multiclass sse12_cvt_p<bits<8> opc, RegisterClass SrcRC, RegisterClass DstRC,
SDNode OpNode, X86MemOperand x86memop, PatFrag ld_frag,
string asm, Domain d, OpndItins itins> {
- def rr : PI<opc, MRMSrcReg, (outs DstRC:$dst), (ins SrcRC:$src), asm,
+ def rr : I<opc, MRMSrcReg, (outs DstRC:$dst), (ins SrcRC:$src), asm,
[(set DstRC:$dst, (OpNode SrcRC:$src))],
itins.rr, d>;
- def rm : PI<opc, MRMSrcMem, (outs DstRC:$dst), (ins x86memop:$src), asm,
+ def rm : I<opc, MRMSrcMem, (outs DstRC:$dst), (ins x86memop:$src), asm,
[(set DstRC:$dst, (OpNode (ld_frag addr:$src)))],
itins.rm, d>;
}
"cvttsd2si{q}", SSE_CVT_SD2SI>,
XD, REX_W;
-let Pattern = []<dag> in {
+let Pattern = []<dag>, neverHasSideEffects = 1 in {
defm VCVTSS2SI : sse12_cvt_s<0x2D, FR32, GR32, undef, f32mem, load,
"cvtss2si{l}\t{$src, $dst|$dst, $src}",
SSE_CVT_SS2SI_32>, XS, VEX, VEX_LIG;
"cvtss2si\t{$src, $dst|$dst, $src}",
SSE_CVT_SS2SI_64>, XS, VEX, VEX_W, VEX_LIG;
defm VCVTDQ2PS : sse12_cvt_p<0x5B, VR128, VR128, undef, i128mem, load,
- "cvtdq2ps\t{$src, $dst|$dst, $src}",
- SSEPackedSingle, SSE_CVT_PS>, TB, VEX;
+ "vcvtdq2ps\t{$src, $dst|$dst, $src}",
+ SSEPackedSingle, SSE_CVT_PS>, TB, VEX,
+ Requires<[HasAVX]>;
defm VCVTDQ2PSY : sse12_cvt_p<0x5B, VR256, VR256, undef, i256mem, load,
- "cvtdq2ps\t{$src, $dst|$dst, $src}",
- SSEPackedSingle, SSE_CVT_PS>, TB, VEX;
+ "vcvtdq2ps\t{$src, $dst|$dst, $src}",
+ SSEPackedSingle, SSE_CVT_PS>, TB, VEX,
+ Requires<[HasAVX]>;
}
-let Pattern = []<dag> in {
+let Pattern = []<dag>, neverHasSideEffects = 1 in {
defm CVTSS2SI : sse12_cvt_s<0x2D, FR32, GR32, undef, f32mem, load /*dummy*/,
"cvtss2si{l}\t{$src, $dst|$dst, $src}",
SSE_CVT_SS2SI_32>, XS;
SSE_CVT_SS2SI_64>, XS, REX_W;
defm CVTDQ2PS : sse12_cvt_p<0x5B, VR128, VR128, undef, i128mem, load /*dummy*/,
"cvtdq2ps\t{$src, $dst|$dst, $src}",
- SSEPackedSingle, SSE_CVT_PS>,
- TB; /* PD SSE3 form is avaiable */
+ SSEPackedSingle, SSE_CVT_PS>, TB,
+ Requires<[HasSSE2]>;
}
let Predicates = [HasAVX] in {
Requires<[HasSSE2]>;
}
-// Convert doubleword to packed single/double fp
-// SSE2 instructions without OpSize prefix
-def Int_VCVTDQ2PSrr : I<0x5B, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
- "vcvtdq2ps\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst, (int_x86_sse2_cvtdq2ps VR128:$src))],
- IIC_SSE_CVT_PS_RR>,
- TB, VEX, Requires<[HasAVX]>;
-def Int_VCVTDQ2PSrm : I<0x5B, MRMSrcMem, (outs VR128:$dst), (ins i128mem:$src),
- "vcvtdq2ps\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst, (int_x86_sse2_cvtdq2ps
- (bitconvert (memopv2i64 addr:$src))))],
- IIC_SSE_CVT_PS_RM>,
- TB, VEX, Requires<[HasAVX]>;
-def Int_CVTDQ2PSrr : I<0x5B, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
- "cvtdq2ps\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst, (int_x86_sse2_cvtdq2ps VR128:$src))],
- IIC_SSE_CVT_PS_RR>,
- TB, Requires<[HasSSE2]>;
-def Int_CVTDQ2PSrm : I<0x5B, MRMSrcMem, (outs VR128:$dst), (ins i128mem:$src),
- "cvtdq2ps\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst, (int_x86_sse2_cvtdq2ps
- (bitconvert (memopv2i64 addr:$src))))],
- IIC_SSE_CVT_PS_RM>,
- TB, Requires<[HasSSE2]>;
-
-// FIXME: why the non-intrinsic version is described as SSE3?
-// SSE2 instructions with XS prefix
-def Int_VCVTDQ2PDrr : I<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
- "vcvtdq2pd\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst, (int_x86_sse2_cvtdq2pd VR128:$src))],
- IIC_SSE_CVT_PD_RR>,
- XS, VEX, Requires<[HasAVX]>;
-def Int_VCVTDQ2PDrm : I<0xE6, MRMSrcMem, (outs VR128:$dst), (ins i64mem:$src),
- "vcvtdq2pd\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst, (int_x86_sse2_cvtdq2pd
- (bitconvert (memopv2i64 addr:$src))))],
- IIC_SSE_CVT_PD_RM>,
- XS, VEX, Requires<[HasAVX]>;
-def Int_CVTDQ2PDrr : I<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
- "cvtdq2pd\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst, (int_x86_sse2_cvtdq2pd VR128:$src))],
- IIC_SSE_CVT_PD_RR>,
- XS, Requires<[HasSSE2]>;
-def Int_CVTDQ2PDrm : I<0xE6, MRMSrcMem, (outs VR128:$dst), (ins i64mem:$src),
- "cvtdq2pd\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst, (int_x86_sse2_cvtdq2pd
- (bitconvert (memopv2i64 addr:$src))))],
- IIC_SSE_CVT_PD_RM>,
- XS, Requires<[HasSSE2]>;
-
-
// Convert packed single/double fp to doubleword
def VCVTPS2DQrr : VPDI<0x5B, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
"cvtps2dq\t{$src, $dst|$dst, $src}", [],
"cvtps2dq\t{$src, $dst|$dst, $src}", [],
IIC_SSE_CVT_PS_RM>;
-def Int_VCVTPS2DQrr : VPDI<0x5B, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
- "cvtps2dq\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst, (int_x86_sse2_cvtps2dq VR128:$src))],
- IIC_SSE_CVT_PS_RR>,
- VEX;
-def Int_VCVTPS2DQrm : VPDI<0x5B, MRMSrcMem, (outs VR128:$dst),
- (ins f128mem:$src),
- "cvtps2dq\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst, (int_x86_sse2_cvtps2dq
- (memop addr:$src)))],
- IIC_SSE_CVT_PS_RM>, VEX;
-def Int_CVTPS2DQrr : PDI<0x5B, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
- "cvtps2dq\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst, (int_x86_sse2_cvtps2dq VR128:$src))],
- IIC_SSE_CVT_PS_RR>;
-def Int_CVTPS2DQrm : PDI<0x5B, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
- "cvtps2dq\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst, (int_x86_sse2_cvtps2dq
- (memop addr:$src)))],
- IIC_SSE_CVT_PS_RM>;
-
-// SSE2 packed instructions with XD prefix
-def Int_VCVTPD2DQrr : I<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
- "vcvtpd2dq\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst, (int_x86_sse2_cvtpd2dq VR128:$src))],
- IIC_SSE_CVT_PD_RR>,
- XD, VEX, Requires<[HasAVX]>;
-def Int_VCVTPD2DQrm : I<0xE6, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
- "vcvtpd2dq\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst, (int_x86_sse2_cvtpd2dq
- (memop addr:$src)))],
- IIC_SSE_CVT_PD_RM>,
- XD, VEX, Requires<[HasAVX]>;
-def Int_CVTPD2DQrr : I<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
- "cvtpd2dq\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst, (int_x86_sse2_cvtpd2dq VR128:$src))],
- IIC_SSE_CVT_PD_RR>,
- XD, Requires<[HasSSE2]>;
-def Int_CVTPD2DQrm : I<0xE6, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
- "cvtpd2dq\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst, (int_x86_sse2_cvtpd2dq
- (memop addr:$src)))],
- IIC_SSE_CVT_PD_RM>,
- XD, Requires<[HasSSE2]>;
+let Predicates = [HasAVX] in {
+ def : Pat<(int_x86_sse2_cvtps2dq VR128:$src),
+ (VCVTPS2DQrr VR128:$src)>;
+ def : Pat<(int_x86_sse2_cvtps2dq (memopv4f32 addr:$src)),
+ (VCVTPS2DQrm addr:$src)>;
+}
+let Predicates = [HasSSE2] in {
+ def : Pat<(int_x86_sse2_cvtps2dq VR128:$src),
+ (CVTPS2DQrr VR128:$src)>;
+ def : Pat<(int_x86_sse2_cvtps2dq (memopv4f32 addr:$src)),
+ (CVTPS2DQrm addr:$src)>;
+}
+
+// Convert Packed Double FP to Packed DW Integers
+let Predicates = [HasAVX] in {
+// The assembler can recognize rr 256-bit instructions by seeing a ymm
+// register, but the same isn't true when using memory operands instead.
+// Provide other assembly rr and rm forms to address this explicitly.
+def VCVTPD2DQrr : SDI<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
+ "vcvtpd2dq\t{$src, $dst|$dst, $src}", []>, VEX;
+
+// XMM only
+def : InstAlias<"vcvtpd2dqx\t{$src, $dst|$dst, $src}",
+ (VCVTPD2DQrr VR128:$dst, VR128:$src)>;
+def VCVTPD2DQXrm : SDI<0xE6, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
+ "vcvtpd2dqx\t{$src, $dst|$dst, $src}", []>, VEX;
+
+// YMM only
+def VCVTPD2DQYrr : SDI<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR256:$src),
+ "vcvtpd2dq{y}\t{$src, $dst|$dst, $src}", []>, VEX;
+def VCVTPD2DQYrm : SDI<0xE6, MRMSrcMem, (outs VR128:$dst), (ins f256mem:$src),
+ "vcvtpd2dq{y}\t{$src, $dst|$dst, $src}", []>, VEX, VEX_L;
+def : InstAlias<"vcvtpd2dq\t{$src, $dst|$dst, $src}",
+ (VCVTPD2DQYrr VR128:$dst, VR256:$src)>;
+}
+
+def CVTPD2DQrm : SDI<0xE6, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
+ "cvtpd2dq\t{$src, $dst|$dst, $src}", [],
+ IIC_SSE_CVT_PD_RM>;
+def CVTPD2DQrr : SDI<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
+ "cvtpd2dq\t{$src, $dst|$dst, $src}", [],
+ IIC_SSE_CVT_PD_RR>;
+
+let Predicates = [HasAVX] in {
+ def : Pat<(int_x86_sse2_cvtpd2dq VR128:$src),
+ (VCVTPD2DQrr VR128:$src)>;
+ def : Pat<(int_x86_sse2_cvtpd2dq (memopv2f64 addr:$src)),
+ (VCVTPD2DQXrm addr:$src)>;
+}
+
+let Predicates = [HasSSE2] in {
+ def : Pat<(int_x86_sse2_cvtpd2dq VR128:$src),
+ (CVTPD2DQrr VR128:$src)>;
+ def : Pat<(int_x86_sse2_cvtpd2dq (memopv2f64 addr:$src)),
+ (CVTPD2DQrm addr:$src)>;
+}
// Convert with truncation packed single/double fp to doubleword
// SSE2 packed instructions with XS prefix
def VCVTTPS2DQrm : VSSI<0x5B, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
"cvttps2dq\t{$src, $dst|$dst, $src}",
[(set VR128:$dst, (int_x86_sse2_cvttps2dq
- (memop addr:$src)))],
+ (memopv4f32 addr:$src)))],
IIC_SSE_CVT_PS_RM>, VEX;
def VCVTTPS2DQYrr : VSSI<0x5B, MRMSrcReg, (outs VR256:$dst), (ins VR256:$src),
"cvttps2dq\t{$src, $dst|$dst, $src}",
def CVTTPS2DQrm : SSI<0x5B, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
"cvttps2dq\t{$src, $dst|$dst, $src}",
[(set VR128:$dst,
- (int_x86_sse2_cvttps2dq (memop addr:$src)))],
+ (int_x86_sse2_cvttps2dq (memopv4f32 addr:$src)))],
IIC_SSE_CVT_PS_RM>;
let Predicates = [HasAVX] in {
def : Pat<(v4f32 (sint_to_fp (v4i32 VR128:$src))),
- (Int_VCVTDQ2PSrr VR128:$src)>;
+ (VCVTDQ2PSrr VR128:$src)>;
def : Pat<(v4f32 (sint_to_fp (bc_v4i32 (memopv2i64 addr:$src)))),
- (Int_VCVTDQ2PSrm 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:$src))),
+ (VCVTDQ2PSrm addr:$src)>;
def : Pat<(v4i32 (fp_to_sint (v4f32 VR128:$src))),
(VCVTTPS2DQrr VR128:$src)>;
let Predicates = [HasSSE2] in {
def : Pat<(v4f32 (sint_to_fp (v4i32 VR128:$src))),
- (Int_CVTDQ2PSrr VR128:$src)>;
+ (CVTDQ2PSrr VR128:$src)>;
def : Pat<(v4f32 (sint_to_fp (bc_v4i32 (memopv2i64 addr:$src)))),
- (Int_CVTDQ2PSrm addr:$src)>;
+ (CVTDQ2PSrm addr:$src)>;
+
+ def : Pat<(int_x86_sse2_cvtdq2ps VR128:$src),
+ (CVTDQ2PSrr VR128:$src)>;
+ def : Pat<(int_x86_sse2_cvtdq2ps (bc_v4i32 (memopv2i64 addr:$src))),
+ (CVTDQ2PSrm addr:$src)>;
def : Pat<(v4i32 (fp_to_sint (v4f32 VR128:$src))),
(CVTTPS2DQrr VR128:$src)>;
[(set VR128:$dst,
(int_x86_sse2_cvttpd2dq VR128:$src))],
IIC_SSE_CVT_PD_RR>, VEX;
-let isCodeGenOnly = 1 in
-def VCVTTPD2DQrm : VPDI<0xE6, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
- "cvttpd2dq\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst, (int_x86_sse2_cvttpd2dq
- (memop addr:$src)))],
- IIC_SSE_CVT_PD_RM>, VEX;
+
def CVTTPD2DQrr : PDI<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
"cvttpd2dq\t{$src, $dst|$dst, $src}",
[(set VR128:$dst, (int_x86_sse2_cvttpd2dq VR128:$src))],
def CVTTPD2DQrm : PDI<0xE6, MRMSrcMem, (outs VR128:$dst),(ins f128mem:$src),
"cvttpd2dq\t{$src, $dst|$dst, $src}",
[(set VR128:$dst, (int_x86_sse2_cvttpd2dq
- (memop addr:$src)))],
+ (memopv2f64 addr:$src)))],
IIC_SSE_CVT_PD_RM>;
// The assembler can recognize rr 256-bit instructions by seeing a ymm
// register, but the same isn't true when using memory operands instead.
// Provide other assembly rr and rm forms to address this explicitly.
-def VCVTTPD2DQXrYr : VPDI<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR256:$src),
- "cvttpd2dq\t{$src, $dst|$dst, $src}", [],
- IIC_SSE_CVT_PD_RR>, VEX;
// XMM only
-def VCVTTPD2DQXrr : VPDI<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
- "cvttpd2dqx\t{$src, $dst|$dst, $src}", [],
- IIC_SSE_CVT_PD_RR>, VEX;
+def : InstAlias<"vcvttpd2dqx\t{$src, $dst|$dst, $src}",
+ (VCVTTPD2DQrr VR128:$dst, VR128:$src)>;
def VCVTTPD2DQXrm : VPDI<0xE6, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
- "cvttpd2dqx\t{$src, $dst|$dst, $src}", [],
+ "cvttpd2dqx\t{$src, $dst|$dst, $src}",
+ [(set VR128:$dst, (int_x86_sse2_cvttpd2dq
+ (memopv2f64 addr:$src)))],
IIC_SSE_CVT_PD_RM>, VEX;
// YMM only
def VCVTTPD2DQYrr : VPDI<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR256:$src),
- "cvttpd2dqy\t{$src, $dst|$dst, $src}", [],
+ "cvttpd2dq{y}\t{$src, $dst|$dst, $src}", [],
IIC_SSE_CVT_PD_RR>, VEX;
def VCVTTPD2DQYrm : VPDI<0xE6, MRMSrcMem, (outs VR128:$dst), (ins f256mem:$src),
- "cvttpd2dqy\t{$src, $dst|$dst, $src}", [],
+ "cvttpd2dq{y}\t{$src, $dst|$dst, $src}", [],
IIC_SSE_CVT_PD_RM>, VEX, VEX_L;
+def : InstAlias<"vcvttpd2dq\t{$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))),
+ (VCVTTPD2DQYrm addr:$src)>;
+} // Predicates = [HasAVX]
// Convert packed single to packed double
let Predicates = [HasAVX] in {
"vcvtps2pd\t{$src, $dst|$dst, $src}", [],
IIC_SSE_CVT_PD_RM>, TB, VEX;
}
+
+let Predicates = [HasSSE2] in {
def CVTPS2PDrr : I<0x5A, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
"cvtps2pd\t{$src, $dst|$dst, $src}", [],
IIC_SSE_CVT_PD_RR>, TB;
def CVTPS2PDrm : I<0x5A, MRMSrcMem, (outs VR128:$dst), (ins f64mem:$src),
"cvtps2pd\t{$src, $dst|$dst, $src}", [],
IIC_SSE_CVT_PD_RM>, TB;
+}
+
+let Predicates = [HasAVX] in {
+ def : Pat<(int_x86_sse2_cvtps2pd VR128:$src),
+ (VCVTPS2PDrr VR128:$src)>;
+}
+
+let Predicates = [HasSSE2] in {
+ def : Pat<(int_x86_sse2_cvtps2pd VR128:$src),
+ (CVTPS2PDrr VR128:$src)>;
+}
+
+// Convert Packed DW Integers to Packed Double FP
+let Predicates = [HasAVX] in {
+def VCVTDQ2PDrm : SSDI<0xE6, MRMSrcMem, (outs VR128:$dst), (ins i64mem:$src),
+ "vcvtdq2pd\t{$src, $dst|$dst, $src}", []>, VEX;
+def VCVTDQ2PDrr : SSDI<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
+ "vcvtdq2pd\t{$src, $dst|$dst, $src}", []>, VEX;
+def VCVTDQ2PDYrm : SSDI<0xE6, MRMSrcMem, (outs VR256:$dst), (ins i128mem:$src),
+ "vcvtdq2pd\t{$src, $dst|$dst, $src}", []>, VEX;
+def VCVTDQ2PDYrr : SSDI<0xE6, MRMSrcReg, (outs VR256:$dst), (ins VR128:$src),
+ "vcvtdq2pd\t{$src, $dst|$dst, $src}", []>, VEX;
+}
+
+def CVTDQ2PDrm : SSDI<0xE6, MRMSrcMem, (outs VR128:$dst), (ins i64mem:$src),
+ "cvtdq2pd\t{$src, $dst|$dst, $src}", [],
+ IIC_SSE_CVT_PD_RR>;
+def CVTDQ2PDrr : SSDI<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
+ "cvtdq2pd\t{$src, $dst|$dst, $src}", [],
+ IIC_SSE_CVT_PD_RM>;
+
+// 128 bit register conversion intrinsics
+let Predicates = [HasAVX] in
+def : Pat<(int_x86_sse2_cvtdq2pd VR128:$src),
+ (VCVTDQ2PDrr VR128:$src)>;
-def Int_VCVTPS2PDrr : I<0x5A, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
- "vcvtps2pd\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst, (int_x86_sse2_cvtps2pd VR128:$src))],
- IIC_SSE_CVT_PD_RR>,
- TB, VEX, Requires<[HasAVX]>;
-def Int_VCVTPS2PDrm : I<0x5A, MRMSrcMem, (outs VR128:$dst), (ins f64mem:$src),
- "vcvtps2pd\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst, (int_x86_sse2_cvtps2pd
- (load addr:$src)))],
- IIC_SSE_CVT_PD_RM>,
- TB, VEX, Requires<[HasAVX]>;
-def Int_CVTPS2PDrr : I<0x5A, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
- "cvtps2pd\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst, (int_x86_sse2_cvtps2pd VR128:$src))],
- IIC_SSE_CVT_PD_RR>,
- TB, Requires<[HasSSE2]>;
-def Int_CVTPS2PDrm : I<0x5A, MRMSrcMem, (outs VR128:$dst), (ins f64mem:$src),
- "cvtps2pd\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst, (int_x86_sse2_cvtps2pd
- (load addr:$src)))],
- IIC_SSE_CVT_PD_RM>,
- TB, Requires<[HasSSE2]>;
+let Predicates = [HasSSE2] in
+def : Pat<(int_x86_sse2_cvtdq2pd VR128:$src),
+ (CVTDQ2PDrr VR128:$src)>;
+
+// AVX 256-bit register conversion intrinsics
+let Predicates = [HasAVX] in {
+ def : Pat<(int_x86_avx_cvtdq2_pd_256 VR128:$src),
+ (VCVTDQ2PDYrr VR128:$src)>;
+ def : Pat<(int_x86_avx_cvtdq2_pd_256 (bitconvert (memopv2i64 addr:$src))),
+ (VCVTDQ2PDYrm addr:$src)>;
+
+ def : Pat<(int_x86_avx_cvt_pd2dq_256 VR256:$src),
+ (VCVTPD2DQYrr VR256:$src)>;
+ def : Pat<(int_x86_avx_cvt_pd2dq_256 (memopv4f64 addr:$src)),
+ (VCVTPD2DQYrm addr:$src)>;
+
+ def : Pat<(v4f64 (sint_to_fp (v4i32 VR128:$src))),
+ (VCVTDQ2PDYrr VR128:$src)>;
+ def : Pat<(v4f64 (sint_to_fp (bc_v4i32 (memopv2i64 addr:$src)))),
+ (VCVTDQ2PDYrm addr:$src)>;
+} // Predicates = [HasAVX]
// Convert packed double to packed single
// The assembler can recognize rr 256-bit instructions by seeing a ymm
def VCVTPD2PSrr : VPDI<0x5A, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
"cvtpd2ps\t{$src, $dst|$dst, $src}", [],
IIC_SSE_CVT_PD_RR>, VEX;
-def VCVTPD2PSXrYr : VPDI<0x5A, MRMSrcReg, (outs VR128:$dst), (ins VR256:$src),
- "cvtpd2ps\t{$src, $dst|$dst, $src}", [],
- IIC_SSE_CVT_PD_RR>, VEX;
// XMM only
-def VCVTPD2PSXrr : VPDI<0x5A, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
- "cvtpd2psx\t{$src, $dst|$dst, $src}", [],
- IIC_SSE_CVT_PD_RR>, VEX;
+def : InstAlias<"vcvtpd2psx\t{$src, $dst|$dst, $src}",
+ (VCVTPD2PSrr VR128:$dst, VR128:$src)>;
def VCVTPD2PSXrm : VPDI<0x5A, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
"cvtpd2psx\t{$src, $dst|$dst, $src}", [],
IIC_SSE_CVT_PD_RM>, VEX;
// YMM only
def VCVTPD2PSYrr : VPDI<0x5A, MRMSrcReg, (outs VR128:$dst), (ins VR256:$src),
- "cvtpd2psy\t{$src, $dst|$dst, $src}", [],
+ "cvtpd2ps{y}\t{$src, $dst|$dst, $src}", [],
IIC_SSE_CVT_PD_RR>, VEX;
def VCVTPD2PSYrm : VPDI<0x5A, MRMSrcMem, (outs VR128:$dst), (ins f256mem:$src),
- "cvtpd2psy\t{$src, $dst|$dst, $src}", [],
+ "cvtpd2ps{y}\t{$src, $dst|$dst, $src}", [],
IIC_SSE_CVT_PD_RM>, VEX, VEX_L;
+def : InstAlias<"vcvtpd2ps\t{$src, $dst|$dst, $src}",
+ (VCVTPD2PSYrr VR128:$dst, VR256:$src)>;
+
def CVTPD2PSrr : PDI<0x5A, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
"cvtpd2ps\t{$src, $dst|$dst, $src}", [],
IIC_SSE_CVT_PD_RR>;
IIC_SSE_CVT_PD_RM>;
-def Int_VCVTPD2PSrr : VPDI<0x5A, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
- "cvtpd2ps\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst, (int_x86_sse2_cvtpd2ps VR128:$src))],
- IIC_SSE_CVT_PD_RR>;
-def Int_VCVTPD2PSrm : VPDI<0x5A, MRMSrcMem, (outs VR128:$dst),
- (ins f128mem:$src),
- "cvtpd2ps\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst, (int_x86_sse2_cvtpd2ps
- (memop addr:$src)))],
- IIC_SSE_CVT_PD_RM>;
-def Int_CVTPD2PSrr : PDI<0x5A, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
- "cvtpd2ps\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst, (int_x86_sse2_cvtpd2ps VR128:$src))],
- IIC_SSE_CVT_PD_RR>;
-def Int_CVTPD2PSrm : PDI<0x5A, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
- "cvtpd2ps\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst, (int_x86_sse2_cvtpd2ps
- (memop addr:$src)))],
- IIC_SSE_CVT_PD_RM>;
+let Predicates = [HasAVX] in {
+ def : Pat<(int_x86_sse2_cvtpd2ps VR128:$src),
+ (VCVTPD2PSrr VR128:$src)>;
+ def : Pat<(int_x86_sse2_cvtpd2ps (memopv2f64 addr:$src)),
+ (VCVTPD2PSXrm addr:$src)>;
+}
+
+let Predicates = [HasSSE2] in {
+ def : Pat<(int_x86_sse2_cvtpd2ps VR128:$src),
+ (CVTPD2PSrr VR128:$src)>;
+ def : Pat<(int_x86_sse2_cvtpd2ps (memopv2f64 addr:$src)),
+ (CVTPD2PSrm addr:$src)>;
+}
// AVX 256-bit register conversion intrinsics
// FIXME: Migrate SSE conversion intrinsics matching to use patterns as below
// whenever possible to avoid declaring two versions of each one.
-def : Pat<(int_x86_avx_cvtdq2_ps_256 VR256:$src),
- (VCVTDQ2PSYrr VR256:$src)>;
-def : Pat<(int_x86_avx_cvtdq2_ps_256 (bitconvert (memopv4i64 addr:$src))),
- (VCVTDQ2PSYrm addr:$src)>;
-
-def : Pat<(int_x86_avx_cvt_pd2_ps_256 VR256:$src),
- (VCVTPD2PSYrr VR256:$src)>;
-def : Pat<(int_x86_avx_cvt_pd2_ps_256 (memopv4f64 addr:$src)),
- (VCVTPD2PSYrm addr:$src)>;
-
-def : Pat<(int_x86_avx_cvt_ps2dq_256 VR256:$src),
- (VCVTPS2DQYrr VR256:$src)>;
-def : Pat<(int_x86_avx_cvt_ps2dq_256 (memopv8f32 addr:$src)),
- (VCVTPS2DQYrm addr:$src)>;
-
-def : Pat<(int_x86_avx_cvt_ps2_pd_256 VR128:$src),
- (VCVTPS2PDYrr VR128:$src)>;
-def : Pat<(int_x86_avx_cvt_ps2_pd_256 (memopv4f32 addr:$src)),
- (VCVTPS2PDYrm addr:$src)>;
-
-def : Pat<(int_x86_avx_cvtt_pd2dq_256 VR256:$src),
- (VCVTTPD2DQYrr VR256:$src)>;
-def : Pat<(int_x86_avx_cvtt_pd2dq_256 (memopv4f64 addr:$src)),
- (VCVTTPD2DQYrm addr:$src)>;
-
-// Match fround and fextend for 128/256-bit conversions
-def : Pat<(v4f32 (fround (v4f64 VR256:$src))),
- (VCVTPD2PSYrr VR256:$src)>;
-def : Pat<(v4f32 (fround (loadv4f64 addr:$src))),
- (VCVTPD2PSYrm addr:$src)>;
-
-def : Pat<(v4f64 (fextend (v4f32 VR128:$src))),
- (VCVTPS2PDYrr VR128:$src)>;
-def : Pat<(v4f64 (fextend (loadv4f32 addr:$src))),
- (VCVTPS2PDYrm addr:$src)>;
+let Predicates = [HasAVX] in {
+ def : Pat<(int_x86_avx_cvtdq2_ps_256 VR256:$src),
+ (VCVTDQ2PSYrr VR256:$src)>;
+ def : Pat<(int_x86_avx_cvtdq2_ps_256 (bitconvert (memopv4i64 addr:$src))),
+ (VCVTDQ2PSYrm addr:$src)>;
+
+ def : Pat<(int_x86_avx_cvt_pd2_ps_256 VR256:$src),
+ (VCVTPD2PSYrr VR256:$src)>;
+ def : Pat<(int_x86_avx_cvt_pd2_ps_256 (memopv4f64 addr:$src)),
+ (VCVTPD2PSYrm addr:$src)>;
+
+ def : Pat<(int_x86_avx_cvt_ps2dq_256 VR256:$src),
+ (VCVTPS2DQYrr VR256:$src)>;
+ def : Pat<(int_x86_avx_cvt_ps2dq_256 (memopv8f32 addr:$src)),
+ (VCVTPS2DQYrm addr:$src)>;
+
+ def : Pat<(int_x86_avx_cvt_ps2_pd_256 VR128:$src),
+ (VCVTPS2PDYrr VR128:$src)>;
+ def : Pat<(int_x86_avx_cvt_ps2_pd_256 (memopv4f32 addr:$src)),
+ (VCVTPS2PDYrm addr:$src)>;
+
+ def : Pat<(int_x86_avx_cvtt_pd2dq_256 VR256:$src),
+ (VCVTTPD2DQYrr VR256:$src)>;
+ def : Pat<(int_x86_avx_cvtt_pd2dq_256 (memopv4f64 addr:$src)),
+ (VCVTTPD2DQYrm addr:$src)>;
+
+ // Match fround and fextend for 128/256-bit conversions
+ def : Pat<(v4f32 (fround (v4f64 VR256:$src))),
+ (VCVTPD2PSYrr VR256:$src)>;
+ def : Pat<(v4f32 (fround (loadv4f64 addr:$src))),
+ (VCVTPD2PSYrm addr:$src)>;
+
+ def : Pat<(v4f64 (fextend (v4f32 VR128:$src))),
+ (VCVTPS2PDYrr VR128:$src)>;
+ def : Pat<(v4f64 (fextend (loadv4f32 addr:$src))),
+ (VCVTPS2PDYrm addr:$src)>;
+}
//===----------------------------------------------------------------------===//
// SSE 1 & 2 - Compare Instructions
IIC_SSE_MOVNT>, VEX;
}
-def : Pat<(int_x86_avx_movnt_dq_256 addr:$dst, VR256:$src),
- (VMOVNTDQYmr addr:$dst, VR256:$src)>;
-def : Pat<(int_x86_avx_movnt_pd_256 addr:$dst, VR256:$src),
- (VMOVNTPDYmr addr:$dst, VR256:$src)>;
-def : Pat<(int_x86_avx_movnt_ps_256 addr:$dst, VR256:$src),
- (VMOVNTPSYmr addr:$dst, VR256:$src)>;
-
let AddedComplexity = 400 in { // Prefer non-temporal versions
def MOVNTPSmr : PSI<0x2B, MRMDestMem, (outs), (ins f128mem:$dst, VR128:$src),
"movntps\t{$src, $dst|$dst, $src}",
// Bitcast FR64 <-> GR64
//
let Predicates = [HasAVX] in
-def VMOV64toSDrm : S3SI<0x7E, MRMSrcMem, (outs FR64:$dst), (ins i64mem:$src),
+def VMOV64toSDrm : SSDI<0x7E, MRMSrcMem, (outs FR64:$dst), (ins i64mem:$src),
"vmovq\t{$src, $dst|$dst, $src}",
[(set FR64:$dst, (bitconvert (loadi64 addr:$src)))]>,
VEX;
[(store (i64 (bitconvert FR64:$src)), addr:$dst)],
IIC_SSE_MOVDQ>, VEX;
-def MOV64toSDrm : S3SI<0x7E, MRMSrcMem, (outs FR64:$dst), (ins i64mem:$src),
+def MOV64toSDrm : SSDI<0x7E, MRMSrcMem, (outs FR64:$dst), (ins i64mem:$src),
"movq\t{$src, $dst|$dst, $src}",
[(set FR64:$dst, (bitconvert (loadi64 addr:$src)))],
IIC_SSE_MOVDQ>;
def MOVQxrxr : I<0x7E, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
"movq\t{$src, $dst|$dst, $src}", [], IIC_SSE_MOVQ_RR>, XS;
-//===---------------------------------------------------------------------===//
-// SSE3 - Conversion Instructions
-//===---------------------------------------------------------------------===//
-
-// Convert Packed Double FP to Packed DW Integers
-let Predicates = [HasAVX] in {
-// The assembler can recognize rr 256-bit instructions by seeing a ymm
-// register, but the same isn't true when using memory operands instead.
-// Provide other assembly rr and rm forms to address this explicitly.
-def VCVTPD2DQrr : S3DI<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
- "vcvtpd2dq\t{$src, $dst|$dst, $src}", []>, VEX;
-def VCVTPD2DQXrYr : S3DI<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR256:$src),
- "vcvtpd2dq\t{$src, $dst|$dst, $src}", []>, VEX;
-
-// XMM only
-def VCVTPD2DQXrr : S3DI<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
- "vcvtpd2dqx\t{$src, $dst|$dst, $src}", []>, VEX;
-def VCVTPD2DQXrm : S3DI<0xE6, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
- "vcvtpd2dqx\t{$src, $dst|$dst, $src}", []>, VEX;
-
-// YMM only
-def VCVTPD2DQYrr : S3DI<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR256:$src),
- "vcvtpd2dqy\t{$src, $dst|$dst, $src}", []>, VEX;
-def VCVTPD2DQYrm : S3DI<0xE6, MRMSrcMem, (outs VR128:$dst), (ins f256mem:$src),
- "vcvtpd2dqy\t{$src, $dst|$dst, $src}", []>, VEX, VEX_L;
-}
-
-def CVTPD2DQrm : S3DI<0xE6, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
- "cvtpd2dq\t{$src, $dst|$dst, $src}", [],
- IIC_SSE_CVT_PD_RM>;
-def CVTPD2DQrr : S3DI<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
- "cvtpd2dq\t{$src, $dst|$dst, $src}", [],
- IIC_SSE_CVT_PD_RR>;
-
-def : Pat<(v4i32 (fp_to_sint (v4f64 VR256:$src))),
- (VCVTTPD2DQYrr VR256:$src)>;
-def : Pat<(v4i32 (fp_to_sint (memopv4f64 addr:$src))),
- (VCVTTPD2DQYrm addr:$src)>;
-
-// Convert Packed DW Integers to Packed Double FP
-let Predicates = [HasAVX] in {
-def VCVTDQ2PDrm : S3SI<0xE6, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
- "vcvtdq2pd\t{$src, $dst|$dst, $src}", []>, VEX;
-def VCVTDQ2PDrr : S3SI<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
- "vcvtdq2pd\t{$src, $dst|$dst, $src}", []>, VEX;
-def VCVTDQ2PDYrm : S3SI<0xE6, MRMSrcMem, (outs VR256:$dst), (ins f128mem:$src),
- "vcvtdq2pd\t{$src, $dst|$dst, $src}", []>, VEX;
-def VCVTDQ2PDYrr : S3SI<0xE6, MRMSrcReg, (outs VR256:$dst), (ins VR128:$src),
- "vcvtdq2pd\t{$src, $dst|$dst, $src}", []>, VEX;
-}
-
-def CVTDQ2PDrm : S3SI<0xE6, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
- "cvtdq2pd\t{$src, $dst|$dst, $src}", [],
- IIC_SSE_CVT_PD_RR>;
-def CVTDQ2PDrr : S3SI<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
- "cvtdq2pd\t{$src, $dst|$dst, $src}", [],
- IIC_SSE_CVT_PD_RM>;
-
-// AVX 256-bit register conversion intrinsics
-def : Pat<(int_x86_avx_cvtdq2_pd_256 VR128:$src),
- (VCVTDQ2PDYrr VR128:$src)>;
-def : Pat<(int_x86_avx_cvtdq2_pd_256 (bitconvert (memopv2i64 addr:$src))),
- (VCVTDQ2PDYrm addr:$src)>;
-
-def : Pat<(int_x86_avx_cvt_pd2dq_256 VR256:$src),
- (VCVTPD2DQYrr VR256:$src)>;
-def : Pat<(int_x86_avx_cvt_pd2dq_256 (memopv4f64 addr:$src)),
- (VCVTPD2DQYrm addr:$src)>;
-
-def : Pat<(v4f64 (sint_to_fp (v4i32 VR128:$src))),
- (VCVTDQ2PDYrr VR128:$src)>;
-def : Pat<(v4f64 (sint_to_fp (bc_v4i32 (memopv2i64 addr:$src)))),
- (VCVTDQ2PDYrm addr:$src)>;
-
//===---------------------------------------------------------------------===//
// SSE3 - Replicate Single FP - MOVSHDUP and MOVSLDUP
//===---------------------------------------------------------------------===//
let isCommutable = 0 in {
let ExeDomain = SSEPackedSingle in {
defm VBLENDPS : SS41I_binop_rmi_int<0x0C, "vblendps", int_x86_sse41_blendps,
- VR128, memopv4f32, i128mem, 0>, VEX_4V;
+ VR128, memopv4f32, f128mem, 0>, VEX_4V;
defm VBLENDPSY : SS41I_binop_rmi_int<0x0C, "vblendps",
- int_x86_avx_blend_ps_256, VR256, memopv8f32, i256mem, 0>, VEX_4V;
+ int_x86_avx_blend_ps_256, VR256, memopv8f32, f256mem, 0>, VEX_4V;
}
let ExeDomain = SSEPackedDouble in {
defm VBLENDPD : SS41I_binop_rmi_int<0x0D, "vblendpd", int_x86_sse41_blendpd,
- VR128, memopv2f64, i128mem, 0>, VEX_4V;
+ VR128, memopv2f64, f128mem, 0>, VEX_4V;
defm VBLENDPDY : SS41I_binop_rmi_int<0x0D, "vblendpd",
- int_x86_avx_blend_pd_256, VR256, memopv4f64, i256mem, 0>, VEX_4V;
+ int_x86_avx_blend_pd_256, VR256, memopv4f64, f256mem, 0>, VEX_4V;
}
defm VPBLENDW : SS41I_binop_rmi_int<0x0E, "vpblendw", int_x86_sse41_pblendw,
VR128, memopv2i64, i128mem, 0>, VEX_4V;
}
let ExeDomain = SSEPackedSingle in
defm VDPPS : SS41I_binop_rmi_int<0x40, "vdpps", int_x86_sse41_dpps,
- VR128, memopv4f32, i128mem, 0>, VEX_4V;
+ VR128, memopv4f32, f128mem, 0>, VEX_4V;
let ExeDomain = SSEPackedDouble in
defm VDPPD : SS41I_binop_rmi_int<0x41, "vdppd", int_x86_sse41_dppd,
- VR128, memopv2f64, i128mem, 0>, VEX_4V;
+ VR128, memopv2f64, 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;
let isCommutable = 0 in {
let ExeDomain = SSEPackedSingle in
defm BLENDPS : SS41I_binop_rmi_int<0x0C, "blendps", int_x86_sse41_blendps,
- VR128, memopv4f32, i128mem>;
+ VR128, memopv4f32, f128mem>;
let ExeDomain = SSEPackedDouble in
defm BLENDPD : SS41I_binop_rmi_int<0x0D, "blendpd", int_x86_sse41_blendpd,
- VR128, memopv2f64, i128mem>;
+ VR128, memopv2f64, f128mem>;
defm PBLENDW : SS41I_binop_rmi_int<0x0E, "pblendw", int_x86_sse41_pblendw,
VR128, memopv2i64, i128mem>;
defm MPSADBW : SS41I_binop_rmi_int<0x42, "mpsadbw", int_x86_sse41_mpsadbw,
}
let ExeDomain = SSEPackedSingle in
defm DPPS : SS41I_binop_rmi_int<0x40, "dpps", int_x86_sse41_dpps,
- VR128, memopv4f32, i128mem>;
+ VR128, memopv4f32, f128mem>;
let ExeDomain = SSEPackedDouble in
defm DPPD : SS41I_binop_rmi_int<0x41, "dppd", int_x86_sse41_dppd,
- VR128, memopv2f64, i128mem>;
+ VR128, memopv2f64, f128mem>;
}
/// 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, i128mem,
+defm VBLENDVPD : SS41I_quaternary_int_avx<0x4B, "vblendvpd", VR128, f128mem,
memopv2f64, int_x86_sse41_blendvpd>;
-defm VBLENDVPDY : SS41I_quaternary_int_avx<0x4B, "vblendvpd", VR256, i256mem,
+defm VBLENDVPDY : SS41I_quaternary_int_avx<0x4B, "vblendvpd", VR256, f256mem,
memopv4f64, int_x86_avx_blendv_pd_256>;
} // ExeDomain = SSEPackedDouble
let ExeDomain = SSEPackedSingle in {
-defm VBLENDVPS : SS41I_quaternary_int_avx<0x4A, "vblendvps", VR128, i128mem,
+defm VBLENDVPS : SS41I_quaternary_int_avx<0x4A, "vblendvps", VR128, f128mem,
memopv4f32, int_x86_sse41_blendvps>;
-defm VBLENDVPSY : SS41I_quaternary_int_avx<0x4A, "vblendvps", VR256, i256mem,
+defm VBLENDVPSY : SS41I_quaternary_int_avx<0x4A, "vblendvps", VR256, f256mem,
memopv8f32, int_x86_avx_blendv_ps_256>;
} // ExeDomain = SSEPackedSingle
defm VPBLENDVB : SS41I_quaternary_int_avx<0x4C, "vpblendvb", VR128, i128mem,
/// SS41I_ternary_int - SSE 4.1 ternary operator
let Uses = [XMM0], Constraints = "$src1 = $dst" in {
multiclass SS41I_ternary_int<bits<8> opc, string OpcodeStr, PatFrag mem_frag,
- Intrinsic IntId> {
+ X86MemOperand x86memop, Intrinsic IntId> {
def rr0 : SS48I<opc, MRMSrcReg, (outs VR128:$dst),
(ins VR128:$src1, VR128:$src2),
!strconcat(OpcodeStr,
OpSize;
def rm0 : SS48I<opc, MRMSrcMem, (outs VR128:$dst),
- (ins VR128:$src1, i128mem:$src2),
+ (ins VR128:$src1, x86memop:$src2),
!strconcat(OpcodeStr,
"\t{$src2, $dst|$dst, $src2}"),
[(set VR128:$dst,
}
let ExeDomain = SSEPackedDouble in
-defm BLENDVPD : SS41I_ternary_int<0x15, "blendvpd", memopv2f64,
+defm BLENDVPD : SS41I_ternary_int<0x15, "blendvpd", memopv2f64, f128mem,
int_x86_sse41_blendvpd>;
let ExeDomain = SSEPackedSingle in
-defm BLENDVPS : SS41I_ternary_int<0x14, "blendvps", memopv4f32,
+defm BLENDVPS : SS41I_ternary_int<0x14, "blendvps", memopv4f32, f128mem,
int_x86_sse41_blendvps>;
-defm PBLENDVB : SS41I_ternary_int<0x10, "pblendvb", memopv2i64,
+defm PBLENDVB : SS41I_ternary_int<0x10, "pblendvb", memopv2i64, i128mem,
int_x86_sse41_pblendvb>;
+// Aliases with the implicit xmm0 argument
+def : InstAlias<"blendvpd\t{%xmm0, $src2, $dst|$dst, $src2, %xmm0}",
+ (BLENDVPDrr0 VR128:$dst, VR128:$src2)>;
+def : InstAlias<"blendvpd\t{%xmm0, $src2, $dst|$dst, $src2, %xmm0}",
+ (BLENDVPDrm0 VR128:$dst, f128mem:$src2)>;
+def : InstAlias<"blendvps\t{%xmm0, $src2, $dst|$dst, $src2, %xmm0}",
+ (BLENDVPSrr0 VR128:$dst, VR128:$src2)>;
+def : InstAlias<"blendvps\t{%xmm0, $src2, $dst|$dst, $src2, %xmm0}",
+ (BLENDVPSrm0 VR128:$dst, f128mem:$src2)>;
+def : InstAlias<"pblendvb\t{%xmm0, $src2, $dst|$dst, $src2, %xmm0}",
+ (PBLENDVBrr0 VR128:$dst, VR128:$src2)>;
+def : InstAlias<"pblendvb\t{%xmm0, $src2, $dst|$dst, $src2, %xmm0}",
+ (PBLENDVBrm0 VR128:$dst, i128mem:$src2)>;
+
let Predicates = [HasSSE41] in {
def : Pat<(v16i8 (vselect (v16i8 XMM0), (v16i8 VR128:$src1),
(v16i8 VR128:$src2))),
OpSize;
//===----------------------------------------------------------------------===//
-// CLMUL Instructions
+// PCLMUL Instructions
//===----------------------------------------------------------------------===//
-// Carry-less Multiplication instructions
-let neverHasSideEffects = 1 in {
// AVX carry-less Multiplication instructions
-def VPCLMULQDQrr : AVXCLMULIi8<0x44, MRMSrcReg, (outs VR128:$dst),
+def VPCLMULQDQrr : AVXPCLMULIi8<0x44, MRMSrcReg, (outs VR128:$dst),
(ins VR128:$src1, VR128:$src2, i8imm:$src3),
"vpclmulqdq\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}",
- []>;
+ [(set VR128:$dst,
+ (int_x86_pclmulqdq VR128:$src1, VR128:$src2, imm:$src3))]>;
-let mayLoad = 1 in
-def VPCLMULQDQrm : AVXCLMULIi8<0x44, MRMSrcMem, (outs VR128:$dst),
+def VPCLMULQDQrm : AVXPCLMULIi8<0x44, MRMSrcMem, (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:$src3))]>;
+// Carry-less Multiplication instructions
let Constraints = "$src1 = $dst" in {
-def PCLMULQDQrr : CLMULIi8<0x44, MRMSrcReg, (outs VR128:$dst),
+def PCLMULQDQrr : PCLMULIi8<0x44, MRMSrcReg, (outs VR128:$dst),
(ins VR128:$src1, VR128:$src2, i8imm:$src3),
"pclmulqdq\t{$src3, $src2, $dst|$dst, $src2, $src3}",
- []>;
+ [(set VR128:$dst,
+ (int_x86_pclmulqdq VR128:$src1, VR128:$src2, imm:$src3))]>;
-let mayLoad = 1 in
-def PCLMULQDQrm : CLMULIi8<0x44, MRMSrcMem, (outs VR128:$dst),
+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:$src3))]>;
} // Constraints = "$src1 = $dst"
-} // neverHasSideEffects = 1
multiclass pclmul_alias<string asm, int immop> {
- def : InstAlias<!strconcat("pclmul", asm,
- "dq {$src, $dst|$dst, $src}"),
+ def : InstAlias<!strconcat("pclmul", asm, "dq {$src, $dst|$dst, $src}"),
(PCLMULQDQrr VR128:$dst, VR128:$src, immop)>;
- def : InstAlias<!strconcat("pclmul", asm,
- "dq {$src, $dst|$dst, $src}"),
+ def : InstAlias<!strconcat("pclmul", asm, "dq {$src, $dst|$dst, $src}"),
(PCLMULQDQrm VR128:$dst, i128mem:$src, immop)>;
- def : InstAlias<!strconcat("vpclmul", asm,
+ def : InstAlias<!strconcat("vpclmul", asm,
"dq {$src2, $src1, $dst|$dst, $src1, $src2}"),
(VPCLMULQDQrr VR128:$dst, VR128:$src1, VR128:$src2, immop)>;
- def : InstAlias<!strconcat("vpclmul", asm,
+ def : InstAlias<!strconcat("vpclmul", asm,
"dq {$src2, $src1, $dst|$dst, $src1, $src2}"),
(VPCLMULQDQrm VR128:$dst, VR128:$src1, i128mem:$src2, immop)>;
}
//===----------------------------------------------------------------------===//
let Predicates = [HasSSE4A] in {
+
+let Constraints = "$src = $dst" in {
+def EXTRQI : Ii8<0x78, MRM0r, (outs VR128:$dst),
+ (ins VR128:$src, i8imm:$len, i8imm:$idx),
+ "extrq\t{$idx, $len, $src|$src, $len, $idx}",
+ [(set VR128:$dst, (int_x86_sse4a_extrqi VR128:$src, imm:$len,
+ imm:$idx))]>, TB, OpSize;
+def EXTRQ : I<0x79, MRMSrcReg, (outs VR128:$dst),
+ (ins VR128:$src, VR128:$mask),
+ "extrq\t{$mask, $src|$src, $mask}",
+ [(set VR128:$dst, (int_x86_sse4a_extrq VR128:$src,
+ VR128:$mask))]>, TB, OpSize;
+
+def INSERTQI : Ii8<0x78, MRMSrcReg, (outs VR128:$dst),
+ (ins VR128:$src, VR128:$src2, i8imm:$len, i8imm:$idx),
+ "insertq\t{$idx, $len, $src2, $src|$src, $src2, $len, $idx}",
+ [(set VR128:$dst, (int_x86_sse4a_insertqi VR128:$src,
+ VR128:$src2, imm:$len, imm:$idx))]>, XD;
+def INSERTQ : I<0x79, MRMSrcReg, (outs VR128:$dst),
+ (ins VR128:$src, VR128:$mask),
+ "insertq\t{$mask, $src|$src, $mask}",
+ [(set VR128:$dst, (int_x86_sse4a_insertq VR128:$src,
+ VR128:$mask))]>, XD;
+}
+
def MOVNTSS : I<0x2B, MRMDestMem, (outs), (ins f32mem:$dst, VR128:$src),
"movntss\t{$src, $dst|$dst, $src}",
[(int_x86_sse4a_movnt_ss addr:$dst, VR128:$src)]>, XS;
int_x86_avx2_vbroadcast_ss_ps_256>;
}
let ExeDomain = SSEPackedDouble in
-def VBROADCASTSDrr : avx2_broadcast_reg<0x19, "vbroadcastsd", VR256,
- int_x86_avx2_vbroadcast_sd_pd_256>;
+def VBROADCASTSDYrr : avx2_broadcast_reg<0x19, "vbroadcastsd", VR256,
+ int_x86_avx2_vbroadcast_sd_pd_256>;
let Predicates = [HasAVX2] in
def VBROADCASTI128 : avx_broadcast<0x5A, "vbroadcasti128", VR256, i128mem,
// Half precision conversion instructions
//===----------------------------------------------------------------------===//
multiclass f16c_ph2ps<RegisterClass RC, X86MemOperand x86memop, Intrinsic Int> {
-let Predicates = [HasAVX, HasF16C] in {
def rr : I<0x13, MRMSrcReg, (outs RC:$dst), (ins VR128:$src),
"vcvtph2ps\t{$src, $dst|$dst, $src}",
[(set RC:$dst, (Int VR128:$src))]>,
def rm : I<0x13, MRMSrcMem, (outs RC:$dst), (ins x86memop:$src),
"vcvtph2ps\t{$src, $dst|$dst, $src}", []>, T8, OpSize, VEX;
}
-}
multiclass f16c_ps2ph<RegisterClass RC, X86MemOperand x86memop, Intrinsic Int> {
-let Predicates = [HasAVX, HasF16C] in {
def rr : Ii8<0x1D, MRMDestReg, (outs VR128:$dst),
(ins RC:$src1, i32i8imm:$src2),
"vcvtps2ph\t{$src2, $src1, $dst|$dst, $src1, $src2}",
[(set VR128:$dst, (Int RC:$src1, imm:$src2))]>,
TA, OpSize, VEX;
- let neverHasSideEffects = 1, mayLoad = 1 in
- def mr : Ii8<0x1D, MRMDestMem, (outs x86memop:$dst),
- (ins RC:$src1, i32i8imm:$src2),
+ let neverHasSideEffects = 1, mayStore = 1 in
+ def mr : Ii8<0x1D, MRMDestMem, (outs),
+ (ins x86memop:$dst, RC:$src1, i32i8imm:$src2),
"vcvtps2ph\t{$src2, $src1, $dst|$dst, $src1, $src2}", []>,
TA, OpSize, VEX;
}
-}
-defm VCVTPH2PS : f16c_ph2ps<VR128, f64mem, int_x86_vcvtph2ps_128>;
-defm VCVTPH2PSY : f16c_ph2ps<VR256, f128mem, int_x86_vcvtph2ps_256>;
-defm VCVTPS2PH : f16c_ps2ph<VR128, f64mem, int_x86_vcvtps2ph_128>;
-defm VCVTPS2PHY : f16c_ps2ph<VR256, f128mem, int_x86_vcvtps2ph_256>;
+let Predicates = [HasAVX, HasF16C] in {
+ defm VCVTPH2PS : f16c_ph2ps<VR128, f64mem, int_x86_vcvtph2ps_128>;
+ defm VCVTPH2PSY : f16c_ph2ps<VR256, f128mem, int_x86_vcvtph2ps_256>;
+ defm VCVTPS2PH : f16c_ps2ph<VR128, f64mem, int_x86_vcvtps2ph_128>;
+ defm VCVTPS2PHY : f16c_ps2ph<VR256, f128mem, int_x86_vcvtps2ph_256>;
+}
//===----------------------------------------------------------------------===//
// AVX2 Instructions
def : Pat<(v4i64 (X86VBroadcast (loadi64 addr:$src))),
(VPBROADCASTQYrm addr:$src)>;
+ def : Pat<(v16i8 (X86VBroadcast (v16i8 VR128:$src))),
+ (VPBROADCASTBrr VR128:$src)>;
+ def : Pat<(v32i8 (X86VBroadcast (v16i8 VR128:$src))),
+ (VPBROADCASTBYrr VR128:$src)>;
+ def : Pat<(v8i16 (X86VBroadcast (v8i16 VR128:$src))),
+ (VPBROADCASTWrr VR128:$src)>;
+ def : Pat<(v16i16 (X86VBroadcast (v8i16 VR128:$src))),
+ (VPBROADCASTWYrr VR128:$src)>;
+ def : Pat<(v4i32 (X86VBroadcast (v4i32 VR128:$src))),
+ (VPBROADCASTDrr VR128:$src)>;
+ def : Pat<(v8i32 (X86VBroadcast (v4i32 VR128:$src))),
+ (VPBROADCASTDYrr VR128:$src)>;
+ def : Pat<(v2i64 (X86VBroadcast (v2i64 VR128:$src))),
+ (VPBROADCASTQrr VR128:$src)>;
+ def : Pat<(v4i64 (X86VBroadcast (v2i64 VR128:$src))),
+ (VPBROADCASTQYrr VR128:$src)>;
+ def : Pat<(v4f32 (X86VBroadcast (v4f32 VR128:$src))),
+ (VBROADCASTSSrr VR128:$src)>;
+ def : Pat<(v8f32 (X86VBroadcast (v4f32 VR128:$src))),
+ (VBROADCASTSSYrr VR128:$src)>;
+ def : Pat<(v2f64 (X86VBroadcast (v2f64 VR128:$src))),
+ (VPBROADCASTQrr VR128:$src)>;
+ def : Pat<(v4f64 (X86VBroadcast (v2f64 VR128:$src))),
+ (VBROADCASTSDYrr VR128:$src)>;
+
// Provide fallback in case the load node that is used in the patterns above
// is used by additional users, which prevents the pattern selection.
let AddedComplexity = 20 in {
(VBROADCASTSSYrr
(INSERT_SUBREG (v4f32 (IMPLICIT_DEF)), FR32:$src, sub_ss))>;
def : Pat<(v4f64 (X86VBroadcast FR64:$src)),
- (VBROADCASTSDrr
+ (VBROADCASTSDYrr
(INSERT_SUBREG (v2f64 (IMPLICIT_DEF)), FR64:$src, sub_sd))>;
def : Pat<(v4i32 (X86VBroadcast GR32:$src)),
(VBROADCASTSSYrr
(INSERT_SUBREG (v4i32 (IMPLICIT_DEF)), GR32:$src, sub_ss))>;
def : Pat<(v4i64 (X86VBroadcast GR64:$src)),
- (VBROADCASTSDrr
+ (VBROADCASTSDYrr
(INSERT_SUBREG (v2i64 (IMPLICIT_DEF)), GR64:$src, sub_sd))>;
}
}
defm VPSRLVD : avx2_var_shift<0x45, "vpsrlvd", srl, v4i32, v8i32>;
defm VPSRLVQ : avx2_var_shift<0x45, "vpsrlvq", srl, v2i64, v4i64>, VEX_W;
defm VPSRAVD : avx2_var_shift<0x46, "vpsravd", sra, v4i32, v8i32>;
+
+//===----------------------------------------------------------------------===//
+// VGATHER - GATHER Operations
+multiclass avx2_gather<bits<8> opc, string OpcodeStr,
+ RegisterClass RC256, X86MemOperand memop256,
+ Intrinsic IntGather128, Intrinsic IntGather256> {
+ def rm : AVX28I<opc, MRMSrcMem, (outs VR128:$dst),
+ (ins VR128:$src1, v128mem:$src2, VR128:$mask),
+ !strconcat(OpcodeStr,
+ "\t{$src1, $src2, $mask|$mask, $src2, $src1}"),
+ []>, VEX_4VOp3;
+ def Yrm : AVX28I<opc, MRMSrcMem, (outs RC256:$dst),
+ (ins RC256:$src1, memop256:$src2, RC256:$mask),
+ !strconcat(OpcodeStr,
+ "\t{$src1, $src2, $mask|$mask, $src2, $src1}"),
+ []>, VEX_4VOp3, VEX_L;
+}
+
+let Constraints = "$src1 = $dst" in {
+ defm VGATHERDPD : avx2_gather<0x92, "vgatherdpd",
+ VR256, v128mem,
+ int_x86_avx2_gather_d_pd,
+ int_x86_avx2_gather_d_pd_256>, VEX_W;
+ defm VGATHERQPD : avx2_gather<0x93, "vgatherqpd",
+ VR256, v256mem,
+ int_x86_avx2_gather_q_pd,
+ int_x86_avx2_gather_q_pd_256>, VEX_W;
+ defm VGATHERDPS : avx2_gather<0x92, "vgatherdps",
+ VR256, v256mem,
+ int_x86_avx2_gather_d_ps,
+ int_x86_avx2_gather_d_ps_256>;
+ defm VGATHERQPS : avx2_gather<0x93, "vgatherqps",
+ VR128, v256mem,
+ int_x86_avx2_gather_q_ps,
+ int_x86_avx2_gather_q_ps_256>;
+ defm VPGATHERDQ : avx2_gather<0x90, "vpgatherdq",
+ VR256, v128mem,
+ int_x86_avx2_gather_d_q,
+ int_x86_avx2_gather_d_q_256>, VEX_W;
+ defm VPGATHERQQ : avx2_gather<0x91, "vpgatherqq",
+ VR256, v256mem,
+ int_x86_avx2_gather_q_q,
+ int_x86_avx2_gather_q_q_256>, VEX_W;
+ defm VPGATHERDD : avx2_gather<0x90, "vpgatherdd",
+ VR256, v256mem,
+ int_x86_avx2_gather_d_d,
+ int_x86_avx2_gather_d_d_256>;
+ defm VPGATHERQD : avx2_gather<0x91, "vpgatherqd",
+ VR128, v256mem,
+ int_x86_avx2_gather_q_d,
+ int_x86_avx2_gather_q_d_256>;
+}
projects / oota-llvm.git / blobdiff