$src2, u32u8imm:$src3),
"\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}")),
[(set RC:$dst,
(IntId RC:$src1,
- (bitconvert (memop_frag addr:$src2)), imm:$src3))]>,
+ (bitconvert (memop_frag addr:$src2)), imm:$src3))], itins.rm>,
OpSize;
}
let isCommutable = 0 in {
let ExeDomain = SSEPackedSingle in {
defm VBLENDPS : SS41I_binop_rmi_int<0x0C, "vblendps", int_x86_sse41_blendps,
- VR128, memopv4f32, f128mem, 0>, VEX_4V;
+ VR128, loadv4f32, f128mem, 0>, VEX_4V;
defm VBLENDPSY : SS41I_binop_rmi_int<0x0C, "vblendps",
- int_x86_avx_blend_ps_256, VR256, memopv8f32,
+ int_x86_avx_blend_ps_256, VR256, loadv8f32,
f256mem, 0>, VEX_4V, VEX_L;
}
let ExeDomain = SSEPackedDouble in {
defm VBLENDPD : SS41I_binop_rmi_int<0x0D, "vblendpd", int_x86_sse41_blendpd,
- VR128, memopv2f64, f128mem, 0>, VEX_4V;
+ VR128, loadv2f64, f128mem, 0>, VEX_4V;
defm VBLENDPDY : SS41I_binop_rmi_int<0x0D, "vblendpd",
- int_x86_avx_blend_pd_256,VR256, memopv4f64,
+ int_x86_avx_blend_pd_256,VR256, loadv4f64,
f256mem, 0>, VEX_4V, VEX_L;
}
defm VPBLENDW : SS41I_binop_rmi_int<0x0E, "vpblendw", int_x86_sse41_pblendw,
- VR128, memopv2i64, i128mem, 0>, VEX_4V;
+ VR128, loadv2i64, i128mem, 0>, VEX_4V;
defm VMPSADBW : SS41I_binop_rmi_int<0x42, "vmpsadbw", int_x86_sse41_mpsadbw,
- VR128, memopv2i64, i128mem, 0>, VEX_4V;
+ VR128, loadv2i64, i128mem, 0>, VEX_4V;
}
let ExeDomain = SSEPackedSingle in
defm VDPPS : SS41I_binop_rmi_int<0x40, "vdpps", int_x86_sse41_dpps,
- VR128, memopv4f32, f128mem, 0>, VEX_4V;
+ VR128, loadv4f32, f128mem, 0>, VEX_4V;
let ExeDomain = SSEPackedDouble in
defm VDPPD : SS41I_binop_rmi_int<0x41, "vdppd", int_x86_sse41_dppd,
- VR128, memopv2f64, f128mem, 0>, VEX_4V;
+ VR128, loadv2f64, f128mem, 0>, VEX_4V;
let ExeDomain = SSEPackedSingle in
defm VDPPSY : SS41I_binop_rmi_int<0x40, "vdpps", int_x86_avx_dp_ps_256,
- VR256, memopv8f32, i256mem, 0>, VEX_4V, VEX_L;
+ VR256, loadv8f32, i256mem, 0>, VEX_4V, VEX_L;
}
let Predicates = [HasAVX2] in {
let isCommutable = 0 in {
defm VPBLENDWY : SS41I_binop_rmi_int<0x0E, "vpblendw", int_x86_avx2_pblendw,
- VR256, memopv4i64, i256mem, 0>, VEX_4V, VEX_L;
+ VR256, loadv4i64, i256mem, 0>, VEX_4V, VEX_L;
defm VMPSADBWY : SS41I_binop_rmi_int<0x42, "vmpsadbw", int_x86_avx2_mpsadbw,
- VR256, memopv4i64, i256mem, 0>, VEX_4V, VEX_L;
+ VR256, loadv4i64, i256mem, 0>, VEX_4V, VEX_L;
}
}
let isCommutable = 0 in {
let ExeDomain = SSEPackedSingle in
defm BLENDPS : SS41I_binop_rmi_int<0x0C, "blendps", int_x86_sse41_blendps,
- VR128, memopv4f32, f128mem>;
+ VR128, memopv4f32, f128mem,
+ 1, SSE_INTALU_ITINS_P>;
let ExeDomain = SSEPackedDouble in
defm BLENDPD : SS41I_binop_rmi_int<0x0D, "blendpd", int_x86_sse41_blendpd,
- VR128, memopv2f64, f128mem>;
+ VR128, memopv2f64, f128mem,
+ 1, SSE_INTALU_ITINS_P>;
defm PBLENDW : SS41I_binop_rmi_int<0x0E, "pblendw", int_x86_sse41_pblendw,
- VR128, memopv2i64, i128mem>;
+ VR128, memopv2i64, i128mem,
+ 1, SSE_INTALU_ITINS_P>;
defm MPSADBW : SS41I_binop_rmi_int<0x42, "mpsadbw", int_x86_sse41_mpsadbw,
- VR128, memopv2i64, i128mem>;
+ VR128, memopv2i64, i128mem,
+ 1, SSE_INTMUL_ITINS_P>;
}
let ExeDomain = SSEPackedSingle in
defm DPPS : SS41I_binop_rmi_int<0x40, "dpps", int_x86_sse41_dpps,
- VR128, memopv4f32, f128mem>;
+ VR128, memopv4f32, f128mem, 1,
+ SSE_DPPS_ITINS>;
let ExeDomain = SSEPackedDouble in
defm DPPD : SS41I_binop_rmi_int<0x41, "dppd", int_x86_sse41_dppd,
- VR128, memopv2f64, f128mem>;
+ VR128, memopv2f64, f128mem, 1,
+ SSE_DPPD_ITINS>;
}
/// SS41I_quaternary_int_avx - AVX SSE 4.1 with 4 operators
let Predicates = [HasAVX] in {
let ExeDomain = SSEPackedDouble in {
defm VBLENDVPD : SS41I_quaternary_int_avx<0x4B, "vblendvpd", VR128, f128mem,
- memopv2f64, int_x86_sse41_blendvpd>;
+ loadv2f64, int_x86_sse41_blendvpd>;
defm VBLENDVPDY : SS41I_quaternary_int_avx<0x4B, "vblendvpd", VR256, f256mem,
- memopv4f64, int_x86_avx_blendv_pd_256>, VEX_L;
+ loadv4f64, int_x86_avx_blendv_pd_256>, VEX_L;
} // ExeDomain = SSEPackedDouble
let ExeDomain = SSEPackedSingle in {
defm VBLENDVPS : SS41I_quaternary_int_avx<0x4A, "vblendvps", VR128, f128mem,
- memopv4f32, int_x86_sse41_blendvps>;
+ loadv4f32, int_x86_sse41_blendvps>;
defm VBLENDVPSY : SS41I_quaternary_int_avx<0x4A, "vblendvps", VR256, f256mem,
- memopv8f32, int_x86_avx_blendv_ps_256>, VEX_L;
+ loadv8f32, int_x86_avx_blendv_ps_256>, VEX_L;
} // ExeDomain = SSEPackedSingle
defm VPBLENDVB : SS41I_quaternary_int_avx<0x4C, "vpblendvb", VR128, i128mem,
- memopv2i64, int_x86_sse41_pblendvb>;
+ loadv2i64, int_x86_sse41_pblendvb>;
}
let Predicates = [HasAVX2] in {
defm VPBLENDVBY : SS41I_quaternary_int_avx<0x4C, "vpblendvb", VR256, i256mem,
- memopv4i64, int_x86_avx2_pblendvb>, VEX_L;
+ loadv4i64, int_x86_avx2_pblendvb>, VEX_L;
}
let Predicates = [HasAVX] in {
let Predicates = [HasAVX2] in {
def : Pat<(v32i8 (vselect (v32i8 VR256:$mask), (v32i8 VR256:$src1),
(v32i8 VR256:$src2))),
- (VPBLENDVBYrr VR256:$src1, VR256:$src2, VR256:$mask)>;
+ (VPBLENDVBYrr VR256:$src2, VR256:$src1, VR256:$mask)>;
def : Pat<(v16i16 (X86Blendi (v16i16 VR256:$src1), (v16i16 VR256:$src2),
(imm:$mask))),
(VPBLENDWYrri VR256:$src1, VR256:$src2, imm:$mask)>;
/// SS41I_ternary_int - SSE 4.1 ternary operator
let Uses = [XMM0], Constraints = "$src1 = $dst" in {
multiclass SS41I_ternary_int<bits<8> opc, string OpcodeStr, PatFrag mem_frag,
- X86MemOperand x86memop, Intrinsic IntId> {
+ X86MemOperand x86memop, Intrinsic IntId,
+ OpndItins itins = DEFAULT_ITINS> {
def rr0 : SS48I<opc, MRMSrcReg, (outs VR128:$dst),
(ins VR128:$src1, VR128:$src2),
!strconcat(OpcodeStr,
"\t{$src2, $dst|$dst, $src2}"),
- [(set VR128:$dst, (IntId VR128:$src1, VR128:$src2, XMM0))]>,
- OpSize;
+ [(set VR128:$dst, (IntId VR128:$src1, VR128:$src2, XMM0))],
+ itins.rr>, OpSize;
def rm0 : SS48I<opc, MRMSrcMem, (outs VR128:$dst),
(ins VR128:$src1, x86memop:$src2),
"\t{$src2, $dst|$dst, $src2}"),
[(set VR128:$dst,
(IntId VR128:$src1,
- (bitconvert (mem_frag addr:$src2)), XMM0))]>, OpSize;
+ (bitconvert (mem_frag addr:$src2)), XMM0))],
+ itins.rm>, OpSize;
}
}
let Predicates = [HasAVX] in
defm VPCMPGTQ : SS42I_binop_rm<0x37, "vpcmpgtq", X86pcmpgt, v2i64, VR128,
- memopv2i64, i128mem, 0>, VEX_4V;
+ loadv2i64, i128mem, 0>, VEX_4V;
let Predicates = [HasAVX2] in
defm VPCMPGTQY : SS42I_binop_rm<0x37, "vpcmpgtq", X86pcmpgt, v4i64, VR256,
- memopv4i64, i256mem, 0>, VEX_4V, VEX_L;
+ loadv4i64, i256mem, 0>, VEX_4V, VEX_L;
let Constraints = "$src1 = $dst" in
defm PCMPGTQ : SS42I_binop_rm<0x37, "pcmpgtq", X86pcmpgt, v2i64, VR128,
// crc intrinsic instruction
// This set of instructions are only rm, the only difference is the size
// of r and m.
+class SS42I_crc32r<bits<8> opc, string asm, RegisterClass RCOut,
+ RegisterClass RCIn, SDPatternOperator Int> :
+ SS42FI<opc, MRMSrcReg, (outs RCOut:$dst), (ins RCOut:$src1, RCIn:$src2),
+ !strconcat(asm, "\t{$src2, $src1|$src1, $src2}"),
+ [(set RCOut:$dst, (Int RCOut:$src1, RCIn:$src2))], IIC_CRC32_REG>;
+
+class SS42I_crc32m<bits<8> opc, string asm, RegisterClass RCOut,
+ X86MemOperand x86memop, SDPatternOperator Int> :
+ SS42FI<opc, MRMSrcMem, (outs RCOut:$dst), (ins RCOut:$src1, x86memop:$src2),
+ !strconcat(asm, "\t{$src2, $src1|$src1, $src2}"),
+ [(set RCOut:$dst, (Int RCOut:$src1, (load addr:$src2)))],
+ IIC_CRC32_MEM>;
+
let Constraints = "$src1 = $dst" in {
- def CRC32r32m8 : SS42FI<0xF0, MRMSrcMem, (outs GR32:$dst),
- (ins GR32:$src1, i8mem:$src2),
- "crc32{b}\t{$src2, $src1|$src1, $src2}",
- [(set GR32:$dst,
- (int_x86_sse42_crc32_32_8 GR32:$src1,
- (load addr:$src2)))]>;
- def CRC32r32r8 : SS42FI<0xF0, MRMSrcReg, (outs GR32:$dst),
- (ins GR32:$src1, GR8:$src2),
- "crc32{b}\t{$src2, $src1|$src1, $src2}",
- [(set GR32:$dst,
- (int_x86_sse42_crc32_32_8 GR32:$src1, GR8:$src2))]>;
- def CRC32r32m16 : SS42FI<0xF1, MRMSrcMem, (outs GR32:$dst),
- (ins GR32:$src1, i16mem:$src2),
- "crc32{w}\t{$src2, $src1|$src1, $src2}",
- [(set GR32:$dst,
- (int_x86_sse42_crc32_32_16 GR32:$src1,
- (load addr:$src2)))]>,
- OpSize;
- def CRC32r32r16 : SS42FI<0xF1, MRMSrcReg, (outs GR32:$dst),
- (ins GR32:$src1, GR16:$src2),
- "crc32{w}\t{$src2, $src1|$src1, $src2}",
- [(set GR32:$dst,
- (int_x86_sse42_crc32_32_16 GR32:$src1, GR16:$src2))]>,
- OpSize;
- def CRC32r32m32 : SS42FI<0xF1, MRMSrcMem, (outs GR32:$dst),
- (ins GR32:$src1, i32mem:$src2),
- "crc32{l}\t{$src2, $src1|$src1, $src2}",
- [(set GR32:$dst,
- (int_x86_sse42_crc32_32_32 GR32:$src1,
- (load addr:$src2)))]>;
- def CRC32r32r32 : SS42FI<0xF1, MRMSrcReg, (outs GR32:$dst),
- (ins GR32:$src1, GR32:$src2),
- "crc32{l}\t{$src2, $src1|$src1, $src2}",
- [(set GR32:$dst,
- (int_x86_sse42_crc32_32_32 GR32:$src1, GR32:$src2))]>;
- def CRC32r64m8 : SS42FI<0xF0, MRMSrcMem, (outs GR64:$dst),
- (ins GR64:$src1, i8mem:$src2),
- "crc32{b}\t{$src2, $src1|$src1, $src2}",
- [(set GR64:$dst,
- (int_x86_sse42_crc32_64_8 GR64:$src1,
- (load addr:$src2)))]>,
- REX_W;
- def CRC32r64r8 : SS42FI<0xF0, MRMSrcReg, (outs GR64:$dst),
- (ins GR64:$src1, GR8:$src2),
- "crc32{b}\t{$src2, $src1|$src1, $src2}",
- [(set GR64:$dst,
- (int_x86_sse42_crc32_64_8 GR64:$src1, GR8:$src2))]>,
- REX_W;
- def CRC32r64m64 : SS42FI<0xF1, MRMSrcMem, (outs GR64:$dst),
- (ins GR64:$src1, i64mem:$src2),
- "crc32{q}\t{$src2, $src1|$src1, $src2}",
- [(set GR64:$dst,
- (int_x86_sse42_crc32_64_64 GR64:$src1,
- (load addr:$src2)))]>,
- REX_W;
- def CRC32r64r64 : SS42FI<0xF1, MRMSrcReg, (outs GR64:$dst),
- (ins GR64:$src1, GR64:$src2),
- "crc32{q}\t{$src2, $src1|$src1, $src2}",
- [(set GR64:$dst,
- (int_x86_sse42_crc32_64_64 GR64:$src1, GR64:$src2))]>,
- REX_W;
+ def CRC32r32m8 : SS42I_crc32m<0xF0, "crc32{b}", GR32, i8mem,
+ int_x86_sse42_crc32_32_8>;
+ def CRC32r32r8 : SS42I_crc32r<0xF0, "crc32{b}", GR32, GR8,
+ int_x86_sse42_crc32_32_8>;
+ def CRC32r32m16 : SS42I_crc32m<0xF1, "crc32{w}", GR32, i16mem,
+ int_x86_sse42_crc32_32_16>, OpSize;
+ def CRC32r32r16 : SS42I_crc32r<0xF1, "crc32{w}", GR32, GR16,
+ int_x86_sse42_crc32_32_16>, OpSize;
+ def CRC32r32m32 : SS42I_crc32m<0xF1, "crc32{l}", GR32, i32mem,
+ int_x86_sse42_crc32_32_32>;
+ def CRC32r32r32 : SS42I_crc32r<0xF1, "crc32{l}", GR32, GR32,
+ int_x86_sse42_crc32_32_32>;
+ def CRC32r64m64 : SS42I_crc32m<0xF1, "crc32{q}", GR64, i64mem,
+ int_x86_sse42_crc32_64_64>, REX_W;
+ def CRC32r64r64 : SS42I_crc32r<0xF1, "crc32{q}", GR64, GR64,
+ int_x86_sse42_crc32_64_64>, REX_W;
+ let hasSideEffects = 0 in {
+ let mayLoad = 1 in
+ def CRC32r64m8 : SS42I_crc32m<0xF0, "crc32{b}", GR64, i8mem,
+ null_frag>, REX_W;
+ def CRC32r64r8 : SS42I_crc32r<0xF0, "crc32{b}", GR64, GR8,
+ null_frag>, REX_W;
+ }
}
+//===----------------------------------------------------------------------===//
+// SHA-NI Instructions
+//===----------------------------------------------------------------------===//
+
+multiclass SHAI_binop<bits<8> Opc, string OpcodeStr, Intrinsic IntId,
+ bit UsesXMM0 = 0> {
+ def rr : I<Opc, MRMSrcReg, (outs VR128:$dst),
+ (ins VR128:$src1, VR128:$src2),
+ !strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"),
+ [!if(UsesXMM0,
+ (set VR128:$dst, (IntId VR128:$src1, VR128:$src2, XMM0)),
+ (set VR128:$dst, (IntId VR128:$src1, VR128:$src2)))]>, T8;
+
+ def rm : I<Opc, MRMSrcMem, (outs VR128:$dst),
+ (ins VR128:$src1, i128mem:$src2),
+ !strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"),
+ [!if(UsesXMM0,
+ (set VR128:$dst, (IntId VR128:$src1,
+ (bc_v4i32 (memopv2i64 addr:$src2)), XMM0)),
+ (set VR128:$dst, (IntId VR128:$src1,
+ (bc_v4i32 (memopv2i64 addr:$src2)))))]>, T8;
+}
+
+let Constraints = "$src1 = $dst", Predicates = [HasSHA] in {
+ def SHA1RNDS4rri : Ii8<0xCC, MRMSrcReg, (outs VR128:$dst),
+ (ins VR128:$src1, VR128:$src2, i8imm:$src3),
+ "sha1rnds4\t{$src3, $src2, $dst|$dst, $src2, $src3}",
+ [(set VR128:$dst,
+ (int_x86_sha1rnds4 VR128:$src1, VR128:$src2,
+ (i8 imm:$src3)))]>, TA;
+ def SHA1RNDS4rmi : Ii8<0xCC, MRMSrcMem, (outs VR128:$dst),
+ (ins VR128:$src1, i128mem:$src2, i8imm:$src3),
+ "sha1rnds4\t{$src3, $src2, $dst|$dst, $src2, $src3}",
+ [(set VR128:$dst,
+ (int_x86_sha1rnds4 VR128:$src1,
+ (bc_v4i32 (memopv2i64 addr:$src2)),
+ (i8 imm:$src3)))]>, TA;
+
+ defm SHA1NEXTE : SHAI_binop<0xC8, "sha1nexte", int_x86_sha1nexte>;
+ defm SHA1MSG1 : SHAI_binop<0xC9, "sha1msg1", int_x86_sha1msg1>;
+ defm SHA1MSG2 : SHAI_binop<0xCA, "sha1msg2", int_x86_sha1msg2>;
+
+ let Uses=[XMM0] in
+ defm SHA256RNDS2 : SHAI_binop<0xCB, "sha256rnds2", int_x86_sha256rnds2, 1>;
+
+ defm SHA256MSG1 : SHAI_binop<0xCC, "sha256msg1", int_x86_sha256msg1>;
+ defm SHA256MSG2 : SHAI_binop<0xCD, "sha256msg2", int_x86_sha256msg2>;
+}
+
+// Aliases with explicit %xmm0
+def : InstAlias<"sha256rnds2\t{%xmm0, $src2, $dst|$dst, $src2, xmm0}",
+ (SHA256RNDS2rr VR128:$dst, VR128:$src2)>;
+def : InstAlias<"sha256rnds2\t{%xmm0, $src2, $dst|$dst, $src2, xmm0}",
+ (SHA256RNDS2rm VR128:$dst, i128mem:$src2)>;
+
//===----------------------------------------------------------------------===//
// AES-NI Instructions
//===----------------------------------------------------------------------===//
def VAESIMCrm : AES8I<0xDB, MRMSrcMem, (outs VR128:$dst),
(ins i128mem:$src1),
"vaesimc\t{$src1, $dst|$dst, $src1}",
- [(set VR128:$dst, (int_x86_aesni_aesimc (memopv2i64 addr:$src1)))]>,
+ [(set VR128:$dst, (int_x86_aesni_aesimc (loadv2i64 addr:$src1)))]>,
OpSize, VEX;
}
def AESIMCrr : AES8I<0xDB, MRMSrcReg, (outs VR128:$dst),
(ins i128mem:$src1, i8imm:$src2),
"vaeskeygenassist\t{$src2, $src1, $dst|$dst, $src1, $src2}",
[(set VR128:$dst,
- (int_x86_aesni_aeskeygenassist (memopv2i64 addr:$src1), imm:$src2))]>,
+ (int_x86_aesni_aeskeygenassist (loadv2i64 addr:$src1), imm:$src2))]>,
OpSize, VEX;
}
def AESKEYGENASSIST128rr : AESAI<0xDF, MRMSrcReg, (outs VR128:$dst),
(ins VR128:$src1, i128mem:$src2, i8imm:$src3),
"vpclmulqdq\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}",
[(set VR128:$dst, (int_x86_pclmulqdq VR128:$src1,
- (memopv2i64 addr:$src2), imm:$src3))]>;
+ (loadv2i64 addr:$src2), imm:$src3))]>;
// Carry-less Multiplication instructions
let Constraints = "$src1 = $dst" in {
(ins VR128:$src1, VR128:$src2, i8imm:$src3),
"pclmulqdq\t{$src3, $src2, $dst|$dst, $src2, $src3}",
[(set VR128:$dst,
- (int_x86_pclmulqdq VR128:$src1, VR128:$src2, imm:$src3))]>;
+ (int_x86_pclmulqdq VR128:$src1, VR128:$src2, imm:$src3))],
+ IIC_SSE_PCLMULQDQ_RR>;
def PCLMULQDQrm : PCLMULIi8<0x44, MRMSrcMem, (outs VR128:$dst),
(ins VR128:$src1, i128mem:$src2, i8imm:$src3),
"pclmulqdq\t{$src3, $src2, $dst|$dst, $src2, $src3}",
[(set VR128:$dst, (int_x86_pclmulqdq VR128:$src1,
- (memopv2i64 addr:$src2), imm:$src3))]>;
+ (memopv2i64 addr:$src2), imm:$src3))],
+ IIC_SSE_PCLMULQDQ_RM>;
} // Constraints = "$src1 = $dst"
(VINSERTF128rr VR256:$src1, VR128:$src2,
(INSERT_get_vinsert128_imm VR256:$ins))>;
-def : Pat<(vinsert128_insert:$ins (v8f32 VR256:$src1), (memopv4f32 addr:$src2),
+def : Pat<(vinsert128_insert:$ins (v8f32 VR256:$src1), (loadv4f32 addr:$src2),
(iPTR imm)),
(VINSERTF128rm VR256:$src1, addr:$src2,
(INSERT_get_vinsert128_imm VR256:$ins))>;
-def : Pat<(vinsert128_insert:$ins (v4f64 VR256:$src1), (memopv2f64 addr:$src2),
+def : Pat<(vinsert128_insert:$ins (v4f64 VR256:$src1), (loadv2f64 addr:$src2),
(iPTR imm)),
(VINSERTF128rm VR256:$src1, addr:$src2,
(INSERT_get_vinsert128_imm VR256:$ins))>;
(VINSERTF128rr VR256:$src1, VR128:$src2,
(INSERT_get_vinsert128_imm VR256:$ins))>;
-def : Pat<(vinsert128_insert:$ins (v4i64 VR256:$src1), (memopv2i64 addr:$src2),
+def : Pat<(vinsert128_insert:$ins (v4i64 VR256:$src1), (loadv2i64 addr:$src2),
(iPTR imm)),
(VINSERTF128rm VR256:$src1, addr:$src2,
(INSERT_get_vinsert128_imm VR256:$ins))>;
def : Pat<(vinsert128_insert:$ins (v8i32 VR256:$src1),
- (bc_v4i32 (memopv2i64 addr:$src2)),
+ (bc_v4i32 (loadv2i64 addr:$src2)),
(iPTR imm)),
(VINSERTF128rm VR256:$src1, addr:$src2,
(INSERT_get_vinsert128_imm VR256:$ins))>;
def : Pat<(vinsert128_insert:$ins (v32i8 VR256:$src1),
- (bc_v16i8 (memopv2i64 addr:$src2)),
+ (bc_v16i8 (loadv2i64 addr:$src2)),
(iPTR imm)),
(VINSERTF128rm VR256:$src1, addr:$src2,
(INSERT_get_vinsert128_imm VR256:$ins))>;
def : Pat<(vinsert128_insert:$ins (v16i16 VR256:$src1),
- (bc_v8i16 (memopv2i64 addr:$src2)),
+ (bc_v8i16 (loadv2i64 addr:$src2)),
(iPTR imm)),
(VINSERTF128rm VR256:$src1, addr:$src2,
(INSERT_get_vinsert128_imm VR256:$ins))>;
(v4f64 VR256:$src1),
(EXTRACT_get_vextract128_imm VR128:$ext)))>;
-def : Pat<(alignedstore (v4f32 (vextract128_extract:$ext (v8f32 VR256:$src1),
- (iPTR imm))), addr:$dst),
+def : Pat<(store (v4f32 (vextract128_extract:$ext (v8f32 VR256:$src1),
+ (iPTR imm))), addr:$dst),
(VEXTRACTF128mr addr:$dst, VR256:$src1,
(EXTRACT_get_vextract128_imm VR128:$ext))>;
-def : Pat<(alignedstore (v2f64 (vextract128_extract:$ext (v4f64 VR256:$src1),
- (iPTR imm))), addr:$dst),
+def : Pat<(store (v2f64 (vextract128_extract:$ext (v4f64 VR256:$src1),
+ (iPTR imm))), addr:$dst),
(VEXTRACTF128mr addr:$dst, VR256:$src1,
(EXTRACT_get_vextract128_imm VR128:$ext))>;
}
let ExeDomain = SSEPackedSingle in {
defm VPERMILPS : avx_permil<0x0C, 0x04, "vpermilps", VR128, f128mem, i128mem,
- memopv2i64, int_x86_avx_vpermilvar_ps, v4f32>;
+ loadv2i64, int_x86_avx_vpermilvar_ps, v4f32>;
defm VPERMILPSY : avx_permil<0x0C, 0x04, "vpermilps", VR256, f256mem, i256mem,
- memopv4i64, int_x86_avx_vpermilvar_ps_256, v8f32>, VEX_L;
+ loadv4i64, int_x86_avx_vpermilvar_ps_256, v8f32>, VEX_L;
}
let ExeDomain = SSEPackedDouble in {
defm VPERMILPD : avx_permil<0x0D, 0x05, "vpermilpd", VR128, f128mem, i128mem,
- memopv2i64, int_x86_avx_vpermilvar_pd, v2f64>;
+ loadv2i64, int_x86_avx_vpermilvar_pd, v2f64>;
defm VPERMILPDY : avx_permil<0x0D, 0x05, "vpermilpd", VR256, f256mem, i256mem,
- memopv4i64, int_x86_avx_vpermilvar_pd_256, v4f64>, VEX_L;
+ loadv4i64, int_x86_avx_vpermilvar_pd_256, v4f64>, VEX_L;
}
let Predicates = [HasAVX] in {
(VPERMILPSYri VR256:$src1, imm:$imm)>;
def : Pat<(v4i64 (X86VPermilp VR256:$src1, (i8 imm:$imm))),
(VPERMILPDYri VR256:$src1, imm:$imm)>;
-def : Pat<(v8i32 (X86VPermilp (bc_v8i32 (memopv4i64 addr:$src1)),
+def : Pat<(v8i32 (X86VPermilp (bc_v8i32 (loadv4i64 addr:$src1)),
(i8 imm:$imm))),
(VPERMILPSYmi addr:$src1, imm:$imm)>;
-def : Pat<(v4i64 (X86VPermilp (memopv4i64 addr:$src1), (i8 imm:$imm))),
+def : Pat<(v4i64 (X86VPermilp (loadv4i64 addr:$src1), (i8 imm:$imm))),
(VPERMILPDYmi addr:$src1, imm:$imm)>;
def : Pat<(v2i64 (X86VPermilp VR128:$src1, (i8 imm:$imm))),
(VPERMILPDri VR128:$src1, imm:$imm)>;
-def : Pat<(v2i64 (X86VPermilp (memopv2i64 addr:$src1), (i8 imm:$imm))),
+def : Pat<(v2i64 (X86VPermilp (loadv2i64 addr:$src1), (i8 imm:$imm))),
(VPERMILPDmi addr:$src1, imm:$imm)>;
}
def VPERM2F128rm : AVXAIi8<0x06, MRMSrcMem, (outs VR256:$dst),
(ins VR256:$src1, f256mem:$src2, i8imm:$src3),
"vperm2f128\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}",
- [(set VR256:$dst, (X86VPerm2x128 VR256:$src1, (memopv8f32 addr:$src2),
+ [(set VR256:$dst, (X86VPerm2x128 VR256:$src1, (loadv8f32 addr:$src2),
(i8 imm:$src3)))]>, VEX_4V, VEX_L;
}
def : Pat<(v4f64 (X86VPerm2x128 VR256:$src1, VR256:$src2, (i8 imm:$imm))),
(VPERM2F128rr VR256:$src1, VR256:$src2, imm:$imm)>;
def : Pat<(v4f64 (X86VPerm2x128 VR256:$src1,
- (memopv4f64 addr:$src2), (i8 imm:$imm))),
+ (loadv4f64 addr:$src2), (i8 imm:$imm))),
(VPERM2F128rm VR256:$src1, addr:$src2, imm:$imm)>;
}
(VPERM2F128rr VR256:$src1, VR256:$src2, imm:$imm)>;
def : Pat<(v8i32 (X86VPerm2x128 VR256:$src1,
- (bc_v8i32 (memopv4i64 addr:$src2)), (i8 imm:$imm))),
+ (bc_v8i32 (loadv4i64 addr:$src2)), (i8 imm:$imm))),
(VPERM2F128rm VR256:$src1, addr:$src2, imm:$imm)>;
def : Pat<(v4i64 (X86VPerm2x128 VR256:$src1,
- (memopv4i64 addr:$src2), (i8 imm:$imm))),
+ (loadv4i64 addr:$src2), (i8 imm:$imm))),
(VPERM2F128rm VR256:$src1, addr:$src2, imm:$imm)>;
def : Pat<(v32i8 (X86VPerm2x128 VR256:$src1,
- (bc_v32i8 (memopv4i64 addr:$src2)), (i8 imm:$imm))),
+ (bc_v32i8 (loadv4i64 addr:$src2)), (i8 imm:$imm))),
(VPERM2F128rm VR256:$src1, addr:$src2, imm:$imm)>;
def : Pat<(v16i16 (X86VPerm2x128 VR256:$src1,
- (bc_v16i16 (memopv4i64 addr:$src2)), (i8 imm:$imm))),
+ (bc_v16i16 (loadv4i64 addr:$src2)), (i8 imm:$imm))),
(VPERM2F128rm VR256:$src1, addr:$src2, imm:$imm)>;
}
TA, OpSize, VEX;
}
-let Predicates = [HasAVX, HasF16C] in {
+let Predicates = [HasF16C] in {
defm VCVTPH2PS : f16c_ph2ps<VR128, f64mem, int_x86_vcvtph2ps_128>;
defm VCVTPH2PSY : f16c_ph2ps<VR256, f128mem, int_x86_vcvtph2ps_256>, VEX_L;
defm VCVTPS2PH : f16c_ps2ph<VR128, f64mem, int_x86_vcvtps2ph_128>;
let isCommutable = 0 in {
defm VPBLENDD : AVX2_binop_rmi_int<0x02, "vpblendd", int_x86_avx2_pblendd_128,
- VR128, memopv2i64, i128mem>;
+ VR128, loadv2i64, i128mem>;
defm VPBLENDDY : AVX2_binop_rmi_int<0x02, "vpblendd", int_x86_avx2_pblendd_256,
- VR256, memopv4i64, i256mem>, VEX_L;
+ VR256, loadv4i64, i256mem>, VEX_L;
}
def : Pat<(v4i32 (X86Blendi (v4i32 VR128:$src1), (v4i32 VR128:$src2),
VEX_4V, VEX_L;
}
-defm VPERMD : avx2_perm<0x36, "vpermd", memopv4i64, v8i32>;
+defm VPERMD : avx2_perm<0x36, "vpermd", loadv4i64, v8i32>;
let ExeDomain = SSEPackedSingle in
-defm VPERMPS : avx2_perm<0x16, "vpermps", memopv8f32, v8f32>;
+defm VPERMPS : avx2_perm<0x16, "vpermps", loadv8f32, v8f32>;
multiclass avx2_perm_imm<bits<8> opc, string OpcodeStr, PatFrag mem_frag,
ValueType OpVT> {
(i8 imm:$src2))))]>, VEX, VEX_L;
}
-defm VPERMQ : avx2_perm_imm<0x00, "vpermq", memopv4i64, v4i64>, VEX_W;
+defm VPERMQ : avx2_perm_imm<0x00, "vpermq", loadv4i64, v4i64>, VEX_W;
let ExeDomain = SSEPackedDouble in
-defm VPERMPD : avx2_perm_imm<0x01, "vpermpd", memopv4f64, v4f64>, VEX_W;
+defm VPERMPD : avx2_perm_imm<0x01, "vpermpd", loadv4f64, v4f64>, VEX_W;
//===----------------------------------------------------------------------===//
// VPERM2I128 - Permute Floating-Point Values in 128-bit chunks
def VPERM2I128rm : AVX2AIi8<0x46, MRMSrcMem, (outs VR256:$dst),
(ins VR256:$src1, f256mem:$src2, i8imm:$src3),
"vperm2i128\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}",
- [(set VR256:$dst, (X86VPerm2x128 VR256:$src1, (memopv4i64 addr:$src2),
+ [(set VR256:$dst, (X86VPerm2x128 VR256:$src1, (loadv4i64 addr:$src2),
(i8 imm:$src3)))]>, VEX_4V, VEX_L;
let Predicates = [HasAVX2] in {
def : Pat<(v16i16 (X86VPerm2x128 VR256:$src1, VR256:$src2, (i8 imm:$imm))),
(VPERM2I128rr VR256:$src1, VR256:$src2, imm:$imm)>;
-def : Pat<(v32i8 (X86VPerm2x128 VR256:$src1, (bc_v32i8 (memopv4i64 addr:$src2)),
+def : Pat<(v32i8 (X86VPerm2x128 VR256:$src1, (bc_v32i8 (loadv4i64 addr:$src2)),
(i8 imm:$imm))),
(VPERM2I128rm VR256:$src1, addr:$src2, imm:$imm)>;
def : Pat<(v16i16 (X86VPerm2x128 VR256:$src1,
- (bc_v16i16 (memopv4i64 addr:$src2)), (i8 imm:$imm))),
+ (bc_v16i16 (loadv4i64 addr:$src2)), (i8 imm:$imm))),
(VPERM2I128rm VR256:$src1, addr:$src2, imm:$imm)>;
-def : Pat<(v8i32 (X86VPerm2x128 VR256:$src1, (bc_v8i32 (memopv4i64 addr:$src2)),
+def : Pat<(v8i32 (X86VPerm2x128 VR256:$src1, (bc_v8i32 (loadv4i64 addr:$src2)),
(i8 imm:$imm))),
(VPERM2I128rm VR256:$src1, addr:$src2, imm:$imm)>;
}
(VINSERTI128rr VR256:$src1, VR128:$src2,
(INSERT_get_vinsert128_imm VR256:$ins))>;
-def : Pat<(vinsert128_insert:$ins (v4i64 VR256:$src1), (memopv2i64 addr:$src2),
+def : Pat<(vinsert128_insert:$ins (v4i64 VR256:$src1), (loadv2i64 addr:$src2),
(iPTR imm)),
(VINSERTI128rm VR256:$src1, addr:$src2,
(INSERT_get_vinsert128_imm VR256:$ins))>;
def : Pat<(vinsert128_insert:$ins (v8i32 VR256:$src1),
- (bc_v4i32 (memopv2i64 addr:$src2)),
+ (bc_v4i32 (loadv2i64 addr:$src2)),
(iPTR imm)),
(VINSERTI128rm VR256:$src1, addr:$src2,
(INSERT_get_vinsert128_imm VR256:$ins))>;
def : Pat<(vinsert128_insert:$ins (v32i8 VR256:$src1),
- (bc_v16i8 (memopv2i64 addr:$src2)),
+ (bc_v16i8 (loadv2i64 addr:$src2)),
(iPTR imm)),
(VINSERTI128rm VR256:$src1, addr:$src2,
(INSERT_get_vinsert128_imm VR256:$ins))>;
def : Pat<(vinsert128_insert:$ins (v16i16 VR256:$src1),
- (bc_v8i16 (memopv2i64 addr:$src2)),
+ (bc_v8i16 (loadv2i64 addr:$src2)),
(iPTR imm)),
(VINSERTI128rm VR256:$src1, addr:$src2,
(INSERT_get_vinsert128_imm VR256:$ins))>;
(v32i8 VR256:$src1),
(EXTRACT_get_vextract128_imm VR128:$ext)))>;
-def : Pat<(alignedstore (v2i64 (vextract128_extract:$ext (v4i64 VR256:$src1),
- (iPTR imm))), addr:$dst),
+def : Pat<(store (v2i64 (vextract128_extract:$ext (v4i64 VR256:$src1),
+ (iPTR imm))), addr:$dst),
(VEXTRACTI128mr addr:$dst, VR256:$src1,
(EXTRACT_get_vextract128_imm VR128:$ext))>;
-def : Pat<(alignedstore (v4i32 (vextract128_extract:$ext (v8i32 VR256:$src1),
- (iPTR imm))), addr:$dst),
+def : Pat<(store (v4i32 (vextract128_extract:$ext (v8i32 VR256:$src1),
+ (iPTR imm))), addr:$dst),
(VEXTRACTI128mr addr:$dst, VR256:$src1,
(EXTRACT_get_vextract128_imm VR128:$ext))>;
-def : Pat<(alignedstore (v8i16 (vextract128_extract:$ext (v16i16 VR256:$src1),
- (iPTR imm))), addr:$dst),
+def : Pat<(store (v8i16 (vextract128_extract:$ext (v16i16 VR256:$src1),
+ (iPTR imm))), addr:$dst),
(VEXTRACTI128mr addr:$dst, VR256:$src1,
(EXTRACT_get_vextract128_imm VR128:$ext))>;
-def : Pat<(alignedstore (v16i8 (vextract128_extract:$ext (v32i8 VR256:$src1),
- (iPTR imm))), addr:$dst),
+def : Pat<(store (v16i8 (vextract128_extract:$ext (v32i8 VR256:$src1),
+ (iPTR imm))), addr:$dst),
(VEXTRACTI128mr addr:$dst, VR256:$src1,
(EXTRACT_get_vextract128_imm VR128:$ext))>;
}
!strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
[(set VR128:$dst,
(vt128 (OpNode VR128:$src1,
- (vt128 (bitconvert (memopv2i64 addr:$src2))))))]>,
+ (vt128 (bitconvert (loadv2i64 addr:$src2))))))]>,
VEX_4V;
def Yrr : AVX28I<opc, MRMSrcReg, (outs VR256:$dst),
(ins VR256:$src1, VR256:$src2),
!strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
[(set VR256:$dst,
(vt256 (OpNode VR256:$src1,
- (vt256 (bitconvert (memopv4i64 addr:$src2))))))]>,
+ (vt256 (bitconvert (loadv4i64 addr:$src2))))))]>,
VEX_4V, VEX_L;
}
projects / oota-llvm.git / blobdiff