multiclass sse12_fp_packed_logical_rm<bits<8> opc, RegisterClass RC, Domain d,
string OpcodeStr, X86MemOperand x86memop,
list<dag> pat_rr, list<dag> pat_rm,
- bit Is2Addr = 1,
- bit rr_hasSideEffects = 0> {
- let isCommutable = 1, neverHasSideEffects = rr_hasSideEffects in
+ bit Is2Addr = 1> {
+ let isCommutable = 1, hasSideEffects = 0 in
def rr : PI<opc, MRMSrcReg, (outs RC:$dst), (ins RC:$src1, RC:$src2),
!if(Is2Addr,
!strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"),
pat_rm, IIC_DEFAULT, d>;
}
-/// sse12_fp_packed_int - SSE 1 & 2 packed instructions intrinsics class
-multiclass sse12_fp_packed_int<bits<8> opc, string OpcodeStr, RegisterClass RC,
- string asm, string SSEVer, string FPSizeStr,
- X86MemOperand x86memop, PatFrag mem_frag,
- Domain d, OpndItins itins, bit Is2Addr = 1> {
- def rr_Int : PI<opc, MRMSrcReg, (outs RC:$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_", SSEVer, "_", OpcodeStr, FPSizeStr))
- RC:$src1, RC:$src2))], IIC_DEFAULT, d>;
- def rm_Int : PI<opc, MRMSrcMem, (outs RC:$dst), (ins RC:$src1,x86memop:$src2),
- !if(Is2Addr,
- !strconcat(asm, "\t{$src2, $dst|$dst, $src2}"),
- !strconcat(asm, "\t{$src2, $src1, $dst|$dst, $src1, $src2}")),
- [(set RC:$dst, (!cast<Intrinsic>(
- !strconcat("int_x86_", SSEVer, "_", OpcodeStr, FPSizeStr))
- RC:$src1, (mem_frag addr:$src2)))], IIC_DEFAULT, d>;
-}
-
//===----------------------------------------------------------------------===//
// Non-instruction patterns
//===----------------------------------------------------------------------===//
// A 128-bit subvector extract from the first 256-bit vector position
// is a subregister copy that needs no instruction.
-def : Pat<(v4i32 (extract_subvector (v8i32 VR256:$src), (i32 0))),
+def : Pat<(v4i32 (extract_subvector (v8i32 VR256:$src), (iPTR 0))),
(v4i32 (EXTRACT_SUBREG (v8i32 VR256:$src), sub_xmm))>;
-def : Pat<(v4f32 (extract_subvector (v8f32 VR256:$src), (i32 0))),
+def : Pat<(v4f32 (extract_subvector (v8f32 VR256:$src), (iPTR 0))),
(v4f32 (EXTRACT_SUBREG (v8f32 VR256:$src), sub_xmm))>;
-def : Pat<(v2i64 (extract_subvector (v4i64 VR256:$src), (i32 0))),
+def : Pat<(v2i64 (extract_subvector (v4i64 VR256:$src), (iPTR 0))),
(v2i64 (EXTRACT_SUBREG (v4i64 VR256:$src), sub_xmm))>;
-def : Pat<(v2f64 (extract_subvector (v4f64 VR256:$src), (i32 0))),
+def : Pat<(v2f64 (extract_subvector (v4f64 VR256:$src), (iPTR 0))),
(v2f64 (EXTRACT_SUBREG (v4f64 VR256:$src), sub_xmm))>;
-def : Pat<(v8i16 (extract_subvector (v16i16 VR256:$src), (i32 0))),
+def : Pat<(v8i16 (extract_subvector (v16i16 VR256:$src), (iPTR 0))),
(v8i16 (EXTRACT_SUBREG (v16i16 VR256:$src), sub_xmm))>;
-def : Pat<(v16i8 (extract_subvector (v32i8 VR256:$src), (i32 0))),
+def : Pat<(v16i8 (extract_subvector (v32i8 VR256:$src), (iPTR 0))),
(v16i8 (EXTRACT_SUBREG (v32i8 VR256:$src), sub_xmm))>;
// A 128-bit subvector insert to the first 256-bit vector position
// is a subregister copy that needs no instruction.
-def : Pat<(insert_subvector undef, (v2i64 VR128:$src), (i32 0)),
+let AddedComplexity = 25 in { // to give priority over vinsertf128rm
+def : Pat<(insert_subvector undef, (v2i64 VR128:$src), (iPTR 0)),
(INSERT_SUBREG (v4i64 (IMPLICIT_DEF)), VR128:$src, sub_xmm)>;
-def : Pat<(insert_subvector undef, (v2f64 VR128:$src), (i32 0)),
+def : Pat<(insert_subvector undef, (v2f64 VR128:$src), (iPTR 0)),
(INSERT_SUBREG (v4f64 (IMPLICIT_DEF)), VR128:$src, sub_xmm)>;
-def : Pat<(insert_subvector undef, (v4i32 VR128:$src), (i32 0)),
+def : Pat<(insert_subvector undef, (v4i32 VR128:$src), (iPTR 0)),
(INSERT_SUBREG (v8i32 (IMPLICIT_DEF)), VR128:$src, sub_xmm)>;
-def : Pat<(insert_subvector undef, (v4f32 VR128:$src), (i32 0)),
+def : Pat<(insert_subvector undef, (v4f32 VR128:$src), (iPTR 0)),
(INSERT_SUBREG (v8f32 (IMPLICIT_DEF)), VR128:$src, sub_xmm)>;
-def : Pat<(insert_subvector undef, (v8i16 VR128:$src), (i32 0)),
+def : Pat<(insert_subvector undef, (v8i16 VR128:$src), (iPTR 0)),
(INSERT_SUBREG (v16i16 (IMPLICIT_DEF)), VR128:$src, sub_xmm)>;
-def : Pat<(insert_subvector undef, (v16i8 VR128:$src), (i32 0)),
+def : Pat<(insert_subvector undef, (v16i8 VR128:$src), (iPTR 0)),
(INSERT_SUBREG (v32i8 (IMPLICIT_DEF)), VR128:$src, sub_xmm)>;
+}
// Implicitly promote a 32-bit scalar to a vector.
def : Pat<(v4f32 (scalar_to_vector FR32:$src)),
def : Pat<(v16i16 (bitconvert (v32i8 VR256:$src))), (v16i16 VR256:$src)>;
}
-// Alias instructions that map fld0 to pxor for sse.
+// Alias instructions that map fld0 to xorps for sse or vxorps for avx.
// This is expanded by ExpandPostRAPseudos.
let isReMaterializable = 1, isAsCheapAsAMove = 1, canFoldAsLoad = 1,
isPseudo = 1 in {
// We set canFoldAsLoad because this can be converted to a constant-pool
// load of an all-zeros value if folding it would be beneficial.
let isReMaterializable = 1, isAsCheapAsAMove = 1, canFoldAsLoad = 1,
- isPseudo = 1, neverHasSideEffects = 1 in {
-def V_SET0 : I<0, Pseudo, (outs VR128:$dst), (ins), "", []>;
+ isPseudo = 1 in {
+def V_SET0 : I<0, Pseudo, (outs VR128:$dst), (ins), "",
+ [(set VR128:$dst, (v4f32 immAllZerosV))]>;
}
-def : Pat<(v4f32 immAllZerosV), (V_SET0)>;
def : Pat<(v2f64 immAllZerosV), (V_SET0)>;
def : Pat<(v4i32 immAllZerosV), (V_SET0)>;
def : Pat<(v2i64 immAllZerosV), (V_SET0)>;
def : Pat<(v16i8 immAllZerosV), (V_SET0)>;
-// The same as done above but for AVX. The 256-bit ISA does not support PI,
+// The same as done above but for AVX. The 256-bit AVX1 ISA doesn't support PI,
// and doesn't need it because on sandy bridge the register is set to zero
// at the rename stage without using any execution unit, so SET0PSY
// and SET0PDY can be used for vector int instructions without penalty
-// FIXME: Change encoding to pseudo! This is blocked right now by the x86
-// JIT implementatioan, it does not expand the instructions below like
-// X86MCInstLower does.
let isReMaterializable = 1, isAsCheapAsAMove = 1, canFoldAsLoad = 1,
- isCodeGenOnly = 1 in {
-let Predicates = [HasAVX] in {
-def AVX_SET0PSY : PSI<0x57, MRMInitReg, (outs VR256:$dst), (ins), "",
- [(set VR256:$dst, (v8f32 immAllZerosV))]>, VEX_4V;
-def AVX_SET0PDY : PDI<0x57, MRMInitReg, (outs VR256:$dst), (ins), "",
- [(set VR256:$dst, (v4f64 immAllZerosV))]>, VEX_4V;
-}
-let Predicates = [HasAVX2], neverHasSideEffects = 1 in
-def AVX2_SET0 : PDI<0xef, MRMInitReg, (outs VR256:$dst), (ins), "",
- []>, VEX_4V;
+ isPseudo = 1, Predicates = [HasAVX] in {
+def AVX_SET0 : I<0, Pseudo, (outs VR256:$dst), (ins), "",
+ [(set VR256:$dst, (v8f32 immAllZerosV))]>;
}
-let Predicates = [HasAVX2], AddedComplexity = 5 in {
- def : Pat<(v4i64 immAllZerosV), (AVX2_SET0)>;
- def : Pat<(v8i32 immAllZerosV), (AVX2_SET0)>;
- def : Pat<(v16i16 immAllZerosV), (AVX2_SET0)>;
- def : Pat<(v32i8 immAllZerosV), (AVX2_SET0)>;
+let Predicates = [HasAVX] in
+ def : Pat<(v4f64 immAllZerosV), (AVX_SET0)>;
+
+let Predicates = [HasAVX2] in {
+ def : Pat<(v4i64 immAllZerosV), (AVX_SET0)>;
+ def : Pat<(v8i32 immAllZerosV), (AVX_SET0)>;
+ def : Pat<(v16i16 immAllZerosV), (AVX_SET0)>;
+ def : Pat<(v32i8 immAllZerosV), (AVX_SET0)>;
}
-// AVX has no support for 256-bit integer instructions, but since the 128-bit
+// AVX1 has no support for 256-bit integer instructions, but since the 128-bit
// VPXOR instruction writes zero to its upper part, it's safe build zeros.
+let Predicates = [HasAVX1Only] in {
def : Pat<(v32i8 immAllZerosV), (SUBREG_TO_REG (i8 0), (V_SET0), sub_xmm)>;
def : Pat<(bc_v32i8 (v8f32 immAllZerosV)),
(SUBREG_TO_REG (i8 0), (V_SET0), sub_xmm)>;
def : Pat<(v4i64 immAllZerosV), (SUBREG_TO_REG (i64 0), (V_SET0), sub_xmm)>;
def : Pat<(bc_v4i64 (v8f32 immAllZerosV)),
(SUBREG_TO_REG (i64 0), (V_SET0), sub_xmm)>;
+}
// We set canFoldAsLoad because this can be converted to a constant-pool
// load of an all-ones value if folding it would be beneficial.
-// FIXME: Change encoding to pseudo! This is blocked right now by the x86
-// JIT implementation, it does not expand the instructions below like
-// X86MCInstLower does.
let isReMaterializable = 1, isAsCheapAsAMove = 1, canFoldAsLoad = 1,
- isCodeGenOnly = 1, ExeDomain = SSEPackedInt in {
- let Predicates = [HasAVX] in
- def AVX_SETALLONES : PDI<0x76, MRMInitReg, (outs VR128:$dst), (ins), "",
- [(set VR128:$dst, (v4i32 immAllOnesV))]>, VEX_4V;
- def V_SETALLONES : PDI<0x76, MRMInitReg, (outs VR128:$dst), (ins), "",
- [(set VR128:$dst, (v4i32 immAllOnesV))]>;
+ isPseudo = 1 in {
+ def V_SETALLONES : I<0, Pseudo, (outs VR128:$dst), (ins), "",
+ [(set VR128:$dst, (v4i32 immAllOnesV))]>;
let Predicates = [HasAVX2] in
- def AVX2_SETALLONES : PDI<0x76, MRMInitReg, (outs VR256:$dst), (ins), "",
- [(set VR256:$dst, (v8i32 immAllOnesV))]>, VEX_4V;
+ def AVX2_SETALLONES : I<0, Pseudo, (outs VR256:$dst), (ins), "",
+ [(set VR256:$dst, (v8i32 immAllOnesV))]>;
}
// in terms of a copy, and just mentioned, we don't use movss/movsd for copies.
//===----------------------------------------------------------------------===//
-class sse12_move_rr<RegisterClass RC, SDNode OpNode, ValueType vt, string asm> :
- SI<0x10, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src1, RC:$src2), asm,
- [(set VR128:$dst, (vt (OpNode VR128:$src1,
- (scalar_to_vector RC:$src2))))],
- IIC_SSE_MOV_S_RR>;
+multiclass sse12_move_rr<RegisterClass RC, SDNode OpNode, ValueType vt,
+ X86MemOperand x86memop, string base_opc,
+ string asm_opr> {
+ def rr : SI<0x10, MRMSrcReg, (outs VR128:$dst),
+ (ins VR128:$src1, RC:$src2),
+ !strconcat(base_opc, asm_opr),
+ [(set VR128:$dst, (vt (OpNode VR128:$src1,
+ (scalar_to_vector RC:$src2))))],
+ IIC_SSE_MOV_S_RR>;
-// Loading from memory automatically zeroing upper bits.
-class sse12_move_rm<RegisterClass RC, X86MemOperand x86memop,
- PatFrag mem_pat, string OpcodeStr> :
- SI<0x10, MRMSrcMem, (outs RC:$dst), (ins x86memop:$src),
- !strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"),
- [(set RC:$dst, (mem_pat addr:$src))],
- IIC_SSE_MOV_S_RM>;
-
-// AVX
-def VMOVSSrr : sse12_move_rr<FR32, X86Movss, v4f32,
- "movss\t{$src2, $src1, $dst|$dst, $src1, $src2}">, XS, VEX_4V,
- VEX_LIG;
-def VMOVSDrr : sse12_move_rr<FR64, X86Movsd, v2f64,
- "movsd\t{$src2, $src1, $dst|$dst, $src1, $src2}">, XD, VEX_4V,
- VEX_LIG;
-
-// For the disassembler
-let isCodeGenOnly = 1 in {
- def VMOVSSrr_REV : SI<0x11, MRMDestReg, (outs VR128:$dst),
- (ins VR128:$src1, FR32:$src2),
- "movss\t{$src2, $src1, $dst|$dst, $src1, $src2}", [],
- IIC_SSE_MOV_S_RR>,
- XS, VEX_4V, VEX_LIG;
- def VMOVSDrr_REV : SI<0x11, MRMDestReg, (outs VR128:$dst),
- (ins VR128:$src1, FR64:$src2),
- "movsd\t{$src2, $src1, $dst|$dst, $src1, $src2}", [],
- IIC_SSE_MOV_S_RR>,
- XD, VEX_4V, VEX_LIG;
+ // For the disassembler
+ let isCodeGenOnly = 1, hasSideEffects = 0 in
+ def rr_REV : SI<0x11, MRMDestReg, (outs VR128:$dst),
+ (ins VR128:$src1, RC:$src2),
+ !strconcat(base_opc, asm_opr),
+ [], IIC_SSE_MOV_S_RR>;
}
-let canFoldAsLoad = 1, isReMaterializable = 1 in {
- def VMOVSSrm : sse12_move_rm<FR32, f32mem, loadf32, "movss">, XS, VEX,
- VEX_LIG;
- let AddedComplexity = 20 in
- def VMOVSDrm : sse12_move_rm<FR64, f64mem, loadf64, "movsd">, XD, VEX,
- VEX_LIG;
-}
+multiclass sse12_move<RegisterClass RC, SDNode OpNode, ValueType vt,
+ X86MemOperand x86memop, string OpcodeStr> {
+ // AVX
+ defm V#NAME : sse12_move_rr<RC, OpNode, vt, x86memop, OpcodeStr,
+ "\t{$src2, $src1, $dst|$dst, $src1, $src2}">,
+ VEX_4V, VEX_LIG;
-def VMOVSSmr : SI<0x11, MRMDestMem, (outs), (ins f32mem:$dst, FR32:$src),
- "movss\t{$src, $dst|$dst, $src}",
- [(store FR32:$src, addr:$dst)], IIC_SSE_MOV_S_MR>,
- XS, VEX, VEX_LIG;
-def VMOVSDmr : SI<0x11, MRMDestMem, (outs), (ins f64mem:$dst, FR64:$src),
- "movsd\t{$src, $dst|$dst, $src}",
- [(store FR64:$src, addr:$dst)], IIC_SSE_MOV_S_MR>,
- XD, VEX, VEX_LIG;
+ def V#NAME#mr : SI<0x11, MRMDestMem, (outs), (ins x86memop:$dst, RC:$src),
+ !strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"),
+ [(store RC:$src, addr:$dst)], IIC_SSE_MOV_S_MR>,
+ VEX, VEX_LIG;
+ // SSE1 & 2
+ let Constraints = "$src1 = $dst" in {
+ defm NAME : sse12_move_rr<RC, OpNode, vt, x86memop, OpcodeStr,
+ "\t{$src2, $dst|$dst, $src2}">;
+ }
-// SSE1 & 2
-let Constraints = "$src1 = $dst" in {
- def MOVSSrr : sse12_move_rr<FR32, X86Movss, v4f32,
- "movss\t{$src2, $dst|$dst, $src2}">, XS;
- def MOVSDrr : sse12_move_rr<FR64, X86Movsd, v2f64,
- "movsd\t{$src2, $dst|$dst, $src2}">, XD;
+ def NAME#mr : SI<0x11, MRMDestMem, (outs), (ins x86memop:$dst, RC:$src),
+ !strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"),
+ [(store RC:$src, addr:$dst)], IIC_SSE_MOV_S_MR>;
+}
- // For the disassembler
- let isCodeGenOnly = 1 in {
- def MOVSSrr_REV : SI<0x11, MRMDestReg, (outs VR128:$dst),
- (ins VR128:$src1, FR32:$src2),
- "movss\t{$src2, $dst|$dst, $src2}", [],
- IIC_SSE_MOV_S_RR>, XS;
- def MOVSDrr_REV : SI<0x11, MRMDestReg, (outs VR128:$dst),
- (ins VR128:$src1, FR64:$src2),
- "movsd\t{$src2, $dst|$dst, $src2}", [],
- IIC_SSE_MOV_S_RR>, XD;
- }
+// Loading from memory automatically zeroing upper bits.
+multiclass sse12_move_rm<RegisterClass RC, X86MemOperand x86memop,
+ PatFrag mem_pat, string OpcodeStr> {
+ def V#NAME#rm : SI<0x10, MRMSrcMem, (outs RC:$dst), (ins x86memop:$src),
+ !strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"),
+ [(set RC:$dst, (mem_pat addr:$src))],
+ IIC_SSE_MOV_S_RM>, VEX, VEX_LIG;
+ def NAME#rm : SI<0x10, MRMSrcMem, (outs RC:$dst), (ins x86memop:$src),
+ !strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"),
+ [(set RC:$dst, (mem_pat addr:$src))],
+ IIC_SSE_MOV_S_RM>;
}
+defm MOVSS : sse12_move<FR32, X86Movss, v4f32, f32mem, "movss">, XS;
+defm MOVSD : sse12_move<FR64, X86Movsd, v2f64, f64mem, "movsd">, XD;
+
let canFoldAsLoad = 1, isReMaterializable = 1 in {
- def MOVSSrm : sse12_move_rm<FR32, f32mem, loadf32, "movss">, XS;
+ defm MOVSS : sse12_move_rm<FR32, f32mem, loadf32, "movss">, XS;
let AddedComplexity = 20 in
- def MOVSDrm : sse12_move_rm<FR64, f64mem, loadf64, "movsd">, XD;
+ defm MOVSD : sse12_move_rm<FR64, f64mem, loadf64, "movsd">, XD;
}
-def MOVSSmr : SSI<0x11, MRMDestMem, (outs), (ins f32mem:$dst, FR32:$src),
- "movss\t{$src, $dst|$dst, $src}",
- [(store FR32:$src, addr:$dst)], IIC_SSE_MOV_S_MR>;
-def MOVSDmr : SDI<0x11, MRMDestMem, (outs), (ins f64mem:$dst, FR64:$src),
- "movsd\t{$src, $dst|$dst, $src}",
- [(store FR64:$src, addr:$dst)], IIC_SSE_MOV_S_MR>;
-
// Patterns
let Predicates = [HasAVX] in {
let AddedComplexity = 15 in {
// Represent the same patterns above but in the form they appear for
// 256-bit types
def : Pat<(v8i32 (X86vzmovl (insert_subvector undef,
- (v4i32 (scalar_to_vector (loadi32 addr:$src))), (i32 0)))),
+ (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:$src))), (i32 0)))),
+ (v4f32 (scalar_to_vector (loadf32 addr:$src))), (iPTR 0)))),
(SUBREG_TO_REG (i32 0), (VMOVSSrm addr:$src), sub_xmm)>;
def : Pat<(v4f64 (X86vzmovl (insert_subvector undef,
- (v2f64 (scalar_to_vector (loadf64 addr:$src))), (i32 0)))),
+ (v2f64 (scalar_to_vector (loadf64 addr:$src))), (iPTR 0)))),
(SUBREG_TO_REG (i32 0), (VMOVSDrm addr:$src), sub_xmm)>;
}
def : Pat<(v8f32 (X86vzmovl (insert_subvector undef,
- (v4f32 (scalar_to_vector FR32:$src)), (i32 0)))),
+ (v4f32 (scalar_to_vector FR32:$src)), (iPTR 0)))),
(SUBREG_TO_REG (i32 0),
(v4f32 (VMOVSSrr (v4f32 (V_SET0)), FR32:$src)),
sub_xmm)>;
def : Pat<(v4f64 (X86vzmovl (insert_subvector undef,
- (v2f64 (scalar_to_vector FR64:$src)), (i32 0)))),
+ (v2f64 (scalar_to_vector FR64:$src)), (iPTR 0)))),
(SUBREG_TO_REG (i64 0),
(v2f64 (VMOVSDrr (v2f64 (V_SET0)), FR64:$src)),
sub_xmm)>;
def : Pat<(v4i64 (X86vzmovl (insert_subvector undef,
- (v2i64 (scalar_to_vector (loadi64 addr:$src))), (i32 0)))),
+ (v2i64 (scalar_to_vector (loadi64 addr:$src))), (iPTR 0)))),
(SUBREG_TO_REG (i64 0), (VMOVSDrm addr:$src), sub_xmm)>;
// Move low f64 and clear high bits.
(VMOVSDrr VR128:$src1, (COPY_TO_REGCLASS VR128:$src2, FR64))>;
}
-let Predicates = [HasSSE1] in {
+let Predicates = [UseSSE1] in {
let AddedComplexity = 15 in {
// Move scalar to XMM zero-extended, zeroing a VR128 then do a
// MOVSS to the lower bits.
(MOVSSrr VR128:$src1, (COPY_TO_REGCLASS VR128:$src2, FR32))>;
}
-let Predicates = [HasSSE2] in {
+let Predicates = [UseSSE2] in {
let AddedComplexity = 15 in {
// Move scalar to XMM zero-extended, zeroing a VR128 then do a
// MOVSD to the lower bits.
defm VMOVAPSY : sse12_mov_packed<0x28, VR256, f256mem, alignedloadv8f32,
"movaps", SSEPackedSingle, SSE_MOVA_ITINS>,
- TB, VEX;
+ TB, VEX, VEX_L;
defm VMOVAPDY : sse12_mov_packed<0x28, VR256, f256mem, alignedloadv4f64,
"movapd", SSEPackedDouble, SSE_MOVA_ITINS>,
- TB, OpSize, VEX;
+ TB, OpSize, VEX, VEX_L;
defm VMOVUPSY : sse12_mov_packed<0x10, VR256, f256mem, loadv8f32,
"movups", SSEPackedSingle, SSE_MOVU_ITINS>,
- TB, VEX;
+ TB, VEX, VEX_L;
defm VMOVUPDY : sse12_mov_packed<0x10, VR256, f256mem, loadv4f64,
"movupd", SSEPackedDouble, SSE_MOVU_ITINS, 0>,
- TB, OpSize, VEX;
+ TB, OpSize, VEX, VEX_L;
defm MOVAPS : sse12_mov_packed<0x28, VR128, f128mem, alignedloadv4f32,
"movaps", SSEPackedSingle, SSE_MOVA_ITINS>,
TB;
def VMOVAPSYmr : VPSI<0x29, MRMDestMem, (outs), (ins f256mem:$dst, VR256:$src),
"movaps\t{$src, $dst|$dst, $src}",
[(alignedstore256 (v8f32 VR256:$src), addr:$dst)],
- IIC_SSE_MOVA_P_MR>, VEX;
+ IIC_SSE_MOVA_P_MR>, VEX, VEX_L;
def VMOVAPDYmr : VPDI<0x29, MRMDestMem, (outs), (ins f256mem:$dst, VR256:$src),
"movapd\t{$src, $dst|$dst, $src}",
[(alignedstore256 (v4f64 VR256:$src), addr:$dst)],
- IIC_SSE_MOVA_P_MR>, VEX;
+ IIC_SSE_MOVA_P_MR>, VEX, VEX_L;
def VMOVUPSYmr : VPSI<0x11, MRMDestMem, (outs), (ins f256mem:$dst, VR256:$src),
"movups\t{$src, $dst|$dst, $src}",
[(store (v8f32 VR256:$src), addr:$dst)],
- IIC_SSE_MOVU_P_MR>, VEX;
+ IIC_SSE_MOVU_P_MR>, VEX, VEX_L;
def VMOVUPDYmr : VPDI<0x11, MRMDestMem, (outs), (ins f256mem:$dst, VR256:$src),
"movupd\t{$src, $dst|$dst, $src}",
[(store (v4f64 VR256:$src), addr:$dst)],
- IIC_SSE_MOVU_P_MR>, VEX;
+ IIC_SSE_MOVU_P_MR>, VEX, VEX_L;
// For disassembler
-let isCodeGenOnly = 1 in {
+let isCodeGenOnly = 1, hasSideEffects = 0 in {
def VMOVAPSrr_REV : VPSI<0x29, MRMDestReg, (outs VR128:$dst),
(ins VR128:$src),
"movaps\t{$src, $dst|$dst, $src}", [],
def VMOVAPSYrr_REV : VPSI<0x29, MRMDestReg, (outs VR256:$dst),
(ins VR256:$src),
"movaps\t{$src, $dst|$dst, $src}", [],
- IIC_SSE_MOVA_P_RR>, VEX;
+ IIC_SSE_MOVA_P_RR>, VEX, VEX_L;
def VMOVAPDYrr_REV : VPDI<0x29, MRMDestReg, (outs VR256:$dst),
(ins VR256:$src),
"movapd\t{$src, $dst|$dst, $src}", [],
- IIC_SSE_MOVA_P_RR>, VEX;
+ IIC_SSE_MOVA_P_RR>, VEX, VEX_L;
def VMOVUPSYrr_REV : VPSI<0x11, MRMDestReg, (outs VR256:$dst),
(ins VR256:$src),
"movups\t{$src, $dst|$dst, $src}", [],
- IIC_SSE_MOVU_P_RR>, VEX;
+ IIC_SSE_MOVU_P_RR>, VEX, VEX_L;
def VMOVUPDYrr_REV : VPDI<0x11, MRMDestReg, (outs VR256:$dst),
(ins VR256:$src),
"movupd\t{$src, $dst|$dst, $src}", [],
- IIC_SSE_MOVU_P_RR>, VEX;
+ IIC_SSE_MOVU_P_RR>, VEX, VEX_L;
}
let Predicates = [HasAVX] in {
def : Pat<(v8i32 (X86vzmovl
- (insert_subvector undef, (v4i32 VR128:$src), (i32 0)))),
+ (insert_subvector undef, (v4i32 VR128:$src), (iPTR 0)))),
(SUBREG_TO_REG (i32 0), (VMOVAPSrr VR128:$src), sub_xmm)>;
def : Pat<(v4i64 (X86vzmovl
- (insert_subvector undef, (v2i64 VR128:$src), (i32 0)))),
+ (insert_subvector undef, (v2i64 VR128:$src), (iPTR 0)))),
(SUBREG_TO_REG (i32 0), (VMOVAPSrr VR128:$src), sub_xmm)>;
def : Pat<(v8f32 (X86vzmovl
- (insert_subvector undef, (v4f32 VR128:$src), (i32 0)))),
+ (insert_subvector undef, (v4f32 VR128:$src), (iPTR 0)))),
(SUBREG_TO_REG (i32 0), (VMOVAPSrr VR128:$src), sub_xmm)>;
def : Pat<(v4f64 (X86vzmovl
- (insert_subvector undef, (v2f64 VR128:$src), (i32 0)))),
+ (insert_subvector undef, (v2f64 VR128:$src), (iPTR 0)))),
(SUBREG_TO_REG (i32 0), (VMOVAPSrr VR128:$src), sub_xmm)>;
}
IIC_SSE_MOVU_P_MR>;
// For disassembler
-let isCodeGenOnly = 1 in {
+let isCodeGenOnly = 1, hasSideEffects = 0 in {
def MOVAPSrr_REV : PSI<0x29, MRMDestReg, (outs VR128:$dst), (ins VR128:$src),
"movaps\t{$src, $dst|$dst, $src}", [],
IIC_SSE_MOVA_P_RR>;
(VMOVUPDmr addr:$dst, VR128:$src)>;
}
-let Predicates = [HasSSE1] in
+let Predicates = [UseSSE1] in
def : Pat<(int_x86_sse_storeu_ps addr:$dst, VR128:$src),
(MOVUPSmr addr:$dst, VR128:$src)>;
-let Predicates = [HasSSE2] in
+let Predicates = [UseSSE2] in
def : Pat<(int_x86_sse2_storeu_pd addr:$dst, VR128:$src),
(MOVUPDmr addr:$dst, VR128:$src)>;
(VMOVUPSYmr addr:$dst, VR256:$src)>;
def : Pat<(store (v32i8 VR256:$src), addr:$dst),
(VMOVUPSYmr addr:$dst, VR256:$src)>;
+
+ // Special patterns for storing subvector extracts of lower 128-bits
+ // Its cheaper to just use VMOVAPS/VMOVUPS instead of VEXTRACTF128mr
+ def : Pat<(alignedstore (v2f64 (extract_subvector
+ (v4f64 VR256:$src), (iPTR 0))), addr:$dst),
+ (VMOVAPDmr addr:$dst, (v2f64 (EXTRACT_SUBREG VR256:$src,sub_xmm)))>;
+ def : Pat<(alignedstore (v4f32 (extract_subvector
+ (v8f32 VR256:$src), (iPTR 0))), addr:$dst),
+ (VMOVAPSmr addr:$dst, (v4f32 (EXTRACT_SUBREG VR256:$src,sub_xmm)))>;
+ def : Pat<(alignedstore (v2i64 (extract_subvector
+ (v4i64 VR256:$src), (iPTR 0))), addr:$dst),
+ (VMOVAPDmr addr:$dst, (v2i64 (EXTRACT_SUBREG VR256:$src,sub_xmm)))>;
+ def : Pat<(alignedstore (v4i32 (extract_subvector
+ (v8i32 VR256:$src), (iPTR 0))), addr:$dst),
+ (VMOVAPSmr addr:$dst, (v4i32 (EXTRACT_SUBREG VR256:$src,sub_xmm)))>;
+ def : Pat<(alignedstore (v8i16 (extract_subvector
+ (v16i16 VR256:$src), (iPTR 0))), addr:$dst),
+ (VMOVAPSmr addr:$dst, (v8i16 (EXTRACT_SUBREG VR256:$src,sub_xmm)))>;
+ def : Pat<(alignedstore (v16i8 (extract_subvector
+ (v32i8 VR256:$src), (iPTR 0))), addr:$dst),
+ (VMOVAPSmr addr:$dst, (v16i8 (EXTRACT_SUBREG VR256:$src,sub_xmm)))>;
+
+ def : Pat<(store (v2f64 (extract_subvector
+ (v4f64 VR256:$src), (iPTR 0))), addr:$dst),
+ (VMOVUPDmr addr:$dst, (v2f64 (EXTRACT_SUBREG VR256:$src,sub_xmm)))>;
+ def : Pat<(store (v4f32 (extract_subvector
+ (v8f32 VR256:$src), (iPTR 0))), addr:$dst),
+ (VMOVUPSmr addr:$dst, (v4f32 (EXTRACT_SUBREG VR256:$src,sub_xmm)))>;
+ def : Pat<(store (v2i64 (extract_subvector
+ (v4i64 VR256:$src), (iPTR 0))), addr:$dst),
+ (VMOVUPDmr addr:$dst, (v2i64 (EXTRACT_SUBREG VR256:$src,sub_xmm)))>;
+ def : Pat<(store (v4i32 (extract_subvector
+ (v8i32 VR256:$src), (iPTR 0))), addr:$dst),
+ (VMOVUPSmr addr:$dst, (v4i32 (EXTRACT_SUBREG VR256:$src,sub_xmm)))>;
+ def : Pat<(store (v8i16 (extract_subvector
+ (v16i16 VR256:$src), (iPTR 0))), addr:$dst),
+ (VMOVAPSmr addr:$dst, (v8i16 (EXTRACT_SUBREG VR256:$src,sub_xmm)))>;
+ def : Pat<(store (v16i8 (extract_subvector
+ (v32i8 VR256:$src), (iPTR 0))), addr:$dst),
+ (VMOVUPSmr addr:$dst, (v16i8 (EXTRACT_SUBREG VR256:$src,sub_xmm)))>;
}
// Use movaps / movups for SSE integer load / store (one byte shorter).
// The instructions selected below are then converted to MOVDQA/MOVDQU
// during the SSE domain pass.
-let Predicates = [HasSSE1] in {
+let Predicates = [UseSSE1] in {
def : Pat<(alignedloadv2i64 addr:$src),
(MOVAPSrm addr:$src)>;
def : Pat<(loadv2i64 addr:$src),
// SSE 1 & 2 - Move Low packed FP Instructions
//===----------------------------------------------------------------------===//
-multiclass sse12_mov_hilo_packed<bits<8>opc, RegisterClass RC,
- SDNode psnode, SDNode pdnode, string base_opc,
- string asm_opr, InstrItinClass itin> {
+multiclass sse12_mov_hilo_packed_base<bits<8>opc, SDNode psnode, SDNode pdnode,
+ string base_opc, string asm_opr,
+ InstrItinClass itin> {
def PSrm : PI<opc, MRMSrcMem,
(outs VR128:$dst), (ins VR128:$src1, f64mem:$src2),
!strconcat(base_opc, "s", asm_opr),
- [(set RC:$dst,
- (psnode RC:$src1,
+ [(set VR128:$dst,
+ (psnode VR128:$src1,
(bc_v4f32 (v2f64 (scalar_to_vector (loadf64 addr:$src2))))))],
itin, SSEPackedSingle>, TB;
def PDrm : PI<opc, MRMSrcMem,
- (outs RC:$dst), (ins RC:$src1, f64mem:$src2),
+ (outs VR128:$dst), (ins VR128:$src1, f64mem:$src2),
!strconcat(base_opc, "d", asm_opr),
- [(set RC:$dst, (v2f64 (pdnode RC:$src1,
+ [(set VR128:$dst, (v2f64 (pdnode VR128:$src1,
(scalar_to_vector (loadf64 addr:$src2)))))],
itin, SSEPackedDouble>, TB, OpSize;
+
}
-let AddedComplexity = 20 in {
- defm VMOVL : sse12_mov_hilo_packed<0x12, VR128, X86Movlps, X86Movlpd, "movlp",
- "\t{$src2, $src1, $dst|$dst, $src1, $src2}",
- IIC_SSE_MOV_LH>, VEX_4V;
+multiclass sse12_mov_hilo_packed<bits<8>opc, SDNode psnode, SDNode pdnode,
+ string base_opc, InstrItinClass itin> {
+ defm V#NAME : sse12_mov_hilo_packed_base<opc, psnode, pdnode, base_opc,
+ "\t{$src2, $src1, $dst|$dst, $src1, $src2}",
+ itin>, VEX_4V;
+
+let Constraints = "$src1 = $dst" in
+ defm NAME : sse12_mov_hilo_packed_base<opc, psnode, pdnode, base_opc,
+ "\t{$src2, $dst|$dst, $src2}",
+ itin>;
}
-let Constraints = "$src1 = $dst", AddedComplexity = 20 in {
- defm MOVL : sse12_mov_hilo_packed<0x12, VR128, X86Movlps, X86Movlpd, "movlp",
- "\t{$src2, $dst|$dst, $src2}",
- IIC_SSE_MOV_LH>;
+
+let AddedComplexity = 20 in {
+ defm MOVL : sse12_mov_hilo_packed<0x12, X86Movlps, X86Movlpd, "movlp",
+ IIC_SSE_MOV_LH>;
}
def VMOVLPSmr : VPSI<0x13, MRMDestMem, (outs), (ins f64mem:$dst, VR128:$src),
(VMOVLPDmr addr:$src1, VR128:$src2)>;
}
-let Predicates = [HasSSE1] in {
+let Predicates = [UseSSE1] in {
// (store (vector_shuffle (load addr), v2, <4, 5, 2, 3>), addr) using MOVLPS
def : Pat<(store (i64 (vector_extract (bc_v2i64 (v4f32 VR128:$src2)),
(iPTR 0))), addr:$src1),
(MOVLPSmr addr:$src1, VR128:$src2)>;
}
-let Predicates = [HasSSE2] in {
+let Predicates = [UseSSE2] in {
// Shuffle with MOVLPD
def : Pat<(v2f64 (X86Movlpd VR128:$src1, (load addr:$src2))),
(MOVLPDrm VR128:$src1, addr:$src2)>;
//===----------------------------------------------------------------------===//
let AddedComplexity = 20 in {
- defm VMOVH : sse12_mov_hilo_packed<0x16, VR128, X86Movlhps, X86Movlhpd, "movhp",
- "\t{$src2, $src1, $dst|$dst, $src1, $src2}",
- IIC_SSE_MOV_LH>, VEX_4V;
-}
-let Constraints = "$src1 = $dst", AddedComplexity = 20 in {
- defm MOVH : sse12_mov_hilo_packed<0x16, VR128, X86Movlhps, X86Movlhpd, "movhp",
- "\t{$src2, $dst|$dst, $src2}",
- IIC_SSE_MOV_LH>;
+ defm MOVH : sse12_mov_hilo_packed<0x16, X86Movlhps, X86Movlhpd, "movhp",
+ IIC_SSE_MOV_LH>;
}
// v2f64 extract element 1 is always custom lowered to unpack high to low
(VMOVHPSrm VR128:$src1, addr:$src2)>;
// FIXME: Instead of X86Unpckl, there should be a X86Movlhpd here, the problem
- // is during lowering, where it's not possible to recognize the load fold
+ // is during lowering, where it's not possible to recognize the load fold
// cause it has two uses through a bitcast. One use disappears at isel time
// and the fold opportunity reappears.
def : Pat<(v2f64 (X86Unpckl VR128:$src1,
(VMOVHPDrm VR128:$src1, addr:$src2)>;
}
-let Predicates = [HasSSE1] in {
+let Predicates = [UseSSE1] in {
// MOVHPS patterns
def : Pat<(X86Movlhps VR128:$src1,
(bc_v4f32 (v2i64 (scalar_to_vector (loadi64 addr:$src2))))),
(MOVHPSrm VR128:$src1, addr:$src2)>;
}
-let Predicates = [HasSSE2] in {
+let Predicates = [UseSSE2] in {
// FIXME: Instead of X86Unpckl, there should be a X86Movlhpd here, the problem
- // is during lowering, where it's not possible to recognize the load fold
+ // is during lowering, where it's not possible to recognize the load fold
// cause it has two uses through a bitcast. One use disappears at isel time
// and the fold opportunity reappears.
def : Pat<(v2f64 (X86Unpckl VR128:$src1,
(VMOVHLPSrr VR128:$src1, VR128:$src2)>;
}
-let Predicates = [HasSSE1] in {
+let Predicates = [UseSSE1] in {
// MOVLHPS patterns
def : Pat<(v4i32 (X86Movlhps VR128:$src1, VR128:$src2)),
(MOVLHPSrr VR128:$src1, VR128:$src2)>;
}
multiclass sse12_cvt_p<bits<8> opc, RegisterClass SrcRC, RegisterClass DstRC,
- SDNode OpNode, X86MemOperand x86memop, PatFrag ld_frag,
- string asm, Domain d, OpndItins itins> {
+ X86MemOperand x86memop, string asm, Domain d,
+ OpndItins itins> {
+let neverHasSideEffects = 1 in {
def rr : I<opc, MRMSrcReg, (outs DstRC:$dst), (ins SrcRC:$src), asm,
- [(set DstRC:$dst, (OpNode SrcRC:$src))],
- itins.rr, d>;
+ [], itins.rr, d>;
+ let mayLoad = 1 in
def rm : I<opc, MRMSrcMem, (outs DstRC:$dst), (ins x86memop:$src), asm,
- [(set DstRC:$dst, (OpNode (ld_frag addr:$src)))],
- itins.rm, d>;
+ [], itins.rm, d>;
+}
}
multiclass sse12_vcvt_avx<bits<8> opc, RegisterClass SrcRC, RegisterClass DstRC,
SSE_CVT_SS2SI_32>,
XS, VEX, VEX_LIG;
defm VCVTTSS2SI64 : sse12_cvt_s<0x2C, FR32, GR64, fp_to_sint, f32mem, loadf32,
- "cvttss2si{q}\t{$src, $dst|$dst, $src}",
+ "cvttss2si\t{$src, $dst|$dst, $src}",
SSE_CVT_SS2SI_64>,
XS, VEX, VEX_W, VEX_LIG;
defm VCVTTSD2SI : sse12_cvt_s<0x2C, FR64, GR32, fp_to_sint, f64mem, loadf64,
SSE_CVT_SD2SI>,
XD, VEX, VEX_LIG;
defm VCVTTSD2SI64 : sse12_cvt_s<0x2C, FR64, GR64, fp_to_sint, f64mem, loadf64,
- "cvttsd2si{q}\t{$src, $dst|$dst, $src}",
+ "cvttsd2si\t{$src, $dst|$dst, $src}",
SSE_CVT_SD2SI>,
XD, VEX, VEX_W, VEX_LIG;
+def : InstAlias<"vcvttss2si{l}\t{$src, $dst|$dst, $src}",
+ (VCVTTSS2SIrr GR32:$dst, FR32:$src), 0>;
+def : InstAlias<"vcvttss2si{l}\t{$src, $dst|$dst, $src}",
+ (VCVTTSS2SIrm GR32:$dst, f32mem:$src), 0>;
+def : InstAlias<"vcvttsd2si{l}\t{$src, $dst|$dst, $src}",
+ (VCVTTSD2SIrr GR32:$dst, FR64:$src), 0>;
+def : InstAlias<"vcvttsd2si{l}\t{$src, $dst|$dst, $src}",
+ (VCVTTSD2SIrm GR32:$dst, f64mem:$src), 0>;
+def : InstAlias<"vcvttss2si{q}\t{$src, $dst|$dst, $src}",
+ (VCVTTSS2SI64rr GR64:$dst, FR32:$src), 0>;
+def : InstAlias<"vcvttss2si{q}\t{$src, $dst|$dst, $src}",
+ (VCVTTSS2SI64rm GR64:$dst, f32mem:$src), 0>;
+def : InstAlias<"vcvttsd2si{q}\t{$src, $dst|$dst, $src}",
+ (VCVTTSD2SI64rr GR64:$dst, FR64:$src), 0>;
+def : InstAlias<"vcvttsd2si{q}\t{$src, $dst|$dst, $src}",
+ (VCVTTSD2SI64rm GR64:$dst, f64mem:$src), 0>;
+
// The assembler can recognize rr 64-bit instructions by seeing a rxx
// register, but the same isn't true when only using memory operands,
// provide other assembly "l" and "q" forms to address this explicitly
// where appropriate to do so.
-defm VCVTSI2SS : sse12_vcvt_avx<0x2A, GR32, FR32, i32mem, "cvtsi2ss">,
+defm VCVTSI2SS : sse12_vcvt_avx<0x2A, GR32, FR32, i32mem, "cvtsi2ss{l}">,
XS, VEX_4V, VEX_LIG;
defm VCVTSI2SS64 : sse12_vcvt_avx<0x2A, GR64, FR32, i64mem, "cvtsi2ss{q}">,
XS, VEX_4V, VEX_W, VEX_LIG;
-defm VCVTSI2SD : sse12_vcvt_avx<0x2A, GR32, FR64, i32mem, "cvtsi2sd">,
+defm VCVTSI2SD : sse12_vcvt_avx<0x2A, GR32, FR64, i32mem, "cvtsi2sd{l}">,
XD, VEX_4V, VEX_LIG;
defm VCVTSI2SD64 : sse12_vcvt_avx<0x2A, GR64, FR64, i64mem, "cvtsi2sd{q}">,
XD, VEX_4V, VEX_W, VEX_LIG;
-def : InstAlias<"vcvtsi2sd{l}\t{$src, $src1, $dst|$dst, $src1, $src}",
- (VCVTSI2SDrr FR64:$dst, FR64:$src1, GR32:$src)>;
-def : InstAlias<"vcvtsi2sd{l}\t{$src, $src1, $dst|$dst, $src1, $src}",
+def : InstAlias<"vcvtsi2ss\t{$src, $src1, $dst|$dst, $src1, $src}",
+ (VCVTSI2SSrm FR64:$dst, FR64:$src1, i32mem:$src)>;
+def : InstAlias<"vcvtsi2sd\t{$src, $src1, $dst|$dst, $src1, $src}",
(VCVTSI2SDrm FR64:$dst, FR64:$src1, i32mem:$src)>;
-let Predicates = [HasAVX], AddedComplexity = 1 in {
+let Predicates = [HasAVX] in {
def : Pat<(f32 (sint_to_fp (loadi32 addr:$src))),
(VCVTSI2SSrm (f32 (IMPLICIT_DEF)), addr:$src)>;
def : Pat<(f32 (sint_to_fp (loadi64 addr:$src))),
"cvttss2si\t{$src, $dst|$dst, $src}",
SSE_CVT_SS2SI_32>, XS;
defm CVTTSS2SI64 : sse12_cvt_s<0x2C, FR32, GR64, fp_to_sint, f32mem, loadf32,
- "cvttss2si{q}\t{$src, $dst|$dst, $src}",
+ "cvttss2si\t{$src, $dst|$dst, $src}",
SSE_CVT_SS2SI_64>, XS, REX_W;
defm CVTTSD2SI : sse12_cvt_s<0x2C, FR64, GR32, fp_to_sint, f64mem, loadf64,
"cvttsd2si\t{$src, $dst|$dst, $src}",
SSE_CVT_SD2SI>, XD;
defm CVTTSD2SI64 : sse12_cvt_s<0x2C, FR64, GR64, fp_to_sint, f64mem, loadf64,
- "cvttsd2si{q}\t{$src, $dst|$dst, $src}",
+ "cvttsd2si\t{$src, $dst|$dst, $src}",
SSE_CVT_SD2SI>, XD, REX_W;
defm CVTSI2SS : sse12_cvt_s<0x2A, GR32, FR32, sint_to_fp, i32mem, loadi32,
- "cvtsi2ss\t{$src, $dst|$dst, $src}",
+ "cvtsi2ss{l}\t{$src, $dst|$dst, $src}",
SSE_CVT_Scalar>, XS;
defm CVTSI2SS64 : sse12_cvt_s<0x2A, GR64, FR32, sint_to_fp, i64mem, loadi64,
"cvtsi2ss{q}\t{$src, $dst|$dst, $src}",
SSE_CVT_Scalar>, XS, REX_W;
defm CVTSI2SD : sse12_cvt_s<0x2A, GR32, FR64, sint_to_fp, i32mem, loadi32,
- "cvtsi2sd\t{$src, $dst|$dst, $src}",
+ "cvtsi2sd{l}\t{$src, $dst|$dst, $src}",
SSE_CVT_Scalar>, XD;
defm CVTSI2SD64 : sse12_cvt_s<0x2A, GR64, FR64, sint_to_fp, i64mem, loadi64,
"cvtsi2sd{q}\t{$src, $dst|$dst, $src}",
SSE_CVT_Scalar>, XD, REX_W;
+def : InstAlias<"cvttss2si{l}\t{$src, $dst|$dst, $src}",
+ (CVTTSS2SIrr GR32:$dst, FR32:$src), 0>;
+def : InstAlias<"cvttss2si{l}\t{$src, $dst|$dst, $src}",
+ (CVTTSS2SIrm GR32:$dst, f32mem:$src), 0>;
+def : InstAlias<"cvttsd2si{l}\t{$src, $dst|$dst, $src}",
+ (CVTTSD2SIrr GR32:$dst, FR64:$src), 0>;
+def : InstAlias<"cvttsd2si{l}\t{$src, $dst|$dst, $src}",
+ (CVTTSD2SIrm GR32:$dst, f64mem:$src), 0>;
+def : InstAlias<"cvttss2si{q}\t{$src, $dst|$dst, $src}",
+ (CVTTSS2SI64rr GR64:$dst, FR32:$src), 0>;
+def : InstAlias<"cvttss2si{q}\t{$src, $dst|$dst, $src}",
+ (CVTTSS2SI64rm GR64:$dst, f32mem:$src), 0>;
+def : InstAlias<"cvttsd2si{q}\t{$src, $dst|$dst, $src}",
+ (CVTTSD2SI64rr GR64:$dst, FR64:$src), 0>;
+def : InstAlias<"cvttsd2si{q}\t{$src, $dst|$dst, $src}",
+ (CVTTSD2SI64rm GR64:$dst, f64mem:$src), 0>;
+
+def : InstAlias<"cvtsi2ss\t{$src, $dst|$dst, $src}",
+ (CVTSI2SSrm FR64:$dst, i32mem:$src)>;
+def : InstAlias<"cvtsi2sd\t{$src, $dst|$dst, $src}",
+ (CVTSI2SDrm FR64:$dst, i32mem:$src)>;
+
// Conversion Instructions Intrinsics - Match intrinsics which expect MM
// and/or XMM operand(s).
multiclass sse12_cvt_sint<bits<8> opc, RegisterClass SrcRC, RegisterClass DstRC,
- Intrinsic Int, X86MemOperand x86memop, PatFrag ld_frag,
+ Intrinsic Int, Operand memop, ComplexPattern mem_cpat,
string asm, OpndItins itins> {
def rr : SI<opc, MRMSrcReg, (outs DstRC:$dst), (ins SrcRC:$src),
!strconcat(asm, "\t{$src, $dst|$dst, $src}"),
[(set DstRC:$dst, (Int SrcRC:$src))], itins.rr>;
- def rm : SI<opc, MRMSrcMem, (outs DstRC:$dst), (ins x86memop:$src),
+ def rm : SI<opc, MRMSrcMem, (outs DstRC:$dst), (ins memop:$src),
!strconcat(asm, "\t{$src, $dst|$dst, $src}"),
- [(set DstRC:$dst, (Int (ld_frag addr:$src)))], itins.rm>;
+ [(set DstRC:$dst, (Int mem_cpat:$src))], itins.rm>;
}
multiclass sse12_cvt_sint_3addr<bits<8> opc, RegisterClass SrcRC,
itins.rm>;
}
-defm VCVTSD2SI : sse12_cvt_sint<0x2D, VR128, GR32, int_x86_sse2_cvtsd2si,
- f128mem, load, "cvtsd2si{l}", SSE_CVT_SD2SI>, XD, VEX, VEX_LIG;
+defm VCVTSD2SI : sse12_cvt_sint<0x2D, VR128, GR32,
+ int_x86_sse2_cvtsd2si, sdmem, sse_load_f64, "cvtsd2si",
+ SSE_CVT_SD2SI>, XD, VEX, VEX_LIG;
defm VCVTSD2SI64 : sse12_cvt_sint<0x2D, VR128, GR64,
- int_x86_sse2_cvtsd2si64, f128mem, load, "cvtsd2si{q}",
- SSE_CVT_SD2SI>, XD, VEX, VEX_W, VEX_LIG;
+ int_x86_sse2_cvtsd2si64, sdmem, sse_load_f64, "cvtsd2si",
+ SSE_CVT_SD2SI>, XD, VEX, VEX_W, VEX_LIG;
defm CVTSD2SI : sse12_cvt_sint<0x2D, VR128, GR32, int_x86_sse2_cvtsd2si,
- f128mem, load, "cvtsd2si{l}", SSE_CVT_SD2SI>, XD;
+ sdmem, sse_load_f64, "cvtsd2si", SSE_CVT_SD2SI>, XD;
defm CVTSD2SI64 : sse12_cvt_sint<0x2D, VR128, GR64, int_x86_sse2_cvtsd2si64,
- f128mem, load, "cvtsd2si{q}", SSE_CVT_SD2SI>, XD, REX_W;
+ sdmem, sse_load_f64, "cvtsd2si", SSE_CVT_SD2SI>, XD, REX_W;
defm Int_VCVTSI2SS : sse12_cvt_sint_3addr<0x2A, GR32, VR128,
- int_x86_sse_cvtsi2ss, i32mem, loadi32, "cvtsi2ss",
+ int_x86_sse_cvtsi2ss, i32mem, loadi32, "cvtsi2ss{l}",
SSE_CVT_Scalar, 0>, XS, VEX_4V;
defm Int_VCVTSI2SS64 : sse12_cvt_sint_3addr<0x2A, GR64, VR128,
int_x86_sse_cvtsi642ss, i64mem, loadi64, "cvtsi2ss{q}",
SSE_CVT_Scalar, 0>, XS, VEX_4V,
VEX_W;
defm Int_VCVTSI2SD : sse12_cvt_sint_3addr<0x2A, GR32, VR128,
- int_x86_sse2_cvtsi2sd, i32mem, loadi32, "cvtsi2sd",
+ int_x86_sse2_cvtsi2sd, i32mem, loadi32, "cvtsi2sd{l}",
SSE_CVT_Scalar, 0>, XD, VEX_4V;
defm Int_VCVTSI2SD64 : sse12_cvt_sint_3addr<0x2A, GR64, VR128,
int_x86_sse2_cvtsi642sd, i64mem, loadi64, "cvtsi2sd{q}",
let Constraints = "$src1 = $dst" in {
defm Int_CVTSI2SS : sse12_cvt_sint_3addr<0x2A, GR32, VR128,
int_x86_sse_cvtsi2ss, i32mem, loadi32,
- "cvtsi2ss", SSE_CVT_Scalar>, XS;
+ "cvtsi2ss{l}", SSE_CVT_Scalar>, XS;
defm Int_CVTSI2SS64 : sse12_cvt_sint_3addr<0x2A, GR64, VR128,
int_x86_sse_cvtsi642ss, i64mem, loadi64,
"cvtsi2ss{q}", SSE_CVT_Scalar>, XS, REX_W;
defm Int_CVTSI2SD : sse12_cvt_sint_3addr<0x2A, GR32, VR128,
int_x86_sse2_cvtsi2sd, i32mem, loadi32,
- "cvtsi2sd", SSE_CVT_Scalar>, XD;
+ "cvtsi2sd{l}", SSE_CVT_Scalar>, XD;
defm Int_CVTSI2SD64 : sse12_cvt_sint_3addr<0x2A, GR64, VR128,
int_x86_sse2_cvtsi642sd, i64mem, loadi64,
"cvtsi2sd{q}", SSE_CVT_Scalar>, XD, REX_W;
// Aliases for intrinsics
defm Int_VCVTTSS2SI : sse12_cvt_sint<0x2C, VR128, GR32, int_x86_sse_cvttss2si,
- f32mem, load, "cvttss2si",
+ ssmem, sse_load_f32, "cvttss2si",
SSE_CVT_SS2SI_32>, XS, VEX;
defm Int_VCVTTSS2SI64 : sse12_cvt_sint<0x2C, VR128, GR64,
- int_x86_sse_cvttss2si64, f32mem, load,
- "cvttss2si{q}", SSE_CVT_SS2SI_64>,
- XS, VEX, VEX_W;
+ int_x86_sse_cvttss2si64, ssmem, sse_load_f32,
+ "cvttss2si", SSE_CVT_SS2SI_64>,
+ XS, VEX, VEX_W;
defm Int_VCVTTSD2SI : sse12_cvt_sint<0x2C, VR128, GR32, int_x86_sse2_cvttsd2si,
- f128mem, load, "cvttsd2si", SSE_CVT_SD2SI>,
- XD, VEX;
+ sdmem, sse_load_f64, "cvttsd2si",
+ SSE_CVT_SD2SI>, XD, VEX;
defm Int_VCVTTSD2SI64 : sse12_cvt_sint<0x2C, VR128, GR64,
- int_x86_sse2_cvttsd2si64, f128mem, load,
- "cvttsd2si{q}", SSE_CVT_SD2SI>,
- XD, VEX, VEX_W;
+ int_x86_sse2_cvttsd2si64, sdmem, sse_load_f64,
+ "cvttsd2si", SSE_CVT_SD2SI>,
+ XD, VEX, VEX_W;
defm Int_CVTTSS2SI : sse12_cvt_sint<0x2C, VR128, GR32, int_x86_sse_cvttss2si,
- f32mem, load, "cvttss2si",
+ ssmem, sse_load_f32, "cvttss2si",
SSE_CVT_SS2SI_32>, XS;
defm Int_CVTTSS2SI64 : sse12_cvt_sint<0x2C, VR128, GR64,
- int_x86_sse_cvttss2si64, f32mem, load,
- "cvttss2si{q}", SSE_CVT_SS2SI_64>,
- XS, REX_W;
+ int_x86_sse_cvttss2si64, ssmem, sse_load_f32,
+ "cvttss2si", SSE_CVT_SS2SI_64>, XS, REX_W;
defm Int_CVTTSD2SI : sse12_cvt_sint<0x2C, VR128, GR32, int_x86_sse2_cvttsd2si,
- f128mem, load, "cvttsd2si", SSE_CVT_SD2SI>,
- XD;
+ sdmem, sse_load_f64, "cvttsd2si",
+ SSE_CVT_SD2SI>, XD;
defm Int_CVTTSD2SI64 : sse12_cvt_sint<0x2C, VR128, GR64,
- int_x86_sse2_cvttsd2si64, f128mem, load,
- "cvttsd2si{q}", SSE_CVT_SD2SI>,
- XD, REX_W;
-
-let Pattern = []<dag>, neverHasSideEffects = 1 in {
-defm VCVTSS2SI : sse12_cvt_s<0x2D, FR32, GR32, undef, f32mem, load,
- "cvtss2si{l}\t{$src, $dst|$dst, $src}",
- SSE_CVT_SS2SI_32>, XS, VEX, VEX_LIG;
-defm VCVTSS2SI64 : sse12_cvt_s<0x2D, FR32, GR64, undef, f32mem, load,
- "cvtss2si{q}\t{$src, $dst|$dst, $src}",
- SSE_CVT_SS2SI_64>, XS, VEX, VEX_W, VEX_LIG;
-defm VCVTDQ2PS : sse12_cvt_p<0x5B, VR128, VR128, undef, i128mem, load,
+ int_x86_sse2_cvttsd2si64, sdmem, sse_load_f64,
+ "cvttsd2si", SSE_CVT_SD2SI>, XD, REX_W;
+
+defm VCVTSS2SI : sse12_cvt_sint<0x2D, VR128, GR32, int_x86_sse_cvtss2si,
+ ssmem, sse_load_f32, "cvtss2si",
+ SSE_CVT_SS2SI_32>, XS, VEX, VEX_LIG;
+defm VCVTSS2SI64 : sse12_cvt_sint<0x2D, VR128, GR64, int_x86_sse_cvtss2si64,
+ ssmem, sse_load_f32, "cvtss2si",
+ SSE_CVT_SS2SI_64>, XS, VEX, VEX_W, VEX_LIG;
+
+defm CVTSS2SI : sse12_cvt_sint<0x2D, VR128, GR32, int_x86_sse_cvtss2si,
+ ssmem, sse_load_f32, "cvtss2si",
+ SSE_CVT_SS2SI_32>, XS;
+defm CVTSS2SI64 : sse12_cvt_sint<0x2D, VR128, GR64, int_x86_sse_cvtss2si64,
+ ssmem, sse_load_f32, "cvtss2si",
+ SSE_CVT_SS2SI_64>, XS, REX_W;
+
+defm VCVTDQ2PS : sse12_cvt_p<0x5B, VR128, VR128, i128mem,
"vcvtdq2ps\t{$src, $dst|$dst, $src}",
- SSEPackedSingle, SSE_CVT_PS>, TB, VEX,
- Requires<[HasAVX]>;
-defm VCVTDQ2PSY : sse12_cvt_p<0x5B, VR256, VR256, undef, i256mem, load,
+ SSEPackedSingle, SSE_CVT_PS>,
+ TB, VEX, Requires<[HasAVX]>;
+defm VCVTDQ2PSY : sse12_cvt_p<0x5B, VR256, VR256, i256mem,
"vcvtdq2ps\t{$src, $dst|$dst, $src}",
- SSEPackedSingle, SSE_CVT_PS>, TB, VEX,
- Requires<[HasAVX]>;
-}
-
-let Pattern = []<dag>, neverHasSideEffects = 1 in {
-defm CVTSS2SI : sse12_cvt_s<0x2D, FR32, GR32, undef, f32mem, load /*dummy*/,
- "cvtss2si{l}\t{$src, $dst|$dst, $src}",
- SSE_CVT_SS2SI_32>, XS;
-defm CVTSS2SI64 : sse12_cvt_s<0x2D, FR32, GR64, undef, f32mem, load /*dummy*/,
- "cvtss2si{q}\t{$src, $dst|$dst, $src}",
- SSE_CVT_SS2SI_64>, XS, REX_W;
-defm CVTDQ2PS : sse12_cvt_p<0x5B, VR128, VR128, undef, i128mem, load /*dummy*/,
- "cvtdq2ps\t{$src, $dst|$dst, $src}",
- SSEPackedSingle, SSE_CVT_PS>, TB,
- Requires<[HasSSE2]>;
-}
-
-let Predicates = [HasAVX] in {
- def : Pat<(int_x86_sse_cvtss2si VR128:$src),
- (VCVTSS2SIrr (COPY_TO_REGCLASS VR128:$src, FR32))>;
- def : Pat<(int_x86_sse_cvtss2si (load addr:$src)),
- (VCVTSS2SIrm addr:$src)>;
- def : Pat<(int_x86_sse_cvtss2si64 VR128:$src),
- (VCVTSS2SI64rr (COPY_TO_REGCLASS VR128:$src, FR32))>;
- def : Pat<(int_x86_sse_cvtss2si64 (load addr:$src)),
- (VCVTSS2SI64rm addr:$src)>;
-}
+ SSEPackedSingle, SSE_CVT_PS>,
+ TB, VEX, VEX_L, Requires<[HasAVX]>;
-let Predicates = [HasSSE1] in {
- def : Pat<(int_x86_sse_cvtss2si VR128:$src),
- (CVTSS2SIrr (COPY_TO_REGCLASS VR128:$src, FR32))>;
- def : Pat<(int_x86_sse_cvtss2si (load addr:$src)),
- (CVTSS2SIrm addr:$src)>;
- def : Pat<(int_x86_sse_cvtss2si64 VR128:$src),
- (CVTSS2SI64rr (COPY_TO_REGCLASS VR128:$src, FR32))>;
- def : Pat<(int_x86_sse_cvtss2si64 (load addr:$src)),
- (CVTSS2SI64rm addr:$src)>;
-}
+defm CVTDQ2PS : sse12_cvt_p<0x5B, VR128, VR128, i128mem,
+ "cvtdq2ps\t{$src, $dst|$dst, $src}",
+ SSEPackedSingle, SSE_CVT_PS>,
+ TB, Requires<[UseSSE2]>;
+
+def : InstAlias<"vcvtss2si{l}\t{$src, $dst|$dst, $src}",
+ (VCVTSS2SIrr GR32:$dst, VR128:$src), 0>;
+def : InstAlias<"vcvtss2si{l}\t{$src, $dst|$dst, $src}",
+ (VCVTSS2SIrm GR32:$dst, ssmem:$src), 0>;
+def : InstAlias<"vcvtsd2si{l}\t{$src, $dst|$dst, $src}",
+ (VCVTSD2SIrr GR32:$dst, VR128:$src), 0>;
+def : InstAlias<"vcvtsd2si{l}\t{$src, $dst|$dst, $src}",
+ (VCVTSD2SIrm GR32:$dst, sdmem:$src), 0>;
+def : InstAlias<"vcvtss2si{q}\t{$src, $dst|$dst, $src}",
+ (VCVTSS2SI64rr GR64:$dst, VR128:$src), 0>;
+def : InstAlias<"vcvtss2si{q}\t{$src, $dst|$dst, $src}",
+ (VCVTSS2SI64rm GR64:$dst, ssmem:$src), 0>;
+def : InstAlias<"vcvtsd2si{q}\t{$src, $dst|$dst, $src}",
+ (VCVTSD2SI64rr GR64:$dst, VR128:$src), 0>;
+def : InstAlias<"vcvtsd2si{q}\t{$src, $dst|$dst, $src}",
+ (VCVTSD2SI64rm GR64:$dst, sdmem:$src), 0>;
+
+def : InstAlias<"cvtss2si{l}\t{$src, $dst|$dst, $src}",
+ (CVTSS2SIrr GR32:$dst, VR128:$src), 0>;
+def : InstAlias<"cvtss2si{l}\t{$src, $dst|$dst, $src}",
+ (CVTSS2SIrm GR32:$dst, ssmem:$src), 0>;
+def : InstAlias<"cvtsd2si{l}\t{$src, $dst|$dst, $src}",
+ (CVTSD2SIrr GR32:$dst, VR128:$src), 0>;
+def : InstAlias<"cvtsd2si{l}\t{$src, $dst|$dst, $src}",
+ (CVTSD2SIrm GR32:$dst, sdmem:$src), 0>;
+def : InstAlias<"cvtss2si{q}\t{$src, $dst|$dst, $src}",
+ (CVTSS2SI64rr GR64:$dst, VR128:$src), 0>;
+def : InstAlias<"cvtss2si{q}\t{$src, $dst|$dst, $src}",
+ (CVTSS2SI64rm GR64:$dst, ssmem:$src), 0>;
+def : InstAlias<"cvtsd2si{q}\t{$src, $dst|$dst, $src}",
+ (CVTSD2SI64rr GR64:$dst, VR128:$src), 0>;
+def : InstAlias<"cvtsd2si{q}\t{$src, $dst|$dst, $src}",
+ (CVTSD2SI64rm GR64:$dst, sdmem:$src)>;
/// SSE 2 Only
// Convert scalar double to scalar single
+let neverHasSideEffects = 1 in {
def VCVTSD2SSrr : VSDI<0x5A, MRMSrcReg, (outs FR32:$dst),
(ins FR64:$src1, FR64:$src2),
"cvtsd2ss\t{$src2, $src1, $dst|$dst, $src1, $src2}", [],
"vcvtsd2ss\t{$src2, $src1, $dst|$dst, $src1, $src2}",
[], IIC_SSE_CVT_Scalar_RM>,
XD, Requires<[HasAVX, OptForSize]>, VEX_4V, VEX_LIG;
+}
def : Pat<(f32 (fround FR64:$src)), (VCVTSD2SSrr FR64:$src, FR64:$src)>,
Requires<[HasAVX]>;
[(set FR32:$dst, (fround (loadf64 addr:$src)))],
IIC_SSE_CVT_Scalar_RM>,
XD,
- Requires<[HasSSE2, OptForSize]>;
+ Requires<[UseSSE2, OptForSize]>;
-defm Int_VCVTSD2SS: sse12_cvt_sint_3addr<0x5A, VR128, VR128,
- int_x86_sse2_cvtsd2ss, f64mem, load, "cvtsd2ss",
- SSE_CVT_Scalar, 0>,
- XS, VEX_4V;
-let Constraints = "$src1 = $dst" in
-defm Int_CVTSD2SS: sse12_cvt_sint_3addr<0x5A, VR128, VR128,
- int_x86_sse2_cvtsd2ss, f64mem, load, "cvtsd2ss",
- SSE_CVT_Scalar>, XS;
+def Int_VCVTSD2SSrr: I<0x5A, MRMSrcReg,
+ (outs VR128:$dst), (ins 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<[HasAVX]>;
+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<[HasAVX]>;
+
+let Constraints = "$src1 = $dst" in {
+def Int_CVTSD2SSrr: I<0x5A, MRMSrcReg,
+ (outs VR128:$dst), (ins VR128:$src1, VR128:$src2),
+ "cvtsd2ss\t{$src2, $src1, $dst|$dst, $src1, $src2}",
+ [(set VR128:$dst,
+ (int_x86_sse2_cvtsd2ss VR128:$src1, VR128:$src2))],
+ IIC_SSE_CVT_Scalar_RR>, XD, Requires<[UseSSE2]>;
+def Int_CVTSD2SSrm: I<0x5A, MRMSrcReg,
+ (outs VR128:$dst), (ins VR128:$src1, sdmem:$src2),
+ "cvtsd2ss\t{$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, Requires<[UseSSE2]>;
+}
// Convert scalar single to scalar double
// SSE2 instructions with XS prefix
+let neverHasSideEffects = 1 in {
def VCVTSS2SDrr : I<0x5A, MRMSrcReg, (outs FR64:$dst),
(ins FR32:$src1, FR32:$src2),
"vcvtss2sd\t{$src2, $src1, $dst|$dst, $src1, $src2}",
"vcvtss2sd\t{$src2, $src1, $dst|$dst, $src1, $src2}",
[], IIC_SSE_CVT_Scalar_RM>,
XS, VEX_4V, VEX_LIG, Requires<[HasAVX, OptForSize]>;
-
-let Predicates = [HasAVX] in {
- def : Pat<(f64 (fextend FR32:$src)),
- (VCVTSS2SDrr FR32:$src, FR32:$src)>;
- def : Pat<(fextend (loadf32 addr:$src)),
- (VCVTSS2SDrm (f32 (IMPLICIT_DEF)), addr:$src)>;
- def : Pat<(extloadf32 addr:$src),
- (VCVTSS2SDrm (f32 (IMPLICIT_DEF)), addr:$src)>;
}
+def : Pat<(f64 (fextend FR32:$src)),
+ (VCVTSS2SDrr FR32:$src, FR32:$src)>, Requires<[HasAVX]>;
+def : Pat<(fextend (loadf32 addr:$src)),
+ (VCVTSS2SDrm (f32 (IMPLICIT_DEF)), addr:$src)>, Requires<[HasAVX]>;
+
+def : Pat<(extloadf32 addr:$src),
+ (VCVTSS2SDrm (f32 (IMPLICIT_DEF)), addr:$src)>,
+ Requires<[HasAVX, OptForSize]>;
def : Pat<(extloadf32 addr:$src),
- (VCVTSS2SDrr (f32 (IMPLICIT_DEF)), (MOVSSrm addr:$src))>,
- Requires<[HasAVX, OptForSpeed]>;
+ (VCVTSS2SDrr (f32 (IMPLICIT_DEF)), (VMOVSSrm addr:$src))>,
+ Requires<[HasAVX, OptForSpeed]>;
def CVTSS2SDrr : I<0x5A, MRMSrcReg, (outs FR64:$dst), (ins FR32:$src),
"cvtss2sd\t{$src, $dst|$dst, $src}",
[(set FR64:$dst, (fextend FR32:$src))],
IIC_SSE_CVT_Scalar_RR>, XS,
- Requires<[HasSSE2]>;
+ Requires<[UseSSE2]>;
def CVTSS2SDrm : I<0x5A, MRMSrcMem, (outs FR64:$dst), (ins f32mem:$src),
"cvtss2sd\t{$src, $dst|$dst, $src}",
[(set FR64:$dst, (extloadf32 addr:$src))],
IIC_SSE_CVT_Scalar_RM>, XS,
- Requires<[HasSSE2, OptForSize]>;
+ Requires<[UseSSE2, OptForSize]>;
// extload f32 -> f64. This matches load+fextend because we have a hack in
// the isel (PreprocessForFPConvert) that can introduce loads after dag
// Since these loads aren't folded into the fextend, we have to match it
// explicitly here.
def : Pat<(fextend (loadf32 addr:$src)),
- (CVTSS2SDrm addr:$src)>, Requires<[HasSSE2]>;
+ (CVTSS2SDrm addr:$src)>, Requires<[UseSSE2]>;
def : Pat<(extloadf32 addr:$src),
- (CVTSS2SDrr (MOVSSrm addr:$src))>, Requires<[HasSSE2, OptForSpeed]>;
+ (CVTSS2SDrr (MOVSSrm addr:$src))>, Requires<[UseSSE2, OptForSpeed]>;
def Int_VCVTSS2SDrr: I<0x5A, MRMSrcReg,
(outs VR128:$dst), (ins VR128:$src1, VR128:$src2),
"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<[HasAVX]>;
+ [(set VR128:$dst,
+ (int_x86_sse2_cvtss2sd VR128:$src1, VR128:$src2))],
+ IIC_SSE_CVT_Scalar_RR>, XS, VEX_4V, Requires<[HasAVX]>;
def Int_VCVTSS2SDrm: I<0x5A, MRMSrcMem,
- (outs VR128:$dst), (ins VR128:$src1, f32mem:$src2),
+ (outs VR128:$dst), (ins VR128:$src1, ssmem:$src2),
"vcvtss2sd\t{$src2, $src1, $dst|$dst, $src1, $src2}",
- [(set VR128:$dst, (int_x86_sse2_cvtss2sd VR128:$src1,
- (load addr:$src2)))],
- IIC_SSE_CVT_Scalar_RM>, XS, VEX_4V,
- Requires<[HasAVX]>;
+ [(set VR128:$dst,
+ (int_x86_sse2_cvtss2sd VR128:$src1, sse_load_f32:$src2))],
+ IIC_SSE_CVT_Scalar_RM>, XS, VEX_4V, Requires<[HasAVX]>;
let Constraints = "$src1 = $dst" in { // SSE2 instructions with XS prefix
def Int_CVTSS2SDrr: I<0x5A, MRMSrcReg,
(outs VR128:$dst), (ins VR128:$src1, VR128:$src2),
"cvtss2sd\t{$src2, $dst|$dst, $src2}",
- [(set VR128:$dst, (int_x86_sse2_cvtss2sd VR128:$src1,
- VR128:$src2))],
- IIC_SSE_CVT_Scalar_RR>, XS,
- Requires<[HasSSE2]>;
+ [(set VR128:$dst,
+ (int_x86_sse2_cvtss2sd VR128:$src1, VR128:$src2))],
+ IIC_SSE_CVT_Scalar_RR>, XS, Requires<[UseSSE2]>;
def Int_CVTSS2SDrm: I<0x5A, MRMSrcMem,
- (outs VR128:$dst), (ins VR128:$src1, f32mem:$src2),
+ (outs VR128:$dst), (ins VR128:$src1, ssmem:$src2),
"cvtss2sd\t{$src2, $dst|$dst, $src2}",
- [(set VR128:$dst, (int_x86_sse2_cvtss2sd VR128:$src1,
- (load addr:$src2)))],
- IIC_SSE_CVT_Scalar_RM>, XS,
- Requires<[HasSSE2]>;
+ [(set VR128:$dst,
+ (int_x86_sse2_cvtss2sd VR128:$src1, sse_load_f32:$src2))],
+ IIC_SSE_CVT_Scalar_RM>, XS, Requires<[UseSSE2]>;
}
// Convert packed single/double fp to doubleword
def VCVTPS2DQrr : VPDI<0x5B, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
- "cvtps2dq\t{$src, $dst|$dst, $src}", [],
+ "cvtps2dq\t{$src, $dst|$dst, $src}",
+ [(set VR128:$dst, (int_x86_sse2_cvtps2dq VR128:$src))],
IIC_SSE_CVT_PS_RR>, VEX;
def VCVTPS2DQrm : VPDI<0x5B, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
- "cvtps2dq\t{$src, $dst|$dst, $src}", [],
+ "cvtps2dq\t{$src, $dst|$dst, $src}",
+ [(set VR128:$dst,
+ (int_x86_sse2_cvtps2dq (memopv4f32 addr:$src)))],
IIC_SSE_CVT_PS_RM>, VEX;
def VCVTPS2DQYrr : VPDI<0x5B, MRMSrcReg, (outs VR256:$dst), (ins VR256:$src),
- "cvtps2dq\t{$src, $dst|$dst, $src}", [],
- IIC_SSE_CVT_PS_RR>, VEX;
+ "cvtps2dq\t{$src, $dst|$dst, $src}",
+ [(set VR256:$dst,
+ (int_x86_avx_cvt_ps2dq_256 VR256:$src))],
+ IIC_SSE_CVT_PS_RR>, VEX, VEX_L;
def VCVTPS2DQYrm : VPDI<0x5B, MRMSrcMem, (outs VR256:$dst), (ins f256mem:$src),
- "cvtps2dq\t{$src, $dst|$dst, $src}", [],
- IIC_SSE_CVT_PS_RM>, VEX;
+ "cvtps2dq\t{$src, $dst|$dst, $src}",
+ [(set VR256:$dst,
+ (int_x86_avx_cvt_ps2dq_256 (memopv8f32 addr:$src)))],
+ IIC_SSE_CVT_PS_RM>, VEX, VEX_L;
def CVTPS2DQrr : PDI<0x5B, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
- "cvtps2dq\t{$src, $dst|$dst, $src}", [],
+ "cvtps2dq\t{$src, $dst|$dst, $src}",
+ [(set VR128:$dst, (int_x86_sse2_cvtps2dq VR128:$src))],
IIC_SSE_CVT_PS_RR>;
def CVTPS2DQrm : PDI<0x5B, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
- "cvtps2dq\t{$src, $dst|$dst, $src}", [],
+ "cvtps2dq\t{$src, $dst|$dst, $src}",
+ [(set VR128:$dst,
+ (int_x86_sse2_cvtps2dq (memopv4f32 addr:$src)))],
IIC_SSE_CVT_PS_RM>;
-let Predicates = [HasAVX] in {
- def : Pat<(int_x86_sse2_cvtps2dq VR128:$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:$src)>;
-}
// Convert Packed Double FP to Packed DW Integers
let Predicates = [HasAVX] in {
// register, but the same isn't true when using memory operands instead.
// Provide other assembly rr and rm forms to address this explicitly.
def VCVTPD2DQrr : SDI<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
- "vcvtpd2dq\t{$src, $dst|$dst, $src}", []>, VEX;
+ "vcvtpd2dq\t{$src, $dst|$dst, $src}",
+ [(set VR128:$dst, (int_x86_sse2_cvtpd2dq VR128:$src))]>,
+ VEX;
// XMM only
def : InstAlias<"vcvtpd2dqx\t{$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, $src}", []>, VEX;
+ "vcvtpd2dqx\t{$src, $dst|$dst, $src}",
+ [(set VR128:$dst,
+ (int_x86_sse2_cvtpd2dq (memopv2f64 addr:$src)))]>, VEX;
// YMM only
def VCVTPD2DQYrr : SDI<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR256:$src),
- "vcvtpd2dq{y}\t{$src, $dst|$dst, $src}", []>, VEX;
+ "vcvtpd2dq{y}\t{$src, $dst|$dst, $src}",
+ [(set VR128:$dst,
+ (int_x86_avx_cvt_pd2dq_256 VR256:$src))]>, VEX, VEX_L;
def VCVTPD2DQYrm : SDI<0xE6, MRMSrcMem, (outs VR128:$dst), (ins f256mem:$src),
- "vcvtpd2dq{y}\t{$src, $dst|$dst, $src}", []>, VEX, VEX_L;
+ "vcvtpd2dq{y}\t{$src, $dst|$dst, $src}",
+ [(set VR128:$dst,
+ (int_x86_avx_cvt_pd2dq_256 (memopv4f64 addr:$src)))]>,
+ VEX, VEX_L;
def : InstAlias<"vcvtpd2dq\t{$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, $src}", [],
+ "cvtpd2dq\t{$src, $dst|$dst, $src}",
+ [(set VR128:$dst,
+ (int_x86_sse2_cvtpd2dq (memopv2f64 addr:$src)))],
IIC_SSE_CVT_PD_RM>;
def CVTPD2DQrr : SDI<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
- "cvtpd2dq\t{$src, $dst|$dst, $src}", [],
+ "cvtpd2dq\t{$src, $dst|$dst, $src}",
+ [(set VR128:$dst, (int_x86_sse2_cvtpd2dq VR128:$src))],
IIC_SSE_CVT_PD_RR>;
-let Predicates = [HasAVX] in {
- def : Pat<(int_x86_sse2_cvtpd2dq VR128:$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:$src)>;
-}
-
// Convert with truncation packed single/double fp to doubleword
// SSE2 packed instructions with XS prefix
-def VCVTTPS2DQrr : VSSI<0x5B, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
- "cvttps2dq\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst,
- (int_x86_sse2_cvttps2dq VR128:$src))],
- IIC_SSE_CVT_PS_RR>, VEX;
-def VCVTTPS2DQrm : VSSI<0x5B, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
- "cvttps2dq\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst, (int_x86_sse2_cvttps2dq
- (memopv4f32 addr:$src)))],
- IIC_SSE_CVT_PS_RM>, VEX;
-def VCVTTPS2DQYrr : VSSI<0x5B, MRMSrcReg, (outs VR256:$dst), (ins VR256:$src),
+def VCVTTPS2DQrr : VS2SI<0x5B, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
"cvttps2dq\t{$src, $dst|$dst, $src}",
- [(set VR256:$dst,
- (int_x86_avx_cvtt_ps2dq_256 VR256:$src))],
- IIC_SSE_CVT_PS_RR>, VEX;
-def VCVTTPS2DQYrm : VSSI<0x5B, MRMSrcMem, (outs VR256:$dst), (ins f256mem:$src),
+ [(set VR128:$dst,
+ (int_x86_sse2_cvttps2dq VR128:$src))],
+ IIC_SSE_CVT_PS_RR>, VEX;
+def VCVTTPS2DQrm : VS2SI<0x5B, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
"cvttps2dq\t{$src, $dst|$dst, $src}",
- [(set VR256:$dst, (int_x86_avx_cvtt_ps2dq_256
- (memopv8f32 addr:$src)))],
- IIC_SSE_CVT_PS_RM>, VEX;
-
-def CVTTPS2DQrr : SSI<0x5B, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
- "cvttps2dq\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst,
- (int_x86_sse2_cvttps2dq VR128:$src))],
- IIC_SSE_CVT_PS_RR>;
-def CVTTPS2DQrm : SSI<0x5B, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
- "cvttps2dq\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst,
- (int_x86_sse2_cvttps2dq (memopv4f32 addr:$src)))],
- IIC_SSE_CVT_PS_RM>;
+ [(set VR128:$dst, (int_x86_sse2_cvttps2dq
+ (memopv4f32 addr:$src)))],
+ IIC_SSE_CVT_PS_RM>, VEX;
+def VCVTTPS2DQYrr : VS2SI<0x5B, MRMSrcReg, (outs VR256:$dst), (ins VR256:$src),
+ "cvttps2dq\t{$src, $dst|$dst, $src}",
+ [(set VR256:$dst,
+ (int_x86_avx_cvtt_ps2dq_256 VR256:$src))],
+ IIC_SSE_CVT_PS_RR>, VEX, VEX_L;
+def VCVTTPS2DQYrm : VS2SI<0x5B, MRMSrcMem, (outs VR256:$dst), (ins f256mem:$src),
+ "cvttps2dq\t{$src, $dst|$dst, $src}",
+ [(set VR256:$dst, (int_x86_avx_cvtt_ps2dq_256
+ (memopv8f32 addr:$src)))],
+ IIC_SSE_CVT_PS_RM>, VEX, VEX_L;
+
+def CVTTPS2DQrr : S2SI<0x5B, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
+ "cvttps2dq\t{$src, $dst|$dst, $src}",
+ [(set VR128:$dst, (int_x86_sse2_cvttps2dq VR128:$src))],
+ IIC_SSE_CVT_PS_RR>;
+def CVTTPS2DQrm : S2SI<0x5B, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
+ "cvttps2dq\t{$src, $dst|$dst, $src}",
+ [(set VR128:$dst,
+ (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))),
(VCVTTPS2DQYrm addr:$src)>;
}
-let Predicates = [HasSSE2] in {
+let Predicates = [UseSSE2] in {
def : Pat<(v4f32 (sint_to_fp (v4i32 VR128:$src))),
(CVTDQ2PSrr VR128:$src)>;
def : Pat<(v4f32 (sint_to_fp (bc_v4i32 (memopv2i64 addr:$src)))),
(int_x86_sse2_cvttpd2dq VR128:$src))],
IIC_SSE_CVT_PD_RR>, VEX;
-def CVTTPD2DQrr : PDI<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
- "cvttpd2dq\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst, (int_x86_sse2_cvttpd2dq VR128:$src))],
- IIC_SSE_CVT_PD_RR>;
-def CVTTPD2DQrm : PDI<0xE6, MRMSrcMem, (outs VR128:$dst),(ins f128mem:$src),
- "cvttpd2dq\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst, (int_x86_sse2_cvttpd2dq
- (memopv2f64 addr:$src)))],
- IIC_SSE_CVT_PD_RM>;
-
// The assembler can recognize rr 256-bit instructions by seeing a ymm
// register, but the same isn't true when using memory operands instead.
// Provide other assembly rr and rm forms to address this explicitly.
// YMM only
def VCVTTPD2DQYrr : VPDI<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR256:$src),
- "cvttpd2dq{y}\t{$src, $dst|$dst, $src}", [],
- IIC_SSE_CVT_PD_RR>, VEX;
+ "cvttpd2dq{y}\t{$src, $dst|$dst, $src}",
+ [(set VR128:$dst,
+ (int_x86_avx_cvtt_pd2dq_256 VR256:$src))],
+ IIC_SSE_CVT_PD_RR>, VEX, VEX_L;
def VCVTTPD2DQYrm : VPDI<0xE6, MRMSrcMem, (outs VR128:$dst), (ins f256mem:$src),
- "cvttpd2dq{y}\t{$src, $dst|$dst, $src}", [],
+ "cvttpd2dq{y}\t{$src, $dst|$dst, $src}",
+ [(set VR128:$dst,
+ (int_x86_avx_cvtt_pd2dq_256 (memopv4f64 addr:$src)))],
IIC_SSE_CVT_PD_RM>, VEX, VEX_L;
def : InstAlias<"vcvttpd2dq\t{$src, $dst|$dst, $src}",
(VCVTTPD2DQYrr VR128:$dst, VR256:$src)>;
(VCVTTPD2DQYrm addr:$src)>;
} // Predicates = [HasAVX]
+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:$src))],
+ IIC_SSE_CVT_PD_RR>;
+def CVTTPD2DQrm : PDI<0xE6, MRMSrcMem, (outs VR128:$dst),(ins f128mem:$src),
+ "cvttpd2dq\t{$src, $dst|$dst, $src}",
+ [(set VR128:$dst, (int_x86_sse2_cvttpd2dq
+ (memopv2f64 addr:$src)))],
+ IIC_SSE_CVT_PD_RM>;
+
// Convert packed single to packed double
let Predicates = [HasAVX] in {
// SSE2 instructions without OpSize prefix
def VCVTPS2PDrr : I<0x5A, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
- "vcvtps2pd\t{$src, $dst|$dst, $src}", [],
+ "vcvtps2pd\t{$src, $dst|$dst, $src}",
+ [(set VR128:$dst, (int_x86_sse2_cvtps2pd VR128:$src))],
IIC_SSE_CVT_PD_RR>, TB, VEX;
def VCVTPS2PDrm : I<0x5A, MRMSrcMem, (outs VR128:$dst), (ins f64mem:$src),
- "vcvtps2pd\t{$src, $dst|$dst, $src}", [],
- IIC_SSE_CVT_PD_RM>, TB, VEX;
+ "vcvtps2pd\t{$src, $dst|$dst, $src}",
+ [(set VR128:$dst, (v2f64 (extloadv2f32 addr:$src)))],
+ IIC_SSE_CVT_PD_RM>, TB, VEX;
def VCVTPS2PDYrr : I<0x5A, MRMSrcReg, (outs VR256:$dst), (ins VR128:$src),
- "vcvtps2pd\t{$src, $dst|$dst, $src}", [],
- IIC_SSE_CVT_PD_RR>, TB, VEX;
+ "vcvtps2pd\t{$src, $dst|$dst, $src}",
+ [(set VR256:$dst,
+ (int_x86_avx_cvt_ps2_pd_256 VR128:$src))],
+ IIC_SSE_CVT_PD_RR>, TB, VEX, VEX_L;
def VCVTPS2PDYrm : I<0x5A, MRMSrcMem, (outs VR256:$dst), (ins f128mem:$src),
- "vcvtps2pd\t{$src, $dst|$dst, $src}", [],
- IIC_SSE_CVT_PD_RM>, TB, VEX;
+ "vcvtps2pd\t{$src, $dst|$dst, $src}",
+ [(set VR256:$dst,
+ (int_x86_avx_cvt_ps2_pd_256 (memopv4f32 addr:$src)))],
+ IIC_SSE_CVT_PD_RM>, TB, VEX, VEX_L;
}
-let Predicates = [HasSSE2] in {
+let Predicates = [UseSSE2] in {
def CVTPS2PDrr : I<0x5A, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
- "cvtps2pd\t{$src, $dst|$dst, $src}", [],
+ "cvtps2pd\t{$src, $dst|$dst, $src}",
+ [(set VR128:$dst, (int_x86_sse2_cvtps2pd VR128:$src))],
IIC_SSE_CVT_PD_RR>, TB;
def CVTPS2PDrm : I<0x5A, MRMSrcMem, (outs VR128:$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:$src)>;
+ "cvtps2pd\t{$src, $dst|$dst, $src}",
+ [(set VR128:$dst, (v2f64 (extloadv2f32 addr:$src)))],
+ IIC_SSE_CVT_PD_RM>, TB;
}
// Convert Packed DW Integers to Packed Double FP
let Predicates = [HasAVX] in {
-def VCVTDQ2PDrm : SSDI<0xE6, MRMSrcMem, (outs VR128:$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, $src}", []>, VEX;
+let neverHasSideEffects = 1, mayLoad = 1 in
+def VCVTDQ2PDrm : S2SI<0xE6, MRMSrcMem, (outs VR128:$dst), (ins i64mem:$src),
+ "vcvtdq2pd\t{$src, $dst|$dst, $src}",
+ []>, VEX;
+def VCVTDQ2PDrr : S2SI<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
+ "vcvtdq2pd\t{$src, $dst|$dst, $src}",
+ [(set VR128:$dst,
+ (int_x86_sse2_cvtdq2pd VR128:$src))]>, VEX;
+def VCVTDQ2PDYrm : S2SI<0xE6, MRMSrcMem, (outs VR256:$dst), (ins i128mem:$src),
+ "vcvtdq2pd\t{$src, $dst|$dst, $src}",
+ [(set VR256:$dst,
+ (int_x86_avx_cvtdq2_pd_256
+ (bitconvert (memopv2i64 addr:$src))))]>, VEX, VEX_L;
+def VCVTDQ2PDYrr : S2SI<0xE6, MRMSrcReg, (outs VR256:$dst), (ins VR128:$src),
+ "vcvtdq2pd\t{$src, $dst|$dst, $src}",
+ [(set VR256:$dst,
+ (int_x86_avx_cvtdq2_pd_256 VR128:$src))]>, VEX, VEX_L;
}
-def CVTDQ2PDrm : SSDI<0xE6, MRMSrcMem, (outs VR128:$dst), (ins i64mem:$src),
+let neverHasSideEffects = 1, mayLoad = 1 in
+def CVTDQ2PDrm : S2SI<0xE6, MRMSrcMem, (outs VR128:$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, $src}", [],
+def CVTDQ2PDrr : S2SI<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
+ "cvtdq2pd\t{$src, $dst|$dst, $src}",
+ [(set VR128:$dst, (int_x86_sse2_cvtdq2pd VR128:$src))],
IIC_SSE_CVT_PD_RM>;
-// 128 bit register conversion intrinsics
-let Predicates = [HasAVX] in
-def : Pat<(int_x86_sse2_cvtdq2pd VR128:$src),
- (VCVTDQ2PDrr VR128:$src)>;
-
-let Predicates = [HasSSE2] in
-def : Pat<(int_x86_sse2_cvtdq2pd VR128:$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)))),
// register, but the same isn't true when using memory operands instead.
// Provide other assembly rr and rm forms to address this explicitly.
def VCVTPD2PSrr : VPDI<0x5A, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
- "cvtpd2ps\t{$src, $dst|$dst, $src}", [],
+ "cvtpd2ps\t{$src, $dst|$dst, $src}",
+ [(set VR128:$dst, (int_x86_sse2_cvtpd2ps VR128:$src))],
IIC_SSE_CVT_PD_RR>, VEX;
// XMM only
def : InstAlias<"vcvtpd2psx\t{$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, $src}", [],
+ "cvtpd2psx\t{$src, $dst|$dst, $src}",
+ [(set VR128:$dst,
+ (int_x86_sse2_cvtpd2ps (memopv2f64 addr:$src)))],
IIC_SSE_CVT_PD_RM>, VEX;
// YMM only
def VCVTPD2PSYrr : VPDI<0x5A, MRMSrcReg, (outs VR128:$dst), (ins VR256:$src),
- "cvtpd2ps{y}\t{$src, $dst|$dst, $src}", [],
- IIC_SSE_CVT_PD_RR>, VEX;
+ "cvtpd2ps{y}\t{$src, $dst|$dst, $src}",
+ [(set VR128:$dst,
+ (int_x86_avx_cvt_pd2_ps_256 VR256:$src))],
+ IIC_SSE_CVT_PD_RR>, VEX, VEX_L;
def VCVTPD2PSYrm : VPDI<0x5A, MRMSrcMem, (outs VR128:$dst), (ins f256mem:$src),
- "cvtpd2ps{y}\t{$src, $dst|$dst, $src}", [],
+ "cvtpd2ps{y}\t{$src, $dst|$dst, $src}",
+ [(set VR128:$dst,
+ (int_x86_avx_cvt_pd2_ps_256 (memopv4f64 addr:$src)))],
IIC_SSE_CVT_PD_RM>, VEX, VEX_L;
def : InstAlias<"vcvtpd2ps\t{$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, $src}", [],
+ "cvtpd2ps\t{$src, $dst|$dst, $src}",
+ [(set VR128:$dst, (int_x86_sse2_cvtpd2ps VR128:$src))],
IIC_SSE_CVT_PD_RR>;
def CVTPD2PSrm : PDI<0x5A, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
- "cvtpd2ps\t{$src, $dst|$dst, $src}", [],
+ "cvtpd2ps\t{$src, $dst|$dst, $src}",
+ [(set VR128:$dst,
+ (int_x86_sse2_cvtpd2ps (memopv2f64 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:$src)>;
-}
-
// AVX 256-bit register conversion intrinsics
// FIXME: Migrate SSE conversion intrinsics matching to use patterns as below
// whenever possible to avoid declaring two versions of each one.
def : Pat<(int_x86_avx_cvtdq2_ps_256 (bitconvert (memopv4i64 addr:$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:$src)>;
-
// Match fround and fextend for 128/256-bit conversions
+ def : Pat<(v4f32 (X86vfpround (v2f64 VR128:$src))),
+ (VCVTPD2PSrr VR128:$src)>;
+ def : Pat<(v4f32 (X86vfpround (memopv2f64 addr:$src))),
+ (VCVTPD2PSXrm addr:$src)>;
def : Pat<(v4f32 (fround (v4f64 VR256:$src))),
(VCVTPD2PSYrr VR256:$src)>;
def : Pat<(v4f32 (fround (loadv4f64 addr:$src))),
(VCVTPD2PSYrm addr:$src)>;
+ def : Pat<(v2f64 (X86vfpext (v4f32 VR128:$src))),
+ (VCVTPS2PDrr VR128:$src)>;
def : Pat<(v4f64 (fextend (v4f32 VR128:$src))),
(VCVTPS2PDYrr VR128:$src)>;
- def : Pat<(v4f64 (fextend (loadv4f32 addr:$src))),
+ def : Pat<(v4f64 (extloadv4f32 addr:$src)),
(VCVTPS2PDYrm addr:$src)>;
}
+let Predicates = [UseSSE2] in {
+ // Match fround and fextend for 128 conversions
+ def : Pat<(v4f32 (X86vfpround (v2f64 VR128:$src))),
+ (CVTPD2PSrr VR128:$src)>;
+ def : Pat<(v4f32 (X86vfpround (memopv2f64 addr:$src))),
+ (CVTPD2PSrm addr:$src)>;
+
+ def : Pat<(v2f64 (X86vfpext (v4f32 VR128:$src))),
+ (CVTPS2PDrr VR128:$src)>;
+}
+
//===----------------------------------------------------------------------===//
// SSE 1 & 2 - Compare Instructions
//===----------------------------------------------------------------------===//
// sse12_cmp_scalar - sse 1 & 2 compare scalar instructions
multiclass sse12_cmp_scalar<RegisterClass RC, X86MemOperand x86memop,
- Operand CC, SDNode OpNode, ValueType VT,
+ Operand CC, SDNode OpNode, ValueType VT,
PatFrag ld_frag, string asm, string asm_alt,
OpndItins itins> {
def rr : SIi8<0xC2, MRMSrcReg,
// sse12_cmp_packed - sse 1 & 2 compare packed instructions
multiclass sse12_cmp_packed<RegisterClass RC, X86MemOperand x86memop,
- Operand CC, Intrinsic Int, string asm,
+ Operand CC, Intrinsic Int, string asm,
string asm_alt, Domain d> {
def rri : PIi8<0xC2, MRMSrcReg,
(outs RC:$dst), (ins RC:$src1, RC:$src2, CC:$cc), asm,
defm VCMPPSY : sse12_cmp_packed<VR256, f256mem, AVXCC, int_x86_avx_cmp_ps_256,
"cmp${cc}ps\t{$src2, $src1, $dst|$dst, $src1, $src2}",
"cmpps\t{$cc, $src2, $src1, $dst|$dst, $src1, $src2, $cc}",
- SSEPackedSingle>, TB, VEX_4V;
+ SSEPackedSingle>, TB, VEX_4V, VEX_L;
defm VCMPPDY : sse12_cmp_packed<VR256, f256mem, AVXCC, int_x86_avx_cmp_pd_256,
"cmp${cc}pd\t{$src2, $src1, $dst|$dst, $src1, $src2}",
"cmppd\t{$cc, $src2, $src1, $dst|$dst, $src1, $src2, $cc}",
- SSEPackedDouble>, TB, OpSize, VEX_4V;
+ SSEPackedDouble>, TB, OpSize, VEX_4V, VEX_L;
let Constraints = "$src1 = $dst" in {
defm CMPPS : sse12_cmp_packed<VR128, f128mem, SSECC, int_x86_sse_cmp_ps,
"cmp${cc}ps\t{$src2, $dst|$dst, $src2}",
(VCMPPDYrmi VR256:$src1, addr:$src2, imm:$cc)>;
}
-let Predicates = [HasSSE1] in {
+let Predicates = [UseSSE1] in {
def : Pat<(v4i32 (X86cmpp (v4f32 VR128:$src1), VR128:$src2, imm:$cc)),
(CMPPSrri (v4f32 VR128:$src1), (v4f32 VR128:$src2), imm:$cc)>;
def : Pat<(v4i32 (X86cmpp (v4f32 VR128:$src1), (memop addr:$src2), imm:$cc)),
(CMPPSrmi (v4f32 VR128:$src1), addr:$src2, imm:$cc)>;
}
-let Predicates = [HasSSE2] in {
+let Predicates = [UseSSE2] in {
def : Pat<(v2i64 (X86cmpp (v2f64 VR128:$src1), VR128:$src2, imm:$cc)),
(CMPPDrri VR128:$src1, VR128:$src2, imm:$cc)>;
def : Pat<(v2i64 (X86cmpp (v2f64 VR128:$src1), (memop addr:$src2), imm:$cc)),
memopv4f32, SSEPackedSingle>, TB, VEX_4V;
defm VSHUFPSY : sse12_shuffle<VR256, f256mem, v8f32,
"shufps\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}",
- memopv8f32, SSEPackedSingle>, TB, VEX_4V;
+ memopv8f32, SSEPackedSingle>, TB, VEX_4V, VEX_L;
defm VSHUFPD : sse12_shuffle<VR128, f128mem, v2f64,
"shufpd\t{$src3, $src2, $src1, $dst|$dst, $src2, $src2, $src3}",
memopv2f64, SSEPackedDouble>, TB, OpSize, VEX_4V;
defm VSHUFPDY : sse12_shuffle<VR256, f256mem, v4f64,
"shufpd\t{$src3, $src2, $src1, $dst|$dst, $src2, $src2, $src3}",
- memopv4f64, SSEPackedDouble>, TB, OpSize, VEX_4V;
+ memopv4f64, SSEPackedDouble>, TB, OpSize, VEX_4V, VEX_L;
let Constraints = "$src1 = $dst" in {
defm SHUFPS : sse12_shuffle<VR128, f128mem, v4f32,
(VSHUFPDYrmi VR256:$src1, addr:$src2, imm:$imm)>;
}
-let Predicates = [HasSSE1] in {
+let Predicates = [UseSSE1] in {
def : Pat<(v4i32 (X86Shufp VR128:$src1,
(bc_v4i32 (memopv2i64 addr:$src2)), (i8 imm:$imm))),
(SHUFPSrmi VR128:$src1, addr:$src2, imm:$imm)>;
(SHUFPSrri VR128:$src1, VR128:$src2, imm:$imm)>;
}
-let Predicates = [HasSSE2] in {
+let Predicates = [UseSSE2] in {
// Generic SHUFPD patterns
def : Pat<(v2i64 (X86Shufp VR128:$src1,
(memopv2i64 addr:$src2), (i8 imm:$imm))),
defm VUNPCKHPSY: sse12_unpack_interleave<0x15, X86Unpckh, v8f32, memopv8f32,
VR256, f256mem, "unpckhps\t{$src2, $src1, $dst|$dst, $src1, $src2}",
- SSEPackedSingle>, TB, VEX_4V;
+ SSEPackedSingle>, TB, VEX_4V, VEX_L;
defm VUNPCKHPDY: sse12_unpack_interleave<0x15, X86Unpckh, v4f64, memopv4f64,
VR256, f256mem, "unpckhpd\t{$src2, $src1, $dst|$dst, $src1, $src2}",
- SSEPackedDouble>, TB, OpSize, VEX_4V;
+ SSEPackedDouble>, TB, OpSize, VEX_4V, VEX_L;
defm VUNPCKLPSY: sse12_unpack_interleave<0x14, X86Unpckl, v8f32, memopv8f32,
VR256, f256mem, "unpcklps\t{$src2, $src1, $dst|$dst, $src1, $src2}",
- SSEPackedSingle>, TB, VEX_4V;
+ SSEPackedSingle>, TB, VEX_4V, VEX_L;
defm VUNPCKLPDY: sse12_unpack_interleave<0x14, X86Unpckl, v4f64, memopv4f64,
VR256, f256mem, "unpcklpd\t{$src2, $src1, $dst|$dst, $src1, $src2}",
- SSEPackedDouble>, TB, OpSize, VEX_4V;
+ SSEPackedDouble>, TB, OpSize, VEX_4V, VEX_L;
let Constraints = "$src1 = $dst" in {
defm UNPCKHPS: sse12_unpack_interleave<0x15, X86Unpckh, v4f32, memopv4f32,
SSEPackedDouble>, TB, OpSize;
} // Constraints = "$src1 = $dst"
-let Predicates = [HasAVX], AddedComplexity = 1 in {
+let Predicates = [HasAVX1Only] in {
+ def : Pat<(v8i32 (X86Unpckl VR256:$src1, (bc_v8i32 (memopv4i64 addr:$src2)))),
+ (VUNPCKLPSYrm VR256:$src1, addr:$src2)>;
+ def : Pat<(v8i32 (X86Unpckl VR256:$src1, VR256:$src2)),
+ (VUNPCKLPSYrr VR256:$src1, VR256:$src2)>;
+ def : Pat<(v8i32 (X86Unpckh VR256:$src1, (bc_v8i32 (memopv4i64 addr:$src2)))),
+ (VUNPCKHPSYrm VR256:$src1, addr:$src2)>;
+ def : Pat<(v8i32 (X86Unpckh VR256:$src1, VR256:$src2)),
+ (VUNPCKHPSYrr VR256:$src1, VR256:$src2)>;
+
+ def : Pat<(v4i64 (X86Unpckl VR256:$src1, (memopv4i64 addr:$src2))),
+ (VUNPCKLPDYrm VR256:$src1, addr:$src2)>;
+ def : Pat<(v4i64 (X86Unpckl VR256:$src1, VR256:$src2)),
+ (VUNPCKLPDYrr VR256:$src1, VR256:$src2)>;
+ def : Pat<(v4i64 (X86Unpckh VR256:$src1, (memopv4i64 addr:$src2))),
+ (VUNPCKHPDYrm VR256:$src1, addr:$src2)>;
+ def : Pat<(v4i64 (X86Unpckh VR256:$src1, VR256:$src2)),
+ (VUNPCKHPDYrr VR256:$src1, VR256:$src2)>;
+}
+
+let Predicates = [HasAVX] in {
// FIXME: Instead of X86Movddup, there should be a X86Unpckl here, the
// problem is during lowering, where it's not possible to recognize the load
// fold cause it has two uses through a bitcast. One use disappears at isel
(VUNPCKLPDrr VR128:$src, VR128:$src)>;
}
-let Predicates = [HasSSE2] in {
+let Predicates = [UseSSE2] in {
// FIXME: Instead of X86Movddup, there should be a X86Unpckl here, the
// problem is during lowering, where it's not possible to recognize the load
// fold cause it has two uses through a bitcast. One use disappears at isel
"movmskpd", SSEPackedDouble>, TB,
OpSize, VEX;
defm VMOVMSKPSY : sse12_extr_sign_mask<VR256, int_x86_avx_movmsk_ps_256,
- "movmskps", SSEPackedSingle>, TB, VEX;
+ "movmskps", SSEPackedSingle>, TB,
+ VEX, VEX_L;
defm VMOVMSKPDY : sse12_extr_sign_mask<VR256, int_x86_avx_movmsk_pd_256,
"movmskpd", SSEPackedDouble>, TB,
- OpSize, VEX;
+ OpSize, VEX, VEX_L;
def : Pat<(i32 (X86fgetsign FR32:$src)),
(VMOVMSKPSrr32 (COPY_TO_REGCLASS FR32:$src, VR128))>;
OpSize, VEX;
def VMOVMSKPSYr64r : PI<0x50, MRMSrcReg, (outs GR64:$dst), (ins VR256:$src),
"movmskps\t{$src, $dst|$dst, $src}", [], IIC_SSE_MOVMSK,
- SSEPackedSingle>, TB, VEX;
+ SSEPackedSingle>, TB, VEX, VEX_L;
def VMOVMSKPDYr64r : PI<0x50, MRMSrcReg, (outs GR64:$dst), (ins VR256:$src),
"movmskpd\t{$src, $dst|$dst, $src}", [], IIC_SSE_MOVMSK,
SSEPackedDouble>, TB,
- OpSize, VEX;
+ OpSize, VEX, VEX_L;
}
defm MOVMSKPS : sse12_extr_sign_mask<VR128, int_x86_sse_movmsk_ps, "movmskps",
def : Pat<(i32 (X86fgetsign FR32:$src)),
(MOVMSKPSrr32 (COPY_TO_REGCLASS FR32:$src, VR128))>,
- Requires<[HasSSE1]>;
+ Requires<[UseSSE1]>;
def : Pat<(i64 (X86fgetsign FR32:$src)),
(MOVMSKPSrr64 (COPY_TO_REGCLASS FR32:$src, VR128))>,
- Requires<[HasSSE1]>;
+ Requires<[UseSSE1]>;
def : Pat<(i32 (X86fgetsign FR64:$src)),
(MOVMSKPDrr32 (COPY_TO_REGCLASS FR64:$src, VR128))>,
- Requires<[HasSSE2]>;
+ Requires<[UseSSE2]>;
def : Pat<(i64 (X86fgetsign FR64:$src)),
(MOVMSKPDrr64 (COPY_TO_REGCLASS FR64:$src, VR128))>,
- Requires<[HasSSE2]>;
+ Requires<[UseSSE2]>;
//===---------------------------------------------------------------------===//
// SSE2 - Packed Integer Logical Instructions
/// PDI_binop_rm - Simple SSE2 binary operator.
multiclass PDI_binop_rm<bits<8> opc, string OpcodeStr, SDNode OpNode,
ValueType OpVT, RegisterClass RC, PatFrag memop_frag,
- X86MemOperand x86memop,
- OpndItins itins,
- bit IsCommutable = 0,
- bit Is2Addr = 1> {
+ X86MemOperand x86memop, OpndItins itins,
+ bit IsCommutable, bit Is2Addr> {
let isCommutable = IsCommutable in
def rr : PDI<opc, MRMSrcReg, (outs RC:$dst),
(ins RC:$src1, RC:$src2),
}
} // ExeDomain = SSEPackedInt
-// These are ordered here for pattern ordering requirements with the fp versions
+multiclass PDI_binop_all<bits<8> opc, string OpcodeStr, SDNode Opcode,
+ ValueType OpVT128, ValueType OpVT256,
+ OpndItins itins, bit IsCommutable = 0> {
+let Predicates = [HasAVX] in
+ defm V#NAME : PDI_binop_rm<opc, !strconcat("v", OpcodeStr), Opcode, OpVT128,
+ VR128, memopv2i64, i128mem, itins, IsCommutable, 0>, VEX_4V;
-let Predicates = [HasAVX] in {
-defm VPAND : PDI_binop_rm<0xDB, "vpand", and, v2i64, VR128, memopv2i64,
- i128mem, SSE_BIT_ITINS_P, 1, 0>, VEX_4V;
-defm VPOR : PDI_binop_rm<0xEB, "vpor" , or, v2i64, VR128, memopv2i64,
- i128mem, SSE_BIT_ITINS_P, 1, 0>, VEX_4V;
-defm VPXOR : PDI_binop_rm<0xEF, "vpxor", xor, v2i64, VR128, memopv2i64,
- i128mem, SSE_BIT_ITINS_P, 1, 0>, VEX_4V;
-defm VPANDN : PDI_binop_rm<0xDF, "vpandn", X86andnp, v2i64, VR128, memopv2i64,
- i128mem, SSE_BIT_ITINS_P, 0, 0>, VEX_4V;
+let Constraints = "$src1 = $dst" in
+ defm NAME : PDI_binop_rm<opc, OpcodeStr, Opcode, OpVT128, VR128,
+ memopv2i64, i128mem, itins, IsCommutable, 1>;
+
+let Predicates = [HasAVX2] in
+ defm V#NAME#Y : PDI_binop_rm<opc, !strconcat("v", OpcodeStr), Opcode,
+ OpVT256, VR256, memopv4i64, i256mem, itins,
+ IsCommutable, 0>, VEX_4V, VEX_L;
}
-let Constraints = "$src1 = $dst" in {
-defm PAND : PDI_binop_rm<0xDB, "pand", and, v2i64, VR128, memopv2i64,
- i128mem, SSE_BIT_ITINS_P, 1>;
-defm POR : PDI_binop_rm<0xEB, "por" , or, v2i64, VR128, memopv2i64,
- i128mem, SSE_BIT_ITINS_P, 1>;
-defm PXOR : PDI_binop_rm<0xEF, "pxor", xor, v2i64, VR128, memopv2i64,
- i128mem, SSE_BIT_ITINS_P, 1>;
-defm PANDN : PDI_binop_rm<0xDF, "pandn", X86andnp, v2i64, VR128, memopv2i64,
- i128mem, SSE_BIT_ITINS_P, 0>;
-} // Constraints = "$src1 = $dst"
+// These are ordered here for pattern ordering requirements with the fp versions
-let Predicates = [HasAVX2] in {
-defm VPANDY : PDI_binop_rm<0xDB, "vpand", and, v4i64, VR256, memopv4i64,
- i256mem, SSE_BIT_ITINS_P, 1, 0>, VEX_4V;
-defm VPORY : PDI_binop_rm<0xEB, "vpor", or, v4i64, VR256, memopv4i64,
- i256mem, SSE_BIT_ITINS_P, 1, 0>, VEX_4V;
-defm VPXORY : PDI_binop_rm<0xEF, "vpxor", xor, v4i64, VR256, memopv4i64,
- i256mem, SSE_BIT_ITINS_P, 1, 0>, VEX_4V;
-defm VPANDNY : PDI_binop_rm<0xDF, "vpandn", X86andnp, v4i64, VR256, memopv4i64,
- i256mem, SSE_BIT_ITINS_P, 0, 0>, VEX_4V;
-}
+defm PAND : PDI_binop_all<0xDB, "pand", and, v2i64, v4i64, SSE_BIT_ITINS_P, 1>;
+defm POR : PDI_binop_all<0xEB, "por", or, v2i64, v4i64, SSE_BIT_ITINS_P, 1>;
+defm PXOR : PDI_binop_all<0xEF, "pxor", xor, v2i64, v4i64, SSE_BIT_ITINS_P, 1>;
+defm PANDN : PDI_binop_all<0xDF, "pandn", X86andnp, v2i64, v4i64,
+ SSE_BIT_ITINS_P, 0>;
//===----------------------------------------------------------------------===//
// SSE 1 & 2 - Logical Instructions
}
// Alias bitwise logical operations using SSE logical ops on packed FP values.
-let mayLoad = 0 in {
- defm FsAND : sse12_fp_alias_pack_logical<0x54, "and", X86fand,
- SSE_BIT_ITINS_P>;
- defm FsOR : sse12_fp_alias_pack_logical<0x56, "or", X86for,
- SSE_BIT_ITINS_P>;
- defm FsXOR : sse12_fp_alias_pack_logical<0x57, "xor", X86fxor,
- SSE_BIT_ITINS_P>;
-}
+defm FsAND : sse12_fp_alias_pack_logical<0x54, "and", X86fand,
+ SSE_BIT_ITINS_P>;
+defm FsOR : sse12_fp_alias_pack_logical<0x56, "or", X86for,
+ SSE_BIT_ITINS_P>;
+defm FsXOR : sse12_fp_alias_pack_logical<0x57, "xor", X86fxor,
+ SSE_BIT_ITINS_P>;
let neverHasSideEffects = 1, Pattern = []<dag>, isCommutable = 0 in
defm FsANDN : sse12_fp_alias_pack_logical<0x55, "andn", undef,
///
multiclass sse12_fp_packed_logical<bits<8> opc, string OpcodeStr,
SDNode OpNode> {
+ defm V#NAME#PSY : sse12_fp_packed_logical_rm<opc, VR256, SSEPackedSingle,
+ !strconcat(OpcodeStr, "ps"), f256mem,
+ [(set VR256:$dst, (v4i64 (OpNode VR256:$src1, VR256:$src2)))],
+ [(set VR256:$dst, (OpNode (bc_v4i64 (v8f32 VR256:$src1)),
+ (memopv4i64 addr:$src2)))], 0>, TB, VEX_4V, VEX_L;
+
+ defm V#NAME#PDY : sse12_fp_packed_logical_rm<opc, VR256, SSEPackedDouble,
+ !strconcat(OpcodeStr, "pd"), f256mem,
+ [(set VR256:$dst, (OpNode (bc_v4i64 (v4f64 VR256:$src1)),
+ (bc_v4i64 (v4f64 VR256:$src2))))],
+ [(set VR256:$dst, (OpNode (bc_v4i64 (v4f64 VR256:$src1)),
+ (memopv4i64 addr:$src2)))], 0>,
+ TB, OpSize, VEX_4V, VEX_L;
+
// In AVX no need to add a pattern for 128-bit logical rr ps, because they
// are all promoted to v2i64, and the patterns are covered by the int
// version. This is needed in SSE only, because v2i64 isn't supported on
defm V#NAME#PS : sse12_fp_packed_logical_rm<opc, VR128, SSEPackedSingle,
!strconcat(OpcodeStr, "ps"), f128mem, [],
[(set VR128:$dst, (OpNode (bc_v2i64 (v4f32 VR128:$src1)),
- (memopv2i64 addr:$src2)))], 0, 1>, TB, VEX_4V;
+ (memopv2i64 addr:$src2)))], 0>, TB, VEX_4V;
defm V#NAME#PD : sse12_fp_packed_logical_rm<opc, VR128, SSEPackedDouble,
!strconcat(OpcodeStr, "pd"), f128mem,
[(set VR128:$dst, (OpNode (bc_v2i64 (v2f64 VR128:$src1)),
(memopv2i64 addr:$src2)))], 0>,
TB, OpSize, VEX_4V;
+
let Constraints = "$src1 = $dst" in {
defm PS : sse12_fp_packed_logical_rm<opc, VR128, SSEPackedSingle,
!strconcat(OpcodeStr, "ps"), f128mem,
}
}
-/// sse12_fp_packed_logical_y - AVX 256-bit SSE 1 & 2 logical ops forms
-///
-multiclass sse12_fp_packed_logical_y<bits<8> opc, string OpcodeStr,
- SDNode OpNode> {
- defm PSY : sse12_fp_packed_logical_rm<opc, VR256, SSEPackedSingle,
- !strconcat(OpcodeStr, "ps"), f256mem,
- [(set VR256:$dst, (v4i64 (OpNode VR256:$src1, VR256:$src2)))],
- [(set VR256:$dst, (OpNode (bc_v4i64 (v8f32 VR256:$src1)),
- (memopv4i64 addr:$src2)))], 0>, TB, VEX_4V;
-
- defm PDY : sse12_fp_packed_logical_rm<opc, VR256, SSEPackedDouble,
- !strconcat(OpcodeStr, "pd"), f256mem,
- [(set VR256:$dst, (OpNode (bc_v4i64 (v4f64 VR256:$src1)),
- (bc_v4i64 (v4f64 VR256:$src2))))],
- [(set VR256:$dst, (OpNode (bc_v4i64 (v4f64 VR256:$src1)),
- (memopv4i64 addr:$src2)))], 0>,
- TB, OpSize, VEX_4V;
-}
-
-// AVX 256-bit packed logical ops forms
-defm VAND : sse12_fp_packed_logical_y<0x54, "and", and>;
-defm VOR : sse12_fp_packed_logical_y<0x56, "or", or>;
-defm VXOR : sse12_fp_packed_logical_y<0x57, "xor", xor>;
-defm VANDN : sse12_fp_packed_logical_y<0x55, "andn", X86andnp>;
-
defm AND : sse12_fp_packed_logical<0x54, "and", and>;
defm OR : sse12_fp_packed_logical<0x56, "or", or>;
defm XOR : sse12_fp_packed_logical<0x57, "xor", xor>;
itins.d, Is2Addr>, XD;
}
-multiclass basic_sse12_fp_binop_p<bits<8> opc, string OpcodeStr, SDNode OpNode,
- SizeItins itins,
- bit Is2Addr = 1> {
- let mayLoad = 0 in {
+multiclass basic_sse12_fp_binop_p<bits<8> opc, string OpcodeStr,
+ SDNode OpNode, SizeItins itins> {
+let Predicates = [HasAVX] in {
+ defm V#NAME#PS : sse12_fp_packed<opc, !strconcat(OpcodeStr, "ps"), OpNode,
+ VR128, v4f32, f128mem, memopv4f32,
+ SSEPackedSingle, itins.s, 0>, TB, VEX_4V;
+ defm V#NAME#PD : sse12_fp_packed<opc, !strconcat(OpcodeStr, "pd"), OpNode,
+ VR128, v2f64, f128mem, memopv2f64,
+ SSEPackedDouble, itins.d, 0>, TB, OpSize, VEX_4V;
+
+ defm V#NAME#PSY : sse12_fp_packed<opc, !strconcat(OpcodeStr, "ps"),
+ OpNode, VR256, v8f32, f256mem, memopv8f32,
+ SSEPackedSingle, itins.s, 0>, TB, VEX_4V, VEX_L;
+ defm V#NAME#PDY : sse12_fp_packed<opc, !strconcat(OpcodeStr, "pd"),
+ OpNode, VR256, v4f64, f256mem, memopv4f64,
+ SSEPackedDouble, itins.d, 0>, TB, OpSize, VEX_4V, VEX_L;
+}
+
+let Constraints = "$src1 = $dst" in {
defm PS : sse12_fp_packed<opc, !strconcat(OpcodeStr, "ps"), OpNode, VR128,
- v4f32, f128mem, memopv4f32, SSEPackedSingle, itins.s, Is2Addr>,
- TB;
+ v4f32, f128mem, memopv4f32, SSEPackedSingle,
+ itins.s, 1>, TB;
defm PD : sse12_fp_packed<opc, !strconcat(OpcodeStr, "pd"), OpNode, VR128,
- v2f64, f128mem, memopv2f64, SSEPackedDouble, itins.d, Is2Addr>,
- TB, OpSize;
- }
+ v2f64, f128mem, memopv2f64, SSEPackedDouble,
+ itins.d, 1>, TB, OpSize;
}
-
-multiclass basic_sse12_fp_binop_p_y<bits<8> opc, string OpcodeStr,
- SDNode OpNode,
- SizeItins itins> {
- let mayLoad = 0 in {
- defm PSY : sse12_fp_packed<opc, !strconcat(OpcodeStr, "ps"), OpNode, VR256,
- v8f32, f256mem, memopv8f32, SSEPackedSingle, itins.s, 0>,
- TB;
- defm PDY : sse12_fp_packed<opc, !strconcat(OpcodeStr, "pd"), OpNode, VR256,
- v4f64, f256mem, memopv4f64, SSEPackedDouble, itins.d, 0>,
- TB, OpSize;
- }
}
multiclass basic_sse12_fp_binop_s_int<bits<8> opc, string OpcodeStr,
itins.d, Is2Addr>, XD;
}
-multiclass basic_sse12_fp_binop_p_int<bits<8> opc, string OpcodeStr,
- SizeItins itins,
- bit Is2Addr = 1> {
- defm PS : sse12_fp_packed_int<opc, OpcodeStr, VR128,
- !strconcat(OpcodeStr, "ps"), "sse", "_ps", f128mem, memopv4f32,
- SSEPackedSingle, itins.s, Is2Addr>,
- TB;
-
- defm PD : sse12_fp_packed_int<opc, OpcodeStr, VR128,
- !strconcat(OpcodeStr, "pd"), "sse2", "_pd", f128mem, memopv2f64,
- SSEPackedDouble, itins.d, Is2Addr>,
- TB, OpSize;
+// Binary Arithmetic instructions
+defm ADD : basic_sse12_fp_binop_p<0x58, "add", fadd, SSE_ALU_ITINS_P>;
+defm MUL : basic_sse12_fp_binop_p<0x59, "mul", fmul, SSE_MUL_ITINS_P>;
+let isCommutable = 0 in {
+ defm SUB : basic_sse12_fp_binop_p<0x5C, "sub", fsub, SSE_ALU_ITINS_P>;
+ defm DIV : basic_sse12_fp_binop_p<0x5E, "div", fdiv, SSE_DIV_ITINS_P>;
+ defm MAX : basic_sse12_fp_binop_p<0x5F, "max", X86fmax, SSE_ALU_ITINS_P>;
+ defm MIN : basic_sse12_fp_binop_p<0x5D, "min", X86fmin, SSE_ALU_ITINS_P>;
}
-multiclass basic_sse12_fp_binop_p_y_int<bits<8> opc, string OpcodeStr,
- SizeItins itins> {
- defm PSY : sse12_fp_packed_int<opc, OpcodeStr, VR256,
- !strconcat(OpcodeStr, "ps"), "avx", "_ps_256", f256mem, memopv8f32,
- SSEPackedSingle, itins.s, 0>, TB;
-
- defm PDY : sse12_fp_packed_int<opc, OpcodeStr, VR256,
- !strconcat(OpcodeStr, "pd"), "avx", "_pd_256", f256mem, memopv4f64,
- SSEPackedDouble, itins.d, 0>, TB, OpSize;
+let isCodeGenOnly = 1 in {
+ defm MAXC: basic_sse12_fp_binop_p<0x5F, "max", X86fmaxc, SSE_ALU_ITINS_P>;
+ defm MINC: basic_sse12_fp_binop_p<0x5D, "min", X86fminc, SSE_ALU_ITINS_P>;
}
-// Binary Arithmetic instructions
defm VADD : basic_sse12_fp_binop_s<0x58, "add", fadd, SSE_ALU_ITINS_S, 0>,
basic_sse12_fp_binop_s_int<0x58, "add", SSE_ALU_ITINS_S, 0>,
VEX_4V, VEX_LIG;
-defm VADD : basic_sse12_fp_binop_p<0x58, "add", fadd, SSE_ALU_ITINS_P, 0>,
- basic_sse12_fp_binop_p_y<0x58, "add", fadd, SSE_ALU_ITINS_P>,
- VEX_4V;
defm VMUL : basic_sse12_fp_binop_s<0x59, "mul", fmul, SSE_MUL_ITINS_S, 0>,
basic_sse12_fp_binop_s_int<0x59, "mul", SSE_MUL_ITINS_S, 0>,
VEX_4V, VEX_LIG;
-defm VMUL : basic_sse12_fp_binop_p<0x59, "mul", fmul, SSE_MUL_ITINS_P, 0>,
- basic_sse12_fp_binop_p_y<0x59, "mul", fmul, SSE_MUL_ITINS_P>,
- VEX_4V;
let isCommutable = 0 in {
defm VSUB : basic_sse12_fp_binop_s<0x5C, "sub", fsub, SSE_ALU_ITINS_S, 0>,
basic_sse12_fp_binop_s_int<0x5C, "sub", SSE_ALU_ITINS_S, 0>,
VEX_4V, VEX_LIG;
- defm VSUB : basic_sse12_fp_binop_p<0x5C, "sub", fsub, SSE_ALU_ITINS_P, 0>,
- basic_sse12_fp_binop_p_y<0x5C, "sub", fsub, SSE_ALU_ITINS_P>, VEX_4V;
defm VDIV : basic_sse12_fp_binop_s<0x5E, "div", fdiv, SSE_DIV_ITINS_S, 0>,
basic_sse12_fp_binop_s_int<0x5E, "div", SSE_DIV_ITINS_S, 0>,
VEX_4V, VEX_LIG;
- defm VDIV : basic_sse12_fp_binop_p<0x5E, "div", fdiv, SSE_ALU_ITINS_P, 0>,
- basic_sse12_fp_binop_p_y<0x5E, "div", fdiv, SSE_DIV_ITINS_P>,
- VEX_4V;
defm VMAX : basic_sse12_fp_binop_s<0x5F, "max", X86fmax, SSE_ALU_ITINS_S, 0>,
basic_sse12_fp_binop_s_int<0x5F, "max", SSE_ALU_ITINS_S, 0>,
VEX_4V, VEX_LIG;
- defm VMAX : basic_sse12_fp_binop_p<0x5F, "max", X86fmax, SSE_ALU_ITINS_P, 0>,
- basic_sse12_fp_binop_p_int<0x5F, "max", SSE_ALU_ITINS_P, 0>,
- basic_sse12_fp_binop_p_y<0x5F, "max", X86fmax, SSE_ALU_ITINS_P>,
- basic_sse12_fp_binop_p_y_int<0x5F, "max", SSE_ALU_ITINS_P>,
- VEX_4V;
defm VMIN : basic_sse12_fp_binop_s<0x5D, "min", X86fmin, SSE_ALU_ITINS_S, 0>,
basic_sse12_fp_binop_s_int<0x5D, "min", SSE_ALU_ITINS_S, 0>,
VEX_4V, VEX_LIG;
- defm VMIN : basic_sse12_fp_binop_p<0x5D, "min", X86fmin, SSE_ALU_ITINS_P, 0>,
- basic_sse12_fp_binop_p_int<0x5D, "min", SSE_ALU_ITINS_P, 0>,
- basic_sse12_fp_binop_p_y_int<0x5D, "min", SSE_ALU_ITINS_P>,
- basic_sse12_fp_binop_p_y<0x5D, "min", X86fmin, SSE_ALU_ITINS_P>,
- VEX_4V;
}
let Constraints = "$src1 = $dst" in {
defm ADD : basic_sse12_fp_binop_s<0x58, "add", fadd, SSE_ALU_ITINS_S>,
- basic_sse12_fp_binop_p<0x58, "add", fadd, SSE_ALU_ITINS_P>,
basic_sse12_fp_binop_s_int<0x58, "add", SSE_ALU_ITINS_S>;
defm MUL : basic_sse12_fp_binop_s<0x59, "mul", fmul, SSE_MUL_ITINS_S>,
- basic_sse12_fp_binop_p<0x59, "mul", fmul, SSE_MUL_ITINS_P>,
basic_sse12_fp_binop_s_int<0x59, "mul", SSE_MUL_ITINS_S>;
let isCommutable = 0 in {
defm SUB : basic_sse12_fp_binop_s<0x5C, "sub", fsub, SSE_ALU_ITINS_S>,
- basic_sse12_fp_binop_p<0x5C, "sub", fsub, SSE_ALU_ITINS_P>,
basic_sse12_fp_binop_s_int<0x5C, "sub", SSE_ALU_ITINS_S>;
defm DIV : basic_sse12_fp_binop_s<0x5E, "div", fdiv, SSE_DIV_ITINS_S>,
- basic_sse12_fp_binop_p<0x5E, "div", fdiv, SSE_DIV_ITINS_P>,
basic_sse12_fp_binop_s_int<0x5E, "div", SSE_DIV_ITINS_S>;
defm MAX : basic_sse12_fp_binop_s<0x5F, "max", X86fmax, SSE_ALU_ITINS_S>,
- basic_sse12_fp_binop_p<0x5F, "max", X86fmax, SSE_ALU_ITINS_P>,
- basic_sse12_fp_binop_s_int<0x5F, "max", SSE_ALU_ITINS_S>,
- basic_sse12_fp_binop_p_int<0x5F, "max", SSE_ALU_ITINS_P>;
+ basic_sse12_fp_binop_s_int<0x5F, "max", SSE_ALU_ITINS_S>;
defm MIN : basic_sse12_fp_binop_s<0x5D, "min", X86fmin, SSE_ALU_ITINS_S>,
- basic_sse12_fp_binop_p<0x5D, "min", X86fmin, SSE_ALU_ITINS_P>,
- basic_sse12_fp_binop_s_int<0x5D, "min", SSE_ALU_ITINS_S>,
- basic_sse12_fp_binop_p_int<0x5D, "min", SSE_ALU_ITINS_P>;
+ basic_sse12_fp_binop_s_int<0x5D, "min", SSE_ALU_ITINS_S>;
+ }
+}
+
+let isCodeGenOnly = 1 in {
+ defm VMAXC: basic_sse12_fp_binop_s<0x5F, "max", X86fmaxc, SSE_ALU_ITINS_S, 0>,
+ VEX_4V, VEX_LIG;
+ defm VMINC: basic_sse12_fp_binop_s<0x5D, "min", X86fminc, SSE_ALU_ITINS_S, 0>,
+ VEX_4V, VEX_LIG;
+ let Constraints = "$src1 = $dst" in {
+ defm MAXC: basic_sse12_fp_binop_s<0x5F, "max", X86fmaxc, SSE_ALU_ITINS_S>;
+ defm MINC: basic_sse12_fp_binop_s<0x5D, "min", X86fminc, SSE_ALU_ITINS_S>;
}
}
/// sse1_fp_unop_s - SSE1 unops in scalar form.
multiclass sse1_fp_unop_s<bits<8> opc, string OpcodeStr,
SDNode OpNode, Intrinsic F32Int, OpndItins itins> {
+let Predicates = [HasAVX], hasSideEffects = 0 in {
+ def V#NAME#SSr : SSI<opc, MRMSrcReg, (outs FR32:$dst),
+ (ins FR32:$src1, FR32:$src2),
+ !strconcat("v", OpcodeStr,
+ "ss\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
+ []>, VEX_4V, VEX_LIG;
+ let mayLoad = 1 in {
+ def V#NAME#SSm : SSI<opc, MRMSrcMem, (outs FR32:$dst),
+ (ins FR32:$src1,f32mem:$src2),
+ !strconcat("v", OpcodeStr,
+ "ss\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
+ []>, VEX_4V, VEX_LIG;
+ def V#NAME#SSm_Int : SSI<opc, MRMSrcMem, (outs VR128:$dst),
+ (ins VR128:$src1, ssmem:$src2),
+ !strconcat("v", OpcodeStr,
+ "ss\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
+ []>, VEX_4V, VEX_LIG;
+ }
+}
+
def SSr : SSI<opc, MRMSrcReg, (outs FR32:$dst), (ins FR32:$src),
!strconcat(OpcodeStr, "ss\t{$src, $dst|$dst, $src}"),
[(set FR32:$dst, (OpNode FR32:$src))]>;
def SSm : I<opc, MRMSrcMem, (outs FR32:$dst), (ins f32mem:$src),
!strconcat(OpcodeStr, "ss\t{$src, $dst|$dst, $src}"),
[(set FR32:$dst, (OpNode (load addr:$src)))], itins.rm>, XS,
- Requires<[HasSSE1, OptForSize]>;
+ Requires<[UseSSE1, OptForSize]>;
def SSr_Int : SSI<opc, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
!strconcat(OpcodeStr, "ss\t{$src, $dst|$dst, $src}"),
[(set VR128:$dst, (F32Int VR128:$src))], itins.rr>;
[(set VR128:$dst, (F32Int sse_load_f32:$src))], itins.rm>;
}
-/// sse1_fp_unop_s_avx - AVX SSE1 unops in scalar form.
-multiclass sse1_fp_unop_s_avx<bits<8> opc, string OpcodeStr> {
- def SSr : SSI<opc, MRMSrcReg, (outs FR32:$dst), (ins FR32:$src1, FR32:$src2),
- !strconcat(OpcodeStr,
- "ss\t{$src2, $src1, $dst|$dst, $src1, $src2}"), []>;
- let mayLoad = 1 in
- def SSm : SSI<opc, MRMSrcMem, (outs FR32:$dst), (ins FR32:$src1,f32mem:$src2),
- !strconcat(OpcodeStr,
- "ss\t{$src2, $src1, $dst|$dst, $src1, $src2}"), []>;
- def SSm_Int : SSI<opc, MRMSrcMem, (outs VR128:$dst),
- (ins VR128:$src1, ssmem:$src2),
- !strconcat(OpcodeStr,
- "ss\t{$src2, $src1, $dst|$dst, $src1, $src2}"), []>;
+/// sse1_fp_unop_s_rw - SSE1 unops where vector form has a read-write operand.
+multiclass sse1_fp_unop_rw<bits<8> opc, string OpcodeStr, SDNode OpNode,
+ OpndItins itins> {
+let Predicates = [HasAVX], hasSideEffects = 0 in {
+ def V#NAME#SSr : SSI<opc, MRMSrcReg, (outs FR32:$dst),
+ (ins FR32:$src1, FR32:$src2),
+ !strconcat("v", OpcodeStr,
+ "ss\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
+ []>, VEX_4V, VEX_LIG;
+ let mayLoad = 1 in {
+ def V#NAME#SSm : SSI<opc, MRMSrcMem, (outs FR32:$dst),
+ (ins FR32:$src1,f32mem:$src2),
+ !strconcat("v", OpcodeStr,
+ "ss\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
+ []>, VEX_4V, VEX_LIG;
+ def V#NAME#SSm_Int : SSI<opc, MRMSrcMem, (outs VR128:$dst),
+ (ins VR128:$src1, ssmem:$src2),
+ !strconcat("v", OpcodeStr,
+ "ss\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
+ []>, VEX_4V, VEX_LIG;
+ }
+}
+
+ def SSr : SSI<opc, MRMSrcReg, (outs FR32:$dst), (ins FR32:$src),
+ !strconcat(OpcodeStr, "ss\t{$src, $dst|$dst, $src}"),
+ [(set FR32:$dst, (OpNode FR32:$src))]>;
+ // For scalar unary operations, fold a load into the operation
+ // only in OptForSize mode. It eliminates an instruction, but it also
+ // eliminates a whole-register clobber (the load), so it introduces a
+ // partial register update condition.
+ def SSm : I<opc, MRMSrcMem, (outs FR32:$dst), (ins f32mem:$src),
+ !strconcat(OpcodeStr, "ss\t{$src, $dst|$dst, $src}"),
+ [(set FR32:$dst, (OpNode (load addr:$src)))], itins.rm>, XS,
+ Requires<[UseSSE1, OptForSize]>;
+ let Constraints = "$src1 = $dst" in {
+ def SSr_Int : SSI<opc, MRMSrcReg, (outs VR128:$dst),
+ (ins VR128:$src1, VR128:$src2),
+ !strconcat(OpcodeStr, "ss\t{$src2, $dst|$dst, $src2}"),
+ [], itins.rr>;
+ let mayLoad = 1, hasSideEffects = 0 in
+ def SSm_Int : SSI<opc, MRMSrcMem, (outs VR128:$dst),
+ (ins VR128:$src1, ssmem:$src2),
+ !strconcat(OpcodeStr, "ss\t{$src2, $dst|$dst, $src2}"),
+ [], itins.rm>;
+ }
}
/// sse1_fp_unop_p - SSE1 unops in packed form.
multiclass sse1_fp_unop_p<bits<8> opc, string OpcodeStr, SDNode OpNode,
OpndItins itins> {
+let Predicates = [HasAVX] in {
+ def V#NAME#PSr : PSI<opc, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
+ !strconcat("v", OpcodeStr,
+ "ps\t{$src, $dst|$dst, $src}"),
+ [(set VR128:$dst, (v4f32 (OpNode VR128:$src)))],
+ itins.rr>, VEX;
+ def V#NAME#PSm : PSI<opc, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
+ !strconcat("v", OpcodeStr,
+ "ps\t{$src, $dst|$dst, $src}"),
+ [(set VR128:$dst, (OpNode (memopv4f32 addr:$src)))],
+ itins.rm>, VEX;
+ def V#NAME#PSYr : PSI<opc, MRMSrcReg, (outs VR256:$dst), (ins VR256:$src),
+ !strconcat("v", OpcodeStr,
+ "ps\t{$src, $dst|$dst, $src}"),
+ [(set VR256:$dst, (v8f32 (OpNode VR256:$src)))],
+ itins.rr>, VEX, VEX_L;
+ def V#NAME#PSYm : PSI<opc, MRMSrcMem, (outs VR256:$dst), (ins f256mem:$src),
+ !strconcat("v", OpcodeStr,
+ "ps\t{$src, $dst|$dst, $src}"),
+ [(set VR256:$dst, (OpNode (memopv8f32 addr:$src)))],
+ itins.rm>, VEX, VEX_L;
+}
+
def PSr : PSI<opc, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
- !strconcat(OpcodeStr, "ps\t{$src, $dst|$dst, $src}"),
- [(set VR128:$dst, (v4f32 (OpNode VR128:$src)))], itins.rr>;
+ !strconcat(OpcodeStr, "ps\t{$src, $dst|$dst, $src}"),
+ [(set VR128:$dst, (v4f32 (OpNode VR128:$src)))], itins.rr>;
def PSm : PSI<opc, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
!strconcat(OpcodeStr, "ps\t{$src, $dst|$dst, $src}"),
[(set VR128:$dst, (OpNode (memopv4f32 addr:$src)))], itins.rm>;
}
-/// sse1_fp_unop_p_y - AVX 256-bit SSE1 unops in packed form.
-multiclass sse1_fp_unop_p_y<bits<8> opc, string OpcodeStr, SDNode OpNode,
- OpndItins itins> {
- def PSYr : PSI<opc, MRMSrcReg, (outs VR256:$dst), (ins VR256:$src),
- !strconcat(OpcodeStr, "ps\t{$src, $dst|$dst, $src}"),
- [(set VR256:$dst, (v8f32 (OpNode VR256:$src)))],
- itins.rr>;
- def PSYm : PSI<opc, MRMSrcMem, (outs VR256:$dst), (ins f256mem:$src),
- !strconcat(OpcodeStr, "ps\t{$src, $dst|$dst, $src}"),
- [(set VR256:$dst, (OpNode (memopv8f32 addr:$src)))],
- itins.rm>;
-}
-
/// sse1_fp_unop_p_int - SSE1 intrinsics unops in packed forms.
multiclass sse1_fp_unop_p_int<bits<8> opc, string OpcodeStr,
- Intrinsic V4F32Int, OpndItins itins> {
+ Intrinsic V4F32Int, Intrinsic V8F32Int,
+ OpndItins itins> {
+let Predicates = [HasAVX] in {
+ def V#NAME#PSr_Int : PSI<opc, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
+ !strconcat("v", OpcodeStr,
+ "ps\t{$src, $dst|$dst, $src}"),
+ [(set VR128:$dst, (V4F32Int VR128:$src))],
+ itins.rr>, VEX;
+ def V#NAME#PSm_Int : PSI<opc, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
+ !strconcat("v", OpcodeStr,
+ "ps\t{$src, $dst|$dst, $src}"),
+ [(set VR128:$dst, (V4F32Int (memopv4f32 addr:$src)))],
+ itins.rm>, VEX;
+ def V#NAME#PSYr_Int : PSI<opc, MRMSrcReg, (outs VR256:$dst), (ins VR256:$src),
+ !strconcat("v", OpcodeStr,
+ "ps\t{$src, $dst|$dst, $src}"),
+ [(set VR256:$dst, (V8F32Int VR256:$src))],
+ itins.rr>, VEX, VEX_L;
+ def V#NAME#PSYm_Int : PSI<opc, MRMSrcMem, (outs VR256:$dst),
+ (ins f256mem:$src),
+ !strconcat("v", OpcodeStr,
+ "ps\t{$src, $dst|$dst, $src}"),
+ [(set VR256:$dst, (V8F32Int (memopv8f32 addr:$src)))],
+ itins.rm>, VEX, VEX_L;
+}
+
def PSr_Int : PSI<opc, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
!strconcat(OpcodeStr, "ps\t{$src, $dst|$dst, $src}"),
[(set VR128:$dst, (V4F32Int VR128:$src))],
itins.rm>;
}
-/// sse1_fp_unop_p_y_int - AVX 256-bit intrinsics unops in packed forms.
-multiclass sse1_fp_unop_p_y_int<bits<8> opc, string OpcodeStr,
- Intrinsic V4F32Int, OpndItins itins> {
- def PSYr_Int : PSI<opc, MRMSrcReg, (outs VR256:$dst), (ins VR256:$src),
- !strconcat(OpcodeStr, "ps\t{$src, $dst|$dst, $src}"),
- [(set VR256:$dst, (V4F32Int VR256:$src))],
- itins.rr>;
- def PSYm_Int : PSI<opc, MRMSrcMem, (outs VR256:$dst), (ins f256mem:$src),
- !strconcat(OpcodeStr, "ps\t{$src, $dst|$dst, $src}"),
- [(set VR256:$dst, (V4F32Int (memopv8f32 addr:$src)))],
- itins.rm>;
-}
-
/// sse2_fp_unop_s - SSE2 unops in scalar form.
multiclass sse2_fp_unop_s<bits<8> opc, string OpcodeStr,
SDNode OpNode, Intrinsic F64Int, OpndItins itins> {
+let Predicates = [HasAVX], hasSideEffects = 0 in {
+ def V#NAME#SDr : SDI<opc, MRMSrcReg, (outs FR64:$dst),
+ (ins FR64:$src1, FR64:$src2),
+ !strconcat("v", OpcodeStr,
+ "sd\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
+ []>, VEX_4V, VEX_LIG;
+ let mayLoad = 1 in {
+ def V#NAME#SDm : SDI<opc, MRMSrcMem, (outs FR64:$dst),
+ (ins FR64:$src1,f64mem:$src2),
+ !strconcat("v", OpcodeStr,
+ "sd\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
+ []>, VEX_4V, VEX_LIG;
+ def V#NAME#SDm_Int : SDI<opc, MRMSrcMem, (outs VR128:$dst),
+ (ins VR128:$src1, sdmem:$src2),
+ !strconcat("v", OpcodeStr,
+ "sd\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
+ []>, VEX_4V, VEX_LIG;
+ }
+}
+
def SDr : SDI<opc, MRMSrcReg, (outs FR64:$dst), (ins FR64:$src),
!strconcat(OpcodeStr, "sd\t{$src, $dst|$dst, $src}"),
[(set FR64:$dst, (OpNode FR64:$src))], itins.rr>;
def SDm : I<opc, MRMSrcMem, (outs FR64:$dst), (ins f64mem:$src),
!strconcat(OpcodeStr, "sd\t{$src, $dst|$dst, $src}"),
[(set FR64:$dst, (OpNode (load addr:$src)))], itins.rm>, XD,
- Requires<[HasSSE2, OptForSize]>;
+ Requires<[UseSSE2, OptForSize]>;
def SDr_Int : SDI<opc, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
!strconcat(OpcodeStr, "sd\t{$src, $dst|$dst, $src}"),
[(set VR128:$dst, (F64Int VR128:$src))], itins.rr>;
[(set VR128:$dst, (F64Int sse_load_f64:$src))], itins.rm>;
}
-/// sse2_fp_unop_s_avx - AVX SSE2 unops in scalar form.
-multiclass sse2_fp_unop_s_avx<bits<8> opc, string OpcodeStr> {
- let neverHasSideEffects = 1 in {
- def SDr : SDI<opc, MRMSrcReg, (outs FR64:$dst), (ins FR64:$src1, FR64:$src2),
- !strconcat(OpcodeStr,
- "sd\t{$src2, $src1, $dst|$dst, $src1, $src2}"), []>;
- let mayLoad = 1 in
- def SDm : SDI<opc, MRMSrcMem, (outs FR64:$dst), (ins FR64:$src1,f64mem:$src2),
- !strconcat(OpcodeStr,
- "sd\t{$src2, $src1, $dst|$dst, $src1, $src2}"), []>;
- }
- def SDm_Int : SDI<opc, MRMSrcMem, (outs VR128:$dst),
- (ins VR128:$src1, sdmem:$src2),
- !strconcat(OpcodeStr,
- "sd\t{$src2, $src1, $dst|$dst, $src1, $src2}"), []>;
-}
-
/// sse2_fp_unop_p - SSE2 unops in vector forms.
multiclass sse2_fp_unop_p<bits<8> opc, string OpcodeStr,
SDNode OpNode, OpndItins itins> {
+let Predicates = [HasAVX] in {
+ def V#NAME#PDr : PDI<opc, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
+ !strconcat("v", OpcodeStr,
+ "pd\t{$src, $dst|$dst, $src}"),
+ [(set VR128:$dst, (v2f64 (OpNode VR128:$src)))],
+ itins.rr>, VEX;
+ def V#NAME#PDm : PDI<opc, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
+ !strconcat("v", OpcodeStr,
+ "pd\t{$src, $dst|$dst, $src}"),
+ [(set VR128:$dst, (OpNode (memopv2f64 addr:$src)))],
+ itins.rm>, VEX;
+ def V#NAME#PDYr : PDI<opc, MRMSrcReg, (outs VR256:$dst), (ins VR256:$src),
+ !strconcat("v", OpcodeStr,
+ "pd\t{$src, $dst|$dst, $src}"),
+ [(set VR256:$dst, (v4f64 (OpNode VR256:$src)))],
+ itins.rr>, VEX, VEX_L;
+ def V#NAME#PDYm : PDI<opc, MRMSrcMem, (outs VR256:$dst), (ins f256mem:$src),
+ !strconcat("v", OpcodeStr,
+ "pd\t{$src, $dst|$dst, $src}"),
+ [(set VR256:$dst, (OpNode (memopv4f64 addr:$src)))],
+ itins.rm>, VEX, VEX_L;
+}
+
def PDr : PDI<opc, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
!strconcat(OpcodeStr, "pd\t{$src, $dst|$dst, $src}"),
[(set VR128:$dst, (v2f64 (OpNode VR128:$src)))], itins.rr>;
[(set VR128:$dst, (OpNode (memopv2f64 addr:$src)))], itins.rm>;
}
-/// sse2_fp_unop_p_y - AVX SSE2 256-bit unops in vector forms.
-multiclass sse2_fp_unop_p_y<bits<8> opc, string OpcodeStr, SDNode OpNode,
- OpndItins itins> {
- def PDYr : PDI<opc, MRMSrcReg, (outs VR256:$dst), (ins VR256:$src),
- !strconcat(OpcodeStr, "pd\t{$src, $dst|$dst, $src}"),
- [(set VR256:$dst, (v4f64 (OpNode VR256:$src)))],
- itins.rr>;
- def PDYm : PDI<opc, MRMSrcMem, (outs VR256:$dst), (ins f256mem:$src),
- !strconcat(OpcodeStr, "pd\t{$src, $dst|$dst, $src}"),
- [(set VR256:$dst, (OpNode (memopv4f64 addr:$src)))],
- itins.rm>;
-}
-
-/// sse2_fp_unop_p_int - SSE2 intrinsic unops in vector forms.
-multiclass sse2_fp_unop_p_int<bits<8> opc, string OpcodeStr,
- Intrinsic V2F64Int, OpndItins itins> {
- def PDr_Int : PDI<opc, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
- !strconcat(OpcodeStr, "pd\t{$src, $dst|$dst, $src}"),
- [(set VR128:$dst, (V2F64Int VR128:$src))],
- itins.rr>;
- def PDm_Int : PDI<opc, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
- !strconcat(OpcodeStr, "pd\t{$src, $dst|$dst, $src}"),
- [(set VR128:$dst, (V2F64Int (memopv2f64 addr:$src)))],
- itins.rm>;
-}
+// Square root.
+defm SQRT : sse1_fp_unop_s<0x51, "sqrt", fsqrt, int_x86_sse_sqrt_ss,
+ SSE_SQRTS>,
+ sse1_fp_unop_p<0x51, "sqrt", fsqrt, SSE_SQRTP>,
+ sse2_fp_unop_s<0x51, "sqrt", fsqrt, int_x86_sse2_sqrt_sd,
+ SSE_SQRTS>,
+ sse2_fp_unop_p<0x51, "sqrt", fsqrt, SSE_SQRTP>;
-/// sse2_fp_unop_p_y_int - AVX 256-bit intrinsic unops in vector forms.
-multiclass sse2_fp_unop_p_y_int<bits<8> opc, string OpcodeStr,
- Intrinsic V2F64Int, OpndItins itins> {
- def PDYr_Int : PDI<opc, MRMSrcReg, (outs VR256:$dst), (ins VR256:$src),
- !strconcat(OpcodeStr, "pd\t{$src, $dst|$dst, $src}"),
- [(set VR256:$dst, (V2F64Int VR256:$src))],
- itins.rr>;
- def PDYm_Int : PDI<opc, MRMSrcMem, (outs VR256:$dst), (ins f256mem:$src),
- !strconcat(OpcodeStr, "pd\t{$src, $dst|$dst, $src}"),
- [(set VR256:$dst, (V2F64Int (memopv4f64 addr:$src)))],
- itins.rm>;
-}
+// Reciprocal approximations. Note that these typically require refinement
+// in order to obtain suitable precision.
+defm RSQRT : sse1_fp_unop_rw<0x52, "rsqrt", X86frsqrt, SSE_SQRTS>,
+ sse1_fp_unop_p<0x52, "rsqrt", X86frsqrt, SSE_SQRTP>,
+ sse1_fp_unop_p_int<0x52, "rsqrt", int_x86_sse_rsqrt_ps,
+ int_x86_avx_rsqrt_ps_256, SSE_SQRTP>;
+defm RCP : sse1_fp_unop_rw<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>;
-let Predicates = [HasAVX] in {
- // Square root.
- defm VSQRT : sse1_fp_unop_s_avx<0x51, "vsqrt">,
- sse2_fp_unop_s_avx<0x51, "vsqrt">, VEX_4V, VEX_LIG;
-
- defm VSQRT : sse1_fp_unop_p<0x51, "vsqrt", fsqrt, SSE_SQRTP>,
- sse2_fp_unop_p<0x51, "vsqrt", fsqrt, SSE_SQRTP>,
- sse1_fp_unop_p_y<0x51, "vsqrt", fsqrt, SSE_SQRTP>,
- sse2_fp_unop_p_y<0x51, "vsqrt", fsqrt, SSE_SQRTP>,
- sse1_fp_unop_p_int<0x51, "vsqrt", int_x86_sse_sqrt_ps,
- SSE_SQRTP>,
- sse2_fp_unop_p_int<0x51, "vsqrt", int_x86_sse2_sqrt_pd,
- SSE_SQRTP>,
- sse1_fp_unop_p_y_int<0x51, "vsqrt", int_x86_avx_sqrt_ps_256,
- SSE_SQRTP>,
- sse2_fp_unop_p_y_int<0x51, "vsqrt", int_x86_avx_sqrt_pd_256,
- SSE_SQRTP>,
- VEX;
-
- // Reciprocal approximations. Note that these typically require refinement
- // in order to obtain suitable precision.
- defm VRSQRT : sse1_fp_unop_s_avx<0x52, "vrsqrt">, VEX_4V, VEX_LIG;
- defm VRSQRT : sse1_fp_unop_p<0x52, "vrsqrt", X86frsqrt, SSE_SQRTP>,
- sse1_fp_unop_p_y<0x52, "vrsqrt", X86frsqrt, SSE_SQRTP>,
- sse1_fp_unop_p_y_int<0x52, "vrsqrt", int_x86_avx_rsqrt_ps_256,
- SSE_SQRTP>,
- sse1_fp_unop_p_int<0x52, "vrsqrt", int_x86_sse_rsqrt_ps,
- SSE_SQRTP>, VEX;
-
- defm VRCP : sse1_fp_unop_s_avx<0x53, "vrcp">, VEX_4V, VEX_LIG;
- defm VRCP : sse1_fp_unop_p<0x53, "vrcp", X86frcp, SSE_RCPP>,
- sse1_fp_unop_p_y<0x53, "vrcp", X86frcp, SSE_RCPP>,
- sse1_fp_unop_p_y_int<0x53, "vrcp", int_x86_avx_rcp_ps_256,
- SSE_RCPP>,
- sse1_fp_unop_p_int<0x53, "vrcp", int_x86_sse_rcp_ps,
- SSE_RCPP>, VEX;
-}
-
-let AddedComplexity = 1 in {
def : Pat<(f32 (fsqrt FR32:$src)),
(VSQRTSSr (f32 (IMPLICIT_DEF)), FR32:$src)>, Requires<[HasAVX]>;
def : Pat<(f32 (fsqrt (load addr:$src))),
def : Pat<(f32 (X86frcp (load addr:$src))),
(VRCPSSm (f32 (IMPLICIT_DEF)), addr:$src)>,
Requires<[HasAVX, OptForSize]>;
-}
-let Predicates = [HasAVX], AddedComplexity = 1 in {
+let Predicates = [HasAVX] in {
def : Pat<(int_x86_sse_sqrt_ss VR128:$src),
(COPY_TO_REGCLASS (VSQRTSSr (f32 (IMPLICIT_DEF)),
(COPY_TO_REGCLASS VR128:$src, FR32)),
(VRCPSSm_Int (v4f32 (IMPLICIT_DEF)), sse_load_f32:$src)>;
}
-// Square root.
-defm SQRT : sse1_fp_unop_s<0x51, "sqrt", fsqrt, int_x86_sse_sqrt_ss,
- SSE_SQRTS>,
- sse1_fp_unop_p<0x51, "sqrt", fsqrt, SSE_SQRTS>,
- sse1_fp_unop_p_int<0x51, "sqrt", int_x86_sse_sqrt_ps, SSE_SQRTS>,
- sse2_fp_unop_s<0x51, "sqrt", fsqrt, int_x86_sse2_sqrt_sd,
- SSE_SQRTS>,
- sse2_fp_unop_p<0x51, "sqrt", fsqrt, SSE_SQRTS>,
- sse2_fp_unop_p_int<0x51, "sqrt", int_x86_sse2_sqrt_pd, SSE_SQRTS>;
-
// Reciprocal approximations. Note that these typically require refinement
// in order to obtain suitable precision.
-defm RSQRT : sse1_fp_unop_s<0x52, "rsqrt", X86frsqrt, int_x86_sse_rsqrt_ss,
- SSE_SQRTS>,
- sse1_fp_unop_p<0x52, "rsqrt", X86frsqrt, SSE_SQRTS>,
- sse1_fp_unop_p_int<0x52, "rsqrt", int_x86_sse_rsqrt_ps,
- SSE_SQRTS>;
-defm RCP : sse1_fp_unop_s<0x53, "rcp", X86frcp, int_x86_sse_rcp_ss,
- SSE_RCPS>,
- sse1_fp_unop_p<0x53, "rcp", X86frcp, SSE_RCPS>,
- sse1_fp_unop_p_int<0x53, "rcp", int_x86_sse_rcp_ps, SSE_RCPS>;
+let Predicates = [UseSSE1] in {
+ def : Pat<(int_x86_sse_rsqrt_ss VR128:$src),
+ (RSQRTSSr_Int VR128:$src, VR128:$src)>;
+ def : Pat<(int_x86_sse_rcp_ss VR128:$src),
+ (RCPSSr_Int VR128:$src, VR128:$src)>;
+}
// There is no f64 version of the reciprocal approximation instructions.
"movntps\t{$src, $dst|$dst, $src}",
[(alignednontemporalstore (v8f32 VR256:$src),
addr:$dst)],
- IIC_SSE_MOVNT>, VEX;
+ IIC_SSE_MOVNT>, VEX, VEX_L;
def VMOVNTPDYmr : VPDI<0x2B, MRMDestMem, (outs),
(ins f256mem:$dst, VR256:$src),
"movntpd\t{$src, $dst|$dst, $src}",
[(alignednontemporalstore (v4f64 VR256:$src),
addr:$dst)],
- IIC_SSE_MOVNT>, VEX;
+ IIC_SSE_MOVNT>, VEX, VEX_L;
let ExeDomain = SSEPackedInt in
def VMOVNTDQYmr : VPDI<0xE7, MRMDestMem, (outs),
(ins f256mem:$dst, VR256:$src),
"movntdq\t{$src, $dst|$dst, $src}",
[(alignednontemporalstore (v4i64 VR256:$src),
addr:$dst)],
- IIC_SSE_MOVNT>, VEX;
+ IIC_SSE_MOVNT>, VEX, VEX_L;
}
let AddedComplexity = 400 in { // Prefer non-temporal versions
IIC_SSE_MOVNT>;
def : Pat<(alignednontemporalstore (v2i64 VR128:$src), addr:$dst),
- (MOVNTDQmr addr:$dst, VR128:$src)>, Requires<[HasSSE2]>;
+ (MOVNTDQmr addr:$dst, VR128:$src)>, Requires<[UseSSE2]>;
// There is no AVX form for instructions below this point
def MOVNTImr : I<0xC3, MRMDestMem, (outs), (ins i32mem:$dst, GR32:$src),
VEX;
def VMOVDQAYrr : VPDI<0x6F, MRMSrcReg, (outs VR256:$dst), (ins VR256:$src),
"movdqa\t{$src, $dst|$dst, $src}", [], IIC_SSE_MOVA_P_RR>,
- VEX;
-}
+ VEX, VEX_L;
def VMOVDQUrr : VSSI<0x6F, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
"movdqu\t{$src, $dst|$dst, $src}", [], IIC_SSE_MOVU_P_RR>,
VEX;
def VMOVDQUYrr : VSSI<0x6F, MRMSrcReg, (outs VR256:$dst), (ins VR256:$src),
"movdqu\t{$src, $dst|$dst, $src}", [], IIC_SSE_MOVU_P_RR>,
- VEX;
+ VEX, VEX_L;
+}
// For Disassembler
-let isCodeGenOnly = 1 in {
+let isCodeGenOnly = 1, hasSideEffects = 0 in {
def VMOVDQArr_REV : VPDI<0x7F, MRMDestReg, (outs VR128:$dst), (ins VR128:$src),
"movdqa\t{$src, $dst|$dst, $src}", [],
IIC_SSE_MOVA_P_RR>,
VEX;
def VMOVDQAYrr_REV : VPDI<0x7F, MRMDestReg, (outs VR256:$dst), (ins VR256:$src),
"movdqa\t{$src, $dst|$dst, $src}", [],
- IIC_SSE_MOVA_P_RR>,
- VEX;
+ IIC_SSE_MOVA_P_RR>, VEX, VEX_L;
def VMOVDQUrr_REV : VSSI<0x7F, MRMDestReg, (outs VR128:$dst), (ins VR128:$src),
"movdqu\t{$src, $dst|$dst, $src}", [],
IIC_SSE_MOVU_P_RR>,
VEX;
def VMOVDQUYrr_REV : VSSI<0x7F, MRMDestReg, (outs VR256:$dst), (ins VR256:$src),
"movdqu\t{$src, $dst|$dst, $src}", [],
- IIC_SSE_MOVU_P_RR>,
- VEX;
+ IIC_SSE_MOVU_P_RR>, VEX, VEX_L;
}
-let canFoldAsLoad = 1, mayLoad = 1 in {
+let canFoldAsLoad = 1, mayLoad = 1, isReMaterializable = 1,
+ neverHasSideEffects = 1 in {
def VMOVDQArm : VPDI<0x6F, MRMSrcMem, (outs VR128:$dst), (ins i128mem:$src),
"movdqa\t{$src, $dst|$dst, $src}", [], IIC_SSE_MOVA_P_RM>,
VEX;
def VMOVDQAYrm : VPDI<0x6F, MRMSrcMem, (outs VR256:$dst), (ins i256mem:$src),
"movdqa\t{$src, $dst|$dst, $src}", [], IIC_SSE_MOVA_P_RM>,
- VEX;
+ VEX, VEX_L;
let Predicates = [HasAVX] in {
def VMOVDQUrm : I<0x6F, MRMSrcMem, (outs VR128:$dst), (ins i128mem:$src),
"vmovdqu\t{$src, $dst|$dst, $src}",[], IIC_SSE_MOVU_P_RM>,
XS, VEX;
def VMOVDQUYrm : I<0x6F, MRMSrcMem, (outs VR256:$dst), (ins i256mem:$src),
"vmovdqu\t{$src, $dst|$dst, $src}",[], IIC_SSE_MOVU_P_RM>,
- XS, VEX;
+ XS, VEX, VEX_L;
}
}
-let mayStore = 1 in {
+let mayStore = 1, neverHasSideEffects = 1 in {
def VMOVDQAmr : VPDI<0x7F, MRMDestMem, (outs),
(ins i128mem:$dst, VR128:$src),
"movdqa\t{$src, $dst|$dst, $src}", [], IIC_SSE_MOVA_P_MR>,
def VMOVDQAYmr : VPDI<0x7F, MRMDestMem, (outs),
(ins i256mem:$dst, VR256:$src),
"movdqa\t{$src, $dst|$dst, $src}", [], IIC_SSE_MOVA_P_MR>,
- VEX;
+ VEX, VEX_L;
let Predicates = [HasAVX] in {
def VMOVDQUmr : I<0x7F, MRMDestMem, (outs), (ins i128mem:$dst, VR128:$src),
"vmovdqu\t{$src, $dst|$dst, $src}",[], IIC_SSE_MOVU_P_MR>,
XS, VEX;
def VMOVDQUYmr : I<0x7F, MRMDestMem, (outs), (ins i256mem:$dst, VR256:$src),
"vmovdqu\t{$src, $dst|$dst, $src}",[], IIC_SSE_MOVU_P_MR>,
- XS, VEX;
+ XS, VEX, VEX_L;
}
}
def MOVDQUrr : I<0x6F, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
"movdqu\t{$src, $dst|$dst, $src}",
- [], IIC_SSE_MOVU_P_RR>, XS, Requires<[HasSSE2]>;
+ [], IIC_SSE_MOVU_P_RR>, XS, Requires<[UseSSE2]>;
// For Disassembler
-let isCodeGenOnly = 1 in {
+let isCodeGenOnly = 1, hasSideEffects = 0 in {
def MOVDQArr_REV : PDI<0x7F, MRMDestReg, (outs VR128:$dst), (ins VR128:$src),
"movdqa\t{$src, $dst|$dst, $src}", [],
IIC_SSE_MOVA_P_RR>;
def MOVDQUrr_REV : I<0x7F, MRMDestReg, (outs VR128:$dst), (ins VR128:$src),
"movdqu\t{$src, $dst|$dst, $src}",
- [], IIC_SSE_MOVU_P_RR>, XS, Requires<[HasSSE2]>;
+ [], IIC_SSE_MOVU_P_RR>, XS, Requires<[UseSSE2]>;
}
-let canFoldAsLoad = 1, mayLoad = 1 in {
+let canFoldAsLoad = 1, mayLoad = 1, isReMaterializable = 1,
+ neverHasSideEffects = 1 in {
def MOVDQArm : PDI<0x6F, MRMSrcMem, (outs VR128:$dst), (ins i128mem:$src),
"movdqa\t{$src, $dst|$dst, $src}",
[/*(set VR128:$dst, (alignedloadv2i64 addr:$src))*/],
"movdqu\t{$src, $dst|$dst, $src}",
[/*(set VR128:$dst, (loadv2i64 addr:$src))*/],
IIC_SSE_MOVU_P_RM>,
- XS, Requires<[HasSSE2]>;
+ XS, Requires<[UseSSE2]>;
}
let mayStore = 1 in {
"movdqu\t{$src, $dst|$dst, $src}",
[/*(store (v2i64 VR128:$src), addr:$dst)*/],
IIC_SSE_MOVU_P_MR>,
- XS, Requires<[HasSSE2]>;
+ XS, Requires<[UseSSE2]>;
}
-// Intrinsic forms of MOVDQU load and store
-def VMOVDQUmr_Int : I<0x7F, MRMDestMem, (outs), (ins i128mem:$dst, VR128:$src),
- "vmovdqu\t{$src, $dst|$dst, $src}",
- [(int_x86_sse2_storeu_dq addr:$dst, VR128:$src)],
- IIC_SSE_MOVU_P_MR>,
- XS, VEX, Requires<[HasAVX]>;
-
-def MOVDQUmr_Int : I<0x7F, MRMDestMem, (outs), (ins i128mem:$dst, VR128:$src),
- "movdqu\t{$src, $dst|$dst, $src}",
- [(int_x86_sse2_storeu_dq addr:$dst, VR128:$src)],
- IIC_SSE_MOVU_P_MR>,
- XS, Requires<[HasSSE2]>;
-
} // ExeDomain = SSEPackedInt
let Predicates = [HasAVX] in {
+ def : Pat<(int_x86_sse2_storeu_dq addr:$dst, VR128:$src),
+ (VMOVDQUmr addr:$dst, VR128:$src)>;
def : Pat<(int_x86_avx_storeu_dq_256 addr:$dst, VR256:$src),
(VMOVDQUYmr addr:$dst, VR256:$src)>;
}
+let Predicates = [UseSSE2] in
+def : Pat<(int_x86_sse2_storeu_dq addr:$dst, VR128:$src),
+ (MOVDQUmr addr:$dst, VR128:$src)>;
//===---------------------------------------------------------------------===//
// SSE2 - Packed Integer Arithmetic Instructions
itins.rm>;
}
+multiclass PDI_binop_all_int<bits<8> opc, string OpcodeStr, Intrinsic IntId128,
+ Intrinsic IntId256, OpndItins itins,
+ bit IsCommutable = 0> {
+let Predicates = [HasAVX] in
+ defm V#NAME : PDI_binop_rm_int<opc, !strconcat("v", OpcodeStr), IntId128,
+ VR128, memopv2i64, i128mem, itins,
+ IsCommutable, 0>, VEX_4V;
+
+let Constraints = "$src1 = $dst" in
+ defm NAME : PDI_binop_rm_int<opc, OpcodeStr, IntId128, VR128, memopv2i64,
+ i128mem, itins, IsCommutable, 1>;
+
+let Predicates = [HasAVX2] in
+ defm V#NAME#Y : PDI_binop_rm_int<opc, !strconcat("v", OpcodeStr), IntId256,
+ VR256, memopv4i64, i256mem, itins,
+ IsCommutable, 0>, VEX_4V, VEX_L;
+}
+
multiclass PDI_binop_rmi<bits<8> opc, bits<8> opc2, Format ImmForm,
string OpcodeStr, SDNode OpNode,
SDNode OpNode2, RegisterClass RC,
[(set RC:$dst, (DstVT (OpNode2 RC:$src1, (i32 imm:$src2))))], itins.ri>;
}
-/// PDI_binop_rm - Simple SSE2 binary operator with different src and dst types
+/// PDI_binop_rm2 - Simple SSE2 binary operator with different src and dst types
multiclass PDI_binop_rm2<bits<8> opc, string OpcodeStr, SDNode OpNode,
ValueType DstVT, ValueType SrcVT, RegisterClass RC,
PatFrag memop_frag, X86MemOperand x86memop,
}
} // ExeDomain = SSEPackedInt
-// 128-bit Integer Arithmetic
+defm PADDB : PDI_binop_all<0xFC, "paddb", add, v16i8, v32i8,
+ SSE_INTALU_ITINS_P, 1>;
+defm PADDW : PDI_binop_all<0xFD, "paddw", add, v8i16, v16i16,
+ SSE_INTALU_ITINS_P, 1>;
+defm PADDD : PDI_binop_all<0xFE, "paddd", add, v4i32, v8i32,
+ SSE_INTALU_ITINS_P, 1>;
+defm PADDQ : PDI_binop_all<0xD4, "paddq", add, v2i64, v4i64,
+ SSE_INTALUQ_ITINS_P, 1>;
+defm PMULLW : PDI_binop_all<0xD5, "pmullw", mul, v8i16, v16i16,
+ SSE_INTMUL_ITINS_P, 1>;
+defm PSUBB : PDI_binop_all<0xF8, "psubb", sub, v16i8, v32i8,
+ SSE_INTALU_ITINS_P, 0>;
+defm PSUBW : PDI_binop_all<0xF9, "psubw", sub, v8i16, v16i16,
+ SSE_INTALU_ITINS_P, 0>;
+defm PSUBD : PDI_binop_all<0xFA, "psubd", sub, v4i32, v8i32,
+ SSE_INTALU_ITINS_P, 0>;
+defm PSUBQ : PDI_binop_all<0xFB, "psubq", sub, v2i64, v4i64,
+ SSE_INTALUQ_ITINS_P, 0>;
+defm PSUBUSB : PDI_binop_all<0xD8, "psubusb", X86subus, v16i8, v32i8,
+ SSE_INTALU_ITINS_P, 0>;
+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>;
-let Predicates = [HasAVX] in {
-defm VPADDB : PDI_binop_rm<0xFC, "vpaddb", add, v16i8, VR128, memopv2i64,
- i128mem, SSE_INTALU_ITINS_P, 1, 0 /*3addr*/>,
- VEX_4V;
-defm VPADDW : PDI_binop_rm<0xFD, "vpaddw", add, v8i16, VR128, memopv2i64,
- i128mem, SSE_INTALU_ITINS_P, 1, 0>, VEX_4V;
-defm VPADDD : PDI_binop_rm<0xFE, "vpaddd", add, v4i32, VR128, memopv2i64,
- i128mem, SSE_INTALU_ITINS_P, 1, 0>, VEX_4V;
-defm VPADDQ : PDI_binop_rm<0xD4, "vpaddq", add, v2i64, VR128, memopv2i64,
- i128mem, SSE_INTALUQ_ITINS_P, 1, 0>, VEX_4V;
-defm VPMULLW : PDI_binop_rm<0xD5, "vpmullw", mul, v8i16, VR128, memopv2i64,
- i128mem, SSE_INTMUL_ITINS_P, 1, 0>, VEX_4V;
-defm VPSUBB : PDI_binop_rm<0xF8, "vpsubb", sub, v16i8, VR128, memopv2i64,
- i128mem, SSE_INTALU_ITINS_P, 0, 0>, VEX_4V;
-defm VPSUBW : PDI_binop_rm<0xF9, "vpsubw", sub, v8i16, VR128, memopv2i64,
- i128mem, SSE_INTALU_ITINS_P, 0, 0>, VEX_4V;
-defm VPSUBD : PDI_binop_rm<0xFA, "vpsubd", sub, v4i32, VR128, memopv2i64,
- i128mem, SSE_INTALU_ITINS_P, 0, 0>, VEX_4V;
-defm VPSUBQ : PDI_binop_rm<0xFB, "vpsubq", sub, v2i64, VR128, memopv2i64,
- i128mem, SSE_INTALUQ_ITINS_P, 0, 0>, VEX_4V;
+// Intrinsic forms
+defm PSUBSB : PDI_binop_all_int<0xE8, "psubsb", int_x86_sse2_psubs_b,
+ int_x86_avx2_psubs_b, SSE_INTALU_ITINS_P, 0>;
+defm PSUBSW : PDI_binop_all_int<0xE9, "psubsw" , int_x86_sse2_psubs_w,
+ int_x86_avx2_psubs_w, SSE_INTALU_ITINS_P, 0>;
+defm PADDSB : PDI_binop_all_int<0xEC, "paddsb" , int_x86_sse2_padds_b,
+ int_x86_avx2_padds_b, SSE_INTALU_ITINS_P, 1>;
+defm PADDSW : PDI_binop_all_int<0xED, "paddsw" , int_x86_sse2_padds_w,
+ int_x86_avx2_padds_w, SSE_INTALU_ITINS_P, 1>;
+defm PADDUSB : PDI_binop_all_int<0xDC, "paddusb", int_x86_sse2_paddus_b,
+ int_x86_avx2_paddus_b, SSE_INTALU_ITINS_P, 1>;
+defm PADDUSW : PDI_binop_all_int<0xDD, "paddusw", int_x86_sse2_paddus_w,
+ int_x86_avx2_paddus_w, SSE_INTALU_ITINS_P, 1>;
+defm PMULHUW : PDI_binop_all_int<0xE4, "pmulhuw", int_x86_sse2_pmulhu_w,
+ int_x86_avx2_pmulhu_w, SSE_INTMUL_ITINS_P, 1>;
+defm PMULHW : PDI_binop_all_int<0xE5, "pmulhw" , int_x86_sse2_pmulh_w,
+ int_x86_avx2_pmulh_w, SSE_INTMUL_ITINS_P, 1>;
+defm PMADDWD : PDI_binop_all_int<0xF5, "pmaddwd", int_x86_sse2_pmadd_wd,
+ int_x86_avx2_pmadd_wd, SSE_PMADD, 1>;
+defm PAVGB : PDI_binop_all_int<0xE0, "pavgb", int_x86_sse2_pavg_b,
+ int_x86_avx2_pavg_b, SSE_INTALU_ITINS_P, 1>;
+defm PAVGW : PDI_binop_all_int<0xE3, "pavgw", int_x86_sse2_pavg_w,
+ int_x86_avx2_pavg_w, SSE_INTALU_ITINS_P, 1>;
+defm PSADBW : PDI_binop_all_int<0xF6, "psadbw", int_x86_sse2_psad_bw,
+ int_x86_avx2_psad_bw, SSE_INTALU_ITINS_P, 1>;
+
+let Predicates = [HasAVX] in
defm VPMULUDQ : PDI_binop_rm2<0xF4, "vpmuludq", X86pmuludq, v2i64, v4i32, VR128,
memopv2i64, i128mem, SSE_INTMUL_ITINS_P, 1, 0>,
VEX_4V;
-
-// Intrinsic forms
-defm VPSUBSB : PDI_binop_rm_int<0xE8, "vpsubsb" , int_x86_sse2_psubs_b,
- VR128, memopv2i64, i128mem,
- SSE_INTALU_ITINS_P, 0, 0>, VEX_4V;
-defm VPSUBSW : PDI_binop_rm_int<0xE9, "vpsubsw" , int_x86_sse2_psubs_w,
- VR128, memopv2i64, i128mem,
- SSE_INTALU_ITINS_P, 0, 0>, VEX_4V;
-defm VPSUBUSB : PDI_binop_rm_int<0xD8, "vpsubusb", int_x86_sse2_psubus_b,
- VR128, memopv2i64, i128mem,
- SSE_INTALU_ITINS_P, 0, 0>, VEX_4V;
-defm VPSUBUSW : PDI_binop_rm_int<0xD9, "vpsubusw", int_x86_sse2_psubus_w,
- VR128, memopv2i64, i128mem,
- SSE_INTALU_ITINS_P, 0, 0>, VEX_4V;
-defm VPADDSB : PDI_binop_rm_int<0xEC, "vpaddsb" , int_x86_sse2_padds_b,
- VR128, memopv2i64, i128mem,
- SSE_INTALU_ITINS_P, 1, 0>, VEX_4V;
-defm VPADDSW : PDI_binop_rm_int<0xED, "vpaddsw" , int_x86_sse2_padds_w,
- VR128, memopv2i64, i128mem,
- SSE_INTALU_ITINS_P, 1, 0>, VEX_4V;
-defm VPADDUSB : PDI_binop_rm_int<0xDC, "vpaddusb", int_x86_sse2_paddus_b,
- VR128, memopv2i64, i128mem,
- SSE_INTALU_ITINS_P, 1, 0>, VEX_4V;
-defm VPADDUSW : PDI_binop_rm_int<0xDD, "vpaddusw", int_x86_sse2_paddus_w,
- VR128, memopv2i64, i128mem,
- SSE_INTALU_ITINS_P, 1, 0>, VEX_4V;
-defm VPMULHUW : PDI_binop_rm_int<0xE4, "vpmulhuw", int_x86_sse2_pmulhu_w,
- VR128, memopv2i64, i128mem,
- SSE_INTMUL_ITINS_P, 1, 0>, VEX_4V;
-defm VPMULHW : PDI_binop_rm_int<0xE5, "vpmulhw" , int_x86_sse2_pmulh_w,
- VR128, memopv2i64, i128mem,
- SSE_INTMUL_ITINS_P, 1, 0>, VEX_4V;
-defm VPMADDWD : PDI_binop_rm_int<0xF5, "vpmaddwd", int_x86_sse2_pmadd_wd,
- VR128, memopv2i64, i128mem,
- SSE_PMADD, 1, 0>, VEX_4V;
-defm VPAVGB : PDI_binop_rm_int<0xE0, "vpavgb", int_x86_sse2_pavg_b,
- VR128, memopv2i64, i128mem,
- SSE_INTALU_ITINS_P, 1, 0>, VEX_4V;
-defm VPAVGW : PDI_binop_rm_int<0xE3, "vpavgw", int_x86_sse2_pavg_w,
- VR128, memopv2i64, i128mem,
- SSE_INTALU_ITINS_P, 1, 0>, VEX_4V;
-defm VPMINUB : PDI_binop_rm_int<0xDA, "vpminub", int_x86_sse2_pminu_b,
- VR128, memopv2i64, i128mem,
- SSE_INTALU_ITINS_P, 1, 0>, VEX_4V;
-defm VPMINSW : PDI_binop_rm_int<0xEA, "vpminsw", int_x86_sse2_pmins_w,
- VR128, memopv2i64, i128mem,
- SSE_INTALU_ITINS_P, 1, 0>, VEX_4V;
-defm VPMAXUB : PDI_binop_rm_int<0xDE, "vpmaxub", int_x86_sse2_pmaxu_b,
- VR128, memopv2i64, i128mem,
- SSE_INTALU_ITINS_P, 1, 0>, VEX_4V;
-defm VPMAXSW : PDI_binop_rm_int<0xEE, "vpmaxsw", int_x86_sse2_pmaxs_w,
- VR128, memopv2i64, i128mem,
- SSE_INTALU_ITINS_P, 1, 0>, VEX_4V;
-defm VPSADBW : PDI_binop_rm_int<0xF6, "vpsadbw", int_x86_sse2_psad_bw,
- VR128, memopv2i64, i128mem,
- SSE_INTALU_ITINS_P, 1, 0>, VEX_4V;
-}
-
-let Predicates = [HasAVX2] in {
-defm VPADDBY : PDI_binop_rm<0xFC, "vpaddb", add, v32i8, VR256, memopv4i64,
- i256mem, SSE_INTALU_ITINS_P, 1, 0>, VEX_4V;
-defm VPADDWY : PDI_binop_rm<0xFD, "vpaddw", add, v16i16, VR256, memopv4i64,
- i256mem, SSE_INTALU_ITINS_P, 1, 0>, VEX_4V;
-defm VPADDDY : PDI_binop_rm<0xFE, "vpaddd", add, v8i32, VR256, memopv4i64,
- i256mem, SSE_INTALU_ITINS_P, 1, 0>, VEX_4V;
-defm VPADDQY : PDI_binop_rm<0xD4, "vpaddq", add, v4i64, VR256, memopv4i64,
- i256mem, SSE_INTALUQ_ITINS_P, 1, 0>, VEX_4V;
-defm VPMULLWY : PDI_binop_rm<0xD5, "vpmullw", mul, v16i16, VR256, memopv4i64,
- i256mem, SSE_INTMUL_ITINS_P, 1, 0>, VEX_4V;
-defm VPSUBBY : PDI_binop_rm<0xF8, "vpsubb", sub, v32i8, VR256, memopv4i64,
- i256mem, SSE_INTALU_ITINS_P, 0, 0>, VEX_4V;
-defm VPSUBWY : PDI_binop_rm<0xF9, "vpsubw", sub, v16i16,VR256, memopv4i64,
- i256mem, SSE_INTALU_ITINS_P, 0, 0>, VEX_4V;
-defm VPSUBDY : PDI_binop_rm<0xFA, "vpsubd", sub, v8i32, VR256, memopv4i64,
- i256mem, SSE_INTALU_ITINS_P, 0, 0>, VEX_4V;
-defm VPSUBQY : PDI_binop_rm<0xFB, "vpsubq", sub, v4i64, VR256, memopv4i64,
- i256mem, SSE_INTALUQ_ITINS_P, 0, 0>, VEX_4V;
+let Predicates = [HasAVX2] in
defm VPMULUDQY : PDI_binop_rm2<0xF4, "vpmuludq", X86pmuludq, v4i64, v8i32,
VR256, memopv4i64, i256mem,
- SSE_INTMUL_ITINS_P, 1, 0>, VEX_4V;
-
-// Intrinsic forms
-defm VPSUBSBY : PDI_binop_rm_int<0xE8, "vpsubsb" , int_x86_avx2_psubs_b,
- VR256, memopv4i64, i256mem,
- SSE_INTALU_ITINS_P, 0, 0>, VEX_4V;
-defm VPSUBSWY : PDI_binop_rm_int<0xE9, "vpsubsw" , int_x86_avx2_psubs_w,
- VR256, memopv4i64, i256mem,
- SSE_INTALU_ITINS_P, 0, 0>, VEX_4V;
-defm VPSUBUSBY : PDI_binop_rm_int<0xD8, "vpsubusb", int_x86_avx2_psubus_b,
- VR256, memopv4i64, i256mem,
- SSE_INTALU_ITINS_P, 0, 0>, VEX_4V;
-defm VPSUBUSWY : PDI_binop_rm_int<0xD9, "vpsubusw", int_x86_avx2_psubus_w,
- VR256, memopv4i64, i256mem,
- SSE_INTALU_ITINS_P, 0, 0>, VEX_4V;
-defm VPADDSBY : PDI_binop_rm_int<0xEC, "vpaddsb" , int_x86_avx2_padds_b,
- VR256, memopv4i64, i256mem,
- SSE_INTALU_ITINS_P, 1, 0>, VEX_4V;
-defm VPADDSWY : PDI_binop_rm_int<0xED, "vpaddsw" , int_x86_avx2_padds_w,
- VR256, memopv4i64, i256mem,
- SSE_INTALU_ITINS_P, 1, 0>, VEX_4V;
-defm VPADDUSBY : PDI_binop_rm_int<0xDC, "vpaddusb", int_x86_avx2_paddus_b,
- VR256, memopv4i64, i256mem,
- SSE_INTALU_ITINS_P, 1, 0>, VEX_4V;
-defm VPADDUSWY : PDI_binop_rm_int<0xDD, "vpaddusw", int_x86_avx2_paddus_w,
- VR256, memopv4i64, i256mem,
- SSE_INTALU_ITINS_P, 1, 0>, VEX_4V;
-defm VPMULHUWY : PDI_binop_rm_int<0xE4, "vpmulhuw", int_x86_avx2_pmulhu_w,
- VR256, memopv4i64, i256mem,
- SSE_INTMUL_ITINS_P, 1, 0>, VEX_4V;
-defm VPMULHWY : PDI_binop_rm_int<0xE5, "vpmulhw" , int_x86_avx2_pmulh_w,
- VR256, memopv4i64, i256mem,
- SSE_INTMUL_ITINS_P, 1, 0>, VEX_4V;
-defm VPMADDWDY : PDI_binop_rm_int<0xF5, "vpmaddwd", int_x86_avx2_pmadd_wd,
- VR256, memopv4i64, i256mem,
- SSE_PMADD, 1, 0>, VEX_4V;
-defm VPAVGBY : PDI_binop_rm_int<0xE0, "vpavgb", int_x86_avx2_pavg_b,
- VR256, memopv4i64, i256mem,
- SSE_INTALU_ITINS_P, 1, 0>, VEX_4V;
-defm VPAVGWY : PDI_binop_rm_int<0xE3, "vpavgw", int_x86_avx2_pavg_w,
- VR256, memopv4i64, i256mem,
- SSE_INTALU_ITINS_P, 1, 0>, VEX_4V;
-defm VPMINUBY : PDI_binop_rm_int<0xDA, "vpminub", int_x86_avx2_pminu_b,
- VR256, memopv4i64, i256mem,
- SSE_INTALU_ITINS_P, 1, 0>, VEX_4V;
-defm VPMINSWY : PDI_binop_rm_int<0xEA, "vpminsw", int_x86_avx2_pmins_w,
- VR256, memopv4i64, i256mem,
- SSE_INTALU_ITINS_P, 1, 0>, VEX_4V;
-defm VPMAXUBY : PDI_binop_rm_int<0xDE, "vpmaxub", int_x86_avx2_pmaxu_b,
- VR256, memopv4i64, i256mem,
- SSE_INTALU_ITINS_P, 1, 0>, VEX_4V;
-defm VPMAXSWY : PDI_binop_rm_int<0xEE, "vpmaxsw", int_x86_avx2_pmaxs_w,
- VR256, memopv4i64, i256mem,
- SSE_INTALU_ITINS_P, 1, 0>, VEX_4V;
-defm VPSADBWY : PDI_binop_rm_int<0xF6, "vpsadbw", int_x86_avx2_psad_bw,
- VR256, memopv4i64, i256mem,
- SSE_INTALU_ITINS_P, 1, 0>, VEX_4V;
-}
-
-let Constraints = "$src1 = $dst" in {
-defm PADDB : PDI_binop_rm<0xFC, "paddb", add, v16i8, VR128, memopv2i64,
- i128mem, SSE_INTALU_ITINS_P, 1>;
-defm PADDW : PDI_binop_rm<0xFD, "paddw", add, v8i16, VR128, memopv2i64,
- i128mem, SSE_INTALU_ITINS_P, 1>;
-defm PADDD : PDI_binop_rm<0xFE, "paddd", add, v4i32, VR128, memopv2i64,
- i128mem, SSE_INTALU_ITINS_P, 1>;
-defm PADDQ : PDI_binop_rm<0xD4, "paddq", add, v2i64, VR128, memopv2i64,
- i128mem, SSE_INTALUQ_ITINS_P, 1>;
-defm PMULLW : PDI_binop_rm<0xD5, "pmullw", mul, v8i16, VR128, memopv2i64,
- i128mem, SSE_INTMUL_ITINS_P, 1>;
-defm PSUBB : PDI_binop_rm<0xF8, "psubb", sub, v16i8, VR128, memopv2i64,
- i128mem, SSE_INTALU_ITINS_P>;
-defm PSUBW : PDI_binop_rm<0xF9, "psubw", sub, v8i16, VR128, memopv2i64,
- i128mem, SSE_INTALU_ITINS_P>;
-defm PSUBD : PDI_binop_rm<0xFA, "psubd", sub, v4i32, VR128, memopv2i64,
- i128mem, SSE_INTALU_ITINS_P>;
-defm PSUBQ : PDI_binop_rm<0xFB, "psubq", sub, v2i64, VR128, memopv2i64,
- i128mem, SSE_INTALUQ_ITINS_P>;
+ SSE_INTMUL_ITINS_P, 1, 0>, VEX_4V, VEX_L;
+let Constraints = "$src1 = $dst" in
defm PMULUDQ : PDI_binop_rm2<0xF4, "pmuludq", X86pmuludq, v2i64, v4i32, VR128,
memopv2i64, i128mem, SSE_INTMUL_ITINS_P, 1>;
-// Intrinsic forms
-defm PSUBSB : PDI_binop_rm_int<0xE8, "psubsb" , int_x86_sse2_psubs_b,
- VR128, memopv2i64, i128mem,
- SSE_INTALU_ITINS_P>;
-defm PSUBSW : PDI_binop_rm_int<0xE9, "psubsw" , int_x86_sse2_psubs_w,
- VR128, memopv2i64, i128mem,
- SSE_INTALU_ITINS_P>;
-defm PSUBUSB : PDI_binop_rm_int<0xD8, "psubusb", int_x86_sse2_psubus_b,
- VR128, memopv2i64, i128mem,
- SSE_INTALU_ITINS_P>;
-defm PSUBUSW : PDI_binop_rm_int<0xD9, "psubusw", int_x86_sse2_psubus_w,
- VR128, memopv2i64, i128mem,
- SSE_INTALU_ITINS_P>;
-defm PADDSB : PDI_binop_rm_int<0xEC, "paddsb" , int_x86_sse2_padds_b,
- VR128, memopv2i64, i128mem,
- SSE_INTALU_ITINS_P, 1>;
-defm PADDSW : PDI_binop_rm_int<0xED, "paddsw" , int_x86_sse2_padds_w,
- VR128, memopv2i64, i128mem,
- SSE_INTALU_ITINS_P, 1>;
-defm PADDUSB : PDI_binop_rm_int<0xDC, "paddusb", int_x86_sse2_paddus_b,
- VR128, memopv2i64, i128mem,
- SSE_INTALU_ITINS_P, 1>;
-defm PADDUSW : PDI_binop_rm_int<0xDD, "paddusw", int_x86_sse2_paddus_w,
- VR128, memopv2i64, i128mem,
- SSE_INTALU_ITINS_P, 1>;
-defm PMULHUW : PDI_binop_rm_int<0xE4, "pmulhuw", int_x86_sse2_pmulhu_w,
- VR128, memopv2i64, i128mem,
- SSE_INTMUL_ITINS_P, 1>;
-defm PMULHW : PDI_binop_rm_int<0xE5, "pmulhw" , int_x86_sse2_pmulh_w,
- VR128, memopv2i64, i128mem,
- SSE_INTMUL_ITINS_P, 1>;
-defm PMADDWD : PDI_binop_rm_int<0xF5, "pmaddwd", int_x86_sse2_pmadd_wd,
- VR128, memopv2i64, i128mem,
- SSE_PMADD, 1>;
-defm PAVGB : PDI_binop_rm_int<0xE0, "pavgb", int_x86_sse2_pavg_b,
- VR128, memopv2i64, i128mem,
- SSE_INTALU_ITINS_P, 1>;
-defm PAVGW : PDI_binop_rm_int<0xE3, "pavgw", int_x86_sse2_pavg_w,
- VR128, memopv2i64, i128mem,
- SSE_INTALU_ITINS_P, 1>;
-defm PMINUB : PDI_binop_rm_int<0xDA, "pminub", int_x86_sse2_pminu_b,
- VR128, memopv2i64, i128mem,
- SSE_INTALU_ITINS_P, 1>;
-defm PMINSW : PDI_binop_rm_int<0xEA, "pminsw", int_x86_sse2_pmins_w,
- VR128, memopv2i64, i128mem,
- SSE_INTALU_ITINS_P, 1>;
-defm PMAXUB : PDI_binop_rm_int<0xDE, "pmaxub", int_x86_sse2_pmaxu_b,
- VR128, memopv2i64, i128mem,
- SSE_INTALU_ITINS_P, 1>;
-defm PMAXSW : PDI_binop_rm_int<0xEE, "pmaxsw", int_x86_sse2_pmaxs_w,
- VR128, memopv2i64, i128mem,
- SSE_INTALU_ITINS_P, 1>;
-defm PSADBW : PDI_binop_rm_int<0xF6, "psadbw", int_x86_sse2_psad_bw,
- VR128, memopv2i64, i128mem,
- SSE_INTALU_ITINS_P, 1>;
-
-} // Constraints = "$src1 = $dst"
-
//===---------------------------------------------------------------------===//
// SSE2 - Packed Integer Logical Instructions
//===---------------------------------------------------------------------===//
let Predicates = [HasAVX2] in {
defm VPSLLWY : PDI_binop_rmi<0xF1, 0x71, MRM6r, "vpsllw", X86vshl, X86vshli,
VR256, v16i16, v8i16, bc_v8i16,
- SSE_INTSHIFT_ITINS_P, 0>, VEX_4V;
+ SSE_INTSHIFT_ITINS_P, 0>, VEX_4V, VEX_L;
defm VPSLLDY : PDI_binop_rmi<0xF2, 0x72, MRM6r, "vpslld", X86vshl, X86vshli,
VR256, v8i32, v4i32, bc_v4i32,
- SSE_INTSHIFT_ITINS_P, 0>, VEX_4V;
+ SSE_INTSHIFT_ITINS_P, 0>, VEX_4V, VEX_L;
defm VPSLLQY : PDI_binop_rmi<0xF3, 0x73, MRM6r, "vpsllq", X86vshl, X86vshli,
VR256, v4i64, v2i64, bc_v2i64,
- SSE_INTSHIFT_ITINS_P, 0>, VEX_4V;
+ SSE_INTSHIFT_ITINS_P, 0>, VEX_4V, VEX_L;
defm VPSRLWY : PDI_binop_rmi<0xD1, 0x71, MRM2r, "vpsrlw", X86vsrl, X86vsrli,
VR256, v16i16, v8i16, bc_v8i16,
- SSE_INTSHIFT_ITINS_P, 0>, VEX_4V;
+ SSE_INTSHIFT_ITINS_P, 0>, VEX_4V, VEX_L;
defm VPSRLDY : PDI_binop_rmi<0xD2, 0x72, MRM2r, "vpsrld", X86vsrl, X86vsrli,
VR256, v8i32, v4i32, bc_v4i32,
- SSE_INTSHIFT_ITINS_P, 0>, VEX_4V;
+ SSE_INTSHIFT_ITINS_P, 0>, VEX_4V, VEX_L;
defm VPSRLQY : PDI_binop_rmi<0xD3, 0x73, MRM2r, "vpsrlq", X86vsrl, X86vsrli,
VR256, v4i64, v2i64, bc_v2i64,
- SSE_INTSHIFT_ITINS_P, 0>, VEX_4V;
+ SSE_INTSHIFT_ITINS_P, 0>, VEX_4V, VEX_L;
defm VPSRAWY : PDI_binop_rmi<0xE1, 0x71, MRM4r, "vpsraw", X86vsra, X86vsrai,
VR256, v16i16, v8i16, bc_v8i16,
- SSE_INTSHIFT_ITINS_P, 0>, VEX_4V;
+ SSE_INTSHIFT_ITINS_P, 0>, VEX_4V, VEX_L;
defm VPSRADY : PDI_binop_rmi<0xE2, 0x72, MRM4r, "vpsrad", X86vsra, X86vsrai,
VR256, v8i32, v4i32, bc_v4i32,
- SSE_INTSHIFT_ITINS_P, 0>, VEX_4V;
+ SSE_INTSHIFT_ITINS_P, 0>, VEX_4V, VEX_L;
let ExeDomain = SSEPackedInt in {
// 256-bit logical shifts.
"vpslldq\t{$src2, $src1, $dst|$dst, $src1, $src2}",
[(set VR256:$dst,
(int_x86_avx2_psll_dq_bs VR256:$src1, imm:$src2))]>,
- VEX_4V;
+ VEX_4V, VEX_L;
def VPSRLDQYri : PDIi8<0x73, MRM3r,
(outs VR256:$dst), (ins VR256:$src1, i32i8imm:$src2),
"vpsrldq\t{$src2, $src1, $dst|$dst, $src1, $src2}",
[(set VR256:$dst,
(int_x86_avx2_psrl_dq_bs VR256:$src1, imm:$src2))]>,
- VEX_4V;
+ VEX_4V, VEX_L;
// PSRADQYri doesn't exist in SSE[1-3].
}
} // Predicates = [HasAVX2]
(VPSRLDQYri VR256:$src1, (BYTE_imm imm:$src2))>;
}
-let Predicates = [HasSSE2] in {
+let Predicates = [UseSSE2] in {
def : Pat<(int_x86_sse2_psll_dq VR128:$src1, imm:$src2),
(PSLLDQri VR128:$src1, (BYTE_imm imm:$src2))>;
def : Pat<(int_x86_sse2_psrl_dq VR128:$src1, imm:$src2),
// SSE2 - Packed Integer Comparison Instructions
//===---------------------------------------------------------------------===//
-let Predicates = [HasAVX] in {
- defm VPCMPEQB : PDI_binop_rm<0x74, "vpcmpeqb", X86pcmpeq, v16i8,
- VR128, memopv2i64, i128mem,
- SSE_INTALU_ITINS_P, 1, 0>, VEX_4V;
- defm VPCMPEQW : PDI_binop_rm<0x75, "vpcmpeqw", X86pcmpeq, v8i16,
- VR128, memopv2i64, i128mem,
- SSE_INTALU_ITINS_P, 1, 0>, VEX_4V;
- defm VPCMPEQD : PDI_binop_rm<0x76, "vpcmpeqd", X86pcmpeq, v4i32,
- VR128, memopv2i64, i128mem,
- SSE_INTALU_ITINS_P, 1, 0>, VEX_4V;
- defm VPCMPGTB : PDI_binop_rm<0x64, "vpcmpgtb", X86pcmpgt, v16i8,
- VR128, memopv2i64, i128mem,
- SSE_INTALU_ITINS_P, 0, 0>, VEX_4V;
- defm VPCMPGTW : PDI_binop_rm<0x65, "vpcmpgtw", X86pcmpgt, v8i16,
- VR128, memopv2i64, i128mem,
- SSE_INTALU_ITINS_P, 0, 0>, VEX_4V;
- defm VPCMPGTD : PDI_binop_rm<0x66, "vpcmpgtd", X86pcmpgt, v4i32,
- VR128, memopv2i64, i128mem,
- SSE_INTALU_ITINS_P, 0, 0>, VEX_4V;
-}
-
-let Predicates = [HasAVX2] in {
- defm VPCMPEQBY : PDI_binop_rm<0x74, "vpcmpeqb", X86pcmpeq, v32i8,
- VR256, memopv4i64, i256mem,
- SSE_INTALU_ITINS_P, 1, 0>, VEX_4V;
- defm VPCMPEQWY : PDI_binop_rm<0x75, "vpcmpeqw", X86pcmpeq, v16i16,
- VR256, memopv4i64, i256mem,
- SSE_INTALU_ITINS_P, 1, 0>, VEX_4V;
- defm VPCMPEQDY : PDI_binop_rm<0x76, "vpcmpeqd", X86pcmpeq, v8i32,
- VR256, memopv4i64, i256mem,
- SSE_INTALU_ITINS_P, 1, 0>, VEX_4V;
- defm VPCMPGTBY : PDI_binop_rm<0x64, "vpcmpgtb", X86pcmpgt, v32i8,
- VR256, memopv4i64, i256mem,
- SSE_INTALU_ITINS_P, 0, 0>, VEX_4V;
- defm VPCMPGTWY : PDI_binop_rm<0x65, "vpcmpgtw", X86pcmpgt, v16i16,
- VR256, memopv4i64, i256mem,
- SSE_INTALU_ITINS_P, 0, 0>, VEX_4V;
- defm VPCMPGTDY : PDI_binop_rm<0x66, "vpcmpgtd", X86pcmpgt, v8i32,
- VR256, memopv4i64, i256mem,
- SSE_INTALU_ITINS_P, 0, 0>, VEX_4V;
-}
-
-let Constraints = "$src1 = $dst" in {
- defm PCMPEQB : PDI_binop_rm<0x74, "pcmpeqb", X86pcmpeq, v16i8,
- VR128, memopv2i64, i128mem,
- SSE_INTALU_ITINS_P, 1>;
- defm PCMPEQW : PDI_binop_rm<0x75, "pcmpeqw", X86pcmpeq, v8i16,
- VR128, memopv2i64, i128mem,
- SSE_INTALU_ITINS_P, 1>;
- defm PCMPEQD : PDI_binop_rm<0x76, "pcmpeqd", X86pcmpeq, v4i32,
- VR128, memopv2i64, i128mem,
- SSE_INTALU_ITINS_P, 1>;
- defm PCMPGTB : PDI_binop_rm<0x64, "pcmpgtb", X86pcmpgt, v16i8,
- VR128, memopv2i64, i128mem,
- SSE_INTALU_ITINS_P>;
- defm PCMPGTW : PDI_binop_rm<0x65, "pcmpgtw", X86pcmpgt, v8i16,
- VR128, memopv2i64, i128mem,
- SSE_INTALU_ITINS_P>;
- defm PCMPGTD : PDI_binop_rm<0x66, "pcmpgtd", X86pcmpgt, v4i32,
- VR128, memopv2i64, i128mem,
- SSE_INTALU_ITINS_P>;
-} // Constraints = "$src1 = $dst"
+defm PCMPEQB : PDI_binop_all<0x74, "pcmpeqb", X86pcmpeq, v16i8, v32i8,
+ SSE_INTALU_ITINS_P, 1>;
+defm PCMPEQW : PDI_binop_all<0x75, "pcmpeqw", X86pcmpeq, v8i16, v16i16,
+ SSE_INTALU_ITINS_P, 1>;
+defm PCMPEQD : PDI_binop_all<0x76, "pcmpeqd", X86pcmpeq, v4i32, v8i32,
+ SSE_INTALU_ITINS_P, 1>;
+defm PCMPGTB : PDI_binop_all<0x64, "pcmpgtb", X86pcmpgt, v16i8, v32i8,
+ SSE_INTALU_ITINS_P, 0>;
+defm PCMPGTW : PDI_binop_all<0x65, "pcmpgtw", X86pcmpgt, v8i16, v16i16,
+ SSE_INTALU_ITINS_P, 0>;
+defm PCMPGTD : PDI_binop_all<0x66, "pcmpgtd", X86pcmpgt, v4i32, v8i32,
+ SSE_INTALU_ITINS_P, 0>;
//===---------------------------------------------------------------------===//
// SSE2 - Packed Integer Pack Instructions
//===---------------------------------------------------------------------===//
-let Predicates = [HasAVX] in {
-defm VPACKSSWB : PDI_binop_rm_int<0x63, "vpacksswb", int_x86_sse2_packsswb_128,
- VR128, memopv2i64, i128mem,
- SSE_INTALU_ITINS_P, 0, 0>, VEX_4V;
-defm VPACKSSDW : PDI_binop_rm_int<0x6B, "vpackssdw", int_x86_sse2_packssdw_128,
- VR128, memopv2i64, i128mem,
- SSE_INTALU_ITINS_P, 0, 0>, VEX_4V;
-defm VPACKUSWB : PDI_binop_rm_int<0x67, "vpackuswb", int_x86_sse2_packuswb_128,
- VR128, memopv2i64, i128mem,
- SSE_INTALU_ITINS_P, 0, 0>, VEX_4V;
-}
-
-let Predicates = [HasAVX2] in {
-defm VPACKSSWBY : PDI_binop_rm_int<0x63, "vpacksswb", int_x86_avx2_packsswb,
- VR256, memopv4i64, i256mem,
- SSE_INTALU_ITINS_P, 0, 0>, VEX_4V;
-defm VPACKSSDWY : PDI_binop_rm_int<0x6B, "vpackssdw", int_x86_avx2_packssdw,
- VR256, memopv4i64, i256mem,
- SSE_INTALU_ITINS_P, 0, 0>, VEX_4V;
-defm VPACKUSWBY : PDI_binop_rm_int<0x67, "vpackuswb", int_x86_avx2_packuswb,
- VR256, memopv4i64, i256mem,
- SSE_INTALU_ITINS_P, 0, 0>, VEX_4V;
-}
-
-let Constraints = "$src1 = $dst" in {
-defm PACKSSWB : PDI_binop_rm_int<0x63, "packsswb", int_x86_sse2_packsswb_128,
- VR128, memopv2i64, i128mem,
- SSE_INTALU_ITINS_P>;
-defm PACKSSDW : PDI_binop_rm_int<0x6B, "packssdw", int_x86_sse2_packssdw_128,
- VR128, memopv2i64, i128mem,
- SSE_INTALU_ITINS_P>;
-defm PACKUSWB : PDI_binop_rm_int<0x67, "packuswb", int_x86_sse2_packuswb_128,
- VR128, memopv2i64, i128mem,
- SSE_INTALU_ITINS_P>;
-} // Constraints = "$src1 = $dst"
+defm PACKSSWB : PDI_binop_all_int<0x63, "packsswb", int_x86_sse2_packsswb_128,
+ int_x86_avx2_packsswb, SSE_INTALU_ITINS_P, 0>;
+defm PACKSSDW : PDI_binop_all_int<0x6B, "packssdw", int_x86_sse2_packssdw_128,
+ int_x86_avx2_packssdw, SSE_INTALU_ITINS_P, 0>;
+defm PACKUSWB : PDI_binop_all_int<0x67, "packuswb", int_x86_sse2_packuswb_128,
+ int_x86_avx2_packuswb, SSE_INTALU_ITINS_P, 0>;
//===---------------------------------------------------------------------===//
// SSE2 - Packed Integer Shuffle Instructions
//===---------------------------------------------------------------------===//
let ExeDomain = SSEPackedInt in {
-multiclass sse2_pshuffle<string OpcodeStr, ValueType vt, SDNode OpNode> {
-def ri : Ii8<0x70, MRMSrcReg,
- (outs VR128:$dst), (ins VR128:$src1, i8imm:$src2),
- !strconcat(OpcodeStr,
- "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
- [(set VR128:$dst, (vt (OpNode VR128:$src1, (i8 imm:$src2))))],
- IIC_SSE_PSHUF>;
-def mi : Ii8<0x70, MRMSrcMem,
- (outs VR128:$dst), (ins i128mem:$src1, i8imm:$src2),
- !strconcat(OpcodeStr,
- "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
- [(set VR128:$dst,
- (vt (OpNode (bitconvert (memopv2i64 addr:$src1)),
- (i8 imm:$src2))))],
- IIC_SSE_PSHUF>;
-}
-
-multiclass sse2_pshuffle_y<string OpcodeStr, ValueType vt, SDNode OpNode> {
-def Yri : Ii8<0x70, MRMSrcReg,
- (outs VR256:$dst), (ins VR256:$src1, i8imm:$src2),
- !strconcat(OpcodeStr,
- "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
- [(set VR256:$dst, (vt (OpNode VR256:$src1, (i8 imm:$src2))))]>;
-def Ymi : Ii8<0x70, MRMSrcMem,
- (outs VR256:$dst), (ins i256mem:$src1, i8imm:$src2),
- !strconcat(OpcodeStr,
- "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
- [(set VR256:$dst,
- (vt (OpNode (bitconvert (memopv4i64 addr:$src1)),
- (i8 imm:$src2))))]>;
-}
-} // ExeDomain = SSEPackedInt
-
+multiclass sse2_pshuffle<string OpcodeStr, ValueType vt128, ValueType vt256,
+ SDNode OpNode> {
let Predicates = [HasAVX] in {
- let AddedComplexity = 5 in
- defm VPSHUFD : sse2_pshuffle<"vpshufd", v4i32, X86PShufd>, TB, OpSize, VEX;
-
- // SSE2 with ImmT == Imm8 and XS prefix.
- defm VPSHUFHW : sse2_pshuffle<"vpshufhw", v8i16, X86PShufhw>, XS, VEX;
-
- // SSE2 with ImmT == Imm8 and XD prefix.
- defm VPSHUFLW : sse2_pshuffle<"vpshuflw", v8i16, X86PShuflw>, XD, VEX;
-
- def : Pat<(v4f32 (X86PShufd (memopv4f32 addr:$src1), (i8 imm:$imm))),
- (VPSHUFDmi addr:$src1, imm:$imm)>;
- def : Pat<(v4f32 (X86PShufd VR128:$src1, (i8 imm:$imm))),
- (VPSHUFDri VR128:$src1, imm:$imm)>;
+ def V#NAME#ri : Ii8<0x70, MRMSrcReg, (outs VR128:$dst),
+ (ins VR128:$src1, i8imm:$src2),
+ !strconcat("v", OpcodeStr,
+ "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
+ [(set VR128:$dst,
+ (vt128 (OpNode VR128:$src1, (i8 imm:$src2))))],
+ IIC_SSE_PSHUF>, VEX;
+ def V#NAME#mi : Ii8<0x70, MRMSrcMem, (outs VR128:$dst),
+ (ins i128mem:$src1, i8imm:$src2),
+ !strconcat("v", OpcodeStr,
+ "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
+ [(set VR128:$dst,
+ (vt128 (OpNode (bitconvert (memopv2i64 addr:$src1)),
+ (i8 imm:$src2))))], IIC_SSE_PSHUF>, VEX;
}
let Predicates = [HasAVX2] in {
- defm VPSHUFD : sse2_pshuffle_y<"vpshufd", v8i32, X86PShufd>, TB, OpSize, VEX;
- defm VPSHUFHW : sse2_pshuffle_y<"vpshufhw", v16i16, X86PShufhw>, XS, VEX;
- defm VPSHUFLW : sse2_pshuffle_y<"vpshuflw", v16i16, X86PShuflw>, XD, VEX;
+ def V#NAME#Yri : Ii8<0x70, MRMSrcReg, (outs VR256:$dst),
+ (ins VR256:$src1, i8imm:$src2),
+ !strconcat("v", OpcodeStr,
+ "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
+ [(set VR256:$dst,
+ (vt256 (OpNode VR256:$src1, (i8 imm:$src2))))],
+ IIC_SSE_PSHUF>, VEX, VEX_L;
+ def V#NAME#Ymi : Ii8<0x70, MRMSrcMem, (outs VR256:$dst),
+ (ins i256mem:$src1, i8imm:$src2),
+ !strconcat("v", OpcodeStr,
+ "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
+ [(set VR256:$dst,
+ (vt256 (OpNode (bitconvert (memopv4i64 addr:$src1)),
+ (i8 imm:$src2))))], IIC_SSE_PSHUF>, VEX, VEX_L;
+}
+
+let Predicates = [UseSSE2] in {
+ def ri : Ii8<0x70, MRMSrcReg,
+ (outs VR128:$dst), (ins VR128:$src1, i8imm:$src2),
+ !strconcat(OpcodeStr,
+ "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
+ [(set VR128:$dst,
+ (vt128 (OpNode VR128:$src1, (i8 imm:$src2))))],
+ IIC_SSE_PSHUF>;
+ def mi : Ii8<0x70, MRMSrcMem,
+ (outs VR128:$dst), (ins i128mem:$src1, i8imm:$src2),
+ !strconcat(OpcodeStr,
+ "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
+ [(set VR128:$dst,
+ (vt128 (OpNode (bitconvert (memopv2i64 addr:$src1)),
+ (i8 imm:$src2))))], IIC_SSE_PSHUF>;
}
+}
+} // ExeDomain = SSEPackedInt
-let Predicates = [HasSSE2] in {
- let AddedComplexity = 5 in
- defm PSHUFD : sse2_pshuffle<"pshufd", v4i32, X86PShufd>, TB, OpSize;
-
- // SSE2 with ImmT == Imm8 and XS prefix.
- defm PSHUFHW : sse2_pshuffle<"pshufhw", v8i16, X86PShufhw>, XS;
+defm PSHUFD : sse2_pshuffle<"pshufd", v4i32, v8i32, X86PShufd>, TB, OpSize;
+defm PSHUFHW : sse2_pshuffle<"pshufhw", v8i16, v16i16, X86PShufhw>, XS;
+defm PSHUFLW : sse2_pshuffle<"pshuflw", v8i16, v16i16, X86PShuflw>, XD;
- // SSE2 with ImmT == Imm8 and XD prefix.
- defm PSHUFLW : sse2_pshuffle<"pshuflw", v8i16, X86PShuflw>, XD;
+let Predicates = [HasAVX] in {
+ def : Pat<(v4f32 (X86PShufd (memopv4f32 addr:$src1), (i8 imm:$imm))),
+ (VPSHUFDmi addr:$src1, imm:$imm)>;
+ def : Pat<(v4f32 (X86PShufd VR128:$src1, (i8 imm:$imm))),
+ (VPSHUFDri VR128:$src1, imm:$imm)>;
+}
- def : Pat<(v4f32 (X86PShufd (memopv4f32 addr:$src1), (i8 imm:$imm))),
- (PSHUFDmi addr:$src1, imm:$imm)>;
- def : Pat<(v4f32 (X86PShufd VR128:$src1, (i8 imm:$imm))),
- (PSHUFDri VR128:$src1, imm:$imm)>;
+let Predicates = [UseSSE2] in {
+ def : Pat<(v4f32 (X86PShufd (memopv4f32 addr:$src1), (i8 imm:$imm))),
+ (PSHUFDmi addr:$src1, imm:$imm)>;
+ def : Pat<(v4f32 (X86PShufd VR128:$src1, (i8 imm:$imm))),
+ (PSHUFDri VR128:$src1, imm:$imm)>;
}
//===---------------------------------------------------------------------===//
let Predicates = [HasAVX2] in {
defm VPUNPCKLBW : sse2_unpack_y<0x60, "vpunpcklbw", v32i8, X86Unpckl,
- bc_v32i8>, VEX_4V;
+ bc_v32i8>, VEX_4V, VEX_L;
defm VPUNPCKLWD : sse2_unpack_y<0x61, "vpunpcklwd", v16i16, X86Unpckl,
- bc_v16i16>, VEX_4V;
+ bc_v16i16>, VEX_4V, VEX_L;
defm VPUNPCKLDQ : sse2_unpack_y<0x62, "vpunpckldq", v8i32, X86Unpckl,
- bc_v8i32>, VEX_4V;
+ bc_v8i32>, VEX_4V, VEX_L;
defm VPUNPCKLQDQ : sse2_unpack_y<0x6C, "vpunpcklqdq", v4i64, X86Unpckl,
- bc_v4i64>, VEX_4V;
+ bc_v4i64>, VEX_4V, VEX_L;
defm VPUNPCKHBW : sse2_unpack_y<0x68, "vpunpckhbw", v32i8, X86Unpckh,
- bc_v32i8>, VEX_4V;
+ bc_v32i8>, VEX_4V, VEX_L;
defm VPUNPCKHWD : sse2_unpack_y<0x69, "vpunpckhwd", v16i16, X86Unpckh,
- bc_v16i16>, VEX_4V;
+ bc_v16i16>, VEX_4V, VEX_L;
defm VPUNPCKHDQ : sse2_unpack_y<0x6A, "vpunpckhdq", v8i32, X86Unpckh,
- bc_v8i32>, VEX_4V;
+ bc_v8i32>, VEX_4V, VEX_L;
defm VPUNPCKHQDQ : sse2_unpack_y<0x6D, "vpunpckhqdq", v4i64, X86Unpckh,
- bc_v4i64>, VEX_4V;
+ bc_v4i64>, VEX_4V, VEX_L;
}
let Constraints = "$src1 = $dst" in {
defm PUNPCKHQDQ : sse2_unpack<0x6D, "punpckhqdq", v2i64, X86Unpckh,
bc_v2i64>;
}
-} // ExeDomain = SSEPackedInt
-
-// Patterns for using AVX1 instructions with integer vectors
-// Here to give AVX2 priority
-let Predicates = [HasAVX] in {
- def : Pat<(v8i32 (X86Unpckl VR256:$src1, (bc_v8i32 (memopv4i64 addr:$src2)))),
- (VUNPCKLPSYrm VR256:$src1, addr:$src2)>;
- def : Pat<(v8i32 (X86Unpckl VR256:$src1, VR256:$src2)),
- (VUNPCKLPSYrr VR256:$src1, VR256:$src2)>;
- def : Pat<(v8i32 (X86Unpckh VR256:$src1, (bc_v8i32 (memopv4i64 addr:$src2)))),
- (VUNPCKHPSYrm VR256:$src1, addr:$src2)>;
- def : Pat<(v8i32 (X86Unpckh VR256:$src1, VR256:$src2)),
- (VUNPCKHPSYrr VR256:$src1, VR256:$src2)>;
-
- def : Pat<(v4i64 (X86Unpckl VR256:$src1, (memopv4i64 addr:$src2))),
- (VUNPCKLPDYrm VR256:$src1, addr:$src2)>;
- def : Pat<(v4i64 (X86Unpckl VR256:$src1, VR256:$src2)),
- (VUNPCKLPDYrr VR256:$src1, VR256:$src2)>;
- def : Pat<(v4i64 (X86Unpckh VR256:$src1, (memopv4i64 addr:$src2))),
- (VUNPCKHPDYrm VR256:$src1, addr:$src2)>;
- def : Pat<(v4i64 (X86Unpckh VR256:$src1, VR256:$src2)),
- (VUNPCKHPDYrr VR256:$src1, VR256:$src2)>;
-}
+} // ExeDomain = SSEPackedInt
//===---------------------------------------------------------------------===//
// SSE2 - Packed Integer Extract and Insert
}
let Constraints = "$src1 = $dst" in
- defm PINSRW : sse2_pinsrw, TB, OpSize, Requires<[HasSSE2]>;
+ defm PINSRW : sse2_pinsrw, TB, OpSize, Requires<[UseSSE2]>;
} // ExeDomain = SSEPackedInt
let Predicates = [HasAVX2] in {
def VPMOVMSKBYrr : VPDI<0xD7, MRMSrcReg, (outs GR32:$dst), (ins VR256:$src),
"pmovmskb\t{$src, $dst|$dst, $src}",
- [(set GR32:$dst, (int_x86_avx2_pmovmskb VR256:$src))]>, VEX;
+ [(set GR32:$dst, (int_x86_avx2_pmovmskb VR256:$src))]>, VEX, VEX_L;
def VPMOVMSKBYr64r : VPDI<0xD7, MRMSrcReg, (outs GR64:$dst), (ins VR256:$src),
- "pmovmskb\t{$src, $dst|$dst, $src}", []>, VEX;
+ "pmovmskb\t{$src, $dst|$dst, $src}", []>, VEX, VEX_L;
}
def PMOVMSKBrr : PDI<0xD7, MRMSrcReg, (outs GR32:$dst), (ins VR128:$src),
// Move Packed Doubleword Int first element to Doubleword Int
//
def VMOVPQIto64rr : I<0x7E, MRMDestReg, (outs GR64:$dst), (ins VR128:$src),
- "mov{d|q}\t{$src, $dst|$dst, $src}",
+ "vmov{d|q}\t{$src, $dst|$dst, $src}",
[(set GR64:$dst, (vector_extract (v2i64 VR128:$src),
(iPTR 0)))],
IIC_SSE_MOVD_ToGP>,
// Bitcast FR64 <-> GR64
//
let Predicates = [HasAVX] in
-def VMOV64toSDrm : SSDI<0x7E, MRMSrcMem, (outs FR64:$dst), (ins i64mem:$src),
+def VMOV64toSDrm : S2SI<0x7E, MRMSrcMem, (outs FR64:$dst), (ins i64mem:$src),
"vmovq\t{$src, $dst|$dst, $src}",
[(set FR64:$dst, (bitconvert (loadi64 addr:$src)))]>,
VEX;
[(store (i64 (bitconvert FR64:$src)), addr:$dst)],
IIC_SSE_MOVDQ>, VEX;
-def MOV64toSDrm : SSDI<0x7E, MRMSrcMem, (outs FR64:$dst), (ins i64mem:$src),
+def MOV64toSDrm : S2SI<0x7E, MRMSrcMem, (outs FR64:$dst), (ins i64mem:$src),
"movq\t{$src, $dst|$dst, $src}",
[(set FR64:$dst, (bitconvert (loadi64 addr:$src)))],
IIC_SSE_MOVDQ>;
}
// Use regular 128-bit instructions to match 256-bit scalar_to_vec+zext.
def : Pat<(v8i32 (X86vzmovl (insert_subvector undef,
- (v4i32 (scalar_to_vector GR32:$src)),(i32 0)))),
+ (v4i32 (scalar_to_vector GR32:$src)),(iPTR 0)))),
(SUBREG_TO_REG (i32 0), (VMOVZDI2PDIrr GR32:$src), sub_xmm)>;
def : Pat<(v4i64 (X86vzmovl (insert_subvector undef,
- (v2i64 (scalar_to_vector GR64:$src)),(i32 0)))),
+ (v2i64 (scalar_to_vector GR64:$src)),(iPTR 0)))),
(SUBREG_TO_REG (i64 0), (VMOVZQI2PQIrr GR64:$src), sub_xmm)>;
}
-let Predicates = [HasSSE2], AddedComplexity = 20 in {
+let Predicates = [UseSSE2], AddedComplexity = 20 in {
def : Pat<(v4i32 (X86vzmovl (bc_v4i32 (loadv4f32 addr:$src)))),
(MOVZDI2PDIrm addr:$src)>;
def : Pat<(v4i32 (X86vzmovl (bc_v4i32 (loadv2i64 addr:$src)))),
[(set VR128:$dst,
(v2i64 (scalar_to_vector (loadi64 addr:$src))))],
IIC_SSE_MOVDQ>, XS,
- Requires<[HasSSE2]>; // SSE2 instruction with XS Prefix
+ Requires<[UseSSE2]>; // SSE2 instruction with XS Prefix
//===---------------------------------------------------------------------===//
// Move Packed Quadword Int to Quadword Int
(v2i64 (X86vzmovl (v2i64 (scalar_to_vector
(loadi64 addr:$src))))))],
IIC_SSE_MOVDQ>,
- XS, Requires<[HasSSE2]>;
+ XS, Requires<[UseSSE2]>;
let Predicates = [HasAVX], AddedComplexity = 20 in {
def : Pat<(v2i64 (X86vzmovl (loadv2i64 addr:$src))),
(VMOVZQI2PQIrm addr:$src)>;
}
-let Predicates = [HasSSE2], AddedComplexity = 20 in {
+let Predicates = [UseSSE2], AddedComplexity = 20 in {
def : Pat<(v2i64 (X86vzmovl (loadv2i64 addr:$src))),
(MOVZQI2PQIrm addr:$src)>;
def : Pat<(v2i64 (X86vzmovl (bc_v2i64 (loadv4f32 addr:$src)))),
"movq\t{$src, $dst|$dst, $src}",
[(set VR128:$dst, (v2i64 (X86vzmovl (v2i64 VR128:$src))))],
IIC_SSE_MOVQ_RR>,
- XS, Requires<[HasSSE2]>;
+ XS, Requires<[UseSSE2]>;
let AddedComplexity = 20 in
def VMOVZPQILo2PQIrm : I<0x7E, MRMSrcMem, (outs VR128:$dst), (ins i128mem:$src),
[(set VR128:$dst, (v2i64 (X86vzmovl
(loadv2i64 addr:$src))))],
IIC_SSE_MOVDQ>,
- XS, Requires<[HasSSE2]>;
+ XS, Requires<[UseSSE2]>;
}
let AddedComplexity = 20 in {
def : Pat<(v2f64 (X86vzmovl (v2f64 VR128:$src))),
(VMOVZPQILo2PQIrr VR128:$src)>;
}
- let Predicates = [HasSSE2] in {
+ let Predicates = [UseSSE2] in {
def : Pat<(v2i64 (X86vzmovl (loadv2i64 addr:$src))),
(MOVZPQILo2PQIrm addr:$src)>;
def : Pat<(v2f64 (X86vzmovl (v2f64 VR128:$src))),
defm VMOVSLDUP : sse3_replicate_sfp<0x12, X86Movsldup, "vmovsldup",
v4f32, VR128, memopv4f32, f128mem>, VEX;
defm VMOVSHDUPY : sse3_replicate_sfp<0x16, X86Movshdup, "vmovshdup",
- v8f32, VR256, memopv8f32, f256mem>, VEX;
+ v8f32, VR256, memopv8f32, f256mem>, VEX, VEX_L;
defm VMOVSLDUPY : sse3_replicate_sfp<0x12, X86Movsldup, "vmovsldup",
- v8f32, VR256, memopv8f32, f256mem>, VEX;
+ v8f32, VR256, memopv8f32, f256mem>, VEX, VEX_L;
}
defm MOVSHDUP : sse3_replicate_sfp<0x16, X86Movshdup, "movshdup", v4f32, VR128,
memopv4f32, f128mem>;
(VMOVSLDUPYrm addr:$src)>;
}
-let Predicates = [HasSSE3] in {
+let Predicates = [UseSSE3] in {
def : Pat<(v4i32 (X86Movshdup VR128:$src)),
(MOVSHDUPrr VR128:$src)>;
def : Pat<(v4i32 (X86Movshdup (bc_v4i32 (memopv2i64 addr:$src)))),
let Predicates = [HasAVX] in {
defm VMOVDDUP : sse3_replicate_dfp<"vmovddup">, VEX;
- defm VMOVDDUPY : sse3_replicate_dfp_y<"vmovddup">, VEX;
+ defm VMOVDDUPY : sse3_replicate_dfp_y<"vmovddup">, VEX, VEX_L;
}
defm MOVDDUP : sse3_replicate_dfp<"movddup">;
(VMOVDDUPYrr VR256:$src)>;
}
-let Predicates = [HasSSE3] in {
+let Predicates = [UseSSE3] in {
def : Pat<(X86Movddup (memopv2f64 addr:$src)),
(MOVDDUPrm addr:$src)>;
def : Pat<(X86Movddup (bc_v2f64 (memopv4f32 addr:$src))),
[(set VR128:$dst, (int_x86_sse3_ldu_dq addr:$src))]>, VEX;
def VLDDQUYrm : S3DI<0xF0, MRMSrcMem, (outs VR256:$dst), (ins i256mem:$src),
"vlddqu\t{$src, $dst|$dst, $src}",
- [(set VR256:$dst, (int_x86_avx_ldu_dq_256 addr:$src))]>, VEX;
+ [(set VR256:$dst, (int_x86_avx_ldu_dq_256 addr:$src))]>,
+ VEX, VEX_L;
}
def LDDQUrm : S3DI<0xF0, MRMSrcMem, (outs VR128:$dst), (ins i128mem:$src),
"lddqu\t{$src, $dst|$dst, $src}",
defm VADDSUBPS : sse3_addsub<int_x86_sse3_addsub_ps, "vaddsubps", VR128,
f128mem, SSE_ALU_F32P, 0>, TB, XD, VEX_4V;
defm VADDSUBPSY : sse3_addsub<int_x86_avx_addsub_ps_256, "vaddsubps", VR256,
- f256mem, SSE_ALU_F32P, 0>, TB, XD, VEX_4V;
+ f256mem, SSE_ALU_F32P, 0>, TB, XD, VEX_4V, VEX_L;
}
let ExeDomain = SSEPackedDouble in {
defm VADDSUBPD : sse3_addsub<int_x86_sse3_addsub_pd, "vaddsubpd", VR128,
f128mem, SSE_ALU_F64P, 0>, TB, OpSize, VEX_4V;
defm VADDSUBPDY : sse3_addsub<int_x86_avx_addsub_pd_256, "vaddsubpd", VR256,
- f256mem, SSE_ALU_F64P, 0>, TB, OpSize, VEX_4V;
+ f256mem, SSE_ALU_F64P, 0>, TB, OpSize, VEX_4V, VEX_L;
}
}
-let Constraints = "$src1 = $dst", Predicates = [HasSSE3] in {
+let Constraints = "$src1 = $dst", Predicates = [UseSSE3] in {
let ExeDomain = SSEPackedSingle in
defm ADDSUBPS : sse3_addsub<int_x86_sse3_addsub_ps, "addsubps", VR128,
f128mem, SSE_ALU_F32P>, TB, XD;
defm VHSUBPS : S3D_Int<0x7D, "vhsubps", v4f32, VR128, f128mem,
X86fhsub, 0>, VEX_4V;
defm VHADDPSY : S3D_Int<0x7C, "vhaddps", v8f32, VR256, f256mem,
- X86fhadd, 0>, VEX_4V;
+ X86fhadd, 0>, VEX_4V, VEX_L;
defm VHSUBPSY : S3D_Int<0x7D, "vhsubps", v8f32, VR256, f256mem,
- X86fhsub, 0>, VEX_4V;
+ X86fhsub, 0>, VEX_4V, VEX_L;
}
let ExeDomain = SSEPackedDouble in {
defm VHADDPD : S3_Int <0x7C, "vhaddpd", v2f64, VR128, f128mem,
defm VHSUBPD : S3_Int <0x7D, "vhsubpd", v2f64, VR128, f128mem,
X86fhsub, 0>, VEX_4V;
defm VHADDPDY : S3_Int <0x7C, "vhaddpd", v4f64, VR256, f256mem,
- X86fhadd, 0>, VEX_4V;
+ X86fhadd, 0>, VEX_4V, VEX_L;
defm VHSUBPDY : S3_Int <0x7D, "vhsubpd", v4f64, VR256, f256mem,
- X86fhsub, 0>, VEX_4V;
+ X86fhsub, 0>, VEX_4V, VEX_L;
}
}
let Predicates = [HasAVX2] in {
defm VPABSB : SS3I_unop_rm_int_y<0x1C, "vpabsb",
- int_x86_avx2_pabs_b>, VEX;
+ int_x86_avx2_pabs_b>, VEX, VEX_L;
defm VPABSW : SS3I_unop_rm_int_y<0x1D, "vpabsw",
- int_x86_avx2_pabs_w>, VEX;
+ int_x86_avx2_pabs_w>, VEX, VEX_L;
defm VPABSD : SS3I_unop_rm_int_y<0x1E, "vpabsd",
- int_x86_avx2_pabs_d>, VEX;
+ int_x86_avx2_pabs_d>, VEX, VEX_L;
}
defm PABSB : SS3I_unop_rm_int<0x1C, "pabsb",
let isCommutable = 0 in {
defm VPHADDWY : SS3I_binop_rm<0x01, "vphaddw", X86hadd, v16i16, VR256,
memopv4i64, i256mem,
- SSE_PHADDSUBW, 0>, VEX_4V;
+ SSE_PHADDSUBW, 0>, VEX_4V, VEX_L;
defm VPHADDDY : SS3I_binop_rm<0x02, "vphaddd", X86hadd, v8i32, VR256,
memopv4i64, i256mem,
- SSE_PHADDSUBW, 0>, VEX_4V;
+ SSE_PHADDSUBW, 0>, VEX_4V, VEX_L;
defm VPHSUBWY : SS3I_binop_rm<0x05, "vphsubw", X86hsub, v16i16, VR256,
memopv4i64, i256mem,
- SSE_PHADDSUBW, 0>, VEX_4V;
+ SSE_PHADDSUBW, 0>, VEX_4V, VEX_L;
defm VPHSUBDY : SS3I_binop_rm<0x06, "vphsubd", X86hsub, v8i32, VR256,
memopv4i64, i256mem,
- SSE_PHADDSUBW, 0>, VEX_4V;
+ SSE_PHADDSUBW, 0>, VEX_4V, VEX_L;
defm VPSIGNBY : SS3I_binop_rm<0x08, "vpsignb", X86psign, v32i8, VR256,
memopv4i64, i256mem,
- SSE_PHADDSUBW, 0>, VEX_4V;
+ SSE_PHADDSUBW, 0>, VEX_4V, VEX_L;
defm VPSIGNWY : SS3I_binop_rm<0x09, "vpsignw", X86psign, v16i16, VR256,
memopv4i64, i256mem,
- SSE_PHADDSUBW, 0>, VEX_4V;
+ SSE_PHADDSUBW, 0>, VEX_4V, VEX_L;
defm VPSIGNDY : SS3I_binop_rm<0x0A, "vpsignd", X86psign, v8i32, VR256,
memopv4i64, i256mem,
- SSE_PHADDSUBW, 0>, VEX_4V;
+ SSE_PHADDSUBW, 0>, VEX_4V, VEX_L;
defm VPSHUFBY : SS3I_binop_rm<0x00, "vpshufb", X86pshufb, v32i8, VR256,
memopv4i64, i256mem,
- SSE_PHADDSUBW, 0>, VEX_4V;
+ SSE_PHADDSUBW, 0>, VEX_4V, VEX_L;
defm VPHADDSW : SS3I_binop_rm_int_y<0x03, "vphaddsw",
- int_x86_avx2_phadd_sw>, VEX_4V;
+ int_x86_avx2_phadd_sw>, VEX_4V, VEX_L;
defm VPHSUBSW : SS3I_binop_rm_int_y<0x07, "vphsubsw",
- int_x86_avx2_phsub_sw>, VEX_4V;
+ int_x86_avx2_phsub_sw>, VEX_4V, VEX_L;
defm VPMADDUBSW : SS3I_binop_rm_int_y<0x04, "vpmaddubsw",
- int_x86_avx2_pmadd_ub_sw>, VEX_4V;
+ int_x86_avx2_pmadd_ub_sw>, VEX_4V, VEX_L;
}
defm VPMULHRSW : SS3I_binop_rm_int_y<0x0B, "vpmulhrsw",
- int_x86_avx2_pmul_hr_sw>, VEX_4V;
+ int_x86_avx2_pmul_hr_sw>, VEX_4V, VEX_L;
}
// None of these have i8 immediate fields.
// SSSE3 - Packed Align Instruction Patterns
//===---------------------------------------------------------------------===//
-multiclass ssse3_palign<string asm, bit Is2Addr = 1> {
+multiclass ssse3_palignr<string asm, bit Is2Addr = 1> {
let neverHasSideEffects = 1 in {
def R128rr : SS3AI<0x0F, MRMSrcReg, (outs VR128:$dst),
(ins VR128:$src1, VR128:$src2, i8imm:$src3),
}
}
-multiclass ssse3_palign_y<string asm, bit Is2Addr = 1> {
+multiclass ssse3_palignr_y<string asm, bit Is2Addr = 1> {
let neverHasSideEffects = 1 in {
def R256rr : SS3AI<0x0F, MRMSrcReg, (outs VR256:$dst),
(ins VR256:$src1, VR256:$src2, i8imm:$src3),
}
let Predicates = [HasAVX] in
- defm VPALIGN : ssse3_palign<"vpalignr", 0>, VEX_4V;
+ defm VPALIGN : ssse3_palignr<"vpalignr", 0>, VEX_4V;
let Predicates = [HasAVX2] in
- defm VPALIGN : ssse3_palign_y<"vpalignr", 0>, VEX_4V;
-let Constraints = "$src1 = $dst", Predicates = [HasSSSE3] in
- defm PALIGN : ssse3_palign<"palignr">;
+ defm VPALIGN : ssse3_palignr_y<"vpalignr", 0>, VEX_4V, VEX_L;
+let Constraints = "$src1 = $dst", Predicates = [UseSSSE3] in
+ defm PALIGN : ssse3_palignr<"palignr">;
let Predicates = [HasAVX2] in {
-def : Pat<(v8i32 (X86PAlign VR256:$src1, VR256:$src2, (i8 imm:$imm))),
+def : Pat<(v8i32 (X86PAlignr VR256:$src1, VR256:$src2, (i8 imm:$imm))),
(VPALIGNR256rr VR256:$src2, VR256:$src1, imm:$imm)>;
-def : Pat<(v8f32 (X86PAlign VR256:$src1, VR256:$src2, (i8 imm:$imm))),
+def : Pat<(v8f32 (X86PAlignr VR256:$src1, VR256:$src2, (i8 imm:$imm))),
(VPALIGNR256rr VR256:$src2, VR256:$src1, imm:$imm)>;
-def : Pat<(v16i16 (X86PAlign VR256:$src1, VR256:$src2, (i8 imm:$imm))),
+def : Pat<(v16i16 (X86PAlignr VR256:$src1, VR256:$src2, (i8 imm:$imm))),
(VPALIGNR256rr VR256:$src2, VR256:$src1, imm:$imm)>;
-def : Pat<(v32i8 (X86PAlign VR256:$src1, VR256:$src2, (i8 imm:$imm))),
+def : Pat<(v32i8 (X86PAlignr VR256:$src1, VR256:$src2, (i8 imm:$imm))),
(VPALIGNR256rr VR256:$src2, VR256:$src1, imm:$imm)>;
}
let Predicates = [HasAVX] in {
-def : Pat<(v4i32 (X86PAlign VR128:$src1, VR128:$src2, (i8 imm:$imm))),
+def : Pat<(v4i32 (X86PAlignr VR128:$src1, VR128:$src2, (i8 imm:$imm))),
(VPALIGNR128rr VR128:$src2, VR128:$src1, imm:$imm)>;
-def : Pat<(v4f32 (X86PAlign VR128:$src1, VR128:$src2, (i8 imm:$imm))),
+def : Pat<(v4f32 (X86PAlignr VR128:$src1, VR128:$src2, (i8 imm:$imm))),
(VPALIGNR128rr VR128:$src2, VR128:$src1, imm:$imm)>;
-def : Pat<(v8i16 (X86PAlign VR128:$src1, VR128:$src2, (i8 imm:$imm))),
+def : Pat<(v8i16 (X86PAlignr VR128:$src1, VR128:$src2, (i8 imm:$imm))),
(VPALIGNR128rr VR128:$src2, VR128:$src1, imm:$imm)>;
-def : Pat<(v16i8 (X86PAlign VR128:$src1, VR128:$src2, (i8 imm:$imm))),
+def : Pat<(v16i8 (X86PAlignr VR128:$src1, VR128:$src2, (i8 imm:$imm))),
(VPALIGNR128rr VR128:$src2, VR128:$src1, imm:$imm)>;
}
-let Predicates = [HasSSSE3] in {
-def : Pat<(v4i32 (X86PAlign VR128:$src1, VR128:$src2, (i8 imm:$imm))),
+let Predicates = [UseSSSE3] in {
+def : Pat<(v4i32 (X86PAlignr VR128:$src1, VR128:$src2, (i8 imm:$imm))),
(PALIGNR128rr VR128:$src2, VR128:$src1, imm:$imm)>;
-def : Pat<(v4f32 (X86PAlign VR128:$src1, VR128:$src2, (i8 imm:$imm))),
+def : Pat<(v4f32 (X86PAlignr VR128:$src1, VR128:$src2, (i8 imm:$imm))),
(PALIGNR128rr VR128:$src2, VR128:$src1, imm:$imm)>;
-def : Pat<(v8i16 (X86PAlign VR128:$src1, VR128:$src2, (i8 imm:$imm))),
+def : Pat<(v8i16 (X86PAlignr VR128:$src1, VR128:$src2, (i8 imm:$imm))),
(PALIGNR128rr VR128:$src2, VR128:$src1, imm:$imm)>;
-def : Pat<(v16i8 (X86PAlign VR128:$src1, VR128:$src2, (i8 imm:$imm))),
+def : Pat<(v16i8 (X86PAlignr VR128:$src1, VR128:$src2, (i8 imm:$imm))),
(PALIGNR128rr VR128:$src2, VR128:$src1, imm:$imm)>;
}
def MONITOR : PseudoI<(outs), (ins i32mem:$src1, GR32:$src2, GR32:$src3),
[(int_x86_sse3_monitor addr:$src1, GR32:$src2, GR32:$src3)]>,
Requires<[HasSSE3]>;
-def MWAIT : PseudoI<(outs), (ins GR32:$src1, GR32:$src2),
- [(int_x86_sse3_mwait GR32:$src1, GR32:$src2)]>,
- Requires<[HasSSE3]>;
}
let Uses = [EAX, ECX, EDX] in
def MONITORrrr : I<0x01, MRM_C8, (outs), (ins), "monitor", [], IIC_SSE_MONITOR>,
TB, Requires<[HasSSE3]>;
let Uses = [ECX, EAX] in
-def MWAITrr : I<0x01, MRM_C9, (outs), (ins), "mwait", [], IIC_SSE_MWAIT>,
+def MWAITrr : I<0x01, MRM_C9, (outs), (ins), "mwait",
+ [(int_x86_sse3_mwait ECX, EAX)], IIC_SSE_MWAIT>,
TB, Requires<[HasSSE3]>;
def : InstAlias<"mwait %eax, %ecx", (MWAITrr)>, Requires<[In32BitMode]>;
let Predicates = [HasAVX2] in {
defm VPMOVSXBW : SS41I_binop_rm_int16_y<0x20, "vpmovsxbw",
- int_x86_avx2_pmovsxbw>, VEX;
+ int_x86_avx2_pmovsxbw>, VEX, VEX_L;
defm VPMOVSXWD : SS41I_binop_rm_int16_y<0x23, "vpmovsxwd",
- int_x86_avx2_pmovsxwd>, VEX;
+ int_x86_avx2_pmovsxwd>, VEX, VEX_L;
defm VPMOVSXDQ : SS41I_binop_rm_int16_y<0x25, "vpmovsxdq",
- int_x86_avx2_pmovsxdq>, VEX;
+ int_x86_avx2_pmovsxdq>, VEX, VEX_L;
defm VPMOVZXBW : SS41I_binop_rm_int16_y<0x30, "vpmovzxbw",
- int_x86_avx2_pmovzxbw>, VEX;
+ int_x86_avx2_pmovzxbw>, VEX, VEX_L;
defm VPMOVZXWD : SS41I_binop_rm_int16_y<0x33, "vpmovzxwd",
- int_x86_avx2_pmovzxwd>, VEX;
+ int_x86_avx2_pmovzxwd>, VEX, VEX_L;
defm VPMOVZXDQ : SS41I_binop_rm_int16_y<0x35, "vpmovzxdq",
- int_x86_avx2_pmovzxdq>, VEX;
+ int_x86_avx2_pmovzxdq>, VEX, VEX_L;
}
defm PMOVSXBW : SS41I_binop_rm_int8<0x20, "pmovsxbw", int_x86_sse41_pmovsxbw>;
(VPMOVSXBWrm addr:$src)>;
def : Pat<(int_x86_sse41_pmovsxbw (vzload_v2i64 addr:$src)),
(VPMOVSXBWrm addr:$src)>;
+ def : Pat<(int_x86_sse41_pmovsxbw (bc_v16i8 (loadv2i64 addr:$src))),
+ (VPMOVSXBWrm addr:$src)>;
def : Pat<(int_x86_sse41_pmovsxwd (vzmovl_v2i64 addr:$src)),
(VPMOVSXWDrm addr:$src)>;
def : Pat<(int_x86_sse41_pmovsxwd (vzload_v2i64 addr:$src)),
(VPMOVSXWDrm addr:$src)>;
+ def : Pat<(int_x86_sse41_pmovsxwd (bc_v8i16 (loadv2i64 addr:$src))),
+ (VPMOVSXWDrm addr:$src)>;
def : Pat<(int_x86_sse41_pmovsxdq (vzmovl_v2i64 addr:$src)),
(VPMOVSXDQrm addr:$src)>;
def : Pat<(int_x86_sse41_pmovsxdq (vzload_v2i64 addr:$src)),
(VPMOVSXDQrm addr:$src)>;
+ def : Pat<(int_x86_sse41_pmovsxdq (bc_v4i32 (loadv2i64 addr:$src))),
+ (VPMOVSXDQrm addr:$src)>;
def : Pat<(int_x86_sse41_pmovzxbw (vzmovl_v2i64 addr:$src)),
(VPMOVZXBWrm addr:$src)>;
def : Pat<(int_x86_sse41_pmovzxbw (vzload_v2i64 addr:$src)),
(VPMOVZXBWrm addr:$src)>;
+ def : Pat<(int_x86_sse41_pmovzxbw (bc_v16i8 (loadv2i64 addr:$src))),
+ (VPMOVZXBWrm addr:$src)>;
def : Pat<(int_x86_sse41_pmovzxwd (vzmovl_v2i64 addr:$src)),
(VPMOVZXWDrm addr:$src)>;
def : Pat<(int_x86_sse41_pmovzxwd (vzload_v2i64 addr:$src)),
(VPMOVZXWDrm addr:$src)>;
+ def : Pat<(int_x86_sse41_pmovzxwd (bc_v8i16 (loadv2i64 addr:$src))),
+ (VPMOVZXWDrm addr:$src)>;
def : Pat<(int_x86_sse41_pmovzxdq (vzmovl_v2i64 addr:$src)),
(VPMOVZXDQrm addr:$src)>;
def : Pat<(int_x86_sse41_pmovzxdq (vzload_v2i64 addr:$src)),
(VPMOVZXDQrm addr:$src)>;
+ def : Pat<(int_x86_sse41_pmovzxdq (bc_v4i32 (loadv2i64 addr:$src))),
+ (VPMOVZXDQrm addr:$src)>;
}
-let Predicates = [HasSSE41] in {
+let Predicates = [UseSSE41] in {
// Common patterns involving scalar load.
def : Pat<(int_x86_sse41_pmovsxbw (vzmovl_v2i64 addr:$src)),
(PMOVSXBWrm addr:$src)>;
def : Pat<(int_x86_sse41_pmovsxbw (vzload_v2i64 addr:$src)),
(PMOVSXBWrm addr:$src)>;
+ def : Pat<(int_x86_sse41_pmovsxbw (bc_v16i8 (loadv2i64 addr:$src))),
+ (PMOVSXBWrm addr:$src)>;
def : Pat<(int_x86_sse41_pmovsxwd (vzmovl_v2i64 addr:$src)),
(PMOVSXWDrm addr:$src)>;
def : Pat<(int_x86_sse41_pmovsxwd (vzload_v2i64 addr:$src)),
(PMOVSXWDrm addr:$src)>;
+ def : Pat<(int_x86_sse41_pmovsxwd (bc_v8i16 (loadv2i64 addr:$src))),
+ (PMOVSXWDrm addr:$src)>;
def : Pat<(int_x86_sse41_pmovsxdq (vzmovl_v2i64 addr:$src)),
(PMOVSXDQrm addr:$src)>;
def : Pat<(int_x86_sse41_pmovsxdq (vzload_v2i64 addr:$src)),
(PMOVSXDQrm addr:$src)>;
+ def : Pat<(int_x86_sse41_pmovsxdq (bc_v4i32 (loadv2i64 addr:$src))),
+ (PMOVSXDQrm addr:$src)>;
def : Pat<(int_x86_sse41_pmovzxbw (vzmovl_v2i64 addr:$src)),
(PMOVZXBWrm addr:$src)>;
def : Pat<(int_x86_sse41_pmovzxbw (vzload_v2i64 addr:$src)),
(PMOVZXBWrm addr:$src)>;
+ def : Pat<(int_x86_sse41_pmovzxbw (bc_v16i8 (loadv2i64 addr:$src))),
+ (PMOVZXBWrm addr:$src)>;
def : Pat<(int_x86_sse41_pmovzxwd (vzmovl_v2i64 addr:$src)),
(PMOVZXWDrm addr:$src)>;
def : Pat<(int_x86_sse41_pmovzxwd (vzload_v2i64 addr:$src)),
(PMOVZXWDrm addr:$src)>;
+ def : Pat<(int_x86_sse41_pmovzxwd (bc_v8i16 (loadv2i64 addr:$src))),
+ (PMOVZXWDrm addr:$src)>;
def : Pat<(int_x86_sse41_pmovzxdq (vzmovl_v2i64 addr:$src)),
(PMOVZXDQrm addr:$src)>;
def : Pat<(int_x86_sse41_pmovzxdq (vzload_v2i64 addr:$src)),
(PMOVZXDQrm addr:$src)>;
+ def : Pat<(int_x86_sse41_pmovzxdq (bc_v4i32 (loadv2i64 addr:$src))),
+ (PMOVZXDQrm addr:$src)>;
}
let Predicates = [HasAVX2] in {
def : Pat<(v4i32 (X86vsmovl (v8i16 VR128:$src))), (VPMOVSXWDrr VR128:$src)>;
}
-let Predicates = [HasSSE41] in {
+let Predicates = [UseSSE41] in {
def : Pat<(v2i64 (X86vsmovl (v4i32 VR128:$src))), (PMOVSXDQrr VR128:$src)>;
def : Pat<(v4i32 (X86vsmovl (v8i16 VR128:$src))), (PMOVSXWDrr VR128:$src)>;
}
let Predicates = [HasAVX2] in {
defm VPMOVSXBD : SS41I_binop_rm_int8_y<0x21, "vpmovsxbd",
- int_x86_avx2_pmovsxbd>, VEX;
+ int_x86_avx2_pmovsxbd>, VEX, VEX_L;
defm VPMOVSXWQ : SS41I_binop_rm_int8_y<0x24, "vpmovsxwq",
- int_x86_avx2_pmovsxwq>, VEX;
+ int_x86_avx2_pmovsxwq>, VEX, VEX_L;
defm VPMOVZXBD : SS41I_binop_rm_int8_y<0x31, "vpmovzxbd",
- int_x86_avx2_pmovzxbd>, VEX;
+ int_x86_avx2_pmovzxbd>, VEX, VEX_L;
defm VPMOVZXWQ : SS41I_binop_rm_int8_y<0x34, "vpmovzxwq",
- int_x86_avx2_pmovzxwq>, VEX;
+ int_x86_avx2_pmovzxwq>, VEX, VEX_L;
}
defm PMOVSXBD : SS41I_binop_rm_int4<0x21, "pmovsxbd", int_x86_sse41_pmovsxbd>;
(VPMOVZXWQrm addr:$src)>;
}
-let Predicates = [HasSSE41] in {
+let Predicates = [UseSSE41] in {
// Common patterns involving scalar load
def : Pat<(int_x86_sse41_pmovsxbd (vzmovl_v4i32 addr:$src)),
(PMOVSXBDrm addr:$src)>;
}
let Predicates = [HasAVX2] in {
defm VPMOVSXBQ : SS41I_binop_rm_int4_y<0x22, "vpmovsxbq",
- int_x86_avx2_pmovsxbq>, VEX;
+ int_x86_avx2_pmovsxbq>, VEX, VEX_L;
defm VPMOVZXBQ : SS41I_binop_rm_int4_y<0x32, "vpmovzxbq",
- int_x86_avx2_pmovzxbq>, VEX;
+ int_x86_avx2_pmovzxbq>, VEX, VEX_L;
}
defm PMOVSXBQ : SS41I_binop_rm_int2<0x22, "pmovsxbq", int_x86_sse41_pmovsxbq>;
defm PMOVZXBQ : SS41I_binop_rm_int2<0x32, "pmovzxbq", int_x86_sse41_pmovzxbq>;
+let Predicates = [HasAVX2] in {
+ def : Pat<(v16i16 (X86vsext (v16i8 VR128:$src))), (VPMOVSXBWYrr VR128:$src)>;
+ def : Pat<(v8i32 (X86vsext (v16i8 VR128:$src))), (VPMOVSXBDYrr VR128:$src)>;
+ def : Pat<(v4i64 (X86vsext (v16i8 VR128:$src))), (VPMOVSXBQYrr VR128:$src)>;
+
+ def : Pat<(v8i32 (X86vsext (v8i16 VR128:$src))), (VPMOVSXWDYrr VR128:$src)>;
+ def : Pat<(v4i64 (X86vsext (v8i16 VR128:$src))), (VPMOVSXWQYrr VR128:$src)>;
+
+ def : Pat<(v4i64 (X86vsext (v4i32 VR128:$src))), (VPMOVSXDQYrr VR128:$src)>;
+
+ def : Pat<(v16i16 (X86vsext (v32i8 VR256:$src))),
+ (VPMOVSXBWYrr (EXTRACT_SUBREG VR256:$src, sub_xmm))>;
+ def : Pat<(v8i32 (X86vsext (v32i8 VR256:$src))),
+ (VPMOVSXBDYrr (EXTRACT_SUBREG VR256:$src, sub_xmm))>;
+ def : Pat<(v4i64 (X86vsext (v32i8 VR256:$src))),
+ (VPMOVSXBQYrr (EXTRACT_SUBREG VR256:$src, sub_xmm))>;
+
+ def : Pat<(v8i32 (X86vsext (v16i16 VR256:$src))),
+ (VPMOVSXWDYrr (EXTRACT_SUBREG VR256:$src, sub_xmm))>;
+ def : Pat<(v4i64 (X86vsext (v16i16 VR256:$src))),
+ (VPMOVSXWQYrr (EXTRACT_SUBREG VR256:$src, sub_xmm))>;
+
+ def : Pat<(v4i64 (X86vsext (v8i32 VR256:$src))),
+ (VPMOVSXDQYrr (EXTRACT_SUBREG VR256:$src, sub_xmm))>;
+
+ def : Pat<(v8i32 (X86vsmovl (v8i16 (bitconvert (v2i64 (load addr:$src)))))),
+ (VPMOVSXWDYrm addr:$src)>;
+ def : Pat<(v4i64 (X86vsmovl (v4i32 (bitconvert (v2i64 (load addr:$src)))))),
+ (VPMOVSXDQYrm addr:$src)>;
+
+ def : Pat<(v8i32 (X86vsext (v16i8 (bitconvert (v2i64
+ (scalar_to_vector (loadi64 addr:$src))))))),
+ (VPMOVSXBDYrm addr:$src)>;
+ def : Pat<(v8i32 (X86vsext (v16i8 (bitconvert (v2f64
+ (scalar_to_vector (loadf64 addr:$src))))))),
+ (VPMOVSXBDYrm addr:$src)>;
+
+ def : Pat<(v4i64 (X86vsext (v8i16 (bitconvert (v2i64
+ (scalar_to_vector (loadi64 addr:$src))))))),
+ (VPMOVSXWQYrm addr:$src)>;
+ def : Pat<(v4i64 (X86vsext (v8i16 (bitconvert (v2f64
+ (scalar_to_vector (loadf64 addr:$src))))))),
+ (VPMOVSXWQYrm addr:$src)>;
+
+ def : Pat<(v4i64 (X86vsext (v16i8 (bitconvert (v4i32
+ (scalar_to_vector (loadi32 addr:$src))))))),
+ (VPMOVSXBQYrm addr:$src)>;
+}
+
let Predicates = [HasAVX] in {
// Common patterns involving scalar load
def : Pat<(int_x86_sse41_pmovsxbq
(VPMOVZXBQrm addr:$src)>;
}
-let Predicates = [HasSSE41] in {
+let Predicates = [UseSSE41] in {
+ def : Pat<(v8i16 (X86vsext (v16i8 VR128:$src))), (PMOVSXBWrr VR128:$src)>;
+ def : Pat<(v4i32 (X86vsext (v16i8 VR128:$src))), (PMOVSXBDrr VR128:$src)>;
+ def : Pat<(v2i64 (X86vsext (v16i8 VR128:$src))), (PMOVSXBQrr VR128:$src)>;
+
+ def : Pat<(v4i32 (X86vsext (v8i16 VR128:$src))), (PMOVSXWDrr VR128:$src)>;
+ def : Pat<(v2i64 (X86vsext (v8i16 VR128:$src))), (PMOVSXWQrr VR128:$src)>;
+
+ def : Pat<(v2i64 (X86vsext (v4i32 VR128:$src))), (PMOVSXDQrr VR128:$src)>;
+
// Common patterns involving scalar load
def : Pat<(int_x86_sse41_pmovsxbq
(bitconvert (v4i32 (X86vzmovl
(bitconvert (v4i32 (X86vzmovl
(v4i32 (scalar_to_vector (loadi32 addr:$src))))))),
(PMOVZXBQrm addr:$src)>;
+
+ def : Pat<(v4i32 (X86vsext (v8i16 (bitconvert (v2i64
+ (scalar_to_vector (loadi64 addr:$src))))))),
+ (PMOVSXWDrm addr:$src)>;
+ def : Pat<(v4i32 (X86vsext (v8i16 (bitconvert (v2f64
+ (scalar_to_vector (loadf64 addr:$src))))))),
+ (PMOVSXWDrm addr:$src)>;
+ def : Pat<(v4i32 (X86vsext (v16i8 (bitconvert (v4i32
+ (scalar_to_vector (loadi32 addr:$src))))))),
+ (PMOVSXBDrm addr:$src)>;
+ def : Pat<(v2i64 (X86vsext (v8i16 (bitconvert (v4i32
+ (scalar_to_vector (loadi32 addr:$src))))))),
+ (PMOVSXWQrm addr:$src)>;
+ def : Pat<(v2i64 (X86vsext (v16i8 (bitconvert (v4i32
+ (scalar_to_vector (extloadi32i16 addr:$src))))))),
+ (PMOVSXBQrm addr:$src)>;
+ def : Pat<(v2i64 (X86vsext (v4i32 (bitconvert (v2i64
+ (scalar_to_vector (loadi64 addr:$src))))))),
+ (PMOVSXDQrm addr:$src)>;
+ def : Pat<(v2i64 (X86vsext (v4i32 (bitconvert (v2f64
+ (scalar_to_vector (loadf64 addr:$src))))))),
+ (PMOVSXDQrm addr:$src)>;
+ def : Pat<(v8i16 (X86vsext (v16i8 (bitconvert (v2i64
+ (scalar_to_vector (loadi64 addr:$src))))))),
+ (PMOVSXBWrm addr:$src)>;
+ def : Pat<(v8i16 (X86vsext (v16i8 (bitconvert (v2f64
+ (scalar_to_vector (loadf64 addr:$src))))))),
+ (PMOVSXBWrm addr:$src)>;
+}
+
+let Predicates = [HasAVX2] in {
+ def : Pat<(v16i16 (X86vzext (v16i8 VR128:$src))), (VPMOVZXBWYrr VR128:$src)>;
+ def : Pat<(v8i32 (X86vzext (v16i8 VR128:$src))), (VPMOVZXBDYrr VR128:$src)>;
+ def : Pat<(v4i64 (X86vzext (v16i8 VR128:$src))), (VPMOVZXBQYrr VR128:$src)>;
+
+ def : Pat<(v8i32 (X86vzext (v8i16 VR128:$src))), (VPMOVZXWDYrr VR128:$src)>;
+ def : Pat<(v4i64 (X86vzext (v8i16 VR128:$src))), (VPMOVZXWQYrr VR128:$src)>;
+
+ def : Pat<(v4i64 (X86vzext (v4i32 VR128:$src))), (VPMOVZXDQYrr VR128:$src)>;
+
+ def : Pat<(v16i16 (X86vzext (v32i8 VR256:$src))),
+ (VPMOVZXBWYrr (EXTRACT_SUBREG VR256:$src, sub_xmm))>;
+ def : Pat<(v8i32 (X86vzext (v32i8 VR256:$src))),
+ (VPMOVZXBDYrr (EXTRACT_SUBREG VR256:$src, sub_xmm))>;
+ def : Pat<(v4i64 (X86vzext (v32i8 VR256:$src))),
+ (VPMOVZXBQYrr (EXTRACT_SUBREG VR256:$src, sub_xmm))>;
+
+ def : Pat<(v8i32 (X86vzext (v16i16 VR256:$src))),
+ (VPMOVZXWDYrr (EXTRACT_SUBREG VR256:$src, sub_xmm))>;
+ def : Pat<(v4i64 (X86vzext (v16i16 VR256:$src))),
+ (VPMOVZXWQYrr (EXTRACT_SUBREG VR256:$src, sub_xmm))>;
+
+ def : Pat<(v4i64 (X86vzext (v8i32 VR256:$src))),
+ (VPMOVZXDQYrr (EXTRACT_SUBREG VR256:$src, sub_xmm))>;
+}
+
+let Predicates = [HasAVX] in {
+ def : Pat<(v8i16 (X86vzext (v16i8 VR128:$src))), (VPMOVZXBWrr VR128:$src)>;
+ def : Pat<(v4i32 (X86vzext (v16i8 VR128:$src))), (VPMOVZXBDrr VR128:$src)>;
+ def : Pat<(v2i64 (X86vzext (v16i8 VR128:$src))), (VPMOVZXBQrr VR128:$src)>;
+
+ def : Pat<(v4i32 (X86vzext (v8i16 VR128:$src))), (VPMOVZXWDrr VR128:$src)>;
+ def : Pat<(v2i64 (X86vzext (v8i16 VR128:$src))), (VPMOVZXWQrr VR128:$src)>;
+
+ def : Pat<(v2i64 (X86vzext (v4i32 VR128:$src))), (VPMOVZXDQrr VR128:$src)>;
+
+ def : Pat<(v8i16 (X86vzext (v16i8 (bitconvert (v2i64 (scalar_to_vector (loadi64 addr:$src))))))),
+ (VPMOVZXBWrm addr:$src)>;
+ def : Pat<(v8i16 (X86vzext (v16i8 (bitconvert (v2f64 (scalar_to_vector (loadf64 addr:$src))))))),
+ (VPMOVZXBWrm addr:$src)>;
+ def : Pat<(v4i32 (X86vzext (v16i8 (bitconvert (v4i32 (scalar_to_vector (loadi32 addr:$src))))))),
+ (VPMOVZXBDrm addr:$src)>;
+ def : Pat<(v2i64 (X86vzext (v16i8 (bitconvert (v4i32 (scalar_to_vector (loadi16_anyext addr:$src))))))),
+ (VPMOVZXBQrm addr:$src)>;
+
+ def : Pat<(v4i32 (X86vzext (v8i16 (bitconvert (v2i64 (scalar_to_vector (loadi64 addr:$src))))))),
+ (VPMOVZXWDrm addr:$src)>;
+ def : Pat<(v4i32 (X86vzext (v8i16 (bitconvert (v2f64 (scalar_to_vector (loadf64 addr:$src))))))),
+ (VPMOVZXWDrm addr:$src)>;
+ def : Pat<(v2i64 (X86vzext (v8i16 (bitconvert (v4i32 (scalar_to_vector (loadi32 addr:$src))))))),
+ (VPMOVZXWQrm addr:$src)>;
+
+ def : Pat<(v2i64 (X86vzext (v4i32 (bitconvert (v2i64 (scalar_to_vector (loadi64 addr:$src))))))),
+ (VPMOVZXDQrm addr:$src)>;
+ def : Pat<(v2i64 (X86vzext (v4i32 (bitconvert (v2f64 (scalar_to_vector (loadf64 addr:$src))))))),
+ (VPMOVZXDQrm addr:$src)>;
+ def : Pat<(v2i64 (X86vzext (v4i32 (bitconvert (v2i64 (X86vzload addr:$src)))))),
+ (VPMOVZXDQrm addr:$src)>;
+
+ def : Pat<(v8i16 (X86vsext (v16i8 VR128:$src))), (VPMOVSXBWrr VR128:$src)>;
+ def : Pat<(v4i32 (X86vsext (v16i8 VR128:$src))), (VPMOVSXBDrr VR128:$src)>;
+ def : Pat<(v2i64 (X86vsext (v16i8 VR128:$src))), (VPMOVSXBQrr VR128:$src)>;
+
+ def : Pat<(v4i32 (X86vsext (v8i16 VR128:$src))), (VPMOVSXWDrr VR128:$src)>;
+ def : Pat<(v2i64 (X86vsext (v8i16 VR128:$src))), (VPMOVSXWQrr VR128:$src)>;
+
+ def : Pat<(v2i64 (X86vsext (v4i32 VR128:$src))), (VPMOVSXDQrr VR128:$src)>;
+
+ def : Pat<(v4i32 (X86vsext (v8i16 (bitconvert (v2i64
+ (scalar_to_vector (loadi64 addr:$src))))))),
+ (VPMOVSXWDrm addr:$src)>;
+ def : Pat<(v2i64 (X86vsext (v4i32 (bitconvert (v2i64
+ (scalar_to_vector (loadi64 addr:$src))))))),
+ (VPMOVSXDQrm addr:$src)>;
+ def : Pat<(v4i32 (X86vsext (v8i16 (bitconvert (v2f64
+ (scalar_to_vector (loadf64 addr:$src))))))),
+ (VPMOVSXWDrm addr:$src)>;
+ def : Pat<(v2i64 (X86vsext (v4i32 (bitconvert (v2f64
+ (scalar_to_vector (loadf64 addr:$src))))))),
+ (VPMOVSXDQrm addr:$src)>;
+ def : Pat<(v8i16 (X86vsext (v16i8 (bitconvert (v2i64
+ (scalar_to_vector (loadi64 addr:$src))))))),
+ (VPMOVSXBWrm addr:$src)>;
+ def : Pat<(v8i16 (X86vsext (v16i8 (bitconvert (v2f64
+ (scalar_to_vector (loadf64 addr:$src))))))),
+ (VPMOVSXBWrm addr:$src)>;
+
+ def : Pat<(v4i32 (X86vsext (v16i8 (bitconvert (v4i32
+ (scalar_to_vector (loadi32 addr:$src))))))),
+ (VPMOVSXBDrm addr:$src)>;
+ def : Pat<(v2i64 (X86vsext (v8i16 (bitconvert (v4i32
+ (scalar_to_vector (loadi32 addr:$src))))))),
+ (VPMOVSXWQrm addr:$src)>;
+ def : Pat<(v2i64 (X86vsext (v16i8 (bitconvert (v4i32
+ (scalar_to_vector (extloadi32i16 addr:$src))))))),
+ (VPMOVSXBQrm addr:$src)>;
+}
+
+let Predicates = [UseSSE41] in {
+ def : Pat<(v8i16 (X86vzext (v16i8 VR128:$src))), (PMOVZXBWrr VR128:$src)>;
+ def : Pat<(v4i32 (X86vzext (v16i8 VR128:$src))), (PMOVZXBDrr VR128:$src)>;
+ def : Pat<(v2i64 (X86vzext (v16i8 VR128:$src))), (PMOVZXBQrr VR128:$src)>;
+
+ def : Pat<(v4i32 (X86vzext (v8i16 VR128:$src))), (PMOVZXWDrr VR128:$src)>;
+ def : Pat<(v2i64 (X86vzext (v8i16 VR128:$src))), (PMOVZXWQrr VR128:$src)>;
+
+ def : Pat<(v2i64 (X86vzext (v4i32 VR128:$src))), (PMOVZXDQrr VR128:$src)>;
+
+ def : Pat<(v8i16 (X86vzext (v16i8 (bitconvert (v2i64 (scalar_to_vector (loadi64 addr:$src))))))),
+ (PMOVZXBWrm addr:$src)>;
+ def : Pat<(v8i16 (X86vzext (v16i8 (bitconvert (v2f64 (scalar_to_vector (loadf64 addr:$src))))))),
+ (PMOVZXBWrm addr:$src)>;
+ def : Pat<(v4i32 (X86vzext (v16i8 (bitconvert (v4i32 (scalar_to_vector (loadi32 addr:$src))))))),
+ (PMOVZXBDrm addr:$src)>;
+ def : Pat<(v2i64 (X86vzext (v16i8 (bitconvert (v4i32 (scalar_to_vector (loadi16_anyext addr:$src))))))),
+ (PMOVZXBQrm addr:$src)>;
+
+ def : Pat<(v4i32 (X86vzext (v8i16 (bitconvert (v2i64 (scalar_to_vector (loadi64 addr:$src))))))),
+ (PMOVZXWDrm addr:$src)>;
+ def : Pat<(v4i32 (X86vzext (v8i16 (bitconvert (v2f64 (scalar_to_vector (loadf64 addr:$src))))))),
+ (PMOVZXWDrm addr:$src)>;
+ def : Pat<(v2i64 (X86vzext (v8i16 (bitconvert (v4i32 (scalar_to_vector (loadi32 addr:$src))))))),
+ (PMOVZXWQrm addr:$src)>;
+
+ def : Pat<(v2i64 (X86vzext (v4i32 (bitconvert (v2i64 (scalar_to_vector (loadi64 addr:$src))))))),
+ (PMOVZXDQrm addr:$src)>;
+ def : Pat<(v2i64 (X86vzext (v4i32 (bitconvert (v2f64 (scalar_to_vector (loadf64 addr:$src))))))),
+ (PMOVZXDQrm addr:$src)>;
+ def : Pat<(v2i64 (X86vzext (v4i32 (bitconvert (v2i64 (X86vzload addr:$src)))))),
+ (PMOVZXDQrm addr:$src)>;
}
//===----------------------------------------------------------------------===//
imm:$src2))),
addr:$dst),
(EXTRACTPSmr addr:$dst, VR128:$src1, imm:$src2)>,
- Requires<[HasSSE41]>;
+ Requires<[UseSSE41]>;
//===----------------------------------------------------------------------===//
// SSE4.1 - Insert Instructions
Intrinsic F64Int, bit Is2Addr = 1> {
let ExeDomain = GenericDomain in {
// Operation, reg.
+ let hasSideEffects = 0 in
def SSr : SS4AIi8<opcss, MRMSrcReg,
(outs FR32:$dst), (ins FR32:$src1, FR32:$src2, i32i8imm:$src3),
!if(Is2Addr,
OpSize;
// Operation, reg.
+ let hasSideEffects = 0 in
def SDr : SS4AIi8<opcsd, MRMSrcReg,
(outs FR64:$dst), (ins FR64:$src1, FR64:$src2, i32i8imm:$src3),
!if(Is2Addr,
defm VROUNDY : sse41_fp_unop_rm<0x08, 0x09, "vround", f256mem, VR256,
memopv8f32, memopv4f64,
int_x86_avx_round_ps_256,
- int_x86_avx_round_pd_256>, VEX;
+ int_x86_avx_round_pd_256>, VEX, VEX_L;
defm VROUND : sse41_fp_binop_rm<0x0A, 0x0B, "vround",
int_x86_sse41_round_ss,
int_x86_sse41_round_sd, 0>, VEX_4V, VEX_LIG;
(VROUNDSSr (f32 (IMPLICIT_DEF)), FR32:$src, (i32 0x3))>;
def : Pat<(f64 (ftrunc FR64:$src)),
(VROUNDSDr (f64 (IMPLICIT_DEF)), FR64:$src, (i32 0x3))>;
+
+ def : Pat<(v4f32 (ffloor VR128:$src)),
+ (VROUNDPSr VR128:$src, (i32 0x1))>;
+ def : Pat<(v4f32 (fnearbyint VR128:$src)),
+ (VROUNDPSr VR128:$src, (i32 0xC))>;
+ def : Pat<(v4f32 (fceil VR128:$src)),
+ (VROUNDPSr VR128:$src, (i32 0x2))>;
+ def : Pat<(v4f32 (frint VR128:$src)),
+ (VROUNDPSr VR128:$src, (i32 0x4))>;
+ def : Pat<(v4f32 (ftrunc VR128:$src)),
+ (VROUNDPSr VR128:$src, (i32 0x3))>;
+
+ def : Pat<(v2f64 (ffloor VR128:$src)),
+ (VROUNDPDr VR128:$src, (i32 0x1))>;
+ def : Pat<(v2f64 (fnearbyint VR128:$src)),
+ (VROUNDPDr VR128:$src, (i32 0xC))>;
+ def : Pat<(v2f64 (fceil VR128:$src)),
+ (VROUNDPDr VR128:$src, (i32 0x2))>;
+ def : Pat<(v2f64 (frint VR128:$src)),
+ (VROUNDPDr VR128:$src, (i32 0x4))>;
+ def : Pat<(v2f64 (ftrunc VR128:$src)),
+ (VROUNDPDr VR128:$src, (i32 0x3))>;
+
+ def : Pat<(v8f32 (ffloor VR256:$src)),
+ (VROUNDYPSr VR256:$src, (i32 0x1))>;
+ def : Pat<(v8f32 (fnearbyint VR256:$src)),
+ (VROUNDYPSr VR256:$src, (i32 0xC))>;
+ def : Pat<(v8f32 (fceil VR256:$src)),
+ (VROUNDYPSr VR256:$src, (i32 0x2))>;
+ def : Pat<(v8f32 (frint VR256:$src)),
+ (VROUNDYPSr VR256:$src, (i32 0x4))>;
+ def : Pat<(v8f32 (ftrunc VR256:$src)),
+ (VROUNDYPSr VR256:$src, (i32 0x3))>;
+
+ def : Pat<(v4f64 (ffloor VR256:$src)),
+ (VROUNDYPDr VR256:$src, (i32 0x1))>;
+ def : Pat<(v4f64 (fnearbyint VR256:$src)),
+ (VROUNDYPDr VR256:$src, (i32 0xC))>;
+ def : Pat<(v4f64 (fceil VR256:$src)),
+ (VROUNDYPDr VR256:$src, (i32 0x2))>;
+ def : Pat<(v4f64 (frint VR256:$src)),
+ (VROUNDYPDr VR256:$src, (i32 0x4))>;
+ def : Pat<(v4f64 (ftrunc VR256:$src)),
+ (VROUNDYPDr VR256:$src, (i32 0x3))>;
}
defm ROUND : sse41_fp_unop_rm<0x08, 0x09, "round", f128mem, VR128,
defm ROUND : sse41_fp_binop_rm<0x0A, 0x0B, "round",
int_x86_sse41_round_ss, int_x86_sse41_round_sd>;
-def : Pat<(ffloor FR32:$src),
- (ROUNDSSr (f32 (IMPLICIT_DEF)), FR32:$src, (i32 0x1))>;
-def : Pat<(f64 (ffloor FR64:$src)),
- (ROUNDSDr (f64 (IMPLICIT_DEF)), FR64:$src, (i32 0x1))>;
-def : Pat<(f32 (fnearbyint FR32:$src)),
- (ROUNDSSr (f32 (IMPLICIT_DEF)), FR32:$src, (i32 0xC))>;
-def : Pat<(f64 (fnearbyint FR64:$src)),
- (ROUNDSDr (f64 (IMPLICIT_DEF)), FR64:$src, (i32 0xC))>;
-def : Pat<(f32 (fceil FR32:$src)),
- (ROUNDSSr (f32 (IMPLICIT_DEF)), FR32:$src, (i32 0x2))>;
-def : Pat<(f64 (fceil FR64:$src)),
- (ROUNDSDr (f64 (IMPLICIT_DEF)), FR64:$src, (i32 0x2))>;
-def : Pat<(f32 (frint FR32:$src)),
- (ROUNDSSr (f32 (IMPLICIT_DEF)), FR32:$src, (i32 0x4))>;
-def : Pat<(f64 (frint FR64:$src)),
- (ROUNDSDr (f64 (IMPLICIT_DEF)), FR64:$src, (i32 0x4))>;
-def : Pat<(f32 (ftrunc FR32:$src)),
- (ROUNDSSr (f32 (IMPLICIT_DEF)), FR32:$src, (i32 0x3))>;
-def : Pat<(f64 (ftrunc FR64:$src)),
- (ROUNDSDr (f64 (IMPLICIT_DEF)), FR64:$src, (i32 0x3))>;
+let Predicates = [UseSSE41] in {
+ def : Pat<(ffloor FR32:$src),
+ (ROUNDSSr (f32 (IMPLICIT_DEF)), FR32:$src, (i32 0x1))>;
+ def : Pat<(f64 (ffloor FR64:$src)),
+ (ROUNDSDr (f64 (IMPLICIT_DEF)), FR64:$src, (i32 0x1))>;
+ def : Pat<(f32 (fnearbyint FR32:$src)),
+ (ROUNDSSr (f32 (IMPLICIT_DEF)), FR32:$src, (i32 0xC))>;
+ def : Pat<(f64 (fnearbyint FR64:$src)),
+ (ROUNDSDr (f64 (IMPLICIT_DEF)), FR64:$src, (i32 0xC))>;
+ def : Pat<(f32 (fceil FR32:$src)),
+ (ROUNDSSr (f32 (IMPLICIT_DEF)), FR32:$src, (i32 0x2))>;
+ def : Pat<(f64 (fceil FR64:$src)),
+ (ROUNDSDr (f64 (IMPLICIT_DEF)), FR64:$src, (i32 0x2))>;
+ def : Pat<(f32 (frint FR32:$src)),
+ (ROUNDSSr (f32 (IMPLICIT_DEF)), FR32:$src, (i32 0x4))>;
+ def : Pat<(f64 (frint FR64:$src)),
+ (ROUNDSDr (f64 (IMPLICIT_DEF)), FR64:$src, (i32 0x4))>;
+ def : Pat<(f32 (ftrunc FR32:$src)),
+ (ROUNDSSr (f32 (IMPLICIT_DEF)), FR32:$src, (i32 0x3))>;
+ def : Pat<(f64 (ftrunc FR64:$src)),
+ (ROUNDSDr (f64 (IMPLICIT_DEF)), FR64:$src, (i32 0x3))>;
+
+ def : Pat<(v4f32 (ffloor VR128:$src)),
+ (ROUNDPSr VR128:$src, (i32 0x1))>;
+ def : Pat<(v4f32 (fnearbyint VR128:$src)),
+ (ROUNDPSr VR128:$src, (i32 0xC))>;
+ def : Pat<(v4f32 (fceil VR128:$src)),
+ (ROUNDPSr VR128:$src, (i32 0x2))>;
+ def : Pat<(v4f32 (frint VR128:$src)),
+ (ROUNDPSr VR128:$src, (i32 0x4))>;
+ def : Pat<(v4f32 (ftrunc VR128:$src)),
+ (ROUNDPSr VR128:$src, (i32 0x3))>;
+
+ def : Pat<(v2f64 (ffloor VR128:$src)),
+ (ROUNDPDr VR128:$src, (i32 0x1))>;
+ def : Pat<(v2f64 (fnearbyint VR128:$src)),
+ (ROUNDPDr VR128:$src, (i32 0xC))>;
+ def : Pat<(v2f64 (fceil VR128:$src)),
+ (ROUNDPDr VR128:$src, (i32 0x2))>;
+ def : Pat<(v2f64 (frint VR128:$src)),
+ (ROUNDPDr VR128:$src, (i32 0x4))>;
+ def : Pat<(v2f64 (ftrunc VR128:$src)),
+ (ROUNDPDr VR128:$src, (i32 0x3))>;
+}
//===----------------------------------------------------------------------===//
// SSE4.1 - Packed Bit Test
def VPTESTYrr : SS48I<0x17, MRMSrcReg, (outs), (ins VR256:$src1, VR256:$src2),
"vptest\t{$src2, $src1|$src1, $src2}",
[(set EFLAGS, (X86ptest VR256:$src1, (v4i64 VR256:$src2)))]>,
- OpSize, VEX;
+ OpSize, VEX, VEX_L;
def VPTESTYrm : SS48I<0x17, MRMSrcMem, (outs), (ins VR256:$src1, i256mem:$src2),
"vptest\t{$src2, $src1|$src1, $src2}",
[(set EFLAGS,(X86ptest VR256:$src1, (memopv4i64 addr:$src2)))]>,
- OpSize, VEX;
+ OpSize, VEX, VEX_L;
}
let Defs = [EFLAGS] in {
let Defs = [EFLAGS], Predicates = [HasAVX] in {
let ExeDomain = SSEPackedSingle in {
defm VTESTPS : avx_bittest<0x0E, "vtestps", VR128, f128mem, memopv4f32, v4f32>;
-defm VTESTPSY : avx_bittest<0x0E, "vtestps", VR256, f256mem, memopv8f32, v8f32>;
+defm VTESTPSY : avx_bittest<0x0E, "vtestps", VR256, f256mem, memopv8f32, v8f32>,
+ VEX_L;
}
let ExeDomain = SSEPackedDouble in {
defm VTESTPD : avx_bittest<0x0F, "vtestpd", VR128, f128mem, memopv2f64, v2f64>;
-defm VTESTPDY : avx_bittest<0x0F, "vtestpd", VR256, f256mem, memopv4f64, v4f64>;
+defm VTESTPDY : avx_bittest<0x0F, "vtestpd", VR256, f256mem, memopv4f64, v4f64>,
+ VEX_L;
}
}
(bitconvert (memopv2i64 addr:$src2))))]>, OpSize;
}
-/// SS41I_binop_rm_int - Simple SSE 4.1 binary operator
+/// SS41I_binop_rm_int_y - Simple SSE 4.1 binary operator
multiclass SS41I_binop_rm_int_y<bits<8> opc, string OpcodeStr,
Intrinsic IntId256> {
let isCommutable = 1 in
(bitconvert (memopv4i64 addr:$src2))))]>, OpSize;
}
-let Predicates = [HasAVX] in {
- let isCommutable = 0 in
- defm VPACKUSDW : SS41I_binop_rm_int<0x2B, "vpackusdw", int_x86_sse41_packusdw,
- 0>, VEX_4V;
- defm VPMINSB : SS41I_binop_rm_int<0x38, "vpminsb", int_x86_sse41_pminsb,
- 0>, VEX_4V;
- defm VPMINSD : SS41I_binop_rm_int<0x39, "vpminsd", int_x86_sse41_pminsd,
- 0>, VEX_4V;
- defm VPMINUD : SS41I_binop_rm_int<0x3B, "vpminud", int_x86_sse41_pminud,
- 0>, VEX_4V;
- defm VPMINUW : SS41I_binop_rm_int<0x3A, "vpminuw", int_x86_sse41_pminuw,
- 0>, VEX_4V;
- defm VPMAXSB : SS41I_binop_rm_int<0x3C, "vpmaxsb", int_x86_sse41_pmaxsb,
- 0>, VEX_4V;
- defm VPMAXSD : SS41I_binop_rm_int<0x3D, "vpmaxsd", int_x86_sse41_pmaxsd,
- 0>, VEX_4V;
- defm VPMAXUD : SS41I_binop_rm_int<0x3F, "vpmaxud", int_x86_sse41_pmaxud,
- 0>, VEX_4V;
- defm VPMAXUW : SS41I_binop_rm_int<0x3E, "vpmaxuw", int_x86_sse41_pmaxuw,
- 0>, VEX_4V;
- defm VPMULDQ : SS41I_binop_rm_int<0x28, "vpmuldq", int_x86_sse41_pmuldq,
- 0>, VEX_4V;
-}
-
-let Predicates = [HasAVX2] in {
- let isCommutable = 0 in
- defm VPACKUSDW : SS41I_binop_rm_int_y<0x2B, "vpackusdw",
- int_x86_avx2_packusdw>, VEX_4V;
- defm VPMINSB : SS41I_binop_rm_int_y<0x38, "vpminsb",
- int_x86_avx2_pmins_b>, VEX_4V;
- defm VPMINSD : SS41I_binop_rm_int_y<0x39, "vpminsd",
- int_x86_avx2_pmins_d>, VEX_4V;
- defm VPMINUD : SS41I_binop_rm_int_y<0x3B, "vpminud",
- int_x86_avx2_pminu_d>, VEX_4V;
- defm VPMINUW : SS41I_binop_rm_int_y<0x3A, "vpminuw",
- int_x86_avx2_pminu_w>, VEX_4V;
- defm VPMAXSB : SS41I_binop_rm_int_y<0x3C, "vpmaxsb",
- int_x86_avx2_pmaxs_b>, VEX_4V;
- defm VPMAXSD : SS41I_binop_rm_int_y<0x3D, "vpmaxsd",
- int_x86_avx2_pmaxs_d>, VEX_4V;
- defm VPMAXUD : SS41I_binop_rm_int_y<0x3F, "vpmaxud",
- int_x86_avx2_pmaxu_d>, VEX_4V;
- defm VPMAXUW : SS41I_binop_rm_int_y<0x3E, "vpmaxuw",
- int_x86_avx2_pmaxu_w>, VEX_4V;
- defm VPMULDQ : SS41I_binop_rm_int_y<0x28, "vpmuldq",
- int_x86_avx2_pmul_dq>, VEX_4V;
-}
-
-let Constraints = "$src1 = $dst" in {
- let isCommutable = 0 in
- defm PACKUSDW : SS41I_binop_rm_int<0x2B, "packusdw", int_x86_sse41_packusdw>;
- defm PMINSB : SS41I_binop_rm_int<0x38, "pminsb", int_x86_sse41_pminsb>;
- defm PMINSD : SS41I_binop_rm_int<0x39, "pminsd", int_x86_sse41_pminsd>;
- defm PMINUD : SS41I_binop_rm_int<0x3B, "pminud", int_x86_sse41_pminud>;
- defm PMINUW : SS41I_binop_rm_int<0x3A, "pminuw", int_x86_sse41_pminuw>;
- defm PMAXSB : SS41I_binop_rm_int<0x3C, "pmaxsb", int_x86_sse41_pmaxsb>;
- defm PMAXSD : SS41I_binop_rm_int<0x3D, "pmaxsd", int_x86_sse41_pmaxsd>;
- defm PMAXUD : SS41I_binop_rm_int<0x3F, "pmaxud", int_x86_sse41_pmaxud>;
- defm PMAXUW : SS41I_binop_rm_int<0x3E, "pmaxuw", int_x86_sse41_pmaxuw>;
- defm PMULDQ : SS41I_binop_rm_int<0x28, "pmuldq", int_x86_sse41_pmuldq>;
-}
/// SS48I_binop_rm - Simple SSE41 binary operator.
multiclass SS48I_binop_rm<bits<8> opc, string OpcodeStr, SDNode OpNode,
(bitconvert (memop_frag addr:$src2)))))]>, OpSize;
}
+let Predicates = [HasAVX] in {
+ let isCommutable = 0 in
+ defm VPACKUSDW : SS41I_binop_rm_int<0x2B, "vpackusdw", int_x86_sse41_packusdw,
+ 0>, VEX_4V;
+ defm VPMINSB : SS48I_binop_rm<0x38, "vpminsb", X86smin, v16i8, VR128,
+ memopv2i64, i128mem, 0>, VEX_4V;
+ defm VPMINSD : SS48I_binop_rm<0x39, "vpminsd", X86smin, v4i32, VR128,
+ memopv2i64, i128mem, 0>, VEX_4V;
+ defm VPMINUD : SS48I_binop_rm<0x3B, "vpminud", X86umin, v4i32, VR128,
+ memopv2i64, i128mem, 0>, VEX_4V;
+ defm VPMINUW : SS48I_binop_rm<0x3A, "vpminuw", X86umin, v8i16, VR128,
+ memopv2i64, i128mem, 0>, VEX_4V;
+ defm VPMAXSB : SS48I_binop_rm<0x3C, "vpmaxsb", X86smax, v16i8, VR128,
+ memopv2i64, i128mem, 0>, VEX_4V;
+ defm VPMAXSD : SS48I_binop_rm<0x3D, "vpmaxsd", X86smax, v4i32, VR128,
+ memopv2i64, i128mem, 0>, VEX_4V;
+ defm VPMAXUD : SS48I_binop_rm<0x3F, "vpmaxud", X86umax, v4i32, VR128,
+ memopv2i64, i128mem, 0>, VEX_4V;
+ defm VPMAXUW : SS48I_binop_rm<0x3E, "vpmaxuw", X86umax, v8i16, VR128,
+ memopv2i64, i128mem, 0>, VEX_4V;
+ defm VPMULDQ : SS41I_binop_rm_int<0x28, "vpmuldq", int_x86_sse41_pmuldq,
+ 0>, VEX_4V;
+}
+
+let Predicates = [HasAVX2] in {
+ let isCommutable = 0 in
+ defm VPACKUSDW : SS41I_binop_rm_int_y<0x2B, "vpackusdw",
+ int_x86_avx2_packusdw>, VEX_4V, VEX_L;
+ defm VPMINSBY : SS48I_binop_rm<0x38, "vpminsb", X86smin, v32i8, VR256,
+ memopv4i64, i256mem, 0>, VEX_4V, VEX_L;
+ defm VPMINSDY : SS48I_binop_rm<0x39, "vpminsd", X86smin, v8i32, VR256,
+ memopv4i64, i256mem, 0>, VEX_4V, VEX_L;
+ defm VPMINUDY : SS48I_binop_rm<0x3B, "vpminud", X86umin, v8i32, VR256,
+ memopv4i64, i256mem, 0>, VEX_4V, VEX_L;
+ defm VPMINUWY : SS48I_binop_rm<0x3A, "vpminuw", X86umin, v16i16, VR256,
+ memopv4i64, i256mem, 0>, VEX_4V, VEX_L;
+ defm VPMAXSBY : SS48I_binop_rm<0x3C, "vpmaxsb", X86smax, v32i8, VR256,
+ memopv4i64, i256mem, 0>, VEX_4V, VEX_L;
+ defm VPMAXSDY : SS48I_binop_rm<0x3D, "vpmaxsd", X86smax, v8i32, VR256,
+ memopv4i64, i256mem, 0>, VEX_4V, VEX_L;
+ defm VPMAXUDY : SS48I_binop_rm<0x3F, "vpmaxud", X86umax, v8i32, VR256,
+ memopv4i64, i256mem, 0>, VEX_4V, VEX_L;
+ defm VPMAXUWY : SS48I_binop_rm<0x3E, "vpmaxuw", X86umax, v16i16, VR256,
+ memopv4i64, i256mem, 0>, VEX_4V, VEX_L;
+ defm VPMULDQ : SS41I_binop_rm_int_y<0x28, "vpmuldq",
+ int_x86_avx2_pmul_dq>, VEX_4V, VEX_L;
+}
+
+let Constraints = "$src1 = $dst" in {
+ let isCommutable = 0 in
+ defm PACKUSDW : SS41I_binop_rm_int<0x2B, "packusdw", int_x86_sse41_packusdw>;
+ defm PMINSB : SS48I_binop_rm<0x38, "pminsb", X86smin, v16i8, VR128,
+ memopv2i64, i128mem>;
+ defm PMINSD : SS48I_binop_rm<0x39, "pminsd", X86smin, v4i32, VR128,
+ memopv2i64, i128mem>;
+ defm PMINUD : SS48I_binop_rm<0x3B, "pminud", X86umin, v4i32, VR128,
+ memopv2i64, i128mem>;
+ defm PMINUW : SS48I_binop_rm<0x3A, "pminuw", X86umin, v8i16, VR128,
+ memopv2i64, i128mem>;
+ defm PMAXSB : SS48I_binop_rm<0x3C, "pmaxsb", X86smax, v16i8, VR128,
+ memopv2i64, i128mem>;
+ defm PMAXSD : SS48I_binop_rm<0x3D, "pmaxsd", X86smax, v4i32, VR128,
+ memopv2i64, i128mem>;
+ defm PMAXUD : SS48I_binop_rm<0x3F, "pmaxud", X86umax, v4i32, VR128,
+ memopv2i64, i128mem>;
+ defm PMAXUW : SS48I_binop_rm<0x3E, "pmaxuw", X86umax, v8i16, VR128,
+ memopv2i64, i128mem>;
+ defm PMULDQ : SS41I_binop_rm_int<0x28, "pmuldq", int_x86_sse41_pmuldq>;
+}
+
let Predicates = [HasAVX] in {
defm VPMULLD : SS48I_binop_rm<0x40, "vpmulld", mul, v4i32, VR128,
memopv2i64, i128mem, 0>, VEX_4V;
}
let Predicates = [HasAVX2] in {
defm VPMULLDY : SS48I_binop_rm<0x40, "vpmulld", mul, v8i32, VR256,
- memopv4i64, i256mem, 0>, VEX_4V;
+ memopv4i64, i256mem, 0>, VEX_4V, VEX_L;
defm VPCMPEQQY : SS48I_binop_rm<0x29, "vpcmpeqq", X86pcmpeq, v4i64, VR256,
- memopv4i64, i256mem, 0>, VEX_4V;
+ memopv4i64, i256mem, 0>, VEX_4V, VEX_L;
}
let Constraints = "$src1 = $dst" in {
defm VBLENDPS : SS41I_binop_rmi_int<0x0C, "vblendps", int_x86_sse41_blendps,
VR128, memopv4f32, f128mem, 0>, VEX_4V;
defm VBLENDPSY : SS41I_binop_rmi_int<0x0C, "vblendps",
- int_x86_avx_blend_ps_256, VR256, memopv8f32, f256mem, 0>, VEX_4V;
+ int_x86_avx_blend_ps_256, VR256, memopv8f32,
+ f256mem, 0>, VEX_4V, VEX_L;
}
let ExeDomain = SSEPackedDouble in {
defm VBLENDPD : SS41I_binop_rmi_int<0x0D, "vblendpd", int_x86_sse41_blendpd,
VR128, memopv2f64, f128mem, 0>, VEX_4V;
defm VBLENDPDY : SS41I_binop_rmi_int<0x0D, "vblendpd",
- int_x86_avx_blend_pd_256, VR256, memopv4f64, f256mem, 0>, VEX_4V;
+ int_x86_avx_blend_pd_256,VR256, memopv4f64,
+ f256mem, 0>, VEX_4V, VEX_L;
}
defm VPBLENDW : SS41I_binop_rmi_int<0x0E, "vpblendw", int_x86_sse41_pblendw,
VR128, memopv2i64, i128mem, 0>, VEX_4V;
VR128, 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;
+ VR256, memopv8f32, i256mem, 0>, VEX_4V, VEX_L;
}
let Predicates = [HasAVX2] in {
let isCommutable = 0 in {
defm VPBLENDWY : SS41I_binop_rmi_int<0x0E, "vpblendw", int_x86_avx2_pblendw,
- VR256, memopv4i64, i256mem, 0>, VEX_4V;
+ VR256, memopv4i64, i256mem, 0>, VEX_4V, VEX_L;
defm VMPSADBWY : SS41I_binop_rmi_int<0x42, "vmpsadbw", int_x86_avx2_mpsadbw,
- VR256, memopv4i64, i256mem, 0>, VEX_4V;
+ VR256, memopv4i64, i256mem, 0>, VEX_4V, VEX_L;
}
}
defm VBLENDVPD : SS41I_quaternary_int_avx<0x4B, "vblendvpd", VR128, f128mem,
memopv2f64, int_x86_sse41_blendvpd>;
defm VBLENDVPDY : SS41I_quaternary_int_avx<0x4B, "vblendvpd", VR256, f256mem,
- memopv4f64, int_x86_avx_blendv_pd_256>;
+ memopv4f64, int_x86_avx_blendv_pd_256>, VEX_L;
} // ExeDomain = SSEPackedDouble
let ExeDomain = SSEPackedSingle in {
defm VBLENDVPS : SS41I_quaternary_int_avx<0x4A, "vblendvps", VR128, f128mem,
memopv4f32, int_x86_sse41_blendvps>;
defm VBLENDVPSY : SS41I_quaternary_int_avx<0x4A, "vblendvps", VR256, f256mem,
- memopv8f32, int_x86_avx_blendv_ps_256>;
+ memopv8f32, int_x86_avx_blendv_ps_256>, VEX_L;
} // ExeDomain = SSEPackedSingle
defm VPBLENDVB : SS41I_quaternary_int_avx<0x4C, "vpblendvb", VR128, i128mem,
memopv2i64, int_x86_sse41_pblendvb>;
let Predicates = [HasAVX2] in {
defm VPBLENDVBY : SS41I_quaternary_int_avx<0x4C, "vpblendvb", VR256, i256mem,
- memopv4i64, int_x86_avx2_pblendvb>;
+ memopv4i64, int_x86_avx2_pblendvb>, VEX_L;
}
let Predicates = [HasAVX] in {
(v4f64 VR256:$src2))),
(VBLENDVPDYrr VR256:$src2, VR256:$src1, VR256:$mask)>;
- def : Pat<(v8f32 (X86Blendps (v8f32 VR256:$src1), (v8f32 VR256:$src2),
+ def : Pat<(v8f32 (X86Blendi (v8f32 VR256:$src1), (v8f32 VR256:$src2),
(imm:$mask))),
- (VBLENDPSYrri VR256:$src2, VR256:$src1, imm:$mask)>;
- def : Pat<(v4f64 (X86Blendpd (v4f64 VR256:$src1), (v4f64 VR256:$src2),
+ (VBLENDPSYrri VR256:$src1, VR256:$src2, imm:$mask)>;
+ def : Pat<(v4f64 (X86Blendi (v4f64 VR256:$src1), (v4f64 VR256:$src2),
(imm:$mask))),
- (VBLENDPDYrri VR256:$src2, VR256:$src1, imm:$mask)>;
+ (VBLENDPDYrri VR256:$src1, VR256:$src2, imm:$mask)>;
- def : Pat<(v8i16 (X86Blendpw (v8i16 VR128:$src1), (v8i16 VR128:$src2),
+ def : Pat<(v8i16 (X86Blendi (v8i16 VR128:$src1), (v8i16 VR128:$src2),
(imm:$mask))),
- (VPBLENDWrri VR128:$src2, VR128:$src1, imm:$mask)>;
- def : Pat<(v4f32 (X86Blendps (v4f32 VR128:$src1), (v4f32 VR128:$src2),
+ (VPBLENDWrri VR128:$src1, VR128:$src2, imm:$mask)>;
+ def : Pat<(v4f32 (X86Blendi (v4f32 VR128:$src1), (v4f32 VR128:$src2),
(imm:$mask))),
- (VBLENDPSrri VR128:$src2, VR128:$src1, imm:$mask)>;
- def : Pat<(v2f64 (X86Blendpd (v2f64 VR128:$src1), (v2f64 VR128:$src2),
+ (VBLENDPSrri VR128:$src1, VR128:$src2, imm:$mask)>;
+ def : Pat<(v2f64 (X86Blendi (v2f64 VR128:$src1), (v2f64 VR128:$src2),
(imm:$mask))),
- (VBLENDPDrri VR128:$src2, VR128:$src1, imm:$mask)>;
+ (VBLENDPDrri VR128:$src1, VR128:$src2, imm:$mask)>;
}
let Predicates = [HasAVX2] in {
def : Pat<(v32i8 (vselect (v32i8 VR256:$mask), (v32i8 VR256:$src1),
(v32i8 VR256:$src2))),
- (VPBLENDVBYrr VR256:$src2, VR256:$src1, VR256:$mask)>;
- def : Pat<(v16i16 (X86Blendpw (v16i16 VR256:$src1), (v16i16 VR256:$src2),
+ (VPBLENDVBYrr VR256:$src1, VR256:$src2, VR256:$mask)>;
+ def : Pat<(v16i16 (X86Blendi (v16i16 VR256:$src1), (v16i16 VR256:$src2),
(imm:$mask))),
- (VPBLENDWYrri VR256:$src2, VR256:$src1, imm:$mask)>;
+ (VPBLENDWYrri VR256:$src1, VR256:$src2, imm:$mask)>;
}
/// SS41I_ternary_int - SSE 4.1 ternary operator
def : InstAlias<"pblendvb\t{%xmm0, $src2, $dst|$dst, $src2, %xmm0}",
(PBLENDVBrm0 VR128:$dst, i128mem:$src2)>;
-let Predicates = [HasSSE41] in {
+let Predicates = [UseSSE41] in {
def : Pat<(v16i8 (vselect (v16i8 XMM0), (v16i8 VR128:$src1),
(v16i8 VR128:$src2))),
(PBLENDVBrr0 VR128:$src2, VR128:$src1)>;
(v2f64 VR128:$src2))),
(BLENDVPDrr0 VR128:$src2, VR128:$src1)>;
- def : Pat<(v8i16 (X86Blendpw (v8i16 VR128:$src1), (v8i16 VR128:$src2),
+ def : Pat<(v8i16 (X86Blendi (v8i16 VR128:$src1), (v8i16 VR128:$src2),
(imm:$mask))),
- (PBLENDWrri VR128:$src2, VR128:$src1, imm:$mask)>;
- def : Pat<(v4f32 (X86Blendps (v4f32 VR128:$src1), (v4f32 VR128:$src2),
+ (PBLENDWrri VR128:$src1, VR128:$src2, imm:$mask)>;
+ def : Pat<(v4f32 (X86Blendi (v4f32 VR128:$src1), (v4f32 VR128:$src2),
(imm:$mask))),
- (BLENDPSrri VR128:$src2, VR128:$src1, imm:$mask)>;
- def : Pat<(v2f64 (X86Blendpd (v2f64 VR128:$src1), (v2f64 VR128:$src2),
+ (BLENDPSrri VR128:$src1, VR128:$src2, imm:$mask)>;
+ def : Pat<(v2f64 (X86Blendi (v2f64 VR128:$src1), (v2f64 VR128:$src2),
(imm:$mask))),
- (BLENDPDrri VR128:$src2, VR128:$src1, imm:$mask)>;
+ (BLENDPDrri VR128:$src1, VR128:$src2, imm:$mask)>;
}
def VMOVNTDQAYrm : SS48I<0x2A, MRMSrcMem, (outs VR256:$dst), (ins i256mem:$src),
"vmovntdqa\t{$src, $dst|$dst, $src}",
[(set VR256:$dst, (int_x86_avx2_movntdqa addr:$src))]>,
- OpSize, VEX;
+ OpSize, VEX, VEX_L;
def MOVNTDQArm : SS48I<0x2A, MRMSrcMem, (outs VR128:$dst), (ins i128mem:$src),
"movntdqa\t{$src, $dst|$dst, $src}",
[(set VR128:$dst, (int_x86_sse41_movntdqa addr:$src))]>,
let Predicates = [HasAVX2] in
defm VPCMPGTQY : SS42I_binop_rm<0x37, "vpcmpgtq", X86pcmpgt, v4i64, VR256,
- memopv4i64, i256mem, 0>, VEX_4V;
+ memopv4i64, i256mem, 0>, VEX_4V, VEX_L;
let Constraints = "$src1 = $dst" in
defm PCMPGTQ : SS42I_binop_rm<0x37, "pcmpgtq", X86pcmpgt, v2i64, VR128,
imm:$src3))]>;
def MEM : PseudoI<(outs VR128:$dst),
(ins VR128:$src1, i128mem:$src2, i8imm:$src3),
- [(set VR128:$dst, (int_x86_sse42_pcmpistrm128
- VR128:$src1, (load addr:$src2), imm:$src3))]>;
+ [(set VR128:$dst, (int_x86_sse42_pcmpistrm128 VR128:$src1,
+ (bc_v16i8 (memopv2i64 addr:$src2)), imm:$src3))]>;
}
let Defs = [EFLAGS], usesCustomInserter = 1 in {
- let AddedComplexity = 1 in
- defm VPCMPISTRM128 : pseudo_pcmpistrm<"#VPCMPISTRM128">, Requires<[HasAVX]>;
- defm PCMPISTRM128 : pseudo_pcmpistrm<"#PCMPISTRM128">, Requires<[HasSSE42]>;
+ defm VPCMPISTRM128 : pseudo_pcmpistrm<"#VPCMPISTRM128">, Requires<[HasAVX]>;
+ defm PCMPISTRM128 : pseudo_pcmpistrm<"#PCMPISTRM128">, Requires<[UseSSE42]>;
}
-let Defs = [XMM0, EFLAGS], neverHasSideEffects = 1, Predicates = [HasAVX] in {
- def VPCMPISTRM128rr : SS42AI<0x62, MRMSrcReg, (outs),
- (ins VR128:$src1, VR128:$src2, i8imm:$src3),
- "vpcmpistrm\t{$src3, $src2, $src1|$src1, $src2, $src3}", []>, OpSize, VEX;
+multiclass pcmpistrm_SS42AI<string asm> {
+ def rr : SS42AI<0x62, MRMSrcReg, (outs),
+ (ins VR128:$src1, VR128:$src2, i8imm:$src3),
+ !strconcat(asm, "\t{$src3, $src2, $src1|$src1, $src2, $src3}"),
+ []>, OpSize;
let mayLoad = 1 in
- def VPCMPISTRM128rm : SS42AI<0x62, MRMSrcMem, (outs),
- (ins VR128:$src1, i128mem:$src2, i8imm:$src3),
- "vpcmpistrm\t{$src3, $src2, $src1|$src1, $src2, $src3}", []>, OpSize, VEX;
+ def rm :SS42AI<0x62, MRMSrcMem, (outs),
+ (ins VR128:$src1, i128mem:$src2, i8imm:$src3),
+ !strconcat(asm, "\t{$src3, $src2, $src1|$src1, $src2, $src3}"),
+ []>, OpSize;
}
let Defs = [XMM0, EFLAGS], neverHasSideEffects = 1 in {
- def PCMPISTRM128rr : SS42AI<0x62, MRMSrcReg, (outs),
- (ins VR128:$src1, VR128:$src2, i8imm:$src3),
- "pcmpistrm\t{$src3, $src2, $src1|$src1, $src2, $src3}", []>, OpSize;
- let mayLoad = 1 in
- def PCMPISTRM128rm : SS42AI<0x62, MRMSrcMem, (outs),
- (ins VR128:$src1, i128mem:$src2, i8imm:$src3),
- "pcmpistrm\t{$src3, $src2, $src1|$src1, $src2, $src3}", []>, OpSize;
+ let Predicates = [HasAVX] in
+ defm VPCMPISTRM128 : pcmpistrm_SS42AI<"vpcmpistrm">, VEX;
+ defm PCMPISTRM128 : pcmpistrm_SS42AI<"pcmpistrm"> ;
}
// Packed Compare Explicit Length Strings, Return Mask
VR128:$src1, EAX, VR128:$src3, EDX, imm:$src5))]>;
def MEM : PseudoI<(outs VR128:$dst),
(ins VR128:$src1, i128mem:$src3, i8imm:$src5),
- [(set VR128:$dst, (int_x86_sse42_pcmpestrm128
- VR128:$src1, EAX, (load addr:$src3), EDX, imm:$src5))]>;
+ [(set VR128:$dst, (int_x86_sse42_pcmpestrm128 VR128:$src1, EAX,
+ (bc_v16i8 (memopv2i64 addr:$src3)), EDX, imm:$src5))]>;
}
let Defs = [EFLAGS], Uses = [EAX, EDX], usesCustomInserter = 1 in {
- let AddedComplexity = 1 in
- defm VPCMPESTRM128 : pseudo_pcmpestrm<"#VPCMPESTRM128">, Requires<[HasAVX]>;
- defm PCMPESTRM128 : pseudo_pcmpestrm<"#PCMPESTRM128">, Requires<[HasSSE42]>;
+ defm VPCMPESTRM128 : pseudo_pcmpestrm<"#VPCMPESTRM128">, Requires<[HasAVX]>;
+ defm PCMPESTRM128 : pseudo_pcmpestrm<"#PCMPESTRM128">, Requires<[UseSSE42]>;
}
-let Predicates = [HasAVX],
- Defs = [XMM0, EFLAGS], Uses = [EAX, EDX], neverHasSideEffects = 1 in {
- def VPCMPESTRM128rr : SS42AI<0x60, MRMSrcReg, (outs),
- (ins VR128:$src1, VR128:$src3, i8imm:$src5),
- "vpcmpestrm\t{$src5, $src3, $src1|$src1, $src3, $src5}", []>, OpSize, VEX;
+multiclass SS42AI_pcmpestrm<string asm> {
+ def rr : SS42AI<0x60, MRMSrcReg, (outs),
+ (ins VR128:$src1, VR128:$src3, i8imm:$src5),
+ !strconcat(asm, "\t{$src5, $src3, $src1|$src1, $src3, $src5}"),
+ []>, OpSize;
let mayLoad = 1 in
- def VPCMPESTRM128rm : SS42AI<0x60, MRMSrcMem, (outs),
- (ins VR128:$src1, i128mem:$src3, i8imm:$src5),
- "vpcmpestrm\t{$src5, $src3, $src1|$src1, $src3, $src5}", []>, OpSize, VEX;
+ def rm : SS42AI<0x60, MRMSrcMem, (outs),
+ (ins VR128:$src1, i128mem:$src3, i8imm:$src5),
+ !strconcat(asm, "\t{$src5, $src3, $src1|$src1, $src3, $src5}"),
+ []>, OpSize;
}
let Defs = [XMM0, EFLAGS], Uses = [EAX, EDX], neverHasSideEffects = 1 in {
- def PCMPESTRM128rr : SS42AI<0x60, MRMSrcReg, (outs),
- (ins VR128:$src1, VR128:$src3, i8imm:$src5),
- "pcmpestrm\t{$src5, $src3, $src1|$src1, $src3, $src5}", []>, OpSize;
- let mayLoad = 1 in
- def PCMPESTRM128rm : SS42AI<0x60, MRMSrcMem, (outs),
- (ins VR128:$src1, i128mem:$src3, i8imm:$src5),
- "pcmpestrm\t{$src5, $src3, $src1|$src1, $src3, $src5}", []>, OpSize;
+ let Predicates = [HasAVX] in
+ defm VPCMPESTRM128 : SS42AI_pcmpestrm<"vpcmpestrm">, VEX;
+ defm PCMPESTRM128 : SS42AI_pcmpestrm<"pcmpestrm">;
}
// Packed Compare Implicit Length Strings, Return Index
-let Defs = [ECX, EFLAGS] in {
- multiclass SS42AI_pcmpistri<Intrinsic IntId128, string asm = "pcmpistri"> {
- def rr : SS42AI<0x63, MRMSrcReg, (outs),
- (ins VR128:$src1, VR128:$src2, i8imm:$src3),
- !strconcat(asm, "\t{$src3, $src2, $src1|$src1, $src2, $src3}"),
- [(set ECX, (IntId128 VR128:$src1, VR128:$src2, imm:$src3)),
- (implicit EFLAGS)]>, OpSize;
- def rm : SS42AI<0x63, MRMSrcMem, (outs),
- (ins VR128:$src1, i128mem:$src2, i8imm:$src3),
- !strconcat(asm, "\t{$src3, $src2, $src1|$src1, $src2, $src3}"),
- [(set ECX, (IntId128 VR128:$src1, (load addr:$src2), imm:$src3)),
- (implicit EFLAGS)]>, OpSize;
- }
+multiclass pseudo_pcmpistri<string asm> {
+ def REG : PseudoI<(outs GR32:$dst),
+ (ins VR128:$src1, VR128:$src2, i8imm:$src3),
+ [(set GR32:$dst, EFLAGS,
+ (X86pcmpistri VR128:$src1, VR128:$src2, imm:$src3))]>;
+ def MEM : PseudoI<(outs GR32:$dst),
+ (ins VR128:$src1, i128mem:$src2, i8imm:$src3),
+ [(set GR32:$dst, EFLAGS, (X86pcmpistri VR128:$src1,
+ (bc_v16i8 (memopv2i64 addr:$src2)), imm:$src3))]>;
}
-let Predicates = [HasAVX] in {
-defm VPCMPISTRI : SS42AI_pcmpistri<int_x86_sse42_pcmpistri128, "vpcmpistri">,
- VEX;
-defm VPCMPISTRIA : SS42AI_pcmpistri<int_x86_sse42_pcmpistria128, "vpcmpistri">,
- VEX;
-defm VPCMPISTRIC : SS42AI_pcmpistri<int_x86_sse42_pcmpistric128, "vpcmpistri">,
- VEX;
-defm VPCMPISTRIO : SS42AI_pcmpistri<int_x86_sse42_pcmpistrio128, "vpcmpistri">,
- VEX;
-defm VPCMPISTRIS : SS42AI_pcmpistri<int_x86_sse42_pcmpistris128, "vpcmpistri">,
- VEX;
-defm VPCMPISTRIZ : SS42AI_pcmpistri<int_x86_sse42_pcmpistriz128, "vpcmpistri">,
- VEX;
-}
-
-defm PCMPISTRI : SS42AI_pcmpistri<int_x86_sse42_pcmpistri128>;
-defm PCMPISTRIA : SS42AI_pcmpistri<int_x86_sse42_pcmpistria128>;
-defm PCMPISTRIC : SS42AI_pcmpistri<int_x86_sse42_pcmpistric128>;
-defm PCMPISTRIO : SS42AI_pcmpistri<int_x86_sse42_pcmpistrio128>;
-defm PCMPISTRIS : SS42AI_pcmpistri<int_x86_sse42_pcmpistris128>;
-defm PCMPISTRIZ : SS42AI_pcmpistri<int_x86_sse42_pcmpistriz128>;
+let Defs = [EFLAGS], usesCustomInserter = 1 in {
+ defm VPCMPISTRI : pseudo_pcmpistri<"#VPCMPISTRI">, Requires<[HasAVX]>;
+ defm PCMPISTRI : pseudo_pcmpistri<"#PCMPISTRI">, Requires<[UseSSE42]>;
+}
+
+multiclass SS42AI_pcmpistri<string asm> {
+ def rr : SS42AI<0x63, MRMSrcReg, (outs),
+ (ins VR128:$src1, VR128:$src2, i8imm:$src3),
+ !strconcat(asm, "\t{$src3, $src2, $src1|$src1, $src2, $src3}"),
+ []>, OpSize;
+ let mayLoad = 1 in
+ def rm : SS42AI<0x63, MRMSrcMem, (outs),
+ (ins VR128:$src1, i128mem:$src2, i8imm:$src3),
+ !strconcat(asm, "\t{$src3, $src2, $src1|$src1, $src2, $src3}"),
+ []>, OpSize;
+}
+
+let Defs = [ECX, EFLAGS], neverHasSideEffects = 1 in {
+ let Predicates = [HasAVX] in
+ defm VPCMPISTRI : SS42AI_pcmpistri<"vpcmpistri">, VEX;
+ defm PCMPISTRI : SS42AI_pcmpistri<"pcmpistri">;
+}
// Packed Compare Explicit Length Strings, Return Index
-let Defs = [ECX, EFLAGS], Uses = [EAX, EDX] in {
- multiclass SS42AI_pcmpestri<Intrinsic IntId128, string asm = "pcmpestri"> {
- def rr : SS42AI<0x61, MRMSrcReg, (outs),
- (ins VR128:$src1, VR128:$src3, i8imm:$src5),
- !strconcat(asm, "\t{$src5, $src3, $src1|$src1, $src3, $src5}"),
- [(set ECX, (IntId128 VR128:$src1, EAX, VR128:$src3, EDX, imm:$src5)),
- (implicit EFLAGS)]>, OpSize;
- def rm : SS42AI<0x61, MRMSrcMem, (outs),
- (ins VR128:$src1, i128mem:$src3, i8imm:$src5),
- !strconcat(asm, "\t{$src5, $src3, $src1|$src1, $src3, $src5}"),
- [(set ECX,
- (IntId128 VR128:$src1, EAX, (load addr:$src3), EDX, imm:$src5)),
- (implicit EFLAGS)]>, OpSize;
- }
+multiclass pseudo_pcmpestri<string asm> {
+ def REG : PseudoI<(outs GR32:$dst),
+ (ins VR128:$src1, VR128:$src3, i8imm:$src5),
+ [(set GR32:$dst, EFLAGS,
+ (X86pcmpestri VR128:$src1, EAX, VR128:$src3, EDX, imm:$src5))]>;
+ def MEM : PseudoI<(outs GR32:$dst),
+ (ins VR128:$src1, i128mem:$src3, i8imm:$src5),
+ [(set GR32:$dst, EFLAGS,
+ (X86pcmpestri VR128:$src1, EAX, (bc_v16i8 (memopv2i64 addr:$src3)), EDX,
+ imm:$src5))]>;
}
-let Predicates = [HasAVX] in {
-defm VPCMPESTRI : SS42AI_pcmpestri<int_x86_sse42_pcmpestri128, "vpcmpestri">,
- VEX;
-defm VPCMPESTRIA : SS42AI_pcmpestri<int_x86_sse42_pcmpestria128, "vpcmpestri">,
- VEX;
-defm VPCMPESTRIC : SS42AI_pcmpestri<int_x86_sse42_pcmpestric128, "vpcmpestri">,
- VEX;
-defm VPCMPESTRIO : SS42AI_pcmpestri<int_x86_sse42_pcmpestrio128, "vpcmpestri">,
- VEX;
-defm VPCMPESTRIS : SS42AI_pcmpestri<int_x86_sse42_pcmpestris128, "vpcmpestri">,
- VEX;
-defm VPCMPESTRIZ : SS42AI_pcmpestri<int_x86_sse42_pcmpestriz128, "vpcmpestri">,
- VEX;
-}
-
-defm PCMPESTRI : SS42AI_pcmpestri<int_x86_sse42_pcmpestri128>;
-defm PCMPESTRIA : SS42AI_pcmpestri<int_x86_sse42_pcmpestria128>;
-defm PCMPESTRIC : SS42AI_pcmpestri<int_x86_sse42_pcmpestric128>;
-defm PCMPESTRIO : SS42AI_pcmpestri<int_x86_sse42_pcmpestrio128>;
-defm PCMPESTRIS : SS42AI_pcmpestri<int_x86_sse42_pcmpestris128>;
-defm PCMPESTRIZ : SS42AI_pcmpestri<int_x86_sse42_pcmpestriz128>;
+let Defs = [EFLAGS], Uses = [EAX, EDX], usesCustomInserter = 1 in {
+ defm VPCMPESTRI : pseudo_pcmpestri<"#VPCMPESTRI">, Requires<[HasAVX]>;
+ defm PCMPESTRI : pseudo_pcmpestri<"#PCMPESTRI">, Requires<[UseSSE42]>;
+}
+
+multiclass SS42AI_pcmpestri<string asm> {
+ def rr : SS42AI<0x61, MRMSrcReg, (outs),
+ (ins VR128:$src1, VR128:$src3, i8imm:$src5),
+ !strconcat(asm, "\t{$src5, $src3, $src1|$src1, $src3, $src5}"),
+ []>, OpSize;
+ let mayLoad = 1 in
+ def rm : SS42AI<0x61, MRMSrcMem, (outs),
+ (ins VR128:$src1, i128mem:$src3, i8imm:$src5),
+ !strconcat(asm, "\t{$src5, $src3, $src1|$src1, $src3, $src5}"),
+ []>, OpSize;
+}
+
+let Defs = [ECX, EFLAGS], Uses = [EAX, EDX], neverHasSideEffects = 1 in {
+ let Predicates = [HasAVX] in
+ defm VPCMPESTRI : SS42AI_pcmpestri<"vpcmpestri">, VEX;
+ defm PCMPESTRI : SS42AI_pcmpestri<"pcmpestri">;
+}
//===----------------------------------------------------------------------===//
// SSE4.2 - CRC Instructions
def VBROADCASTSSrm : avx_broadcast<0x18, "vbroadcastss", VR128, f32mem,
int_x86_avx_vbroadcast_ss>;
def VBROADCASTSSYrm : avx_broadcast<0x18, "vbroadcastss", VR256, f32mem,
- int_x86_avx_vbroadcast_ss_256>;
+ int_x86_avx_vbroadcast_ss_256>, VEX_L;
}
let ExeDomain = SSEPackedDouble in
def VBROADCASTSDYrm : avx_broadcast<0x19, "vbroadcastsd", VR256, f64mem,
- int_x86_avx_vbroadcast_sd_256>;
+ int_x86_avx_vbroadcast_sd_256>, VEX_L;
def VBROADCASTF128 : avx_broadcast<0x1A, "vbroadcastf128", VR256, f128mem,
- int_x86_avx_vbroadcastf128_pd_256>;
+ int_x86_avx_vbroadcastf128_pd_256>, VEX_L;
let ExeDomain = SSEPackedSingle in {
def VBROADCASTSSrr : avx2_broadcast_reg<0x18, "vbroadcastss", VR128,
int_x86_avx2_vbroadcast_ss_ps>;
def VBROADCASTSSYrr : avx2_broadcast_reg<0x18, "vbroadcastss", VR256,
- int_x86_avx2_vbroadcast_ss_ps_256>;
+ int_x86_avx2_vbroadcast_ss_ps_256>, VEX_L;
}
let ExeDomain = SSEPackedDouble in
def VBROADCASTSDYrr : avx2_broadcast_reg<0x19, "vbroadcastsd", VR256,
- int_x86_avx2_vbroadcast_sd_pd_256>;
+ int_x86_avx2_vbroadcast_sd_pd_256>, VEX_L;
let Predicates = [HasAVX2] in
def VBROADCASTI128 : avx_broadcast<0x5A, "vbroadcasti128", VR256, i128mem,
- int_x86_avx2_vbroadcasti128>;
+ int_x86_avx2_vbroadcasti128>, VEX_L;
let Predicates = [HasAVX] in
def : Pat<(int_x86_avx_vbroadcastf128_ps_256 addr:$src),
def VINSERTF128rr : AVXAIi8<0x18, MRMSrcReg, (outs VR256:$dst),
(ins VR256:$src1, VR128:$src2, i8imm:$src3),
"vinsertf128\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}",
- []>, VEX_4V;
+ []>, VEX_4V, VEX_L;
let mayLoad = 1 in
def VINSERTF128rm : AVXAIi8<0x18, MRMSrcMem, (outs VR256:$dst),
(ins VR256:$src1, f128mem:$src2, i8imm:$src3),
"vinsertf128\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}",
- []>, VEX_4V;
+ []>, VEX_4V, VEX_L;
}
let Predicates = [HasAVX] in {
def : Pat<(vinsertf128_insert:$ins (v8f32 VR256:$src1), (v4f32 VR128:$src2),
- (i32 imm)),
+ (iPTR imm)),
(VINSERTF128rr VR256:$src1, VR128:$src2,
(INSERT_get_vinsertf128_imm VR256:$ins))>;
def : Pat<(vinsertf128_insert:$ins (v4f64 VR256:$src1), (v2f64 VR128:$src2),
- (i32 imm)),
+ (iPTR imm)),
(VINSERTF128rr VR256:$src1, VR128:$src2,
(INSERT_get_vinsertf128_imm VR256:$ins))>;
+
+def : Pat<(vinsertf128_insert:$ins (v8f32 VR256:$src1), (memopv4f32 addr:$src2),
+ (iPTR imm)),
+ (VINSERTF128rm VR256:$src1, addr:$src2,
+ (INSERT_get_vinsertf128_imm VR256:$ins))>;
+def : Pat<(vinsertf128_insert:$ins (v4f64 VR256:$src1), (memopv2f64 addr:$src2),
+ (iPTR imm)),
+ (VINSERTF128rm VR256:$src1, addr:$src2,
+ (INSERT_get_vinsertf128_imm VR256:$ins))>;
+}
+
+let Predicates = [HasAVX1Only] in {
def : Pat<(vinsertf128_insert:$ins (v4i64 VR256:$src1), (v2i64 VR128:$src2),
- (i32 imm)),
+ (iPTR imm)),
(VINSERTF128rr VR256:$src1, VR128:$src2,
(INSERT_get_vinsertf128_imm VR256:$ins))>;
def : Pat<(vinsertf128_insert:$ins (v8i32 VR256:$src1), (v4i32 VR128:$src2),
- (i32 imm)),
+ (iPTR imm)),
(VINSERTF128rr VR256:$src1, VR128:$src2,
(INSERT_get_vinsertf128_imm VR256:$ins))>;
def : Pat<(vinsertf128_insert:$ins (v32i8 VR256:$src1), (v16i8 VR128:$src2),
- (i32 imm)),
+ (iPTR imm)),
(VINSERTF128rr VR256:$src1, VR128:$src2,
(INSERT_get_vinsertf128_imm VR256:$ins))>;
def : Pat<(vinsertf128_insert:$ins (v16i16 VR256:$src1), (v8i16 VR128:$src2),
- (i32 imm)),
+ (iPTR imm)),
(VINSERTF128rr VR256:$src1, VR128:$src2,
(INSERT_get_vinsertf128_imm VR256:$ins))>;
-def : Pat<(vinsertf128_insert:$ins (v8f32 VR256:$src1), (loadv4f32 addr:$src2),
- (i32 imm)),
+def : Pat<(vinsertf128_insert:$ins (v4i64 VR256:$src1), (memopv2i64 addr:$src2),
+ (iPTR imm)),
+ (VINSERTF128rm VR256:$src1, addr:$src2,
+ (INSERT_get_vinsertf128_imm VR256:$ins))>;
+def : Pat<(vinsertf128_insert:$ins (v8i32 VR256:$src1),
+ (bc_v4i32 (memopv2i64 addr:$src2)),
+ (iPTR imm)),
(VINSERTF128rm VR256:$src1, addr:$src2,
(INSERT_get_vinsertf128_imm VR256:$ins))>;
-def : Pat<(vinsertf128_insert:$ins (v4f64 VR256:$src1), (loadv2f64 addr:$src2),
- (i32 imm)),
+def : Pat<(vinsertf128_insert:$ins (v32i8 VR256:$src1),
+ (bc_v16i8 (memopv2i64 addr:$src2)),
+ (iPTR imm)),
(VINSERTF128rm VR256:$src1, addr:$src2,
(INSERT_get_vinsertf128_imm VR256:$ins))>;
-def : Pat<(vinsertf128_insert:$ins (v4i64 VR256:$src1), (loadv2i64 addr:$src2),
- (i32 imm)),
+def : Pat<(vinsertf128_insert:$ins (v16i16 VR256:$src1),
+ (bc_v8i16 (memopv2i64 addr:$src2)),
+ (iPTR imm)),
(VINSERTF128rm VR256:$src1, addr:$src2,
(INSERT_get_vinsertf128_imm VR256:$ins))>;
}
def VEXTRACTF128rr : AVXAIi8<0x19, MRMDestReg, (outs VR128:$dst),
(ins VR256:$src1, i8imm:$src2),
"vextractf128\t{$src2, $src1, $dst|$dst, $src1, $src2}",
- []>, VEX;
+ []>, VEX, VEX_L;
let mayStore = 1 in
def VEXTRACTF128mr : AVXAIi8<0x19, MRMDestMem, (outs),
(ins f128mem:$dst, VR256:$src1, i8imm:$src2),
"vextractf128\t{$src2, $src1, $dst|$dst, $src1, $src2}",
- []>, VEX;
-}
-
-// Extract and store.
-let Predicates = [HasAVX] in {
- def : Pat<(alignedstore (int_x86_avx_vextractf128_ps_256 VR256:$src1, imm:$src2), addr:$dst),
- (VEXTRACTF128mr addr:$dst, VR256:$src1, imm:$src2)>;
- def : Pat<(alignedstore (int_x86_avx_vextractf128_pd_256 VR256:$src1, imm:$src2), addr:$dst),
- (VEXTRACTF128mr addr:$dst, VR256:$src1, imm:$src2)>;
- def : Pat<(alignedstore (int_x86_avx_vextractf128_si_256 VR256:$src1, imm:$src2), addr:$dst),
- (VEXTRACTF128mr addr:$dst, VR256:$src1, imm:$src2)>;
-
- def : Pat<(int_x86_sse_storeu_ps addr:$dst, (int_x86_avx_vextractf128_ps_256 VR256:$src1, imm:$src2)),
- (VEXTRACTF128mr addr:$dst, VR256:$src1, imm:$src2)>;
- def : Pat<(int_x86_sse2_storeu_pd addr:$dst, (int_x86_avx_vextractf128_pd_256 VR256:$src1, imm:$src2)),
- (VEXTRACTF128mr addr:$dst, VR256:$src1, imm:$src2)>;
- def : Pat<(int_x86_sse2_storeu_dq addr:$dst, (bc_v16i8 (int_x86_avx_vextractf128_si_256 VR256:$src1, imm:$src2))),
- (VEXTRACTF128mr addr:$dst, VR256:$src1, imm:$src2)>;
+ []>, VEX, VEX_L;
}
// AVX1 patterns
let Predicates = [HasAVX] in {
-def : Pat<(int_x86_avx_vextractf128_pd_256 VR256:$src1, imm:$src2),
- (VEXTRACTF128rr VR256:$src1, imm:$src2)>;
-def : Pat<(int_x86_avx_vextractf128_ps_256 VR256:$src1, imm:$src2),
- (VEXTRACTF128rr VR256:$src1, imm:$src2)>;
-def : Pat<(int_x86_avx_vextractf128_si_256 VR256:$src1, imm:$src2),
- (VEXTRACTF128rr VR256:$src1, imm:$src2)>;
-
-def : Pat<(vextractf128_extract:$ext VR256:$src1, (i32 imm)),
+def : Pat<(vextractf128_extract:$ext VR256:$src1, (iPTR imm)),
(v4f32 (VEXTRACTF128rr
(v8f32 VR256:$src1),
(EXTRACT_get_vextractf128_imm VR128:$ext)))>;
-def : Pat<(vextractf128_extract:$ext VR256:$src1, (i32 imm)),
+def : Pat<(vextractf128_extract:$ext VR256:$src1, (iPTR imm)),
(v2f64 (VEXTRACTF128rr
(v4f64 VR256:$src1),
(EXTRACT_get_vextractf128_imm VR128:$ext)))>;
-def : Pat<(vextractf128_extract:$ext VR256:$src1, (i32 imm)),
+
+def : Pat<(alignedstore (v4f32 (vextractf128_extract:$ext (v8f32 VR256:$src1),
+ (iPTR imm))), addr:$dst),
+ (VEXTRACTF128mr addr:$dst, VR256:$src1,
+ (EXTRACT_get_vextractf128_imm VR128:$ext))>;
+def : Pat<(alignedstore (v2f64 (vextractf128_extract:$ext (v4f64 VR256:$src1),
+ (iPTR imm))), addr:$dst),
+ (VEXTRACTF128mr addr:$dst, VR256:$src1,
+ (EXTRACT_get_vextractf128_imm VR128:$ext))>;
+}
+
+let Predicates = [HasAVX1Only] in {
+def : Pat<(vextractf128_extract:$ext VR256:$src1, (iPTR imm)),
(v2i64 (VEXTRACTF128rr
- (v4i64 VR256:$src1),
- (EXTRACT_get_vextractf128_imm VR128:$ext)))>;
-def : Pat<(vextractf128_extract:$ext VR256:$src1, (i32 imm)),
+ (v4i64 VR256:$src1),
+ (EXTRACT_get_vextractf128_imm VR128:$ext)))>;
+def : Pat<(vextractf128_extract:$ext VR256:$src1, (iPTR imm)),
(v4i32 (VEXTRACTF128rr
- (v8i32 VR256:$src1),
- (EXTRACT_get_vextractf128_imm VR128:$ext)))>;
-def : Pat<(vextractf128_extract:$ext VR256:$src1, (i32 imm)),
+ (v8i32 VR256:$src1),
+ (EXTRACT_get_vextractf128_imm VR128:$ext)))>;
+def : Pat<(vextractf128_extract:$ext VR256:$src1, (iPTR imm)),
(v8i16 (VEXTRACTF128rr
- (v16i16 VR256:$src1),
- (EXTRACT_get_vextractf128_imm VR128:$ext)))>;
-def : Pat<(vextractf128_extract:$ext VR256:$src1, (i32 imm)),
+ (v16i16 VR256:$src1),
+ (EXTRACT_get_vextractf128_imm VR128:$ext)))>;
+def : Pat<(vextractf128_extract:$ext VR256:$src1, (iPTR imm)),
(v16i8 (VEXTRACTF128rr
- (v32i8 VR256:$src1),
- (EXTRACT_get_vextractf128_imm VR128:$ext)))>;
+ (v32i8 VR256:$src1),
+ (EXTRACT_get_vextractf128_imm VR128:$ext)))>;
+
+def : Pat<(alignedstore (v2i64 (vextractf128_extract:$ext (v4i64 VR256:$src1),
+ (iPTR imm))), addr:$dst),
+ (VEXTRACTF128mr addr:$dst, VR256:$src1,
+ (EXTRACT_get_vextractf128_imm VR128:$ext))>;
+def : Pat<(alignedstore (v4i32 (vextractf128_extract:$ext (v8i32 VR256:$src1),
+ (iPTR imm))), addr:$dst),
+ (VEXTRACTF128mr addr:$dst, VR256:$src1,
+ (EXTRACT_get_vextractf128_imm VR128:$ext))>;
+def : Pat<(alignedstore (v8i16 (vextractf128_extract:$ext (v16i16 VR256:$src1),
+ (iPTR imm))), addr:$dst),
+ (VEXTRACTF128mr addr:$dst, VR256:$src1,
+ (EXTRACT_get_vextractf128_imm VR128:$ext))>;
+def : Pat<(alignedstore (v16i8 (vextractf128_extract:$ext (v32i8 VR256:$src1),
+ (iPTR imm))), addr:$dst),
+ (VEXTRACTF128mr addr:$dst, VR256:$src1,
+ (EXTRACT_get_vextractf128_imm VR128:$ext))>;
}
//===----------------------------------------------------------------------===//
(ins VR256:$src1, f256mem:$src2),
!strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
[(set VR256:$dst, (IntLd256 addr:$src2, VR256:$src1))]>,
- VEX_4V;
+ VEX_4V, VEX_L;
def mr : AVX8I<opc_mr, MRMDestMem, (outs),
(ins f128mem:$dst, VR128:$src1, VR128:$src2),
!strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
def Ymr : AVX8I<opc_mr, MRMDestMem, (outs),
(ins f256mem:$dst, VR256:$src1, VR256:$src2),
!strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
- [(IntSt256 addr:$dst, VR256:$src1, VR256:$src2)]>, VEX_4V;
+ [(IntSt256 addr:$dst, VR256:$src1, VR256:$src2)]>, VEX_4V, VEX_L;
}
let ExeDomain = SSEPackedSingle in
defm VPERMILPS : avx_permil<0x0C, 0x04, "vpermilps", VR128, f128mem, i128mem,
memopv2i64, int_x86_avx_vpermilvar_ps, v4f32>;
defm VPERMILPSY : avx_permil<0x0C, 0x04, "vpermilps", VR256, f256mem, i256mem,
- memopv4i64, int_x86_avx_vpermilvar_ps_256, v8f32>;
+ memopv4i64, int_x86_avx_vpermilvar_ps_256, v8f32>, VEX_L;
}
let ExeDomain = SSEPackedDouble in {
defm VPERMILPD : avx_permil<0x0D, 0x05, "vpermilpd", VR128, f128mem, i128mem,
memopv2i64, int_x86_avx_vpermilvar_pd, v2f64>;
defm VPERMILPDY : avx_permil<0x0D, 0x05, "vpermilpd", VR256, f256mem, i256mem,
- memopv4i64, int_x86_avx_vpermilvar_pd_256, v4f64>;
+ memopv4i64, int_x86_avx_vpermilvar_pd_256, v4f64>, VEX_L;
}
let Predicates = [HasAVX] in {
(ins VR256:$src1, VR256:$src2, i8imm:$src3),
"vperm2f128\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}",
[(set VR256:$dst, (v8f32 (X86VPerm2x128 VR256:$src1, VR256:$src2,
- (i8 imm:$src3))))]>, VEX_4V;
+ (i8 imm:$src3))))]>, VEX_4V, VEX_L;
def VPERM2F128rm : AVXAIi8<0x06, MRMSrcMem, (outs VR256:$dst),
(ins VR256:$src1, f256mem:$src2, i8imm:$src3),
"vperm2f128\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}",
[(set VR256:$dst, (X86VPerm2x128 VR256:$src1, (memopv8f32 addr:$src2),
- (i8 imm:$src3)))]>, VEX_4V;
+ (i8 imm:$src3)))]>, VEX_4V, VEX_L;
}
let Predicates = [HasAVX] in {
+def : Pat<(v4f64 (X86VPerm2x128 VR256:$src1, VR256:$src2, (i8 imm:$imm))),
+ (VPERM2F128rr VR256:$src1, VR256:$src2, imm:$imm)>;
+def : Pat<(v4f64 (X86VPerm2x128 VR256:$src1,
+ (memopv4f64 addr:$src2), (i8 imm:$imm))),
+ (VPERM2F128rm VR256:$src1, addr:$src2, imm:$imm)>;
+}
+
+let Predicates = [HasAVX1Only] in {
def : Pat<(v8i32 (X86VPerm2x128 VR256:$src1, VR256:$src2, (i8 imm:$imm))),
(VPERM2F128rr VR256:$src1, VR256:$src2, imm:$imm)>;
def : Pat<(v4i64 (X86VPerm2x128 VR256:$src1, VR256:$src2, (i8 imm:$imm))),
(VPERM2F128rr VR256:$src1, VR256:$src2, imm:$imm)>;
-def : Pat<(v4f64 (X86VPerm2x128 VR256:$src1, VR256:$src2, (i8 imm:$imm))),
- (VPERM2F128rr VR256:$src1, VR256:$src2, imm:$imm)>;
def : Pat<(v32i8 (X86VPerm2x128 VR256:$src1, VR256:$src2, (i8 imm:$imm))),
(VPERM2F128rr VR256:$src1, VR256:$src2, imm:$imm)>;
def : Pat<(v16i16 (X86VPerm2x128 VR256:$src1, VR256:$src2, (i8 imm:$imm))),
(VPERM2F128rr VR256:$src1, VR256:$src2, imm:$imm)>;
-def : Pat<(v8f32 (X86VPerm2x128 VR256:$src1,
- (memopv8f32 addr:$src2), (i8 imm:$imm))),
- (VPERM2F128rm VR256:$src1, addr:$src2, imm:$imm)>;
def : Pat<(v8i32 (X86VPerm2x128 VR256:$src1,
(bc_v8i32 (memopv4i64 addr:$src2)), (i8 imm:$imm))),
(VPERM2F128rm VR256:$src1, addr:$src2, imm:$imm)>;
def : Pat<(v4i64 (X86VPerm2x128 VR256:$src1,
(memopv4i64 addr:$src2), (i8 imm:$imm))),
(VPERM2F128rm VR256:$src1, addr:$src2, imm:$imm)>;
-def : Pat<(v4f64 (X86VPerm2x128 VR256:$src1,
- (memopv4f64 addr:$src2), (i8 imm:$imm))),
- (VPERM2F128rm VR256:$src1, addr:$src2, imm:$imm)>;
def : Pat<(v32i8 (X86VPerm2x128 VR256:$src1,
(bc_v32i8 (memopv4i64 addr:$src2)), (i8 imm:$imm))),
(VPERM2F128rm VR256:$src1, addr:$src2, imm:$imm)>;
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 VCVTPH2PSY : f16c_ph2ps<VR256, f128mem, int_x86_vcvtph2ps_256>, VEX_L;
defm VCVTPS2PH : f16c_ps2ph<VR128, f64mem, int_x86_vcvtps2ph_128>;
- defm VCVTPS2PHY : f16c_ps2ph<VR256, f128mem, int_x86_vcvtps2ph_256>;
+ defm VCVTPS2PHY : f16c_ps2ph<VR256, f128mem, int_x86_vcvtps2ph_256>, VEX_L;
}
//===----------------------------------------------------------------------===//
defm VPBLENDD : AVX2_binop_rmi_int<0x02, "vpblendd", int_x86_avx2_pblendd_128,
VR128, memopv2i64, i128mem>;
defm VPBLENDDY : AVX2_binop_rmi_int<0x02, "vpblendd", int_x86_avx2_pblendd_256,
- VR256, memopv4i64, i256mem>;
+ VR256, memopv4i64, i256mem>, VEX_L;
}
+def : Pat<(v4i32 (X86Blendi (v4i32 VR128:$src1), (v4i32 VR128:$src2),
+ imm:$mask)),
+ (VPBLENDDrri VR128:$src1, VR128:$src2, imm:$mask)>;
+def : Pat<(v8i32 (X86Blendi (v8i32 VR256:$src1), (v8i32 VR256:$src2),
+ imm:$mask)),
+ (VPBLENDDYrri VR256:$src1, VR256:$src2, imm:$mask)>;
+
//===----------------------------------------------------------------------===//
// VPBROADCAST - Load from memory and broadcast to all elements of the
// destination operand
(Int128 (scalar_to_vector (ld_frag addr:$src))))]>, VEX;
def Yrr : AVX28I<opc, MRMSrcReg, (outs VR256:$dst), (ins VR128:$src),
!strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"),
- [(set VR256:$dst, (Int256 VR128:$src))]>, VEX;
+ [(set VR256:$dst, (Int256 VR128:$src))]>, VEX, VEX_L;
def Yrm : AVX28I<opc, MRMSrcMem, (outs VR256:$dst), (ins x86memop:$src),
!strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"),
[(set VR256:$dst,
- (Int256 (scalar_to_vector (ld_frag addr:$src))))]>, VEX;
+ (Int256 (scalar_to_vector (ld_frag addr:$src))))]>,
+ VEX, VEX_L;
}
defm VPBROADCASTB : avx2_broadcast<0x78, "vpbroadcastb", i8mem, loadi8,
}
// AVX1 broadcast patterns
-let Predicates = [HasAVX] in {
+let Predicates = [HasAVX1Only] in {
def : Pat<(v8i32 (X86VBroadcast (loadi32 addr:$src))),
(VBROADCASTSSYrm addr:$src)>;
def : Pat<(v4i64 (X86VBroadcast (loadi64 addr:$src))),
(VBROADCASTSDYrm addr:$src)>;
+def : Pat<(v4i32 (X86VBroadcast (loadi32 addr:$src))),
+ (VBROADCASTSSrm addr:$src)>;
+}
+
+let Predicates = [HasAVX] in {
def : Pat<(v8f32 (X86VBroadcast (loadf32 addr:$src))),
(VBROADCASTSSYrm addr:$src)>;
def : Pat<(v4f64 (X86VBroadcast (loadf64 addr:$src))),
(VBROADCASTSDYrm addr:$src)>;
def : Pat<(v4f32 (X86VBroadcast (loadf32 addr:$src))),
(VBROADCASTSSrm addr:$src)>;
-def : Pat<(v4i32 (X86VBroadcast (loadi32 addr:$src))),
- (VBROADCASTSSrm addr:$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.
!strconcat(OpcodeStr,
"\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
[(set VR256:$dst,
- (OpVT (X86VPermv VR256:$src1, VR256:$src2)))]>, VEX_4V;
+ (OpVT (X86VPermv VR256:$src1, VR256:$src2)))]>,
+ VEX_4V, VEX_L;
def Yrm : AVX28I<opc, MRMSrcMem, (outs VR256:$dst),
(ins VR256:$src1, i256mem:$src2),
!strconcat(OpcodeStr,
[(set VR256:$dst,
(OpVT (X86VPermv VR256:$src1,
(bitconvert (mem_frag addr:$src2)))))]>,
- VEX_4V;
+ VEX_4V, VEX_L;
}
defm VPERMD : avx2_perm<0x36, "vpermd", memopv4i64, v8i32>;
!strconcat(OpcodeStr,
"\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
[(set VR256:$dst,
- (OpVT (X86VPermi VR256:$src1, (i8 imm:$src2))))]>, VEX;
+ (OpVT (X86VPermi VR256:$src1, (i8 imm:$src2))))]>,
+ VEX, VEX_L;
def Ymi : AVX2AIi8<opc, MRMSrcMem, (outs VR256:$dst),
(ins i256mem:$src1, i8imm:$src2),
!strconcat(OpcodeStr,
"\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
[(set VR256:$dst,
(OpVT (X86VPermi (mem_frag addr:$src1),
- (i8 imm:$src2))))]>, VEX;
+ (i8 imm:$src2))))]>, VEX, VEX_L;
}
defm VPERMQ : avx2_perm_imm<0x00, "vpermq", memopv4i64, v4i64>, VEX_W;
//===----------------------------------------------------------------------===//
// VPERM2I128 - Permute Floating-Point Values in 128-bit chunks
//
-let AddedComplexity = 1 in {
def VPERM2I128rr : AVX2AIi8<0x46, MRMSrcReg, (outs VR256:$dst),
(ins VR256:$src1, VR256:$src2, i8imm:$src3),
"vperm2i128\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}",
[(set VR256:$dst, (v4i64 (X86VPerm2x128 VR256:$src1, VR256:$src2,
- (i8 imm:$src3))))]>, VEX_4V;
+ (i8 imm:$src3))))]>, VEX_4V, VEX_L;
def VPERM2I128rm : AVX2AIi8<0x46, MRMSrcMem, (outs VR256:$dst),
(ins VR256:$src1, f256mem:$src2, i8imm:$src3),
"vperm2i128\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}",
[(set VR256:$dst, (X86VPerm2x128 VR256:$src1, (memopv4i64 addr:$src2),
- (i8 imm:$src3)))]>, VEX_4V;
-}
+ (i8 imm:$src3)))]>, VEX_4V, VEX_L;
-let Predicates = [HasAVX2], AddedComplexity = 1 in {
+let Predicates = [HasAVX2] in {
def : Pat<(v8i32 (X86VPerm2x128 VR256:$src1, VR256:$src2, (i8 imm:$imm))),
(VPERM2I128rr VR256:$src1, VR256:$src2, imm:$imm)>;
def : Pat<(v32i8 (X86VPerm2x128 VR256:$src1, VR256:$src2, (i8 imm:$imm))),
def VINSERTI128rr : AVX2AIi8<0x38, MRMSrcReg, (outs VR256:$dst),
(ins VR256:$src1, VR128:$src2, i8imm:$src3),
"vinserti128\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}",
- []>, VEX_4V;
+ []>, VEX_4V, VEX_L;
let mayLoad = 1 in
def VINSERTI128rm : AVX2AIi8<0x38, MRMSrcMem, (outs VR256:$dst),
(ins VR256:$src1, i128mem:$src2, i8imm:$src3),
"vinserti128\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}",
- []>, VEX_4V;
+ []>, VEX_4V, VEX_L;
}
-let Predicates = [HasAVX2], AddedComplexity = 1 in {
+let Predicates = [HasAVX2] in {
def : Pat<(vinsertf128_insert:$ins (v4i64 VR256:$src1), (v2i64 VR128:$src2),
- (i32 imm)),
+ (iPTR imm)),
(VINSERTI128rr VR256:$src1, VR128:$src2,
(INSERT_get_vinsertf128_imm VR256:$ins))>;
def : Pat<(vinsertf128_insert:$ins (v8i32 VR256:$src1), (v4i32 VR128:$src2),
- (i32 imm)),
+ (iPTR imm)),
(VINSERTI128rr VR256:$src1, VR128:$src2,
(INSERT_get_vinsertf128_imm VR256:$ins))>;
def : Pat<(vinsertf128_insert:$ins (v32i8 VR256:$src1), (v16i8 VR128:$src2),
- (i32 imm)),
+ (iPTR imm)),
(VINSERTI128rr VR256:$src1, VR128:$src2,
(INSERT_get_vinsertf128_imm VR256:$ins))>;
def : Pat<(vinsertf128_insert:$ins (v16i16 VR256:$src1), (v8i16 VR128:$src2),
- (i32 imm)),
+ (iPTR imm)),
(VINSERTI128rr VR256:$src1, VR128:$src2,
(INSERT_get_vinsertf128_imm VR256:$ins))>;
+
+def : Pat<(vinsertf128_insert:$ins (v4i64 VR256:$src1), (memopv2i64 addr:$src2),
+ (iPTR imm)),
+ (VINSERTI128rm VR256:$src1, addr:$src2,
+ (INSERT_get_vinsertf128_imm VR256:$ins))>;
+def : Pat<(vinsertf128_insert:$ins (v8i32 VR256:$src1),
+ (bc_v4i32 (memopv2i64 addr:$src2)),
+ (iPTR imm)),
+ (VINSERTI128rm VR256:$src1, addr:$src2,
+ (INSERT_get_vinsertf128_imm VR256:$ins))>;
+def : Pat<(vinsertf128_insert:$ins (v32i8 VR256:$src1),
+ (bc_v16i8 (memopv2i64 addr:$src2)),
+ (iPTR imm)),
+ (VINSERTI128rm VR256:$src1, addr:$src2,
+ (INSERT_get_vinsertf128_imm VR256:$ins))>;
+def : Pat<(vinsertf128_insert:$ins (v16i16 VR256:$src1),
+ (bc_v8i16 (memopv2i64 addr:$src2)),
+ (iPTR imm)),
+ (VINSERTI128rm VR256:$src1, addr:$src2,
+ (INSERT_get_vinsertf128_imm VR256:$ins))>;
}
//===----------------------------------------------------------------------===//
"vextracti128\t{$src2, $src1, $dst|$dst, $src1, $src2}",
[(set VR128:$dst,
(int_x86_avx2_vextracti128 VR256:$src1, imm:$src2))]>,
- VEX;
+ VEX, VEX_L;
let neverHasSideEffects = 1, mayStore = 1 in
def VEXTRACTI128mr : AVX2AIi8<0x39, MRMDestMem, (outs),
(ins i128mem:$dst, VR256:$src1, i8imm:$src2),
- "vextracti128\t{$src2, $src1, $dst|$dst, $src1, $src2}", []>, VEX;
+ "vextracti128\t{$src2, $src1, $dst|$dst, $src1, $src2}", []>,
+ VEX, VEX_L;
-let Predicates = [HasAVX2], AddedComplexity = 1 in {
-def : Pat<(vextractf128_extract:$ext VR256:$src1, (i32 imm)),
+let Predicates = [HasAVX2] in {
+def : Pat<(vextractf128_extract:$ext VR256:$src1, (iPTR imm)),
(v2i64 (VEXTRACTI128rr
(v4i64 VR256:$src1),
(EXTRACT_get_vextractf128_imm VR128:$ext)))>;
-def : Pat<(vextractf128_extract:$ext VR256:$src1, (i32 imm)),
+def : Pat<(vextractf128_extract:$ext VR256:$src1, (iPTR imm)),
(v4i32 (VEXTRACTI128rr
(v8i32 VR256:$src1),
(EXTRACT_get_vextractf128_imm VR128:$ext)))>;
-def : Pat<(vextractf128_extract:$ext VR256:$src1, (i32 imm)),
+def : Pat<(vextractf128_extract:$ext VR256:$src1, (iPTR imm)),
(v8i16 (VEXTRACTI128rr
(v16i16 VR256:$src1),
(EXTRACT_get_vextractf128_imm VR128:$ext)))>;
-def : Pat<(vextractf128_extract:$ext VR256:$src1, (i32 imm)),
+def : Pat<(vextractf128_extract:$ext VR256:$src1, (iPTR imm)),
(v16i8 (VEXTRACTI128rr
(v32i8 VR256:$src1),
(EXTRACT_get_vextractf128_imm VR128:$ext)))>;
+
+def : Pat<(alignedstore (v2i64 (vextractf128_extract:$ext (v4i64 VR256:$src1),
+ (iPTR imm))), addr:$dst),
+ (VEXTRACTI128mr addr:$dst, VR256:$src1,
+ (EXTRACT_get_vextractf128_imm VR128:$ext))>;
+def : Pat<(alignedstore (v4i32 (vextractf128_extract:$ext (v8i32 VR256:$src1),
+ (iPTR imm))), addr:$dst),
+ (VEXTRACTI128mr addr:$dst, VR256:$src1,
+ (EXTRACT_get_vextractf128_imm VR128:$ext))>;
+def : Pat<(alignedstore (v8i16 (vextractf128_extract:$ext (v16i16 VR256:$src1),
+ (iPTR imm))), addr:$dst),
+ (VEXTRACTI128mr addr:$dst, VR256:$src1,
+ (EXTRACT_get_vextractf128_imm VR128:$ext))>;
+def : Pat<(alignedstore (v16i8 (vextractf128_extract:$ext (v32i8 VR256:$src1),
+ (iPTR imm))), addr:$dst),
+ (VEXTRACTI128mr addr:$dst, VR256:$src1,
+ (EXTRACT_get_vextractf128_imm VR128:$ext))>;
}
//===----------------------------------------------------------------------===//
def Yrm : AVX28I<0x8c, MRMSrcMem, (outs VR256:$dst),
(ins VR256:$src1, i256mem:$src2),
!strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
- [(set VR256:$dst, (IntLd256 addr:$src2, VR256:$src1))]>, VEX_4V;
+ [(set VR256:$dst, (IntLd256 addr:$src2, VR256:$src1))]>,
+ VEX_4V, VEX_L;
def mr : AVX28I<0x8e, MRMDestMem, (outs),
(ins i128mem:$dst, VR128:$src1, VR128:$src2),
!strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
def Ymr : AVX28I<0x8e, MRMDestMem, (outs),
(ins i256mem:$dst, VR256:$src1, VR256:$src2),
!strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
- [(IntSt256 addr:$dst, VR256:$src1, VR256:$src2)]>, VEX_4V;
+ [(IntSt256 addr:$dst, VR256:$src1, VR256:$src2)]>, VEX_4V, VEX_L;
}
defm VPMASKMOVD : avx2_pmovmask<"vpmaskmovd",
!strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
[(set VR256:$dst,
(vt256 (OpNode VR256:$src1, (vt256 VR256:$src2))))]>,
- VEX_4V;
+ VEX_4V, VEX_L;
def Yrm : AVX28I<opc, MRMSrcMem, (outs VR256:$dst),
(ins VR256:$src1, i256mem:$src2),
!strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
[(set VR256:$dst,
(vt256 (OpNode VR256:$src1,
(vt256 (bitconvert (memopv4i64 addr:$src2))))))]>,
- VEX_4V;
+ VEX_4V, VEX_L;
}
defm VPSLLVD : avx2_var_shift<0x47, "vpsllvd", shl, v4i32, v8i32>;
[]>, VEX_4VOp3, VEX_L;
}
-let Constraints = "$src1 = $dst, $mask = $mask_wb" in {
+let mayLoad = 1, Constraints = "$src1 = $dst, $mask = $mask_wb" in {
defm VGATHERDPD : avx2_gather<0x92, "vgatherdpd", VR256, vx64mem, vx64mem>, VEX_W;
defm VGATHERQPD : avx2_gather<0x93, "vgatherqpd", VR256, vx64mem, vy64mem>, VEX_W;
defm VGATHERDPS : avx2_gather<0x92, "vgatherdps", VR256, vx32mem, vy32mem>;