//
//===----------------------------------------------------------------------===//
+class OpndItins<InstrItinClass arg_rr, InstrItinClass arg_rm> {
+ InstrItinClass rr = arg_rr;
+ InstrItinClass rm = arg_rm;
+}
+
+class SizeItins<OpndItins arg_s, OpndItins arg_d> {
+ OpndItins s = arg_s;
+ OpndItins d = arg_d;
+}
+
+
+class ShiftOpndItins<InstrItinClass arg_rr, InstrItinClass arg_rm,
+ InstrItinClass arg_ri> {
+ InstrItinClass rr = arg_rr;
+ InstrItinClass rm = arg_rm;
+ InstrItinClass ri = arg_ri;
+}
+
+
+// scalar
+def SSE_ALU_F32S : OpndItins<
+ IIC_SSE_ALU_F32S_RR, IIC_SSE_ALU_F32S_RM
+>;
+
+def SSE_ALU_F64S : OpndItins<
+ IIC_SSE_ALU_F64S_RR, IIC_SSE_ALU_F64S_RM
+>;
+
+def SSE_ALU_ITINS_S : SizeItins<
+ SSE_ALU_F32S, SSE_ALU_F64S
+>;
+
+def SSE_MUL_F32S : OpndItins<
+ IIC_SSE_MUL_F32S_RR, IIC_SSE_MUL_F64S_RM
+>;
+
+def SSE_MUL_F64S : OpndItins<
+ IIC_SSE_MUL_F64S_RR, IIC_SSE_MUL_F64S_RM
+>;
+
+def SSE_MUL_ITINS_S : SizeItins<
+ SSE_MUL_F32S, SSE_MUL_F64S
+>;
+
+def SSE_DIV_F32S : OpndItins<
+ IIC_SSE_DIV_F32S_RR, IIC_SSE_DIV_F64S_RM
+>;
+
+def SSE_DIV_F64S : OpndItins<
+ IIC_SSE_DIV_F64S_RR, IIC_SSE_DIV_F64S_RM
+>;
+
+def SSE_DIV_ITINS_S : SizeItins<
+ SSE_DIV_F32S, SSE_DIV_F64S
+>;
+
+// parallel
+def SSE_ALU_F32P : OpndItins<
+ IIC_SSE_ALU_F32P_RR, IIC_SSE_ALU_F32P_RM
+>;
+
+def SSE_ALU_F64P : OpndItins<
+ IIC_SSE_ALU_F64P_RR, IIC_SSE_ALU_F64P_RM
+>;
+
+def SSE_ALU_ITINS_P : SizeItins<
+ SSE_ALU_F32P, SSE_ALU_F64P
+>;
+
+def SSE_MUL_F32P : OpndItins<
+ IIC_SSE_MUL_F32P_RR, IIC_SSE_MUL_F64P_RM
+>;
+
+def SSE_MUL_F64P : OpndItins<
+ IIC_SSE_MUL_F64P_RR, IIC_SSE_MUL_F64P_RM
+>;
+
+def SSE_MUL_ITINS_P : SizeItins<
+ SSE_MUL_F32P, SSE_MUL_F64P
+>;
+
+def SSE_DIV_F32P : OpndItins<
+ IIC_SSE_DIV_F32P_RR, IIC_SSE_DIV_F64P_RM
+>;
+
+def SSE_DIV_F64P : OpndItins<
+ IIC_SSE_DIV_F64P_RR, IIC_SSE_DIV_F64P_RM
+>;
+
+def SSE_DIV_ITINS_P : SizeItins<
+ SSE_DIV_F32P, SSE_DIV_F64P
+>;
+
+def SSE_BIT_ITINS_P : OpndItins<
+ IIC_SSE_BIT_P_RR, IIC_SSE_BIT_P_RM
+>;
+
+def SSE_INTALU_ITINS_P : OpndItins<
+ IIC_SSE_INTALU_P_RR, IIC_SSE_INTALU_P_RM
+>;
+
+def SSE_INTALUQ_ITINS_P : OpndItins<
+ IIC_SSE_INTALUQ_P_RR, IIC_SSE_INTALUQ_P_RM
+>;
+
+def SSE_INTMUL_ITINS_P : OpndItins<
+ IIC_SSE_INTMUL_P_RR, IIC_SSE_INTMUL_P_RM
+>;
+
+def SSE_INTSHIFT_ITINS_P : ShiftOpndItins<
+ IIC_SSE_INTSH_P_RR, IIC_SSE_INTSH_P_RM, IIC_SSE_INTSH_P_RI
+>;
+
+def SSE_MOVA_ITINS : OpndItins<
+ IIC_SSE_MOVA_P_RR, IIC_SSE_MOVA_P_RM
+>;
+
+def SSE_MOVU_ITINS : OpndItins<
+ IIC_SSE_MOVU_P_RR, IIC_SSE_MOVU_P_RM
+>;
//===----------------------------------------------------------------------===//
// SSE 1 & 2 Instructions Classes
/// sse12_fp_scalar - SSE 1 & 2 scalar instructions class
multiclass sse12_fp_scalar<bits<8> opc, string OpcodeStr, SDNode OpNode,
RegisterClass RC, X86MemOperand x86memop,
+ OpndItins itins,
bit Is2Addr = 1> {
let isCommutable = 1 in {
def rr : SI<opc, MRMSrcReg, (outs RC:$dst), (ins RC:$src1, RC:$src2),
!if(Is2Addr,
!strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"),
!strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}")),
- [(set RC:$dst, (OpNode RC:$src1, RC:$src2))]>;
+ [(set RC:$dst, (OpNode RC:$src1, RC:$src2))], itins.rr>;
}
def rm : SI<opc, MRMSrcMem, (outs RC:$dst), (ins RC:$src1, x86memop:$src2),
!if(Is2Addr,
!strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"),
!strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}")),
- [(set RC:$dst, (OpNode RC:$src1, (load addr:$src2)))]>;
+ [(set RC:$dst, (OpNode RC:$src1, (load addr:$src2)))], itins.rm>;
}
/// sse12_fp_scalar_int - SSE 1 & 2 scalar instructions intrinsics class
multiclass sse12_fp_scalar_int<bits<8> opc, string OpcodeStr, RegisterClass RC,
string asm, string SSEVer, string FPSizeStr,
Operand memopr, ComplexPattern mem_cpat,
+ OpndItins itins,
bit Is2Addr = 1> {
def rr_Int : SI<opc, MRMSrcReg, (outs RC:$dst), (ins RC:$src1, RC:$src2),
!if(Is2Addr,
!strconcat(asm, "\t{$src2, $src1, $dst|$dst, $src1, $src2}")),
[(set RC:$dst, (!cast<Intrinsic>(
!strconcat("int_x86_sse", SSEVer, "_", OpcodeStr, FPSizeStr))
- RC:$src1, RC:$src2))]>;
+ RC:$src1, RC:$src2))], itins.rr>;
def rm_Int : SI<opc, MRMSrcMem, (outs RC:$dst), (ins RC:$src1, memopr:$src2),
!if(Is2Addr,
!strconcat(asm, "\t{$src2, $dst|$dst, $src2}"),
!strconcat(asm, "\t{$src2, $src1, $dst|$dst, $src1, $src2}")),
[(set RC:$dst, (!cast<Intrinsic>(!strconcat("int_x86_sse",
SSEVer, "_", OpcodeStr, FPSizeStr))
- RC:$src1, mem_cpat:$src2))]>;
+ RC:$src1, mem_cpat:$src2))], itins.rm>;
}
/// sse12_fp_packed - SSE 1 & 2 packed instructions class
multiclass sse12_fp_packed<bits<8> opc, string OpcodeStr, SDNode OpNode,
RegisterClass RC, ValueType vt,
X86MemOperand x86memop, PatFrag mem_frag,
- Domain d, bit Is2Addr = 1> {
+ Domain d, OpndItins itins, bit Is2Addr = 1> {
let isCommutable = 1 in
def rr : PI<opc, MRMSrcReg, (outs RC:$dst), (ins RC:$src1, RC:$src2),
!if(Is2Addr,
!strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"),
!strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}")),
- [(set RC:$dst, (vt (OpNode RC:$src1, RC:$src2)))], IIC_DEFAULT, d>;
+ [(set RC:$dst, (vt (OpNode RC:$src1, RC:$src2)))], itins.rr, d>;
let mayLoad = 1 in
def rm : PI<opc, MRMSrcMem, (outs RC:$dst), (ins RC:$src1, x86memop:$src2),
!if(Is2Addr,
!strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"),
!strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}")),
[(set RC:$dst, (OpNode RC:$src1, (mem_frag addr:$src2)))],
- IIC_DEFAULT, d>;
+ itins.rm, d>;
}
/// sse12_fp_packed_logical_rm - SSE 1 & 2 packed instructions 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, bit Is2Addr = 1> {
+ 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, $src2}"),
// A vector extract of the first f32/f64 position is a subregister copy
def : Pat<(f32 (vector_extract (v4f32 VR128:$src), (iPTR 0))),
- (f32 (EXTRACT_SUBREG (v4f32 VR128:$src), sub_ss))>;
+ (COPY_TO_REGCLASS (v4f32 VR128:$src), FR32)>;
def : Pat<(f64 (vector_extract (v2f64 VR128:$src), (iPTR 0))),
- (f64 (EXTRACT_SUBREG (v2f64 VR128:$src), sub_sd))>;
+ (COPY_TO_REGCLASS (v2f64 VR128:$src), FR64)>;
// A 128-bit subvector extract from the first 256-bit vector position
// is a subregister copy that needs no instruction.
// A 128-bit subvector insert to the first 256-bit vector position
// is a subregister copy that needs no instruction.
+let AddedComplexity = 25 in { // to give priority over vinsertf128rm
def : Pat<(insert_subvector undef, (v2i64 VR128:$src), (i32 0)),
(INSERT_SUBREG (v4i64 (IMPLICIT_DEF)), VR128:$src, sub_xmm)>;
def : Pat<(insert_subvector undef, (v2f64 VR128:$src), (i32 0)),
(INSERT_SUBREG (v16i16 (IMPLICIT_DEF)), VR128:$src, sub_xmm)>;
def : Pat<(insert_subvector undef, (v16i8 VR128:$src), (i32 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)),
- (INSERT_SUBREG (v4f32 (IMPLICIT_DEF)), FR32:$src, sub_ss)>;
+ (COPY_TO_REGCLASS FR32:$src, VR128)>;
def : Pat<(v8f32 (scalar_to_vector FR32:$src)),
- (INSERT_SUBREG (v8f32 (IMPLICIT_DEF)), FR32:$src, sub_ss)>;
+ (COPY_TO_REGCLASS FR32:$src, VR128)>;
// Implicitly promote a 64-bit scalar to a vector.
def : Pat<(v2f64 (scalar_to_vector FR64:$src)),
- (INSERT_SUBREG (v2f64 (IMPLICIT_DEF)), FR64:$src, sub_sd)>;
+ (COPY_TO_REGCLASS FR64:$src, VR128)>;
def : Pat<(v4f64 (scalar_to_vector FR64:$src)),
- (INSERT_SUBREG (v4f64 (IMPLICIT_DEF)), FR64:$src, sub_sd)>;
+ (COPY_TO_REGCLASS FR64:$src, VR128)>;
// Bitcasts between 128-bit vector types. Return the original type since
// no instruction is needed for the conversion
def : Pat<(v16i16 (bitconvert (v32i8 VR256:$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:$dst, (v8i32 immAllOnesV))]>;
}
class sse12_move_rr<RegisterClass RC, SDNode OpNode, ValueType vt, string asm> :
SI<0x10, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src1, RC:$src2), asm,
[(set VR128:$dst, (vt (OpNode VR128:$src1,
- (scalar_to_vector RC:$src2))))]>;
+ (scalar_to_vector RC:$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:$src))]>;
+ [(set RC:$dst, (mem_pat addr:$src))],
+ IIC_SSE_MOV_S_RM>;
// AVX
def VMOVSSrr : sse12_move_rr<FR32, X86Movss, v4f32,
let isCodeGenOnly = 1 in {
def VMOVSSrr_REV : SI<0x11, MRMDestReg, (outs VR128:$dst),
(ins VR128:$src1, FR32:$src2),
- "movss\t{$src2, $src1, $dst|$dst, $src1, $src2}", []>,
+ "movss\t{$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, $src2}", []>,
+ "movsd\t{$src2, $src1, $dst|$dst, $src1, $src2}", [],
+ IIC_SSE_MOV_S_RR>,
XD, VEX_4V, VEX_LIG;
}
def VMOVSSmr : SI<0x11, MRMDestMem, (outs), (ins f32mem:$dst, FR32:$src),
"movss\t{$src, $dst|$dst, $src}",
- [(store FR32:$src, addr:$dst)]>, XS, VEX, VEX_LIG;
+ [(store FR32:$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:$dst)]>, XD, VEX, VEX_LIG;
+ [(store FR64:$src, addr:$dst)], IIC_SSE_MOV_S_MR>,
+ XD, VEX, VEX_LIG;
// SSE1 & 2
let Constraints = "$src1 = $dst" in {
let isCodeGenOnly = 1 in {
def MOVSSrr_REV : SI<0x11, MRMDestReg, (outs VR128:$dst),
(ins VR128:$src1, FR32:$src2),
- "movss\t{$src2, $dst|$dst, $src2}", []>, XS;
+ "movss\t{$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, $src2}", []>, XD;
+ "movsd\t{$src2, $dst|$dst, $src2}", [],
+ IIC_SSE_MOV_S_RR>, XD;
}
}
def MOVSSmr : SSI<0x11, MRMDestMem, (outs), (ins f32mem:$dst, FR32:$src),
"movss\t{$src, $dst|$dst, $src}",
- [(store FR32:$src, addr:$dst)]>;
+ [(store FR32:$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:$dst)]>;
+ [(store FR64:$src, addr:$dst)], IIC_SSE_MOV_S_MR>;
// Patterns
let Predicates = [HasAVX] in {
def : Pat<(v4f32 (X86vzmovl (v4f32 (scalar_to_vector FR32:$src)))),
(VMOVSSrr (v4f32 (V_SET0)), FR32:$src)>;
def : Pat<(v4f32 (X86vzmovl (v4f32 VR128:$src))),
- (VMOVSSrr (v4f32 (V_SET0)),
- (f32 (EXTRACT_SUBREG (v4f32 VR128:$src), sub_ss)))>;
+ (VMOVSSrr (v4f32 (V_SET0)), (COPY_TO_REGCLASS VR128:$src, FR32))>;
def : Pat<(v4i32 (X86vzmovl (v4i32 VR128:$src))),
- (VMOVSSrr (v4i32 (V_SET0)),
- (EXTRACT_SUBREG (v4i32 VR128:$src), sub_ss))>;
+ (VMOVSSrr (v4i32 (V_SET0)), (COPY_TO_REGCLASS VR128:$src, FR32))>;
def : Pat<(v2f64 (X86vzmovl (v2f64 (scalar_to_vector FR64:$src)))),
(VMOVSDrr (v2f64 (V_SET0)), FR64:$src)>;
// Move low f32 and clear high bits.
def : Pat<(v8f32 (X86vzmovl (v8f32 VR256:$src))),
(SUBREG_TO_REG (i32 0),
- (VMOVSSrr (v4f32 (V_SET0)),
- (EXTRACT_SUBREG (v8f32 VR256:$src), sub_ss)), sub_xmm)>;
+ (VMOVSSrr (v4f32 (V_SET0)),
+ (EXTRACT_SUBREG (v8f32 VR256:$src), sub_xmm)), sub_xmm)>;
def : Pat<(v8i32 (X86vzmovl (v8i32 VR256:$src))),
(SUBREG_TO_REG (i32 0),
- (VMOVSSrr (v4i32 (V_SET0)),
- (EXTRACT_SUBREG (v8i32 VR256:$src), sub_ss)), sub_xmm)>;
+ (VMOVSSrr (v4i32 (V_SET0)),
+ (EXTRACT_SUBREG (v8i32 VR256:$src), sub_xmm)), sub_xmm)>;
}
let AddedComplexity = 20 in {
// MOVSSrm zeros the high parts of the register; represent this
// with SUBREG_TO_REG. The AVX versions also write: DST[255:128] <- 0
def : Pat<(v4f32 (X86vzmovl (v4f32 (scalar_to_vector (loadf32 addr:$src))))),
- (SUBREG_TO_REG (i32 0), (VMOVSSrm addr:$src), sub_ss)>;
+ (COPY_TO_REGCLASS (VMOVSSrm addr:$src), VR128)>;
def : Pat<(v4f32 (scalar_to_vector (loadf32 addr:$src))),
- (SUBREG_TO_REG (i32 0), (VMOVSSrm addr:$src), sub_ss)>;
+ (COPY_TO_REGCLASS (VMOVSSrm addr:$src), VR128)>;
def : Pat<(v4f32 (X86vzmovl (loadv4f32 addr:$src))),
- (SUBREG_TO_REG (i32 0), (VMOVSSrm addr:$src), sub_ss)>;
+ (COPY_TO_REGCLASS (VMOVSSrm addr:$src), VR128)>;
// MOVSDrm zeros the high parts of the register; represent this
// with SUBREG_TO_REG. The AVX versions also write: DST[255:128] <- 0
def : Pat<(v2f64 (X86vzmovl (v2f64 (scalar_to_vector (loadf64 addr:$src))))),
- (SUBREG_TO_REG (i64 0), (VMOVSDrm addr:$src), sub_sd)>;
+ (COPY_TO_REGCLASS (VMOVSDrm addr:$src), VR128)>;
def : Pat<(v2f64 (scalar_to_vector (loadf64 addr:$src))),
- (SUBREG_TO_REG (i64 0), (VMOVSDrm addr:$src), sub_sd)>;
+ (COPY_TO_REGCLASS (VMOVSDrm addr:$src), VR128)>;
def : Pat<(v2f64 (X86vzmovl (loadv2f64 addr:$src))),
- (SUBREG_TO_REG (i64 0), (VMOVSDrm addr:$src), sub_sd)>;
+ (COPY_TO_REGCLASS (VMOVSDrm addr:$src), VR128)>;
def : Pat<(v2f64 (X86vzmovl (bc_v2f64 (loadv4f32 addr:$src)))),
- (SUBREG_TO_REG (i64 0), (VMOVSDrm addr:$src), sub_sd)>;
+ (COPY_TO_REGCLASS (VMOVSDrm addr:$src), VR128)>;
def : Pat<(v2f64 (X86vzload addr:$src)),
- (SUBREG_TO_REG (i64 0), (VMOVSDrm addr:$src), sub_sd)>;
+ (COPY_TO_REGCLASS (VMOVSDrm addr:$src), VR128)>;
// Represent the same patterns above but in the form they appear for
// 256-bit types
def : Pat<(v8i32 (X86vzmovl (insert_subvector undef,
(v4i32 (scalar_to_vector (loadi32 addr:$src))), (i32 0)))),
- (SUBREG_TO_REG (i32 0), (VMOVSSrm addr:$src), sub_ss)>;
+ (SUBREG_TO_REG (i32 0), (VMOVSSrm addr:$src), sub_xmm)>;
def : Pat<(v8f32 (X86vzmovl (insert_subvector undef,
(v4f32 (scalar_to_vector (loadf32 addr:$src))), (i32 0)))),
- (SUBREG_TO_REG (i32 0), (VMOVSSrm addr:$src), sub_ss)>;
+ (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)))),
- (SUBREG_TO_REG (i32 0), (VMOVSDrm addr:$src), sub_sd)>;
+ (SUBREG_TO_REG (i32 0), (VMOVSDrm addr:$src), sub_xmm)>;
}
def : Pat<(v8f32 (X86vzmovl (insert_subvector undef,
(v4f32 (scalar_to_vector FR32:$src)), (i32 0)))),
sub_xmm)>;
def : Pat<(v4i64 (X86vzmovl (insert_subvector undef,
(v2i64 (scalar_to_vector (loadi64 addr:$src))), (i32 0)))),
- (SUBREG_TO_REG (i64 0), (VMOVSDrm addr:$src), sub_sd)>;
+ (SUBREG_TO_REG (i64 0), (VMOVSDrm addr:$src), sub_xmm)>;
// Move low f64 and clear high bits.
def : Pat<(v4f64 (X86vzmovl (v4f64 VR256:$src))),
(SUBREG_TO_REG (i32 0),
- (VMOVSDrr (v2f64 (V_SET0)),
- (EXTRACT_SUBREG (v4f64 VR256:$src), sub_sd)), sub_xmm)>;
+ (VMOVSDrr (v2f64 (V_SET0)),
+ (EXTRACT_SUBREG (v4f64 VR256:$src), sub_xmm)), sub_xmm)>;
def : Pat<(v4i64 (X86vzmovl (v4i64 VR256:$src))),
(SUBREG_TO_REG (i32 0),
- (VMOVSDrr (v2i64 (V_SET0)),
- (EXTRACT_SUBREG (v4i64 VR256:$src), sub_sd)), sub_xmm)>;
+ (VMOVSDrr (v2i64 (V_SET0)),
+ (EXTRACT_SUBREG (v4i64 VR256:$src), sub_xmm)), sub_xmm)>;
-// Extract and store.
+ // Extract and store.
def : Pat<(store (f32 (vector_extract (v4f32 VR128:$src), (iPTR 0))),
addr:$dst),
- (VMOVSSmr addr:$dst,
- (EXTRACT_SUBREG (v4f32 VR128:$src), sub_ss))>;
+ (VMOVSSmr addr:$dst, (COPY_TO_REGCLASS (v4f32 VR128:$src), FR32))>;
def : Pat<(store (f64 (vector_extract (v2f64 VR128:$src), (iPTR 0))),
addr:$dst),
- (VMOVSDmr addr:$dst,
- (EXTRACT_SUBREG (v2f64 VR128:$src), sub_sd))>;
+ (VMOVSDmr addr:$dst, (COPY_TO_REGCLASS (v2f64 VR128:$src), FR64))>;
// Shuffle with VMOVSS
def : Pat<(v4i32 (X86Movss VR128:$src1, VR128:$src2)),
(VMOVSSrr (v4i32 VR128:$src1),
- (EXTRACT_SUBREG (v4i32 VR128:$src2), sub_ss))>;
+ (COPY_TO_REGCLASS (v4i32 VR128:$src2), FR32))>;
def : Pat<(v4f32 (X86Movss VR128:$src1, VR128:$src2)),
(VMOVSSrr (v4f32 VR128:$src1),
- (EXTRACT_SUBREG (v4f32 VR128:$src2), sub_ss))>;
+ (COPY_TO_REGCLASS (v4f32 VR128:$src2), FR32))>;
// 256-bit variants
def : Pat<(v8i32 (X86Movss VR256:$src1, VR256:$src2)),
(SUBREG_TO_REG (i32 0),
- (VMOVSSrr (EXTRACT_SUBREG (v8i32 VR256:$src1), sub_ss),
- (EXTRACT_SUBREG (v8i32 VR256:$src2), sub_ss)), sub_xmm)>;
+ (VMOVSSrr (EXTRACT_SUBREG (v8i32 VR256:$src1), sub_xmm),
+ (EXTRACT_SUBREG (v8i32 VR256:$src2), sub_xmm)),
+ sub_xmm)>;
def : Pat<(v8f32 (X86Movss VR256:$src1, VR256:$src2)),
(SUBREG_TO_REG (i32 0),
- (VMOVSSrr (EXTRACT_SUBREG (v8f32 VR256:$src1), sub_ss),
- (EXTRACT_SUBREG (v8f32 VR256:$src2), sub_ss)), sub_xmm)>;
+ (VMOVSSrr (EXTRACT_SUBREG (v8f32 VR256:$src1), sub_xmm),
+ (EXTRACT_SUBREG (v8f32 VR256:$src2), sub_xmm)),
+ sub_xmm)>;
// Shuffle with VMOVSD
def : Pat<(v2i64 (X86Movsd VR128:$src1, VR128:$src2)),
- (VMOVSDrr (v2i64 VR128:$src1),
- (EXTRACT_SUBREG (v2i64 VR128:$src2), sub_sd))>;
+ (VMOVSDrr VR128:$src1, (COPY_TO_REGCLASS VR128:$src2, FR64))>;
def : Pat<(v2f64 (X86Movsd VR128:$src1, VR128:$src2)),
- (VMOVSDrr (v2f64 VR128:$src1),
- (EXTRACT_SUBREG (v2f64 VR128:$src2), sub_sd))>;
+ (VMOVSDrr VR128:$src1, (COPY_TO_REGCLASS VR128:$src2, FR64))>;
def : Pat<(v4f32 (X86Movsd VR128:$src1, VR128:$src2)),
- (VMOVSDrr VR128:$src1, (EXTRACT_SUBREG (v4f32 VR128:$src2),
- sub_sd))>;
+ (VMOVSDrr VR128:$src1, (COPY_TO_REGCLASS VR128:$src2, FR64))>;
def : Pat<(v4i32 (X86Movsd VR128:$src1, VR128:$src2)),
- (VMOVSDrr VR128:$src1, (EXTRACT_SUBREG (v4i32 VR128:$src2),
- sub_sd))>;
+ (VMOVSDrr VR128:$src1, (COPY_TO_REGCLASS VR128:$src2, FR64))>;
// 256-bit variants
def : Pat<(v4i64 (X86Movsd VR256:$src1, VR256:$src2)),
(SUBREG_TO_REG (i32 0),
- (VMOVSDrr (EXTRACT_SUBREG (v4i64 VR256:$src1), sub_sd),
- (EXTRACT_SUBREG (v4i64 VR256:$src2), sub_sd)), sub_xmm)>;
+ (VMOVSDrr (EXTRACT_SUBREG (v4i64 VR256:$src1), sub_xmm),
+ (EXTRACT_SUBREG (v4i64 VR256:$src2), sub_xmm)),
+ sub_xmm)>;
def : Pat<(v4f64 (X86Movsd VR256:$src1, VR256:$src2)),
(SUBREG_TO_REG (i32 0),
- (VMOVSDrr (EXTRACT_SUBREG (v4f64 VR256:$src1), sub_sd),
- (EXTRACT_SUBREG (v4f64 VR256:$src2), sub_sd)), sub_xmm)>;
+ (VMOVSDrr (EXTRACT_SUBREG (v4f64 VR256:$src1), sub_xmm),
+ (EXTRACT_SUBREG (v4f64 VR256:$src2), sub_xmm)),
+ sub_xmm)>;
// FIXME: Instead of a X86Movlps there should be a X86Movsd here, the problem
// it has two uses through a bitcast. One use disappears at isel time and the
// fold opportunity reappears.
def : Pat<(v2f64 (X86Movlpd VR128:$src1, VR128:$src2)),
- (VMOVSDrr VR128:$src1, (EXTRACT_SUBREG (v2f64 VR128:$src2),
- sub_sd))>;
+ (VMOVSDrr VR128:$src1, (COPY_TO_REGCLASS VR128:$src2, FR64))>;
def : Pat<(v2i64 (X86Movlpd VR128:$src1, VR128:$src2)),
- (VMOVSDrr VR128:$src1, (EXTRACT_SUBREG (v2i64 VR128:$src2),
- sub_sd))>;
+ (VMOVSDrr VR128:$src1, (COPY_TO_REGCLASS VR128:$src2, FR64))>;
def : Pat<(v4f32 (X86Movlps VR128:$src1, VR128:$src2)),
- (VMOVSDrr VR128:$src1, (EXTRACT_SUBREG (v4f32 VR128:$src2),
- sub_sd))>;
+ (VMOVSDrr VR128:$src1, (COPY_TO_REGCLASS VR128:$src2, FR64))>;
def : Pat<(v4i32 (X86Movlps VR128:$src1, VR128:$src2)),
- (VMOVSDrr VR128:$src1, (EXTRACT_SUBREG (v4i32 VR128:$src2),
- sub_sd))>;
+ (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.
def : Pat<(v4f32 (X86vzmovl (v4f32 (scalar_to_vector FR32:$src)))),
(MOVSSrr (v4f32 (V_SET0)), FR32:$src)>;
def : Pat<(v4f32 (X86vzmovl (v4f32 VR128:$src))),
- (MOVSSrr (v4f32 (V_SET0)),
- (f32 (EXTRACT_SUBREG (v4f32 VR128:$src), sub_ss)))>;
+ (MOVSSrr (v4f32 (V_SET0)), (COPY_TO_REGCLASS VR128:$src, FR32))>;
def : Pat<(v4i32 (X86vzmovl (v4i32 VR128:$src))),
- (MOVSSrr (v4i32 (V_SET0)),
- (EXTRACT_SUBREG (v4i32 VR128:$src), sub_ss))>;
+ (MOVSSrr (v4i32 (V_SET0)), (COPY_TO_REGCLASS VR128:$src, FR32))>;
}
let AddedComplexity = 20 in {
- // MOVSSrm zeros the high parts of the register; represent this
- // with SUBREG_TO_REG.
+ // MOVSSrm already zeros the high parts of the register.
def : Pat<(v4f32 (X86vzmovl (v4f32 (scalar_to_vector (loadf32 addr:$src))))),
- (SUBREG_TO_REG (i32 0), (MOVSSrm addr:$src), sub_ss)>;
+ (COPY_TO_REGCLASS (MOVSSrm addr:$src), VR128)>;
def : Pat<(v4f32 (scalar_to_vector (loadf32 addr:$src))),
- (SUBREG_TO_REG (i32 0), (MOVSSrm addr:$src), sub_ss)>;
+ (COPY_TO_REGCLASS (MOVSSrm addr:$src), VR128)>;
def : Pat<(v4f32 (X86vzmovl (loadv4f32 addr:$src))),
- (SUBREG_TO_REG (i32 0), (MOVSSrm addr:$src), sub_ss)>;
+ (COPY_TO_REGCLASS (MOVSSrm addr:$src), VR128)>;
}
// Extract and store.
def : Pat<(store (f32 (vector_extract (v4f32 VR128:$src), (iPTR 0))),
addr:$dst),
- (MOVSSmr addr:$dst,
- (EXTRACT_SUBREG (v4f32 VR128:$src), sub_ss))>;
+ (MOVSSmr addr:$dst, (COPY_TO_REGCLASS VR128:$src, FR32))>;
// Shuffle with MOVSS
def : Pat<(v4i32 (X86Movss VR128:$src1, VR128:$src2)),
- (MOVSSrr (v4i32 VR128:$src1),
- (EXTRACT_SUBREG (v4i32 VR128:$src2), sub_ss))>;
+ (MOVSSrr VR128:$src1, (COPY_TO_REGCLASS VR128:$src2, FR32))>;
def : Pat<(v4f32 (X86Movss VR128:$src1, VR128:$src2)),
- (MOVSSrr (v4f32 VR128:$src1),
- (EXTRACT_SUBREG (v4f32 VR128:$src2), sub_ss))>;
+ (MOVSSrr VR128:$src1, (COPY_TO_REGCLASS VR128:$src2, FR32))>;
}
-let Predicates = [HasSSE2] in {
+let Predicates = [UseSSE2] in {
let AddedComplexity = 15 in {
// Move scalar to XMM zero-extended, zeroing a VR128 then do a
// MOVSD to the lower bits.
}
let AddedComplexity = 20 in {
- // MOVSDrm zeros the high parts of the register; represent this
- // with SUBREG_TO_REG.
+ // MOVSDrm already zeros the high parts of the register.
def : Pat<(v2f64 (X86vzmovl (v2f64 (scalar_to_vector (loadf64 addr:$src))))),
- (SUBREG_TO_REG (i64 0), (MOVSDrm addr:$src), sub_sd)>;
+ (COPY_TO_REGCLASS (MOVSDrm addr:$src), VR128)>;
def : Pat<(v2f64 (scalar_to_vector (loadf64 addr:$src))),
- (SUBREG_TO_REG (i64 0), (MOVSDrm addr:$src), sub_sd)>;
+ (COPY_TO_REGCLASS (MOVSDrm addr:$src), VR128)>;
def : Pat<(v2f64 (X86vzmovl (loadv2f64 addr:$src))),
- (SUBREG_TO_REG (i64 0), (MOVSDrm addr:$src), sub_sd)>;
+ (COPY_TO_REGCLASS (MOVSDrm addr:$src), VR128)>;
def : Pat<(v2f64 (X86vzmovl (bc_v2f64 (loadv4f32 addr:$src)))),
- (SUBREG_TO_REG (i64 0), (MOVSDrm addr:$src), sub_sd)>;
+ (COPY_TO_REGCLASS (MOVSDrm addr:$src), VR128)>;
def : Pat<(v2f64 (X86vzload addr:$src)),
- (SUBREG_TO_REG (i64 0), (MOVSDrm addr:$src), sub_sd)>;
+ (COPY_TO_REGCLASS (MOVSDrm addr:$src), VR128)>;
}
// Extract and store.
def : Pat<(store (f64 (vector_extract (v2f64 VR128:$src), (iPTR 0))),
addr:$dst),
- (MOVSDmr addr:$dst,
- (EXTRACT_SUBREG (v2f64 VR128:$src), sub_sd))>;
+ (MOVSDmr addr:$dst, (COPY_TO_REGCLASS VR128:$src, FR64))>;
// Shuffle with MOVSD
def : Pat<(v2i64 (X86Movsd VR128:$src1, VR128:$src2)),
- (MOVSDrr (v2i64 VR128:$src1),
- (EXTRACT_SUBREG (v2i64 VR128:$src2), sub_sd))>;
+ (MOVSDrr VR128:$src1, (COPY_TO_REGCLASS VR128:$src2, FR64))>;
def : Pat<(v2f64 (X86Movsd VR128:$src1, VR128:$src2)),
- (MOVSDrr (v2f64 VR128:$src1),
- (EXTRACT_SUBREG (v2f64 VR128:$src2), sub_sd))>;
+ (MOVSDrr VR128:$src1, (COPY_TO_REGCLASS VR128:$src2, FR64))>;
def : Pat<(v4f32 (X86Movsd VR128:$src1, VR128:$src2)),
- (MOVSDrr VR128:$src1, (EXTRACT_SUBREG (v4f32 VR128:$src2),sub_sd))>;
+ (MOVSDrr VR128:$src1, (COPY_TO_REGCLASS VR128:$src2, FR64))>;
def : Pat<(v4i32 (X86Movsd VR128:$src1, VR128:$src2)),
- (MOVSDrr VR128:$src1, (EXTRACT_SUBREG (v4i32 VR128:$src2),sub_sd))>;
+ (MOVSDrr VR128:$src1, (COPY_TO_REGCLASS VR128:$src2, FR64))>;
// FIXME: Instead of a X86Movlps there should be a X86Movsd here, the problem
// is during lowering, where it's not possible to recognize the fold cause
// it has two uses through a bitcast. One use disappears at isel time and the
// fold opportunity reappears.
def : Pat<(v2f64 (X86Movlpd VR128:$src1, VR128:$src2)),
- (MOVSDrr VR128:$src1, (EXTRACT_SUBREG (v2f64 VR128:$src2),sub_sd))>;
+ (MOVSDrr VR128:$src1, (COPY_TO_REGCLASS VR128:$src2, FR64))>;
def : Pat<(v2i64 (X86Movlpd VR128:$src1, VR128:$src2)),
- (MOVSDrr VR128:$src1, (EXTRACT_SUBREG (v2i64 VR128:$src2),sub_sd))>;
+ (MOVSDrr VR128:$src1, (COPY_TO_REGCLASS VR128:$src2, FR64))>;
def : Pat<(v4f32 (X86Movlps VR128:$src1, VR128:$src2)),
- (MOVSDrr VR128:$src1, (EXTRACT_SUBREG (v4f32 VR128:$src2),sub_sd))>;
+ (MOVSDrr VR128:$src1, (COPY_TO_REGCLASS VR128:$src2, FR64))>;
def : Pat<(v4i32 (X86Movlps VR128:$src1, VR128:$src2)),
- (MOVSDrr VR128:$src1, (EXTRACT_SUBREG (v4i32 VR128:$src2),sub_sd))>;
+ (MOVSDrr VR128:$src1, (COPY_TO_REGCLASS VR128:$src2, FR64))>;
}
//===----------------------------------------------------------------------===//
multiclass sse12_mov_packed<bits<8> opc, RegisterClass RC,
X86MemOperand x86memop, PatFrag ld_frag,
string asm, Domain d,
+ OpndItins itins,
bit IsReMaterializable = 1> {
let neverHasSideEffects = 1 in
def rr : PI<opc, MRMSrcReg, (outs RC:$dst), (ins RC:$src),
- !strconcat(asm, "\t{$src, $dst|$dst, $src}"), [], IIC_DEFAULT, d>;
+ !strconcat(asm, "\t{$src, $dst|$dst, $src}"), [], itins.rr, d>;
let canFoldAsLoad = 1, isReMaterializable = IsReMaterializable in
def rm : PI<opc, MRMSrcMem, (outs RC:$dst), (ins x86memop:$src),
!strconcat(asm, "\t{$src, $dst|$dst, $src}"),
- [(set RC:$dst, (ld_frag addr:$src))], IIC_DEFAULT, d>;
+ [(set RC:$dst, (ld_frag addr:$src))], itins.rm, d>;
}
defm VMOVAPS : sse12_mov_packed<0x28, VR128, f128mem, alignedloadv4f32,
- "movaps", SSEPackedSingle>, TB, VEX;
+ "movaps", SSEPackedSingle, SSE_MOVA_ITINS>,
+ TB, VEX;
defm VMOVAPD : sse12_mov_packed<0x28, VR128, f128mem, alignedloadv2f64,
- "movapd", SSEPackedDouble>, TB, OpSize, VEX;
+ "movapd", SSEPackedDouble, SSE_MOVA_ITINS>,
+ TB, OpSize, VEX;
defm VMOVUPS : sse12_mov_packed<0x10, VR128, f128mem, loadv4f32,
- "movups", SSEPackedSingle>, TB, VEX;
+ "movups", SSEPackedSingle, SSE_MOVU_ITINS>,
+ TB, VEX;
defm VMOVUPD : sse12_mov_packed<0x10, VR128, f128mem, loadv2f64,
- "movupd", SSEPackedDouble, 0>, TB, OpSize, VEX;
+ "movupd", SSEPackedDouble, SSE_MOVU_ITINS, 0>,
+ TB, OpSize, VEX;
defm VMOVAPSY : sse12_mov_packed<0x28, VR256, f256mem, alignedloadv8f32,
- "movaps", SSEPackedSingle>, TB, VEX;
+ "movaps", SSEPackedSingle, SSE_MOVA_ITINS>,
+ TB, VEX;
defm VMOVAPDY : sse12_mov_packed<0x28, VR256, f256mem, alignedloadv4f64,
- "movapd", SSEPackedDouble>, TB, OpSize, VEX;
+ "movapd", SSEPackedDouble, SSE_MOVA_ITINS>,
+ TB, OpSize, VEX;
defm VMOVUPSY : sse12_mov_packed<0x10, VR256, f256mem, loadv8f32,
- "movups", SSEPackedSingle>, TB, VEX;
+ "movups", SSEPackedSingle, SSE_MOVU_ITINS>,
+ TB, VEX;
defm VMOVUPDY : sse12_mov_packed<0x10, VR256, f256mem, loadv4f64,
- "movupd", SSEPackedDouble, 0>, TB, OpSize, VEX;
+ "movupd", SSEPackedDouble, SSE_MOVU_ITINS, 0>,
+ TB, OpSize, VEX;
defm MOVAPS : sse12_mov_packed<0x28, VR128, f128mem, alignedloadv4f32,
- "movaps", SSEPackedSingle>, TB;
+ "movaps", SSEPackedSingle, SSE_MOVA_ITINS>,
+ TB;
defm MOVAPD : sse12_mov_packed<0x28, VR128, f128mem, alignedloadv2f64,
- "movapd", SSEPackedDouble>, TB, OpSize;
+ "movapd", SSEPackedDouble, SSE_MOVA_ITINS>,
+ TB, OpSize;
defm MOVUPS : sse12_mov_packed<0x10, VR128, f128mem, loadv4f32,
- "movups", SSEPackedSingle>, TB;
+ "movups", SSEPackedSingle, SSE_MOVU_ITINS>,
+ TB;
defm MOVUPD : sse12_mov_packed<0x10, VR128, f128mem, loadv2f64,
- "movupd", SSEPackedDouble, 0>, TB, OpSize;
+ "movupd", SSEPackedDouble, SSE_MOVU_ITINS, 0>,
+ TB, OpSize;
def VMOVAPSmr : VPSI<0x29, MRMDestMem, (outs), (ins f128mem:$dst, VR128:$src),
"movaps\t{$src, $dst|$dst, $src}",
- [(alignedstore (v4f32 VR128:$src), addr:$dst)]>, VEX;
+ [(alignedstore (v4f32 VR128:$src), addr:$dst)],
+ IIC_SSE_MOVA_P_MR>, VEX;
def VMOVAPDmr : VPDI<0x29, MRMDestMem, (outs), (ins f128mem:$dst, VR128:$src),
"movapd\t{$src, $dst|$dst, $src}",
- [(alignedstore (v2f64 VR128:$src), addr:$dst)]>, VEX;
+ [(alignedstore (v2f64 VR128:$src), addr:$dst)],
+ IIC_SSE_MOVA_P_MR>, VEX;
def VMOVUPSmr : VPSI<0x11, MRMDestMem, (outs), (ins f128mem:$dst, VR128:$src),
"movups\t{$src, $dst|$dst, $src}",
- [(store (v4f32 VR128:$src), addr:$dst)]>, VEX;
+ [(store (v4f32 VR128:$src), addr:$dst)],
+ IIC_SSE_MOVU_P_MR>, VEX;
def VMOVUPDmr : VPDI<0x11, MRMDestMem, (outs), (ins f128mem:$dst, VR128:$src),
"movupd\t{$src, $dst|$dst, $src}",
- [(store (v2f64 VR128:$src), addr:$dst)]>, VEX;
+ [(store (v2f64 VR128:$src), addr:$dst)],
+ IIC_SSE_MOVU_P_MR>, VEX;
def VMOVAPSYmr : VPSI<0x29, MRMDestMem, (outs), (ins f256mem:$dst, VR256:$src),
"movaps\t{$src, $dst|$dst, $src}",
- [(alignedstore256 (v8f32 VR256:$src), addr:$dst)]>, VEX;
+ [(alignedstore256 (v8f32 VR256:$src), addr:$dst)],
+ IIC_SSE_MOVA_P_MR>, VEX;
def VMOVAPDYmr : VPDI<0x29, MRMDestMem, (outs), (ins f256mem:$dst, VR256:$src),
"movapd\t{$src, $dst|$dst, $src}",
- [(alignedstore256 (v4f64 VR256:$src), addr:$dst)]>, VEX;
+ [(alignedstore256 (v4f64 VR256:$src), addr:$dst)],
+ IIC_SSE_MOVA_P_MR>, VEX;
def VMOVUPSYmr : VPSI<0x11, MRMDestMem, (outs), (ins f256mem:$dst, VR256:$src),
"movups\t{$src, $dst|$dst, $src}",
- [(store (v8f32 VR256:$src), addr:$dst)]>, VEX;
+ [(store (v8f32 VR256:$src), addr:$dst)],
+ IIC_SSE_MOVU_P_MR>, VEX;
def VMOVUPDYmr : VPDI<0x11, MRMDestMem, (outs), (ins f256mem:$dst, VR256:$src),
"movupd\t{$src, $dst|$dst, $src}",
- [(store (v4f64 VR256:$src), addr:$dst)]>, VEX;
+ [(store (v4f64 VR256:$src), addr:$dst)],
+ IIC_SSE_MOVU_P_MR>, VEX;
// For disassembler
let isCodeGenOnly = 1 in {
def VMOVAPSrr_REV : VPSI<0x29, MRMDestReg, (outs VR128:$dst),
(ins VR128:$src),
- "movaps\t{$src, $dst|$dst, $src}", []>, VEX;
+ "movaps\t{$src, $dst|$dst, $src}", [],
+ IIC_SSE_MOVA_P_RR>, VEX;
def VMOVAPDrr_REV : VPDI<0x29, MRMDestReg, (outs VR128:$dst),
(ins VR128:$src),
- "movapd\t{$src, $dst|$dst, $src}", []>, VEX;
+ "movapd\t{$src, $dst|$dst, $src}", [],
+ IIC_SSE_MOVA_P_RR>, VEX;
def VMOVUPSrr_REV : VPSI<0x11, MRMDestReg, (outs VR128:$dst),
(ins VR128:$src),
- "movups\t{$src, $dst|$dst, $src}", []>, VEX;
+ "movups\t{$src, $dst|$dst, $src}", [],
+ IIC_SSE_MOVU_P_RR>, VEX;
def VMOVUPDrr_REV : VPDI<0x11, MRMDestReg, (outs VR128:$dst),
(ins VR128:$src),
- "movupd\t{$src, $dst|$dst, $src}", []>, VEX;
+ "movupd\t{$src, $dst|$dst, $src}", [],
+ IIC_SSE_MOVU_P_RR>, VEX;
def VMOVAPSYrr_REV : VPSI<0x29, MRMDestReg, (outs VR256:$dst),
(ins VR256:$src),
- "movaps\t{$src, $dst|$dst, $src}", []>, VEX;
+ "movaps\t{$src, $dst|$dst, $src}", [],
+ IIC_SSE_MOVA_P_RR>, VEX;
def VMOVAPDYrr_REV : VPDI<0x29, MRMDestReg, (outs VR256:$dst),
(ins VR256:$src),
- "movapd\t{$src, $dst|$dst, $src}", []>, VEX;
+ "movapd\t{$src, $dst|$dst, $src}", [],
+ IIC_SSE_MOVA_P_RR>, VEX;
def VMOVUPSYrr_REV : VPSI<0x11, MRMDestReg, (outs VR256:$dst),
(ins VR256:$src),
- "movups\t{$src, $dst|$dst, $src}", []>, VEX;
+ "movups\t{$src, $dst|$dst, $src}", [],
+ IIC_SSE_MOVU_P_RR>, VEX;
def VMOVUPDYrr_REV : VPDI<0x11, MRMDestReg, (outs VR256:$dst),
(ins VR256:$src),
- "movupd\t{$src, $dst|$dst, $src}", []>, VEX;
+ "movupd\t{$src, $dst|$dst, $src}", [],
+ IIC_SSE_MOVU_P_RR>, VEX;
}
let Predicates = [HasAVX] in {
def MOVAPSmr : PSI<0x29, MRMDestMem, (outs), (ins f128mem:$dst, VR128:$src),
"movaps\t{$src, $dst|$dst, $src}",
- [(alignedstore (v4f32 VR128:$src), addr:$dst)]>;
+ [(alignedstore (v4f32 VR128:$src), addr:$dst)],
+ IIC_SSE_MOVA_P_MR>;
def MOVAPDmr : PDI<0x29, MRMDestMem, (outs), (ins f128mem:$dst, VR128:$src),
"movapd\t{$src, $dst|$dst, $src}",
- [(alignedstore (v2f64 VR128:$src), addr:$dst)]>;
+ [(alignedstore (v2f64 VR128:$src), addr:$dst)],
+ IIC_SSE_MOVA_P_MR>;
def MOVUPSmr : PSI<0x11, MRMDestMem, (outs), (ins f128mem:$dst, VR128:$src),
"movups\t{$src, $dst|$dst, $src}",
- [(store (v4f32 VR128:$src), addr:$dst)]>;
+ [(store (v4f32 VR128:$src), addr:$dst)],
+ IIC_SSE_MOVU_P_MR>;
def MOVUPDmr : PDI<0x11, MRMDestMem, (outs), (ins f128mem:$dst, VR128:$src),
"movupd\t{$src, $dst|$dst, $src}",
- [(store (v2f64 VR128:$src), addr:$dst)]>;
+ [(store (v2f64 VR128:$src), addr:$dst)],
+ IIC_SSE_MOVU_P_MR>;
// For disassembler
let isCodeGenOnly = 1 in {
def MOVAPSrr_REV : PSI<0x29, MRMDestReg, (outs VR128:$dst), (ins VR128:$src),
- "movaps\t{$src, $dst|$dst, $src}", []>;
+ "movaps\t{$src, $dst|$dst, $src}", [],
+ IIC_SSE_MOVA_P_RR>;
def MOVAPDrr_REV : PDI<0x29, MRMDestReg, (outs VR128:$dst), (ins VR128:$src),
- "movapd\t{$src, $dst|$dst, $src}", []>;
+ "movapd\t{$src, $dst|$dst, $src}", [],
+ IIC_SSE_MOVA_P_RR>;
def MOVUPSrr_REV : PSI<0x11, MRMDestReg, (outs VR128:$dst), (ins VR128:$src),
- "movups\t{$src, $dst|$dst, $src}", []>;
+ "movups\t{$src, $dst|$dst, $src}", [],
+ IIC_SSE_MOVU_P_RR>;
def MOVUPDrr_REV : PDI<0x11, MRMDestReg, (outs VR128:$dst), (ins VR128:$src),
- "movupd\t{$src, $dst|$dst, $src}", []>;
+ "movupd\t{$src, $dst|$dst, $src}", [],
+ IIC_SSE_MOVU_P_RR>;
}
let Predicates = [HasAVX] in {
(VMOVUPDmr addr:$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), (i32 0))), addr:$dst),
+ (VMOVAPDmr addr:$dst, (v2f64 (EXTRACT_SUBREG VR256:$src,sub_xmm)))>;
+ def : Pat<(alignedstore (v4f32 (extract_subvector
+ (v8f32 VR256:$src), (i32 0))), addr:$dst),
+ (VMOVAPSmr addr:$dst, (v4f32 (EXTRACT_SUBREG VR256:$src,sub_xmm)))>;
+ def : Pat<(alignedstore (v2i64 (extract_subvector
+ (v4i64 VR256:$src), (i32 0))), addr:$dst),
+ (VMOVAPDmr addr:$dst, (v2i64 (EXTRACT_SUBREG VR256:$src,sub_xmm)))>;
+ def : Pat<(alignedstore (v4i32 (extract_subvector
+ (v8i32 VR256:$src), (i32 0))), addr:$dst),
+ (VMOVAPSmr addr:$dst, (v4i32 (EXTRACT_SUBREG VR256:$src,sub_xmm)))>;
+ def : Pat<(alignedstore (v8i16 (extract_subvector
+ (v16i16 VR256:$src), (i32 0))), addr:$dst),
+ (VMOVAPSmr addr:$dst, (v8i16 (EXTRACT_SUBREG VR256:$src,sub_xmm)))>;
+ def : Pat<(alignedstore (v16i8 (extract_subvector
+ (v32i8 VR256:$src), (i32 0))), addr:$dst),
+ (VMOVAPSmr addr:$dst, (v16i8 (EXTRACT_SUBREG VR256:$src,sub_xmm)))>;
+
+ def : Pat<(store (v2f64 (extract_subvector
+ (v4f64 VR256:$src), (i32 0))), addr:$dst),
+ (VMOVUPDmr addr:$dst, (v2f64 (EXTRACT_SUBREG VR256:$src,sub_xmm)))>;
+ def : Pat<(store (v4f32 (extract_subvector
+ (v8f32 VR256:$src), (i32 0))), addr:$dst),
+ (VMOVUPSmr addr:$dst, (v4f32 (EXTRACT_SUBREG VR256:$src,sub_xmm)))>;
+ def : Pat<(store (v2i64 (extract_subvector
+ (v4i64 VR256:$src), (i32 0))), addr:$dst),
+ (VMOVUPDmr addr:$dst, (v2i64 (EXTRACT_SUBREG VR256:$src,sub_xmm)))>;
+ def : Pat<(store (v4i32 (extract_subvector
+ (v8i32 VR256:$src), (i32 0))), addr:$dst),
+ (VMOVUPSmr addr:$dst, (v4i32 (EXTRACT_SUBREG VR256:$src,sub_xmm)))>;
+ def : Pat<(store (v8i16 (extract_subvector
+ (v16i16 VR256:$src), (i32 0))), addr:$dst),
+ (VMOVAPSmr addr:$dst, (v8i16 (EXTRACT_SUBREG VR256:$src,sub_xmm)))>;
+ def : Pat<(store (v16i8 (extract_subvector
+ (v32i8 VR256:$src), (i32 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),
// bits are disregarded. FIXME: Set encoding to pseudo!
let neverHasSideEffects = 1 in {
def FsVMOVAPSrr : VPSI<0x28, MRMSrcReg, (outs FR32:$dst), (ins FR32:$src),
- "movaps\t{$src, $dst|$dst, $src}", []>, VEX;
+ "movaps\t{$src, $dst|$dst, $src}", [],
+ IIC_SSE_MOVA_P_RR>, VEX;
def FsVMOVAPDrr : VPDI<0x28, MRMSrcReg, (outs FR64:$dst), (ins FR64:$src),
- "movapd\t{$src, $dst|$dst, $src}", []>, VEX;
+ "movapd\t{$src, $dst|$dst, $src}", [],
+ IIC_SSE_MOVA_P_RR>, VEX;
def FsMOVAPSrr : PSI<0x28, MRMSrcReg, (outs FR32:$dst), (ins FR32:$src),
- "movaps\t{$src, $dst|$dst, $src}", []>;
+ "movaps\t{$src, $dst|$dst, $src}", [],
+ IIC_SSE_MOVA_P_RR>;
def FsMOVAPDrr : PDI<0x28, MRMSrcReg, (outs FR64:$dst), (ins FR64:$src),
- "movapd\t{$src, $dst|$dst, $src}", []>;
+ "movapd\t{$src, $dst|$dst, $src}", [],
+ IIC_SSE_MOVA_P_RR>;
}
// Alias instruction to load FR32 or FR64 from f128mem using movaps. Upper
let isCodeGenOnly = 1 in {
def FsVMOVAPSrm : VPSI<0x28, MRMSrcMem, (outs FR32:$dst), (ins f128mem:$src),
"movaps\t{$src, $dst|$dst, $src}",
- [(set FR32:$dst, (alignedloadfsf32 addr:$src))]>, VEX;
+ [(set FR32:$dst, (alignedloadfsf32 addr:$src))],
+ IIC_SSE_MOVA_P_RM>, VEX;
def FsVMOVAPDrm : VPDI<0x28, MRMSrcMem, (outs FR64:$dst), (ins f128mem:$src),
"movapd\t{$src, $dst|$dst, $src}",
- [(set FR64:$dst, (alignedloadfsf64 addr:$src))]>, VEX;
+ [(set FR64:$dst, (alignedloadfsf64 addr:$src))],
+ IIC_SSE_MOVA_P_RM>, VEX;
}
def FsMOVAPSrm : PSI<0x28, MRMSrcMem, (outs FR32:$dst), (ins f128mem:$src),
"movaps\t{$src, $dst|$dst, $src}",
- [(set FR32:$dst, (alignedloadfsf32 addr:$src))]>;
+ [(set FR32:$dst, (alignedloadfsf32 addr:$src))],
+ IIC_SSE_MOVA_P_RM>;
def FsMOVAPDrm : PDI<0x28, MRMSrcMem, (outs FR64:$dst), (ins f128mem:$src),
"movapd\t{$src, $dst|$dst, $src}",
- [(set FR64:$dst, (alignedloadfsf64 addr:$src))]>;
+ [(set FR64:$dst, (alignedloadfsf64 addr:$src))],
+ IIC_SSE_MOVA_P_RM>;
}
//===----------------------------------------------------------------------===//
multiclass sse12_mov_hilo_packed<bits<8>opc, RegisterClass RC,
SDNode psnode, SDNode pdnode, string base_opc,
- string asm_opr> {
+ string asm_opr, InstrItinClass itin> {
def PSrm : PI<opc, MRMSrcMem,
(outs VR128:$dst), (ins VR128:$src1, f64mem:$src2),
!strconcat(base_opc, "s", asm_opr),
[(set RC:$dst,
(psnode RC:$src1,
(bc_v4f32 (v2f64 (scalar_to_vector (loadf64 addr:$src2))))))],
- IIC_DEFAULT, SSEPackedSingle>, TB;
+ itin, SSEPackedSingle>, TB;
def PDrm : PI<opc, MRMSrcMem,
(outs RC:$dst), (ins RC:$src1, f64mem:$src2),
!strconcat(base_opc, "d", asm_opr),
[(set RC:$dst, (v2f64 (pdnode RC:$src1,
(scalar_to_vector (loadf64 addr:$src2)))))],
- IIC_DEFAULT, SSEPackedDouble>, TB, OpSize;
+ itin, SSEPackedDouble>, TB, OpSize;
}
let AddedComplexity = 20 in {
defm VMOVL : sse12_mov_hilo_packed<0x12, VR128, X86Movlps, X86Movlpd, "movlp",
- "\t{$src2, $src1, $dst|$dst, $src1, $src2}">, VEX_4V;
+ "\t{$src2, $src1, $dst|$dst, $src1, $src2}",
+ IIC_SSE_MOV_LH>, VEX_4V;
}
let Constraints = "$src1 = $dst", AddedComplexity = 20 in {
defm MOVL : sse12_mov_hilo_packed<0x12, VR128, X86Movlps, X86Movlpd, "movlp",
- "\t{$src2, $dst|$dst, $src2}">;
+ "\t{$src2, $dst|$dst, $src2}",
+ IIC_SSE_MOV_LH>;
}
def VMOVLPSmr : VPSI<0x13, MRMDestMem, (outs), (ins f64mem:$dst, VR128:$src),
"movlps\t{$src, $dst|$dst, $src}",
[(store (f64 (vector_extract (bc_v2f64 (v4f32 VR128:$src)),
- (iPTR 0))), addr:$dst)]>, VEX;
+ (iPTR 0))), addr:$dst)],
+ IIC_SSE_MOV_LH>, VEX;
def VMOVLPDmr : VPDI<0x13, MRMDestMem, (outs), (ins f64mem:$dst, VR128:$src),
"movlpd\t{$src, $dst|$dst, $src}",
[(store (f64 (vector_extract (v2f64 VR128:$src),
- (iPTR 0))), addr:$dst)]>, VEX;
+ (iPTR 0))), addr:$dst)],
+ IIC_SSE_MOV_LH>, VEX;
def MOVLPSmr : PSI<0x13, MRMDestMem, (outs), (ins f64mem:$dst, VR128:$src),
"movlps\t{$src, $dst|$dst, $src}",
[(store (f64 (vector_extract (bc_v2f64 (v4f32 VR128:$src)),
- (iPTR 0))), addr:$dst)]>;
+ (iPTR 0))), addr:$dst)],
+ IIC_SSE_MOV_LH>;
def MOVLPDmr : PDI<0x13, MRMDestMem, (outs), (ins f64mem:$dst, VR128:$src),
"movlpd\t{$src, $dst|$dst, $src}",
[(store (f64 (vector_extract (v2f64 VR128:$src),
- (iPTR 0))), addr:$dst)]>;
+ (iPTR 0))), addr:$dst)],
+ IIC_SSE_MOV_LH>;
let Predicates = [HasAVX] in {
// Shuffle with VMOVLPS
(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, $src2}">, VEX_4V;
+ "\t{$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, $src2}">;
+ "\t{$src2, $dst|$dst, $src2}",
+ IIC_SSE_MOV_LH>;
}
// v2f64 extract element 1 is always custom lowered to unpack high to low
[(store (f64 (vector_extract
(X86Unpckh (bc_v2f64 (v4f32 VR128:$src)),
(bc_v2f64 (v4f32 VR128:$src))),
- (iPTR 0))), addr:$dst)]>, VEX;
+ (iPTR 0))), addr:$dst)], IIC_SSE_MOV_LH>, VEX;
def VMOVHPDmr : VPDI<0x17, MRMDestMem, (outs), (ins f64mem:$dst, VR128:$src),
"movhpd\t{$src, $dst|$dst, $src}",
[(store (f64 (vector_extract
(v2f64 (X86Unpckh VR128:$src, VR128:$src)),
- (iPTR 0))), addr:$dst)]>, VEX;
+ (iPTR 0))), addr:$dst)], IIC_SSE_MOV_LH>, VEX;
def MOVHPSmr : PSI<0x17, MRMDestMem, (outs), (ins f64mem:$dst, VR128:$src),
"movhps\t{$src, $dst|$dst, $src}",
[(store (f64 (vector_extract
(X86Unpckh (bc_v2f64 (v4f32 VR128:$src)),
(bc_v2f64 (v4f32 VR128:$src))),
- (iPTR 0))), addr:$dst)]>;
+ (iPTR 0))), addr:$dst)], IIC_SSE_MOV_LH>;
def MOVHPDmr : PDI<0x17, MRMDestMem, (outs), (ins f64mem:$dst, VR128:$src),
"movhpd\t{$src, $dst|$dst, $src}",
[(store (f64 (vector_extract
(v2f64 (X86Unpckh VR128:$src, VR128:$src)),
- (iPTR 0))), addr:$dst)]>;
+ (iPTR 0))), addr:$dst)], IIC_SSE_MOV_LH>;
let Predicates = [HasAVX] in {
// VMOVHPS patterns
(VMOVHPDrm VR128:$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:$src2)>;
}
-let Predicates = [HasSSE2] in {
+let Predicates = [UseSSE2] in {
// FIXME: Instead of X86Unpckl, there should be a X86Movlhpd here, the problem
// is during lowering, where it's not possible to recognize the load fold
// cause it has two uses through a bitcast. One use disappears at isel time
(ins VR128:$src1, VR128:$src2),
"movlhps\t{$src2, $src1, $dst|$dst, $src1, $src2}",
[(set VR128:$dst,
- (v4f32 (X86Movlhps VR128:$src1, VR128:$src2)))]>,
+ (v4f32 (X86Movlhps VR128:$src1, VR128:$src2)))],
+ IIC_SSE_MOV_LH>,
VEX_4V;
def VMOVHLPSrr : VPSI<0x12, MRMSrcReg, (outs VR128:$dst),
(ins VR128:$src1, VR128:$src2),
"movhlps\t{$src2, $src1, $dst|$dst, $src1, $src2}",
[(set VR128:$dst,
- (v4f32 (X86Movhlps VR128:$src1, VR128:$src2)))]>,
+ (v4f32 (X86Movhlps VR128:$src1, VR128:$src2)))],
+ IIC_SSE_MOV_LH>,
VEX_4V;
}
let Constraints = "$src1 = $dst", AddedComplexity = 20 in {
(ins VR128:$src1, VR128:$src2),
"movlhps\t{$src2, $dst|$dst, $src2}",
[(set VR128:$dst,
- (v4f32 (X86Movlhps VR128:$src1, VR128:$src2)))]>;
+ (v4f32 (X86Movlhps VR128:$src1, VR128:$src2)))],
+ IIC_SSE_MOV_LH>;
def MOVHLPSrr : PSI<0x12, MRMSrcReg, (outs VR128:$dst),
(ins VR128:$src1, VR128:$src2),
"movhlps\t{$src2, $dst|$dst, $src2}",
[(set VR128:$dst,
- (v4f32 (X86Movhlps VR128:$src1, VR128:$src2)))]>;
+ (v4f32 (X86Movhlps VR128:$src1, VR128:$src2)))],
+ IIC_SSE_MOV_LH>;
}
let Predicates = [HasAVX] in {
(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)>;
// SSE 1 & 2 - Conversion Instructions
//===----------------------------------------------------------------------===//
+def SSE_CVT_PD : OpndItins<
+ IIC_SSE_CVT_PD_RR, IIC_SSE_CVT_PD_RM
+>;
+
+def SSE_CVT_PS : OpndItins<
+ IIC_SSE_CVT_PS_RR, IIC_SSE_CVT_PS_RM
+>;
+
+def SSE_CVT_Scalar : OpndItins<
+ IIC_SSE_CVT_Scalar_RR, IIC_SSE_CVT_Scalar_RM
+>;
+
+def SSE_CVT_SS2SI_32 : OpndItins<
+ IIC_SSE_CVT_SS2SI32_RR, IIC_SSE_CVT_SS2SI32_RM
+>;
+
+def SSE_CVT_SS2SI_64 : OpndItins<
+ IIC_SSE_CVT_SS2SI64_RR, IIC_SSE_CVT_SS2SI64_RM
+>;
+
+def SSE_CVT_SD2SI : OpndItins<
+ IIC_SSE_CVT_SD2SI_RR, IIC_SSE_CVT_SD2SI_RM
+>;
+
multiclass sse12_cvt_s<bits<8> opc, RegisterClass SrcRC, RegisterClass DstRC,
SDNode OpNode, X86MemOperand x86memop, PatFrag ld_frag,
- string asm> {
+ string asm, OpndItins itins> {
def rr : SI<opc, MRMSrcReg, (outs DstRC:$dst), (ins SrcRC:$src), asm,
- [(set DstRC:$dst, (OpNode SrcRC:$src))]>;
+ [(set DstRC:$dst, (OpNode SrcRC:$src))],
+ itins.rr>;
def rm : SI<opc, MRMSrcMem, (outs DstRC:$dst), (ins x86memop:$src), asm,
- [(set DstRC:$dst, (OpNode (ld_frag addr:$src)))]>;
+ [(set DstRC:$dst, (OpNode (ld_frag addr:$src)))],
+ itins.rm>;
}
multiclass sse12_cvt_p<bits<8> opc, RegisterClass SrcRC, RegisterClass DstRC,
- SDNode OpNode, X86MemOperand x86memop, PatFrag ld_frag,
- string asm, Domain d> {
- def rr : PI<opc, MRMSrcReg, (outs DstRC:$dst), (ins SrcRC:$src), asm,
- [(set DstRC:$dst, (OpNode SrcRC:$src))],
- IIC_DEFAULT, d>;
- def rm : PI<opc, MRMSrcMem, (outs DstRC:$dst), (ins x86memop:$src), asm,
- [(set DstRC:$dst, (OpNode (ld_frag addr:$src)))],
- IIC_DEFAULT, d>;
+ X86MemOperand x86memop, string asm, Domain d,
+ OpndItins itins> {
+let neverHasSideEffects = 1 in {
+ def rr : I<opc, MRMSrcReg, (outs DstRC:$dst), (ins SrcRC:$src), asm,
+ [], itins.rr, d>;
+ let mayLoad = 1 in
+ def rm : I<opc, MRMSrcMem, (outs DstRC:$dst), (ins x86memop:$src), asm,
+ [], itins.rm, d>;
+}
}
multiclass sse12_vcvt_avx<bits<8> opc, RegisterClass SrcRC, RegisterClass DstRC,
}
defm VCVTTSS2SI : sse12_cvt_s<0x2C, FR32, GR32, fp_to_sint, f32mem, loadf32,
- "cvttss2si\t{$src, $dst|$dst, $src}">, XS, VEX,
- VEX_LIG;
+ "cvttss2si\t{$src, $dst|$dst, $src}",
+ SSE_CVT_SS2SI_32>,
+ XS, VEX, VEX_LIG;
defm VCVTTSS2SI64 : sse12_cvt_s<0x2C, FR32, GR64, fp_to_sint, f32mem, loadf32,
- "cvttss2si\t{$src, $dst|$dst, $src}">, XS, VEX,
- VEX_W, VEX_LIG;
+ "cvttss2si{q}\t{$src, $dst|$dst, $src}",
+ SSE_CVT_SS2SI_64>,
+ XS, VEX, VEX_W, VEX_LIG;
defm VCVTTSD2SI : sse12_cvt_s<0x2C, FR64, GR32, fp_to_sint, f64mem, loadf64,
- "cvttsd2si\t{$src, $dst|$dst, $src}">, XD, VEX,
- VEX_LIG;
+ "cvttsd2si\t{$src, $dst|$dst, $src}",
+ SSE_CVT_SD2SI>,
+ XD, VEX, VEX_LIG;
defm VCVTTSD2SI64 : sse12_cvt_s<0x2C, FR64, GR64, fp_to_sint, f64mem, loadf64,
- "cvttsd2si\t{$src, $dst|$dst, $src}">, XD,
- VEX, VEX_W, VEX_LIG;
+ "cvttsd2si{q}\t{$src, $dst|$dst, $src}",
+ SSE_CVT_SD2SI>,
+ XD, VEX, VEX_W, VEX_LIG;
// The assembler can recognize rr 64-bit instructions by seeing a rxx
// register, but the same isn't true when only using memory operands,
// provide other assembly "l" and "q" forms to address this explicitly
// where appropriate to do so.
-defm VCVTSI2SS : sse12_vcvt_avx<0x2A, GR32, FR32, i32mem, "cvtsi2ss">, XS,
- VEX_4V, VEX_LIG;
-defm VCVTSI2SS64 : sse12_vcvt_avx<0x2A, GR64, FR32, i64mem, "cvtsi2ss{q}">, XS,
- VEX_4V, VEX_W, VEX_LIG;
-defm VCVTSI2SD : sse12_vcvt_avx<0x2A, GR32, FR64, i32mem, "cvtsi2sd">, XD,
- VEX_4V, VEX_LIG;
-defm VCVTSI2SDL : sse12_vcvt_avx<0x2A, GR32, FR64, i32mem, "cvtsi2sd{l}">, XD,
- VEX_4V, VEX_LIG;
-defm VCVTSI2SD64 : sse12_vcvt_avx<0x2A, GR64, FR64, i64mem, "cvtsi2sd{q}">, XD,
- VEX_4V, VEX_W, VEX_LIG;
-
-let Predicates = [HasAVX], AddedComplexity = 1 in {
+defm VCVTSI2SS : sse12_vcvt_avx<0x2A, GR32, FR32, i32mem, "cvtsi2ss">,
+ XS, VEX_4V, VEX_LIG;
+defm VCVTSI2SS64 : sse12_vcvt_avx<0x2A, GR64, FR32, i64mem, "cvtsi2ss{q}">,
+ XS, VEX_4V, VEX_W, VEX_LIG;
+defm VCVTSI2SD : sse12_vcvt_avx<0x2A, GR32, FR64, i32mem, "cvtsi2sd">,
+ XD, VEX_4V, VEX_LIG;
+defm VCVTSI2SD64 : sse12_vcvt_avx<0x2A, GR64, FR64, i64mem, "cvtsi2sd{q}">,
+ XD, VEX_4V, VEX_W, VEX_LIG;
+
+def : InstAlias<"vcvtsi2sd{l}\t{$src, $src1, $dst|$dst, $src1, $src}",
+ (VCVTSI2SDrr FR64:$dst, FR64:$src1, GR32:$src)>;
+def : InstAlias<"vcvtsi2sd{l}\t{$src, $src1, $dst|$dst, $src1, $src}",
+ (VCVTSI2SDrm FR64:$dst, FR64:$src1, i32mem:$src)>;
+
+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:$src))),
}
defm CVTTSS2SI : sse12_cvt_s<0x2C, FR32, GR32, fp_to_sint, f32mem, loadf32,
- "cvttss2si\t{$src, $dst|$dst, $src}">, XS;
+ "cvttss2si\t{$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, $src}">, XS, REX_W;
+ "cvttss2si{q}\t{$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, $src}">, XD;
+ "cvttsd2si\t{$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, $src}">, XD, REX_W;
+ "cvttsd2si{q}\t{$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, $src}">, XS;
+ "cvtsi2ss\t{$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, $src}">, XS, REX_W;
+ "cvtsi2ss{q}\t{$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, $src}">, XD;
+ "cvtsi2sd\t{$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, $src}">, XD, REX_W;
+ "cvtsi2sd{q}\t{$src, $dst|$dst, $src}",
+ SSE_CVT_Scalar>, XD, REX_W;
// Conversion Instructions Intrinsics - Match intrinsics which expect MM
// and/or XMM operand(s).
multiclass sse12_cvt_sint<bits<8> opc, RegisterClass SrcRC, RegisterClass DstRC,
- Intrinsic Int, X86MemOperand x86memop, PatFrag ld_frag,
- string asm> {
+ Intrinsic Int, Operand memop, ComplexPattern mem_cpat,
+ string asm, OpndItins itins> {
def rr : SI<opc, MRMSrcReg, (outs DstRC:$dst), (ins SrcRC:$src),
!strconcat(asm, "\t{$src, $dst|$dst, $src}"),
- [(set DstRC:$dst, (Int SrcRC:$src))]>;
- def rm : SI<opc, MRMSrcMem, (outs DstRC:$dst), (ins x86memop:$src),
+ [(set DstRC:$dst, (Int SrcRC:$src))], itins.rr>;
+ def rm : SI<opc, MRMSrcMem, (outs DstRC:$dst), (ins memop:$src),
!strconcat(asm, "\t{$src, $dst|$dst, $src}"),
- [(set DstRC:$dst, (Int (ld_frag addr:$src)))]>;
+ [(set DstRC:$dst, (Int mem_cpat:$src))], itins.rm>;
}
multiclass sse12_cvt_sint_3addr<bits<8> opc, RegisterClass SrcRC,
RegisterClass DstRC, Intrinsic Int, X86MemOperand x86memop,
- PatFrag ld_frag, string asm, bit Is2Addr = 1> {
+ PatFrag ld_frag, string asm, OpndItins itins,
+ bit Is2Addr = 1> {
def rr : SI<opc, MRMSrcReg, (outs DstRC:$dst), (ins DstRC:$src1, SrcRC:$src2),
!if(Is2Addr,
!strconcat(asm, "\t{$src2, $dst|$dst, $src2}"),
!strconcat(asm, "\t{$src2, $src1, $dst|$dst, $src1, $src2}")),
- [(set DstRC:$dst, (Int DstRC:$src1, SrcRC:$src2))]>;
+ [(set DstRC:$dst, (Int DstRC:$src1, SrcRC:$src2))],
+ itins.rr>;
def rm : SI<opc, MRMSrcMem, (outs DstRC:$dst),
(ins DstRC:$src1, x86memop:$src2),
!if(Is2Addr,
!strconcat(asm, "\t{$src2, $dst|$dst, $src2}"),
!strconcat(asm, "\t{$src2, $src1, $dst|$dst, $src1, $src2}")),
- [(set DstRC:$dst, (Int DstRC:$src1, (ld_frag addr:$src2)))]>;
+ [(set DstRC:$dst, (Int DstRC:$src1, (ld_frag addr:$src2)))],
+ itins.rm>;
}
-defm VCVTSD2SI : sse12_cvt_sint<0x2D, VR128, GR32, int_x86_sse2_cvtsd2si,
- f128mem, load, "cvtsd2si">, XD, VEX, VEX_LIG;
+defm VCVTSD2SI : sse12_cvt_sint<0x2D, VR128, GR32,
+ int_x86_sse2_cvtsd2si, sdmem, sse_load_f64, "cvtsd2si{l}",
+ SSE_CVT_SD2SI>, XD, VEX, VEX_LIG;
defm VCVTSD2SI64 : sse12_cvt_sint<0x2D, VR128, GR64,
- int_x86_sse2_cvtsd2si64, f128mem, load, "cvtsd2si">,
- XD, VEX, VEX_W, VEX_LIG;
+ int_x86_sse2_cvtsd2si64, sdmem, sse_load_f64, "cvtsd2si{q}",
+ SSE_CVT_SD2SI>, XD, VEX, VEX_W, VEX_LIG;
defm CVTSD2SI : sse12_cvt_sint<0x2D, VR128, GR32, int_x86_sse2_cvtsd2si,
- f128mem, load, "cvtsd2si{l}">, XD;
+ sdmem, sse_load_f64, "cvtsd2si{l}", SSE_CVT_SD2SI>, XD;
defm CVTSD2SI64 : sse12_cvt_sint<0x2D, VR128, GR64, int_x86_sse2_cvtsd2si64,
- f128mem, load, "cvtsd2si{q}">, XD, REX_W;
+ sdmem, sse_load_f64, "cvtsd2si{q}", SSE_CVT_SD2SI>, XD, REX_W;
defm Int_VCVTSI2SS : sse12_cvt_sint_3addr<0x2A, GR32, VR128,
- int_x86_sse_cvtsi2ss, i32mem, loadi32, "cvtsi2ss", 0>, XS, VEX_4V;
+ int_x86_sse_cvtsi2ss, i32mem, loadi32, "cvtsi2ss",
+ SSE_CVT_Scalar, 0>, XS, VEX_4V;
defm Int_VCVTSI2SS64 : sse12_cvt_sint_3addr<0x2A, GR64, VR128,
- int_x86_sse_cvtsi642ss, i64mem, loadi64, "cvtsi2ss", 0>, XS, VEX_4V,
+ int_x86_sse_cvtsi642ss, i64mem, loadi64, "cvtsi2ss{q}",
+ SSE_CVT_Scalar, 0>, XS, VEX_4V,
VEX_W;
defm Int_VCVTSI2SD : sse12_cvt_sint_3addr<0x2A, GR32, VR128,
- int_x86_sse2_cvtsi2sd, i32mem, loadi32, "cvtsi2sd", 0>, XD, VEX_4V;
+ int_x86_sse2_cvtsi2sd, i32mem, loadi32, "cvtsi2sd",
+ SSE_CVT_Scalar, 0>, XD, VEX_4V;
defm Int_VCVTSI2SD64 : sse12_cvt_sint_3addr<0x2A, GR64, VR128,
- int_x86_sse2_cvtsi642sd, i64mem, loadi64, "cvtsi2sd", 0>, XD,
+ int_x86_sse2_cvtsi642sd, i64mem, loadi64, "cvtsi2sd{q}",
+ SSE_CVT_Scalar, 0>, XD,
VEX_4V, VEX_W;
let Constraints = "$src1 = $dst" in {
defm Int_CVTSI2SS : sse12_cvt_sint_3addr<0x2A, GR32, VR128,
int_x86_sse_cvtsi2ss, i32mem, loadi32,
- "cvtsi2ss">, XS;
+ "cvtsi2ss", SSE_CVT_Scalar>, XS;
defm Int_CVTSI2SS64 : sse12_cvt_sint_3addr<0x2A, GR64, VR128,
int_x86_sse_cvtsi642ss, i64mem, loadi64,
- "cvtsi2ss{q}">, XS, REX_W;
+ "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">, XD;
+ "cvtsi2sd", SSE_CVT_Scalar>, XD;
defm Int_CVTSI2SD64 : sse12_cvt_sint_3addr<0x2A, GR64, VR128,
int_x86_sse2_cvtsi642sd, i64mem, loadi64,
- "cvtsi2sd">, XD, REX_W;
+ "cvtsi2sd{q}", SSE_CVT_Scalar>, XD, REX_W;
}
/// SSE 1 Only
// Aliases for intrinsics
defm Int_VCVTTSS2SI : sse12_cvt_sint<0x2C, VR128, GR32, int_x86_sse_cvttss2si,
- f32mem, load, "cvttss2si">, XS, VEX;
+ 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">, XS, VEX, VEX_W;
+ int_x86_sse_cvttss2si64, ssmem, sse_load_f32,
+ "cvttss2si{q}", SSE_CVT_SS2SI_64>,
+ XS, VEX, VEX_W;
defm Int_VCVTTSD2SI : sse12_cvt_sint<0x2C, VR128, GR32, int_x86_sse2_cvttsd2si,
- f128mem, load, "cvttsd2si">, 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">, XD, VEX, VEX_W;
+ int_x86_sse2_cvttsd2si64, sdmem, sse_load_f64,
+ "cvttsd2si{q}", SSE_CVT_SD2SI>,
+ XD, VEX, VEX_W;
defm Int_CVTTSS2SI : sse12_cvt_sint<0x2C, VR128, GR32, int_x86_sse_cvttss2si,
- f32mem, load, "cvttss2si">, XS;
+ 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}">, XS, REX_W;
+ int_x86_sse_cvttss2si64, ssmem, sse_load_f32,
+ "cvttss2si{q}", SSE_CVT_SS2SI_64>, XS, REX_W;
defm Int_CVTTSD2SI : sse12_cvt_sint<0x2C, VR128, GR32, int_x86_sse2_cvttsd2si,
- f128mem, load, "cvttsd2si">, 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}">, XD, REX_W;
-
-let Pattern = []<dag> in {
-defm VCVTSS2SI : sse12_cvt_s<0x2D, FR32, GR32, undef, f32mem, load,
- "cvtss2si{l}\t{$src, $dst|$dst, $src}">, XS,
- VEX, VEX_LIG;
-defm VCVTSS2SI64 : sse12_cvt_s<0x2D, FR32, GR64, undef, f32mem, load,
- "cvtss2si\t{$src, $dst|$dst, $src}">, XS, VEX,
- VEX_W, VEX_LIG;
-defm VCVTDQ2PS : sse12_cvt_p<0x5B, VR128, VR128, undef, i128mem, load,
- "cvtdq2ps\t{$src, $dst|$dst, $src}",
- SSEPackedSingle>, TB, VEX;
-defm VCVTDQ2PSY : sse12_cvt_p<0x5B, VR256, VR256, undef, i256mem, load,
- "cvtdq2ps\t{$src, $dst|$dst, $src}",
- SSEPackedSingle>, TB, VEX;
-}
-
-let Pattern = []<dag> in {
-defm CVTSS2SI : sse12_cvt_s<0x2D, FR32, GR32, undef, f32mem, load /*dummy*/,
- "cvtss2si{l}\t{$src, $dst|$dst, $src}">, XS;
-defm CVTSS2SI64 : sse12_cvt_s<0x2D, FR32, GR64, undef, f32mem, load /*dummy*/,
- "cvtss2si{q}\t{$src, $dst|$dst, $src}">, XS, REX_W;
-defm CVTDQ2PS : sse12_cvt_p<0x5B, VR128, VR128, undef, i128mem, load /*dummy*/,
+ int_x86_sse2_cvttsd2si64, sdmem, sse_load_f64,
+ "cvttsd2si{q}", SSE_CVT_SD2SI>, XD, REX_W;
+
+defm VCVTSS2SI : sse12_cvt_sint<0x2D, VR128, GR32, int_x86_sse_cvtss2si,
+ ssmem, sse_load_f32, "cvtss2si{l}",
+ SSE_CVT_SS2SI_32>, XS, VEX, VEX_LIG;
+defm VCVTSS2SI64 : sse12_cvt_sint<0x2D, VR128, GR64, int_x86_sse_cvtss2si64,
+ ssmem, sse_load_f32, "cvtss2si{q}",
+ SSE_CVT_SS2SI_64>, XS, VEX, VEX_W, VEX_LIG;
+
+defm CVTSS2SI : sse12_cvt_sint<0x2D, VR128, GR32, int_x86_sse_cvtss2si,
+ ssmem, sse_load_f32, "cvtss2si{l}",
+ SSE_CVT_SS2SI_32>, XS;
+defm CVTSS2SI64 : sse12_cvt_sint<0x2D, VR128, GR64, int_x86_sse_cvtss2si64,
+ ssmem, sse_load_f32, "cvtss2si{q}",
+ SSE_CVT_SS2SI_64>, XS, REX_W;
+
+defm VCVTDQ2PS : sse12_cvt_p<0x5B, VR128, VR128, i128mem,
+ "vcvtdq2ps\t{$src, $dst|$dst, $src}",
+ SSEPackedSingle, SSE_CVT_PS>,
+ TB, VEX, Requires<[HasAVX]>;
+defm VCVTDQ2PSY : sse12_cvt_p<0x5B, VR256, VR256, i256mem,
+ "vcvtdq2ps\t{$src, $dst|$dst, $src}",
+ SSEPackedSingle, SSE_CVT_PS>,
+ TB, VEX, Requires<[HasAVX]>;
+
+defm CVTDQ2PS : sse12_cvt_p<0x5B, VR128, VR128, i128mem,
"cvtdq2ps\t{$src, $dst|$dst, $src}",
- SSEPackedSingle>, TB; /* PD SSE3 form is avaiable */
-}
-
-let Predicates = [HasAVX] in {
- def : Pat<(int_x86_sse_cvtss2si VR128:$src),
- (VCVTSS2SIrr (EXTRACT_SUBREG (v4f32 VR128:$src), sub_ss))>;
- def : Pat<(int_x86_sse_cvtss2si (load addr:$src)),
- (VCVTSS2SIrm addr:$src)>;
- def : Pat<(int_x86_sse_cvtss2si64 VR128:$src),
- (VCVTSS2SI64rr (EXTRACT_SUBREG (v4f32 VR128:$src), sub_ss))>;
- def : Pat<(int_x86_sse_cvtss2si64 (load addr:$src)),
- (VCVTSS2SI64rm addr:$src)>;
-}
-
-let Predicates = [HasSSE1] in {
- def : Pat<(int_x86_sse_cvtss2si VR128:$src),
- (CVTSS2SIrr (EXTRACT_SUBREG (v4f32 VR128:$src), sub_ss))>;
- def : Pat<(int_x86_sse_cvtss2si (load addr:$src)),
- (CVTSS2SIrm addr:$src)>;
- def : Pat<(int_x86_sse_cvtss2si64 VR128:$src),
- (CVTSS2SI64rr (EXTRACT_SUBREG (v4f32 VR128:$src), sub_ss))>;
- def : Pat<(int_x86_sse_cvtss2si64 (load addr:$src)),
- (CVTSS2SI64rm addr:$src)>;
-}
+ SSEPackedSingle, SSE_CVT_PS>,
+ TB, Requires<[UseSSE2]>;
/// SSE 2 Only
// Convert scalar double to scalar single
+let neverHasSideEffects = 1 in {
def VCVTSD2SSrr : VSDI<0x5A, MRMSrcReg, (outs FR32:$dst),
(ins FR64:$src1, FR64:$src2),
- "cvtsd2ss\t{$src2, $src1, $dst|$dst, $src1, $src2}", []>,
- VEX_4V, VEX_LIG;
+ "cvtsd2ss\t{$src2, $src1, $dst|$dst, $src1, $src2}", [],
+ IIC_SSE_CVT_Scalar_RR>, VEX_4V, VEX_LIG;
let mayLoad = 1 in
def VCVTSD2SSrm : I<0x5A, MRMSrcMem, (outs FR32:$dst),
(ins FR64:$src1, f64mem:$src2),
"vcvtsd2ss\t{$src2, $src1, $dst|$dst, $src1, $src2}",
- []>, XD, Requires<[HasAVX, OptForSize]>, VEX_4V, VEX_LIG;
+ [], IIC_SSE_CVT_Scalar_RM>,
+ XD, Requires<[HasAVX, OptForSize]>, VEX_4V, VEX_LIG;
+}
def : Pat<(f32 (fround FR64:$src)), (VCVTSD2SSrr FR64:$src, FR64:$src)>,
Requires<[HasAVX]>;
def CVTSD2SSrr : SDI<0x5A, MRMSrcReg, (outs FR32:$dst), (ins FR64:$src),
"cvtsd2ss\t{$src, $dst|$dst, $src}",
- [(set FR32:$dst, (fround FR64:$src))]>;
+ [(set FR32:$dst, (fround FR64:$src))],
+ IIC_SSE_CVT_Scalar_RR>;
def CVTSD2SSrm : I<0x5A, MRMSrcMem, (outs FR32:$dst), (ins f64mem:$src),
"cvtsd2ss\t{$src, $dst|$dst, $src}",
- [(set FR32:$dst, (fround (loadf64 addr:$src)))]>, XD,
- Requires<[HasSSE2, OptForSize]>;
+ [(set FR32:$dst, (fround (loadf64 addr:$src)))],
+ IIC_SSE_CVT_Scalar_RM>,
+ XD,
+ Requires<[UseSSE2, OptForSize]>;
+
+def Int_VCVTSD2SSrr: I<0x5A, MRMSrcReg,
+ (outs VR128:$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]>;
-defm Int_VCVTSD2SS: sse12_cvt_sint_3addr<0x5A, VR128, VR128,
- int_x86_sse2_cvtsd2ss, f64mem, load, "cvtsd2ss", 0>,
- XS, VEX_4V;
-let Constraints = "$src1 = $dst" in
-defm Int_CVTSD2SS: sse12_cvt_sint_3addr<0x5A, VR128, VR128,
- int_x86_sse2_cvtsd2ss, f64mem, load, "cvtsd2ss">, XS;
+let Constraints = "$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, $src2}",
- []>, XS, Requires<[HasAVX]>, VEX_4V, VEX_LIG;
+ [], IIC_SSE_CVT_Scalar_RR>,
+ XS, Requires<[HasAVX]>, VEX_4V, VEX_LIG;
let mayLoad = 1 in
def VCVTSS2SDrm : I<0x5A, MRMSrcMem, (outs FR64:$dst),
(ins FR32:$src1, f32mem:$src2),
"vcvtss2sd\t{$src2, $src1, $dst|$dst, $src1, $src2}",
- []>, XS, VEX_4V, VEX_LIG, Requires<[HasAVX, OptForSize]>;
-
-let Predicates = [HasAVX] in {
- def : Pat<(f64 (fextend FR32:$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)>;
+ [], IIC_SSE_CVT_Scalar_RM>,
+ XS, VEX_4V, VEX_LIG, Requires<[HasAVX, OptForSize]>;
}
+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:$src))]>, XS,
- Requires<[HasSSE2]>;
+ [(set FR64:$dst, (fextend FR32:$src))],
+ IIC_SSE_CVT_Scalar_RR>, XS,
+ Requires<[UseSSE2]>;
def CVTSS2SDrm : I<0x5A, MRMSrcMem, (outs FR64:$dst), (ins f32mem:$src),
"cvtss2sd\t{$src, $dst|$dst, $src}",
- [(set FR64:$dst, (extloadf32 addr:$src))]>, XS,
- Requires<[HasSSE2, OptForSize]>;
+ [(set FR64:$dst, (extloadf32 addr:$src))],
+ IIC_SSE_CVT_Scalar_RM>, XS,
+ Requires<[UseSSE2, OptForSize]>;
// extload f32 -> f64. This matches load+fextend because we have a hack in
// the isel (PreprocessForFPConvert) that can introduce loads after dag
// Since these loads aren't folded into the fextend, we have to match it
// explicitly here.
def : Pat<(fextend (loadf32 addr:$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))]>, 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)))]>, 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))]>, 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)))]>, XS,
- Requires<[HasSSE2]>;
-}
-
-// Convert doubleword to packed single/double fp
-// SSE2 instructions without OpSize prefix
-def Int_VCVTDQ2PSrr : I<0x5B, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
- "vcvtdq2ps\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst, (int_x86_sse2_cvtdq2ps VR128:$src))]>,
- TB, VEX, Requires<[HasAVX]>;
-def Int_VCVTDQ2PSrm : I<0x5B, MRMSrcMem, (outs VR128:$dst), (ins i128mem:$src),
- "vcvtdq2ps\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst, (int_x86_sse2_cvtdq2ps
- (bitconvert (memopv2i64 addr:$src))))]>,
- TB, VEX, Requires<[HasAVX]>;
-def Int_CVTDQ2PSrr : I<0x5B, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
- "cvtdq2ps\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst, (int_x86_sse2_cvtdq2ps VR128:$src))]>,
- TB, Requires<[HasSSE2]>;
-def Int_CVTDQ2PSrm : I<0x5B, MRMSrcMem, (outs VR128:$dst), (ins i128mem:$src),
- "cvtdq2ps\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst, (int_x86_sse2_cvtdq2ps
- (bitconvert (memopv2i64 addr:$src))))]>,
- TB, Requires<[HasSSE2]>;
-
-// FIXME: why the non-intrinsic version is described as SSE3?
-// SSE2 instructions with XS prefix
-def Int_VCVTDQ2PDrr : I<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
- "vcvtdq2pd\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst, (int_x86_sse2_cvtdq2pd VR128:$src))]>,
- XS, VEX, Requires<[HasAVX]>;
-def Int_VCVTDQ2PDrm : I<0xE6, MRMSrcMem, (outs VR128:$dst), (ins i64mem:$src),
- "vcvtdq2pd\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst, (int_x86_sse2_cvtdq2pd
- (bitconvert (memopv2i64 addr:$src))))]>,
- XS, VEX, Requires<[HasAVX]>;
-def Int_CVTDQ2PDrr : I<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
- "cvtdq2pd\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst, (int_x86_sse2_cvtdq2pd VR128:$src))]>,
- XS, Requires<[HasSSE2]>;
-def Int_CVTDQ2PDrm : I<0xE6, MRMSrcMem, (outs VR128:$dst), (ins i64mem:$src),
- "cvtdq2pd\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst, (int_x86_sse2_cvtdq2pd
- (bitconvert (memopv2i64 addr:$src))))]>,
- XS, Requires<[HasSSE2]>;
-
+ [(set VR128:$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, $src}", []>, VEX;
+ "cvtps2dq\t{$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, $src}", []>, VEX;
+ "cvtps2dq\t{$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, $src}", []>, VEX;
+ "cvtps2dq\t{$src, $dst|$dst, $src}",
+ [(set VR256:$dst,
+ (int_x86_avx_cvt_ps2dq_256 VR256:$src))],
+ IIC_SSE_CVT_PS_RR>, VEX;
def VCVTPS2DQYrm : VPDI<0x5B, MRMSrcMem, (outs VR256:$dst), (ins f256mem:$src),
- "cvtps2dq\t{$src, $dst|$dst, $src}", []>, VEX;
+ "cvtps2dq\t{$src, $dst|$dst, $src}",
+ [(set VR256:$dst,
+ (int_x86_avx_cvt_ps2dq_256 (memopv8f32 addr:$src)))],
+ IIC_SSE_CVT_PS_RM>, VEX;
def CVTPS2DQrr : PDI<0x5B, MRMSrcReg, (outs VR128:$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:$src)))],
+ IIC_SSE_CVT_PS_RM>;
-def Int_VCVTPS2DQrr : VPDI<0x5B, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
- "cvtps2dq\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst, (int_x86_sse2_cvtps2dq VR128:$src))]>,
- VEX;
-def Int_VCVTPS2DQrm : VPDI<0x5B, MRMSrcMem, (outs VR128:$dst),
- (ins f128mem:$src),
- "cvtps2dq\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst, (int_x86_sse2_cvtps2dq
- (memop addr:$src)))]>, VEX;
-def Int_CVTPS2DQrr : PDI<0x5B, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
- "cvtps2dq\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst, (int_x86_sse2_cvtps2dq VR128:$src))]>;
-def Int_CVTPS2DQrm : PDI<0x5B, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
- "cvtps2dq\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst, (int_x86_sse2_cvtps2dq
- (memop addr:$src)))]>;
-
-// SSE2 packed instructions with XD prefix
-def Int_VCVTPD2DQrr : I<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
- "vcvtpd2dq\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst, (int_x86_sse2_cvtpd2dq VR128:$src))]>,
- XD, VEX, Requires<[HasAVX]>;
-def Int_VCVTPD2DQrm : I<0xE6, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
+
+// Convert Packed Double FP to Packed DW Integers
+let Predicates = [HasAVX] in {
+// The assembler can recognize rr 256-bit instructions by seeing a ymm
+// register, but the same isn't true when using memory operands instead.
+// Provide other assembly rr and rm forms to address this explicitly.
+def VCVTPD2DQrr : SDI<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
"vcvtpd2dq\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst, (int_x86_sse2_cvtpd2dq
- (memop addr:$src)))]>,
- XD, VEX, Requires<[HasAVX]>;
-def Int_CVTPD2DQrr : I<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
- "cvtpd2dq\t{$src, $dst|$dst, $src}",
[(set VR128:$dst, (int_x86_sse2_cvtpd2dq VR128:$src))]>,
- XD, Requires<[HasSSE2]>;
-def Int_CVTPD2DQrm : I<0xE6, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
- "cvtpd2dq\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst, (int_x86_sse2_cvtpd2dq
- (memop addr:$src)))]>,
- XD, Requires<[HasSSE2]>;
+ VEX;
+// XMM only
+def : InstAlias<"vcvtpd2dqx\t{$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, $src}",
+ [(set VR128:$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, $src}",
+ [(set VR128:$dst,
+ (int_x86_avx_cvt_pd2dq_256 VR256:$src))]>, VEX;
+def VCVTPD2DQYrm : SDI<0xE6, MRMSrcMem, (outs VR128:$dst), (ins f256mem:$src),
+ "vcvtpd2dq{y}\t{$src, $dst|$dst, $src}",
+ [(set VR128:$dst,
+ (int_x86_avx_cvt_pd2dq_256 (memopv4f64 addr:$src)))]>,
+ VEX, VEX_L;
+def : InstAlias<"vcvtpd2dq\t{$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, $src}",
+ [(set VR128:$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, $src}",
+ [(set VR128:$dst, (int_x86_sse2_cvtpd2dq VR128:$src))],
+ IIC_SSE_CVT_PD_RR>;
// Convert with truncation packed single/double fp to doubleword
// SSE2 packed instructions with XS prefix
-def VCVTTPS2DQrr : VSSI<0x5B, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
- "cvttps2dq\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst,
- (int_x86_sse2_cvttps2dq VR128:$src))]>, VEX;
-def VCVTTPS2DQrm : VSSI<0x5B, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
- "cvttps2dq\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst, (int_x86_sse2_cvttps2dq
- (memop addr:$src)))]>, VEX;
-def VCVTTPS2DQYrr : VSSI<0x5B, MRMSrcReg, (outs VR256:$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:$src))]>, VEX;
-def VCVTTPS2DQYrm : VSSI<0x5B, MRMSrcMem, (outs VR256:$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:$src)))]>, VEX;
-
-def CVTTPS2DQrr : SSI<0x5B, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
- "cvttps2dq\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst,
- (int_x86_sse2_cvttps2dq VR128:$src))]>;
-def CVTTPS2DQrm : SSI<0x5B, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
- "cvttps2dq\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst,
- (int_x86_sse2_cvttps2dq (memop addr:$src)))]>;
+ [(set VR128:$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:$src))],
+ IIC_SSE_CVT_PS_RR>, VEX;
+def VCVTTPS2DQYrm : VS2SI<0x5B, MRMSrcMem, (outs VR256:$dst), (ins f256mem:$src),
+ "cvttps2dq\t{$src, $dst|$dst, $src}",
+ [(set VR256:$dst, (int_x86_avx_cvtt_ps2dq_256
+ (memopv8f32 addr:$src)))],
+ IIC_SSE_CVT_PS_RM>, VEX;
+
+def CVTTPS2DQrr : S2SI<0x5B, MRMSrcReg, (outs VR128:$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))),
- (Int_VCVTDQ2PSrr VR128:$src)>;
+ (VCVTDQ2PSrr VR128:$src)>;
def : Pat<(v4f32 (sint_to_fp (bc_v4i32 (memopv2i64 addr:$src)))),
- (Int_VCVTDQ2PSrm addr:$src)>;
+ (VCVTDQ2PSrm addr:$src)>;
+
+ def : Pat<(int_x86_sse2_cvtdq2ps VR128:$src),
+ (VCVTDQ2PSrr VR128:$src)>;
+ def : Pat<(int_x86_sse2_cvtdq2ps (bc_v4i32 (memopv2i64 addr:$src))),
+ (VCVTDQ2PSrm addr:$src)>;
def : Pat<(v4i32 (fp_to_sint (v4f32 VR128:$src))),
(VCVTTPS2DQrr VR128:$src)>;
(VCVTTPS2DQYrm addr:$src)>;
}
-let Predicates = [HasSSE2] in {
+let Predicates = [UseSSE2] in {
def : Pat<(v4f32 (sint_to_fp (v4i32 VR128:$src))),
- (Int_CVTDQ2PSrr VR128:$src)>;
+ (CVTDQ2PSrr VR128:$src)>;
def : Pat<(v4f32 (sint_to_fp (bc_v4i32 (memopv2i64 addr:$src)))),
- (Int_CVTDQ2PSrm addr:$src)>;
+ (CVTDQ2PSrm addr:$src)>;
+
+ def : Pat<(int_x86_sse2_cvtdq2ps VR128:$src),
+ (CVTDQ2PSrr VR128:$src)>;
+ def : Pat<(int_x86_sse2_cvtdq2ps (bc_v4i32 (memopv2i64 addr:$src))),
+ (CVTDQ2PSrm addr:$src)>;
def : Pat<(v4i32 (fp_to_sint (v4f32 VR128:$src))),
(CVTTPS2DQrr VR128:$src)>;
def VCVTTPD2DQrr : VPDI<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
"cvttpd2dq\t{$src, $dst|$dst, $src}",
[(set VR128:$dst,
- (int_x86_sse2_cvttpd2dq VR128:$src))]>, VEX;
-let isCodeGenOnly = 1 in
-def VCVTTPD2DQrm : VPDI<0xE6, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
- "cvttpd2dq\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst, (int_x86_sse2_cvttpd2dq
- (memop addr:$src)))]>, VEX;
-def CVTTPD2DQrr : PDI<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
- "cvttpd2dq\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst, (int_x86_sse2_cvttpd2dq VR128:$src))]>;
-def CVTTPD2DQrm : PDI<0xE6, MRMSrcMem, (outs VR128:$dst),(ins f128mem:$src),
- "cvttpd2dq\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst, (int_x86_sse2_cvttpd2dq
- (memop addr:$src)))]>;
+ (int_x86_sse2_cvttpd2dq VR128:$src))],
+ IIC_SSE_CVT_PD_RR>, VEX;
// The assembler can recognize rr 256-bit instructions by seeing a ymm
// register, but the same isn't true when using memory operands instead.
// Provide other assembly rr and rm forms to address this explicitly.
-def VCVTTPD2DQXrYr : VPDI<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR256:$src),
- "cvttpd2dq\t{$src, $dst|$dst, $src}", []>, VEX;
// XMM only
-def VCVTTPD2DQXrr : VPDI<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
- "cvttpd2dqx\t{$src, $dst|$dst, $src}", []>, VEX;
+def : InstAlias<"vcvttpd2dqx\t{$src, $dst|$dst, $src}",
+ (VCVTTPD2DQrr VR128:$dst, VR128:$src)>;
def VCVTTPD2DQXrm : VPDI<0xE6, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
- "cvttpd2dqx\t{$src, $dst|$dst, $src}", []>, VEX;
+ "cvttpd2dqx\t{$src, $dst|$dst, $src}",
+ [(set VR128:$dst, (int_x86_sse2_cvttpd2dq
+ (memopv2f64 addr:$src)))],
+ IIC_SSE_CVT_PD_RM>, VEX;
// YMM only
def VCVTTPD2DQYrr : VPDI<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR256:$src),
- "cvttpd2dqy\t{$src, $dst|$dst, $src}", []>, VEX;
+ "cvttpd2dq{y}\t{$src, $dst|$dst, $src}",
+ [(set VR128:$dst,
+ (int_x86_avx_cvtt_pd2dq_256 VR256:$src))],
+ IIC_SSE_CVT_PD_RR>, VEX;
def VCVTTPD2DQYrm : VPDI<0xE6, MRMSrcMem, (outs VR128:$dst), (ins f256mem:$src),
- "cvttpd2dqy\t{$src, $dst|$dst, $src}", []>, VEX, VEX_L;
+ "cvttpd2dq{y}\t{$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:$src)>;
+
+let Predicates = [HasAVX] in {
+ def : Pat<(v4i32 (fp_to_sint (v4f64 VR256:$src))),
+ (VCVTTPD2DQYrr VR256:$src)>;
+ def : Pat<(v4i32 (fp_to_sint (memopv4f64 addr:$src))),
+ (VCVTTPD2DQYrm addr:$src)>;
+} // Predicates = [HasAVX]
+
+def CVTTPD2DQrr : PDI<0xE6, MRMSrcReg, (outs VR128:$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, $src}", []>, TB, VEX;
+ "vcvtps2pd\t{$src, $dst|$dst, $src}",
+ [(set VR128:$dst, (int_x86_sse2_cvtps2pd VR128:$src))],
+ IIC_SSE_CVT_PD_RR>, TB, VEX;
+let neverHasSideEffects = 1, mayLoad = 1 in
def VCVTPS2PDrm : I<0x5A, MRMSrcMem, (outs VR128:$dst), (ins f64mem:$src),
- "vcvtps2pd\t{$src, $dst|$dst, $src}", []>, TB, VEX;
+ "vcvtps2pd\t{$src, $dst|$dst, $src}", [],
+ IIC_SSE_CVT_PD_RM>, TB, VEX;
def VCVTPS2PDYrr : I<0x5A, MRMSrcReg, (outs VR256:$dst), (ins VR128:$src),
- "vcvtps2pd\t{$src, $dst|$dst, $src}", []>, TB, VEX;
+ "vcvtps2pd\t{$src, $dst|$dst, $src}",
+ [(set VR256:$dst,
+ (int_x86_avx_cvt_ps2_pd_256 VR128:$src))],
+ IIC_SSE_CVT_PD_RR>, TB, VEX;
def VCVTPS2PDYrm : I<0x5A, MRMSrcMem, (outs VR256:$dst), (ins f128mem:$src),
- "vcvtps2pd\t{$src, $dst|$dst, $src}", []>, TB, VEX;
+ "vcvtps2pd\t{$src, $dst|$dst, $src}",
+ [(set VR256:$dst,
+ (int_x86_avx_cvt_ps2_pd_256 (memopv4f32 addr:$src)))],
+ IIC_SSE_CVT_PD_RM>, TB, VEX;
}
+
+let Predicates = [UseSSE2] in {
def CVTPS2PDrr : I<0x5A, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
- "cvtps2pd\t{$src, $dst|$dst, $src}", []>, TB;
-def CVTPS2PDrm : I<0x5A, MRMSrcMem, (outs VR128:$dst), (ins f64mem:$src),
- "cvtps2pd\t{$src, $dst|$dst, $src}", []>, TB;
-
-def Int_VCVTPS2PDrr : I<0x5A, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
- "vcvtps2pd\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst, (int_x86_sse2_cvtps2pd VR128:$src))]>,
- TB, VEX, Requires<[HasAVX]>;
-def Int_VCVTPS2PDrm : I<0x5A, MRMSrcMem, (outs VR128:$dst), (ins f64mem:$src),
- "vcvtps2pd\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst, (int_x86_sse2_cvtps2pd
- (load addr:$src)))]>,
- TB, VEX, Requires<[HasAVX]>;
-def Int_CVTPS2PDrr : I<0x5A, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
"cvtps2pd\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst, (int_x86_sse2_cvtps2pd VR128:$src))]>,
- TB, Requires<[HasSSE2]>;
-def Int_CVTPS2PDrm : I<0x5A, MRMSrcMem, (outs VR128:$dst), (ins f64mem:$src),
- "cvtps2pd\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst, (int_x86_sse2_cvtps2pd
- (load addr:$src)))]>,
- TB, Requires<[HasSSE2]>;
+ [(set VR128:$dst, (int_x86_sse2_cvtps2pd VR128:$src))],
+ IIC_SSE_CVT_PD_RR>, TB;
+let neverHasSideEffects = 1, mayLoad = 1 in
+def CVTPS2PDrm : I<0x5A, MRMSrcMem, (outs VR128:$dst), (ins f64mem:$src),
+ "cvtps2pd\t{$src, $dst|$dst, $src}", [],
+ IIC_SSE_CVT_PD_RM>, TB;
+}
+
+// Convert Packed DW Integers to Packed Double FP
+let Predicates = [HasAVX] in {
+let neverHasSideEffects = 1, mayLoad = 1 in
+def VCVTDQ2PDrm : S2SI<0xE6, MRMSrcMem, (outs VR128:$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:$src))))]>, VEX;
+def VCVTDQ2PDYrr : S2SI<0xE6, MRMSrcReg, (outs VR256:$dst), (ins VR128:$src),
+ "vcvtdq2pd\t{$src, $dst|$dst, $src}",
+ [(set VR256:$dst,
+ (int_x86_avx_cvtdq2_pd_256 VR128:$src))]>, VEX;
+}
+
+let neverHasSideEffects = 1, mayLoad = 1 in
+def CVTDQ2PDrm : S2SI<0xE6, MRMSrcMem, (outs VR128:$dst), (ins i64mem:$src),
+ "cvtdq2pd\t{$src, $dst|$dst, $src}", [],
+ IIC_SSE_CVT_PD_RR>;
+def CVTDQ2PDrr : S2SI<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
+ "cvtdq2pd\t{$src, $dst|$dst, $src}",
+ [(set VR128:$dst, (int_x86_sse2_cvtdq2pd VR128:$src))],
+ IIC_SSE_CVT_PD_RM>;
+
+// AVX 256-bit register conversion intrinsics
+let Predicates = [HasAVX] in {
+ def : Pat<(v4f64 (sint_to_fp (v4i32 VR128:$src))),
+ (VCVTDQ2PDYrr VR128:$src)>;
+ def : Pat<(v4f64 (sint_to_fp (bc_v4i32 (memopv2i64 addr:$src)))),
+ (VCVTDQ2PDYrm addr:$src)>;
+} // Predicates = [HasAVX]
// Convert packed double to packed single
// The assembler can recognize rr 256-bit instructions by seeing a ymm
// register, but the same isn't true when using memory operands instead.
// Provide other assembly rr and rm forms to address this explicitly.
def VCVTPD2PSrr : VPDI<0x5A, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
- "cvtpd2ps\t{$src, $dst|$dst, $src}", []>, VEX;
-def VCVTPD2PSXrYr : VPDI<0x5A, MRMSrcReg, (outs VR128:$dst), (ins VR256:$src),
- "cvtpd2ps\t{$src, $dst|$dst, $src}", []>, VEX;
+ "cvtpd2ps\t{$src, $dst|$dst, $src}",
+ [(set VR128:$dst, (int_x86_sse2_cvtpd2ps VR128:$src))],
+ IIC_SSE_CVT_PD_RR>, VEX;
// XMM only
-def VCVTPD2PSXrr : VPDI<0x5A, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
- "cvtpd2psx\t{$src, $dst|$dst, $src}", []>, VEX;
+def : InstAlias<"vcvtpd2psx\t{$src, $dst|$dst, $src}",
+ (VCVTPD2PSrr VR128:$dst, VR128:$src)>;
def VCVTPD2PSXrm : VPDI<0x5A, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
- "cvtpd2psx\t{$src, $dst|$dst, $src}", []>, VEX;
+ "cvtpd2psx\t{$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:$src),
- "cvtpd2psy\t{$src, $dst|$dst, $src}", []>, VEX;
+ "cvtpd2ps{y}\t{$src, $dst|$dst, $src}",
+ [(set VR128:$dst,
+ (int_x86_avx_cvt_pd2_ps_256 VR256:$src))],
+ IIC_SSE_CVT_PD_RR>, VEX;
def VCVTPD2PSYrm : VPDI<0x5A, MRMSrcMem, (outs VR128:$dst), (ins f256mem:$src),
- "cvtpd2psy\t{$src, $dst|$dst, $src}", []>, VEX, VEX_L;
+ "cvtpd2ps{y}\t{$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, $src}", []>;
-
-
-def Int_VCVTPD2PSrr : VPDI<0x5A, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
- "cvtpd2ps\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst, (int_x86_sse2_cvtpd2ps VR128:$src))]>;
-def Int_VCVTPD2PSrm : VPDI<0x5A, MRMSrcMem, (outs VR128:$dst),
- (ins f128mem:$src),
- "cvtpd2ps\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst, (int_x86_sse2_cvtpd2ps
- (memop addr:$src)))]>;
-def Int_CVTPD2PSrr : PDI<0x5A, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
- "cvtpd2ps\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst, (int_x86_sse2_cvtpd2ps VR128:$src))]>;
-def Int_CVTPD2PSrm : PDI<0x5A, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
- "cvtpd2ps\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst, (int_x86_sse2_cvtpd2ps
- (memop addr:$src)))]>;
+ "cvtpd2ps\t{$src, $dst|$dst, $src}",
+ [(set VR128:$dst,
+ (int_x86_sse2_cvtpd2ps (memopv2f64 addr:$src)))],
+ IIC_SSE_CVT_PD_RM>;
+
// AVX 256-bit register conversion intrinsics
// FIXME: Migrate SSE conversion intrinsics matching to use patterns as below
// whenever possible to avoid declaring two versions of each one.
-def : Pat<(int_x86_avx_cvtdq2_ps_256 VR256:$src),
- (VCVTDQ2PSYrr VR256:$src)>;
-def : Pat<(int_x86_avx_cvtdq2_ps_256 (bitconvert (memopv4i64 addr:$src))),
- (VCVTDQ2PSYrm addr:$src)>;
-
-def : Pat<(int_x86_avx_cvt_pd2_ps_256 VR256:$src),
- (VCVTPD2PSYrr VR256:$src)>;
-def : Pat<(int_x86_avx_cvt_pd2_ps_256 (memopv4f64 addr:$src)),
- (VCVTPD2PSYrm addr:$src)>;
-
-def : Pat<(int_x86_avx_cvt_ps2dq_256 VR256:$src),
- (VCVTPS2DQYrr VR256:$src)>;
-def : Pat<(int_x86_avx_cvt_ps2dq_256 (memopv8f32 addr:$src)),
- (VCVTPS2DQYrm addr:$src)>;
-
-def : Pat<(int_x86_avx_cvt_ps2_pd_256 VR128:$src),
- (VCVTPS2PDYrr VR128:$src)>;
-def : Pat<(int_x86_avx_cvt_ps2_pd_256 (memopv4f32 addr:$src)),
- (VCVTPS2PDYrm addr:$src)>;
-
-def : Pat<(int_x86_avx_cvtt_pd2dq_256 VR256:$src),
- (VCVTTPD2DQYrr VR256:$src)>;
-def : Pat<(int_x86_avx_cvtt_pd2dq_256 (memopv4f64 addr:$src)),
- (VCVTTPD2DQYrm addr:$src)>;
-
-// Match fround and fextend for 128/256-bit conversions
-def : Pat<(v4f32 (fround (v4f64 VR256:$src))),
- (VCVTPD2PSYrr VR256:$src)>;
-def : Pat<(v4f32 (fround (loadv4f64 addr:$src))),
- (VCVTPD2PSYrm addr:$src)>;
-
-def : Pat<(v4f64 (fextend (v4f32 VR128:$src))),
- (VCVTPS2PDYrr VR128:$src)>;
-def : Pat<(v4f64 (fextend (loadv4f32 addr:$src))),
- (VCVTPS2PDYrm addr:$src)>;
+let Predicates = [HasAVX] in {
+ def : Pat<(int_x86_avx_cvtdq2_ps_256 VR256:$src),
+ (VCVTDQ2PSYrr VR256:$src)>;
+ def : Pat<(int_x86_avx_cvtdq2_ps_256 (bitconvert (memopv4i64 addr:$src))),
+ (VCVTDQ2PSYrm addr:$src)>;
+
+ // Match fround and fextend for 128/256-bit conversions
+ def : Pat<(v4f32 (fround (v4f64 VR256:$src))),
+ (VCVTPD2PSYrr VR256:$src)>;
+ def : Pat<(v4f32 (fround (loadv4f64 addr:$src))),
+ (VCVTPD2PSYrm addr:$src)>;
+
+ def : Pat<(v2f64 (X86vfpext (v4f32 VR128:$src))),
+ (VCVTPS2PDrr VR128:$src)>;
+ def : Pat<(v4f64 (fextend (v4f32 VR128:$src))),
+ (VCVTPS2PDYrr VR128:$src)>;
+ def : Pat<(v4f64 (fextend (loadv4f32 addr:$src))),
+ (VCVTPS2PDYrm addr:$src)>;
+}
+
+let Predicates = [UseSSE2] in {
+ // Match fextend for 128 conversions
+ def : Pat<(v2f64 (X86vfpext (v4f32 VR128:$src))),
+ (CVTPS2PDrr VR128:$src)>;
+}
//===----------------------------------------------------------------------===//
// SSE 1 & 2 - Compare Instructions
// sse12_cmp_scalar - sse 1 & 2 compare scalar instructions
multiclass sse12_cmp_scalar<RegisterClass RC, X86MemOperand x86memop,
- SDNode OpNode, ValueType VT, PatFrag ld_frag,
- string asm, string asm_alt> {
+ Operand CC, SDNode OpNode, ValueType VT,
+ PatFrag ld_frag, string asm, string asm_alt,
+ OpndItins itins> {
def rr : SIi8<0xC2, MRMSrcReg,
- (outs RC:$dst), (ins RC:$src1, RC:$src2, SSECC:$cc), asm,
- [(set RC:$dst, (OpNode (VT RC:$src1), RC:$src2, imm:$cc))]>;
+ (outs RC:$dst), (ins RC:$src1, RC:$src2, CC:$cc), asm,
+ [(set RC:$dst, (OpNode (VT RC:$src1), RC:$src2, imm:$cc))],
+ itins.rr>;
def rm : SIi8<0xC2, MRMSrcMem,
- (outs RC:$dst), (ins RC:$src1, x86memop:$src2, SSECC:$cc), asm,
+ (outs RC:$dst), (ins RC:$src1, x86memop:$src2, CC:$cc), asm,
[(set RC:$dst, (OpNode (VT RC:$src1),
- (ld_frag addr:$src2), imm:$cc))]>;
+ (ld_frag addr:$src2), imm:$cc))],
+ itins.rm>;
// Accept explicit immediate argument form instead of comparison code.
let neverHasSideEffects = 1 in {
def rr_alt : SIi8<0xC2, MRMSrcReg, (outs RC:$dst),
- (ins RC:$src1, RC:$src2, i8imm:$cc), asm_alt, []>;
+ (ins RC:$src1, RC:$src2, i8imm:$cc), asm_alt, [],
+ IIC_SSE_ALU_F32S_RR>;
let mayLoad = 1 in
def rm_alt : SIi8<0xC2, MRMSrcMem, (outs RC:$dst),
- (ins RC:$src1, x86memop:$src2, i8imm:$cc), asm_alt, []>;
+ (ins RC:$src1, x86memop:$src2, i8imm:$cc), asm_alt, [],
+ IIC_SSE_ALU_F32S_RM>;
}
}
-defm VCMPSS : sse12_cmp_scalar<FR32, f32mem, X86cmpss, f32, loadf32,
+defm VCMPSS : sse12_cmp_scalar<FR32, f32mem, AVXCC, X86cmpss, f32, loadf32,
"cmp${cc}ss\t{$src2, $src1, $dst|$dst, $src1, $src2}",
- "cmpss\t{$cc, $src2, $src1, $dst|$dst, $src1, $src2, $cc}">,
+ "cmpss\t{$cc, $src2, $src1, $dst|$dst, $src1, $src2, $cc}",
+ SSE_ALU_F32S>,
XS, VEX_4V, VEX_LIG;
-defm VCMPSD : sse12_cmp_scalar<FR64, f64mem, X86cmpsd, f64, loadf64,
+defm VCMPSD : sse12_cmp_scalar<FR64, f64mem, AVXCC, X86cmpsd, f64, loadf64,
"cmp${cc}sd\t{$src2, $src1, $dst|$dst, $src1, $src2}",
- "cmpsd\t{$cc, $src2, $src1, $dst|$dst, $src1, $src2, $cc}">,
+ "cmpsd\t{$cc, $src2, $src1, $dst|$dst, $src1, $src2, $cc}",
+ SSE_ALU_F32S>, // same latency as 32 bit compare
XD, VEX_4V, VEX_LIG;
let Constraints = "$src1 = $dst" in {
- defm CMPSS : sse12_cmp_scalar<FR32, f32mem, X86cmpss, f32, loadf32,
+ defm CMPSS : sse12_cmp_scalar<FR32, f32mem, SSECC, X86cmpss, f32, loadf32,
"cmp${cc}ss\t{$src2, $dst|$dst, $src2}",
- "cmpss\t{$cc, $src2, $dst|$dst, $src2, $cc}">,
+ "cmpss\t{$cc, $src2, $dst|$dst, $src2, $cc}", SSE_ALU_F32S>,
XS;
- defm CMPSD : sse12_cmp_scalar<FR64, f64mem, X86cmpsd, f64, loadf64,
+ defm CMPSD : sse12_cmp_scalar<FR64, f64mem, SSECC, X86cmpsd, f64, loadf64,
"cmp${cc}sd\t{$src2, $dst|$dst, $src2}",
- "cmpsd\t{$cc, $src2, $dst|$dst, $src2, $cc}">,
+ "cmpsd\t{$cc, $src2, $dst|$dst, $src2, $cc}",
+ SSE_ALU_F32S>, // same latency as 32 bit compare
XD;
}
-multiclass sse12_cmp_scalar_int<RegisterClass RC, X86MemOperand x86memop,
- Intrinsic Int, string asm> {
+multiclass sse12_cmp_scalar_int<X86MemOperand x86memop, Operand CC,
+ Intrinsic Int, string asm, OpndItins itins> {
def rr : SIi8<0xC2, MRMSrcReg, (outs VR128:$dst),
- (ins VR128:$src1, VR128:$src, SSECC:$cc), asm,
+ (ins VR128:$src1, VR128:$src, CC:$cc), asm,
[(set VR128:$dst, (Int VR128:$src1,
- VR128:$src, imm:$cc))]>;
+ VR128:$src, imm:$cc))],
+ itins.rr>;
def rm : SIi8<0xC2, MRMSrcMem, (outs VR128:$dst),
- (ins VR128:$src1, x86memop:$src, SSECC:$cc), asm,
+ (ins VR128:$src1, x86memop:$src, CC:$cc), asm,
[(set VR128:$dst, (Int VR128:$src1,
- (load addr:$src), imm:$cc))]>;
+ (load addr:$src), imm:$cc))],
+ itins.rm>;
}
// Aliases to match intrinsics which expect XMM operand(s).
-defm Int_VCMPSS : sse12_cmp_scalar_int<VR128, f32mem, int_x86_sse_cmp_ss,
- "cmp${cc}ss\t{$src, $src1, $dst|$dst, $src1, $src}">,
+defm Int_VCMPSS : sse12_cmp_scalar_int<f32mem, AVXCC, int_x86_sse_cmp_ss,
+ "cmp${cc}ss\t{$src, $src1, $dst|$dst, $src1, $src}",
+ SSE_ALU_F32S>,
XS, VEX_4V;
-defm Int_VCMPSD : sse12_cmp_scalar_int<VR128, f64mem, int_x86_sse2_cmp_sd,
- "cmp${cc}sd\t{$src, $src1, $dst|$dst, $src1, $src}">,
+defm Int_VCMPSD : sse12_cmp_scalar_int<f64mem, AVXCC, int_x86_sse2_cmp_sd,
+ "cmp${cc}sd\t{$src, $src1, $dst|$dst, $src1, $src}",
+ SSE_ALU_F32S>, // same latency as f32
XD, VEX_4V;
let Constraints = "$src1 = $dst" in {
- defm Int_CMPSS : sse12_cmp_scalar_int<VR128, f32mem, int_x86_sse_cmp_ss,
- "cmp${cc}ss\t{$src, $dst|$dst, $src}">, XS;
- defm Int_CMPSD : sse12_cmp_scalar_int<VR128, f64mem, int_x86_sse2_cmp_sd,
- "cmp${cc}sd\t{$src, $dst|$dst, $src}">, XD;
+ defm Int_CMPSS : sse12_cmp_scalar_int<f32mem, SSECC, int_x86_sse_cmp_ss,
+ "cmp${cc}ss\t{$src, $dst|$dst, $src}",
+ SSE_ALU_F32S>, XS;
+ defm Int_CMPSD : sse12_cmp_scalar_int<f64mem, SSECC, int_x86_sse2_cmp_sd,
+ "cmp${cc}sd\t{$src, $dst|$dst, $src}",
+ SSE_ALU_F32S>, // same latency as f32
+ XD;
}
def rr: PI<opc, MRMSrcReg, (outs), (ins RC:$src1, RC:$src2),
!strconcat(OpcodeStr, "\t{$src2, $src1|$src1, $src2}"),
[(set EFLAGS, (OpNode (vt RC:$src1), RC:$src2))],
- IIC_DEFAULT, d>;
+ IIC_SSE_COMIS_RR, d>;
def rm: PI<opc, MRMSrcMem, (outs), (ins RC:$src1, x86memop:$src2),
!strconcat(OpcodeStr, "\t{$src2, $src1|$src1, $src2}"),
[(set EFLAGS, (OpNode (vt RC:$src1),
(ld_frag addr:$src2)))],
- IIC_DEFAULT, d>;
+ IIC_SSE_COMIS_RM, d>;
}
let Defs = [EFLAGS] in {
"comisd", SSEPackedDouble>, TB, OpSize;
} // Defs = [EFLAGS]
-// sse12_cmp_packed - sse 1 & 2 compared packed instructions
+// sse12_cmp_packed - sse 1 & 2 compare packed instructions
multiclass sse12_cmp_packed<RegisterClass RC, X86MemOperand x86memop,
- Intrinsic Int, string asm, string asm_alt,
- Domain d> {
- let isAsmParserOnly = 1 in {
- def rri : PIi8<0xC2, MRMSrcReg,
- (outs RC:$dst), (ins RC:$src1, RC:$src2, SSECC:$cc), asm,
- [(set RC:$dst, (Int RC:$src1, RC:$src2, imm:$cc))],
- IIC_DEFAULT, d>;
- def rmi : PIi8<0xC2, MRMSrcMem,
- (outs RC:$dst), (ins RC:$src1, x86memop:$src2, SSECC:$cc), asm,
- [(set RC:$dst, (Int RC:$src1, (memop addr:$src2), imm:$cc))],
- IIC_DEFAULT, d>;
- }
+ Operand CC, Intrinsic Int, string asm,
+ string asm_alt, Domain d> {
+ def rri : PIi8<0xC2, MRMSrcReg,
+ (outs RC:$dst), (ins RC:$src1, RC:$src2, CC:$cc), asm,
+ [(set RC:$dst, (Int RC:$src1, RC:$src2, imm:$cc))],
+ IIC_SSE_CMPP_RR, d>;
+ def rmi : PIi8<0xC2, MRMSrcMem,
+ (outs RC:$dst), (ins RC:$src1, x86memop:$src2, CC:$cc), asm,
+ [(set RC:$dst, (Int RC:$src1, (memop addr:$src2), imm:$cc))],
+ IIC_SSE_CMPP_RM, d>;
// Accept explicit immediate argument form instead of comparison code.
- def rri_alt : PIi8<0xC2, MRMSrcReg,
- (outs RC:$dst), (ins RC:$src1, RC:$src2, i8imm:$cc),
- asm_alt, [], IIC_DEFAULT, d>;
- def rmi_alt : PIi8<0xC2, MRMSrcMem,
- (outs RC:$dst), (ins RC:$src1, x86memop:$src2, i8imm:$cc),
- asm_alt, [], IIC_DEFAULT, d>;
+ let neverHasSideEffects = 1 in {
+ def rri_alt : PIi8<0xC2, MRMSrcReg,
+ (outs RC:$dst), (ins RC:$src1, RC:$src2, i8imm:$cc),
+ asm_alt, [], IIC_SSE_CMPP_RR, d>;
+ def rmi_alt : PIi8<0xC2, MRMSrcMem,
+ (outs RC:$dst), (ins RC:$src1, x86memop:$src2, i8imm:$cc),
+ asm_alt, [], IIC_SSE_CMPP_RM, d>;
+ }
}
-defm VCMPPS : sse12_cmp_packed<VR128, f128mem, int_x86_sse_cmp_ps,
+defm VCMPPS : sse12_cmp_packed<VR128, f128mem, AVXCC, int_x86_sse_cmp_ps,
"cmp${cc}ps\t{$src2, $src1, $dst|$dst, $src1, $src2}",
"cmpps\t{$cc, $src2, $src1, $dst|$dst, $src1, $src2, $cc}",
SSEPackedSingle>, TB, VEX_4V;
-defm VCMPPD : sse12_cmp_packed<VR128, f128mem, int_x86_sse2_cmp_pd,
+defm VCMPPD : sse12_cmp_packed<VR128, f128mem, AVXCC, int_x86_sse2_cmp_pd,
"cmp${cc}pd\t{$src2, $src1, $dst|$dst, $src1, $src2}",
"cmppd\t{$cc, $src2, $src1, $dst|$dst, $src1, $src2, $cc}",
SSEPackedDouble>, TB, OpSize, VEX_4V;
-defm VCMPPSY : sse12_cmp_packed<VR256, f256mem, int_x86_avx_cmp_ps_256,
+defm VCMPPSY : sse12_cmp_packed<VR256, f256mem, AVXCC, int_x86_avx_cmp_ps_256,
"cmp${cc}ps\t{$src2, $src1, $dst|$dst, $src1, $src2}",
"cmpps\t{$cc, $src2, $src1, $dst|$dst, $src1, $src2, $cc}",
SSEPackedSingle>, TB, VEX_4V;
-defm VCMPPDY : sse12_cmp_packed<VR256, f256mem, int_x86_avx_cmp_pd_256,
+defm VCMPPDY : sse12_cmp_packed<VR256, f256mem, AVXCC, int_x86_avx_cmp_pd_256,
"cmp${cc}pd\t{$src2, $src1, $dst|$dst, $src1, $src2}",
"cmppd\t{$cc, $src2, $src1, $dst|$dst, $src1, $src2, $cc}",
SSEPackedDouble>, TB, OpSize, VEX_4V;
let Constraints = "$src1 = $dst" in {
- defm CMPPS : sse12_cmp_packed<VR128, f128mem, int_x86_sse_cmp_ps,
+ defm CMPPS : sse12_cmp_packed<VR128, f128mem, SSECC, int_x86_sse_cmp_ps,
"cmp${cc}ps\t{$src2, $dst|$dst, $src2}",
"cmpps\t{$cc, $src2, $dst|$dst, $src2, $cc}",
SSEPackedSingle>, TB;
- defm CMPPD : sse12_cmp_packed<VR128, f128mem, int_x86_sse2_cmp_pd,
+ defm CMPPD : sse12_cmp_packed<VR128, f128mem, SSECC, int_x86_sse2_cmp_pd,
"cmp${cc}pd\t{$src2, $dst|$dst, $src2}",
"cmppd\t{$cc, $src2, $dst|$dst, $src2, $cc}",
SSEPackedDouble>, TB, OpSize;
(VCMPPDYrmi VR256:$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:$cc)),
def rmi : PIi8<0xC6, MRMSrcMem, (outs RC:$dst),
(ins RC:$src1, x86memop:$src2, i8imm:$src3), asm,
[(set RC:$dst, (vt (X86Shufp RC:$src1, (mem_frag addr:$src2),
- (i8 imm:$src3))))], IIC_DEFAULT, d>;
+ (i8 imm:$src3))))], IIC_SSE_SHUFP, d>;
let isConvertibleToThreeAddress = IsConvertibleToThreeAddress in
def rri : PIi8<0xC6, MRMSrcReg, (outs RC:$dst),
(ins RC:$src1, RC:$src2, i8imm:$src3), asm,
[(set RC:$dst, (vt (X86Shufp RC:$src1, RC:$src2,
- (i8 imm:$src3))))], IIC_DEFAULT, d>;
+ (i8 imm:$src3))))], IIC_SSE_SHUFP, d>;
}
defm VSHUFPS : sse12_shuffle<VR128, f128mem, v4f32,
(VSHUFPDYrmi VR256:$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))),
(outs RC:$dst), (ins RC:$src1, RC:$src2),
asm, [(set RC:$dst,
(vt (OpNode RC:$src1, RC:$src2)))],
- IIC_DEFAULT, d>;
+ IIC_SSE_UNPCK, d>;
def rm : PI<opc, MRMSrcMem,
(outs RC:$dst), (ins RC:$src1, x86memop:$src2),
asm, [(set RC:$dst,
(vt (OpNode RC:$src1,
(mem_frag addr:$src2))))],
- IIC_DEFAULT, d>;
+ IIC_SSE_UNPCK, d>;
}
defm VUNPCKHPS: 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:$src)>;
}
-let Predicates = [HasSSE2] in {
+let Predicates = [UseSSE2] in {
// FIXME: Instead of X86Movddup, there should be a X86Unpckl here, the
// problem is during lowering, where it's not possible to recognize the load
// fold cause it has two uses through a bitcast. One use disappears at isel
Domain d> {
def rr32 : PI<0x50, MRMSrcReg, (outs GR32:$dst), (ins RC:$src),
!strconcat(asm, "\t{$src, $dst|$dst, $src}"),
- [(set GR32:$dst, (Int RC:$src))], IIC_DEFAULT, d>;
+ [(set GR32:$dst, (Int RC:$src))], IIC_SSE_MOVMSK, d>;
def rr64 : PI<0x50, MRMSrcReg, (outs GR64:$dst), (ins RC:$src),
!strconcat(asm, "\t{$src, $dst|$dst, $src}"), [],
- IIC_DEFAULT, d>, REX_W;
+ IIC_SSE_MOVMSK, d>, REX_W;
}
let Predicates = [HasAVX] in {
OpSize, VEX;
def : Pat<(i32 (X86fgetsign FR32:$src)),
- (VMOVMSKPSrr32 (INSERT_SUBREG (v4f32 (IMPLICIT_DEF)), FR32:$src,
- sub_ss))>;
+ (VMOVMSKPSrr32 (COPY_TO_REGCLASS FR32:$src, VR128))>;
def : Pat<(i64 (X86fgetsign FR32:$src)),
- (VMOVMSKPSrr64 (INSERT_SUBREG (v4f32 (IMPLICIT_DEF)), FR32:$src,
- sub_ss))>;
+ (VMOVMSKPSrr64 (COPY_TO_REGCLASS FR32:$src, VR128))>;
def : Pat<(i32 (X86fgetsign FR64:$src)),
- (VMOVMSKPDrr32 (INSERT_SUBREG (v2f64 (IMPLICIT_DEF)), FR64:$src,
- sub_sd))>;
+ (VMOVMSKPDrr32 (COPY_TO_REGCLASS FR64:$src, VR128))>;
def : Pat<(i64 (X86fgetsign FR64:$src)),
- (VMOVMSKPDrr64 (INSERT_SUBREG (v2f64 (IMPLICIT_DEF)), FR64:$src,
- sub_sd))>;
+ (VMOVMSKPDrr64 (COPY_TO_REGCLASS FR64:$src, VR128))>;
// Assembler Only
def VMOVMSKPSr64r : PI<0x50, MRMSrcReg, (outs GR64:$dst), (ins VR128:$src),
- "movmskps\t{$src, $dst|$dst, $src}", [], IIC_DEFAULT,
+ "movmskps\t{$src, $dst|$dst, $src}", [], IIC_SSE_MOVMSK,
SSEPackedSingle>, TB, VEX;
def VMOVMSKPDr64r : PI<0x50, MRMSrcReg, (outs GR64:$dst), (ins VR128:$src),
- "movmskpd\t{$src, $dst|$dst, $src}", [], IIC_DEFAULT,
+ "movmskpd\t{$src, $dst|$dst, $src}", [], IIC_SSE_MOVMSK,
SSEPackedDouble>, TB,
OpSize, VEX;
def VMOVMSKPSYr64r : PI<0x50, MRMSrcReg, (outs GR64:$dst), (ins VR256:$src),
- "movmskps\t{$src, $dst|$dst, $src}", [], IIC_DEFAULT,
+ "movmskps\t{$src, $dst|$dst, $src}", [], IIC_SSE_MOVMSK,
SSEPackedSingle>, TB, VEX;
def VMOVMSKPDYr64r : PI<0x50, MRMSrcReg, (outs GR64:$dst), (ins VR256:$src),
- "movmskpd\t{$src, $dst|$dst, $src}", [], IIC_DEFAULT,
+ "movmskpd\t{$src, $dst|$dst, $src}", [], IIC_SSE_MOVMSK,
SSEPackedDouble>, TB,
OpSize, VEX;
}
SSEPackedDouble>, TB, OpSize;
def : Pat<(i32 (X86fgetsign FR32:$src)),
- (MOVMSKPSrr32 (INSERT_SUBREG (v4f32 (IMPLICIT_DEF)), FR32:$src,
- sub_ss))>, Requires<[HasSSE1]>;
+ (MOVMSKPSrr32 (COPY_TO_REGCLASS FR32:$src, VR128))>,
+ Requires<[UseSSE1]>;
def : Pat<(i64 (X86fgetsign FR32:$src)),
- (MOVMSKPSrr64 (INSERT_SUBREG (v4f32 (IMPLICIT_DEF)), FR32:$src,
- sub_ss))>, Requires<[HasSSE1]>;
+ (MOVMSKPSrr64 (COPY_TO_REGCLASS FR32:$src, VR128))>,
+ Requires<[UseSSE1]>;
def : Pat<(i32 (X86fgetsign FR64:$src)),
- (MOVMSKPDrr32 (INSERT_SUBREG (v2f64 (IMPLICIT_DEF)), FR64:$src,
- sub_sd))>, Requires<[HasSSE2]>;
+ (MOVMSKPDrr32 (COPY_TO_REGCLASS FR64:$src, VR128))>,
+ Requires<[UseSSE2]>;
def : Pat<(i64 (X86fgetsign FR64:$src)),
- (MOVMSKPDrr64 (INSERT_SUBREG (v2f64 (IMPLICIT_DEF)), FR64:$src,
- sub_sd))>, Requires<[HasSSE2]>;
+ (MOVMSKPDrr64 (COPY_TO_REGCLASS FR64:$src, VR128))>,
+ Requires<[UseSSE2]>;
//===---------------------------------------------------------------------===//
// SSE2 - Packed Integer Logical Instructions
/// PDI_binop_rm - Simple SSE2 binary operator.
multiclass PDI_binop_rm<bits<8> opc, string OpcodeStr, SDNode OpNode,
ValueType OpVT, RegisterClass RC, PatFrag memop_frag,
- X86MemOperand x86memop, bit IsCommutable = 0,
+ X86MemOperand x86memop,
+ OpndItins itins,
+ bit IsCommutable = 0,
bit Is2Addr = 1> {
let isCommutable = IsCommutable in
def rr : PDI<opc, MRMSrcReg, (outs RC:$dst),
!if(Is2Addr,
!strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"),
!strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}")),
- [(set RC:$dst, (OpVT (OpNode RC:$src1, RC:$src2)))]>;
+ [(set RC:$dst, (OpVT (OpNode RC:$src1, RC:$src2)))], itins.rr>;
def rm : PDI<opc, MRMSrcMem, (outs RC:$dst),
(ins RC:$src1, x86memop:$src2),
!if(Is2Addr,
!strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"),
!strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}")),
[(set RC:$dst, (OpVT (OpNode RC:$src1,
- (bitconvert (memop_frag addr:$src2)))))]>;
+ (bitconvert (memop_frag addr:$src2)))))],
+ itins.rm>;
}
} // ExeDomain = SSEPackedInt
let Predicates = [HasAVX] in {
defm VPAND : PDI_binop_rm<0xDB, "vpand", and, v2i64, VR128, memopv2i64,
- i128mem, 1, 0>, VEX_4V;
+ i128mem, SSE_BIT_ITINS_P, 1, 0>, VEX_4V;
defm VPOR : PDI_binop_rm<0xEB, "vpor" , or, v2i64, VR128, memopv2i64,
- i128mem, 1, 0>, VEX_4V;
+ i128mem, SSE_BIT_ITINS_P, 1, 0>, VEX_4V;
defm VPXOR : PDI_binop_rm<0xEF, "vpxor", xor, v2i64, VR128, memopv2i64,
- i128mem, 1, 0>, VEX_4V;
+ i128mem, SSE_BIT_ITINS_P, 1, 0>, VEX_4V;
defm VPANDN : PDI_binop_rm<0xDF, "vpandn", X86andnp, v2i64, VR128, memopv2i64,
- i128mem, 0, 0>, VEX_4V;
+ i128mem, SSE_BIT_ITINS_P, 0, 0>, VEX_4V;
}
let Constraints = "$src1 = $dst" in {
defm PAND : PDI_binop_rm<0xDB, "pand", and, v2i64, VR128, memopv2i64,
- i128mem, 1>;
+ i128mem, SSE_BIT_ITINS_P, 1>;
defm POR : PDI_binop_rm<0xEB, "por" , or, v2i64, VR128, memopv2i64,
- i128mem, 1>;
+ i128mem, SSE_BIT_ITINS_P, 1>;
defm PXOR : PDI_binop_rm<0xEF, "pxor", xor, v2i64, VR128, memopv2i64,
- i128mem, 1>;
+ i128mem, SSE_BIT_ITINS_P, 1>;
defm PANDN : PDI_binop_rm<0xDF, "pandn", X86andnp, v2i64, VR128, memopv2i64,
- i128mem, 0>;
+ i128mem, SSE_BIT_ITINS_P, 0>;
} // Constraints = "$src1 = $dst"
let Predicates = [HasAVX2] in {
defm VPANDY : PDI_binop_rm<0xDB, "vpand", and, v4i64, VR256, memopv4i64,
- i256mem, 1, 0>, VEX_4V;
+ i256mem, SSE_BIT_ITINS_P, 1, 0>, VEX_4V;
defm VPORY : PDI_binop_rm<0xEB, "vpor", or, v4i64, VR256, memopv4i64,
- i256mem, 1, 0>, VEX_4V;
+ i256mem, SSE_BIT_ITINS_P, 1, 0>, VEX_4V;
defm VPXORY : PDI_binop_rm<0xEF, "vpxor", xor, v4i64, VR256, memopv4i64,
- i256mem, 1, 0>, VEX_4V;
+ i256mem, SSE_BIT_ITINS_P, 1, 0>, VEX_4V;
defm VPANDNY : PDI_binop_rm<0xDF, "vpandn", X86andnp, v4i64, VR256, memopv4i64,
- i256mem, 0, 0>, VEX_4V;
+ i256mem, SSE_BIT_ITINS_P, 0, 0>, VEX_4V;
}
//===----------------------------------------------------------------------===//
/// sse12_fp_alias_pack_logical - SSE 1 & 2 aliased packed FP logical ops
///
multiclass sse12_fp_alias_pack_logical<bits<8> opc, string OpcodeStr,
- SDNode OpNode> {
+ SDNode OpNode, OpndItins itins> {
defm V#NAME#PS : sse12_fp_packed<opc, !strconcat(OpcodeStr, "ps"), OpNode,
- FR32, f32, f128mem, memopfsf32, SSEPackedSingle, 0>, TB, VEX_4V;
+ FR32, f32, f128mem, memopfsf32, SSEPackedSingle, itins, 0>,
+ TB, VEX_4V;
defm V#NAME#PD : sse12_fp_packed<opc, !strconcat(OpcodeStr, "pd"), OpNode,
- FR64, f64, f128mem, memopfsf64, SSEPackedDouble, 0>, TB, OpSize, VEX_4V;
+ FR64, f64, f128mem, memopfsf64, SSEPackedDouble, itins, 0>,
+ TB, OpSize, VEX_4V;
let Constraints = "$src1 = $dst" in {
defm PS : sse12_fp_packed<opc, !strconcat(OpcodeStr, "ps"), OpNode, FR32,
- f32, f128mem, memopfsf32, SSEPackedSingle>, TB;
+ f32, f128mem, memopfsf32, SSEPackedSingle, itins>,
+ TB;
defm PD : sse12_fp_packed<opc, !strconcat(OpcodeStr, "pd"), OpNode, FR64,
- f64, f128mem, memopfsf64, SSEPackedDouble>, TB, OpSize;
+ f64, f128mem, memopfsf64, SSEPackedDouble, itins>,
+ TB, OpSize;
}
}
// 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>;
- defm FsOR : sse12_fp_alias_pack_logical<0x56, "or", X86for>;
- defm FsXOR : sse12_fp_alias_pack_logical<0x57, "xor", X86fxor>;
-}
+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>;
+ defm FsANDN : sse12_fp_alias_pack_logical<0x55, "andn", undef,
+ SSE_BIT_ITINS_P>;
/// sse12_fp_packed_logical - SSE 1 & 2 packed FP logical ops
///
/// FIXME: once all 256-bit intrinsics are matched, cleanup and refactor those
/// classes below
multiclass basic_sse12_fp_binop_s<bits<8> opc, string OpcodeStr, SDNode OpNode,
+ SizeItins itins,
bit Is2Addr = 1> {
defm SS : sse12_fp_scalar<opc, !strconcat(OpcodeStr, "ss"),
- OpNode, FR32, f32mem, Is2Addr>, XS;
+ OpNode, FR32, f32mem,
+ itins.s, Is2Addr>, XS;
defm SD : sse12_fp_scalar<opc, !strconcat(OpcodeStr, "sd"),
- OpNode, FR64, f64mem, Is2Addr>, XD;
+ OpNode, FR64, f64mem,
+ 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 {
defm PS : sse12_fp_packed<opc, !strconcat(OpcodeStr, "ps"), OpNode, VR128,
- v4f32, f128mem, memopv4f32, SSEPackedSingle, Is2Addr>, TB;
+ v4f32, f128mem, memopv4f32, SSEPackedSingle, itins.s, Is2Addr>,
+ TB;
defm PD : sse12_fp_packed<opc, !strconcat(OpcodeStr, "pd"), OpNode, VR128,
- v2f64, f128mem, memopv2f64, SSEPackedDouble, Is2Addr>, TB, OpSize;
- }
+ v2f64, f128mem, memopv2f64, SSEPackedDouble, itins.d, Is2Addr>,
+ TB, OpSize;
}
multiclass basic_sse12_fp_binop_p_y<bits<8> opc, string OpcodeStr,
- SDNode OpNode> {
- let mayLoad = 0 in {
- defm PSY : sse12_fp_packed<opc, !strconcat(OpcodeStr, "ps"), OpNode, VR256,
- v8f32, f256mem, memopv8f32, SSEPackedSingle, 0>, TB;
- defm PDY : sse12_fp_packed<opc, !strconcat(OpcodeStr, "pd"), OpNode, VR256,
- v4f64, f256mem, memopv4f64, SSEPackedDouble, 0>, TB, OpSize;
- }
+ SDNode OpNode,
+ SizeItins itins> {
+ 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,
+ SizeItins itins,
bit Is2Addr = 1> {
defm SS : sse12_fp_scalar_int<opc, OpcodeStr, VR128,
- !strconcat(OpcodeStr, "ss"), "", "_ss", ssmem, sse_load_f32, Is2Addr>, XS;
+ !strconcat(OpcodeStr, "ss"), "", "_ss", ssmem, sse_load_f32,
+ itins.s, Is2Addr>, XS;
defm SD : sse12_fp_scalar_int<opc, OpcodeStr, VR128,
- !strconcat(OpcodeStr, "sd"), "2", "_sd", sdmem, sse_load_f64, Is2Addr>, XD;
+ !strconcat(OpcodeStr, "sd"), "2", "_sd", sdmem, sse_load_f64,
+ 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, Is2Addr>, TB;
+ SSEPackedSingle, itins.s, Is2Addr>,
+ TB;
defm PD : sse12_fp_packed_int<opc, OpcodeStr, VR128,
!strconcat(OpcodeStr, "pd"), "sse2", "_pd", f128mem, memopv2f64,
- SSEPackedDouble, Is2Addr>, TB, OpSize;
+ SSEPackedDouble, itins.d, Is2Addr>,
+ TB, OpSize;
}
-multiclass basic_sse12_fp_binop_p_y_int<bits<8> opc, string OpcodeStr> {
+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, 0>, TB;
+ SSEPackedSingle, itins.s, 0>, TB;
defm PDY : sse12_fp_packed_int<opc, OpcodeStr, VR256,
!strconcat(OpcodeStr, "pd"), "avx", "_pd_256", f256mem, memopv4f64,
- SSEPackedDouble, 0>, TB, OpSize;
+ SSEPackedDouble, itins.d, 0>, TB, OpSize;
}
// Binary Arithmetic instructions
-defm VADD : basic_sse12_fp_binop_s<0x58, "add", fadd, 0>,
- basic_sse12_fp_binop_s_int<0x58, "add", 0>, VEX_4V, VEX_LIG;
-defm VADD : basic_sse12_fp_binop_p<0x58, "add", fadd, 0>,
- basic_sse12_fp_binop_p_y<0x58, "add", fadd>, VEX_4V;
-defm VMUL : basic_sse12_fp_binop_s<0x59, "mul", fmul, 0>,
- basic_sse12_fp_binop_s_int<0x59, "mul", 0>, VEX_4V, VEX_LIG;
-defm VMUL : basic_sse12_fp_binop_p<0x59, "mul", fmul, 0>,
- basic_sse12_fp_binop_p_y<0x59, "mul", fmul>, VEX_4V;
+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, 0>,
- basic_sse12_fp_binop_s_int<0x5C, "sub", 0>, VEX_4V, VEX_LIG;
- defm VSUB : basic_sse12_fp_binop_p<0x5C, "sub", fsub, 0>,
- basic_sse12_fp_binop_p_y<0x5C, "sub", fsub>, VEX_4V;
- defm VDIV : basic_sse12_fp_binop_s<0x5E, "div", fdiv, 0>,
- basic_sse12_fp_binop_s_int<0x5E, "div", 0>, VEX_4V, VEX_LIG;
- defm VDIV : basic_sse12_fp_binop_p<0x5E, "div", fdiv, 0>,
- basic_sse12_fp_binop_p_y<0x5E, "div", fdiv>, VEX_4V;
- defm VMAX : basic_sse12_fp_binop_s<0x5F, "max", X86fmax, 0>,
- basic_sse12_fp_binop_s_int<0x5F, "max", 0>, VEX_4V, VEX_LIG;
- defm VMAX : basic_sse12_fp_binop_p<0x5F, "max", X86fmax, 0>,
- basic_sse12_fp_binop_p_int<0x5F, "max", 0>,
- basic_sse12_fp_binop_p_y<0x5F, "max", X86fmax>,
- basic_sse12_fp_binop_p_y_int<0x5F, "max">, VEX_4V;
- defm VMIN : basic_sse12_fp_binop_s<0x5D, "min", X86fmin, 0>,
- basic_sse12_fp_binop_s_int<0x5D, "min", 0>, VEX_4V, VEX_LIG;
- defm VMIN : basic_sse12_fp_binop_p<0x5D, "min", X86fmin, 0>,
- basic_sse12_fp_binop_p_int<0x5D, "min", 0>,
- basic_sse12_fp_binop_p_y_int<0x5D, "min">,
- basic_sse12_fp_binop_p_y<0x5D, "min", X86fmin>, VEX_4V;
+ 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>,
- basic_sse12_fp_binop_p<0x58, "add", fadd>,
- basic_sse12_fp_binop_s_int<0x58, "add">;
- defm MUL : basic_sse12_fp_binop_s<0x59, "mul", fmul>,
- basic_sse12_fp_binop_p<0x59, "mul", fmul>,
- basic_sse12_fp_binop_s_int<0x59, "mul">;
+ 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>,
- basic_sse12_fp_binop_p<0x5C, "sub", fsub>,
- basic_sse12_fp_binop_s_int<0x5C, "sub">;
- defm DIV : basic_sse12_fp_binop_s<0x5E, "div", fdiv>,
- basic_sse12_fp_binop_p<0x5E, "div", fdiv>,
- basic_sse12_fp_binop_s_int<0x5E, "div">;
- defm MAX : basic_sse12_fp_binop_s<0x5F, "max", X86fmax>,
- basic_sse12_fp_binop_p<0x5F, "max", X86fmax>,
- basic_sse12_fp_binop_s_int<0x5F, "max">,
- basic_sse12_fp_binop_p_int<0x5F, "max">;
- defm MIN : basic_sse12_fp_binop_s<0x5D, "min", X86fmin>,
- basic_sse12_fp_binop_p<0x5D, "min", X86fmin>,
- basic_sse12_fp_binop_s_int<0x5D, "min">,
- basic_sse12_fp_binop_p_int<0x5D, "min">;
+ 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>;
+ 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>;
+ }
+}
+
+let isCodeGenOnly = 1 in {
+ defm VMAXC: basic_sse12_fp_binop_s<0x5F, "max", X86fmaxc, SSE_ALU_ITINS_S, 0>,
+ VEX_4V, VEX_LIG;
+ defm VMAXC: basic_sse12_fp_binop_p<0x5F, "max", X86fmaxc, SSE_ALU_ITINS_P, 0>,
+ basic_sse12_fp_binop_p_y<0x5F, "max", X86fmaxc, SSE_ALU_ITINS_P>, VEX_4V;
+ defm VMINC: basic_sse12_fp_binop_s<0x5D, "min", X86fminc, SSE_ALU_ITINS_S, 0>,
+ VEX_4V, VEX_LIG;
+ defm VMINC: basic_sse12_fp_binop_p<0x5D, "min", X86fminc, SSE_ALU_ITINS_P, 0>,
+ basic_sse12_fp_binop_p_y<0x5D, "min", X86fminc, SSE_ALU_ITINS_P>, VEX_4V;
+ let Constraints = "$src1 = $dst" in {
+ defm MAXC: basic_sse12_fp_binop_s<0x5F, "max", X86fmaxc, SSE_ALU_ITINS_S>,
+ basic_sse12_fp_binop_p<0x5F, "max", X86fmaxc, SSE_ALU_ITINS_P>;
+ defm MINC: basic_sse12_fp_binop_s<0x5D, "min", X86fminc, SSE_ALU_ITINS_S>,
+ basic_sse12_fp_binop_p<0x5D, "min", X86fminc, SSE_ALU_ITINS_P>;
}
}
///
/// And, we have a special variant form for a full-vector intrinsic form.
+def SSE_SQRTP : OpndItins<
+ IIC_SSE_SQRTP_RR, IIC_SSE_SQRTP_RM
+>;
+
+def SSE_SQRTS : OpndItins<
+ IIC_SSE_SQRTS_RR, IIC_SSE_SQRTS_RM
+>;
+
+def SSE_RCPP : OpndItins<
+ IIC_SSE_RCPP_RR, IIC_SSE_RCPP_RM
+>;
+
+def SSE_RCPS : OpndItins<
+ IIC_SSE_RCPS_RR, IIC_SSE_RCPS_RM
+>;
+
/// sse1_fp_unop_s - SSE1 unops in scalar form.
multiclass sse1_fp_unop_s<bits<8> opc, string OpcodeStr,
- SDNode OpNode, Intrinsic F32Int> {
+ SDNode OpNode, Intrinsic F32Int, OpndItins itins> {
def SSr : SSI<opc, MRMSrcReg, (outs FR32:$dst), (ins FR32:$src),
!strconcat(OpcodeStr, "ss\t{$src, $dst|$dst, $src}"),
[(set FR32:$dst, (OpNode FR32:$src))]>;
// partial register update condition.
def SSm : I<opc, MRMSrcMem, (outs FR32:$dst), (ins f32mem:$src),
!strconcat(OpcodeStr, "ss\t{$src, $dst|$dst, $src}"),
- [(set FR32:$dst, (OpNode (load addr:$src)))]>, XS,
- Requires<[HasSSE1, OptForSize]>;
+ [(set FR32:$dst, (OpNode (load addr:$src)))], itins.rm>, XS,
+ Requires<[UseSSE1, OptForSize]>;
def SSr_Int : SSI<opc, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
!strconcat(OpcodeStr, "ss\t{$src, $dst|$dst, $src}"),
- [(set VR128:$dst, (F32Int VR128:$src))]>;
+ [(set VR128:$dst, (F32Int VR128:$src))], itins.rr>;
def SSm_Int : SSI<opc, MRMSrcMem, (outs VR128:$dst), (ins ssmem:$src),
!strconcat(OpcodeStr, "ss\t{$src, $dst|$dst, $src}"),
- [(set VR128:$dst, (F32Int sse_load_f32:$src))]>;
+ [(set VR128:$dst, (F32Int sse_load_f32:$src))], itins.rm>;
}
/// sse1_fp_unop_s_avx - AVX SSE1 unops in scalar form.
def SSr : SSI<opc, MRMSrcReg, (outs FR32:$dst), (ins FR32:$src1, FR32:$src2),
!strconcat(OpcodeStr,
"ss\t{$src2, $src1, $dst|$dst, $src1, $src2}"), []>;
- let mayLoad = 1 in
+ let mayLoad = 1 in {
def SSm : SSI<opc, MRMSrcMem, (outs FR32:$dst), (ins FR32:$src1,f32mem:$src2),
!strconcat(OpcodeStr,
"ss\t{$src2, $src1, $dst|$dst, $src1, $src2}"), []>;
(ins VR128:$src1, ssmem:$src2),
!strconcat(OpcodeStr,
"ss\t{$src2, $src1, $dst|$dst, $src1, $src2}"), []>;
+ }
}
/// sse1_fp_unop_p - SSE1 unops in packed form.
-multiclass sse1_fp_unop_p<bits<8> opc, string OpcodeStr, SDNode OpNode> {
+multiclass sse1_fp_unop_p<bits<8> opc, string OpcodeStr, SDNode OpNode,
+ OpndItins itins> {
def PSr : PSI<opc, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
!strconcat(OpcodeStr, "ps\t{$src, $dst|$dst, $src}"),
- [(set VR128:$dst, (v4f32 (OpNode VR128:$src)))]>;
+ [(set VR128:$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:$src)))]>;
+ [(set VR128:$dst, (OpNode (memopv4f32 addr:$src)))], itins.rm>;
}
/// sse1_fp_unop_p_y - AVX 256-bit SSE1 unops in packed form.
-multiclass sse1_fp_unop_p_y<bits<8> opc, string OpcodeStr, SDNode OpNode> {
+multiclass sse1_fp_unop_p_y<bits<8> opc, string OpcodeStr, SDNode OpNode,
+ OpndItins itins> {
def PSYr : PSI<opc, MRMSrcReg, (outs VR256:$dst), (ins VR256:$src),
!strconcat(OpcodeStr, "ps\t{$src, $dst|$dst, $src}"),
- [(set VR256:$dst, (v8f32 (OpNode VR256:$src)))]>;
+ [(set VR256:$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:$src)))]>;
+ [(set VR256:$dst, (OpNode (memopv8f32 addr:$src)))],
+ itins.rm>;
}
/// sse1_fp_unop_p_int - SSE1 intrinsics unops in packed forms.
multiclass sse1_fp_unop_p_int<bits<8> opc, string OpcodeStr,
- Intrinsic V4F32Int> {
+ Intrinsic V4F32Int, OpndItins itins> {
def PSr_Int : PSI<opc, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
!strconcat(OpcodeStr, "ps\t{$src, $dst|$dst, $src}"),
- [(set VR128:$dst, (V4F32Int VR128:$src))]>;
+ [(set VR128:$dst, (V4F32Int VR128:$src))],
+ itins.rr>;
def PSm_Int : PSI<opc, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
!strconcat(OpcodeStr, "ps\t{$src, $dst|$dst, $src}"),
- [(set VR128:$dst, (V4F32Int (memopv4f32 addr:$src)))]>;
+ [(set VR128:$dst, (V4F32Int (memopv4f32 addr:$src)))],
+ itins.rm>;
}
/// sse1_fp_unop_p_y_int - AVX 256-bit intrinsics unops in packed forms.
multiclass sse1_fp_unop_p_y_int<bits<8> opc, string OpcodeStr,
- Intrinsic V4F32Int> {
+ Intrinsic V4F32Int, OpndItins itins> {
def PSYr_Int : PSI<opc, MRMSrcReg, (outs VR256:$dst), (ins VR256:$src),
!strconcat(OpcodeStr, "ps\t{$src, $dst|$dst, $src}"),
- [(set VR256:$dst, (V4F32Int VR256:$src))]>;
+ [(set VR256:$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:$src)))]>;
+ [(set VR256:$dst, (V4F32Int (memopv8f32 addr:$src)))],
+ itins.rm>;
}
/// sse2_fp_unop_s - SSE2 unops in scalar form.
multiclass sse2_fp_unop_s<bits<8> opc, string OpcodeStr,
- SDNode OpNode, Intrinsic F64Int> {
+ SDNode OpNode, Intrinsic F64Int, OpndItins itins> {
def SDr : SDI<opc, MRMSrcReg, (outs FR64:$dst), (ins FR64:$src),
!strconcat(OpcodeStr, "sd\t{$src, $dst|$dst, $src}"),
- [(set FR64:$dst, (OpNode FR64:$src))]>;
+ [(set FR64:$dst, (OpNode FR64:$src))], itins.rr>;
// See the comments in sse1_fp_unop_s for why this is OptForSize.
def SDm : I<opc, MRMSrcMem, (outs FR64:$dst), (ins f64mem:$src),
!strconcat(OpcodeStr, "sd\t{$src, $dst|$dst, $src}"),
- [(set FR64:$dst, (OpNode (load addr:$src)))]>, XD,
- Requires<[HasSSE2, OptForSize]>;
+ [(set FR64:$dst, (OpNode (load addr:$src)))], itins.rm>, XD,
+ Requires<[UseSSE2, OptForSize]>;
def SDr_Int : SDI<opc, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
!strconcat(OpcodeStr, "sd\t{$src, $dst|$dst, $src}"),
- [(set VR128:$dst, (F64Int VR128:$src))]>;
+ [(set VR128:$dst, (F64Int VR128:$src))], itins.rr>;
def SDm_Int : SDI<opc, MRMSrcMem, (outs VR128:$dst), (ins sdmem:$src),
!strconcat(OpcodeStr, "sd\t{$src, $dst|$dst, $src}"),
- [(set VR128:$dst, (F64Int sse_load_f64:$src))]>;
+ [(set VR128:$dst, (F64Int sse_load_f64:$src))], itins.rm>;
}
/// sse2_fp_unop_s_avx - AVX SSE2 unops in scalar form.
+let hasSideEffects = 0 in
multiclass sse2_fp_unop_s_avx<bits<8> opc, string OpcodeStr> {
- let neverHasSideEffects = 1 in {
def SDr : SDI<opc, MRMSrcReg, (outs FR64:$dst), (ins FR64:$src1, FR64:$src2),
!strconcat(OpcodeStr,
"sd\t{$src2, $src1, $dst|$dst, $src1, $src2}"), []>;
- let mayLoad = 1 in
+ let mayLoad = 1 in {
def SDm : SDI<opc, MRMSrcMem, (outs FR64:$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, $src2}"), []>;
+ }
}
/// sse2_fp_unop_p - SSE2 unops in vector forms.
multiclass sse2_fp_unop_p<bits<8> opc, string OpcodeStr,
- SDNode OpNode> {
+ SDNode OpNode, OpndItins itins> {
def PDr : PDI<opc, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
!strconcat(OpcodeStr, "pd\t{$src, $dst|$dst, $src}"),
- [(set VR128:$dst, (v2f64 (OpNode VR128:$src)))]>;
+ [(set VR128:$dst, (v2f64 (OpNode VR128:$src)))], itins.rr>;
def PDm : PDI<opc, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
!strconcat(OpcodeStr, "pd\t{$src, $dst|$dst, $src}"),
- [(set VR128:$dst, (OpNode (memopv2f64 addr:$src)))]>;
+ [(set VR128:$dst, (OpNode (memopv2f64 addr:$src)))], itins.rm>;
}
/// sse2_fp_unop_p_y - AVX SSE2 256-bit unops in vector forms.
-multiclass sse2_fp_unop_p_y<bits<8> opc, string OpcodeStr, SDNode OpNode> {
+multiclass sse2_fp_unop_p_y<bits<8> opc, string OpcodeStr, SDNode OpNode,
+ OpndItins itins> {
def PDYr : PDI<opc, MRMSrcReg, (outs VR256:$dst), (ins VR256:$src),
!strconcat(OpcodeStr, "pd\t{$src, $dst|$dst, $src}"),
- [(set VR256:$dst, (v4f64 (OpNode VR256:$src)))]>;
+ [(set VR256:$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:$src)))]>;
+ [(set VR256:$dst, (OpNode (memopv4f64 addr:$src)))],
+ itins.rm>;
}
/// sse2_fp_unop_p_int - SSE2 intrinsic unops in vector forms.
multiclass sse2_fp_unop_p_int<bits<8> opc, string OpcodeStr,
- Intrinsic V2F64Int> {
+ Intrinsic V2F64Int, OpndItins itins> {
def PDr_Int : PDI<opc, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
!strconcat(OpcodeStr, "pd\t{$src, $dst|$dst, $src}"),
- [(set VR128:$dst, (V2F64Int VR128:$src))]>;
+ [(set VR128:$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:$src)))]>;
+ [(set VR128:$dst, (V2F64Int (memopv2f64 addr:$src)))],
+ itins.rm>;
}
/// sse2_fp_unop_p_y_int - AVX 256-bit intrinsic unops in vector forms.
multiclass sse2_fp_unop_p_y_int<bits<8> opc, string OpcodeStr,
- Intrinsic V2F64Int> {
+ Intrinsic V2F64Int, OpndItins itins> {
def PDYr_Int : PDI<opc, MRMSrcReg, (outs VR256:$dst), (ins VR256:$src),
!strconcat(OpcodeStr, "pd\t{$src, $dst|$dst, $src}"),
- [(set VR256:$dst, (V2F64Int VR256:$src))]>;
+ [(set VR256:$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:$src)))]>;
+ [(set VR256:$dst, (V2F64Int (memopv4f64 addr:$src)))],
+ itins.rm>;
}
let Predicates = [HasAVX] in {
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>,
- sse2_fp_unop_p<0x51, "vsqrt", fsqrt>,
- sse1_fp_unop_p_y<0x51, "vsqrt", fsqrt>,
- sse2_fp_unop_p_y<0x51, "vsqrt", fsqrt>,
- sse1_fp_unop_p_int<0x51, "vsqrt", int_x86_sse_sqrt_ps>,
- sse2_fp_unop_p_int<0x51, "vsqrt", int_x86_sse2_sqrt_pd>,
- sse1_fp_unop_p_y_int<0x51, "vsqrt", int_x86_avx_sqrt_ps_256>,
- sse2_fp_unop_p_y_int<0x51, "vsqrt", int_x86_avx_sqrt_pd_256>,
+ 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>,
- sse1_fp_unop_p_y<0x52, "vrsqrt", X86frsqrt>,
- sse1_fp_unop_p_y_int<0x52, "vrsqrt", int_x86_avx_rsqrt_ps_256>,
- sse1_fp_unop_p_int<0x52, "vrsqrt", int_x86_sse_rsqrt_ps>, VEX;
+ 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>,
- sse1_fp_unop_p_y<0x53, "vrcp", X86frcp>,
- sse1_fp_unop_p_y_int<0x53, "vrcp", int_x86_avx_rcp_ps_256>,
- sse1_fp_unop_p_int<0x53, "vrcp", int_x86_sse_rcp_ps>, VEX;
+ 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:$src),
- (INSERT_SUBREG (v4f32 (IMPLICIT_DEF)),
- (VSQRTSSr (f32 (IMPLICIT_DEF)),
- (EXTRACT_SUBREG (v4f32 VR128:$src), sub_ss)),
- sub_ss)>;
+ (COPY_TO_REGCLASS (VSQRTSSr (f32 (IMPLICIT_DEF)),
+ (COPY_TO_REGCLASS VR128:$src, FR32)),
+ VR128)>;
def : Pat<(int_x86_sse_sqrt_ss sse_load_f32:$src),
(VSQRTSSm_Int (v4f32 (IMPLICIT_DEF)), sse_load_f32:$src)>;
def : Pat<(int_x86_sse2_sqrt_sd VR128:$src),
- (INSERT_SUBREG (v2f64 (IMPLICIT_DEF)),
- (VSQRTSDr (f64 (IMPLICIT_DEF)),
- (EXTRACT_SUBREG (v2f64 VR128:$src), sub_sd)),
- sub_sd)>;
+ (COPY_TO_REGCLASS (VSQRTSDr (f64 (IMPLICIT_DEF)),
+ (COPY_TO_REGCLASS VR128:$src, FR64)),
+ VR128)>;
def : Pat<(int_x86_sse2_sqrt_sd sse_load_f64:$src),
(VSQRTSDm_Int (v2f64 (IMPLICIT_DEF)), sse_load_f64:$src)>;
def : Pat<(int_x86_sse_rsqrt_ss VR128:$src),
- (INSERT_SUBREG (v4f32 (IMPLICIT_DEF)),
- (VRSQRTSSr (f32 (IMPLICIT_DEF)),
- (EXTRACT_SUBREG (v4f32 VR128:$src), sub_ss)),
- sub_ss)>;
+ (COPY_TO_REGCLASS (VRSQRTSSr (f32 (IMPLICIT_DEF)),
+ (COPY_TO_REGCLASS VR128:$src, FR32)),
+ VR128)>;
def : Pat<(int_x86_sse_rsqrt_ss sse_load_f32:$src),
(VRSQRTSSm_Int (v4f32 (IMPLICIT_DEF)), sse_load_f32:$src)>;
def : Pat<(int_x86_sse_rcp_ss VR128:$src),
- (INSERT_SUBREG (v4f32 (IMPLICIT_DEF)),
- (VRCPSSr (f32 (IMPLICIT_DEF)),
- (EXTRACT_SUBREG (v4f32 VR128:$src), sub_ss)),
- sub_ss)>;
+ (COPY_TO_REGCLASS (VRCPSSr (f32 (IMPLICIT_DEF)),
+ (COPY_TO_REGCLASS VR128:$src, FR32)),
+ VR128)>;
def : Pat<(int_x86_sse_rcp_ss sse_load_f32:$src),
(VRCPSSm_Int (v4f32 (IMPLICIT_DEF)), sse_load_f32:$src)>;
}
// Square root.
-defm SQRT : sse1_fp_unop_s<0x51, "sqrt", fsqrt, int_x86_sse_sqrt_ss>,
- sse1_fp_unop_p<0x51, "sqrt", fsqrt>,
- sse1_fp_unop_p_int<0x51, "sqrt", int_x86_sse_sqrt_ps>,
- sse2_fp_unop_s<0x51, "sqrt", fsqrt, int_x86_sse2_sqrt_sd>,
- sse2_fp_unop_p<0x51, "sqrt", fsqrt>,
- sse2_fp_unop_p_int<0x51, "sqrt", int_x86_sse2_sqrt_pd>;
+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>,
- sse1_fp_unop_p<0x52, "rsqrt", X86frsqrt>,
- sse1_fp_unop_p_int<0x52, "rsqrt", int_x86_sse_rsqrt_ps>;
-defm RCP : sse1_fp_unop_s<0x53, "rcp", X86frcp, int_x86_sse_rcp_ss>,
- sse1_fp_unop_p<0x53, "rcp", X86frcp>,
- sse1_fp_unop_p_int<0x53, "rcp", int_x86_sse_rcp_ps>;
+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>;
// There is no f64 version of the reciprocal approximation instructions.
(ins f128mem:$dst, VR128:$src),
"movntps\t{$src, $dst|$dst, $src}",
[(alignednontemporalstore (v4f32 VR128:$src),
- addr:$dst)]>, VEX;
+ addr:$dst)],
+ IIC_SSE_MOVNT>, VEX;
def VMOVNTPDmr : VPDI<0x2B, MRMDestMem, (outs),
(ins f128mem:$dst, VR128:$src),
"movntpd\t{$src, $dst|$dst, $src}",
[(alignednontemporalstore (v2f64 VR128:$src),
- addr:$dst)]>, VEX;
+ addr:$dst)],
+ IIC_SSE_MOVNT>, VEX;
let ExeDomain = SSEPackedInt in
def VMOVNTDQmr : VPDI<0xE7, MRMDestMem, (outs),
(ins f128mem:$dst, VR128:$src),
"movntdq\t{$src, $dst|$dst, $src}",
[(alignednontemporalstore (v2i64 VR128:$src),
- addr:$dst)]>, VEX;
+ addr:$dst)],
+ IIC_SSE_MOVNT>, VEX;
def : Pat<(alignednontemporalstore (v2i64 VR128:$src), addr:$dst),
(VMOVNTDQmr addr:$dst, VR128:$src)>, Requires<[HasAVX]>;
(ins f256mem:$dst, VR256:$src),
"movntps\t{$src, $dst|$dst, $src}",
[(alignednontemporalstore (v8f32 VR256:$src),
- addr:$dst)]>, VEX;
+ addr:$dst)],
+ IIC_SSE_MOVNT>, VEX;
def VMOVNTPDYmr : VPDI<0x2B, MRMDestMem, (outs),
(ins f256mem:$dst, VR256:$src),
"movntpd\t{$src, $dst|$dst, $src}",
[(alignednontemporalstore (v4f64 VR256:$src),
- addr:$dst)]>, VEX;
+ addr:$dst)],
+ IIC_SSE_MOVNT>, VEX;
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:$dst)]>, VEX;
+ addr:$dst)],
+ IIC_SSE_MOVNT>, VEX;
}
-def : Pat<(int_x86_avx_movnt_dq_256 addr:$dst, VR256:$src),
- (VMOVNTDQYmr addr:$dst, VR256:$src)>;
-def : Pat<(int_x86_avx_movnt_pd_256 addr:$dst, VR256:$src),
- (VMOVNTPDYmr addr:$dst, VR256:$src)>;
-def : Pat<(int_x86_avx_movnt_ps_256 addr:$dst, VR256:$src),
- (VMOVNTPSYmr addr:$dst, VR256:$src)>;
-
let AddedComplexity = 400 in { // Prefer non-temporal versions
def MOVNTPSmr : PSI<0x2B, MRMDestMem, (outs), (ins f128mem:$dst, VR128:$src),
"movntps\t{$src, $dst|$dst, $src}",
- [(alignednontemporalstore (v4f32 VR128:$src), addr:$dst)]>;
+ [(alignednontemporalstore (v4f32 VR128:$src), addr:$dst)],
+ IIC_SSE_MOVNT>;
def MOVNTPDmr : PDI<0x2B, MRMDestMem, (outs), (ins f128mem:$dst, VR128:$src),
"movntpd\t{$src, $dst|$dst, $src}",
- [(alignednontemporalstore(v2f64 VR128:$src), addr:$dst)]>;
+ [(alignednontemporalstore(v2f64 VR128:$src), addr:$dst)],
+ IIC_SSE_MOVNT>;
let ExeDomain = SSEPackedInt in
def MOVNTDQmr : PDI<0xE7, MRMDestMem, (outs), (ins f128mem:$dst, VR128:$src),
"movntdq\t{$src, $dst|$dst, $src}",
- [(alignednontemporalstore (v2i64 VR128:$src), addr:$dst)]>;
+ [(alignednontemporalstore (v2i64 VR128:$src), addr:$dst)],
+ IIC_SSE_MOVNT>;
def : Pat<(alignednontemporalstore (v2i64 VR128:$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:$src),
"movnti{l}\t{$src, $dst|$dst, $src}",
- [(nontemporalstore (i32 GR32:$src), addr:$dst)]>,
+ [(nontemporalstore (i32 GR32:$src), addr:$dst)],
+ IIC_SSE_MOVNT>,
TB, Requires<[HasSSE2]>;
def MOVNTI_64mr : RI<0xC3, MRMDestMem, (outs), (ins i64mem:$dst, GR64:$src),
"movnti{q}\t{$src, $dst|$dst, $src}",
- [(nontemporalstore (i64 GR64:$src), addr:$dst)]>,
+ [(nontemporalstore (i64 GR64:$src), addr:$dst)],
+ IIC_SSE_MOVNT>,
TB, Requires<[HasSSE2]>;
}
// Prefetch intrinsic.
let Predicates = [HasSSE1] in {
def PREFETCHT0 : I<0x18, MRM1m, (outs), (ins i8mem:$src),
- "prefetcht0\t$src", [(prefetch addr:$src, imm, (i32 3), (i32 1))]>, TB;
+ "prefetcht0\t$src", [(prefetch addr:$src, imm, (i32 3), (i32 1))],
+ IIC_SSE_PREFETCH>, TB;
def PREFETCHT1 : I<0x18, MRM2m, (outs), (ins i8mem:$src),
- "prefetcht1\t$src", [(prefetch addr:$src, imm, (i32 2), (i32 1))]>, TB;
+ "prefetcht1\t$src", [(prefetch addr:$src, imm, (i32 2), (i32 1))],
+ IIC_SSE_PREFETCH>, TB;
def PREFETCHT2 : I<0x18, MRM3m, (outs), (ins i8mem:$src),
- "prefetcht2\t$src", [(prefetch addr:$src, imm, (i32 1), (i32 1))]>, TB;
+ "prefetcht2\t$src", [(prefetch addr:$src, imm, (i32 1), (i32 1))],
+ IIC_SSE_PREFETCH>, TB;
def PREFETCHNTA : I<0x18, MRM0m, (outs), (ins i8mem:$src),
- "prefetchnta\t$src", [(prefetch addr:$src, imm, (i32 0), (i32 1))]>, TB;
+ "prefetchnta\t$src", [(prefetch addr:$src, imm, (i32 0), (i32 1))],
+ IIC_SSE_PREFETCH>, TB;
}
// Flush cache
def CLFLUSH : I<0xAE, MRM7m, (outs), (ins i8mem:$src),
- "clflush\t$src", [(int_x86_sse2_clflush addr:$src)]>,
- TB, Requires<[HasSSE2]>;
+ "clflush\t$src", [(int_x86_sse2_clflush addr:$src)],
+ IIC_SSE_PREFETCH>, TB, Requires<[HasSSE2]>;
// Pause. This "instruction" is encoded as "rep; nop", so even though it
// was introduced with SSE2, it's backward compatible.
-def PAUSE : I<0x90, RawFrm, (outs), (ins), "pause", []>, REP;
+def PAUSE : I<0x90, RawFrm, (outs), (ins), "pause", [], IIC_SSE_PAUSE>, REP;
// Load, store, and memory fence
def SFENCE : I<0xAE, MRM_F8, (outs), (ins),
- "sfence", [(int_x86_sse_sfence)]>, TB, Requires<[HasSSE1]>;
+ "sfence", [(int_x86_sse_sfence)], IIC_SSE_SFENCE>,
+ TB, Requires<[HasSSE1]>;
def LFENCE : I<0xAE, MRM_E8, (outs), (ins),
- "lfence", [(int_x86_sse2_lfence)]>, TB, Requires<[HasSSE2]>;
+ "lfence", [(int_x86_sse2_lfence)], IIC_SSE_LFENCE>,
+ TB, Requires<[HasSSE2]>;
def MFENCE : I<0xAE, MRM_F0, (outs), (ins),
- "mfence", [(int_x86_sse2_mfence)]>, TB, Requires<[HasSSE2]>;
+ "mfence", [(int_x86_sse2_mfence)], IIC_SSE_MFENCE>,
+ TB, Requires<[HasSSE2]>;
def : Pat<(X86SFence), (SFENCE)>;
def : Pat<(X86LFence), (LFENCE)>;
//===----------------------------------------------------------------------===//
def VLDMXCSR : VPSI<0xAE, MRM2m, (outs), (ins i32mem:$src),
- "ldmxcsr\t$src", [(int_x86_sse_ldmxcsr addr:$src)]>, VEX;
+ "ldmxcsr\t$src", [(int_x86_sse_ldmxcsr addr:$src)],
+ IIC_SSE_LDMXCSR>, VEX;
def VSTMXCSR : VPSI<0xAE, MRM3m, (outs), (ins i32mem:$dst),
- "stmxcsr\t$dst", [(int_x86_sse_stmxcsr addr:$dst)]>, VEX;
+ "stmxcsr\t$dst", [(int_x86_sse_stmxcsr addr:$dst)],
+ IIC_SSE_STMXCSR>, VEX;
def LDMXCSR : PSI<0xAE, MRM2m, (outs), (ins i32mem:$src),
- "ldmxcsr\t$src", [(int_x86_sse_ldmxcsr addr:$src)]>;
+ "ldmxcsr\t$src", [(int_x86_sse_ldmxcsr addr:$src)],
+ IIC_SSE_LDMXCSR>;
def STMXCSR : PSI<0xAE, MRM3m, (outs), (ins i32mem:$dst),
- "stmxcsr\t$dst", [(int_x86_sse_stmxcsr addr:$dst)]>;
+ "stmxcsr\t$dst", [(int_x86_sse_stmxcsr addr:$dst)],
+ IIC_SSE_STMXCSR>;
//===---------------------------------------------------------------------===//
// SSE2 - Move Aligned/Unaligned Packed Integer Instructions
let neverHasSideEffects = 1 in {
def VMOVDQArr : VPDI<0x6F, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
- "movdqa\t{$src, $dst|$dst, $src}", []>, VEX;
+ "movdqa\t{$src, $dst|$dst, $src}", [], IIC_SSE_MOVA_P_RR>,
+ VEX;
def VMOVDQAYrr : VPDI<0x6F, MRMSrcReg, (outs VR256:$dst), (ins VR256:$src),
- "movdqa\t{$src, $dst|$dst, $src}", []>, VEX;
+ "movdqa\t{$src, $dst|$dst, $src}", [], IIC_SSE_MOVA_P_RR>,
+ VEX;
}
def VMOVDQUrr : VSSI<0x6F, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
- "movdqu\t{$src, $dst|$dst, $src}", []>, VEX;
+ "movdqu\t{$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, $src}", []>, VEX;
+ "movdqu\t{$src, $dst|$dst, $src}", [], IIC_SSE_MOVU_P_RR>,
+ VEX;
// For Disassembler
let isCodeGenOnly = 1 in {
def VMOVDQArr_REV : VPDI<0x7F, MRMDestReg, (outs VR128:$dst), (ins VR128:$src),
- "movdqa\t{$src, $dst|$dst, $src}", []>, VEX;
+ "movdqa\t{$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, $src}", []>, VEX;
+ "movdqa\t{$src, $dst|$dst, $src}", [],
+ IIC_SSE_MOVA_P_RR>,
+ VEX;
def VMOVDQUrr_REV : VSSI<0x7F, MRMDestReg, (outs VR128:$dst), (ins VR128:$src),
- "movdqu\t{$src, $dst|$dst, $src}", []>, VEX;
+ "movdqu\t{$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, $src}", []>, VEX;
+ "movdqu\t{$src, $dst|$dst, $src}", [],
+ IIC_SSE_MOVU_P_RR>,
+ VEX;
}
let canFoldAsLoad = 1, mayLoad = 1 in {
def VMOVDQArm : VPDI<0x6F, MRMSrcMem, (outs VR128:$dst), (ins i128mem:$src),
- "movdqa\t{$src, $dst|$dst, $src}", []>, VEX;
+ "movdqa\t{$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, $src}", []>, VEX;
+ "movdqa\t{$src, $dst|$dst, $src}", [], IIC_SSE_MOVA_P_RM>,
+ VEX;
let Predicates = [HasAVX] in {
def VMOVDQUrm : I<0x6F, MRMSrcMem, (outs VR128:$dst), (ins i128mem:$src),
- "vmovdqu\t{$src, $dst|$dst, $src}",[]>, XS, VEX;
+ "vmovdqu\t{$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, $src}",[]>, XS, VEX;
+ "vmovdqu\t{$src, $dst|$dst, $src}",[], IIC_SSE_MOVU_P_RM>,
+ XS, VEX;
}
}
let mayStore = 1 in {
def VMOVDQAmr : VPDI<0x7F, MRMDestMem, (outs),
(ins i128mem:$dst, VR128:$src),
- "movdqa\t{$src, $dst|$dst, $src}", []>, VEX;
+ "movdqa\t{$src, $dst|$dst, $src}", [], IIC_SSE_MOVA_P_MR>,
+ VEX;
def VMOVDQAYmr : VPDI<0x7F, MRMDestMem, (outs),
(ins i256mem:$dst, VR256:$src),
- "movdqa\t{$src, $dst|$dst, $src}", []>, VEX;
+ "movdqa\t{$src, $dst|$dst, $src}", [], IIC_SSE_MOVA_P_MR>,
+ VEX;
let Predicates = [HasAVX] in {
def VMOVDQUmr : I<0x7F, MRMDestMem, (outs), (ins i128mem:$dst, VR128:$src),
- "vmovdqu\t{$src, $dst|$dst, $src}",[]>, XS, VEX;
+ "vmovdqu\t{$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, $src}",[]>, XS, VEX;
+ "vmovdqu\t{$src, $dst|$dst, $src}",[], IIC_SSE_MOVU_P_MR>,
+ XS, VEX;
}
}
let neverHasSideEffects = 1 in
def MOVDQArr : PDI<0x6F, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
- "movdqa\t{$src, $dst|$dst, $src}", []>;
+ "movdqa\t{$src, $dst|$dst, $src}", [], IIC_SSE_MOVA_P_RR>;
def MOVDQUrr : I<0x6F, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
"movdqu\t{$src, $dst|$dst, $src}",
- []>, XS, Requires<[HasSSE2]>;
+ [], IIC_SSE_MOVU_P_RR>, XS, Requires<[UseSSE2]>;
// For Disassembler
let isCodeGenOnly = 1 in {
def MOVDQArr_REV : PDI<0x7F, MRMDestReg, (outs VR128:$dst), (ins VR128:$src),
- "movdqa\t{$src, $dst|$dst, $src}", []>;
+ "movdqa\t{$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, $src}",
- []>, XS, Requires<[HasSSE2]>;
+ [], IIC_SSE_MOVU_P_RR>, XS, Requires<[UseSSE2]>;
}
let canFoldAsLoad = 1, mayLoad = 1 in {
def MOVDQArm : PDI<0x6F, MRMSrcMem, (outs VR128:$dst), (ins i128mem:$src),
"movdqa\t{$src, $dst|$dst, $src}",
- [/*(set VR128:$dst, (alignedloadv2i64 addr:$src))*/]>;
+ [/*(set VR128:$dst, (alignedloadv2i64 addr:$src))*/],
+ IIC_SSE_MOVA_P_RM>;
def MOVDQUrm : I<0x6F, MRMSrcMem, (outs VR128:$dst), (ins i128mem:$src),
"movdqu\t{$src, $dst|$dst, $src}",
- [/*(set VR128:$dst, (loadv2i64 addr:$src))*/]>,
- XS, Requires<[HasSSE2]>;
+ [/*(set VR128:$dst, (loadv2i64 addr:$src))*/],
+ IIC_SSE_MOVU_P_RM>,
+ XS, Requires<[UseSSE2]>;
}
let mayStore = 1 in {
def MOVDQAmr : PDI<0x7F, MRMDestMem, (outs), (ins i128mem:$dst, VR128:$src),
"movdqa\t{$src, $dst|$dst, $src}",
- [/*(alignedstore (v2i64 VR128:$src), addr:$dst)*/]>;
+ [/*(alignedstore (v2i64 VR128:$src), addr:$dst)*/],
+ IIC_SSE_MOVA_P_MR>;
def MOVDQUmr : I<0x7F, MRMDestMem, (outs), (ins i128mem:$dst, VR128:$src),
"movdqu\t{$src, $dst|$dst, $src}",
- [/*(store (v2i64 VR128:$src), addr:$dst)*/]>,
- XS, Requires<[HasSSE2]>;
+ [/*(store (v2i64 VR128:$src), addr:$dst)*/],
+ IIC_SSE_MOVU_P_MR>,
+ XS, Requires<[UseSSE2]>;
}
// Intrinsic forms of MOVDQU load and store
def VMOVDQUmr_Int : I<0x7F, MRMDestMem, (outs), (ins i128mem:$dst, VR128:$src),
"vmovdqu\t{$src, $dst|$dst, $src}",
- [(int_x86_sse2_storeu_dq addr:$dst, VR128:$src)]>,
+ [(int_x86_sse2_storeu_dq addr:$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:$src)]>,
- XS, Requires<[HasSSE2]>;
+ [(int_x86_sse2_storeu_dq addr:$dst, VR128:$src)],
+ IIC_SSE_MOVU_P_MR>,
+ XS, Requires<[UseSSE2]>;
} // ExeDomain = SSEPackedInt
// SSE2 - Packed Integer Arithmetic Instructions
//===---------------------------------------------------------------------===//
+def SSE_PMADD : OpndItins<
+ IIC_SSE_PMADD, IIC_SSE_PMADD
+>;
+
let ExeDomain = SSEPackedInt in { // SSE integer instructions
multiclass PDI_binop_rm_int<bits<8> opc, string OpcodeStr, Intrinsic IntId,
RegisterClass RC, PatFrag memop_frag,
- X86MemOperand x86memop, bit IsCommutable = 0,
+ X86MemOperand x86memop,
+ OpndItins itins,
+ bit IsCommutable = 0,
bit Is2Addr = 1> {
let isCommutable = IsCommutable in
def rr : PDI<opc, MRMSrcReg, (outs RC:$dst),
!if(Is2Addr,
!strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"),
!strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}")),
- [(set RC:$dst, (IntId RC:$src1, RC:$src2))]>;
+ [(set RC:$dst, (IntId RC:$src1, RC:$src2))], itins.rr>;
def rm : PDI<opc, MRMSrcMem, (outs RC:$dst),
(ins RC:$src1, x86memop:$src2),
!if(Is2Addr,
!strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"),
!strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}")),
- [(set RC:$dst, (IntId RC:$src1, (bitconvert (memop_frag addr:$src2))))]>;
+ [(set RC:$dst, (IntId RC:$src1, (bitconvert (memop_frag addr:$src2))))],
+ itins.rm>;
}
multiclass PDI_binop_rmi<bits<8> opc, bits<8> opc2, Format ImmForm,
string OpcodeStr, SDNode OpNode,
SDNode OpNode2, RegisterClass RC,
ValueType DstVT, ValueType SrcVT, PatFrag bc_frag,
+ ShiftOpndItins itins,
bit Is2Addr = 1> {
// src2 is always 128-bit
def rr : PDI<opc, MRMSrcReg, (outs RC:$dst),
!if(Is2Addr,
!strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"),
!strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}")),
- [(set RC:$dst, (DstVT (OpNode RC:$src1, (SrcVT VR128:$src2))))]>;
+ [(set RC:$dst, (DstVT (OpNode RC:$src1, (SrcVT VR128:$src2))))],
+ itins.rr>;
def rm : PDI<opc, MRMSrcMem, (outs RC:$dst),
(ins RC:$src1, i128mem:$src2),
!if(Is2Addr,
!strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"),
!strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}")),
[(set RC:$dst, (DstVT (OpNode RC:$src1,
- (bc_frag (memopv2i64 addr:$src2)))))]>;
+ (bc_frag (memopv2i64 addr:$src2)))))], itins.rm>;
def ri : PDIi8<opc2, ImmForm, (outs RC:$dst),
(ins RC:$src1, i32i8imm:$src2),
!if(Is2Addr,
!strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"),
!strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}")),
- [(set RC:$dst, (DstVT (OpNode2 RC:$src1, (i32 imm:$src2))))]>;
+ [(set RC:$dst, (DstVT (OpNode2 RC:$src1, (i32 imm:$src2))))], itins.ri>;
}
/// PDI_binop_rm - Simple SSE2 binary operator with different src and dst types
multiclass PDI_binop_rm2<bits<8> opc, string OpcodeStr, SDNode OpNode,
ValueType DstVT, ValueType SrcVT, RegisterClass RC,
PatFrag memop_frag, X86MemOperand x86memop,
+ OpndItins itins,
bit IsCommutable = 0, bit Is2Addr = 1> {
let isCommutable = IsCommutable in
def rr : PDI<opc, MRMSrcReg, (outs RC:$dst),
let Predicates = [HasAVX] in {
defm VPADDB : PDI_binop_rm<0xFC, "vpaddb", add, v16i8, VR128, memopv2i64,
- i128mem, 1, 0 /*3addr*/>, VEX_4V;
+ i128mem, SSE_INTALU_ITINS_P, 1, 0 /*3addr*/>,
+ VEX_4V;
defm VPADDW : PDI_binop_rm<0xFD, "vpaddw", add, v8i16, VR128, memopv2i64,
- i128mem, 1, 0>, VEX_4V;
+ i128mem, SSE_INTALU_ITINS_P, 1, 0>, VEX_4V;
defm VPADDD : PDI_binop_rm<0xFE, "vpaddd", add, v4i32, VR128, memopv2i64,
- i128mem, 1, 0>, VEX_4V;
+ i128mem, SSE_INTALU_ITINS_P, 1, 0>, VEX_4V;
defm VPADDQ : PDI_binop_rm<0xD4, "vpaddq", add, v2i64, VR128, memopv2i64,
- i128mem, 1, 0>, VEX_4V;
+ i128mem, SSE_INTALUQ_ITINS_P, 1, 0>, VEX_4V;
defm VPMULLW : PDI_binop_rm<0xD5, "vpmullw", mul, v8i16, VR128, memopv2i64,
- i128mem, 1, 0>, VEX_4V;
+ i128mem, SSE_INTMUL_ITINS_P, 1, 0>, VEX_4V;
defm VPSUBB : PDI_binop_rm<0xF8, "vpsubb", sub, v16i8, VR128, memopv2i64,
- i128mem, 0, 0>, VEX_4V;
+ i128mem, SSE_INTALU_ITINS_P, 0, 0>, VEX_4V;
defm VPSUBW : PDI_binop_rm<0xF9, "vpsubw", sub, v8i16, VR128, memopv2i64,
- i128mem, 0, 0>, VEX_4V;
+ i128mem, SSE_INTALU_ITINS_P, 0, 0>, VEX_4V;
defm VPSUBD : PDI_binop_rm<0xFA, "vpsubd", sub, v4i32, VR128, memopv2i64,
- i128mem, 0, 0>, VEX_4V;
+ i128mem, SSE_INTALU_ITINS_P, 0, 0>, VEX_4V;
defm VPSUBQ : PDI_binop_rm<0xFB, "vpsubq", sub, v2i64, VR128, memopv2i64,
- i128mem, 0, 0>, VEX_4V;
+ i128mem, SSE_INTALUQ_ITINS_P, 0, 0>, VEX_4V;
defm VPMULUDQ : PDI_binop_rm2<0xF4, "vpmuludq", X86pmuludq, v2i64, v4i32, VR128,
- memopv2i64, i128mem, 1, 0>, VEX_4V;
+ 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, 0, 0>, VEX_4V;
+ 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, 0, 0>, VEX_4V;
+ 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, 0, 0>, VEX_4V;
+ 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, 0, 0>, VEX_4V;
+ 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, 1, 0>, VEX_4V;
+ 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, 1, 0>, VEX_4V;
+ 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, 1, 0>, VEX_4V;
+ 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, 1, 0>, VEX_4V;
+ 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, 1, 0>, VEX_4V;
+ 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, 1, 0>, VEX_4V;
+ 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, 1, 0>, VEX_4V;
+ 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, 1, 0>, VEX_4V;
+ 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, 1, 0>, VEX_4V;
+ 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, 1, 0>, VEX_4V;
+ 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, 1, 0>, VEX_4V;
+ 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, 1, 0>, VEX_4V;
+ 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, 1, 0>, VEX_4V;
+ 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, 1, 0>, VEX_4V;
+ 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, 1, 0>, VEX_4V;
+ i256mem, SSE_INTALU_ITINS_P, 1, 0>, VEX_4V;
defm VPADDWY : PDI_binop_rm<0xFD, "vpaddw", add, v16i16, VR256, memopv4i64,
- i256mem, 1, 0>, VEX_4V;
+ i256mem, SSE_INTALU_ITINS_P, 1, 0>, VEX_4V;
defm VPADDDY : PDI_binop_rm<0xFE, "vpaddd", add, v8i32, VR256, memopv4i64,
- i256mem, 1, 0>, VEX_4V;
+ i256mem, SSE_INTALU_ITINS_P, 1, 0>, VEX_4V;
defm VPADDQY : PDI_binop_rm<0xD4, "vpaddq", add, v4i64, VR256, memopv4i64,
- i256mem, 1, 0>, VEX_4V;
+ i256mem, SSE_INTALUQ_ITINS_P, 1, 0>, VEX_4V;
defm VPMULLWY : PDI_binop_rm<0xD5, "vpmullw", mul, v16i16, VR256, memopv4i64,
- i256mem, 1, 0>, VEX_4V;
+ i256mem, SSE_INTMUL_ITINS_P, 1, 0>, VEX_4V;
defm VPSUBBY : PDI_binop_rm<0xF8, "vpsubb", sub, v32i8, VR256, memopv4i64,
- i256mem, 0, 0>, VEX_4V;
+ i256mem, SSE_INTALU_ITINS_P, 0, 0>, VEX_4V;
defm VPSUBWY : PDI_binop_rm<0xF9, "vpsubw", sub, v16i16,VR256, memopv4i64,
- i256mem, 0, 0>, VEX_4V;
+ i256mem, SSE_INTALU_ITINS_P, 0, 0>, VEX_4V;
defm VPSUBDY : PDI_binop_rm<0xFA, "vpsubd", sub, v8i32, VR256, memopv4i64,
- i256mem, 0, 0>, VEX_4V;
+ i256mem, SSE_INTALU_ITINS_P, 0, 0>, VEX_4V;
defm VPSUBQY : PDI_binop_rm<0xFB, "vpsubq", sub, v4i64, VR256, memopv4i64,
- i256mem, 0, 0>, VEX_4V;
+ i256mem, SSE_INTALUQ_ITINS_P, 0, 0>, VEX_4V;
defm VPMULUDQY : PDI_binop_rm2<0xF4, "vpmuludq", X86pmuludq, v4i64, v8i32,
- VR256, memopv4i64, i256mem, 1, 0>, VEX_4V;
+ 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, 0, 0>, VEX_4V;
+ 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, 0, 0>, VEX_4V;
+ 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, 0, 0>, VEX_4V;
+ 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, 0, 0>, VEX_4V;
+ 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, 1, 0>, VEX_4V;
+ 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, 1, 0>, VEX_4V;
+ 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, 1, 0>, VEX_4V;
+ 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, 1, 0>, VEX_4V;
+ 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, 1, 0>, VEX_4V;
+ 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, 1, 0>, VEX_4V;
+ 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, 1, 0>, VEX_4V;
+ 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, 1, 0>, VEX_4V;
+ 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, 1, 0>, VEX_4V;
+ 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, 1, 0>, VEX_4V;
+ 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, 1, 0>, VEX_4V;
+ 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, 1, 0>, VEX_4V;
+ 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, 1, 0>, VEX_4V;
+ 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, 1, 0>, VEX_4V;
+ 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, 1>;
+ i128mem, SSE_INTALU_ITINS_P, 1>;
defm PADDW : PDI_binop_rm<0xFD, "paddw", add, v8i16, VR128, memopv2i64,
- i128mem, 1>;
+ i128mem, SSE_INTALU_ITINS_P, 1>;
defm PADDD : PDI_binop_rm<0xFE, "paddd", add, v4i32, VR128, memopv2i64,
- i128mem, 1>;
+ i128mem, SSE_INTALU_ITINS_P, 1>;
defm PADDQ : PDI_binop_rm<0xD4, "paddq", add, v2i64, VR128, memopv2i64,
- i128mem, 1>;
+ i128mem, SSE_INTALUQ_ITINS_P, 1>;
defm PMULLW : PDI_binop_rm<0xD5, "pmullw", mul, v8i16, VR128, memopv2i64,
- i128mem, 1>;
+ i128mem, SSE_INTMUL_ITINS_P, 1>;
defm PSUBB : PDI_binop_rm<0xF8, "psubb", sub, v16i8, VR128, memopv2i64,
- i128mem>;
+ i128mem, SSE_INTALU_ITINS_P>;
defm PSUBW : PDI_binop_rm<0xF9, "psubw", sub, v8i16, VR128, memopv2i64,
- i128mem>;
+ i128mem, SSE_INTALU_ITINS_P>;
defm PSUBD : PDI_binop_rm<0xFA, "psubd", sub, v4i32, VR128, memopv2i64,
- i128mem>;
+ i128mem, SSE_INTALU_ITINS_P>;
defm PSUBQ : PDI_binop_rm<0xFB, "psubq", sub, v2i64, VR128, memopv2i64,
- i128mem>;
+ i128mem, SSE_INTALUQ_ITINS_P>;
defm PMULUDQ : PDI_binop_rm2<0xF4, "pmuludq", X86pmuludq, v2i64, v4i32, VR128,
- memopv2i64, i128mem, 1>;
+ 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>;
+ VR128, memopv2i64, i128mem,
+ SSE_INTALU_ITINS_P>;
defm PSUBSW : PDI_binop_rm_int<0xE9, "psubsw" , int_x86_sse2_psubs_w,
- VR128, memopv2i64, i128mem>;
+ VR128, memopv2i64, i128mem,
+ SSE_INTALU_ITINS_P>;
defm PSUBUSB : PDI_binop_rm_int<0xD8, "psubusb", int_x86_sse2_psubus_b,
- VR128, memopv2i64, i128mem>;
+ VR128, memopv2i64, i128mem,
+ SSE_INTALU_ITINS_P>;
defm PSUBUSW : PDI_binop_rm_int<0xD9, "psubusw", int_x86_sse2_psubus_w,
- VR128, memopv2i64, i128mem>;
+ VR128, memopv2i64, i128mem,
+ SSE_INTALU_ITINS_P>;
defm PADDSB : PDI_binop_rm_int<0xEC, "paddsb" , int_x86_sse2_padds_b,
- VR128, memopv2i64, i128mem, 1>;
+ VR128, memopv2i64, i128mem,
+ SSE_INTALU_ITINS_P, 1>;
defm PADDSW : PDI_binop_rm_int<0xED, "paddsw" , int_x86_sse2_padds_w,
- VR128, memopv2i64, i128mem, 1>;
+ VR128, memopv2i64, i128mem,
+ SSE_INTALU_ITINS_P, 1>;
defm PADDUSB : PDI_binop_rm_int<0xDC, "paddusb", int_x86_sse2_paddus_b,
- VR128, memopv2i64, i128mem, 1>;
+ VR128, memopv2i64, i128mem,
+ SSE_INTALU_ITINS_P, 1>;
defm PADDUSW : PDI_binop_rm_int<0xDD, "paddusw", int_x86_sse2_paddus_w,
- VR128, memopv2i64, i128mem, 1>;
+ VR128, memopv2i64, i128mem,
+ SSE_INTALU_ITINS_P, 1>;
defm PMULHUW : PDI_binop_rm_int<0xE4, "pmulhuw", int_x86_sse2_pmulhu_w,
- VR128, memopv2i64, i128mem, 1>;
+ VR128, memopv2i64, i128mem,
+ SSE_INTMUL_ITINS_P, 1>;
defm PMULHW : PDI_binop_rm_int<0xE5, "pmulhw" , int_x86_sse2_pmulh_w,
- VR128, memopv2i64, i128mem, 1>;
+ VR128, memopv2i64, i128mem,
+ SSE_INTMUL_ITINS_P, 1>;
defm PMADDWD : PDI_binop_rm_int<0xF5, "pmaddwd", int_x86_sse2_pmadd_wd,
- VR128, memopv2i64, i128mem, 1>;
+ VR128, memopv2i64, i128mem,
+ SSE_PMADD, 1>;
defm PAVGB : PDI_binop_rm_int<0xE0, "pavgb", int_x86_sse2_pavg_b,
- VR128, memopv2i64, i128mem, 1>;
+ VR128, memopv2i64, i128mem,
+ SSE_INTALU_ITINS_P, 1>;
defm PAVGW : PDI_binop_rm_int<0xE3, "pavgw", int_x86_sse2_pavg_w,
- VR128, memopv2i64, i128mem, 1>;
+ VR128, memopv2i64, i128mem,
+ SSE_INTALU_ITINS_P, 1>;
defm PMINUB : PDI_binop_rm_int<0xDA, "pminub", int_x86_sse2_pminu_b,
- VR128, memopv2i64, i128mem, 1>;
+ VR128, memopv2i64, i128mem,
+ SSE_INTALU_ITINS_P, 1>;
defm PMINSW : PDI_binop_rm_int<0xEA, "pminsw", int_x86_sse2_pmins_w,
- VR128, memopv2i64, i128mem, 1>;
+ VR128, memopv2i64, i128mem,
+ SSE_INTALU_ITINS_P, 1>;
defm PMAXUB : PDI_binop_rm_int<0xDE, "pmaxub", int_x86_sse2_pmaxu_b,
- VR128, memopv2i64, i128mem, 1>;
+ VR128, memopv2i64, i128mem,
+ SSE_INTALU_ITINS_P, 1>;
defm PMAXSW : PDI_binop_rm_int<0xEE, "pmaxsw", int_x86_sse2_pmaxs_w,
- VR128, memopv2i64, i128mem, 1>;
+ VR128, memopv2i64, i128mem,
+ SSE_INTALU_ITINS_P, 1>;
defm PSADBW : PDI_binop_rm_int<0xF6, "psadbw", int_x86_sse2_psad_bw,
- VR128, memopv2i64, i128mem, 1>;
+ VR128, memopv2i64, i128mem,
+ SSE_INTALU_ITINS_P, 1>;
} // Constraints = "$src1 = $dst"
let Predicates = [HasAVX] in {
defm VPSLLW : PDI_binop_rmi<0xF1, 0x71, MRM6r, "vpsllw", X86vshl, X86vshli,
- VR128, v8i16, v8i16, bc_v8i16, 0>, VEX_4V;
+ VR128, v8i16, v8i16, bc_v8i16,
+ SSE_INTSHIFT_ITINS_P, 0>, VEX_4V;
defm VPSLLD : PDI_binop_rmi<0xF2, 0x72, MRM6r, "vpslld", X86vshl, X86vshli,
- VR128, v4i32, v4i32, bc_v4i32, 0>, VEX_4V;
+ VR128, v4i32, v4i32, bc_v4i32,
+ SSE_INTSHIFT_ITINS_P, 0>, VEX_4V;
defm VPSLLQ : PDI_binop_rmi<0xF3, 0x73, MRM6r, "vpsllq", X86vshl, X86vshli,
- VR128, v2i64, v2i64, bc_v2i64, 0>, VEX_4V;
+ VR128, v2i64, v2i64, bc_v2i64,
+ SSE_INTSHIFT_ITINS_P, 0>, VEX_4V;
defm VPSRLW : PDI_binop_rmi<0xD1, 0x71, MRM2r, "vpsrlw", X86vsrl, X86vsrli,
- VR128, v8i16, v8i16, bc_v8i16, 0>, VEX_4V;
+ VR128, v8i16, v8i16, bc_v8i16,
+ SSE_INTSHIFT_ITINS_P, 0>, VEX_4V;
defm VPSRLD : PDI_binop_rmi<0xD2, 0x72, MRM2r, "vpsrld", X86vsrl, X86vsrli,
- VR128, v4i32, v4i32, bc_v4i32, 0>, VEX_4V;
+ VR128, v4i32, v4i32, bc_v4i32,
+ SSE_INTSHIFT_ITINS_P, 0>, VEX_4V;
defm VPSRLQ : PDI_binop_rmi<0xD3, 0x73, MRM2r, "vpsrlq", X86vsrl, X86vsrli,
- VR128, v2i64, v2i64, bc_v2i64, 0>, VEX_4V;
+ VR128, v2i64, v2i64, bc_v2i64,
+ SSE_INTSHIFT_ITINS_P, 0>, VEX_4V;
defm VPSRAW : PDI_binop_rmi<0xE1, 0x71, MRM4r, "vpsraw", X86vsra, X86vsrai,
- VR128, v8i16, v8i16, bc_v8i16, 0>, VEX_4V;
+ VR128, v8i16, v8i16, bc_v8i16,
+ SSE_INTSHIFT_ITINS_P, 0>, VEX_4V;
defm VPSRAD : PDI_binop_rmi<0xE2, 0x72, MRM4r, "vpsrad", X86vsra, X86vsrai,
- VR128, v4i32, v4i32, bc_v4i32, 0>, VEX_4V;
+ VR128, v4i32, v4i32, bc_v4i32,
+ SSE_INTSHIFT_ITINS_P, 0>, VEX_4V;
let ExeDomain = SSEPackedInt in {
// 128-bit logical shifts.
let Predicates = [HasAVX2] in {
defm VPSLLWY : PDI_binop_rmi<0xF1, 0x71, MRM6r, "vpsllw", X86vshl, X86vshli,
- VR256, v16i16, v8i16, bc_v8i16, 0>, VEX_4V;
+ VR256, v16i16, v8i16, bc_v8i16,
+ SSE_INTSHIFT_ITINS_P, 0>, VEX_4V;
defm VPSLLDY : PDI_binop_rmi<0xF2, 0x72, MRM6r, "vpslld", X86vshl, X86vshli,
- VR256, v8i32, v4i32, bc_v4i32, 0>, VEX_4V;
+ VR256, v8i32, v4i32, bc_v4i32,
+ SSE_INTSHIFT_ITINS_P, 0>, VEX_4V;
defm VPSLLQY : PDI_binop_rmi<0xF3, 0x73, MRM6r, "vpsllq", X86vshl, X86vshli,
- VR256, v4i64, v2i64, bc_v2i64, 0>, VEX_4V;
+ VR256, v4i64, v2i64, bc_v2i64,
+ SSE_INTSHIFT_ITINS_P, 0>, VEX_4V;
defm VPSRLWY : PDI_binop_rmi<0xD1, 0x71, MRM2r, "vpsrlw", X86vsrl, X86vsrli,
- VR256, v16i16, v8i16, bc_v8i16, 0>, VEX_4V;
+ VR256, v16i16, v8i16, bc_v8i16,
+ SSE_INTSHIFT_ITINS_P, 0>, VEX_4V;
defm VPSRLDY : PDI_binop_rmi<0xD2, 0x72, MRM2r, "vpsrld", X86vsrl, X86vsrli,
- VR256, v8i32, v4i32, bc_v4i32, 0>, VEX_4V;
+ VR256, v8i32, v4i32, bc_v4i32,
+ SSE_INTSHIFT_ITINS_P, 0>, VEX_4V;
defm VPSRLQY : PDI_binop_rmi<0xD3, 0x73, MRM2r, "vpsrlq", X86vsrl, X86vsrli,
- VR256, v4i64, v2i64, bc_v2i64, 0>, VEX_4V;
+ VR256, v4i64, v2i64, bc_v2i64,
+ SSE_INTSHIFT_ITINS_P, 0>, VEX_4V;
defm VPSRAWY : PDI_binop_rmi<0xE1, 0x71, MRM4r, "vpsraw", X86vsra, X86vsrai,
- VR256, v16i16, v8i16, bc_v8i16, 0>, VEX_4V;
+ VR256, v16i16, v8i16, bc_v8i16,
+ SSE_INTSHIFT_ITINS_P, 0>, VEX_4V;
defm VPSRADY : PDI_binop_rmi<0xE2, 0x72, MRM4r, "vpsrad", X86vsra, X86vsrai,
- VR256, v8i32, v4i32, bc_v4i32, 0>, VEX_4V;
+ VR256, v8i32, v4i32, bc_v4i32,
+ SSE_INTSHIFT_ITINS_P, 0>, VEX_4V;
let ExeDomain = SSEPackedInt in {
// 256-bit logical shifts.
let Constraints = "$src1 = $dst" in {
defm PSLLW : PDI_binop_rmi<0xF1, 0x71, MRM6r, "psllw", X86vshl, X86vshli,
- VR128, v8i16, v8i16, bc_v8i16>;
+ VR128, v8i16, v8i16, bc_v8i16,
+ SSE_INTSHIFT_ITINS_P>;
defm PSLLD : PDI_binop_rmi<0xF2, 0x72, MRM6r, "pslld", X86vshl, X86vshli,
- VR128, v4i32, v4i32, bc_v4i32>;
+ VR128, v4i32, v4i32, bc_v4i32,
+ SSE_INTSHIFT_ITINS_P>;
defm PSLLQ : PDI_binop_rmi<0xF3, 0x73, MRM6r, "psllq", X86vshl, X86vshli,
- VR128, v2i64, v2i64, bc_v2i64>;
+ VR128, v2i64, v2i64, bc_v2i64,
+ SSE_INTSHIFT_ITINS_P>;
defm PSRLW : PDI_binop_rmi<0xD1, 0x71, MRM2r, "psrlw", X86vsrl, X86vsrli,
- VR128, v8i16, v8i16, bc_v8i16>;
+ VR128, v8i16, v8i16, bc_v8i16,
+ SSE_INTSHIFT_ITINS_P>;
defm PSRLD : PDI_binop_rmi<0xD2, 0x72, MRM2r, "psrld", X86vsrl, X86vsrli,
- VR128, v4i32, v4i32, bc_v4i32>;
+ VR128, v4i32, v4i32, bc_v4i32,
+ SSE_INTSHIFT_ITINS_P>;
defm PSRLQ : PDI_binop_rmi<0xD3, 0x73, MRM2r, "psrlq", X86vsrl, X86vsrli,
- VR128, v2i64, v2i64, bc_v2i64>;
+ VR128, v2i64, v2i64, bc_v2i64,
+ SSE_INTSHIFT_ITINS_P>;
defm PSRAW : PDI_binop_rmi<0xE1, 0x71, MRM4r, "psraw", X86vsra, X86vsrai,
- VR128, v8i16, v8i16, bc_v8i16>;
+ VR128, v8i16, v8i16, bc_v8i16,
+ SSE_INTSHIFT_ITINS_P>;
defm PSRAD : PDI_binop_rmi<0xE2, 0x72, MRM4r, "psrad", X86vsra, X86vsrai,
- VR128, v4i32, v4i32, bc_v4i32>;
+ VR128, v4i32, v4i32, bc_v4i32,
+ SSE_INTSHIFT_ITINS_P>;
let ExeDomain = SSEPackedInt in {
// 128-bit logical shifts.
(VPSRLDQYri VR256:$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),
let Predicates = [HasAVX] in {
defm VPCMPEQB : PDI_binop_rm<0x74, "vpcmpeqb", X86pcmpeq, v16i8,
- VR128, memopv2i64, i128mem, 1, 0>, VEX_4V;
+ VR128, memopv2i64, i128mem,
+ SSE_INTALU_ITINS_P, 1, 0>, VEX_4V;
defm VPCMPEQW : PDI_binop_rm<0x75, "vpcmpeqw", X86pcmpeq, v8i16,
- VR128, memopv2i64, i128mem, 1, 0>, VEX_4V;
+ VR128, memopv2i64, i128mem,
+ SSE_INTALU_ITINS_P, 1, 0>, VEX_4V;
defm VPCMPEQD : PDI_binop_rm<0x76, "vpcmpeqd", X86pcmpeq, v4i32,
- VR128, memopv2i64, i128mem, 1, 0>, VEX_4V;
+ VR128, memopv2i64, i128mem,
+ SSE_INTALU_ITINS_P, 1, 0>, VEX_4V;
defm VPCMPGTB : PDI_binop_rm<0x64, "vpcmpgtb", X86pcmpgt, v16i8,
- VR128, memopv2i64, i128mem, 0, 0>, VEX_4V;
+ VR128, memopv2i64, i128mem,
+ SSE_INTALU_ITINS_P, 0, 0>, VEX_4V;
defm VPCMPGTW : PDI_binop_rm<0x65, "vpcmpgtw", X86pcmpgt, v8i16,
- VR128, memopv2i64, i128mem, 0, 0>, VEX_4V;
+ VR128, memopv2i64, i128mem,
+ SSE_INTALU_ITINS_P, 0, 0>, VEX_4V;
defm VPCMPGTD : PDI_binop_rm<0x66, "vpcmpgtd", X86pcmpgt, v4i32,
- VR128, memopv2i64, i128mem, 0, 0>, VEX_4V;
+ 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, 1, 0>, VEX_4V;
+ VR256, memopv4i64, i256mem,
+ SSE_INTALU_ITINS_P, 1, 0>, VEX_4V;
defm VPCMPEQWY : PDI_binop_rm<0x75, "vpcmpeqw", X86pcmpeq, v16i16,
- VR256, memopv4i64, i256mem, 1, 0>, VEX_4V;
+ VR256, memopv4i64, i256mem,
+ SSE_INTALU_ITINS_P, 1, 0>, VEX_4V;
defm VPCMPEQDY : PDI_binop_rm<0x76, "vpcmpeqd", X86pcmpeq, v8i32,
- VR256, memopv4i64, i256mem, 1, 0>, VEX_4V;
+ VR256, memopv4i64, i256mem,
+ SSE_INTALU_ITINS_P, 1, 0>, VEX_4V;
defm VPCMPGTBY : PDI_binop_rm<0x64, "vpcmpgtb", X86pcmpgt, v32i8,
- VR256, memopv4i64, i256mem, 0, 0>, VEX_4V;
+ VR256, memopv4i64, i256mem,
+ SSE_INTALU_ITINS_P, 0, 0>, VEX_4V;
defm VPCMPGTWY : PDI_binop_rm<0x65, "vpcmpgtw", X86pcmpgt, v16i16,
- VR256, memopv4i64, i256mem, 0, 0>, VEX_4V;
+ VR256, memopv4i64, i256mem,
+ SSE_INTALU_ITINS_P, 0, 0>, VEX_4V;
defm VPCMPGTDY : PDI_binop_rm<0x66, "vpcmpgtd", X86pcmpgt, v8i32,
- VR256, memopv4i64, i256mem, 0, 0>, VEX_4V;
+ VR256, memopv4i64, i256mem,
+ SSE_INTALU_ITINS_P, 0, 0>, VEX_4V;
}
let Constraints = "$src1 = $dst" in {
defm PCMPEQB : PDI_binop_rm<0x74, "pcmpeqb", X86pcmpeq, v16i8,
- VR128, memopv2i64, i128mem, 1>;
+ VR128, memopv2i64, i128mem,
+ SSE_INTALU_ITINS_P, 1>;
defm PCMPEQW : PDI_binop_rm<0x75, "pcmpeqw", X86pcmpeq, v8i16,
- VR128, memopv2i64, i128mem, 1>;
+ VR128, memopv2i64, i128mem,
+ SSE_INTALU_ITINS_P, 1>;
defm PCMPEQD : PDI_binop_rm<0x76, "pcmpeqd", X86pcmpeq, v4i32,
- VR128, memopv2i64, i128mem, 1>;
+ VR128, memopv2i64, i128mem,
+ SSE_INTALU_ITINS_P, 1>;
defm PCMPGTB : PDI_binop_rm<0x64, "pcmpgtb", X86pcmpgt, v16i8,
- VR128, memopv2i64, i128mem>;
+ VR128, memopv2i64, i128mem,
+ SSE_INTALU_ITINS_P>;
defm PCMPGTW : PDI_binop_rm<0x65, "pcmpgtw", X86pcmpgt, v8i16,
- VR128, memopv2i64, i128mem>;
+ VR128, memopv2i64, i128mem,
+ SSE_INTALU_ITINS_P>;
defm PCMPGTD : PDI_binop_rm<0x66, "pcmpgtd", X86pcmpgt, v4i32,
- VR128, memopv2i64, i128mem>;
+ VR128, memopv2i64, i128mem,
+ SSE_INTALU_ITINS_P>;
} // Constraints = "$src1 = $dst"
//===---------------------------------------------------------------------===//
let Predicates = [HasAVX] in {
defm VPACKSSWB : PDI_binop_rm_int<0x63, "vpacksswb", int_x86_sse2_packsswb_128,
- VR128, memopv2i64, i128mem, 0, 0>, VEX_4V;
+ VR128, memopv2i64, i128mem,
+ SSE_INTALU_ITINS_P, 0, 0>, VEX_4V;
defm VPACKSSDW : PDI_binop_rm_int<0x6B, "vpackssdw", int_x86_sse2_packssdw_128,
- VR128, memopv2i64, i128mem, 0, 0>, VEX_4V;
+ VR128, memopv2i64, i128mem,
+ SSE_INTALU_ITINS_P, 0, 0>, VEX_4V;
defm VPACKUSWB : PDI_binop_rm_int<0x67, "vpackuswb", int_x86_sse2_packuswb_128,
- VR128, memopv2i64, i128mem, 0, 0>, VEX_4V;
+ VR128, memopv2i64, i128mem,
+ SSE_INTALU_ITINS_P, 0, 0>, VEX_4V;
}
let Predicates = [HasAVX2] in {
defm VPACKSSWBY : PDI_binop_rm_int<0x63, "vpacksswb", int_x86_avx2_packsswb,
- VR256, memopv4i64, i256mem, 0, 0>, VEX_4V;
+ VR256, memopv4i64, i256mem,
+ SSE_INTALU_ITINS_P, 0, 0>, VEX_4V;
defm VPACKSSDWY : PDI_binop_rm_int<0x6B, "vpackssdw", int_x86_avx2_packssdw,
- VR256, memopv4i64, i256mem, 0, 0>, VEX_4V;
+ VR256, memopv4i64, i256mem,
+ SSE_INTALU_ITINS_P, 0, 0>, VEX_4V;
defm VPACKUSWBY : PDI_binop_rm_int<0x67, "vpackuswb", int_x86_avx2_packuswb,
- VR256, memopv4i64, i256mem, 0, 0>, VEX_4V;
+ VR256, memopv4i64, i256mem,
+ SSE_INTALU_ITINS_P, 0, 0>, VEX_4V;
}
let Constraints = "$src1 = $dst" in {
defm PACKSSWB : PDI_binop_rm_int<0x63, "packsswb", int_x86_sse2_packsswb_128,
- VR128, memopv2i64, i128mem>;
+ VR128, memopv2i64, i128mem,
+ SSE_INTALU_ITINS_P>;
defm PACKSSDW : PDI_binop_rm_int<0x6B, "packssdw", int_x86_sse2_packssdw_128,
- VR128, memopv2i64, i128mem>;
+ VR128, memopv2i64, i128mem,
+ SSE_INTALU_ITINS_P>;
defm PACKUSWB : PDI_binop_rm_int<0x67, "packuswb", int_x86_sse2_packuswb_128,
- VR128, memopv2i64, i128mem>;
+ VR128, memopv2i64, i128mem,
+ SSE_INTALU_ITINS_P>;
} // Constraints = "$src1 = $dst"
//===---------------------------------------------------------------------===//
(outs VR128:$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))))]>;
+ [(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))))]>;
+ (i8 imm:$src2))))],
+ IIC_SSE_PSHUF>;
}
multiclass sse2_pshuffle_y<string OpcodeStr, ValueType vt, SDNode OpNode> {
defm VPSHUFLW : sse2_pshuffle_y<"vpshuflw", v16i16, X86PShuflw>, XD, VEX;
}
-let Predicates = [HasSSE2] in {
+let Predicates = [UseSSE2] in {
let AddedComplexity = 5 in
defm PSHUFD : sse2_pshuffle<"pshufd", v4i32, X86PShufd>, TB, OpSize;
!if(Is2Addr,
!strconcat(OpcodeStr,"\t{$src2, $dst|$dst, $src2}"),
!strconcat(OpcodeStr,"\t{$src2, $src1, $dst|$dst, $src1, $src2}")),
- [(set VR128:$dst, (vt (OpNode VR128:$src1, VR128:$src2)))]>;
+ [(set VR128:$dst, (vt (OpNode VR128:$src1, VR128:$src2)))],
+ IIC_SSE_UNPCK>;
def rm : PDI<opc, MRMSrcMem,
(outs VR128:$dst), (ins VR128:$src1, i128mem:$src2),
!if(Is2Addr,
!strconcat(OpcodeStr,"\t{$src2, $src1, $dst|$dst, $src1, $src2}")),
[(set VR128:$dst, (OpNode VR128:$src1,
(bc_frag (memopv2i64
- addr:$src2))))]>;
+ addr:$src2))))],
+ IIC_SSE_UNPCK>;
}
multiclass sse2_unpack_y<bits<8> opc, string OpcodeStr, ValueType vt,
}
} // ExeDomain = SSEPackedInt
-// Patterns for using AVX1 instructions with integer vectors
-// Here to give AVX2 priority
-let Predicates = [HasAVX] in {
- def : Pat<(v8i32 (X86Unpckl VR256:$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:$src2)>;
-}
-
//===---------------------------------------------------------------------===//
// SSE2 - Packed Integer Extract and Insert
//===---------------------------------------------------------------------===//
"pinsrw\t{$src3, $src2, $dst|$dst, $src2, $src3}",
"vpinsrw\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}"),
[(set VR128:$dst,
- (X86pinsrw VR128:$src1, GR32:$src2, imm:$src3))]>;
+ (X86pinsrw VR128:$src1, GR32:$src2, imm:$src3))], IIC_SSE_PINSRW>;
def rmi : Ii8<0xC4, MRMSrcMem,
(outs VR128:$dst), (ins VR128:$src1,
i16mem:$src2, i32i8imm:$src3),
"vpinsrw\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}"),
[(set VR128:$dst,
(X86pinsrw VR128:$src1, (extloadi16 addr:$src2),
- imm:$src3))]>;
+ imm:$src3))], IIC_SSE_PINSRW>;
}
// Extract
(outs GR32:$dst), (ins VR128:$src1, i32i8imm:$src2),
"pextrw\t{$src2, $src1, $dst|$dst, $src1, $src2}",
[(set GR32:$dst, (X86pextrw (v8i16 VR128:$src1),
- imm:$src2))]>;
+ imm:$src2))], IIC_SSE_PEXTRW>;
// Insert
let Predicates = [HasAVX] in {
}
let Constraints = "$src1 = $dst" in
- defm PINSRW : sse2_pinsrw, TB, OpSize, Requires<[HasSSE2]>;
+ defm PINSRW : sse2_pinsrw, TB, OpSize, Requires<[UseSSE2]>;
} // ExeDomain = SSEPackedInt
def VPMOVMSKBrr : VPDI<0xD7, MRMSrcReg, (outs GR32:$dst), (ins VR128:$src),
"pmovmskb\t{$src, $dst|$dst, $src}",
- [(set GR32:$dst, (int_x86_sse2_pmovmskb_128 VR128:$src))]>, VEX;
+ [(set GR32:$dst, (int_x86_sse2_pmovmskb_128 VR128:$src))],
+ IIC_SSE_MOVMSK>, VEX;
def VPMOVMSKBr64r : VPDI<0xD7, MRMSrcReg, (outs GR64:$dst), (ins VR128:$src),
- "pmovmskb\t{$src, $dst|$dst, $src}", []>, VEX;
+ "pmovmskb\t{$src, $dst|$dst, $src}", [], IIC_SSE_MOVMSK>, VEX;
let Predicates = [HasAVX2] in {
def VPMOVMSKBYrr : VPDI<0xD7, MRMSrcReg, (outs GR32:$dst), (ins VR256:$src),
def PMOVMSKBrr : PDI<0xD7, MRMSrcReg, (outs GR32:$dst), (ins VR128:$src),
"pmovmskb\t{$src, $dst|$dst, $src}",
- [(set GR32:$dst, (int_x86_sse2_pmovmskb_128 VR128:$src))]>;
+ [(set GR32:$dst, (int_x86_sse2_pmovmskb_128 VR128:$src))],
+ IIC_SSE_MOVMSK>;
} // ExeDomain = SSEPackedInt
def VMASKMOVDQU : VPDI<0xF7, MRMSrcReg, (outs),
(ins VR128:$src, VR128:$mask),
"maskmovdqu\t{$mask, $src|$src, $mask}",
- [(int_x86_sse2_maskmov_dqu VR128:$src, VR128:$mask, EDI)]>, VEX;
+ [(int_x86_sse2_maskmov_dqu VR128:$src, VR128:$mask, EDI)],
+ IIC_SSE_MASKMOV>, VEX;
let Uses = [RDI] in
def VMASKMOVDQU64 : VPDI<0xF7, MRMSrcReg, (outs),
(ins VR128:$src, VR128:$mask),
"maskmovdqu\t{$mask, $src|$src, $mask}",
- [(int_x86_sse2_maskmov_dqu VR128:$src, VR128:$mask, RDI)]>, VEX;
+ [(int_x86_sse2_maskmov_dqu VR128:$src, VR128:$mask, RDI)],
+ IIC_SSE_MASKMOV>, VEX;
let Uses = [EDI] in
def MASKMOVDQU : PDI<0xF7, MRMSrcReg, (outs), (ins VR128:$src, VR128:$mask),
"maskmovdqu\t{$mask, $src|$src, $mask}",
- [(int_x86_sse2_maskmov_dqu VR128:$src, VR128:$mask, EDI)]>;
+ [(int_x86_sse2_maskmov_dqu VR128:$src, VR128:$mask, EDI)],
+ IIC_SSE_MASKMOV>;
let Uses = [RDI] in
def MASKMOVDQU64 : PDI<0xF7, MRMSrcReg, (outs), (ins VR128:$src, VR128:$mask),
"maskmovdqu\t{$mask, $src|$src, $mask}",
- [(int_x86_sse2_maskmov_dqu VR128:$src, VR128:$mask, RDI)]>;
+ [(int_x86_sse2_maskmov_dqu VR128:$src, VR128:$mask, RDI)],
+ IIC_SSE_MASKMOV>;
} // ExeDomain = SSEPackedInt
def VMOVDI2PDIrr : VPDI<0x6E, MRMSrcReg, (outs VR128:$dst), (ins GR32:$src),
"movd\t{$src, $dst|$dst, $src}",
[(set VR128:$dst,
- (v4i32 (scalar_to_vector GR32:$src)))]>, VEX;
+ (v4i32 (scalar_to_vector GR32:$src)))], IIC_SSE_MOVDQ>,
+ VEX;
def VMOVDI2PDIrm : VPDI<0x6E, MRMSrcMem, (outs VR128:$dst), (ins i32mem:$src),
"movd\t{$src, $dst|$dst, $src}",
[(set VR128:$dst,
- (v4i32 (scalar_to_vector (loadi32 addr:$src))))]>,
+ (v4i32 (scalar_to_vector (loadi32 addr:$src))))],
+ IIC_SSE_MOVDQ>,
VEX;
def VMOV64toPQIrr : VRPDI<0x6E, MRMSrcReg, (outs VR128:$dst), (ins GR64:$src),
"mov{d|q}\t{$src, $dst|$dst, $src}",
[(set VR128:$dst,
- (v2i64 (scalar_to_vector GR64:$src)))]>, VEX;
+ (v2i64 (scalar_to_vector GR64:$src)))],
+ IIC_SSE_MOVDQ>, VEX;
def VMOV64toSDrr : VRPDI<0x6E, MRMSrcReg, (outs FR64:$dst), (ins GR64:$src),
"mov{d|q}\t{$src, $dst|$dst, $src}",
- [(set FR64:$dst, (bitconvert GR64:$src))]>, VEX;
+ [(set FR64:$dst, (bitconvert GR64:$src))],
+ IIC_SSE_MOVDQ>, VEX;
def MOVDI2PDIrr : PDI<0x6E, MRMSrcReg, (outs VR128:$dst), (ins GR32:$src),
"movd\t{$src, $dst|$dst, $src}",
[(set VR128:$dst,
- (v4i32 (scalar_to_vector GR32:$src)))]>;
+ (v4i32 (scalar_to_vector GR32:$src)))], IIC_SSE_MOVDQ>;
def MOVDI2PDIrm : PDI<0x6E, MRMSrcMem, (outs VR128:$dst), (ins i32mem:$src),
"movd\t{$src, $dst|$dst, $src}",
[(set VR128:$dst,
- (v4i32 (scalar_to_vector (loadi32 addr:$src))))]>;
+ (v4i32 (scalar_to_vector (loadi32 addr:$src))))],
+ IIC_SSE_MOVDQ>;
def MOV64toPQIrr : RPDI<0x6E, MRMSrcReg, (outs VR128:$dst), (ins GR64:$src),
"mov{d|q}\t{$src, $dst|$dst, $src}",
[(set VR128:$dst,
- (v2i64 (scalar_to_vector GR64:$src)))]>;
+ (v2i64 (scalar_to_vector GR64:$src)))],
+ IIC_SSE_MOVDQ>;
def MOV64toSDrr : RPDI<0x6E, MRMSrcReg, (outs FR64:$dst), (ins GR64:$src),
"mov{d|q}\t{$src, $dst|$dst, $src}",
- [(set FR64:$dst, (bitconvert GR64:$src))]>;
+ [(set FR64:$dst, (bitconvert GR64:$src))],
+ IIC_SSE_MOVDQ>;
//===---------------------------------------------------------------------===//
// Move Int Doubleword to Single Scalar
//
def VMOVDI2SSrr : VPDI<0x6E, MRMSrcReg, (outs FR32:$dst), (ins GR32:$src),
"movd\t{$src, $dst|$dst, $src}",
- [(set FR32:$dst, (bitconvert GR32:$src))]>, VEX;
+ [(set FR32:$dst, (bitconvert GR32:$src))],
+ IIC_SSE_MOVDQ>, VEX;
def VMOVDI2SSrm : VPDI<0x6E, MRMSrcMem, (outs FR32:$dst), (ins i32mem:$src),
"movd\t{$src, $dst|$dst, $src}",
- [(set FR32:$dst, (bitconvert (loadi32 addr:$src)))]>,
+ [(set FR32:$dst, (bitconvert (loadi32 addr:$src)))],
+ IIC_SSE_MOVDQ>,
VEX;
def MOVDI2SSrr : PDI<0x6E, MRMSrcReg, (outs FR32:$dst), (ins GR32:$src),
"movd\t{$src, $dst|$dst, $src}",
- [(set FR32:$dst, (bitconvert GR32:$src))]>;
+ [(set FR32:$dst, (bitconvert GR32:$src))],
+ IIC_SSE_MOVDQ>;
def MOVDI2SSrm : PDI<0x6E, MRMSrcMem, (outs FR32:$dst), (ins i32mem:$src),
"movd\t{$src, $dst|$dst, $src}",
- [(set FR32:$dst, (bitconvert (loadi32 addr:$src)))]>;
+ [(set FR32:$dst, (bitconvert (loadi32 addr:$src)))],
+ IIC_SSE_MOVDQ>;
//===---------------------------------------------------------------------===//
// Move Packed Doubleword Int to Packed Double Int
def VMOVPDI2DIrr : VPDI<0x7E, MRMDestReg, (outs GR32:$dst), (ins VR128:$src),
"movd\t{$src, $dst|$dst, $src}",
[(set GR32:$dst, (vector_extract (v4i32 VR128:$src),
- (iPTR 0)))]>, VEX;
+ (iPTR 0)))], IIC_SSE_MOVD_ToGP>, VEX;
def VMOVPDI2DImr : VPDI<0x7E, MRMDestMem, (outs),
(ins i32mem:$dst, VR128:$src),
"movd\t{$src, $dst|$dst, $src}",
[(store (i32 (vector_extract (v4i32 VR128:$src),
- (iPTR 0))), addr:$dst)]>, VEX;
+ (iPTR 0))), addr:$dst)], IIC_SSE_MOVDQ>,
+ VEX;
def MOVPDI2DIrr : PDI<0x7E, MRMDestReg, (outs GR32:$dst), (ins VR128:$src),
"movd\t{$src, $dst|$dst, $src}",
[(set GR32:$dst, (vector_extract (v4i32 VR128:$src),
- (iPTR 0)))]>;
+ (iPTR 0)))], IIC_SSE_MOVD_ToGP>;
def MOVPDI2DImr : PDI<0x7E, MRMDestMem, (outs), (ins i32mem:$dst, VR128:$src),
"movd\t{$src, $dst|$dst, $src}",
[(store (i32 (vector_extract (v4i32 VR128:$src),
- (iPTR 0))), addr:$dst)]>;
+ (iPTR 0))), addr:$dst)],
+ IIC_SSE_MOVDQ>;
//===---------------------------------------------------------------------===//
// 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:$src),
- (iPTR 0)))]>,
+ (iPTR 0)))],
+ IIC_SSE_MOVD_ToGP>,
TB, OpSize, VEX, VEX_W, Requires<[HasAVX, In64BitMode]>;
def MOVPQIto64rr : RPDI<0x7E, MRMDestReg, (outs GR64:$dst), (ins VR128:$src),
"mov{d|q}\t{$src, $dst|$dst, $src}",
[(set GR64:$dst, (vector_extract (v2i64 VR128:$src),
- (iPTR 0)))]>;
+ (iPTR 0)))],
+ IIC_SSE_MOVD_ToGP>;
//===---------------------------------------------------------------------===//
// Bitcast FR64 <-> GR64
//
let Predicates = [HasAVX] in
-def VMOV64toSDrm : S3SI<0x7E, MRMSrcMem, (outs FR64:$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:$src)))]>,
VEX;
def VMOVSDto64rr : VRPDI<0x7E, MRMDestReg, (outs GR64:$dst), (ins FR64:$src),
"mov{d|q}\t{$src, $dst|$dst, $src}",
- [(set GR64:$dst, (bitconvert FR64:$src))]>, VEX;
+ [(set GR64:$dst, (bitconvert FR64:$src))],
+ IIC_SSE_MOVDQ>, VEX;
def VMOVSDto64mr : VRPDI<0x7E, MRMDestMem, (outs), (ins i64mem:$dst, FR64:$src),
"movq\t{$src, $dst|$dst, $src}",
- [(store (i64 (bitconvert FR64:$src)), addr:$dst)]>,
- VEX;
+ [(store (i64 (bitconvert FR64:$src)), addr:$dst)],
+ IIC_SSE_MOVDQ>, VEX;
-def MOV64toSDrm : S3SI<0x7E, MRMSrcMem, (outs FR64:$dst), (ins i64mem:$src),
+def MOV64toSDrm : S2SI<0x7E, MRMSrcMem, (outs FR64:$dst), (ins i64mem:$src),
"movq\t{$src, $dst|$dst, $src}",
- [(set FR64:$dst, (bitconvert (loadi64 addr:$src)))]>;
+ [(set FR64:$dst, (bitconvert (loadi64 addr:$src)))],
+ IIC_SSE_MOVDQ>;
def MOVSDto64rr : RPDI<0x7E, MRMDestReg, (outs GR64:$dst), (ins FR64:$src),
"mov{d|q}\t{$src, $dst|$dst, $src}",
- [(set GR64:$dst, (bitconvert FR64:$src))]>;
+ [(set GR64:$dst, (bitconvert FR64:$src))],
+ IIC_SSE_MOVD_ToGP>;
def MOVSDto64mr : RPDI<0x7E, MRMDestMem, (outs), (ins i64mem:$dst, FR64:$src),
"movq\t{$src, $dst|$dst, $src}",
- [(store (i64 (bitconvert FR64:$src)), addr:$dst)]>;
+ [(store (i64 (bitconvert FR64:$src)), addr:$dst)],
+ IIC_SSE_MOVDQ>;
//===---------------------------------------------------------------------===//
// Move Scalar Single to Double Int
//
def VMOVSS2DIrr : VPDI<0x7E, MRMDestReg, (outs GR32:$dst), (ins FR32:$src),
"movd\t{$src, $dst|$dst, $src}",
- [(set GR32:$dst, (bitconvert FR32:$src))]>, VEX;
+ [(set GR32:$dst, (bitconvert FR32:$src))],
+ IIC_SSE_MOVD_ToGP>, VEX;
def VMOVSS2DImr : VPDI<0x7E, MRMDestMem, (outs), (ins i32mem:$dst, FR32:$src),
"movd\t{$src, $dst|$dst, $src}",
- [(store (i32 (bitconvert FR32:$src)), addr:$dst)]>, VEX;
+ [(store (i32 (bitconvert FR32:$src)), addr:$dst)],
+ IIC_SSE_MOVDQ>, VEX;
def MOVSS2DIrr : PDI<0x7E, MRMDestReg, (outs GR32:$dst), (ins FR32:$src),
"movd\t{$src, $dst|$dst, $src}",
- [(set GR32:$dst, (bitconvert FR32:$src))]>;
+ [(set GR32:$dst, (bitconvert FR32:$src))],
+ IIC_SSE_MOVD_ToGP>;
def MOVSS2DImr : PDI<0x7E, MRMDestMem, (outs), (ins i32mem:$dst, FR32:$src),
"movd\t{$src, $dst|$dst, $src}",
- [(store (i32 (bitconvert FR32:$src)), addr:$dst)]>;
+ [(store (i32 (bitconvert FR32:$src)), addr:$dst)],
+ IIC_SSE_MOVDQ>;
//===---------------------------------------------------------------------===//
// Patterns and instructions to describe movd/movq to XMM register zero-extends
def VMOVZDI2PDIrr : VPDI<0x6E, MRMSrcReg, (outs VR128:$dst), (ins GR32:$src),
"movd\t{$src, $dst|$dst, $src}",
[(set VR128:$dst, (v4i32 (X86vzmovl
- (v4i32 (scalar_to_vector GR32:$src)))))]>,
- VEX;
+ (v4i32 (scalar_to_vector GR32:$src)))))],
+ IIC_SSE_MOVDQ>, VEX;
def VMOVZQI2PQIrr : VPDI<0x6E, MRMSrcReg, (outs VR128:$dst), (ins GR64:$src),
"mov{d|q}\t{$src, $dst|$dst, $src}", // X86-64 only
[(set VR128:$dst, (v2i64 (X86vzmovl
- (v2i64 (scalar_to_vector GR64:$src)))))]>,
+ (v2i64 (scalar_to_vector GR64:$src)))))],
+ IIC_SSE_MOVDQ>,
VEX, VEX_W;
}
let AddedComplexity = 15 in {
def MOVZDI2PDIrr : PDI<0x6E, MRMSrcReg, (outs VR128:$dst), (ins GR32:$src),
"movd\t{$src, $dst|$dst, $src}",
[(set VR128:$dst, (v4i32 (X86vzmovl
- (v4i32 (scalar_to_vector GR32:$src)))))]>;
+ (v4i32 (scalar_to_vector GR32:$src)))))],
+ IIC_SSE_MOVDQ>;
def MOVZQI2PQIrr : RPDI<0x6E, MRMSrcReg, (outs VR128:$dst), (ins GR64:$src),
"mov{d|q}\t{$src, $dst|$dst, $src}", // X86-64 only
[(set VR128:$dst, (v2i64 (X86vzmovl
- (v2i64 (scalar_to_vector GR64:$src)))))]>;
+ (v2i64 (scalar_to_vector GR64:$src)))))],
+ IIC_SSE_MOVDQ>;
}
let AddedComplexity = 20 in {
"movd\t{$src, $dst|$dst, $src}",
[(set VR128:$dst,
(v4i32 (X86vzmovl (v4i32 (scalar_to_vector
- (loadi32 addr:$src))))))]>,
- VEX;
+ (loadi32 addr:$src))))))],
+ IIC_SSE_MOVDQ>, VEX;
def MOVZDI2PDIrm : PDI<0x6E, MRMSrcMem, (outs VR128:$dst), (ins i32mem:$src),
"movd\t{$src, $dst|$dst, $src}",
[(set VR128:$dst,
(v4i32 (X86vzmovl (v4i32 (scalar_to_vector
- (loadi32 addr:$src))))))]>;
+ (loadi32 addr:$src))))))],
+ IIC_SSE_MOVDQ>;
}
let Predicates = [HasAVX] in {
(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)))),
def MOVQI2PQIrm : I<0x7E, MRMSrcMem, (outs VR128:$dst), (ins i64mem:$src),
"movq\t{$src, $dst|$dst, $src}",
[(set VR128:$dst,
- (v2i64 (scalar_to_vector (loadi64 addr:$src))))]>, XS,
- Requires<[HasSSE2]>; // SSE2 instruction with XS Prefix
+ (v2i64 (scalar_to_vector (loadi64 addr:$src))))],
+ IIC_SSE_MOVDQ>, XS,
+ Requires<[UseSSE2]>; // SSE2 instruction with XS Prefix
//===---------------------------------------------------------------------===//
// Move Packed Quadword Int to Quadword Int
def VMOVPQI2QImr : VPDI<0xD6, MRMDestMem, (outs), (ins i64mem:$dst, VR128:$src),
"movq\t{$src, $dst|$dst, $src}",
[(store (i64 (vector_extract (v2i64 VR128:$src),
- (iPTR 0))), addr:$dst)]>, VEX;
+ (iPTR 0))), addr:$dst)],
+ IIC_SSE_MOVDQ>, VEX;
def MOVPQI2QImr : PDI<0xD6, MRMDestMem, (outs), (ins i64mem:$dst, VR128:$src),
"movq\t{$src, $dst|$dst, $src}",
[(store (i64 (vector_extract (v2i64 VR128:$src),
- (iPTR 0))), addr:$dst)]>;
+ (iPTR 0))), addr:$dst)],
+ IIC_SSE_MOVDQ>;
//===---------------------------------------------------------------------===//
// Store / copy lower 64-bits of a XMM register.
[(int_x86_sse2_storel_dq addr:$dst, VR128:$src)]>, VEX;
def MOVLQ128mr : PDI<0xD6, MRMDestMem, (outs), (ins i64mem:$dst, VR128:$src),
"movq\t{$src, $dst|$dst, $src}",
- [(int_x86_sse2_storel_dq addr:$dst, VR128:$src)]>;
+ [(int_x86_sse2_storel_dq addr:$dst, VR128:$src)],
+ IIC_SSE_MOVDQ>;
let AddedComplexity = 20 in
def VMOVZQI2PQIrm : I<0x7E, MRMSrcMem, (outs VR128:$dst), (ins i64mem:$src),
"vmovq\t{$src, $dst|$dst, $src}",
[(set VR128:$dst,
(v2i64 (X86vzmovl (v2i64 (scalar_to_vector
- (loadi64 addr:$src))))))]>,
+ (loadi64 addr:$src))))))],
+ IIC_SSE_MOVDQ>,
XS, VEX, Requires<[HasAVX]>;
let AddedComplexity = 20 in
"movq\t{$src, $dst|$dst, $src}",
[(set VR128:$dst,
(v2i64 (X86vzmovl (v2i64 (scalar_to_vector
- (loadi64 addr:$src))))))]>,
- XS, Requires<[HasSSE2]>;
+ (loadi64 addr:$src))))))],
+ IIC_SSE_MOVDQ>,
+ 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:$src)))),
}
let Predicates = [HasAVX] in {
-def : Pat<(v4i64 (X86vzload addr:$src)),
+def : Pat<(v4i64 (alignedX86vzload addr:$src)),
(SUBREG_TO_REG (i32 0), (VMOVAPSrm addr:$src), sub_xmm)>;
+def : Pat<(v4i64 (X86vzload addr:$src)),
+ (SUBREG_TO_REG (i32 0), (VMOVUPSrm addr:$src), sub_xmm)>;
}
//===---------------------------------------------------------------------===//
let AddedComplexity = 15 in
def VMOVZPQILo2PQIrr : I<0x7E, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
"vmovq\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst, (v2i64 (X86vzmovl (v2i64 VR128:$src))))]>,
+ [(set VR128:$dst, (v2i64 (X86vzmovl (v2i64 VR128:$src))))],
+ IIC_SSE_MOVQ_RR>,
XS, VEX, Requires<[HasAVX]>;
let AddedComplexity = 15 in
def MOVZPQILo2PQIrr : I<0x7E, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
"movq\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst, (v2i64 (X86vzmovl (v2i64 VR128:$src))))]>,
- XS, Requires<[HasSSE2]>;
+ [(set VR128:$dst, (v2i64 (X86vzmovl (v2i64 VR128:$src))))],
+ IIC_SSE_MOVQ_RR>,
+ XS, Requires<[UseSSE2]>;
let AddedComplexity = 20 in
def VMOVZPQILo2PQIrm : I<0x7E, MRMSrcMem, (outs VR128:$dst), (ins i128mem:$src),
"vmovq\t{$src, $dst|$dst, $src}",
[(set VR128:$dst, (v2i64 (X86vzmovl
- (loadv2i64 addr:$src))))]>,
+ (loadv2i64 addr:$src))))],
+ IIC_SSE_MOVDQ>,
XS, VEX, Requires<[HasAVX]>;
let AddedComplexity = 20 in {
def MOVZPQILo2PQIrm : I<0x7E, MRMSrcMem, (outs VR128:$dst), (ins i128mem:$src),
"movq\t{$src, $dst|$dst, $src}",
[(set VR128:$dst, (v2i64 (X86vzmovl
- (loadv2i64 addr:$src))))]>,
- XS, Requires<[HasSSE2]>;
+ (loadv2i64 addr:$src))))],
+ IIC_SSE_MOVDQ>,
+ 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:$src))),
// Instructions to match in the assembler
def VMOVQs64rr : VPDI<0x6E, MRMSrcReg, (outs VR128:$dst), (ins GR64:$src),
- "movq\t{$src, $dst|$dst, $src}", []>, VEX, VEX_W;
+ "movq\t{$src, $dst|$dst, $src}", [],
+ IIC_SSE_MOVDQ>, VEX, VEX_W;
def VMOVQd64rr : VPDI<0x7E, MRMDestReg, (outs GR64:$dst), (ins VR128:$src),
- "movq\t{$src, $dst|$dst, $src}", []>, VEX, VEX_W;
+ "movq\t{$src, $dst|$dst, $src}", [],
+ IIC_SSE_MOVDQ>, VEX, VEX_W;
// Recognize "movd" with GR64 destination, but encode as a "movq"
def VMOVQd64rr_alt : VPDI<0x7E, MRMDestReg, (outs GR64:$dst), (ins VR128:$src),
- "movd\t{$src, $dst|$dst, $src}", []>, VEX, VEX_W;
+ "movd\t{$src, $dst|$dst, $src}", [],
+ IIC_SSE_MOVDQ>, VEX, VEX_W;
// Instructions for the disassembler
// xr = XMM register
def VMOVQxrxr: I<0x7E, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
"vmovq\t{$src, $dst|$dst, $src}", []>, VEX, XS;
def MOVQxrxr : I<0x7E, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
- "movq\t{$src, $dst|$dst, $src}", []>, XS;
-
-//===---------------------------------------------------------------------===//
-// SSE3 - Conversion Instructions
-//===---------------------------------------------------------------------===//
-
-// Convert Packed Double FP to Packed DW Integers
-let Predicates = [HasAVX] in {
-// The assembler can recognize rr 256-bit instructions by seeing a ymm
-// register, but the same isn't true when using memory operands instead.
-// Provide other assembly rr and rm forms to address this explicitly.
-def VCVTPD2DQrr : S3DI<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
- "vcvtpd2dq\t{$src, $dst|$dst, $src}", []>, VEX;
-def VCVTPD2DQXrYr : S3DI<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR256:$src),
- "vcvtpd2dq\t{$src, $dst|$dst, $src}", []>, VEX;
-
-// XMM only
-def VCVTPD2DQXrr : S3DI<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
- "vcvtpd2dqx\t{$src, $dst|$dst, $src}", []>, VEX;
-def VCVTPD2DQXrm : S3DI<0xE6, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
- "vcvtpd2dqx\t{$src, $dst|$dst, $src}", []>, VEX;
-
-// YMM only
-def VCVTPD2DQYrr : S3DI<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR256:$src),
- "vcvtpd2dqy\t{$src, $dst|$dst, $src}", []>, VEX;
-def VCVTPD2DQYrm : S3DI<0xE6, MRMSrcMem, (outs VR128:$dst), (ins f256mem:$src),
- "vcvtpd2dqy\t{$src, $dst|$dst, $src}", []>, VEX, VEX_L;
-}
-
-def CVTPD2DQrm : S3DI<0xE6, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
- "cvtpd2dq\t{$src, $dst|$dst, $src}", []>;
-def CVTPD2DQrr : S3DI<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
- "cvtpd2dq\t{$src, $dst|$dst, $src}", []>;
-
-def : Pat<(v4i32 (fp_to_sint (v4f64 VR256:$src))),
- (VCVTTPD2DQYrr VR256:$src)>;
-def : Pat<(v4i32 (fp_to_sint (memopv4f64 addr:$src))),
- (VCVTTPD2DQYrm addr:$src)>;
-
-// Convert Packed DW Integers to Packed Double FP
-let Predicates = [HasAVX] in {
-def VCVTDQ2PDrm : S3SI<0xE6, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
- "vcvtdq2pd\t{$src, $dst|$dst, $src}", []>, VEX;
-def VCVTDQ2PDrr : S3SI<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
- "vcvtdq2pd\t{$src, $dst|$dst, $src}", []>, VEX;
-def VCVTDQ2PDYrm : S3SI<0xE6, MRMSrcMem, (outs VR256:$dst), (ins f128mem:$src),
- "vcvtdq2pd\t{$src, $dst|$dst, $src}", []>, VEX;
-def VCVTDQ2PDYrr : S3SI<0xE6, MRMSrcReg, (outs VR256:$dst), (ins VR128:$src),
- "vcvtdq2pd\t{$src, $dst|$dst, $src}", []>, VEX;
-}
-
-def CVTDQ2PDrm : S3SI<0xE6, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
- "cvtdq2pd\t{$src, $dst|$dst, $src}", []>;
-def CVTDQ2PDrr : S3SI<0xE6, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
- "cvtdq2pd\t{$src, $dst|$dst, $src}", []>;
-
-// AVX 256-bit register conversion intrinsics
-def : Pat<(int_x86_avx_cvtdq2_pd_256 VR128:$src),
- (VCVTDQ2PDYrr VR128:$src)>;
-def : Pat<(int_x86_avx_cvtdq2_pd_256 (bitconvert (memopv2i64 addr:$src))),
- (VCVTDQ2PDYrm addr:$src)>;
-
-def : Pat<(int_x86_avx_cvt_pd2dq_256 VR256:$src),
- (VCVTPD2DQYrr VR256:$src)>;
-def : Pat<(int_x86_avx_cvt_pd2dq_256 (memopv4f64 addr:$src)),
- (VCVTPD2DQYrm addr:$src)>;
-
-def : Pat<(v4f64 (sint_to_fp (v4i32 VR128:$src))),
- (VCVTDQ2PDYrr VR128:$src)>;
-def : Pat<(v4f64 (sint_to_fp (bc_v4i32 (memopv2i64 addr:$src)))),
- (VCVTDQ2PDYrm addr:$src)>;
+ "movq\t{$src, $dst|$dst, $src}", [], IIC_SSE_MOVQ_RR>, XS;
//===---------------------------------------------------------------------===//
// SSE3 - Replicate Single FP - MOVSHDUP and MOVSLDUP
X86MemOperand x86memop> {
def rr : S3SI<op, MRMSrcReg, (outs RC:$dst), (ins RC:$src),
!strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"),
- [(set RC:$dst, (vt (OpNode RC:$src)))]>;
+ [(set RC:$dst, (vt (OpNode RC:$src)))],
+ IIC_SSE_MOV_LH>;
def rm : S3SI<op, MRMSrcMem, (outs RC:$dst), (ins x86memop:$src),
!strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"),
- [(set RC:$dst, (OpNode (mem_frag addr:$src)))]>;
+ [(set RC:$dst, (OpNode (mem_frag addr:$src)))],
+ IIC_SSE_MOV_LH>;
}
let Predicates = [HasAVX] in {
(VMOVSLDUPYrm addr:$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:$src)))),
let neverHasSideEffects = 1 in
def rr : S3DI<0x12, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
!strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"),
- []>;
+ [], IIC_SSE_MOV_LH>;
def rm : S3DI<0x12, MRMSrcMem, (outs VR128:$dst), (ins f64mem:$src),
!strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"),
[(set VR128:$dst,
(v2f64 (X86Movddup
- (scalar_to_vector (loadf64 addr:$src)))))]>;
+ (scalar_to_vector (loadf64 addr:$src)))))],
+ IIC_SSE_MOV_LH>;
}
// FIXME: Merge with above classe when there're patterns for the ymm version
(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))),
}
def LDDQUrm : S3DI<0xF0, MRMSrcMem, (outs VR128:$dst), (ins i128mem:$src),
"lddqu\t{$src, $dst|$dst, $src}",
- [(set VR128:$dst, (int_x86_sse3_ldu_dq addr:$src))]>;
+ [(set VR128:$dst, (int_x86_sse3_ldu_dq addr:$src))],
+ IIC_SSE_LDDQU>;
//===---------------------------------------------------------------------===//
// SSE3 - Arithmetic
//===---------------------------------------------------------------------===//
multiclass sse3_addsub<Intrinsic Int, string OpcodeStr, RegisterClass RC,
- X86MemOperand x86memop, bit Is2Addr = 1> {
+ X86MemOperand x86memop, OpndItins itins,
+ bit Is2Addr = 1> {
def rr : I<0xD0, MRMSrcReg,
(outs RC:$dst), (ins RC:$src1, RC:$src2),
!if(Is2Addr,
!strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"),
!strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}")),
- [(set RC:$dst, (Int RC:$src1, RC:$src2))]>;
+ [(set RC:$dst, (Int RC:$src1, RC:$src2))], itins.rr>;
def rm : I<0xD0, MRMSrcMem,
(outs RC:$dst), (ins RC:$src1, x86memop:$src2),
!if(Is2Addr,
!strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"),
!strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}")),
- [(set RC:$dst, (Int RC:$src1, (memop addr:$src2)))]>;
+ [(set RC:$dst, (Int RC:$src1, (memop addr:$src2)))], itins.rr>;
}
let Predicates = [HasAVX] in {
let ExeDomain = SSEPackedSingle in {
defm VADDSUBPS : sse3_addsub<int_x86_sse3_addsub_ps, "vaddsubps", VR128,
- f128mem, 0>, TB, XD, VEX_4V;
+ f128mem, SSE_ALU_F32P, 0>, TB, XD, VEX_4V;
defm VADDSUBPSY : sse3_addsub<int_x86_avx_addsub_ps_256, "vaddsubps", VR256,
- f256mem, 0>, TB, XD, VEX_4V;
+ f256mem, SSE_ALU_F32P, 0>, TB, XD, VEX_4V;
}
let ExeDomain = SSEPackedDouble in {
defm VADDSUBPD : sse3_addsub<int_x86_sse3_addsub_pd, "vaddsubpd", VR128,
- f128mem, 0>, TB, OpSize, VEX_4V;
+ f128mem, SSE_ALU_F64P, 0>, TB, OpSize, VEX_4V;
defm VADDSUBPDY : sse3_addsub<int_x86_avx_addsub_pd_256, "vaddsubpd", VR256,
- f256mem, 0>, TB, OpSize, VEX_4V;
+ f256mem, SSE_ALU_F64P, 0>, TB, OpSize, VEX_4V;
}
}
-let Constraints = "$src1 = $dst", Predicates = [HasSSE3] in {
+let Constraints = "$src1 = $dst", Predicates = [UseSSE3] in {
let ExeDomain = SSEPackedSingle in
defm ADDSUBPS : sse3_addsub<int_x86_sse3_addsub_ps, "addsubps", VR128,
- f128mem>, TB, XD;
+ f128mem, SSE_ALU_F32P>, TB, XD;
let ExeDomain = SSEPackedDouble in
defm ADDSUBPD : sse3_addsub<int_x86_sse3_addsub_pd, "addsubpd", VR128,
- f128mem>, TB, OpSize;
+ f128mem, SSE_ALU_F64P>, TB, OpSize;
}
//===---------------------------------------------------------------------===//
!if(Is2Addr,
!strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"),
!strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}")),
- [(set RC:$dst, (vt (OpNode RC:$src1, RC:$src2)))]>;
+ [(set RC:$dst, (vt (OpNode RC:$src1, RC:$src2)))], IIC_SSE_HADDSUB_RR>;
def rm : S3DI<o, MRMSrcMem, (outs RC:$dst), (ins RC:$src1, x86memop:$src2),
!if(Is2Addr,
!strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"),
!strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}")),
- [(set RC:$dst, (vt (OpNode RC:$src1, (memop addr:$src2))))]>;
+ [(set RC:$dst, (vt (OpNode RC:$src1, (memop addr:$src2))))],
+ IIC_SSE_HADDSUB_RM>;
}
multiclass S3_Int<bits<8> o, string OpcodeStr, ValueType vt, RegisterClass RC,
X86MemOperand x86memop, SDNode OpNode, bit Is2Addr = 1> {
!if(Is2Addr,
!strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"),
!strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}")),
- [(set RC:$dst, (vt (OpNode RC:$src1, RC:$src2)))]>;
+ [(set RC:$dst, (vt (OpNode RC:$src1, RC:$src2)))], IIC_SSE_HADDSUB_RR>;
def rm : S3I<o, MRMSrcMem, (outs RC:$dst), (ins RC:$src1, x86memop:$src2),
!if(Is2Addr,
!strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"),
!strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}")),
- [(set RC:$dst, (vt (OpNode RC:$src1, (memop addr:$src2))))]>;
+ [(set RC:$dst, (vt (OpNode RC:$src1, (memop addr:$src2))))],
+ IIC_SSE_HADDSUB_RM>;
}
let Predicates = [HasAVX] in {
def rr128 : SS38I<opc, MRMSrcReg, (outs VR128:$dst),
(ins VR128:$src),
!strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"),
- [(set VR128:$dst, (IntId128 VR128:$src))]>,
+ [(set VR128:$dst, (IntId128 VR128:$src))], IIC_SSE_PABS_RR>,
OpSize;
def rm128 : SS38I<opc, MRMSrcMem, (outs VR128:$dst),
!strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"),
[(set VR128:$dst,
(IntId128
- (bitconvert (memopv2i64 addr:$src))))]>, OpSize;
+ (bitconvert (memopv2i64 addr:$src))))], IIC_SSE_PABS_RM>,
+ OpSize;
}
/// SS3I_unop_rm_int_y - Simple SSSE3 unary op whose type can be v*{i8,i16,i32}.
// SSSE3 - Packed Binary Operator Instructions
//===---------------------------------------------------------------------===//
+def SSE_PHADDSUBD : OpndItins<
+ IIC_SSE_PHADDSUBD_RR, IIC_SSE_PHADDSUBD_RM
+>;
+def SSE_PHADDSUBSW : OpndItins<
+ IIC_SSE_PHADDSUBSW_RR, IIC_SSE_PHADDSUBSW_RM
+>;
+def SSE_PHADDSUBW : OpndItins<
+ IIC_SSE_PHADDSUBW_RR, IIC_SSE_PHADDSUBW_RM
+>;
+def SSE_PSHUFB : OpndItins<
+ IIC_SSE_PSHUFB_RR, IIC_SSE_PSHUFB_RM
+>;
+def SSE_PSIGN : OpndItins<
+ IIC_SSE_PSIGN_RR, IIC_SSE_PSIGN_RM
+>;
+def SSE_PMULHRSW : OpndItins<
+ IIC_SSE_PMULHRSW, IIC_SSE_PMULHRSW
+>;
+
/// SS3I_binop_rm - Simple SSSE3 bin op
multiclass SS3I_binop_rm<bits<8> opc, string OpcodeStr, SDNode OpNode,
ValueType OpVT, RegisterClass RC, PatFrag memop_frag,
- X86MemOperand x86memop, bit Is2Addr = 1> {
+ X86MemOperand x86memop, OpndItins itins,
+ bit Is2Addr = 1> {
let isCommutable = 1 in
def rr : SS38I<opc, MRMSrcReg, (outs RC:$dst),
(ins RC:$src1, RC:$src2),
!if(Is2Addr,
!strconcat(OpcodeStr, "\t{$src2, $dst|$dst, $src2}"),
!strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}")),
- [(set RC:$dst, (OpVT (OpNode RC:$src1, RC:$src2)))]>,
+ [(set RC:$dst, (OpVT (OpNode RC:$src1, RC:$src2)))], itins.rr>,
OpSize;
def rm : SS38I<opc, MRMSrcMem, (outs RC:$dst),
(ins RC:$src1, x86memop:$src2),
!strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}")),
[(set RC:$dst,
(OpVT (OpNode RC:$src1,
- (bitconvert (memop_frag addr:$src2)))))]>, OpSize;
+ (bitconvert (memop_frag addr:$src2)))))], itins.rm>, OpSize;
}
/// SS3I_binop_rm_int - Simple SSSE3 bin op whose type can be v*{i8,i16,i32}.
multiclass SS3I_binop_rm_int<bits<8> opc, string OpcodeStr,
- Intrinsic IntId128, bit Is2Addr = 1> {
+ Intrinsic IntId128, OpndItins itins,
+ bit Is2Addr = 1> {
let isCommutable = 1 in
def rr128 : SS38I<opc, MRMSrcReg, (outs VR128:$dst),
(ins VR128:$src1, VR128:$src2),
let ImmT = NoImm, Predicates = [HasAVX] in {
let isCommutable = 0 in {
defm VPHADDW : SS3I_binop_rm<0x01, "vphaddw", X86hadd, v8i16, VR128,
- memopv2i64, i128mem, 0>, VEX_4V;
+ memopv2i64, i128mem,
+ SSE_PHADDSUBW, 0>, VEX_4V;
defm VPHADDD : SS3I_binop_rm<0x02, "vphaddd", X86hadd, v4i32, VR128,
- memopv2i64, i128mem, 0>, VEX_4V;
+ memopv2i64, i128mem,
+ SSE_PHADDSUBD, 0>, VEX_4V;
defm VPHSUBW : SS3I_binop_rm<0x05, "vphsubw", X86hsub, v8i16, VR128,
- memopv2i64, i128mem, 0>, VEX_4V;
+ memopv2i64, i128mem,
+ SSE_PHADDSUBW, 0>, VEX_4V;
defm VPHSUBD : SS3I_binop_rm<0x06, "vphsubd", X86hsub, v4i32, VR128,
- memopv2i64, i128mem, 0>, VEX_4V;
+ memopv2i64, i128mem,
+ SSE_PHADDSUBD, 0>, VEX_4V;
defm VPSIGNB : SS3I_binop_rm<0x08, "vpsignb", X86psign, v16i8, VR128,
- memopv2i64, i128mem, 0>, VEX_4V;
+ memopv2i64, i128mem,
+ SSE_PSIGN, 0>, VEX_4V;
defm VPSIGNW : SS3I_binop_rm<0x09, "vpsignw", X86psign, v8i16, VR128,
- memopv2i64, i128mem, 0>, VEX_4V;
+ memopv2i64, i128mem,
+ SSE_PSIGN, 0>, VEX_4V;
defm VPSIGND : SS3I_binop_rm<0x0A, "vpsignd", X86psign, v4i32, VR128,
- memopv2i64, i128mem, 0>, VEX_4V;
+ memopv2i64, i128mem,
+ SSE_PSIGN, 0>, VEX_4V;
defm VPSHUFB : SS3I_binop_rm<0x00, "vpshufb", X86pshufb, v16i8, VR128,
- memopv2i64, i128mem, 0>, VEX_4V;
+ memopv2i64, i128mem,
+ SSE_PSHUFB, 0>, VEX_4V;
defm VPHADDSW : SS3I_binop_rm_int<0x03, "vphaddsw",
- int_x86_ssse3_phadd_sw_128, 0>, VEX_4V;
+ int_x86_ssse3_phadd_sw_128,
+ SSE_PHADDSUBSW, 0>, VEX_4V;
defm VPHSUBSW : SS3I_binop_rm_int<0x07, "vphsubsw",
- int_x86_ssse3_phsub_sw_128, 0>, VEX_4V;
+ int_x86_ssse3_phsub_sw_128,
+ SSE_PHADDSUBSW, 0>, VEX_4V;
defm VPMADDUBSW : SS3I_binop_rm_int<0x04, "vpmaddubsw",
- int_x86_ssse3_pmadd_ub_sw_128, 0>, VEX_4V;
+ int_x86_ssse3_pmadd_ub_sw_128,
+ SSE_PMADD, 0>, VEX_4V;
}
defm VPMULHRSW : SS3I_binop_rm_int<0x0B, "vpmulhrsw",
- int_x86_ssse3_pmul_hr_sw_128, 0>, VEX_4V;
+ int_x86_ssse3_pmul_hr_sw_128,
+ SSE_PMULHRSW, 0>, VEX_4V;
}
let ImmT = NoImm, Predicates = [HasAVX2] in {
let isCommutable = 0 in {
defm VPHADDWY : SS3I_binop_rm<0x01, "vphaddw", X86hadd, v16i16, VR256,
- memopv4i64, i256mem, 0>, VEX_4V;
+ memopv4i64, i256mem,
+ SSE_PHADDSUBW, 0>, VEX_4V;
defm VPHADDDY : SS3I_binop_rm<0x02, "vphaddd", X86hadd, v8i32, VR256,
- memopv4i64, i256mem, 0>, VEX_4V;
+ memopv4i64, i256mem,
+ SSE_PHADDSUBW, 0>, VEX_4V;
defm VPHSUBWY : SS3I_binop_rm<0x05, "vphsubw", X86hsub, v16i16, VR256,
- memopv4i64, i256mem, 0>, VEX_4V;
+ memopv4i64, i256mem,
+ SSE_PHADDSUBW, 0>, VEX_4V;
defm VPHSUBDY : SS3I_binop_rm<0x06, "vphsubd", X86hsub, v8i32, VR256,
- memopv4i64, i256mem, 0>, VEX_4V;
+ memopv4i64, i256mem,
+ SSE_PHADDSUBW, 0>, VEX_4V;
defm VPSIGNBY : SS3I_binop_rm<0x08, "vpsignb", X86psign, v32i8, VR256,
- memopv4i64, i256mem, 0>, VEX_4V;
+ memopv4i64, i256mem,
+ SSE_PHADDSUBW, 0>, VEX_4V;
defm VPSIGNWY : SS3I_binop_rm<0x09, "vpsignw", X86psign, v16i16, VR256,
- memopv4i64, i256mem, 0>, VEX_4V;
+ memopv4i64, i256mem,
+ SSE_PHADDSUBW, 0>, VEX_4V;
defm VPSIGNDY : SS3I_binop_rm<0x0A, "vpsignd", X86psign, v8i32, VR256,
- memopv4i64, i256mem, 0>, VEX_4V;
+ memopv4i64, i256mem,
+ SSE_PHADDSUBW, 0>, VEX_4V;
defm VPSHUFBY : SS3I_binop_rm<0x00, "vpshufb", X86pshufb, v32i8, VR256,
- memopv4i64, i256mem, 0>, VEX_4V;
+ memopv4i64, i256mem,
+ SSE_PHADDSUBW, 0>, VEX_4V;
defm VPHADDSW : SS3I_binop_rm_int_y<0x03, "vphaddsw",
int_x86_avx2_phadd_sw>, VEX_4V;
defm VPHSUBSW : SS3I_binop_rm_int_y<0x07, "vphsubsw",
let ImmT = NoImm, Constraints = "$src1 = $dst" in {
let isCommutable = 0 in {
defm PHADDW : SS3I_binop_rm<0x01, "phaddw", X86hadd, v8i16, VR128,
- memopv2i64, i128mem>;
+ memopv2i64, i128mem, SSE_PHADDSUBW>;
defm PHADDD : SS3I_binop_rm<0x02, "phaddd", X86hadd, v4i32, VR128,
- memopv2i64, i128mem>;
+ memopv2i64, i128mem, SSE_PHADDSUBD>;
defm PHSUBW : SS3I_binop_rm<0x05, "phsubw", X86hsub, v8i16, VR128,
- memopv2i64, i128mem>;
+ memopv2i64, i128mem, SSE_PHADDSUBW>;
defm PHSUBD : SS3I_binop_rm<0x06, "phsubd", X86hsub, v4i32, VR128,
- memopv2i64, i128mem>;
+ memopv2i64, i128mem, SSE_PHADDSUBD>;
defm PSIGNB : SS3I_binop_rm<0x08, "psignb", X86psign, v16i8, VR128,
- memopv2i64, i128mem>;
+ memopv2i64, i128mem, SSE_PSIGN>;
defm PSIGNW : SS3I_binop_rm<0x09, "psignw", X86psign, v8i16, VR128,
- memopv2i64, i128mem>;
+ memopv2i64, i128mem, SSE_PSIGN>;
defm PSIGND : SS3I_binop_rm<0x0A, "psignd", X86psign, v4i32, VR128,
- memopv2i64, i128mem>;
+ memopv2i64, i128mem, SSE_PSIGN>;
defm PSHUFB : SS3I_binop_rm<0x00, "pshufb", X86pshufb, v16i8, VR128,
- memopv2i64, i128mem>;
+ memopv2i64, i128mem, SSE_PSHUFB>;
defm PHADDSW : SS3I_binop_rm_int<0x03, "phaddsw",
- int_x86_ssse3_phadd_sw_128>;
+ int_x86_ssse3_phadd_sw_128,
+ SSE_PHADDSUBSW>;
defm PHSUBSW : SS3I_binop_rm_int<0x07, "phsubsw",
- int_x86_ssse3_phsub_sw_128>;
+ int_x86_ssse3_phsub_sw_128,
+ SSE_PHADDSUBSW>;
defm PMADDUBSW : SS3I_binop_rm_int<0x04, "pmaddubsw",
- int_x86_ssse3_pmadd_ub_sw_128>;
+ int_x86_ssse3_pmadd_ub_sw_128, SSE_PMADD>;
}
defm PMULHRSW : SS3I_binop_rm_int<0x0B, "pmulhrsw",
- int_x86_ssse3_pmul_hr_sw_128>;
+ int_x86_ssse3_pmul_hr_sw_128,
+ SSE_PMULHRSW>;
}
//===---------------------------------------------------------------------===//
!strconcat(asm, "\t{$src3, $src2, $dst|$dst, $src2, $src3}"),
!strconcat(asm,
"\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}")),
- []>, OpSize;
+ [], IIC_SSE_PALIGNR>, OpSize;
let mayLoad = 1 in
def R128rm : SS3AI<0x0F, MRMSrcMem, (outs VR128:$dst),
(ins VR128:$src1, i128mem:$src2, i8imm:$src3),
!strconcat(asm, "\t{$src3, $src2, $dst|$dst, $src2, $src3}"),
!strconcat(asm,
"\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}")),
- []>, OpSize;
+ [], IIC_SSE_PALIGNR>, OpSize;
}
}
defm VPALIGN : ssse3_palign<"vpalignr", 0>, VEX_4V;
let Predicates = [HasAVX2] in
defm VPALIGN : ssse3_palign_y<"vpalignr", 0>, VEX_4V;
-let Constraints = "$src1 = $dst", Predicates = [HasSSSE3] in
+let Constraints = "$src1 = $dst", Predicates = [UseSSSE3] in
defm PALIGN : ssse3_palign<"palignr">;
let Predicates = [HasAVX2] in {
(VPALIGNR128rr VR128:$src2, VR128:$src1, imm:$imm)>;
}
-let Predicates = [HasSSSE3] in {
+let Predicates = [UseSSSE3] in {
def : Pat<(v4i32 (X86PAlign VR128:$src1, VR128:$src2, (i8 imm:$imm))),
(PALIGNR128rr VR128:$src2, VR128:$src1, imm:$imm)>;
def : Pat<(v4f32 (X86PAlign VR128:$src1, VR128:$src2, (i8 imm:$imm))),
def MONITOR : PseudoI<(outs), (ins i32mem:$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:$src2)]>,
- Requires<[HasSSE3]>;
}
let Uses = [EAX, ECX, EDX] in
-def MONITORrrr : I<0x01, MRM_C8, (outs), (ins), "monitor", []>, TB,
- Requires<[HasSSE3]>;
+def MONITORrrr : I<0x01, MRM_C8, (outs), (ins), "monitor", [], IIC_SSE_MONITOR>,
+ TB, Requires<[HasSSE3]>;
let Uses = [ECX, EAX] in
-def MWAITrr : I<0x01, MRM_C9, (outs), (ins), "mwait", []>, TB,
- Requires<[HasSSE3]>;
+def MWAITrr : I<0x01, MRM_C9, (outs), (ins), "mwait",
+ [(int_x86_sse3_mwait ECX, EAX)], IIC_SSE_MWAIT>,
+ TB, Requires<[HasSSE3]>;
def : InstAlias<"mwait %eax, %ecx", (MWAITrr)>, Requires<[In32BitMode]>;
def : InstAlias<"mwait %rax, %rcx", (MWAITrr)>, Requires<[In64BitMode]>;
(VPMOVZXDQrm addr:$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)>;
(PMOVZXDQrm addr:$src)>;
}
+let Predicates = [HasAVX2] in {
+ let AddedComplexity = 15 in {
+ def : Pat<(v4i64 (X86vzmovly (v4i32 VR128:$src))),
+ (VPMOVZXDQYrr VR128:$src)>;
+ def : Pat<(v8i32 (X86vzmovly (v8i16 VR128:$src))),
+ (VPMOVZXWDYrr VR128:$src)>;
+ }
+
+ def : Pat<(v4i64 (X86vsmovl (v4i32 VR128:$src))), (VPMOVSXDQYrr VR128:$src)>;
+ def : Pat<(v8i32 (X86vsmovl (v8i16 VR128:$src))), (VPMOVSXWDYrr VR128:$src)>;
+}
+
let Predicates = [HasAVX] in {
-def : Pat<(v2i64 (X86vsmovl (v4i32 VR128:$src))), (VPMOVSXDQrr VR128:$src)>;
-def : Pat<(v4i32 (X86vsmovl (v8i16 VR128:$src))), (VPMOVSXWDrr VR128:$src)>;
+ def : Pat<(v2i64 (X86vsmovl (v4i32 VR128:$src))), (VPMOVSXDQrr VR128:$src)>;
+ def : Pat<(v4i32 (X86vsmovl (v8i16 VR128:$src))), (VPMOVSXWDrr VR128:$src)>;
}
-let Predicates = [HasSSE41] in {
-def : Pat<(v2i64 (X86vsmovl (v4i32 VR128:$src))), (PMOVSXDQrr VR128:$src)>;
-def : Pat<(v4i32 (X86vsmovl (v8i16 VR128:$src))), (PMOVSXWDrr VR128:$src)>;
+let Predicates = [UseSSE41] in {
+ def : Pat<(v2i64 (X86vsmovl (v4i32 VR128:$src))), (PMOVSXDQrr VR128:$src)>;
+ def : Pat<(v4i32 (X86vsmovl (v8i16 VR128:$src))), (PMOVSXWDrr VR128:$src)>;
}
(VPMOVZXWQrm addr:$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)>;
(VPMOVZXBQrm addr:$src)>;
}
-let Predicates = [HasSSE41] in {
+let Predicates = [UseSSE41] in {
// Common patterns involving scalar load
def : Pat<(int_x86_sse41_pmovsxbq
(bitconvert (v4i32 (X86vzmovl
imm:$src2))),
addr:$dst),
(EXTRACTPSmr addr:$dst, VR128:$src1, imm:$src2)>,
- Requires<[HasSSE41]>;
+ Requires<[UseSSE41]>;
//===----------------------------------------------------------------------===//
// SSE4.1 - Insert Instructions
let Defs = [EFLAGS], Predicates = [HasAVX] in {
def VPTESTrr : SS48I<0x17, MRMSrcReg, (outs), (ins VR128:$src1, VR128:$src2),
"vptest\t{$src2, $src1|$src1, $src2}",
- [(set EFLAGS, (X86ptest VR128:$src1, (v4f32 VR128:$src2)))]>,
+ [(set EFLAGS, (X86ptest VR128:$src1, (v2i64 VR128:$src2)))]>,
OpSize, VEX;
def VPTESTrm : SS48I<0x17, MRMSrcMem, (outs), (ins VR128:$src1, f128mem:$src2),
"vptest\t{$src2, $src1|$src1, $src2}",
- [(set EFLAGS,(X86ptest VR128:$src1, (memopv4f32 addr:$src2)))]>,
+ [(set EFLAGS,(X86ptest VR128:$src1, (memopv2i64 addr:$src2)))]>,
OpSize, VEX;
def VPTESTYrr : SS48I<0x17, MRMSrcReg, (outs), (ins VR256:$src1, VR256:$src2),
let Defs = [EFLAGS] in {
def PTESTrr : SS48I<0x17, MRMSrcReg, (outs), (ins VR128:$src1, VR128:$src2),
"ptest\t{$src2, $src1|$src1, $src2}",
- [(set EFLAGS, (X86ptest VR128:$src1, (v4f32 VR128:$src2)))]>,
+ [(set EFLAGS, (X86ptest VR128:$src1, (v2i64 VR128:$src2)))]>,
OpSize;
def PTESTrm : SS48I<0x17, MRMSrcMem, (outs), (ins VR128:$src1, f128mem:$src2),
"ptest\t{$src2, $src1|$src1, $src2}",
- [(set EFLAGS, (X86ptest VR128:$src1, (memopv4f32 addr:$src2)))]>,
+ [(set EFLAGS, (X86ptest VR128:$src1, (memopv2i64 addr:$src2)))]>,
OpSize;
}
(bitconvert (memopv2i64 addr:$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
let isCommutable = 0 in {
let ExeDomain = SSEPackedSingle in {
defm VBLENDPS : SS41I_binop_rmi_int<0x0C, "vblendps", int_x86_sse41_blendps,
- VR128, memopv4f32, i128mem, 0>, VEX_4V;
+ VR128, memopv4f32, f128mem, 0>, VEX_4V;
defm VBLENDPSY : SS41I_binop_rmi_int<0x0C, "vblendps",
- int_x86_avx_blend_ps_256, VR256, memopv8f32, i256mem, 0>, VEX_4V;
+ int_x86_avx_blend_ps_256, VR256, memopv8f32, f256mem, 0>, VEX_4V;
}
let ExeDomain = SSEPackedDouble in {
defm VBLENDPD : SS41I_binop_rmi_int<0x0D, "vblendpd", int_x86_sse41_blendpd,
- VR128, memopv2f64, i128mem, 0>, VEX_4V;
+ VR128, memopv2f64, f128mem, 0>, VEX_4V;
defm VBLENDPDY : SS41I_binop_rmi_int<0x0D, "vblendpd",
- int_x86_avx_blend_pd_256, VR256, memopv4f64, i256mem, 0>, VEX_4V;
+ int_x86_avx_blend_pd_256, VR256, memopv4f64, f256mem, 0>, VEX_4V;
}
defm VPBLENDW : SS41I_binop_rmi_int<0x0E, "vpblendw", int_x86_sse41_pblendw,
VR128, memopv2i64, i128mem, 0>, VEX_4V;
}
let ExeDomain = SSEPackedSingle in
defm VDPPS : SS41I_binop_rmi_int<0x40, "vdpps", int_x86_sse41_dpps,
- VR128, memopv4f32, i128mem, 0>, VEX_4V;
+ VR128, memopv4f32, f128mem, 0>, VEX_4V;
let ExeDomain = SSEPackedDouble in
defm VDPPD : SS41I_binop_rmi_int<0x41, "vdppd", int_x86_sse41_dppd,
- VR128, memopv2f64, i128mem, 0>, VEX_4V;
+ VR128, memopv2f64, f128mem, 0>, VEX_4V;
let ExeDomain = SSEPackedSingle in
defm VDPPSY : SS41I_binop_rmi_int<0x40, "vdpps", int_x86_avx_dp_ps_256,
VR256, memopv8f32, i256mem, 0>, VEX_4V;
let isCommutable = 0 in {
let ExeDomain = SSEPackedSingle in
defm BLENDPS : SS41I_binop_rmi_int<0x0C, "blendps", int_x86_sse41_blendps,
- VR128, memopv4f32, i128mem>;
+ VR128, memopv4f32, f128mem>;
let ExeDomain = SSEPackedDouble in
defm BLENDPD : SS41I_binop_rmi_int<0x0D, "blendpd", int_x86_sse41_blendpd,
- VR128, memopv2f64, i128mem>;
+ VR128, memopv2f64, f128mem>;
defm PBLENDW : SS41I_binop_rmi_int<0x0E, "pblendw", int_x86_sse41_pblendw,
VR128, memopv2i64, i128mem>;
defm MPSADBW : SS41I_binop_rmi_int<0x42, "mpsadbw", int_x86_sse41_mpsadbw,
}
let ExeDomain = SSEPackedSingle in
defm DPPS : SS41I_binop_rmi_int<0x40, "dpps", int_x86_sse41_dpps,
- VR128, memopv4f32, i128mem>;
+ VR128, memopv4f32, f128mem>;
let ExeDomain = SSEPackedDouble in
defm DPPD : SS41I_binop_rmi_int<0x41, "dppd", int_x86_sse41_dppd,
- VR128, memopv2f64, i128mem>;
+ VR128, memopv2f64, f128mem>;
}
/// SS41I_quaternary_int_avx - AVX SSE 4.1 with 4 operators
let Predicates = [HasAVX] in {
let ExeDomain = SSEPackedDouble in {
-defm VBLENDVPD : SS41I_quaternary_int_avx<0x4B, "vblendvpd", VR128, i128mem,
+defm VBLENDVPD : SS41I_quaternary_int_avx<0x4B, "vblendvpd", VR128, f128mem,
memopv2f64, int_x86_sse41_blendvpd>;
-defm VBLENDVPDY : SS41I_quaternary_int_avx<0x4B, "vblendvpd", VR256, i256mem,
+defm VBLENDVPDY : SS41I_quaternary_int_avx<0x4B, "vblendvpd", VR256, f256mem,
memopv4f64, int_x86_avx_blendv_pd_256>;
} // ExeDomain = SSEPackedDouble
let ExeDomain = SSEPackedSingle in {
-defm VBLENDVPS : SS41I_quaternary_int_avx<0x4A, "vblendvps", VR128, i128mem,
+defm VBLENDVPS : SS41I_quaternary_int_avx<0x4A, "vblendvps", VR128, f128mem,
memopv4f32, int_x86_sse41_blendvps>;
-defm VBLENDVPSY : SS41I_quaternary_int_avx<0x4A, "vblendvps", VR256, i256mem,
+defm VBLENDVPSY : SS41I_quaternary_int_avx<0x4A, "vblendvps", VR256, f256mem,
memopv8f32, int_x86_avx_blendv_ps_256>;
} // ExeDomain = SSEPackedSingle
defm VPBLENDVB : SS41I_quaternary_int_avx<0x4C, "vpblendvb", VR128, i128mem,
def : Pat<(v4f64 (vselect (v4i64 VR256:$mask), (v4f64 VR256:$src1),
(v4f64 VR256:$src2))),
(VBLENDVPDYrr VR256:$src2, VR256:$src1, VR256:$mask)>;
+
+ def : Pat<(v8f32 (X86Blendps (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),
+ (imm:$mask))),
+ (VBLENDPDYrri VR256:$src2, VR256:$src1, imm:$mask)>;
+
+ def : Pat<(v8i16 (X86Blendpw (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),
+ (imm:$mask))),
+ (VBLENDPSrri VR128:$src2, VR128:$src1, imm:$mask)>;
+ def : Pat<(v2f64 (X86Blendpd (v2f64 VR128:$src1), (v2f64 VR128:$src2),
+ (imm:$mask))),
+ (VBLENDPDrri VR128:$src2, VR128:$src1, imm:$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),
+ (imm:$mask))),
+ (VPBLENDWYrri VR256:$src2, VR256:$src1, imm:$mask)>;
}
/// SS41I_ternary_int - SSE 4.1 ternary operator
let Uses = [XMM0], Constraints = "$src1 = $dst" in {
multiclass SS41I_ternary_int<bits<8> opc, string OpcodeStr, PatFrag mem_frag,
- Intrinsic IntId> {
+ X86MemOperand x86memop, Intrinsic IntId> {
def rr0 : SS48I<opc, MRMSrcReg, (outs VR128:$dst),
(ins VR128:$src1, VR128:$src2),
!strconcat(OpcodeStr,
OpSize;
def rm0 : SS48I<opc, MRMSrcMem, (outs VR128:$dst),
- (ins VR128:$src1, i128mem:$src2),
+ (ins VR128:$src1, x86memop:$src2),
!strconcat(OpcodeStr,
"\t{$src2, $dst|$dst, $src2}"),
[(set VR128:$dst,
}
let ExeDomain = SSEPackedDouble in
-defm BLENDVPD : SS41I_ternary_int<0x15, "blendvpd", memopv2f64,
+defm BLENDVPD : SS41I_ternary_int<0x15, "blendvpd", memopv2f64, f128mem,
int_x86_sse41_blendvpd>;
let ExeDomain = SSEPackedSingle in
-defm BLENDVPS : SS41I_ternary_int<0x14, "blendvps", memopv4f32,
+defm BLENDVPS : SS41I_ternary_int<0x14, "blendvps", memopv4f32, f128mem,
int_x86_sse41_blendvps>;
-defm PBLENDVB : SS41I_ternary_int<0x10, "pblendvb", memopv2i64,
+defm PBLENDVB : SS41I_ternary_int<0x10, "pblendvb", memopv2i64, i128mem,
int_x86_sse41_pblendvb>;
-let Predicates = [HasSSE41] in {
+// Aliases with the implicit xmm0 argument
+def : InstAlias<"blendvpd\t{%xmm0, $src2, $dst|$dst, $src2, %xmm0}",
+ (BLENDVPDrr0 VR128:$dst, VR128:$src2)>;
+def : InstAlias<"blendvpd\t{%xmm0, $src2, $dst|$dst, $src2, %xmm0}",
+ (BLENDVPDrm0 VR128:$dst, f128mem:$src2)>;
+def : InstAlias<"blendvps\t{%xmm0, $src2, $dst|$dst, $src2, %xmm0}",
+ (BLENDVPSrr0 VR128:$dst, VR128:$src2)>;
+def : InstAlias<"blendvps\t{%xmm0, $src2, $dst|$dst, $src2, %xmm0}",
+ (BLENDVPSrm0 VR128:$dst, f128mem:$src2)>;
+def : InstAlias<"pblendvb\t{%xmm0, $src2, $dst|$dst, $src2, %xmm0}",
+ (PBLENDVBrr0 VR128:$dst, VR128:$src2)>;
+def : InstAlias<"pblendvb\t{%xmm0, $src2, $dst|$dst, $src2, %xmm0}",
+ (PBLENDVBrm0 VR128:$dst, i128mem:$src2)>;
+
+let Predicates = [UseSSE41] in {
def : Pat<(v16i8 (vselect (v16i8 XMM0), (v16i8 VR128:$src1),
(v16i8 VR128:$src2))),
(PBLENDVBrr0 VR128:$src2, VR128:$src1)>;
def : Pat<(v2f64 (vselect (v2i64 XMM0), (v2f64 VR128:$src1),
(v2f64 VR128:$src2))),
(BLENDVPDrr0 VR128:$src2, VR128:$src1)>;
+
+ def : Pat<(v8i16 (X86Blendpw (v8i16 VR128:$src1), (v8i16 VR128:$src2),
+ (imm:$mask))),
+ (PBLENDWrri VR128:$src2, VR128:$src1, imm:$mask)>;
+ def : Pat<(v4f32 (X86Blendps (v4f32 VR128:$src1), (v4f32 VR128:$src2),
+ (imm:$mask))),
+ (BLENDPSrri VR128:$src2, VR128:$src1, imm:$mask)>;
+ def : Pat<(v2f64 (X86Blendpd (v2f64 VR128:$src1), (v2f64 VR128:$src2),
+ (imm:$mask))),
+ (BLENDPDrri VR128:$src2, VR128:$src1, imm:$mask)>;
+
}
let Predicates = [HasAVX] in
}
let Defs = [EFLAGS], usesCustomInserter = 1 in {
- let AddedComplexity = 1 in
- defm VPCMPISTRM128 : pseudo_pcmpistrm<"#VPCMPISTRM128">, Requires<[HasAVX]>;
- defm PCMPISTRM128 : pseudo_pcmpistrm<"#PCMPISTRM128">, Requires<[HasSSE42]>;
+ defm VPCMPISTRM128 : pseudo_pcmpistrm<"#VPCMPISTRM128">, Requires<[HasAVX]>;
+ defm PCMPISTRM128 : pseudo_pcmpistrm<"#PCMPISTRM128">, Requires<[UseSSE42]>;
}
let Defs = [XMM0, EFLAGS], neverHasSideEffects = 1, Predicates = [HasAVX] in {
}
let Defs = [EFLAGS], Uses = [EAX, EDX], usesCustomInserter = 1 in {
- let AddedComplexity = 1 in
- defm VPCMPESTRM128 : pseudo_pcmpestrm<"#VPCMPESTRM128">, Requires<[HasAVX]>;
- defm PCMPESTRM128 : pseudo_pcmpestrm<"#PCMPESTRM128">, Requires<[HasSSE42]>;
+ defm VPCMPESTRM128 : pseudo_pcmpestrm<"#VPCMPESTRM128">, Requires<[HasAVX]>;
+ defm PCMPESTRM128 : pseudo_pcmpestrm<"#PCMPESTRM128">, Requires<[UseSSE42]>;
}
let Predicates = [HasAVX],
}
// Packed Compare Implicit Length Strings, Return Index
-let Defs = [ECX, EFLAGS] in {
- multiclass SS42AI_pcmpistri<Intrinsic IntId128, string asm = "pcmpistri"> {
+let Defs = [ECX, EFLAGS], neverHasSideEffects = 1 in {
+ multiclass SS42AI_pcmpistri<string asm> {
def rr : SS42AI<0x63, MRMSrcReg, (outs),
(ins VR128:$src1, VR128:$src2, i8imm:$src3),
!strconcat(asm, "\t{$src3, $src2, $src1|$src1, $src2, $src3}"),
- [(set ECX, (IntId128 VR128:$src1, VR128:$src2, imm:$src3)),
- (implicit EFLAGS)]>, OpSize;
+ []>, 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, $src3}"),
- [(set ECX, (IntId128 VR128:$src1, (load addr:$src2), imm:$src3)),
- (implicit EFLAGS)]>, OpSize;
+ []>, OpSize;
}
}
-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 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"> {
+let Defs = [ECX, EFLAGS], Uses = [EAX, EDX], neverHasSideEffects = 1 in {
+ 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, $src5}"),
- [(set ECX, (IntId128 VR128:$src1, EAX, VR128:$src3, EDX, imm:$src5)),
- (implicit EFLAGS)]>, OpSize;
+ []>, 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, $src5}"),
- [(set ECX,
- (IntId128 VR128:$src1, EAX, (load addr:$src3), EDX, imm:$src5)),
- (implicit EFLAGS)]>, OpSize;
+ []>, OpSize;
}
}
-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 Predicates = [HasAVX] in
+defm VPCMPESTRI : SS42AI_pcmpestri<"vpcmpestri">, VEX;
+defm PCMPESTRI : SS42AI_pcmpestri<"pcmpestri">;
//===----------------------------------------------------------------------===//
// SSE4.2 - CRC Instructions
OpSize;
//===----------------------------------------------------------------------===//
-// CLMUL Instructions
+// PCLMUL Instructions
//===----------------------------------------------------------------------===//
-// Carry-less Multiplication instructions
-let neverHasSideEffects = 1 in {
// AVX carry-less Multiplication instructions
-def VPCLMULQDQrr : AVXCLMULIi8<0x44, MRMSrcReg, (outs VR128:$dst),
+def VPCLMULQDQrr : AVXPCLMULIi8<0x44, MRMSrcReg, (outs VR128:$dst),
(ins VR128:$src1, VR128:$src2, i8imm:$src3),
"vpclmulqdq\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}",
- []>;
+ [(set VR128:$dst,
+ (int_x86_pclmulqdq VR128:$src1, VR128:$src2, imm:$src3))]>;
-let mayLoad = 1 in
-def VPCLMULQDQrm : AVXCLMULIi8<0x44, MRMSrcMem, (outs VR128:$dst),
+def VPCLMULQDQrm : AVXPCLMULIi8<0x44, MRMSrcMem, (outs VR128:$dst),
(ins VR128:$src1, i128mem:$src2, i8imm:$src3),
"vpclmulqdq\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}",
- []>;
+ [(set VR128:$dst, (int_x86_pclmulqdq VR128:$src1,
+ (memopv2i64 addr:$src2), imm:$src3))]>;
+// Carry-less Multiplication instructions
let Constraints = "$src1 = $dst" in {
-def PCLMULQDQrr : CLMULIi8<0x44, MRMSrcReg, (outs VR128:$dst),
+def PCLMULQDQrr : PCLMULIi8<0x44, MRMSrcReg, (outs VR128:$dst),
(ins VR128:$src1, VR128:$src2, i8imm:$src3),
"pclmulqdq\t{$src3, $src2, $dst|$dst, $src2, $src3}",
- []>;
+ [(set VR128:$dst,
+ (int_x86_pclmulqdq VR128:$src1, VR128:$src2, imm:$src3))]>;
-let mayLoad = 1 in
-def PCLMULQDQrm : CLMULIi8<0x44, MRMSrcMem, (outs VR128:$dst),
+def PCLMULQDQrm : PCLMULIi8<0x44, MRMSrcMem, (outs VR128:$dst),
(ins VR128:$src1, i128mem:$src2, i8imm:$src3),
"pclmulqdq\t{$src3, $src2, $dst|$dst, $src2, $src3}",
- []>;
+ [(set VR128:$dst, (int_x86_pclmulqdq VR128:$src1,
+ (memopv2i64 addr:$src2), imm:$src3))]>;
} // Constraints = "$src1 = $dst"
-} // neverHasSideEffects = 1
multiclass pclmul_alias<string asm, int immop> {
- def : InstAlias<!strconcat("pclmul", asm,
- "dq {$src, $dst|$dst, $src}"),
+ def : InstAlias<!strconcat("pclmul", asm, "dq {$src, $dst|$dst, $src}"),
(PCLMULQDQrr VR128:$dst, VR128:$src, immop)>;
- def : InstAlias<!strconcat("pclmul", asm,
- "dq {$src, $dst|$dst, $src}"),
+ def : InstAlias<!strconcat("pclmul", asm, "dq {$src, $dst|$dst, $src}"),
(PCLMULQDQrm VR128:$dst, i128mem:$src, immop)>;
- def : InstAlias<!strconcat("vpclmul", asm,
+ def : InstAlias<!strconcat("vpclmul", asm,
"dq {$src2, $src1, $dst|$dst, $src1, $src2}"),
(VPCLMULQDQrr VR128:$dst, VR128:$src1, VR128:$src2, immop)>;
- def : InstAlias<!strconcat("vpclmul", asm,
+ def : InstAlias<!strconcat("vpclmul", asm,
"dq {$src2, $src1, $dst|$dst, $src1, $src2}"),
(VPCLMULQDQrm VR128:$dst, VR128:$src1, i128mem:$src2, immop)>;
}
defm : pclmul_alias<"lqhq", 0x10>;
defm : pclmul_alias<"lqlq", 0x00>;
+//===----------------------------------------------------------------------===//
+// SSE4A Instructions
+//===----------------------------------------------------------------------===//
+
+let Predicates = [HasSSE4A] in {
+
+let Constraints = "$src = $dst" in {
+def EXTRQI : Ii8<0x78, MRM0r, (outs VR128:$dst),
+ (ins VR128:$src, i8imm:$len, i8imm:$idx),
+ "extrq\t{$idx, $len, $src|$src, $len, $idx}",
+ [(set VR128:$dst, (int_x86_sse4a_extrqi VR128:$src, imm:$len,
+ imm:$idx))]>, TB, OpSize;
+def EXTRQ : I<0x79, MRMSrcReg, (outs VR128:$dst),
+ (ins VR128:$src, VR128:$mask),
+ "extrq\t{$mask, $src|$src, $mask}",
+ [(set VR128:$dst, (int_x86_sse4a_extrq VR128:$src,
+ VR128:$mask))]>, TB, OpSize;
+
+def INSERTQI : Ii8<0x78, MRMSrcReg, (outs VR128:$dst),
+ (ins VR128:$src, VR128:$src2, i8imm:$len, i8imm:$idx),
+ "insertq\t{$idx, $len, $src2, $src|$src, $src2, $len, $idx}",
+ [(set VR128:$dst, (int_x86_sse4a_insertqi VR128:$src,
+ VR128:$src2, imm:$len, imm:$idx))]>, XD;
+def INSERTQ : I<0x79, MRMSrcReg, (outs VR128:$dst),
+ (ins VR128:$src, VR128:$mask),
+ "insertq\t{$mask, $src|$src, $mask}",
+ [(set VR128:$dst, (int_x86_sse4a_insertq VR128:$src,
+ VR128:$mask))]>, XD;
+}
+
+def MOVNTSS : I<0x2B, MRMDestMem, (outs), (ins f32mem:$dst, VR128:$src),
+ "movntss\t{$src, $dst|$dst, $src}",
+ [(int_x86_sse4a_movnt_ss addr:$dst, VR128:$src)]>, XS;
+
+def MOVNTSD : I<0x2B, MRMDestMem, (outs), (ins f64mem:$dst, VR128:$src),
+ "movntsd\t{$src, $dst|$dst, $src}",
+ [(int_x86_sse4a_movnt_sd addr:$dst, VR128:$src)]>, XD;
+}
+
//===----------------------------------------------------------------------===//
// AVX Instructions
//===----------------------------------------------------------------------===//
int_x86_avx_vbroadcast_ss_256>;
}
let ExeDomain = SSEPackedDouble in
-def VBROADCASTSDrm : avx_broadcast<0x19, "vbroadcastsd", VR256, f64mem,
+def VBROADCASTSDYrm : avx_broadcast<0x19, "vbroadcastsd", VR256, f64mem,
int_x86_avx_vbroadcast_sd_256>;
def VBROADCASTF128 : avx_broadcast<0x1A, "vbroadcastf128", VR256, f128mem,
int_x86_avx_vbroadcastf128_pd_256>;
int_x86_avx2_vbroadcast_ss_ps_256>;
}
let ExeDomain = SSEPackedDouble in
-def VBROADCASTSDrr : avx2_broadcast_reg<0x19, "vbroadcastsd", VR256,
- int_x86_avx2_vbroadcast_sd_pd_256>;
+def VBROADCASTSDYrr : avx2_broadcast_reg<0x19, "vbroadcastsd", VR256,
+ int_x86_avx2_vbroadcast_sd_pd_256>;
let Predicates = [HasAVX2] in
def VBROADCASTI128 : avx_broadcast<0x5A, "vbroadcasti128", VR256, i128mem,
[]>, VEX_4V;
}
+let Predicates = [HasAVX] in {
+def : Pat<(vinsertf128_insert:$ins (v8f32 VR256:$src1), (v4f32 VR128:$src2),
+ (i32 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)),
+ (VINSERTF128rr VR256:$src1, VR128:$src2,
+ (INSERT_get_vinsertf128_imm VR256:$ins))>;
+
+def : Pat<(vinsertf128_insert:$ins (v8f32 VR256:$src1), (memopv4f32 addr:$src2),
+ (i32 imm)),
+ (VINSERTF128rm VR256:$src1, addr:$src2,
+ (INSERT_get_vinsertf128_imm VR256:$ins))>;
+def : Pat<(vinsertf128_insert:$ins (v4f64 VR256:$src1), (memopv2f64 addr:$src2),
+ (i32 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)),
+ (VINSERTF128rr VR256:$src1, VR128:$src2,
+ (INSERT_get_vinsertf128_imm VR256:$ins))>;
+def : Pat<(vinsertf128_insert:$ins (v8i32 VR256:$src1), (v4i32 VR128:$src2),
+ (i32 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)),
+ (VINSERTF128rr VR256:$src1, VR128:$src2,
+ (INSERT_get_vinsertf128_imm VR256:$ins))>;
+def : Pat<(vinsertf128_insert:$ins (v16i16 VR256:$src1), (v8i16 VR128:$src2),
+ (i32 imm)),
+ (VINSERTF128rr VR256:$src1, VR128:$src2,
+ (INSERT_get_vinsertf128_imm VR256:$ins))>;
+
+def : Pat<(vinsertf128_insert:$ins (v4i64 VR256:$src1), (memopv2i64 addr:$src2),
+ (i32 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)),
+ (i32 imm)),
+ (VINSERTF128rm VR256:$src1, addr:$src2,
+ (INSERT_get_vinsertf128_imm VR256:$ins))>;
+def : Pat<(vinsertf128_insert:$ins (v32i8 VR256:$src1),
+ (bc_v16i8 (memopv2i64 addr:$src2)),
+ (i32 imm)),
+ (VINSERTF128rm VR256:$src1, addr:$src2,
+ (INSERT_get_vinsertf128_imm VR256:$ins))>;
+def : Pat<(vinsertf128_insert:$ins (v16i16 VR256:$src1),
+ (bc_v8i16 (memopv2i64 addr:$src2)),
+ (i32 imm)),
+ (VINSERTF128rm VR256:$src1, addr:$src2,
+ (INSERT_get_vinsertf128_imm VR256:$ins))>;
+}
+
//===----------------------------------------------------------------------===//
// VEXTRACTF128 - Extract packed floating-point values
//
[]>, VEX;
}
+// AVX1 patterns
let Predicates = [HasAVX] in {
-def : Pat<(int_x86_avx_vextractf128_pd_256 VR256:$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:$src1, (i32 imm)),
+ (v4f32 (VEXTRACTF128rr
+ (v8f32 VR256:$src1),
+ (EXTRACT_get_vextractf128_imm VR128:$ext)))>;
+def : Pat<(vextractf128_extract:$ext VR256:$src1, (i32 imm)),
+ (v2f64 (VEXTRACTF128rr
+ (v4f64 VR256:$src1),
+ (EXTRACT_get_vextractf128_imm VR128:$ext)))>;
+
+def : Pat<(alignedstore (v4f32 (vextractf128_extract:$ext (v8f32 VR256:$src1),
+ (i32 imm))), addr:$dst),
+ (VEXTRACTF128mr addr:$dst, VR256:$src1,
+ (EXTRACT_get_vextractf128_imm VR128:$ext))>;
+def : Pat<(alignedstore (v2f64 (vextractf128_extract:$ext (v4f64 VR256:$src1),
+ (i32 imm))), addr:$dst),
+ (VEXTRACTF128mr addr:$dst, VR256:$src1,
+ (EXTRACT_get_vextractf128_imm VR128:$ext))>;
+}
+
+let Predicates = [HasAVX1Only] in {
+def : Pat<(vextractf128_extract:$ext VR256:$src1, (i32 imm)),
+ (v2i64 (VEXTRACTF128rr
+ (v4i64 VR256:$src1),
+ (EXTRACT_get_vextractf128_imm VR128:$ext)))>;
+def : Pat<(vextractf128_extract:$ext VR256:$src1, (i32 imm)),
+ (v4i32 (VEXTRACTF128rr
+ (v8i32 VR256:$src1),
+ (EXTRACT_get_vextractf128_imm VR128:$ext)))>;
+def : Pat<(vextractf128_extract:$ext VR256:$src1, (i32 imm)),
+ (v8i16 (VEXTRACTF128rr
+ (v16i16 VR256:$src1),
+ (EXTRACT_get_vextractf128_imm VR128:$ext)))>;
+def : Pat<(vextractf128_extract:$ext VR256:$src1, (i32 imm)),
+ (v16i8 (VEXTRACTF128rr
+ (v32i8 VR256:$src1),
+ (EXTRACT_get_vextractf128_imm VR128:$ext)))>;
+
+def : Pat<(alignedstore (v2i64 (vextractf128_extract:$ext (v4i64 VR256:$src1),
+ (i32 imm))), addr:$dst),
+ (VEXTRACTF128mr addr:$dst, VR256:$src1,
+ (EXTRACT_get_vextractf128_imm VR128:$ext))>;
+def : Pat<(alignedstore (v4i32 (vextractf128_extract:$ext (v8i32 VR256:$src1),
+ (i32 imm))), addr:$dst),
+ (VEXTRACTF128mr addr:$dst, VR256:$src1,
+ (EXTRACT_get_vextractf128_imm VR128:$ext))>;
+def : Pat<(alignedstore (v8i16 (vextractf128_extract:$ext (v16i16 VR256:$src1),
+ (i32 imm))), addr:$dst),
+ (VEXTRACTF128mr addr:$dst, VR256:$src1,
+ (EXTRACT_get_vextractf128_imm VR128:$ext))>;
+def : Pat<(alignedstore (v16i8 (vextractf128_extract:$ext (v32i8 VR256:$src1),
+ (i32 imm))), addr:$dst),
+ (VEXTRACTF128mr addr:$dst, VR256:$src1,
+ (EXTRACT_get_vextractf128_imm VR128:$ext))>;
}
//===----------------------------------------------------------------------===//
}
let Predicates = [HasAVX] in {
+def : Pat<(v4f64 (X86VPerm2x128 VR256:$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:$imm)>;
// Half precision conversion instructions
//===----------------------------------------------------------------------===//
multiclass f16c_ph2ps<RegisterClass RC, X86MemOperand x86memop, Intrinsic Int> {
-let Predicates = [HasAVX, HasF16C] in {
def rr : I<0x13, MRMSrcReg, (outs RC:$dst), (ins VR128:$src),
"vcvtph2ps\t{$src, $dst|$dst, $src}",
[(set RC:$dst, (Int VR128:$src))]>,
def rm : I<0x13, MRMSrcMem, (outs RC:$dst), (ins x86memop:$src),
"vcvtph2ps\t{$src, $dst|$dst, $src}", []>, T8, OpSize, VEX;
}
-}
multiclass f16c_ps2ph<RegisterClass RC, X86MemOperand x86memop, Intrinsic Int> {
-let Predicates = [HasAVX, HasF16C] in {
def rr : Ii8<0x1D, MRMDestReg, (outs VR128:$dst),
(ins RC:$src1, i32i8imm:$src2),
"vcvtps2ph\t{$src2, $src1, $dst|$dst, $src1, $src2}",
[(set VR128:$dst, (Int RC:$src1, imm:$src2))]>,
TA, OpSize, VEX;
- let neverHasSideEffects = 1, mayLoad = 1 in
- def mr : Ii8<0x1D, MRMDestMem, (outs x86memop:$dst),
- (ins RC:$src1, i32i8imm:$src2),
+ let neverHasSideEffects = 1, mayStore = 1 in
+ def mr : Ii8<0x1D, MRMDestMem, (outs),
+ (ins x86memop:$dst, RC:$src1, i32i8imm:$src2),
"vcvtps2ph\t{$src2, $src1, $dst|$dst, $src1, $src2}", []>,
TA, OpSize, VEX;
}
-}
-defm VCVTPH2PS : f16c_ph2ps<VR128, f64mem, int_x86_vcvtph2ps_128>;
-defm VCVTPH2PSY : f16c_ph2ps<VR256, f128mem, int_x86_vcvtph2ps_256>;
-defm VCVTPS2PH : f16c_ps2ph<VR128, f64mem, int_x86_vcvtps2ph_128>;
-defm VCVTPS2PHY : f16c_ps2ph<VR256, f128mem, int_x86_vcvtps2ph_256>;
+let Predicates = [HasAVX, HasF16C] in {
+ defm VCVTPH2PS : f16c_ph2ps<VR128, f64mem, int_x86_vcvtph2ps_128>;
+ defm VCVTPH2PSY : f16c_ph2ps<VR256, f128mem, int_x86_vcvtph2ps_256>;
+ defm VCVTPS2PH : f16c_ps2ph<VR128, f64mem, int_x86_vcvtps2ph_128>;
+ defm VCVTPS2PHY : f16c_ps2ph<VR256, f128mem, int_x86_vcvtps2ph_256>;
+}
//===----------------------------------------------------------------------===//
// AVX2 Instructions
(VPBROADCASTQrm addr:$src)>;
def : Pat<(v4i64 (X86VBroadcast (loadi64 addr:$src))),
(VPBROADCASTQYrm addr:$src)>;
+
+ def : Pat<(v16i8 (X86VBroadcast (v16i8 VR128:$src))),
+ (VPBROADCASTBrr VR128:$src)>;
+ def : Pat<(v32i8 (X86VBroadcast (v16i8 VR128:$src))),
+ (VPBROADCASTBYrr VR128:$src)>;
+ def : Pat<(v8i16 (X86VBroadcast (v8i16 VR128:$src))),
+ (VPBROADCASTWrr VR128:$src)>;
+ def : Pat<(v16i16 (X86VBroadcast (v8i16 VR128:$src))),
+ (VPBROADCASTWYrr VR128:$src)>;
+ def : Pat<(v4i32 (X86VBroadcast (v4i32 VR128:$src))),
+ (VPBROADCASTDrr VR128:$src)>;
+ def : Pat<(v8i32 (X86VBroadcast (v4i32 VR128:$src))),
+ (VPBROADCASTDYrr VR128:$src)>;
+ def : Pat<(v2i64 (X86VBroadcast (v2i64 VR128:$src))),
+ (VPBROADCASTQrr VR128:$src)>;
+ def : Pat<(v4i64 (X86VBroadcast (v2i64 VR128:$src))),
+ (VPBROADCASTQYrr VR128:$src)>;
+ def : Pat<(v4f32 (X86VBroadcast (v4f32 VR128:$src))),
+ (VBROADCASTSSrr VR128:$src)>;
+ def : Pat<(v8f32 (X86VBroadcast (v4f32 VR128:$src))),
+ (VBROADCASTSSYrr VR128:$src)>;
+ def : Pat<(v2f64 (X86VBroadcast (v2f64 VR128:$src))),
+ (VPBROADCASTQrr VR128:$src)>;
+ def : Pat<(v4f64 (X86VBroadcast (v2f64 VR128:$src))),
+ (VBROADCASTSDYrr VR128:$src)>;
+
+ // Provide fallback in case the load node that is used in the patterns above
+ // is used by additional users, which prevents the pattern selection.
+ let AddedComplexity = 20 in {
+ def : Pat<(v4f32 (X86VBroadcast FR32:$src)),
+ (VBROADCASTSSrr (COPY_TO_REGCLASS FR32:$src, VR128))>;
+ def : Pat<(v8f32 (X86VBroadcast FR32:$src)),
+ (VBROADCASTSSYrr (COPY_TO_REGCLASS FR32:$src, VR128))>;
+ def : Pat<(v4f64 (X86VBroadcast FR64:$src)),
+ (VBROADCASTSDYrr (COPY_TO_REGCLASS FR64:$src, VR128))>;
+
+ def : Pat<(v4i32 (X86VBroadcast GR32:$src)),
+ (VBROADCASTSSrr (COPY_TO_REGCLASS GR32:$src, VR128))>;
+ def : Pat<(v8i32 (X86VBroadcast GR32:$src)),
+ (VBROADCASTSSYrr (COPY_TO_REGCLASS GR32:$src, VR128))>;
+ def : Pat<(v4i64 (X86VBroadcast GR64:$src)),
+ (VBROADCASTSDYrr (COPY_TO_REGCLASS GR64:$src, VR128))>;
+ }
}
// 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:$src))),
- (VBROADCASTSDrm addr:$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))),
- (VBROADCASTSDrm 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.
+ let AddedComplexity = 20 in {
+ // 128bit broadcasts:
+ def : Pat<(v4f32 (X86VBroadcast FR32:$src)),
+ (VPSHUFDri (COPY_TO_REGCLASS FR32:$src, VR128), 0)>;
+ def : Pat<(v8f32 (X86VBroadcast FR32:$src)),
+ (VINSERTF128rr (INSERT_SUBREG (v8f32 (IMPLICIT_DEF)),
+ (VPSHUFDri (COPY_TO_REGCLASS FR32:$src, VR128), 0), sub_xmm),
+ (VPSHUFDri (COPY_TO_REGCLASS FR32:$src, VR128), 0), 1)>;
+ def : Pat<(v4f64 (X86VBroadcast FR64:$src)),
+ (VINSERTF128rr (INSERT_SUBREG (v4f64 (IMPLICIT_DEF)),
+ (VPSHUFDri (COPY_TO_REGCLASS FR64:$src, VR128), 0x44), sub_xmm),
+ (VPSHUFDri (COPY_TO_REGCLASS FR64:$src, VR128), 0x44), 1)>;
+
+ def : Pat<(v4i32 (X86VBroadcast GR32:$src)),
+ (VPSHUFDri (COPY_TO_REGCLASS GR32:$src, VR128), 0)>;
+ def : Pat<(v8i32 (X86VBroadcast GR32:$src)),
+ (VINSERTF128rr (INSERT_SUBREG (v8i32 (IMPLICIT_DEF)),
+ (VPSHUFDri (COPY_TO_REGCLASS GR32:$src, VR128), 0), sub_xmm),
+ (VPSHUFDri (COPY_TO_REGCLASS GR32:$src, VR128), 0), 1)>;
+ def : Pat<(v4i64 (X86VBroadcast GR64:$src)),
+ (VINSERTF128rr (INSERT_SUBREG (v4i64 (IMPLICIT_DEF)),
+ (VPSHUFDri (COPY_TO_REGCLASS GR64:$src, VR128), 0x44), sub_xmm),
+ (VPSHUFDri (COPY_TO_REGCLASS GR64:$src, VR128), 0x44), 1)>;
+ }
}
//===----------------------------------------------------------------------===//
//
multiclass avx2_perm<bits<8> opc, string OpcodeStr, PatFrag mem_frag,
- Intrinsic Int> {
+ ValueType OpVT> {
def Yrr : AVX28I<opc, MRMSrcReg, (outs VR256:$dst),
(ins VR256:$src1, VR256:$src2),
!strconcat(OpcodeStr,
"\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
- [(set VR256:$dst, (Int VR256:$src1, VR256:$src2))]>, VEX_4V;
+ [(set VR256:$dst,
+ (OpVT (X86VPermv VR256:$src1, VR256:$src2)))]>, VEX_4V;
def Yrm : AVX28I<opc, MRMSrcMem, (outs VR256:$dst),
(ins VR256:$src1, i256mem:$src2),
!strconcat(OpcodeStr,
"\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
- [(set VR256:$dst, (Int VR256:$src1,
- (bitconvert (mem_frag addr:$src2))))]>,
+ [(set VR256:$dst,
+ (OpVT (X86VPermv VR256:$src1,
+ (bitconvert (mem_frag addr:$src2)))))]>,
VEX_4V;
}
-defm VPERMD : avx2_perm<0x36, "vpermd", memopv4i64, int_x86_avx2_permd>;
+defm VPERMD : avx2_perm<0x36, "vpermd", memopv4i64, v8i32>;
let ExeDomain = SSEPackedSingle in
-defm VPERMPS : avx2_perm<0x16, "vpermps", memopv8f32, int_x86_avx2_permps>;
+defm VPERMPS : avx2_perm<0x16, "vpermps", memopv8f32, v8f32>;
multiclass avx2_perm_imm<bits<8> opc, string OpcodeStr, PatFrag mem_frag,
- Intrinsic Int> {
- def Yrr : AVX2AIi8<opc, MRMSrcReg, (outs VR256:$dst),
+ ValueType OpVT> {
+ def Yri : AVX2AIi8<opc, MRMSrcReg, (outs VR256:$dst),
(ins VR256:$src1, i8imm:$src2),
!strconcat(OpcodeStr,
"\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
- [(set VR256:$dst, (Int VR256:$src1, imm:$src2))]>, VEX;
- def Yrm : AVX2AIi8<opc, MRMSrcMem, (outs VR256:$dst),
+ [(set VR256:$dst,
+ (OpVT (X86VPermi VR256:$src1, (i8 imm:$src2))))]>, VEX;
+ 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, (Int (mem_frag addr:$src1), imm:$src2))]>,
- VEX;
+ [(set VR256:$dst,
+ (OpVT (X86VPermi (mem_frag addr:$src1),
+ (i8 imm:$src2))))]>, VEX;
}
-defm VPERMQ : avx2_perm_imm<0x00, "vpermq", memopv4i64, int_x86_avx2_permq>,
- VEX_W;
+defm VPERMQ : avx2_perm_imm<0x00, "vpermq", memopv4i64, v4i64>, VEX_W;
let ExeDomain = SSEPackedDouble in
-defm VPERMPD : avx2_perm_imm<0x01, "vpermpd", memopv4f64, int_x86_avx2_permpd>,
- VEX_W;
+defm VPERMPD : avx2_perm_imm<0x01, "vpermpd", memopv4f64, v4f64>, VEX_W;
//===----------------------------------------------------------------------===//
// VPERM2I128 - Permute Floating-Point Values in 128-bit chunks
//
-let AddedComplexity = 1 in {
def VPERM2I128rr : AVX2AIi8<0x46, MRMSrcReg, (outs VR256:$dst),
(ins VR256:$src1, VR256:$src2, i8imm:$src3),
"vperm2i128\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}",
"vperm2i128\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}",
[(set VR256:$dst, (X86VPerm2x128 VR256:$src1, (memopv4i64 addr:$src2),
(i8 imm:$src3)))]>, VEX_4V;
-}
-let Predicates = [HasAVX2], AddedComplexity = 1 in {
+let Predicates = [HasAVX2] in {
def : Pat<(v8i32 (X86VPerm2x128 VR256:$src1, VR256:$src2, (i8 imm:$imm))),
(VPERM2I128rr VR256:$src1, VR256:$src2, imm:$imm)>;
def : Pat<(v32i8 (X86VPerm2x128 VR256:$src1, VR256:$src2, (i8 imm:$imm))),
//===----------------------------------------------------------------------===//
// VINSERTI128 - Insert packed integer values
//
+let neverHasSideEffects = 1 in {
def VINSERTI128rr : AVX2AIi8<0x38, MRMSrcReg, (outs VR256:$dst),
(ins VR256:$src1, VR128:$src2, i8imm:$src3),
"vinserti128\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}",
- [(set VR256:$dst,
- (int_x86_avx2_vinserti128 VR256:$src1, VR128:$src2, imm:$src3))]>,
- VEX_4V;
+ []>, VEX_4V;
+let mayLoad = 1 in
def VINSERTI128rm : AVX2AIi8<0x38, MRMSrcMem, (outs VR256:$dst),
(ins VR256:$src1, i128mem:$src2, i8imm:$src3),
"vinserti128\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}",
- [(set VR256:$dst,
- (int_x86_avx2_vinserti128 VR256:$src1, (memopv2i64 addr:$src2),
- imm:$src3))]>, VEX_4V;
+ []>, VEX_4V;
+}
let Predicates = [HasAVX2] in {
def : Pat<(vinsertf128_insert:$ins (v4i64 VR256:$src1), (v2i64 VR128:$src2),
(i32 imm)),
(VINSERTI128rr VR256:$src1, VR128:$src2,
(INSERT_get_vinsertf128_imm VR256:$ins))>;
-}
-// AVX1 patterns
-let Predicates = [HasAVX] in {
-def : Pat<(vinsertf128_insert:$ins (v8f32 VR256:$src1), (v4f32 VR128:$src2),
+def : Pat<(vinsertf128_insert:$ins (v4i64 VR256:$src1), (memopv2i64 addr:$src2),
(i32 imm)),
- (VINSERTF128rr VR256:$src1, VR128:$src2,
+ (VINSERTI128rm VR256:$src1, addr:$src2,
(INSERT_get_vinsertf128_imm VR256:$ins))>;
-def : Pat<(vinsertf128_insert:$ins (v4f64 VR256:$src1), (v2f64 VR128:$src2),
+def : Pat<(vinsertf128_insert:$ins (v8i32 VR256:$src1),
+ (bc_v4i32 (memopv2i64 addr:$src2)),
(i32 imm)),
- (VINSERTF128rr VR256:$src1, VR128:$src2,
+ (VINSERTI128rm VR256:$src1, addr:$src2,
(INSERT_get_vinsertf128_imm VR256:$ins))>;
-def : Pat<(vinsertf128_insert:$ins (v4i64 VR256:$src1), (v2i64 VR128:$src2),
- (i32 imm)),
- (VINSERTF128rr VR256:$src1, VR128:$src2,
- (INSERT_get_vinsertf128_imm VR256:$ins))>;
-def : Pat<(vinsertf128_insert:$ins (v8i32 VR256:$src1), (v4i32 VR128:$src2),
+def : Pat<(vinsertf128_insert:$ins (v32i8 VR256:$src1),
+ (bc_v16i8 (memopv2i64 addr:$src2)),
(i32 imm)),
- (VINSERTF128rr VR256:$src1, VR128:$src2,
+ (VINSERTI128rm VR256:$src1, addr:$src2,
(INSERT_get_vinsertf128_imm VR256:$ins))>;
-def : Pat<(vinsertf128_insert:$ins (v32i8 VR256:$src1), (v16i8 VR128:$src2),
+def : Pat<(vinsertf128_insert:$ins (v16i16 VR256:$src1),
+ (bc_v8i16 (memopv2i64 addr:$src2)),
(i32 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)),
- (VINSERTF128rr VR256:$src1, VR128:$src2,
+ (VINSERTI128rm VR256:$src1, addr:$src2,
(INSERT_get_vinsertf128_imm VR256:$ins))>;
}
(v16i8 (VEXTRACTI128rr
(v32i8 VR256:$src1),
(EXTRACT_get_vextractf128_imm VR128:$ext)))>;
-}
-// AVX1 patterns
-let Predicates = [HasAVX] in {
-def : Pat<(vextractf128_extract:$ext VR256:$src1, (i32 imm)),
- (v4f32 (VEXTRACTF128rr
- (v8f32 VR256:$src1),
- (EXTRACT_get_vextractf128_imm VR128:$ext)))>;
-def : Pat<(vextractf128_extract:$ext VR256:$src1, (i32 imm)),
- (v2f64 (VEXTRACTF128rr
- (v4f64 VR256:$src1),
- (EXTRACT_get_vextractf128_imm VR128:$ext)))>;
-def : Pat<(vextractf128_extract:$ext VR256:$src1, (i32 imm)),
- (v2i64 (VEXTRACTF128rr
- (v4i64 VR256:$src1),
- (EXTRACT_get_vextractf128_imm VR128:$ext)))>;
-def : Pat<(vextractf128_extract:$ext VR256:$src1, (i32 imm)),
- (v4i32 (VEXTRACTF128rr
- (v8i32 VR256:$src1),
- (EXTRACT_get_vextractf128_imm VR128:$ext)))>;
-def : Pat<(vextractf128_extract:$ext VR256:$src1, (i32 imm)),
- (v8i16 (VEXTRACTF128rr
- (v16i16 VR256:$src1),
- (EXTRACT_get_vextractf128_imm VR128:$ext)))>;
-def : Pat<(vextractf128_extract:$ext VR256:$src1, (i32 imm)),
- (v16i8 (VEXTRACTF128rr
- (v32i8 VR256:$src1),
- (EXTRACT_get_vextractf128_imm VR128:$ext)))>;
+def : Pat<(alignedstore (v2i64 (vextractf128_extract:$ext (v4i64 VR256:$src1),
+ (i32 imm))), addr:$dst),
+ (VEXTRACTI128mr addr:$dst, VR256:$src1,
+ (EXTRACT_get_vextractf128_imm VR128:$ext))>;
+def : Pat<(alignedstore (v4i32 (vextractf128_extract:$ext (v8i32 VR256:$src1),
+ (i32 imm))), addr:$dst),
+ (VEXTRACTI128mr addr:$dst, VR256:$src1,
+ (EXTRACT_get_vextractf128_imm VR128:$ext))>;
+def : Pat<(alignedstore (v8i16 (vextractf128_extract:$ext (v16i16 VR256:$src1),
+ (i32 imm))), addr:$dst),
+ (VEXTRACTI128mr addr:$dst, VR256:$src1,
+ (EXTRACT_get_vextractf128_imm VR128:$ext))>;
+def : Pat<(alignedstore (v16i8 (vextractf128_extract:$ext (v32i8 VR256:$src1),
+ (i32 imm))), addr:$dst),
+ (VEXTRACTI128mr addr:$dst, VR256:$src1,
+ (EXTRACT_get_vextractf128_imm VR128:$ext))>;
}
//===----------------------------------------------------------------------===//
defm VPSRLVD : avx2_var_shift<0x45, "vpsrlvd", srl, v4i32, v8i32>;
defm VPSRLVQ : avx2_var_shift<0x45, "vpsrlvq", srl, v2i64, v4i64>, VEX_W;
defm VPSRAVD : avx2_var_shift<0x46, "vpsravd", sra, v4i32, v8i32>;
+
+//===----------------------------------------------------------------------===//
+// VGATHER - GATHER Operations
+multiclass avx2_gather<bits<8> opc, string OpcodeStr, RegisterClass RC256,
+ X86MemOperand memop128, X86MemOperand memop256> {
+ def rm : AVX28I<opc, MRMSrcMem, (outs VR128:$dst, VR128:$mask_wb),
+ (ins VR128:$src1, memop128:$src2, VR128:$mask),
+ !strconcat(OpcodeStr,
+ "\t{$mask, $src2, $dst|$dst, $src2, $mask}"),
+ []>, VEX_4VOp3;
+ def Yrm : AVX28I<opc, MRMSrcMem, (outs RC256:$dst, RC256:$mask_wb),
+ (ins RC256:$src1, memop256:$src2, RC256:$mask),
+ !strconcat(OpcodeStr,
+ "\t{$mask, $src2, $dst|$dst, $src2, $mask}"),
+ []>, VEX_4VOp3, VEX_L;
+}
+
+let mayLoad = 1, Constraints = "$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>;
+ defm VGATHERQPS : avx2_gather<0x93, "vgatherqps", VR128, vx32mem, vy32mem>;
+ defm VPGATHERDQ : avx2_gather<0x90, "vpgatherdq", VR256, vx64mem, vx64mem>, VEX_W;
+ defm VPGATHERQQ : avx2_gather<0x91, "vpgatherqq", VR256, vx64mem, vy64mem>, VEX_W;
+ defm VPGATHERDD : avx2_gather<0x90, "vpgatherdd", VR256, vx32mem, vy32mem>;
+ defm VPGATHERQD : avx2_gather<0x91, "vpgatherqd", VR128, vx32mem, vy32mem>;
+}