/// sse12_fp_scalar - SSE 1 & 2 scalar instructions class
multiclass sse12_fp_scalar<bits<8> opc, string OpcodeStr, SDNode OpNode,
RegisterClass RC, X86MemOperand x86memop,
- OpndItins itins,
- bit Is2Addr = 1> {
+ Domain d, OpndItins itins, bit Is2Addr = 1> {
let isCommutable = 1 in {
def rr : SI<opc, MRMSrcReg, (outs RC:$dst), (ins RC:$src1, RC:$src2),
!if(Is2Addr,
!strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"),
!strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}")),
- [(set RC:$dst, (OpNode RC:$src1, RC:$src2))], itins.rr>,
+ [(set RC:$dst, (OpNode RC:$src1, RC:$src2))], itins.rr, d>,
Sched<[itins.Sched]>;
}
def rm : SI<opc, MRMSrcMem, (outs RC:$dst), (ins RC:$src1, x86memop:$src2),
!if(Is2Addr,
!strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"),
!strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}")),
- [(set RC:$dst, (OpNode RC:$src1, (load addr:$src2)))], itins.rm>,
+ [(set RC:$dst, (OpNode RC:$src1, (load addr:$src2)))], itins.rm, d>,
Sched<[itins.Sched.Folded, ReadAfterLd]>;
}
multiclass sse12_fp_scalar_int<bits<8> opc, string OpcodeStr, RegisterClass RC,
string asm, string SSEVer, string FPSizeStr,
Operand memopr, ComplexPattern mem_cpat,
- OpndItins itins,
- bit Is2Addr = 1> {
+ Domain d, OpndItins itins, bit Is2Addr = 1> {
let isCodeGenOnly = 1 in {
- def rr_Int : SI<opc, MRMSrcReg, (outs RC:$dst), (ins RC:$src1, RC:$src2),
+ def rr_Int : SI_Int<opc, MRMSrcReg, (outs RC:$dst), (ins RC:$src1, RC:$src2),
!if(Is2Addr,
!strconcat(asm, "\t{$src2, $dst|$dst, $src2}"),
!strconcat(asm, "\t{$src2, $src1, $dst|$dst, $src1, $src2}")),
[(set RC:$dst, (!cast<Intrinsic>(
!strconcat("int_x86_sse", SSEVer, "_", OpcodeStr, FPSizeStr))
- RC:$src1, RC:$src2))], itins.rr>,
+ RC:$src1, RC:$src2))], itins.rr, d>,
Sched<[itins.Sched]>;
- def rm_Int : SI<opc, MRMSrcMem, (outs RC:$dst), (ins RC:$src1, memopr:$src2),
+ def rm_Int : SI_Int<opc, MRMSrcMem, (outs RC:$dst), (ins RC:$src1, memopr:$src2),
!if(Is2Addr,
!strconcat(asm, "\t{$src2, $dst|$dst, $src2}"),
!strconcat(asm, "\t{$src2, $src1, $dst|$dst, $src1, $src2}")),
[(set RC:$dst, (!cast<Intrinsic>(!strconcat("int_x86_sse",
SSEVer, "_", OpcodeStr, FPSizeStr))
- RC:$src1, mem_cpat:$src2))], itins.rm>,
+ RC:$src1, mem_cpat:$src2))], itins.rm, d>,
Sched<[itins.Sched.Folded, ReadAfterLd]>;
}
}
// Represent the same patterns above but in the form they appear for
// 256-bit types
- def : Pat<(v8i32 (X86vzmovl (insert_subvector undef,
- (v4i32 (scalar_to_vector (loadi32 addr:$src))), (iPTR 0)))),
- (SUBREG_TO_REG (i32 0), (VMOVSSrm addr:$src), sub_xmm)>;
def : Pat<(v8f32 (X86vzmovl (insert_subvector undef,
(v4f32 (scalar_to_vector (loadf32 addr:$src))), (iPTR 0)))),
(SUBREG_TO_REG (i32 0), (VMOVSSrm addr:$src), sub_xmm)>;
(v2f64 (scalar_to_vector (loadf64 addr:$src))), (iPTR 0)))),
(SUBREG_TO_REG (i32 0), (VMOVSDrm addr:$src), sub_xmm)>;
}
- def : Pat<(v4i64 (X86vzmovl (insert_subvector undef,
- (v2i64 (scalar_to_vector (loadi64 addr:$src))), (iPTR 0)))),
- (SUBREG_TO_REG (i64 0), (VMOVSDrm addr:$src), sub_xmm)>;
// Extract and store.
def : Pat<(store (f32 (vector_extract (v4f32 VR128:$src), (iPTR 0))),
(MOVSDrr VR128:$src1, (COPY_TO_REGCLASS VR128:$src2, FR64))>;
// FIXME: Instead of a X86Movlps there should be a X86Movsd here, the problem
- // is during lowering, where it's not possible to recognize the fold cause
+ // is during lowering, where it's not possible to recognize the fold because
// it has two uses through a bitcast. One use disappears at isel time and the
// fold opportunity reappears.
def : Pat<(v2f64 (X86Movlpd VR128:$src1, VR128:$src2)),
"vcvtsd2ss\t{$src2, $src1, $dst|$dst, $src1, $src2}",
[(set VR128:$dst,
(int_x86_sse2_cvtsd2ss VR128:$src1, VR128:$src2))],
- IIC_SSE_CVT_Scalar_RR>, XD, VEX_4V, Requires<[UseAVX]>,
+ IIC_SSE_CVT_Scalar_RR>, XD, VEX_4V, Requires<[HasAVX]>,
Sched<[WriteCvtF2F]>;
def Int_VCVTSD2SSrm: I<0x5A, MRMSrcReg,
(outs VR128:$dst), (ins VR128:$src1, sdmem:$src2),
"vcvtsd2ss\t{$src2, $src1, $dst|$dst, $src1, $src2}",
[(set VR128:$dst, (int_x86_sse2_cvtsd2ss
VR128:$src1, sse_load_f64:$src2))],
- IIC_SSE_CVT_Scalar_RM>, XD, VEX_4V, Requires<[UseAVX]>,
+ IIC_SSE_CVT_Scalar_RM>, XD, VEX_4V, Requires<[HasAVX]>,
Sched<[WriteCvtF2FLd, ReadAfterLd]>;
let Constraints = "$src1 = $dst" in {
"vcvtss2sd\t{$src2, $src1, $dst|$dst, $src1, $src2}",
[(set VR128:$dst,
(int_x86_sse2_cvtss2sd VR128:$src1, VR128:$src2))],
- IIC_SSE_CVT_Scalar_RR>, XS, VEX_4V, Requires<[UseAVX]>,
+ IIC_SSE_CVT_Scalar_RR>, XS, VEX_4V, Requires<[HasAVX]>,
Sched<[WriteCvtF2F]>;
def Int_VCVTSS2SDrm: I<0x5A, MRMSrcMem,
(outs VR128:$dst), (ins VR128:$src1, ssmem:$src2),
"vcvtss2sd\t{$src2, $src1, $dst|$dst, $src1, $src2}",
[(set VR128:$dst,
(int_x86_sse2_cvtss2sd VR128:$src1, sse_load_f32:$src2))],
- IIC_SSE_CVT_Scalar_RM>, XS, VEX_4V, Requires<[UseAVX]>,
+ IIC_SSE_CVT_Scalar_RM>, XS, VEX_4V, Requires<[HasAVX]>,
Sched<[WriteCvtF2FLd, ReadAfterLd]>;
let Constraints = "$src1 = $dst" in { // SSE2 instructions with XS prefix
def Int_CVTSS2SDrr: I<0x5A, MRMSrcReg,
IIC_SSE_CVT_PS_RM>, Sched<[WriteCvtF2ILd]>;
let Predicates = [HasAVX] in {
- def : Pat<(v4f32 (sint_to_fp (v4i32 VR128:$src))),
+ def : Pat<(int_x86_sse2_cvtdq2ps VR128:$src),
(VCVTDQ2PSrr VR128:$src)>;
- def : Pat<(v4f32 (sint_to_fp (bc_v4i32 (loadv2i64 addr:$src)))),
+ def : Pat<(int_x86_sse2_cvtdq2ps (bc_v4i32 (loadv2i64 addr:$src))),
(VCVTDQ2PSrm addr:$src)>;
+}
- def : Pat<(int_x86_sse2_cvtdq2ps VR128:$src),
+let Predicates = [HasAVX, NoVLX] in {
+ def : Pat<(v4f32 (sint_to_fp (v4i32 VR128:$src))),
(VCVTDQ2PSrr VR128:$src)>;
- def : Pat<(int_x86_sse2_cvtdq2ps (bc_v4i32 (loadv2i64 addr:$src))),
+ def : Pat<(v4f32 (sint_to_fp (bc_v4i32 (loadv2i64 addr:$src)))),
(VCVTDQ2PSrm addr:$src)>;
def : Pat<(v4i32 (fp_to_sint (v4f32 VR128:$src))),
def : InstAlias<"vcvttpd2dq\t{$src, $dst|$dst, $src}",
(VCVTTPD2DQYrr VR128:$dst, VR256:$src), 0>;
-let Predicates = [HasAVX] in {
+let Predicates = [HasAVX, NoVLX] in {
def : Pat<(v4i32 (fp_to_sint (v4f64 VR256:$src))),
(VCVTTPD2DQYrr VR256:$src)>;
def : Pat<(v4i32 (fp_to_sint (loadv4f64 addr:$src))),
[(set VR128:$dst, (int_x86_sse2_cvtdq2pd VR128:$src))],
IIC_SSE_CVT_PD_RM>, Sched<[WriteCvtI2F]>;
-// AVX 256-bit register conversion intrinsics
+// AVX register conversion intrinsics
let Predicates = [HasAVX] in {
+ def : Pat<(v2f64 (X86cvtdq2pd (v4i32 VR128:$src))),
+ (VCVTDQ2PDrr VR128:$src)>;
+ def : Pat<(v2f64 (X86cvtdq2pd (bc_v4i32 (loadv2i64 addr:$src)))),
+ (VCVTDQ2PDrm addr:$src)>;
+
def : Pat<(v4f64 (sint_to_fp (v4i32 VR128:$src))),
(VCVTDQ2PDYrr VR128:$src)>;
def : Pat<(v4f64 (sint_to_fp (bc_v4i32 (loadv2i64 addr:$src)))),
(VCVTDQ2PDYrm addr:$src)>;
} // Predicates = [HasAVX]
+// SSE2 register conversion intrinsics
+let Predicates = [HasSSE2] in {
+ def : Pat<(v2f64 (X86cvtdq2pd (v4i32 VR128:$src))),
+ (CVTDQ2PDrr VR128:$src)>;
+ def : Pat<(v2f64 (X86cvtdq2pd (bc_v4i32 (loadv2i64 addr:$src)))),
+ (CVTDQ2PDrm addr:$src)>;
+} // Predicates = [HasSSE2]
+
// Convert packed double to packed single
// The assembler can recognize rr 256-bit instructions by seeing a ymm
// register, but the same isn't true when using memory operands instead.
(VCVTDQ2PSYrr VR256:$src)>;
def : Pat<(int_x86_avx_cvtdq2_ps_256 (bitconvert (loadv4i64 addr:$src))),
(VCVTDQ2PSYrm addr:$src)>;
+}
+let Predicates = [HasAVX, NoVLX] in {
// Match fround and fextend for 128/256-bit conversions
def : Pat<(v4f32 (X86vfpround (v2f64 VR128:$src))),
(VCVTPD2PSrr VR128:$src)>;
defm VUCOMISD : sse12_ord_cmp<0x2E, FR64, X86cmp, f64, f64mem, loadf64,
"ucomisd">, PD, VEX, VEX_LIG;
let Pattern = []<dag> in {
- defm VCOMISS : sse12_ord_cmp<0x2F, VR128, undef, v4f32, f128mem, load,
+ defm VCOMISS : sse12_ord_cmp<0x2F, FR32, undef, f32, f32mem, loadf32,
"comiss">, PS, VEX, VEX_LIG;
- defm VCOMISD : sse12_ord_cmp<0x2F, VR128, undef, v2f64, f128mem, load,
+ defm VCOMISD : sse12_ord_cmp<0x2F, FR64, undef, f64, f64mem, loadf64,
"comisd">, PD, VEX, VEX_LIG;
}
"ucomisd">, PD;
let Pattern = []<dag> in {
- defm COMISS : sse12_ord_cmp<0x2F, VR128, undef, v4f32, f128mem, load,
+ defm COMISS : sse12_ord_cmp<0x2F, FR32, undef, f32, f32mem, loadf32,
"comiss">, PS;
- defm COMISD : sse12_ord_cmp<0x2F, VR128, undef, v2f64, f128mem, load,
+ defm COMISD : sse12_ord_cmp<0x2F, FR64, undef, f64, f64mem, loadf64,
"comisd">, PD;
}
Sched<[WriteFShuffleLd, ReadAfterLd]>;
}
+let Predicates = [HasAVX, NoVLX] in {
defm VUNPCKHPS: sse12_unpack_interleave<0x15, X86Unpckh, v4f32, loadv4f32,
VR128, f128mem, "unpckhps\t{$src2, $src1, $dst|$dst, $src1, $src2}",
SSEPackedSingle>, PS, VEX_4V;
defm VUNPCKLPDY: sse12_unpack_interleave<0x14, X86Unpckl, v4f64, loadv4f64,
VR256, f256mem, "unpcklpd\t{$src2, $src1, $dst|$dst, $src1, $src2}",
SSEPackedDouble>, PD, VEX_4V, VEX_L;
-
+}// Predicates = [HasAVX, NoVLX]
let Constraints = "$src1 = $dst" in {
defm UNPCKHPS: sse12_unpack_interleave<0x15, X86Unpckh, v4f32, memopv4f32,
VR128, f128mem, "unpckhps\t{$src2, $dst|$dst, $src2}",
multiclass PDI_binop_all<bits<8> opc, string OpcodeStr, SDNode Opcode,
ValueType OpVT128, ValueType OpVT256,
- OpndItins itins, bit IsCommutable = 0> {
-let Predicates = [HasAVX, NoVLX] in
+ OpndItins itins, bit IsCommutable = 0, Predicate prd> {
+let Predicates = [HasAVX, prd] in
defm V#NAME : PDI_binop_rm<opc, !strconcat("v", OpcodeStr), Opcode, OpVT128,
VR128, loadv2i64, i128mem, itins, IsCommutable, 0>, VEX_4V;
defm NAME : PDI_binop_rm<opc, OpcodeStr, Opcode, OpVT128, VR128,
memopv2i64, i128mem, itins, IsCommutable, 1>;
-let Predicates = [HasAVX2, NoVLX] in
+let Predicates = [HasAVX2, prd] in
defm V#NAME#Y : PDI_binop_rm<opc, !strconcat("v", OpcodeStr), Opcode,
OpVT256, VR256, loadv4i64, i256mem, itins,
IsCommutable, 0>, VEX_4V, VEX_L;
// These are ordered here for pattern ordering requirements with the fp versions
defm PAND : PDI_binop_all<0xDB, "pand", and, v2i64, v4i64,
- SSE_VEC_BIT_ITINS_P, 1>;
+ SSE_VEC_BIT_ITINS_P, 1, NoVLX>;
defm POR : PDI_binop_all<0xEB, "por", or, v2i64, v4i64,
- SSE_VEC_BIT_ITINS_P, 1>;
+ SSE_VEC_BIT_ITINS_P, 1, NoVLX>;
defm PXOR : PDI_binop_all<0xEF, "pxor", xor, v2i64, v4i64,
- SSE_VEC_BIT_ITINS_P, 1>;
+ SSE_VEC_BIT_ITINS_P, 1, NoVLX>;
defm PANDN : PDI_binop_all<0xDF, "pandn", X86andnp, v2i64, v4i64,
- SSE_VEC_BIT_ITINS_P, 0>;
+ SSE_VEC_BIT_ITINS_P, 0, NoVLX>;
//===----------------------------------------------------------------------===//
// SSE 1 & 2 - Logical Instructions
defm V#NAME#PD : sse12_fp_packed<opc, !strconcat(OpcodeStr, "pd"), OpNode,
VR128, v2f64, f128mem, loadv2f64, SSEPackedDouble, itins, 0>,
PD, VEX_4V;
+
+ defm V#NAME#PSY : sse12_fp_packed<opc, !strconcat(OpcodeStr, "ps"), OpNode,
+ VR256, v8f32, f256mem, loadv8f32, SSEPackedSingle, itins, 0>,
+ PS, VEX_4V, VEX_L;
+
+ defm V#NAME#PDY : sse12_fp_packed<opc, !strconcat(OpcodeStr, "pd"), OpNode,
+ VR256, v4f64, f256mem, loadv4f64, SSEPackedDouble, itins, 0>,
+ PD, VEX_4V, VEX_L;
}
let Constraints = "$src1 = $dst" in {
multiclass basic_sse12_fp_binop_s<bits<8> opc, string OpcodeStr, SDNode OpNode,
SizeItins itins> {
defm V#NAME#SS : sse12_fp_scalar<opc, !strconcat(OpcodeStr, "ss"),
- OpNode, FR32, f32mem, itins.s, 0>, XS, VEX_4V, VEX_LIG;
+ OpNode, FR32, f32mem, SSEPackedSingle, itins.s, 0>,
+ XS, VEX_4V, VEX_LIG;
defm V#NAME#SD : sse12_fp_scalar<opc, !strconcat(OpcodeStr, "sd"),
- OpNode, FR64, f64mem, itins.d, 0>, XD, VEX_4V, VEX_LIG;
+ OpNode, FR64, f64mem, SSEPackedDouble, itins.d, 0>,
+ XD, VEX_4V, VEX_LIG;
let Constraints = "$src1 = $dst" in {
defm SS : sse12_fp_scalar<opc, !strconcat(OpcodeStr, "ss"),
- OpNode, FR32, f32mem, itins.s>, XS;
+ OpNode, FR32, f32mem, SSEPackedSingle,
+ itins.s>, XS;
defm SD : sse12_fp_scalar<opc, !strconcat(OpcodeStr, "sd"),
- OpNode, FR64, f64mem, itins.d>, XD;
+ OpNode, FR64, f64mem, SSEPackedDouble,
+ itins.d>, XD;
}
}
SizeItins itins> {
defm V#NAME#SS : sse12_fp_scalar_int<opc, OpcodeStr, VR128,
!strconcat(OpcodeStr, "ss"), "", "_ss", ssmem, sse_load_f32,
- itins.s, 0>, XS, VEX_4V, VEX_LIG;
+ SSEPackedSingle, itins.s, 0>, XS, VEX_4V, VEX_LIG;
defm V#NAME#SD : sse12_fp_scalar_int<opc, OpcodeStr, VR128,
!strconcat(OpcodeStr, "sd"), "2", "_sd", sdmem, sse_load_f64,
- itins.d, 0>, XD, VEX_4V, VEX_LIG;
+ SSEPackedDouble, itins.d, 0>, XD, VEX_4V, VEX_LIG;
let Constraints = "$src1 = $dst" in {
defm SS : sse12_fp_scalar_int<opc, OpcodeStr, VR128,
!strconcat(OpcodeStr, "ss"), "", "_ss", ssmem, sse_load_f32,
- itins.s>, XS;
+ SSEPackedSingle, itins.s>, XS;
defm SD : sse12_fp_scalar_int<opc, OpcodeStr, VR128,
!strconcat(OpcodeStr, "sd"), "2", "_sd", sdmem, sse_load_f64,
- itins.d>, XD;
+ SSEPackedDouble, itins.d>, XD;
}
}
(Op (v4f32 VR128:$dst), (v4f32 VR128:$src)))),
(!cast<I>(OpcPrefix#SSrr_Int) v4f32:$dst, v4f32:$src)>;
}
-
- // With SSE 4.1, insertps/blendi are preferred to movsd, so match those too.
- let Predicates = [UseSSE41] in {
- // extracted scalar math op with insert via insertps
- def : Pat<(v4f32 (X86insertps (v4f32 VR128:$dst), (v4f32 (scalar_to_vector
- (Op (f32 (vector_extract (v4f32 VR128:$dst), (iPTR 0))),
- FR32:$src))), (iPTR 0))),
- (!cast<I>(OpcPrefix#SSrr_Int) v4f32:$dst,
- (COPY_TO_REGCLASS FR32:$src, VR128))>;
+ // With SSE 4.1, blendi is preferred to movsd, so match that too.
+ let Predicates = [UseSSE41] in {
// extracted scalar math op with insert via blend
def : Pat<(v4f32 (X86Blendi (v4f32 VR128:$dst), (v4f32 (scalar_to_vector
(Op (f32 (vector_extract (v4f32 VR128:$dst), (iPTR 0))),
// Repeat everything for AVX, except for the movss + scalar combo...
// because that one shouldn't occur with AVX codegen?
let Predicates = [HasAVX] in {
- // extracted scalar math op with insert via insertps
- def : Pat<(v4f32 (X86insertps (v4f32 VR128:$dst), (v4f32 (scalar_to_vector
- (Op (f32 (vector_extract (v4f32 VR128:$dst), (iPTR 0))),
- FR32:$src))), (iPTR 0))),
- (!cast<I>("V"#OpcPrefix#SSrr_Int) v4f32:$dst,
- (COPY_TO_REGCLASS FR32:$src, VR128))>;
-
// extracted scalar math op with insert via blend
def : Pat<(v4f32 (X86Blendi (v4f32 VR128:$dst), (v4f32 (scalar_to_vector
(Op (f32 (vector_extract (v4f32 VR128:$dst), (iPTR 0))),
FR64:$src))), (i8 1))),
(!cast<I>(OpcPrefix#SDrr_Int) v2f64:$dst,
(COPY_TO_REGCLASS FR64:$src, VR128))>;
-
+
// vector math op with insert via blend
def : Pat<(v2f64 (X86Blendi (v2f64 VR128:$dst),
(Op (v2f64 VR128:$dst), (v2f64 VR128:$src)), (i8 1))),
ValueType vt, ValueType ScalarVT,
X86MemOperand x86memop, Operand vec_memop,
ComplexPattern mem_cpat, Intrinsic Intr,
- SDNode OpNode, OpndItins itins, Predicate target,
- string Suffix> {
+ SDNode OpNode, Domain d, OpndItins itins,
+ Predicate target, string Suffix> {
let hasSideEffects = 0 in {
def r : I<opc, MRMSrcReg, (outs RC:$dst), (ins RC:$src1),
!strconcat(OpcodeStr, "\t{$src1, $dst|$dst, $src1}"),
- [(set RC:$dst, (OpNode RC:$src1))], itins.rr>, Sched<[itins.Sched]>,
+ [(set RC:$dst, (OpNode RC:$src1))], itins.rr, d>, Sched<[itins.Sched]>,
Requires<[target]>;
let mayLoad = 1 in
def m : I<opc, MRMSrcMem, (outs RC:$dst), (ins x86memop:$src1),
!strconcat(OpcodeStr, "\t{$src1, $dst|$dst, $src1}"),
- [(set RC:$dst, (OpNode (load addr:$src1)))], itins.rm>,
+ [(set RC:$dst, (OpNode (load addr:$src1)))], itins.rm, d>,
Sched<[itins.Sched.Folded, ReadAfterLd]>,
Requires<[target, OptForSize]>;
// because the high elements of the destination are unchanged in SSE.
def : Pat<(Intr VR128:$src),
(!cast<Instruction>(NAME#Suffix##r_Int) VR128:$src, VR128:$src)>;
- def : Pat<(Intr (load addr:$src)),
+ def : Pat<(Intr (load addr:$src)),
(vt (COPY_TO_REGCLASS(!cast<Instruction>(NAME#Suffix##m)
addr:$src), VR128))>;
- def : Pat<(Intr mem_cpat:$src),
+ def : Pat<(Intr mem_cpat:$src),
(!cast<Instruction>(NAME#Suffix##m_Int)
(vt (IMPLICIT_DEF)), mem_cpat:$src)>;
}
ValueType vt, ValueType ScalarVT,
X86MemOperand x86memop, Operand vec_memop,
ComplexPattern mem_cpat,
- Intrinsic Intr, SDNode OpNode, OpndItins itins,
- Predicate target, string Suffix> {
+ Intrinsic Intr, SDNode OpNode, Domain d,
+ OpndItins itins, string Suffix> {
let hasSideEffects = 0 in {
def r : I<opc, MRMSrcReg, (outs RC:$dst), (ins RC:$src1, RC:$src2),
!strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
- [], itins.rr>, Sched<[itins.Sched]>;
- let mayLoad = 1 in
+ [], itins.rr, d>, Sched<[itins.Sched]>;
+ let mayLoad = 1 in
def m : I<opc, MRMSrcMem, (outs RC:$dst), (ins RC:$src1, x86memop:$src2),
!strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
- [], itins.rm>, Sched<[itins.Sched.Folded, ReadAfterLd]>;
+ [], itins.rm, d>, Sched<[itins.Sched.Folded, ReadAfterLd]>;
let isCodeGenOnly = 1 in {
- // todo: uncomment when all r_Int forms will be added to X86InstrInfo.cpp
- //def r_Int : I<opc, MRMSrcReg, (outs VR128:$dst),
- // (ins VR128:$src1, VR128:$src2),
- // !strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
- // []>, Sched<[itins.Sched.Folded]>;
+ def r_Int : I<opc, MRMSrcReg, (outs VR128:$dst),
+ (ins VR128:$src1, VR128:$src2),
+ !strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
+ []>, Sched<[itins.Sched.Folded]>;
let mayLoad = 1 in
- def m_Int : I<opc, MRMSrcMem, (outs VR128:$dst),
+ def m_Int : I<opc, MRMSrcMem, (outs VR128:$dst),
(ins VR128:$src1, vec_memop:$src2),
!strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
- []>, Sched<[itins.Sched.Folded, ReadAfterLd]>;
+ []>, Sched<[itins.Sched.Folded, ReadAfterLd]>;
}
}
- let Predicates = [target] in {
+ let Predicates = [UseAVX] in {
def : Pat<(OpNode RC:$src), (!cast<Instruction>("V"#NAME#Suffix##r)
(ScalarVT (IMPLICIT_DEF)), RC:$src)>;
- def : Pat<(vt (OpNode mem_cpat:$src)),
+ def : Pat<(vt (OpNode mem_cpat:$src)),
(!cast<Instruction>("V"#NAME#Suffix##m_Int) (vt (IMPLICIT_DEF)),
mem_cpat:$src)>;
- // todo: use r_Int form when it will be ready
- //def : Pat<(Intr VR128:$src), (!cast<Instruction>("V"#NAME#Suffix##r_Int)
- // (VT (IMPLICIT_DEF)), VR128:$src)>;
+ }
+ let Predicates = [HasAVX] in {
def : Pat<(Intr VR128:$src),
- (vt (COPY_TO_REGCLASS(
- !cast<Instruction>("V"#NAME#Suffix##r) (ScalarVT (IMPLICIT_DEF)),
- (ScalarVT (COPY_TO_REGCLASS VR128:$src, RC))), VR128))>;
+ (!cast<Instruction>("V"#NAME#Suffix##r_Int) (vt (IMPLICIT_DEF)),
+ VR128:$src)>;
+
def : Pat<(Intr mem_cpat:$src),
(!cast<Instruction>("V"#NAME#Suffix##m_Int)
(vt (IMPLICIT_DEF)), mem_cpat:$src)>;
}
- let Predicates = [target, OptForSize] in
- def : Pat<(ScalarVT (OpNode (load addr:$src))),
+ let Predicates = [UseAVX, OptForSize] in
+ def : Pat<(ScalarVT (OpNode (load addr:$src))),
(!cast<Instruction>("V"#NAME#Suffix##m) (ScalarVT (IMPLICIT_DEF)),
addr:$src)>;
}
Sched<[itins.Sched.Folded]>;
}
-/// sse1_fp_unop_p_int - SSE1 intrinsics unops in packed forms.
-multiclass sse1_fp_unop_p_int<bits<8> opc, string OpcodeStr,
- Intrinsic V4F32Int, Intrinsic V8F32Int,
- OpndItins itins> {
-let isCodeGenOnly = 1 in {
-let Predicates = [HasAVX] in {
- def V#NAME#PSr_Int : PSI<opc, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
- !strconcat("v", OpcodeStr,
- "ps\t{$src, $dst|$dst, $src}"),
- [(set VR128:$dst, (V4F32Int VR128:$src))],
- itins.rr>, VEX, Sched<[itins.Sched]>;
- def V#NAME#PSm_Int : PSI<opc, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
- !strconcat("v", OpcodeStr,
- "ps\t{$src, $dst|$dst, $src}"),
- [(set VR128:$dst, (V4F32Int (loadv4f32 addr:$src)))],
- itins.rm>, VEX, Sched<[itins.Sched.Folded]>;
- def V#NAME#PSYr_Int : PSI<opc, MRMSrcReg, (outs VR256:$dst), (ins VR256:$src),
- !strconcat("v", OpcodeStr,
- "ps\t{$src, $dst|$dst, $src}"),
- [(set VR256:$dst, (V8F32Int VR256:$src))],
- itins.rr>, VEX, VEX_L, Sched<[itins.Sched]>;
- def V#NAME#PSYm_Int : PSI<opc, MRMSrcMem, (outs VR256:$dst),
- (ins f256mem:$src),
- !strconcat("v", OpcodeStr,
- "ps\t{$src, $dst|$dst, $src}"),
- [(set VR256:$dst, (V8F32Int (loadv8f32 addr:$src)))],
- itins.rm>, VEX, VEX_L, Sched<[itins.Sched.Folded]>;
-}
-
- def PSr_Int : PSI<opc, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
- !strconcat(OpcodeStr, "ps\t{$src, $dst|$dst, $src}"),
- [(set VR128:$dst, (V4F32Int VR128:$src))],
- itins.rr>, Sched<[itins.Sched]>;
- def PSm_Int : PSI<opc, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
- !strconcat(OpcodeStr, "ps\t{$src, $dst|$dst, $src}"),
- [(set VR128:$dst, (V4F32Int (memopv4f32 addr:$src)))],
- itins.rm>, Sched<[itins.Sched.Folded]>;
-} // isCodeGenOnly = 1
-}
-
/// sse2_fp_unop_p - SSE2 unops in vector forms.
multiclass sse2_fp_unop_p<bits<8> opc, string OpcodeStr,
SDNode OpNode, OpndItins itins> {
defm SS : sse_fp_unop_s<opc, OpcodeStr##ss, FR32, v4f32, f32, f32mem,
ssmem, sse_load_f32,
!cast<Intrinsic>("int_x86_sse_"##OpcodeStr##_ss), OpNode,
- itins, UseSSE1, "SS">, XS;
+ SSEPackedSingle, itins, UseSSE1, "SS">, XS;
defm V#NAME#SS : avx_fp_unop_s<opc, "v"#OpcodeStr##ss, FR32, v4f32, f32,
f32mem, ssmem, sse_load_f32,
!cast<Intrinsic>("int_x86_sse_"##OpcodeStr##_ss), OpNode,
- itins, UseAVX, "SS">, XS, VEX_4V, VEX_LIG;
+ SSEPackedSingle, itins, "SS">, XS, VEX_4V, VEX_LIG;
}
multiclass sse2_fp_unop_s<bits<8> opc, string OpcodeStr, SDNode OpNode,
defm SD : sse_fp_unop_s<opc, OpcodeStr##sd, FR64, v2f64, f64, f64mem,
sdmem, sse_load_f64,
!cast<Intrinsic>("int_x86_sse2_"##OpcodeStr##_sd),
- OpNode, itins, UseSSE2, "SD">, XD;
+ OpNode, SSEPackedDouble, itins, UseSSE2, "SD">, XD;
defm V#NAME#SD : avx_fp_unop_s<opc, "v"#OpcodeStr##sd, FR64, v2f64, f64,
f64mem, sdmem, sse_load_f64,
!cast<Intrinsic>("int_x86_sse2_"##OpcodeStr##_sd),
- OpNode, itins, UseAVX, "SD">, XD, VEX_4V, VEX_LIG;
+ OpNode, SSEPackedDouble, itins, "SD">,
+ XD, VEX_4V, VEX_LIG;
}
// Square root.
// Reciprocal approximations. Note that these typically require refinement
// in order to obtain suitable precision.
defm RSQRT : sse1_fp_unop_s<0x52, "rsqrt", X86frsqrt, SSE_RSQRTSS>,
- sse1_fp_unop_p<0x52, "rsqrt", X86frsqrt, SSE_RSQRTPS>,
- sse1_fp_unop_p_int<0x52, "rsqrt", int_x86_sse_rsqrt_ps,
- int_x86_avx_rsqrt_ps_256, SSE_RSQRTPS>;
+ sse1_fp_unop_p<0x52, "rsqrt", X86frsqrt, SSE_RSQRTPS>;
defm RCP : sse1_fp_unop_s<0x53, "rcp", X86frcp, SSE_RCPS>,
- sse1_fp_unop_p<0x53, "rcp", X86frcp, SSE_RCPP>,
- sse1_fp_unop_p_int<0x53, "rcp", int_x86_sse_rcp_ps,
- int_x86_avx_rcp_ps_256, SSE_RCPP>;
+ sse1_fp_unop_p<0x53, "rcp", X86frcp, SSE_RCPP>;
// There is no f64 version of the reciprocal approximation instructions.
+// TODO: We should add *scalar* op patterns for these just like we have for
+// the binops above. If the binop and unop patterns could all be unified
+// that would be even better.
+
+multiclass scalar_unary_math_patterns<Intrinsic Intr, string OpcPrefix,
+ SDNode Move, ValueType VT,
+ Predicate BasePredicate> {
+ let Predicates = [BasePredicate] in {
+ def : Pat<(VT (Move VT:$dst, (Intr VT:$src))),
+ (!cast<I>(OpcPrefix#r_Int) VT:$dst, VT:$src)>;
+ }
+
+ // With SSE 4.1, blendi is preferred to movs*, so match that too.
+ let Predicates = [UseSSE41] in {
+ def : Pat<(VT (X86Blendi VT:$dst, (Intr VT:$src), (i8 1))),
+ (!cast<I>(OpcPrefix#r_Int) VT:$dst, VT:$src)>;
+ }
+
+ // Repeat for AVX versions of the instructions.
+ let Predicates = [HasAVX] in {
+ def : Pat<(VT (Move VT:$dst, (Intr VT:$src))),
+ (!cast<I>("V"#OpcPrefix#r_Int) VT:$dst, VT:$src)>;
+
+ def : Pat<(VT (X86Blendi VT:$dst, (Intr VT:$src), (i8 1))),
+ (!cast<I>("V"#OpcPrefix#r_Int) VT:$dst, VT:$src)>;
+ }
+}
+
+defm : scalar_unary_math_patterns<int_x86_sse_rcp_ss, "RCPSS", X86Movss,
+ v4f32, UseSSE1>;
+defm : scalar_unary_math_patterns<int_x86_sse_rsqrt_ss, "RSQRTSS", X86Movss,
+ v4f32, UseSSE1>;
+defm : scalar_unary_math_patterns<int_x86_sse_sqrt_ss, "SQRTSS", X86Movss,
+ v4f32, UseSSE1>;
+defm : scalar_unary_math_patterns<int_x86_sse2_sqrt_sd, "SQRTSD", X86Movsd,
+ v2f64, UseSSE2>;
+
+
//===----------------------------------------------------------------------===//
// SSE 1 & 2 - Non-temporal stores
//===----------------------------------------------------------------------===//
PS, Requires<[HasSSE2]>;
} // SchedRW = [WriteStore]
+let Predicates = [HasAVX2, NoVLX] in {
+ def : Pat<(alignednontemporalstore (v8i32 VR256:$src), addr:$dst),
+ (VMOVNTDQYmr addr:$dst, VR256:$src)>;
+ def : Pat<(alignednontemporalstore (v16i16 VR256:$src), addr:$dst),
+ (VMOVNTDQYmr addr:$dst, VR256:$src)>;
+ def : Pat<(alignednontemporalstore (v32i8 VR256:$src), addr:$dst),
+ (VMOVNTDQYmr addr:$dst, VR256:$src)>;
+}
+
let Predicates = [HasAVX, NoVLX] in {
def : Pat<(alignednontemporalstore (v4i32 VR128:$src), addr:$dst),
- (VMOVNTPSmr addr:$dst, VR128:$src)>;
+ (VMOVNTDQmr addr:$dst, VR128:$src)>;
+ def : Pat<(alignednontemporalstore (v8i16 VR128:$src), addr:$dst),
+ (VMOVNTDQmr addr:$dst, VR128:$src)>;
+ def : Pat<(alignednontemporalstore (v16i8 VR128:$src), addr:$dst),
+ (VMOVNTDQmr addr:$dst, VR128:$src)>;
}
def : Pat<(alignednontemporalstore (v4i32 VR128:$src), addr:$dst),
- (MOVNTPSmr addr:$dst, VR128:$src)>;
+ (MOVNTDQmr addr:$dst, VR128:$src)>;
+def : Pat<(alignednontemporalstore (v8i16 VR128:$src), addr:$dst),
+ (MOVNTDQmr addr:$dst, VR128:$src)>;
+def : Pat<(alignednontemporalstore (v16i8 VR128:$src), addr:$dst),
+ (MOVNTDQmr addr:$dst, VR128:$src)>;
} // AddedComplexity
} // ExeDomain = SSEPackedInt
defm PADDB : PDI_binop_all<0xFC, "paddb", add, v16i8, v32i8,
- SSE_INTALU_ITINS_P, 1>;
+ SSE_INTALU_ITINS_P, 1, NoVLX_Or_NoBWI>;
defm PADDW : PDI_binop_all<0xFD, "paddw", add, v8i16, v16i16,
- SSE_INTALU_ITINS_P, 1>;
+ SSE_INTALU_ITINS_P, 1, NoVLX_Or_NoBWI>;
defm PADDD : PDI_binop_all<0xFE, "paddd", add, v4i32, v8i32,
- SSE_INTALU_ITINS_P, 1>;
+ SSE_INTALU_ITINS_P, 1, NoVLX>;
defm PADDQ : PDI_binop_all<0xD4, "paddq", add, v2i64, v4i64,
- SSE_INTALUQ_ITINS_P, 1>;
+ SSE_INTALUQ_ITINS_P, 1, NoVLX>;
defm PMULLW : PDI_binop_all<0xD5, "pmullw", mul, v8i16, v16i16,
- SSE_INTMUL_ITINS_P, 1>;
+ SSE_INTMUL_ITINS_P, 1, NoVLX_Or_NoBWI>;
defm PMULHUW : PDI_binop_all<0xE4, "pmulhuw", mulhu, v8i16, v16i16,
- SSE_INTMUL_ITINS_P, 1>;
+ SSE_INTMUL_ITINS_P, 1, NoVLX_Or_NoBWI>;
defm PMULHW : PDI_binop_all<0xE5, "pmulhw", mulhs, v8i16, v16i16,
- SSE_INTMUL_ITINS_P, 1>;
+ SSE_INTMUL_ITINS_P, 1, NoVLX_Or_NoBWI>;
defm PSUBB : PDI_binop_all<0xF8, "psubb", sub, v16i8, v32i8,
- SSE_INTALU_ITINS_P, 0>;
+ SSE_INTALU_ITINS_P, 0, NoVLX_Or_NoBWI>;
defm PSUBW : PDI_binop_all<0xF9, "psubw", sub, v8i16, v16i16,
- SSE_INTALU_ITINS_P, 0>;
+ SSE_INTALU_ITINS_P, 0, NoVLX_Or_NoBWI>;
defm PSUBD : PDI_binop_all<0xFA, "psubd", sub, v4i32, v8i32,
- SSE_INTALU_ITINS_P, 0>;
+ SSE_INTALU_ITINS_P, 0, NoVLX>;
defm PSUBQ : PDI_binop_all<0xFB, "psubq", sub, v2i64, v4i64,
- SSE_INTALUQ_ITINS_P, 0>;
+ SSE_INTALUQ_ITINS_P, 0, NoVLX>;
defm PSUBUSB : PDI_binop_all<0xD8, "psubusb", X86subus, v16i8, v32i8,
- SSE_INTALU_ITINS_P, 0>;
+ SSE_INTALU_ITINS_P, 0, NoVLX_Or_NoBWI>;
defm PSUBUSW : PDI_binop_all<0xD9, "psubusw", X86subus, v8i16, v16i16,
- SSE_INTALU_ITINS_P, 0>;
-defm PMINUB : PDI_binop_all<0xDA, "pminub", X86umin, v16i8, v32i8,
- SSE_INTALU_ITINS_P, 1>;
-defm PMINSW : PDI_binop_all<0xEA, "pminsw", X86smin, v8i16, v16i16,
- SSE_INTALU_ITINS_P, 1>;
-defm PMAXUB : PDI_binop_all<0xDE, "pmaxub", X86umax, v16i8, v32i8,
- SSE_INTALU_ITINS_P, 1>;
-defm PMAXSW : PDI_binop_all<0xEE, "pmaxsw", X86smax, v8i16, v16i16,
- SSE_INTALU_ITINS_P, 1>;
+ SSE_INTALU_ITINS_P, 0, NoVLX_Or_NoBWI>;
+defm PMINUB : PDI_binop_all<0xDA, "pminub", umin, v16i8, v32i8,
+ SSE_INTALU_ITINS_P, 1, NoVLX_Or_NoBWI>;
+defm PMINSW : PDI_binop_all<0xEA, "pminsw", smin, v8i16, v16i16,
+ SSE_INTALU_ITINS_P, 1, NoVLX_Or_NoBWI>;
+defm PMAXUB : PDI_binop_all<0xDE, "pmaxub", umax, v16i8, v32i8,
+ SSE_INTALU_ITINS_P, 1, NoVLX_Or_NoBWI>;
+defm PMAXSW : PDI_binop_all<0xEE, "pmaxsw", smax, v8i16, v16i16,
+ SSE_INTALU_ITINS_P, 1, NoVLX_Or_NoBWI>;
// Intrinsic forms
defm PSUBSB : PDI_binop_all_int<0xE8, "psubsb", int_x86_sse2_psubs_b,
defm PSADBW : PDI_binop_all_int<0xF6, "psadbw", int_x86_sse2_psad_bw,
int_x86_avx2_psad_bw, SSE_PMADD, 1>;
+let Predicates = [HasAVX2] in
+ def : Pat<(v32i8 (X86psadbw (v32i8 VR256:$src1),
+ (v32i8 VR256:$src2))),
+ (VPSADBWYrr VR256:$src2, VR256:$src1)>;
+
+let Predicates = [HasAVX] in
+ def : Pat<(v16i8 (X86psadbw (v16i8 VR128:$src1),
+ (v16i8 VR128:$src2))),
+ (VPSADBWrr VR128:$src2, VR128:$src1)>;
+
+def : Pat<(v16i8 (X86psadbw (v16i8 VR128:$src1),
+ (v16i8 VR128:$src2))),
+ (PSADBWrr VR128:$src2, VR128:$src1)>;
+
let Predicates = [HasAVX] in
defm VPMULUDQ : PDI_binop_rm2<0xF4, "vpmuludq", X86pmuludq, v2i64, v4i32, VR128,
loadv2i64, i128mem, SSE_INTMUL_ITINS_P, 1, 0>,
}
} // Constraints = "$src1 = $dst"
-let Predicates = [HasAVX] in {
- def : Pat<(v2f64 (X86fsrl VR128:$src1, i32immSExt8:$src2)),
- (VPSRLDQri VR128:$src1, (BYTE_imm imm:$src2))>;
-}
-
-let Predicates = [UseSSE2] in {
- def : Pat<(v2f64 (X86fsrl VR128:$src1, i32immSExt8:$src2)),
- (PSRLDQri VR128:$src1, (BYTE_imm imm:$src2))>;
-}
-
//===---------------------------------------------------------------------===//
// SSE2 - Packed Integer Comparison Instructions
//===---------------------------------------------------------------------===//
defm PCMPEQB : PDI_binop_all<0x74, "pcmpeqb", X86pcmpeq, v16i8, v32i8,
- SSE_INTALU_ITINS_P, 1>;
+ SSE_INTALU_ITINS_P, 1, NoVLX_Or_NoBWI>;
defm PCMPEQW : PDI_binop_all<0x75, "pcmpeqw", X86pcmpeq, v8i16, v16i16,
- SSE_INTALU_ITINS_P, 1>;
+ SSE_INTALU_ITINS_P, 1, NoVLX_Or_NoBWI>;
defm PCMPEQD : PDI_binop_all<0x76, "pcmpeqd", X86pcmpeq, v4i32, v8i32,
- SSE_INTALU_ITINS_P, 1>;
+ SSE_INTALU_ITINS_P, 1, NoVLX>;
defm PCMPGTB : PDI_binop_all<0x64, "pcmpgtb", X86pcmpgt, v16i8, v32i8,
- SSE_INTALU_ITINS_P, 0>;
+ SSE_INTALU_ITINS_P, 0, NoVLX_Or_NoBWI>;
defm PCMPGTW : PDI_binop_all<0x65, "pcmpgtw", X86pcmpgt, v8i16, v16i16,
- SSE_INTALU_ITINS_P, 0>;
+ SSE_INTALU_ITINS_P, 0, NoVLX_Or_NoBWI>;
defm PCMPGTD : PDI_binop_all<0x66, "pcmpgtd", X86pcmpgt, v4i32, v8i32,
- SSE_INTALU_ITINS_P, 0>;
+ SSE_INTALU_ITINS_P, 0, NoVLX>;
//===---------------------------------------------------------------------===//
// SSE2 - Packed Integer Shuffle Instructions
Sched<[WriteShuffleLd, ReadAfterLd]>;
}
-let Predicates = [HasAVX] in {
+
+let Predicates = [HasAVX, NoVLX_Or_NoBWI] in {
defm VPUNPCKLBW : sse2_unpack<0x60, "vpunpcklbw", v16i8, X86Unpckl,
bc_v16i8, loadv2i64, 0>, VEX_4V;
defm VPUNPCKLWD : sse2_unpack<0x61, "vpunpcklwd", v8i16, X86Unpckl,
bc_v8i16, loadv2i64, 0>, VEX_4V;
- defm VPUNPCKLDQ : sse2_unpack<0x62, "vpunpckldq", v4i32, X86Unpckl,
- bc_v4i32, loadv2i64, 0>, VEX_4V;
- defm VPUNPCKLQDQ : sse2_unpack<0x6C, "vpunpcklqdq", v2i64, X86Unpckl,
- bc_v2i64, loadv2i64, 0>, VEX_4V;
-
defm VPUNPCKHBW : sse2_unpack<0x68, "vpunpckhbw", v16i8, X86Unpckh,
bc_v16i8, loadv2i64, 0>, VEX_4V;
defm VPUNPCKHWD : sse2_unpack<0x69, "vpunpckhwd", v8i16, X86Unpckh,
bc_v8i16, loadv2i64, 0>, VEX_4V;
+}
+let Predicates = [HasAVX, NoVLX] in {
+ defm VPUNPCKLDQ : sse2_unpack<0x62, "vpunpckldq", v4i32, X86Unpckl,
+ bc_v4i32, loadv2i64, 0>, VEX_4V;
+ defm VPUNPCKLQDQ : sse2_unpack<0x6C, "vpunpcklqdq", v2i64, X86Unpckl,
+ bc_v2i64, loadv2i64, 0>, VEX_4V;
defm VPUNPCKHDQ : sse2_unpack<0x6A, "vpunpckhdq", v4i32, X86Unpckh,
bc_v4i32, loadv2i64, 0>, VEX_4V;
defm VPUNPCKHQDQ : sse2_unpack<0x6D, "vpunpckhqdq", v2i64, X86Unpckh,
bc_v2i64, loadv2i64, 0>, VEX_4V;
}
-let Predicates = [HasAVX2] in {
+let Predicates = [HasAVX2, NoVLX_Or_NoBWI] in {
defm VPUNPCKLBW : sse2_unpack_y<0x60, "vpunpcklbw", v32i8, X86Unpckl,
bc_v32i8>, VEX_4V, VEX_L;
defm VPUNPCKLWD : sse2_unpack_y<0x61, "vpunpcklwd", v16i16, X86Unpckl,
bc_v16i16>, VEX_4V, VEX_L;
- defm VPUNPCKLDQ : sse2_unpack_y<0x62, "vpunpckldq", v8i32, X86Unpckl,
- bc_v8i32>, VEX_4V, VEX_L;
- defm VPUNPCKLQDQ : sse2_unpack_y<0x6C, "vpunpcklqdq", v4i64, X86Unpckl,
- bc_v4i64>, VEX_4V, VEX_L;
-
defm VPUNPCKHBW : sse2_unpack_y<0x68, "vpunpckhbw", v32i8, X86Unpckh,
bc_v32i8>, VEX_4V, VEX_L;
defm VPUNPCKHWD : sse2_unpack_y<0x69, "vpunpckhwd", v16i16, X86Unpckh,
bc_v16i16>, VEX_4V, VEX_L;
+}
+let Predicates = [HasAVX2, NoVLX] in {
+ defm VPUNPCKLDQ : sse2_unpack_y<0x62, "vpunpckldq", v8i32, X86Unpckl,
+ bc_v8i32>, VEX_4V, VEX_L;
+ defm VPUNPCKLQDQ : sse2_unpack_y<0x6C, "vpunpcklqdq", v4i64, X86Unpckl,
+ bc_v4i64>, VEX_4V, VEX_L;
defm VPUNPCKHDQ : sse2_unpack_y<0x6A, "vpunpckhdq", v8i32, X86Unpckh,
bc_v8i32>, VEX_4V, VEX_L;
defm VPUNPCKHQDQ : sse2_unpack_y<0x6D, "vpunpckhqdq", v4i64, X86Unpckh,
def : Pat<(v4i32 (X86vzmovl (v4i32 (scalar_to_vector GR32:$src)))),
(VMOVDI2PDIrr GR32:$src)>;
- // AVX 128-bit movd/movq instruction write zeros in the high 128-bit part.
+ // AVX 128-bit movd/movq instructions write zeros in the high 128-bit part.
+ // These instructions also write zeros in the high part of a 256-bit register.
let AddedComplexity = 20 in {
def : Pat<(v4i32 (X86vzmovl (v4i32 (scalar_to_vector (loadi32 addr:$src))))),
(VMOVDI2PDIrm addr:$src)>;
(VMOVDI2PDIrm addr:$src)>;
def : Pat<(v4i32 (X86vzmovl (bc_v4i32 (loadv2i64 addr:$src)))),
(VMOVDI2PDIrm addr:$src)>;
+ def : Pat<(v8i32 (X86vzmovl (insert_subvector undef,
+ (v4i32 (scalar_to_vector (loadi32 addr:$src))), (iPTR 0)))),
+ (SUBREG_TO_REG (i32 0), (VMOVDI2PDIrm addr:$src), sub_xmm)>;
}
// Use regular 128-bit instructions to match 256-bit scalar_to_vec+zext.
def : Pat<(v8i32 (X86vzmovl (insert_subvector undef,
//===---------------------------------------------------------------------===//
// Store / copy lower 64-bits of a XMM register.
//
-let Predicates = [UseAVX] in
+let Predicates = [HasAVX] in
def : Pat<(int_x86_sse2_storel_dq addr:$dst, VR128:$src),
(VMOVPQI2QImr addr:$dst, VR128:$src)>;
let Predicates = [UseSSE2] in
(VMOVZQI2PQIrm addr:$src)>;
def : Pat<(v2i64 (X86vzload addr:$src)),
(VMOVZQI2PQIrm addr:$src)>;
+ def : Pat<(v4i64 (X86vzmovl (insert_subvector undef,
+ (v2i64 (scalar_to_vector (loadi64 addr:$src))), (iPTR 0)))),
+ (SUBREG_TO_REG (i64 0), (VMOVZQI2PQIrm addr:$src), sub_xmm)>;
}
let Predicates = [UseSSE2], AddedComplexity = 20 in {
let Predicates = [UseAVX, OptForSize] in {
def : Pat<(v2f64 (X86VBroadcast (loadf64 addr:$src))),
- (VMOVDDUPrm addr:$src)>;
+ (VMOVDDUPrm addr:$src)>;
def : Pat<(v2i64 (X86VBroadcast (loadi64 addr:$src))),
- (VMOVDDUPrm addr:$src)>;
+ (VMOVDDUPrm addr:$src)>;
}
let Predicates = [UseSSE3] in {
OpndItins SSEItins, OpndItins AVXItins,
OpndItins AVX2Itins> {
defm NAME : SS41I_pmovx_rrrm<opc, OpcodeStr, MemOp, VR128, VR128, SSEItins>;
- let Predicates = [HasAVX] in
+ let Predicates = [HasAVX, NoVLX] in
defm V#NAME : SS41I_pmovx_rrrm<opc, !strconcat("v", OpcodeStr), MemOp,
VR128, VR128, AVXItins>, VEX;
- let Predicates = [HasAVX2] in
+ let Predicates = [HasAVX2, NoVLX] in
defm V#NAME#Y : SS41I_pmovx_rrrm<opc, !strconcat("v", OpcodeStr), MemYOp,
VR256, VR128, AVX2Itins>, VEX, VEX_L;
}
(!cast<I>(OpcPrefix#DQYrm) addr:$src)>;
}
-let Predicates = [HasAVX2] in {
+let Predicates = [HasAVX2, NoVLX] in {
defm : SS41I_pmovx_avx2_patterns<"VPMOVSX", "s", X86vsext>;
defm : SS41I_pmovx_avx2_patterns<"VPMOVZX", "z", X86vzext>;
}
(!cast<I>(OpcPrefix#DQrm) addr:$src)>;
}
-let Predicates = [HasAVX] in {
+let Predicates = [HasAVX, NoVLX] in {
defm : SS41I_pmovx_patterns<"VPMOVSX", "s", X86vsext, extloadi32i16>;
defm : SS41I_pmovx_patterns<"VPMOVZX", "z", X86vzext, loadi16_anyext>;
}
}
let Predicates = [HasAVX, NoVLX] in {
- let isCommutable = 0 in
- defm VPMINSB : SS48I_binop_rm<0x38, "vpminsb", X86smin, v16i8, VR128,
+ defm VPMINSB : SS48I_binop_rm<0x38, "vpminsb", smin, v16i8, VR128,
loadv2i64, i128mem, 0, SSE_INTALU_ITINS_P>,
VEX_4V;
- defm VPMINSD : SS48I_binop_rm<0x39, "vpminsd", X86smin, v4i32, VR128,
+ defm VPMINSD : SS48I_binop_rm<0x39, "vpminsd", smin, v4i32, VR128,
loadv2i64, i128mem, 0, SSE_INTALU_ITINS_P>,
VEX_4V;
- defm VPMINUD : SS48I_binop_rm<0x3B, "vpminud", X86umin, v4i32, VR128,
+ defm VPMINUD : SS48I_binop_rm<0x3B, "vpminud", umin, v4i32, VR128,
loadv2i64, i128mem, 0, SSE_INTALU_ITINS_P>,
VEX_4V;
- defm VPMINUW : SS48I_binop_rm<0x3A, "vpminuw", X86umin, v8i16, VR128,
+ defm VPMINUW : SS48I_binop_rm<0x3A, "vpminuw", umin, v8i16, VR128,
loadv2i64, i128mem, 0, SSE_INTALU_ITINS_P>,
VEX_4V;
- defm VPMAXSB : SS48I_binop_rm<0x3C, "vpmaxsb", X86smax, v16i8, VR128,
+ defm VPMAXSB : SS48I_binop_rm<0x3C, "vpmaxsb", smax, v16i8, VR128,
loadv2i64, i128mem, 0, SSE_INTALU_ITINS_P>,
VEX_4V;
- defm VPMAXSD : SS48I_binop_rm<0x3D, "vpmaxsd", X86smax, v4i32, VR128,
+ defm VPMAXSD : SS48I_binop_rm<0x3D, "vpmaxsd", smax, v4i32, VR128,
loadv2i64, i128mem, 0, SSE_INTALU_ITINS_P>,
VEX_4V;
- defm VPMAXUD : SS48I_binop_rm<0x3F, "vpmaxud", X86umax, v4i32, VR128,
+ defm VPMAXUD : SS48I_binop_rm<0x3F, "vpmaxud", umax, v4i32, VR128,
loadv2i64, i128mem, 0, SSE_INTALU_ITINS_P>,
VEX_4V;
- defm VPMAXUW : SS48I_binop_rm<0x3E, "vpmaxuw", X86umax, v8i16, VR128,
+ defm VPMAXUW : SS48I_binop_rm<0x3E, "vpmaxuw", umax, v8i16, VR128,
loadv2i64, i128mem, 0, SSE_INTALU_ITINS_P>,
VEX_4V;
defm VPMULDQ : SS48I_binop_rm2<0x28, "vpmuldq", X86pmuldq, v2i64, v4i32,
}
let Predicates = [HasAVX2, NoVLX] in {
- let isCommutable = 0 in
- defm VPMINSBY : SS48I_binop_rm<0x38, "vpminsb", X86smin, v32i8, VR256,
+ defm VPMINSBY : SS48I_binop_rm<0x38, "vpminsb", smin, v32i8, VR256,
loadv4i64, i256mem, 0, SSE_INTALU_ITINS_P>,
VEX_4V, VEX_L;
- defm VPMINSDY : SS48I_binop_rm<0x39, "vpminsd", X86smin, v8i32, VR256,
+ defm VPMINSDY : SS48I_binop_rm<0x39, "vpminsd", smin, v8i32, VR256,
loadv4i64, i256mem, 0, SSE_INTALU_ITINS_P>,
VEX_4V, VEX_L;
- defm VPMINUDY : SS48I_binop_rm<0x3B, "vpminud", X86umin, v8i32, VR256,
+ defm VPMINUDY : SS48I_binop_rm<0x3B, "vpminud", umin, v8i32, VR256,
loadv4i64, i256mem, 0, SSE_INTALU_ITINS_P>,
VEX_4V, VEX_L;
- defm VPMINUWY : SS48I_binop_rm<0x3A, "vpminuw", X86umin, v16i16, VR256,
+ defm VPMINUWY : SS48I_binop_rm<0x3A, "vpminuw", umin, v16i16, VR256,
loadv4i64, i256mem, 0, SSE_INTALU_ITINS_P>,
VEX_4V, VEX_L;
- defm VPMAXSBY : SS48I_binop_rm<0x3C, "vpmaxsb", X86smax, v32i8, VR256,
+ defm VPMAXSBY : SS48I_binop_rm<0x3C, "vpmaxsb", smax, v32i8, VR256,
loadv4i64, i256mem, 0, SSE_INTALU_ITINS_P>,
VEX_4V, VEX_L;
- defm VPMAXSDY : SS48I_binop_rm<0x3D, "vpmaxsd", X86smax, v8i32, VR256,
+ defm VPMAXSDY : SS48I_binop_rm<0x3D, "vpmaxsd", smax, v8i32, VR256,
loadv4i64, i256mem, 0, SSE_INTALU_ITINS_P>,
VEX_4V, VEX_L;
- defm VPMAXUDY : SS48I_binop_rm<0x3F, "vpmaxud", X86umax, v8i32, VR256,
+ defm VPMAXUDY : SS48I_binop_rm<0x3F, "vpmaxud", umax, v8i32, VR256,
loadv4i64, i256mem, 0, SSE_INTALU_ITINS_P>,
VEX_4V, VEX_L;
- defm VPMAXUWY : SS48I_binop_rm<0x3E, "vpmaxuw", X86umax, v16i16, VR256,
+ defm VPMAXUWY : SS48I_binop_rm<0x3E, "vpmaxuw", umax, v16i16, VR256,
loadv4i64, i256mem, 0, SSE_INTALU_ITINS_P>,
VEX_4V, VEX_L;
defm VPMULDQY : SS48I_binop_rm2<0x28, "vpmuldq", X86pmuldq, v4i64, v8i32,
}
let Constraints = "$src1 = $dst" in {
- let isCommutable = 0 in
- defm PMINSB : SS48I_binop_rm<0x38, "pminsb", X86smin, v16i8, VR128,
+ defm PMINSB : SS48I_binop_rm<0x38, "pminsb", smin, v16i8, VR128,
memopv2i64, i128mem, 1, SSE_INTALU_ITINS_P>;
- defm PMINSD : SS48I_binop_rm<0x39, "pminsd", X86smin, v4i32, VR128,
+ defm PMINSD : SS48I_binop_rm<0x39, "pminsd", smin, v4i32, VR128,
memopv2i64, i128mem, 1, SSE_INTALU_ITINS_P>;
- defm PMINUD : SS48I_binop_rm<0x3B, "pminud", X86umin, v4i32, VR128,
+ defm PMINUD : SS48I_binop_rm<0x3B, "pminud", umin, v4i32, VR128,
memopv2i64, i128mem, 1, SSE_INTALU_ITINS_P>;
- defm PMINUW : SS48I_binop_rm<0x3A, "pminuw", X86umin, v8i16, VR128,
+ defm PMINUW : SS48I_binop_rm<0x3A, "pminuw", umin, v8i16, VR128,
memopv2i64, i128mem, 1, SSE_INTALU_ITINS_P>;
- defm PMAXSB : SS48I_binop_rm<0x3C, "pmaxsb", X86smax, v16i8, VR128,
+ defm PMAXSB : SS48I_binop_rm<0x3C, "pmaxsb", smax, v16i8, VR128,
memopv2i64, i128mem, 1, SSE_INTALU_ITINS_P>;
- defm PMAXSD : SS48I_binop_rm<0x3D, "pmaxsd", X86smax, v4i32, VR128,
+ defm PMAXSD : SS48I_binop_rm<0x3D, "pmaxsd", smax, v4i32, VR128,
memopv2i64, i128mem, 1, SSE_INTALU_ITINS_P>;
- defm PMAXUD : SS48I_binop_rm<0x3F, "pmaxud", X86umax, v4i32, VR128,
+ defm PMAXUD : SS48I_binop_rm<0x3F, "pmaxud", umax, v4i32, VR128,
memopv2i64, i128mem, 1, SSE_INTALU_ITINS_P>;
- defm PMAXUW : SS48I_binop_rm<0x3E, "pmaxuw", X86umax, v8i16, VR128,
+ defm PMAXUW : SS48I_binop_rm<0x3E, "pmaxuw", umax, v8i16, VR128,
memopv2i64, i128mem, 1, SSE_INTALU_ITINS_P>;
defm PMULDQ : SS48I_binop_rm2<0x28, "pmuldq", X86pmuldq, v2i64, v4i32,
VR128, memopv2i64, i128mem,
}
// Patterns
+// FIXME: Prefer a movss or movsd over a blendps when optimizing for size or
+// on targets where they have equal performance. These were changed to use
+// blends because blends have better throughput on SandyBridge and Haswell, but
+// movs[s/d] are 1-2 byte shorter instructions.
let Predicates = [UseAVX] in {
let AddedComplexity = 15 in {
// Move scalar to XMM zero-extended, zeroing a VR128 then do a
// Move low f32 and clear high bits.
def : Pat<(v8f32 (X86vzmovl (v8f32 VR256:$src))),
(VBLENDPSYrri (v8f32 (AVX_SET0)), VR256:$src, (i8 1))>;
- def : Pat<(v8i32 (X86vzmovl (v8i32 VR256:$src))),
- (VBLENDPSYrri (v8i32 (AVX_SET0)), VR256:$src, (i8 1))>;
+
+ // Move low f64 and clear high bits.
+ def : Pat<(v4f64 (X86vzmovl (v4f64 VR256:$src))),
+ (VBLENDPDYrri (v4f64 (AVX_SET0)), VR256:$src, (i8 1))>;
}
def : Pat<(v8f32 (X86vzmovl (insert_subvector undef,
(v2f64 (VMOVSDrr (v2f64 (V_SET0)), FR64:$src)),
sub_xmm)>;
- // Move low f64 and clear high bits.
- def : Pat<(v4f64 (X86vzmovl (v4f64 VR256:$src))),
- (VBLENDPDYrri (v4f64 (AVX_SET0)), VR256:$src, (i8 1))>;
-
+ // These will incur an FP/int domain crossing penalty, but it may be the only
+ // way without AVX2. Do not add any complexity because we may be able to match
+ // more optimal patterns defined earlier in this file.
+ def : Pat<(v8i32 (X86vzmovl (v8i32 VR256:$src))),
+ (VBLENDPSYrri (v8i32 (AVX_SET0)), VR256:$src, (i8 1))>;
def : Pat<(v4i64 (X86vzmovl (v4i64 VR256:$src))),
(VBLENDPDYrri (v4i64 (AVX_SET0)), VR256:$src, (i8 1))>;
}
+// FIXME: Prefer a movss or movsd over a blendps when optimizing for size or
+// on targets where they have equal performance. These were changed to use
+// blends because blends have better throughput on SandyBridge and Haswell, but
+// movs[s/d] are 1-2 byte shorter instructions.
let Predicates = [UseSSE41] in {
// With SSE41 we can use blends for these patterns.
def : Pat<(v4f32 (X86vzmovl (v4f32 VR128:$src))),
def EXTRQI : Ii8<0x78, MRMXr, (outs VR128:$dst),
(ins VR128:$src, u8imm:$len, u8imm:$idx),
"extrq\t{$idx, $len, $src|$src, $len, $idx}",
- [(set VR128:$dst, (int_x86_sse4a_extrqi VR128:$src, imm:$len,
+ [(set VR128:$dst, (X86extrqi VR128:$src, imm:$len,
imm:$idx))]>, PD;
def EXTRQ : I<0x79, MRMSrcReg, (outs VR128:$dst),
(ins VR128:$src, VR128:$mask),
def INSERTQI : Ii8<0x78, MRMSrcReg, (outs VR128:$dst),
(ins VR128:$src, VR128:$src2, u8imm:$len, u8imm:$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;
+ [(set VR128:$dst, (X86insertqi VR128:$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, $mask}",
// VBROADCAST - Load from memory and broadcast to all elements of the
// destination operand
//
-class avx_broadcast<bits<8> opc, string OpcodeStr, RegisterClass RC,
- X86MemOperand x86memop, Intrinsic Int, SchedWrite Sched> :
- AVX8I<opc, MRMSrcMem, (outs RC:$dst), (ins x86memop:$src),
- !strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"),
- [(set RC:$dst, (Int addr:$src))]>, Sched<[Sched]>, VEX;
-
-class avx_broadcast_no_int<bits<8> opc, string OpcodeStr, RegisterClass RC,
+class avx_broadcast_rm<bits<8> opc, string OpcodeStr, RegisterClass RC,
X86MemOperand x86memop, ValueType VT,
PatFrag ld_frag, SchedWrite Sched> :
AVX8I<opc, MRMSrcMem, (outs RC:$dst), (ins x86memop:$src),
}
// AVX2 adds register forms
-class avx2_broadcast_reg<bits<8> opc, string OpcodeStr, RegisterClass RC,
- Intrinsic Int, SchedWrite Sched> :
+class avx2_broadcast_rr<bits<8> opc, string OpcodeStr, RegisterClass RC,
+ ValueType ResVT, ValueType OpVT, SchedWrite Sched> :
AVX28I<opc, MRMSrcReg, (outs RC:$dst), (ins VR128:$src),
!strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"),
- [(set RC:$dst, (Int VR128:$src))]>, Sched<[Sched]>, VEX;
+ [(set RC:$dst, (ResVT (X86VBroadcast (OpVT VR128:$src))))]>,
+ Sched<[Sched]>, VEX;
let ExeDomain = SSEPackedSingle in {
- def VBROADCASTSSrm : avx_broadcast_no_int<0x18, "vbroadcastss", VR128,
+ def VBROADCASTSSrm : avx_broadcast_rm<0x18, "vbroadcastss", VR128,
f32mem, v4f32, loadf32, WriteLoad>;
- def VBROADCASTSSYrm : avx_broadcast_no_int<0x18, "vbroadcastss", VR256,
+ def VBROADCASTSSYrm : avx_broadcast_rm<0x18, "vbroadcastss", VR256,
f32mem, v8f32, loadf32,
WriteFShuffleLd>, VEX_L;
}
let ExeDomain = SSEPackedDouble in
-def VBROADCASTSDYrm : avx_broadcast_no_int<0x19, "vbroadcastsd", VR256, f64mem,
+def VBROADCASTSDYrm : avx_broadcast_rm<0x19, "vbroadcastsd", VR256, f64mem,
v4f64, loadf64, WriteFShuffleLd>, VEX_L;
-def VBROADCASTF128 : avx_broadcast<0x1A, "vbroadcastf128", VR256, f128mem,
- int_x86_avx_vbroadcastf128_pd_256,
- WriteFShuffleLd>, VEX_L;
let ExeDomain = SSEPackedSingle in {
- def VBROADCASTSSrr : avx2_broadcast_reg<0x18, "vbroadcastss", VR128,
- int_x86_avx2_vbroadcast_ss_ps,
- WriteFShuffle>;
- def VBROADCASTSSYrr : avx2_broadcast_reg<0x18, "vbroadcastss", VR256,
- int_x86_avx2_vbroadcast_ss_ps_256,
- WriteFShuffle256>, VEX_L;
+ def VBROADCASTSSrr : avx2_broadcast_rr<0x18, "vbroadcastss", VR128,
+ v4f32, v4f32, WriteFShuffle>;
+ def VBROADCASTSSYrr : avx2_broadcast_rr<0x18, "vbroadcastss", VR256,
+ v8f32, v4f32, WriteFShuffle256>, VEX_L;
}
let ExeDomain = SSEPackedDouble in
-def VBROADCASTSDYrr : avx2_broadcast_reg<0x19, "vbroadcastsd", VR256,
- int_x86_avx2_vbroadcast_sd_pd_256,
- WriteFShuffle256>, VEX_L;
+def VBROADCASTSDYrr : avx2_broadcast_rr<0x19, "vbroadcastsd", VR256,
+ v4f64, v2f64, WriteFShuffle256>, VEX_L;
+
+let mayLoad = 1, Predicates = [HasAVX2] in
+def VBROADCASTI128 : AVX8I<0x5A, MRMSrcMem, (outs VR256:$dst),
+ (ins i128mem:$src),
+ "vbroadcasti128\t{$src, $dst|$dst, $src}", []>,
+ Sched<[WriteLoad]>, VEX, VEX_L;
+
+def VBROADCASTF128 : AVX8I<0x1A, MRMSrcMem, (outs VR256:$dst),
+ (ins f128mem:$src),
+ "vbroadcastf128\t{$src, $dst|$dst, $src}",
+ [(set VR256:$dst,
+ (int_x86_avx_vbroadcastf128_pd_256 addr:$src))]>,
+ Sched<[WriteFShuffleLd]>, VEX, VEX_L;
let Predicates = [HasAVX] in
def : Pat<(int_x86_avx_vbroadcastf128_ps_256 addr:$src),
(VCVTPH2PSrm addr:$src)>;
def : Pat<(int_x86_vcvtph2ps_128 (vzload_v2i64 addr:$src)),
(VCVTPH2PSrm addr:$src)>;
+
+ def : Pat<(store (f64 (vector_extract (bc_v2f64 (v8i16
+ (int_x86_vcvtps2ph_128 VR128:$src1, i32:$src2))), (iPTR 0))),
+ addr:$dst),
+ (VCVTPS2PHmr addr:$dst, VR128:$src1, imm:$src2)>;
+ def : Pat<(store (i64 (vector_extract (bc_v2i64 (v8i16
+ (int_x86_vcvtps2ph_128 VR128:$src1, i32:$src2))), (iPTR 0))),
+ addr:$dst),
+ (VCVTPS2PHmr addr:$dst, VR128:$src1, imm:$src2)>;
+ def : Pat<(store (v8i16 (int_x86_vcvtps2ph_256 VR256:$src1, i32:$src2)),
+ addr:$dst),
+ (VCVTPS2PHYmr addr:$dst, VR256:$src1, imm:$src2)>;
}
// Patterns for matching conversions from float to half-float and vice versa.
//
multiclass avx2_broadcast<bits<8> opc, string OpcodeStr,
X86MemOperand x86memop, PatFrag ld_frag,
- Intrinsic Int128, Intrinsic Int256> {
+ ValueType OpVT128, ValueType OpVT256> {
def rr : AVX28I<opc, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
!strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"),
- [(set VR128:$dst, (Int128 VR128:$src))]>,
+ [(set VR128:$dst, (OpVT128 (X86VBroadcast (OpVT128 VR128:$src))))]>,
Sched<[WriteShuffle]>, VEX;
def rm : AVX28I<opc, MRMSrcMem, (outs VR128:$dst), (ins x86memop:$src),
!strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"),
- [(set VR128:$dst,
- (Int128 (scalar_to_vector (ld_frag addr:$src))))]>,
+ [(set VR128:$dst, (OpVT128 (X86VBroadcast (ld_frag addr:$src))))]>,
Sched<[WriteLoad]>, VEX;
def Yrr : AVX28I<opc, MRMSrcReg, (outs VR256:$dst), (ins VR128:$src),
!strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"),
- [(set VR256:$dst, (Int256 VR128:$src))]>,
+ [(set VR256:$dst, (OpVT256 (X86VBroadcast (OpVT128 VR128:$src))))]>,
Sched<[WriteShuffle256]>, VEX, VEX_L;
def Yrm : AVX28I<opc, MRMSrcMem, (outs VR256:$dst), (ins x86memop:$src),
!strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"),
- [(set VR256:$dst,
- (Int256 (scalar_to_vector (ld_frag addr:$src))))]>,
+ [(set VR256:$dst, (OpVT256 (X86VBroadcast (ld_frag addr:$src))))]>,
Sched<[WriteLoad]>, VEX, VEX_L;
}
-defm VPBROADCASTB : avx2_broadcast<0x78, "vpbroadcastb", i8mem, loadi8,
- int_x86_avx2_pbroadcastb_128,
- int_x86_avx2_pbroadcastb_256>;
-defm VPBROADCASTW : avx2_broadcast<0x79, "vpbroadcastw", i16mem, loadi16,
- int_x86_avx2_pbroadcastw_128,
- int_x86_avx2_pbroadcastw_256>;
-defm VPBROADCASTD : avx2_broadcast<0x58, "vpbroadcastd", i32mem, loadi32,
- int_x86_avx2_pbroadcastd_128,
- int_x86_avx2_pbroadcastd_256>;
-defm VPBROADCASTQ : avx2_broadcast<0x59, "vpbroadcastq", i64mem, loadi64,
- int_x86_avx2_pbroadcastq_128,
- int_x86_avx2_pbroadcastq_256>;
+defm VPBROADCASTB : avx2_broadcast<0x78, "vpbroadcastb", i8mem, loadi8, v16i8, v32i8>;
+defm VPBROADCASTW : avx2_broadcast<0x79, "vpbroadcastw", i16mem, loadi16, v8i16, v16i16>;
+defm VPBROADCASTD : avx2_broadcast<0x58, "vpbroadcastd", i32mem, loadi32, v4i32, v8i32>;
+defm VPBROADCASTQ : avx2_broadcast<0x59, "vpbroadcastq", i64mem, loadi64, v2i64, v4i64>;
let Predicates = [HasAVX2] in {
- def : Pat<(v16i8 (X86VBroadcast (loadi8 addr:$src))),
- (VPBROADCASTBrm addr:$src)>;
- def : Pat<(v32i8 (X86VBroadcast (loadi8 addr:$src))),
- (VPBROADCASTBYrm addr:$src)>;
- def : Pat<(v8i16 (X86VBroadcast (loadi16 addr:$src))),
- (VPBROADCASTWrm addr:$src)>;
- def : Pat<(v16i16 (X86VBroadcast (loadi16 addr:$src))),
- (VPBROADCASTWYrm addr:$src)>;
- def : Pat<(v4i32 (X86VBroadcast (loadi32 addr:$src))),
- (VPBROADCASTDrm addr:$src)>;
- def : Pat<(v8i32 (X86VBroadcast (loadi32 addr:$src))),
- (VPBROADCASTDYrm addr:$src)>;
- def : Pat<(v2i64 (X86VBroadcast (loadi64 addr:$src))),
- (VPBROADCASTQrm addr:$src)>;
- def : Pat<(v4i64 (X86VBroadcast (loadi64 addr:$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:$src)>;
-
- // Provide aliases for broadcast from the same regitser class that
+ // loadi16 is tricky to fold, because !isTypeDesirableForOp, justifiably.
+ // This means we'll encounter truncated i32 loads; match that here.
+ def : Pat<(v8i16 (X86VBroadcast (i16 (trunc (i32 (load addr:$src)))))),
+ (VPBROADCASTWrm addr:$src)>;
+ def : Pat<(v16i16 (X86VBroadcast (i16 (trunc (i32 (load addr:$src)))))),
+ (VPBROADCASTWYrm addr:$src)>;
+
+ // Provide aliases for broadcast from the same register class that
// automatically does the extract.
def : Pat<(v32i8 (X86VBroadcast (v32i8 VR256:$src))),
(VPBROADCASTBYrr (v16i8 (EXTRACT_SUBREG (v32i8 VR256:$src),
def : Pat<(v2f64 (X86VBroadcast f64:$src)),
(VMOVDDUPrr (COPY_TO_REGCLASS FR64:$src, VR128))>;
+ def : Pat<(v2i64 (X86VBroadcast i64:$src)),
+ (VMOVDDUPrr (COPY_TO_REGCLASS GR64:$src, VR128))>;
}
//===----------------------------------------------------------------------===//
//
def VEXTRACTI128rr : AVX2AIi8<0x39, MRMDestReg, (outs VR128:$dst),
(ins VR256:$src1, u8imm:$src2),
- "vextracti128\t{$src2, $src1, $dst|$dst, $src1, $src2}",
- [(set VR128:$dst,
- (int_x86_avx2_vextracti128 VR256:$src1, imm:$src2))]>,
+ "vextracti128\t{$src2, $src1, $dst|$dst, $src1, $src2}", []>,
Sched<[WriteShuffle256]>, VEX, VEX_L;
let hasSideEffects = 0, mayStore = 1 in
def VEXTRACTI128mr : AVX2AIi8<0x39, MRMDestMem, (outs),
int_x86_avx2_maskstore_q,
int_x86_avx2_maskstore_q_256>, VEX_W;
-def: Pat<(masked_store addr:$ptr, (v8i32 VR256:$mask), (v8f32 VR256:$src)),
+def: Pat<(X86mstore addr:$ptr, (v8i32 VR256:$mask), (v8f32 VR256:$src)),
(VMASKMOVPSYmr addr:$ptr, VR256:$mask, VR256:$src)>;
-def: Pat<(masked_store addr:$ptr, (v8i32 VR256:$mask), (v8i32 VR256:$src)),
+def: Pat<(X86mstore addr:$ptr, (v8i32 VR256:$mask), (v8i32 VR256:$src)),
(VPMASKMOVDYmr addr:$ptr, VR256:$mask, VR256:$src)>;
-def: Pat<(masked_store addr:$ptr, (v4i32 VR128:$mask), (v4f32 VR128:$src)),
+def: Pat<(X86mstore addr:$ptr, (v4i32 VR128:$mask), (v4f32 VR128:$src)),
(VMASKMOVPSmr addr:$ptr, VR128:$mask, VR128:$src)>;
-def: Pat<(masked_store addr:$ptr, (v4i32 VR128:$mask), (v4i32 VR128:$src)),
+def: Pat<(X86mstore addr:$ptr, (v4i32 VR128:$mask), (v4i32 VR128:$src)),
(VPMASKMOVDmr addr:$ptr, VR128:$mask, VR128:$src)>;
def: Pat<(v8f32 (masked_load addr:$ptr, (v8i32 VR256:$mask), undef)),
(VBLENDVPSrr VR128:$src0, (VPMASKMOVDrm VR128:$mask, addr:$ptr),
VR128:$mask)>;
-def: Pat<(masked_store addr:$ptr, (v4i64 VR256:$mask), (v4f64 VR256:$src)),
+def: Pat<(X86mstore addr:$ptr, (v4i64 VR256:$mask), (v4f64 VR256:$src)),
(VMASKMOVPDYmr addr:$ptr, VR256:$mask, VR256:$src)>;
-def: Pat<(masked_store addr:$ptr, (v4i64 VR256:$mask), (v4i64 VR256:$src)),
+def: Pat<(X86mstore addr:$ptr, (v4i64 VR256:$mask), (v4i64 VR256:$src)),
(VPMASKMOVQYmr addr:$ptr, VR256:$mask, VR256:$src)>;
def: Pat<(v4f64 (masked_load addr:$ptr, (v4i64 VR256:$mask), undef)),
(VBLENDVPDYrr VR256:$src0, (VPMASKMOVQYrm VR256:$mask, addr:$ptr),
VR256:$mask)>;
-def: Pat<(masked_store addr:$ptr, (v2i64 VR128:$mask), (v2f64 VR128:$src)),
+def: Pat<(X86mstore addr:$ptr, (v2i64 VR128:$mask), (v2f64 VR128:$src)),
(VMASKMOVPDmr addr:$ptr, VR128:$mask, VR128:$src)>;
-def: Pat<(masked_store addr:$ptr, (v2i64 VR128:$mask), (v2i64 VR128:$src)),
+def: Pat<(X86mstore addr:$ptr, (v2i64 VR128:$mask), (v2i64 VR128:$src)),
(VPMASKMOVQmr addr:$ptr, VR128:$mask, VR128:$src)>;
def: Pat<(v2f64 (masked_load addr:$ptr, (v2i64 VR128:$mask), undef)),
projects / oota-llvm.git / blobdiff