// Patterns
let Predicates = [UseAVX] in {
- let AddedComplexity = 15 in {
- // Move scalar to XMM zero-extended, zeroing a VR128 then do a
- // MOVS{S,D} to the lower bits.
- def : Pat<(v4f32 (X86vzmovl (v4f32 (scalar_to_vector FR32:$src)))),
- (VMOVSSrr (v4f32 (V_SET0)), FR32:$src)>;
- def : Pat<(v4f32 (X86vzmovl (v4f32 VR128:$src))),
- (VMOVSSrr (v4f32 (V_SET0)), (COPY_TO_REGCLASS VR128:$src, FR32))>;
- def : Pat<(v4i32 (X86vzmovl (v4i32 VR128:$src))),
- (VMOVSSrr (v4i32 (V_SET0)), (COPY_TO_REGCLASS VR128:$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_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_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
(v2f64 (scalar_to_vector (loadf64 addr:$src))), (iPTR 0)))),
(SUBREG_TO_REG (i32 0), (VMOVSDrm addr:$src), sub_xmm)>;
}
- def : Pat<(v8f32 (X86vzmovl (insert_subvector undef,
- (v4f32 (scalar_to_vector FR32:$src)), (iPTR 0)))),
- (SUBREG_TO_REG (i32 0),
- (v4f32 (VMOVSSrr (v4f32 (V_SET0)), FR32:$src)),
- sub_xmm)>;
- def : Pat<(v4f64 (X86vzmovl (insert_subvector undef,
- (v2f64 (scalar_to_vector FR64:$src)), (iPTR 0)))),
- (SUBREG_TO_REG (i64 0),
- (v2f64 (VMOVSDrr (v2f64 (V_SET0)), FR64:$src)),
- sub_xmm)>;
def : Pat<(v4i64 (X86vzmovl (insert_subvector undef,
(v2i64 (scalar_to_vector (loadi64 addr:$src))), (iPTR 0)))),
(SUBREG_TO_REG (i64 0), (VMOVSDrm addr:$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_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_xmm)), sub_xmm)>;
-
// Extract and store.
def : Pat<(store (f32 (vector_extract (v4f32 VR128:$src), (iPTR 0))),
addr:$dst),
(EXTRACT_SUBREG (v4f64 VR256:$src2), sub_xmm)),
sub_xmm)>;
-
// FIXME: Instead of a X86Movlps there should be a X86Movsd here, the problem
// is during lowering, where it's not possible to recognize the fold cause
// it has two uses through a bitcast. One use disappears at isel time and the
}
let Predicates = [UseSSE1] in {
- let AddedComplexity = 15 in {
+ let Predicates = [NoSSE41], AddedComplexity = 15 in {
// Move scalar to XMM zero-extended, zeroing a VR128 then do a
// MOVSS to the lower bits.
def : Pat<(v4f32 (X86vzmovl (v4f32 (scalar_to_vector FR32:$src)))),
}
let Predicates = [UseSSE2] in {
- let AddedComplexity = 15 in {
+ let Predicates = [NoSSE41], AddedComplexity = 15 in {
// Move scalar to XMM zero-extended, zeroing a VR128 then do a
// MOVSD to the lower bits.
def : Pat<(v2f64 (X86vzmovl (v2f64 (scalar_to_vector FR64:$src)))),
string asm, Domain d,
OpndItins itins,
bit IsReMaterializable = 1> {
-let neverHasSideEffects = 1 in
+let hasSideEffects = 0 in
def rr : PI<opc, MRMSrcReg, (outs RC:$dst), (ins RC:$src),
!strconcat(asm, "\t{$src, $dst|$dst, $src}"), [], itins.rr, d>,
Sched<[WriteFShuffle]>;
(VMOVLPDrm VR128:$src1, addr:$src2)>;
def : Pat<(v2i64 (X86Movlpd VR128:$src1, (load addr:$src2))),
(VMOVLPDrm VR128:$src1, addr:$src2)>;
+ def : Pat<(v2f64 (X86Movsd VR128:$src1,
+ (v2f64 (scalar_to_vector (loadf64 addr:$src2))))),
+ (VMOVLPDrm VR128:$src1, addr:$src2)>;
// Store patterns
def : Pat<(store (v4f32 (X86Movlps (load addr:$src1), VR128:$src2)),
(MOVLPDrm VR128:$src1, addr:$src2)>;
def : Pat<(v2i64 (X86Movlpd VR128:$src1, (load addr:$src2))),
(MOVLPDrm VR128:$src1, addr:$src2)>;
+ def : Pat<(v2f64 (X86Movsd VR128:$src1,
+ (v2f64 (scalar_to_vector (loadf64 addr:$src2))))),
+ (MOVLPDrm VR128:$src1, addr:$src2)>;
// Store patterns
def : Pat<(store (v2f64 (X86Movlpd (load addr:$src1), VR128:$src2)),
(bc_v4i32 (v2i64 (X86vzload addr:$src2)))),
(VMOVHPSrm VR128:$src1, addr:$src2)>;
+ // VMOVHPD patterns
+
// 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
def : Pat<(v2f64 (X86Unpckl VR128:$src1,
(scalar_to_vector (loadf64 addr:$src2)))),
(VMOVHPDrm VR128:$src1, addr:$src2)>;
+ // Also handle an i64 load because that may get selected as a faster way to
+ // load the data.
+ def : Pat<(v2f64 (X86Unpckl VR128:$src1,
+ (bc_v2f64 (v2i64 (scalar_to_vector (loadi64 addr:$src2)))))),
+ (VMOVHPDrm VR128:$src1, addr:$src2)>;
+
+ def : Pat<(store (f64 (vector_extract
+ (v2f64 (X86VPermilpi VR128:$src, (i8 1))),
+ (iPTR 0))), addr:$dst),
+ (VMOVHPDmr addr:$dst, VR128:$src)>;
}
let Predicates = [UseSSE1] in {
}
let Predicates = [UseSSE2] in {
+ // MOVHPD patterns
+
// 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
def : Pat<(v2f64 (X86Unpckl VR128:$src1,
(scalar_to_vector (loadf64 addr:$src2)))),
(MOVHPDrm VR128:$src1, addr:$src2)>;
+ // Also handle an i64 load because that may get selected as a faster way to
+ // load the data.
+ def : Pat<(v2f64 (X86Unpckl VR128:$src1,
+ (bc_v2f64 (v2i64 (scalar_to_vector (loadi64 addr:$src2)))))),
+ (MOVHPDrm VR128:$src1, addr:$src2)>;
+
+ def : Pat<(store (f64 (vector_extract
+ (v2f64 (X86Shufp VR128:$src, VR128:$src, (i8 1))),
+ (iPTR 0))), addr:$dst),
+ (MOVHPDmr addr:$dst, VR128:$src)>;
}
//===----------------------------------------------------------------------===//
multiclass sse12_cvt_p<bits<8> opc, RegisterClass SrcRC, RegisterClass DstRC,
X86MemOperand x86memop, string asm, Domain d,
OpndItins itins> {
-let neverHasSideEffects = 1 in {
+let hasSideEffects = 0 in {
def rr : I<opc, MRMSrcReg, (outs DstRC:$dst), (ins SrcRC:$src), asm,
[], itins.rr, d>, Sched<[itins.Sched]>;
let mayLoad = 1 in
multiclass sse12_vcvt_avx<bits<8> opc, RegisterClass SrcRC, RegisterClass DstRC,
X86MemOperand x86memop, string asm> {
-let neverHasSideEffects = 1, Predicates = [UseAVX] in {
+let hasSideEffects = 0, Predicates = [UseAVX] in {
def rr : SI<opc, MRMSrcReg, (outs DstRC:$dst), (ins DstRC:$src1, SrcRC:$src),
!strconcat(asm,"\t{$src, $src1, $dst|$dst, $src1, $src}"), []>,
Sched<[WriteCvtI2F]>;
(ins DstRC:$src1, x86memop:$src),
!strconcat(asm,"\t{$src, $src1, $dst|$dst, $src1, $src}"), []>,
Sched<[WriteCvtI2FLd, ReadAfterLd]>;
-} // neverHasSideEffects = 1
+} // hasSideEffects = 0
}
let Predicates = [UseAVX] in {
/// SSE 2 Only
// Convert scalar double to scalar single
-let neverHasSideEffects = 1, Predicates = [UseAVX] in {
+let hasSideEffects = 0, Predicates = [UseAVX] in {
def VCVTSD2SSrr : VSDI<0x5A, MRMSrcReg, (outs FR32:$dst),
(ins FR64:$src1, FR64:$src2),
"cvtsd2ss\t{$src2, $src1, $dst|$dst, $src1, $src2}", [],
// Convert scalar single to scalar double
// SSE2 instructions with XS prefix
-let neverHasSideEffects = 1, Predicates = [UseAVX] in {
+let hasSideEffects = 0, Predicates = [UseAVX] in {
def VCVTSS2SDrr : I<0x5A, MRMSrcReg, (outs FR64:$dst),
(ins FR32:$src1, FR32:$src2),
"vcvtss2sd\t{$src2, $src1, $dst|$dst, $src1, $src2}",
// Convert Packed DW Integers to Packed Double FP
let Predicates = [HasAVX] in {
-let neverHasSideEffects = 1, mayLoad = 1 in
+let hasSideEffects = 0, mayLoad = 1 in
def VCVTDQ2PDrm : S2SI<0xE6, MRMSrcMem, (outs VR128:$dst), (ins i64mem:$src),
"vcvtdq2pd\t{$src, $dst|$dst, $src}",
[]>, VEX, Sched<[WriteCvtI2FLd]>;
Sched<[WriteCvtI2F]>;
}
-let neverHasSideEffects = 1, mayLoad = 1 in
+let hasSideEffects = 0, mayLoad = 1 in
def CVTDQ2PDrm : S2SI<0xE6, MRMSrcMem, (outs VR128:$dst), (ins i64mem:$src),
"cvtdq2pd\t{$src, $dst|$dst, $src}", [],
IIC_SSE_CVT_PD_RR>, Sched<[WriteCvtI2FLd]>;
/// sse12_shuffle - sse 1 & 2 fp shuffle instructions
multiclass sse12_shuffle<RegisterClass RC, X86MemOperand x86memop,
ValueType vt, string asm, PatFrag mem_frag,
- Domain d, bit IsConvertibleToThreeAddress = 0> {
+ Domain d> {
def rmi : PIi8<0xC6, MRMSrcMem, (outs RC:$dst),
(ins RC:$src1, x86memop:$src2, i8imm:$src3), asm,
[(set RC:$dst, (vt (X86Shufp RC:$src1, (mem_frag addr:$src2),
(i8 imm:$src3))))], IIC_SSE_SHUFP, d>,
Sched<[WriteFShuffleLd, ReadAfterLd]>;
- let isConvertibleToThreeAddress = IsConvertibleToThreeAddress in
- def rri : PIi8<0xC6, MRMSrcReg, (outs RC:$dst),
- (ins RC:$src1, RC:$src2, i8imm:$src3), asm,
- [(set RC:$dst, (vt (X86Shufp RC:$src1, RC:$src2,
- (i8 imm:$src3))))], IIC_SSE_SHUFP, d>,
- Sched<[WriteFShuffle]>;
+ def rri : PIi8<0xC6, MRMSrcReg, (outs RC:$dst),
+ (ins RC:$src1, RC:$src2, i8imm:$src3), asm,
+ [(set RC:$dst, (vt (X86Shufp RC:$src1, RC:$src2,
+ (i8 imm:$src3))))], IIC_SSE_SHUFP, d>,
+ Sched<[WriteFShuffle]>;
}
defm VSHUFPS : sse12_shuffle<VR128, f128mem, v4f32,
let Constraints = "$src1 = $dst" in {
defm SHUFPS : sse12_shuffle<VR128, f128mem, v4f32,
"shufps\t{$src3, $src2, $dst|$dst, $src2, $src3}",
- memopv4f32, SSEPackedSingle, 1 /* cvt to pshufd */>, PS;
+ memopv4f32, SSEPackedSingle>, PS;
defm SHUFPD : sse12_shuffle<VR128, f128mem, v2f64,
"shufpd\t{$src3, $src2, $dst|$dst, $src2, $src3}",
- memopv2f64, SSEPackedDouble, 1 /* cvt to pshufd */>, PD;
+ memopv2f64, SSEPackedDouble>, PD;
}
let Predicates = [HasAVX] in {
multiclass PDI_binop_all<bits<8> opc, string OpcodeStr, SDNode Opcode,
ValueType OpVT128, ValueType OpVT256,
OpndItins itins, bit IsCommutable = 0> {
-let Predicates = [HasAVX] in
+let Predicates = [HasAVX, NoVLX] in
defm V#NAME : PDI_binop_rm<opc, !strconcat("v", OpcodeStr), Opcode, OpVT128,
VR128, loadv2i64, i128mem, itins, IsCommutable, 0>, VEX_4V;
defm NAME : PDI_binop_rm<opc, OpcodeStr, Opcode, OpVT128, VR128,
memopv2i64, i128mem, itins, IsCommutable, 1>;
-let Predicates = [HasAVX2] in
+let Predicates = [HasAVX2, NoVLX] in
defm V#NAME#Y : PDI_binop_rm<opc, !strconcat("v", OpcodeStr), Opcode,
OpVT256, VR256, loadv4i64, i256mem, itins,
IsCommutable, 0>, VEX_4V, VEX_L;
///
multiclass sse12_fp_packed_logical<bits<8> opc, string OpcodeStr,
SDNode OpNode> {
+ let Predicates = [HasAVX, NoVLX] in {
defm V#NAME#PSY : sse12_fp_packed_logical_rm<opc, VR256, SSEPackedSingle,
!strconcat(OpcodeStr, "ps"), f256mem,
[(set VR256:$dst, (v4i64 (OpNode VR256:$src1, VR256:$src2)))],
[(set VR128:$dst, (OpNode (bc_v2i64 (v2f64 VR128:$src1)),
(loadv2i64 addr:$src2)))], 0>,
PD, VEX_4V;
+ }
let Constraints = "$src1 = $dst" in {
defm PS : sse12_fp_packed_logical_rm<opc, VR128, SSEPackedSingle,
/// classes below
multiclass basic_sse12_fp_binop_p<bits<8> opc, string OpcodeStr,
SDNode OpNode, SizeItins itins> {
+ let Predicates = [HasAVX, NoVLX] in {
defm V#NAME#PS : sse12_fp_packed<opc, !strconcat(OpcodeStr, "ps"), OpNode,
VR128, v4f32, f128mem, loadv4f32,
SSEPackedSingle, itins.s, 0>, PS, VEX_4V;
defm V#NAME#PDY : sse12_fp_packed<opc, !strconcat(OpcodeStr, "pd"),
OpNode, VR256, v4f64, f256mem, loadv4f64,
SSEPackedDouble, itins.d, 0>, PD, VEX_4V, VEX_L;
+ }
let Constraints = "$src1 = $dst" in {
defm PS : sse12_fp_packed<opc, !strconcat(OpcodeStr, "ps"), OpNode, VR128,
// previously we generated:
// addss %xmm0, %xmm1
// movss %xmm1, %xmm0
-//
+//
// we now generate:
// addss %xmm1, %xmm0
let Predicates = [UseSSE2] in {
// SSE2 patterns to select scalar double-precision fp arithmetic instructions
-
def : Pat<(v2f64 (X86Movsd (v2f64 VR128:$dst), (v2f64 (scalar_to_vector (fadd
(f64 (vector_extract (v2f64 VR128:$dst), (iPTR 0))),
FR64:$src))))),
}
let Predicates = [UseSSE41] in {
- // If the subtarget has SSE4.1 but not AVX, the vector insert
- // instruction is lowered into a X86insertps rather than a X86Movss.
- // When selecting SSE scalar single-precision fp arithmetic instructions,
- // make sure that we correctly match the X86insertps.
+ // If the subtarget has SSE4.1 but not AVX, the vector insert instruction is
+ // lowered into a X86insertps or a X86Blendi rather than a X86Movss. When
+ // selecting SSE scalar single-precision fp arithmetic instructions, make
+ // sure that we correctly match them.
def : Pat<(v4f32 (X86insertps (v4f32 VR128:$dst), (v4f32 (scalar_to_vector
(fadd (f32 (vector_extract (v4f32 VR128:$dst), (iPTR 0))),
(fdiv (f32 (vector_extract (v4f32 VR128:$dst), (iPTR 0))),
FR32:$src))), (iPTR 0))),
(DIVSSrr_Int v4f32:$dst, (COPY_TO_REGCLASS FR32:$src, VR128))>;
+
+ def : Pat<(v4f32 (X86Blendi (v4f32 VR128:$dst), (v4f32 (scalar_to_vector (fadd
+ (f32 (vector_extract (v4f32 VR128:$dst), (iPTR 0))),
+ FR32:$src))), (i8 1))),
+ (ADDSSrr_Int v4f32:$dst, (COPY_TO_REGCLASS FR32:$src, VR128))>;
+ def : Pat<(v4f32 (X86Blendi (v4f32 VR128:$dst), (v4f32 (scalar_to_vector (fsub
+ (f32 (vector_extract (v4f32 VR128:$dst), (iPTR 0))),
+ FR32:$src))), (i8 1))),
+ (SUBSSrr_Int v4f32:$dst, (COPY_TO_REGCLASS FR32:$src, VR128))>;
+ def : Pat<(v4f32 (X86Blendi (v4f32 VR128:$dst), (v4f32 (scalar_to_vector (fmul
+ (f32 (vector_extract (v4f32 VR128:$dst), (iPTR 0))),
+ FR32:$src))), (i8 1))),
+ (MULSSrr_Int v4f32:$dst, (COPY_TO_REGCLASS FR32:$src, VR128))>;
+ def : Pat<(v4f32 (X86Blendi (v4f32 VR128:$dst), (v4f32 (scalar_to_vector (fdiv
+ (f32 (vector_extract (v4f32 VR128:$dst), (iPTR 0))),
+ FR32:$src))), (i8 1))),
+ (DIVSSrr_Int v4f32:$dst, (COPY_TO_REGCLASS FR32:$src, VR128))>;
+
+ def : Pat<(v2f64 (X86Blendi (v2f64 VR128:$dst), (v2f64 (scalar_to_vector (fadd
+ (f64 (vector_extract (v2f64 VR128:$dst), (iPTR 0))),
+ FR64:$src))), (i8 1))),
+ (ADDSDrr_Int v2f64:$dst, (COPY_TO_REGCLASS FR64:$src, VR128))>;
+ def : Pat<(v2f64 (X86Blendi (v2f64 VR128:$dst), (v2f64 (scalar_to_vector (fsub
+ (f64 (vector_extract (v2f64 VR128:$dst), (iPTR 0))),
+ FR64:$src))), (i8 1))),
+ (SUBSDrr_Int v2f64:$dst, (COPY_TO_REGCLASS FR64:$src, VR128))>;
+ def : Pat<(v2f64 (X86Blendi (v2f64 VR128:$dst), (v2f64 (scalar_to_vector (fmul
+ (f64 (vector_extract (v2f64 VR128:$dst), (iPTR 0))),
+ FR64:$src))), (i8 1))),
+ (MULSDrr_Int v2f64:$dst, (COPY_TO_REGCLASS FR64:$src, VR128))>;
+ def : Pat<(v2f64 (X86Blendi (v2f64 VR128:$dst), (v2f64 (scalar_to_vector (fdiv
+ (f64 (vector_extract (v2f64 VR128:$dst), (iPTR 0))),
+ FR64:$src))), (i8 1))),
+ (DIVSDrr_Int v2f64:$dst, (COPY_TO_REGCLASS FR64:$src, VR128))>;
+
+ def : Pat<(v2f64 (X86Blendi (v2f64 (scalar_to_vector (fadd
+ (f64 (vector_extract (v2f64 VR128:$dst), (iPTR 0))),
+ FR64:$src))), (v2f64 VR128:$dst), (i8 2))),
+ (ADDSDrr_Int v2f64:$dst, (COPY_TO_REGCLASS FR64:$src, VR128))>;
+ def : Pat<(v2f64 (X86Blendi (v2f64 (scalar_to_vector (fsub
+ (f64 (vector_extract (v2f64 VR128:$dst), (iPTR 0))),
+ FR64:$src))), (v2f64 VR128:$dst), (i8 2))),
+ (SUBSDrr_Int v2f64:$dst, (COPY_TO_REGCLASS FR64:$src, VR128))>;
+ def : Pat<(v2f64 (X86Blendi (v2f64 (scalar_to_vector (fmul
+ (f64 (vector_extract (v2f64 VR128:$dst), (iPTR 0))),
+ FR64:$src))), (v2f64 VR128:$dst), (i8 2))),
+ (MULSDrr_Int v2f64:$dst, (COPY_TO_REGCLASS FR64:$src, VR128))>;
+ def : Pat<(v2f64 (X86Blendi (v2f64 (scalar_to_vector (fdiv
+ (f64 (vector_extract (v2f64 VR128:$dst), (iPTR 0))),
+ FR64:$src))), (v2f64 VR128:$dst), (i8 2))),
+ (DIVSDrr_Int v2f64:$dst, (COPY_TO_REGCLASS FR64:$src, VR128))>;
}
let Predicates = [HasAVX] in {
(fdiv (f32 (vector_extract (v4f32 VR128:$dst), (iPTR 0))),
FR32:$src))), (iPTR 0))),
(VDIVSSrr_Int v4f32:$dst, (COPY_TO_REGCLASS FR32:$src, VR128))>;
+
+ def : Pat<(v4f32 (X86Blendi (v4f32 VR128:$dst), (v4f32 (scalar_to_vector (fadd
+ (f32 (vector_extract (v4f32 VR128:$dst), (iPTR 0))),
+ FR32:$src))), (i8 1))),
+ (VADDSSrr_Int v4f32:$dst, (COPY_TO_REGCLASS FR32:$src, VR128))>;
+ def : Pat<(v4f32 (X86Blendi (v4f32 VR128:$dst), (v4f32 (scalar_to_vector (fsub
+ (f32 (vector_extract (v4f32 VR128:$dst), (iPTR 0))),
+ FR32:$src))), (i8 1))),
+ (VSUBSSrr_Int v4f32:$dst, (COPY_TO_REGCLASS FR32:$src, VR128))>;
+ def : Pat<(v4f32 (X86Blendi (v4f32 VR128:$dst), (v4f32 (scalar_to_vector (fmul
+ (f32 (vector_extract (v4f32 VR128:$dst), (iPTR 0))),
+ FR32:$src))), (i8 1))),
+ (VMULSSrr_Int v4f32:$dst, (COPY_TO_REGCLASS FR32:$src, VR128))>;
+ def : Pat<(v4f32 (X86Blendi (v4f32 VR128:$dst), (v4f32 (scalar_to_vector (fdiv
+ (f32 (vector_extract (v4f32 VR128:$dst), (iPTR 0))),
+ FR32:$src))), (i8 1))),
+ (VDIVSSrr_Int v4f32:$dst, (COPY_TO_REGCLASS FR32:$src, VR128))>;
+
+ def : Pat<(v2f64 (X86Blendi (v2f64 VR128:$dst), (v2f64 (scalar_to_vector (fadd
+ (f64 (vector_extract (v2f64 VR128:$dst), (iPTR 0))),
+ FR64:$src))), (i8 1))),
+ (VADDSDrr_Int v2f64:$dst, (COPY_TO_REGCLASS FR64:$src, VR128))>;
+ def : Pat<(v2f64 (X86Blendi (v2f64 VR128:$dst), (v2f64 (scalar_to_vector (fsub
+ (f64 (vector_extract (v2f64 VR128:$dst), (iPTR 0))),
+ FR64:$src))), (i8 1))),
+ (VSUBSDrr_Int v2f64:$dst, (COPY_TO_REGCLASS FR64:$src, VR128))>;
+ def : Pat<(v2f64 (X86Blendi (v2f64 VR128:$dst), (v2f64 (scalar_to_vector (fmul
+ (f64 (vector_extract (v2f64 VR128:$dst), (iPTR 0))),
+ FR64:$src))), (i8 1))),
+ (VMULSDrr_Int v2f64:$dst, (COPY_TO_REGCLASS FR64:$src, VR128))>;
+ def : Pat<(v2f64 (X86Blendi (v2f64 VR128:$dst), (v2f64 (scalar_to_vector (fdiv
+ (f64 (vector_extract (v2f64 VR128:$dst), (iPTR 0))),
+ FR64:$src))), (i8 1))),
+ (VDIVSDrr_Int v2f64:$dst, (COPY_TO_REGCLASS FR64:$src, VR128))>;
+
+ def : Pat<(v2f64 (X86Blendi (v2f64 (scalar_to_vector (fadd
+ (f64 (vector_extract (v2f64 VR128:$dst), (iPTR 0))),
+ FR64:$src))), (v2f64 VR128:$dst), (i8 2))),
+ (VADDSDrr_Int v2f64:$dst, (COPY_TO_REGCLASS FR64:$src, VR128))>;
+ def : Pat<(v2f64 (X86Blendi (v2f64 (scalar_to_vector (fsub
+ (f64 (vector_extract (v2f64 VR128:$dst), (iPTR 0))),
+ FR64:$src))), (v2f64 VR128:$dst), (i8 2))),
+ (VSUBSDrr_Int v2f64:$dst, (COPY_TO_REGCLASS FR64:$src, VR128))>;
+ def : Pat<(v2f64 (X86Blendi (v2f64 (scalar_to_vector (fmul
+ (f64 (vector_extract (v2f64 VR128:$dst), (iPTR 0))),
+ FR64:$src))), (v2f64 VR128:$dst), (i8 2))),
+ (VMULSDrr_Int v2f64:$dst, (COPY_TO_REGCLASS FR64:$src, VR128))>;
+ def : Pat<(v2f64 (X86Blendi (v2f64 (scalar_to_vector (fdiv
+ (f64 (vector_extract (v2f64 VR128:$dst), (iPTR 0))),
+ FR64:$src))), (v2f64 VR128:$dst), (i8 2))),
+ (VDIVSDrr_Int v2f64:$dst, (COPY_TO_REGCLASS FR64:$src, VR128))>;
}
// Patterns used to select SSE scalar fp arithmetic instructions from
// previously we generated:
// addps %xmm0, %xmm1
// movss %xmm1, %xmm0
-//
+//
// we now generate:
// addss %xmm1, %xmm0
def : Pat<(v4f32 (X86Movss (v4f32 VR128:$dst),
(fadd (v4f32 VR128:$dst), (v4f32 VR128:$src)))),
(ADDSSrr_Int v4f32:$dst, v4f32:$src)>;
- def : Pat<(v4f32 (X86Movss (v4f32 VR128:$dst),
+ def : Pat<(v4f32 (X86Movss (v4f32 VR128:$dst),
(fsub (v4f32 VR128:$dst), (v4f32 VR128:$src)))),
(SUBSSrr_Int v4f32:$dst, v4f32:$src)>;
def : Pat<(v4f32 (X86Movss (v4f32 VR128:$dst),
(fmul (v4f32 VR128:$dst), (v4f32 VR128:$src)))),
(MULSSrr_Int v4f32:$dst, v4f32:$src)>;
- def : Pat<(v4f32 (X86Movss (v4f32 VR128:$dst),
+ def : Pat<(v4f32 (X86Movss (v4f32 VR128:$dst),
(fdiv (v4f32 VR128:$dst), (v4f32 VR128:$src)))),
(DIVSSrr_Int v4f32:$dst, v4f32:$src)>;
}
(DIVSDrr_Int v2f64:$dst, v2f64:$src)>;
}
+let Predicates = [UseSSE41] in {
+ // With SSE4.1 we may see these operations using X86Blendi rather than
+ // X86Movs{s,d}.
+ def : Pat<(v4f32 (X86Blendi (v4f32 VR128:$dst),
+ (fadd (v4f32 VR128:$dst), (v4f32 VR128:$src)), (i8 1))),
+ (ADDSSrr_Int v4f32:$dst, v4f32:$src)>;
+ def : Pat<(v4f32 (X86Blendi (v4f32 VR128:$dst),
+ (fsub (v4f32 VR128:$dst), (v4f32 VR128:$src)), (i8 1))),
+ (SUBSSrr_Int v4f32:$dst, v4f32:$src)>;
+ def : Pat<(v4f32 (X86Blendi (v4f32 VR128:$dst),
+ (fmul (v4f32 VR128:$dst), (v4f32 VR128:$src)), (i8 1))),
+ (MULSSrr_Int v4f32:$dst, v4f32:$src)>;
+ def : Pat<(v4f32 (X86Blendi (v4f32 VR128:$dst),
+ (fdiv (v4f32 VR128:$dst), (v4f32 VR128:$src)), (i8 1))),
+ (DIVSSrr_Int v4f32:$dst, v4f32:$src)>;
+
+ def : Pat<(v2f64 (X86Blendi (v2f64 VR128:$dst),
+ (fadd (v2f64 VR128:$dst), (v2f64 VR128:$src)), (i8 1))),
+ (ADDSDrr_Int v2f64:$dst, v2f64:$src)>;
+ def : Pat<(v2f64 (X86Blendi (v2f64 VR128:$dst),
+ (fsub (v2f64 VR128:$dst), (v2f64 VR128:$src)), (i8 1))),
+ (SUBSDrr_Int v2f64:$dst, v2f64:$src)>;
+ def : Pat<(v2f64 (X86Blendi (v2f64 VR128:$dst),
+ (fmul (v2f64 VR128:$dst), (v2f64 VR128:$src)), (i8 1))),
+ (MULSDrr_Int v2f64:$dst, v2f64:$src)>;
+ def : Pat<(v2f64 (X86Blendi (v2f64 VR128:$dst),
+ (fdiv (v2f64 VR128:$dst), (v2f64 VR128:$src)), (i8 1))),
+ (DIVSDrr_Int v2f64:$dst, v2f64:$src)>;
+
+ def : Pat<(v2f64 (X86Blendi (fadd (v2f64 VR128:$dst), (v2f64 VR128:$src)),
+ (v2f64 VR128:$dst), (i8 2))),
+ (ADDSDrr_Int v2f64:$dst, v2f64:$src)>;
+ def : Pat<(v2f64 (X86Blendi (fsub (v2f64 VR128:$dst), (v2f64 VR128:$src)),
+ (v2f64 VR128:$dst), (i8 2))),
+ (SUBSDrr_Int v2f64:$dst, v2f64:$src)>;
+ def : Pat<(v2f64 (X86Blendi (fmul (v2f64 VR128:$dst), (v2f64 VR128:$src)),
+ (v2f64 VR128:$dst), (i8 2))),
+ (MULSDrr_Int v2f64:$dst, v2f64:$src)>;
+ def : Pat<(v2f64 (X86Blendi (fdiv (v2f64 VR128:$dst), (v2f64 VR128:$src)),
+ (v2f64 VR128:$dst), (i8 2))),
+ (DIVSDrr_Int v2f64:$dst, v2f64:$src)>;
+}
+
let Predicates = [HasAVX] in {
// The following patterns select AVX Scalar single/double precision fp
// arithmetic instructions from a packed single precision fp instruction
def : Pat<(v2f64 (X86Movsd (v2f64 VR128:$dst),
(fdiv (v2f64 VR128:$dst), (v2f64 VR128:$src)))),
(VDIVSDrr_Int v2f64:$dst, v2f64:$src)>;
+
+ // Also handle X86Blendi-based patterns.
+ def : Pat<(v4f32 (X86Blendi (v4f32 VR128:$dst),
+ (fadd (v4f32 VR128:$dst), (v4f32 VR128:$src)), (i8 1))),
+ (VADDSSrr_Int v4f32:$dst, v4f32:$src)>;
+ def : Pat<(v4f32 (X86Blendi (v4f32 VR128:$dst),
+ (fsub (v4f32 VR128:$dst), (v4f32 VR128:$src)), (i8 1))),
+ (VSUBSSrr_Int v4f32:$dst, v4f32:$src)>;
+ def : Pat<(v4f32 (X86Blendi (v4f32 VR128:$dst),
+ (fmul (v4f32 VR128:$dst), (v4f32 VR128:$src)), (i8 1))),
+ (VMULSSrr_Int v4f32:$dst, v4f32:$src)>;
+ def : Pat<(v4f32 (X86Blendi (v4f32 VR128:$dst),
+ (fdiv (v4f32 VR128:$dst), (v4f32 VR128:$src)), (i8 1))),
+ (VDIVSSrr_Int v4f32:$dst, v4f32:$src)>;
+
+ def : Pat<(v2f64 (X86Blendi (v2f64 VR128:$dst),
+ (fadd (v2f64 VR128:$dst), (v2f64 VR128:$src)), (i8 1))),
+ (VADDSDrr_Int v2f64:$dst, v2f64:$src)>;
+ def : Pat<(v2f64 (X86Blendi (v2f64 VR128:$dst),
+ (fsub (v2f64 VR128:$dst), (v2f64 VR128:$src)), (i8 1))),
+ (VSUBSDrr_Int v2f64:$dst, v2f64:$src)>;
+ def : Pat<(v2f64 (X86Blendi (v2f64 VR128:$dst),
+ (fmul (v2f64 VR128:$dst), (v2f64 VR128:$src)), (i8 1))),
+ (VMULSDrr_Int v2f64:$dst, v2f64:$src)>;
+ def : Pat<(v2f64 (X86Blendi (v2f64 VR128:$dst),
+ (fdiv (v2f64 VR128:$dst), (v2f64 VR128:$src)), (i8 1))),
+ (VDIVSDrr_Int v2f64:$dst, v2f64:$src)>;
+
+ def : Pat<(v2f64 (X86Blendi (fadd (v2f64 VR128:$dst), (v2f64 VR128:$src)),
+ (v2f64 VR128:$dst), (i8 2))),
+ (VADDSDrr_Int v2f64:$dst, v2f64:$src)>;
+ def : Pat<(v2f64 (X86Blendi (fsub (v2f64 VR128:$dst), (v2f64 VR128:$src)),
+ (v2f64 VR128:$dst), (i8 2))),
+ (VSUBSDrr_Int v2f64:$dst, v2f64:$src)>;
+ def : Pat<(v2f64 (X86Blendi (fmul (v2f64 VR128:$dst), (v2f64 VR128:$src)),
+ (v2f64 VR128:$dst), (i8 2))),
+ (VMULSDrr_Int v2f64:$dst, v2f64:$src)>;
+ def : Pat<(v2f64 (X86Blendi (fdiv (v2f64 VR128:$dst), (v2f64 VR128:$src)),
+ (v2f64 VR128:$dst), (i8 2))),
+ (VDIVSDrr_Int v2f64:$dst, v2f64:$src)>;
}
/// Unop Arithmetic
>;
}
+let Sched = WriteFRsqrt in {
+def SSE_RSQRTPS : OpndItins<
+ IIC_SSE_RSQRTPS_RR, IIC_SSE_RSQRTPS_RM
+>;
+
+def SSE_RSQRTSS : OpndItins<
+ IIC_SSE_RSQRTSS_RR, IIC_SSE_RSQRTSS_RM
+>;
+}
+
let Sched = WriteFRcp in {
def SSE_RCPP : OpndItins<
IIC_SSE_RCPP_RR, IIC_SSE_RCPP_RM
>;
}
-/// sse1_fp_unop_s - SSE1 unops in scalar form.
-multiclass sse1_fp_unop_s<bits<8> opc, string OpcodeStr,
- SDNode OpNode, Intrinsic F32Int, OpndItins itins> {
-let Predicates = [HasAVX], hasSideEffects = 0 in {
- def V#NAME#SSr : SSI<opc, MRMSrcReg, (outs FR32:$dst),
- (ins FR32:$src1, FR32:$src2),
- !strconcat("v", OpcodeStr,
- "ss\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
- []>, VEX_4V, VEX_LIG, Sched<[itins.Sched]>;
- let mayLoad = 1 in {
- def V#NAME#SSm : SSI<opc, MRMSrcMem, (outs FR32:$dst),
- (ins FR32:$src1,f32mem:$src2),
- !strconcat("v", OpcodeStr,
- "ss\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
- []>, VEX_4V, VEX_LIG,
- Sched<[itins.Sched.Folded, ReadAfterLd]>;
- let isCodeGenOnly = 1 in
- def V#NAME#SSm_Int : SSI<opc, MRMSrcMem, (outs VR128:$dst),
- (ins VR128:$src1, ssmem:$src2),
- !strconcat("v", OpcodeStr,
- "ss\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
- []>, VEX_4V, VEX_LIG,
- Sched<[itins.Sched.Folded, ReadAfterLd]>;
- }
-}
-
- def SSr : SSI<opc, MRMSrcReg, (outs FR32:$dst), (ins FR32:$src),
- !strconcat(OpcodeStr, "ss\t{$src, $dst|$dst, $src}"),
- [(set FR32:$dst, (OpNode FR32:$src))]>, Sched<[itins.Sched]>;
- // For scalar unary operations, fold a load into the operation
- // only in OptForSize mode. It eliminates an instruction, but it also
- // eliminates a whole-register clobber (the load), so it introduces a
- // partial register update condition.
- def SSm : I<opc, MRMSrcMem, (outs FR32:$dst), (ins f32mem:$src),
- !strconcat(OpcodeStr, "ss\t{$src, $dst|$dst, $src}"),
- [(set FR32:$dst, (OpNode (load addr:$src)))], itins.rm>, XS,
- Requires<[UseSSE1, OptForSize]>, Sched<[itins.Sched.Folded]>;
-let isCodeGenOnly = 1 in {
- def SSr_Int : SSI<opc, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
- !strconcat(OpcodeStr, "ss\t{$src, $dst|$dst, $src}"),
- [(set VR128:$dst, (F32Int VR128:$src))], itins.rr>,
- Sched<[itins.Sched]>;
- def SSm_Int : SSI<opc, MRMSrcMem, (outs VR128:$dst), (ins ssmem:$src),
- !strconcat(OpcodeStr, "ss\t{$src, $dst|$dst, $src}"),
- [(set VR128:$dst, (F32Int sse_load_f32:$src))], itins.rm>,
- Sched<[itins.Sched.Folded]>;
-}
-}
-
-/// sse1_fp_unop_s_rw - SSE1 unops where vector form has a read-write operand.
-multiclass sse1_fp_unop_rw<bits<8> opc, string OpcodeStr, SDNode OpNode,
+/// sse1_fp_unop_s - SSE1 unops in scalar form
+/// For the non-AVX defs, we need $src1 to be tied to $dst because
+/// the HW instructions are 2 operand / destructive.
+multiclass sse1_fp_unop_s<bits<8> opc, string OpcodeStr, SDNode OpNode,
OpndItins itins> {
let Predicates = [HasAVX], hasSideEffects = 0 in {
def V#NAME#SSr : SSI<opc, MRMSrcReg, (outs FR32:$dst),
}
// Square root.
-defm SQRT : sse1_fp_unop_s<0x51, "sqrt", fsqrt, int_x86_sse_sqrt_ss,
- SSE_SQRTSS>,
+defm SQRT : sse1_fp_unop_s<0x51, "sqrt", fsqrt, SSE_SQRTSS>,
sse1_fp_unop_p<0x51, "sqrt", fsqrt, SSE_SQRTPS>,
- sse2_fp_unop_s<0x51, "sqrt", fsqrt, int_x86_sse2_sqrt_sd,
+ sse2_fp_unop_s<0x51, "sqrt", fsqrt, int_x86_sse2_sqrt_sd,
SSE_SQRTSD>,
sse2_fp_unop_p<0x51, "sqrt", fsqrt, SSE_SQRTPD>;
// Reciprocal approximations. Note that these typically require refinement
// in order to obtain suitable precision.
-defm RSQRT : sse1_fp_unop_rw<0x52, "rsqrt", X86frsqrt, SSE_SQRTSS>,
- sse1_fp_unop_p<0x52, "rsqrt", X86frsqrt, SSE_SQRTPS>,
+defm RSQRT : sse1_fp_unop_s<0x52, "rsqrt", X86frsqrt, SSE_RSQRTSS>,
+ sse1_fp_unop_p<0x52, "rsqrt", X86frsqrt, SSE_RSQRTPS>,
sse1_fp_unop_p_int<0x52, "rsqrt", int_x86_sse_rsqrt_ps,
- int_x86_avx_rsqrt_ps_256, SSE_SQRTPS>;
-defm RCP : sse1_fp_unop_rw<0x53, "rcp", X86frcp, SSE_RCPS>,
+ int_x86_avx_rsqrt_ps_256, SSE_RSQRTPS>;
+defm RCP : sse1_fp_unop_s<0x53, "rcp", X86frcp, SSE_RCPS>,
sse1_fp_unop_p<0x53, "rcp", X86frcp, SSE_RCPP>,
sse1_fp_unop_p_int<0x53, "rcp", int_x86_sse_rcp_ps,
int_x86_avx_rcp_ps_256, SSE_RCPP>;
(VRCPSSm_Int (v4f32 (IMPLICIT_DEF)), sse_load_f32:$src)>;
}
-// Reciprocal approximations. Note that these typically require refinement
-// in order to obtain suitable precision.
+// These are unary operations, but they are modeled as having 2 source operands
+// because the high elements of the destination are unchanged in SSE.
let Predicates = [UseSSE1] in {
def : Pat<(int_x86_sse_rsqrt_ss VR128:$src),
(RSQRTSSr_Int VR128:$src, VR128:$src)>;
def : Pat<(int_x86_sse_rcp_ss VR128:$src),
(RCPSSr_Int VR128:$src, VR128:$src)>;
+ def : Pat<(int_x86_sse_sqrt_ss VR128:$src),
+ (SQRTSSr_Int VR128:$src, VR128:$src)>;
}
// There is no f64 version of the reciprocal approximation instructions.
PS, Requires<[HasSSE2]>;
} // SchedRW = [WriteStore]
+let Predicates = [HasAVX, NoVLX] in {
+ def : Pat<(alignednontemporalstore (v4i32 VR128:$src), addr:$dst),
+ (VMOVNTPSmr addr:$dst, VR128:$src)>;
+}
+
+def : Pat<(alignednontemporalstore (v4i32 VR128:$src), addr:$dst),
+ (MOVNTPSmr addr:$dst, VR128:$src)>;
+
} // AddedComplexity
//===----------------------------------------------------------------------===//
let SchedRW = [WriteNop] in {
// 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", [(int_x86_sse2_pause)], IIC_SSE_PAUSE>,
+def PAUSE : I<0x90, RawFrm, (outs), (ins),
+ "pause", [(int_x86_sse2_pause)], IIC_SSE_PAUSE>,
OBXS, Requires<[HasSSE2]>;
}
"stmxcsr\t$dst", [(int_x86_sse_stmxcsr addr:$dst)],
IIC_SSE_STMXCSR>, VEX, Sched<[WriteStore]>;
-def LDMXCSR : PSI<0xAE, MRM2m, (outs), (ins i32mem:$src),
- "ldmxcsr\t$src", [(int_x86_sse_ldmxcsr addr:$src)],
- IIC_SSE_LDMXCSR>, Sched<[WriteLoad]>;
-def STMXCSR : PSI<0xAE, MRM3m, (outs), (ins i32mem:$dst),
- "stmxcsr\t$dst", [(int_x86_sse_stmxcsr addr:$dst)],
- IIC_SSE_STMXCSR>, Sched<[WriteStore]>;
+let Predicates = [UseSSE1] in {
+def LDMXCSR : I<0xAE, MRM2m, (outs), (ins i32mem:$src),
+ "ldmxcsr\t$src", [(int_x86_sse_ldmxcsr addr:$src)],
+ IIC_SSE_LDMXCSR>, TB, Sched<[WriteLoad]>;
+def STMXCSR : I<0xAE, MRM3m, (outs), (ins i32mem:$dst),
+ "stmxcsr\t$dst", [(int_x86_sse_stmxcsr addr:$dst)],
+ IIC_SSE_STMXCSR>, TB, Sched<[WriteStore]>;
+}
//===---------------------------------------------------------------------===//
// SSE2 - Move Aligned/Unaligned Packed Integer Instructions
let ExeDomain = SSEPackedInt in { // SSE integer instructions
-let neverHasSideEffects = 1, SchedRW = [WriteMove] in {
+let hasSideEffects = 0, SchedRW = [WriteMove] in {
def VMOVDQArr : VPDI<0x6F, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
"movdqa\t{$src, $dst|$dst, $src}", [], IIC_SSE_MOVA_P_RR>,
VEX;
}
let canFoldAsLoad = 1, mayLoad = 1, isReMaterializable = 1,
- neverHasSideEffects = 1, SchedRW = [WriteLoad] in {
+ hasSideEffects = 0, SchedRW = [WriteLoad] in {
def VMOVDQArm : VPDI<0x6F, MRMSrcMem, (outs VR128:$dst), (ins i128mem:$src),
"movdqa\t{$src, $dst|$dst, $src}", [], IIC_SSE_MOVA_P_RM>,
VEX;
}
}
-let mayStore = 1, neverHasSideEffects = 1, SchedRW = [WriteStore] in {
+let mayStore = 1, hasSideEffects = 0, SchedRW = [WriteStore] in {
def VMOVDQAmr : VPDI<0x7F, MRMDestMem, (outs),
(ins i128mem:$dst, VR128:$src),
"movdqa\t{$src, $dst|$dst, $src}", [], IIC_SSE_MOVA_P_MR>,
}
let SchedRW = [WriteMove] in {
-let neverHasSideEffects = 1 in
+let hasSideEffects = 0 in
def MOVDQArr : PDI<0x6F, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
"movdqa\t{$src, $dst|$dst, $src}", [], IIC_SSE_MOVA_P_RR>;
} // SchedRW
let canFoldAsLoad = 1, mayLoad = 1, isReMaterializable = 1,
- neverHasSideEffects = 1, SchedRW = [WriteLoad] in {
+ hasSideEffects = 0, SchedRW = [WriteLoad] in {
def MOVDQArm : PDI<0x6F, MRMSrcMem, (outs VR128:$dst), (ins i128mem:$src),
"movdqa\t{$src, $dst|$dst, $src}",
[/*(set VR128:$dst, (alignedloadv2i64 addr:$src))*/],
XS, Requires<[UseSSE2]>;
}
-let mayStore = 1, neverHasSideEffects = 1, SchedRW = [WriteStore] in {
+let mayStore = 1, hasSideEffects = 0, SchedRW = [WriteStore] in {
def MOVDQAmr : PDI<0x7F, MRMDestMem, (outs), (ins i128mem:$dst, VR128:$src),
"movdqa\t{$src, $dst|$dst, $src}",
[/*(alignedstore (v2i64 VR128:$src), addr:$dst)*/],
//===---------------------------------------------------------------------===//
multiclass sse3_replicate_dfp<string OpcodeStr> {
-let neverHasSideEffects = 1 in
+let hasSideEffects = 0 in
def rr : S3DI<0x12, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
!strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"),
[], IIC_SSE_MOV_LH>, Sched<[WriteFShuffle]>;
//===---------------------------------------------------------------------===//
multiclass ssse3_palignr<string asm, bit Is2Addr = 1> {
- let neverHasSideEffects = 1 in {
+ let hasSideEffects = 0 in {
def R128rr : SS3AI<0x0F, MRMSrcReg, (outs VR128:$dst),
(ins VR128:$src1, VR128:$src2, i8imm:$src3),
!if(Is2Addr,
}
multiclass ssse3_palignr_y<string asm, bit Is2Addr = 1> {
- let neverHasSideEffects = 1 in {
+ let hasSideEffects = 0 in {
def R256rr : SS3AI<0x0F, MRMSrcReg, (outs VR256:$dst),
(ins VR256:$src1, VR256:$src2, i8imm:$src3),
!strconcat(asm,
// SSE4.1 - Packed Move with Sign/Zero Extend
//===----------------------------------------------------------------------===//
-multiclass SS41I_binop_rm_int8<bits<8> opc, string OpcodeStr, Intrinsic IntId,
- OpndItins itins = DEFAULT_ITINS> {
- def rr : SS48I<opc, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
- !strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"),
- [(set VR128:$dst, (IntId VR128:$src))], itins.rr>,
- Sched<[itins.Sched]>;
-
- def rm : SS48I<opc, MRMSrcMem, (outs VR128:$dst), (ins i64mem:$src),
- !strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"),
- [(set VR128:$dst,
- (IntId (bitconvert (v2i64 (scalar_to_vector (loadi64 addr:$src))))))],
- itins.rm>, Sched<[itins.Sched.Folded]>;
-}
-
-multiclass SS41I_binop_rm_int16_y<bits<8> opc, string OpcodeStr,
- Intrinsic IntId, X86FoldableSchedWrite Sched> {
- def Yrr : SS48I<opc, MRMSrcReg, (outs VR256:$dst), (ins VR128:$src),
- !strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"),
- [(set VR256:$dst, (IntId VR128:$src))]>, Sched<[Sched]>;
-
- def Yrm : SS48I<opc, MRMSrcMem, (outs VR256:$dst), (ins i128mem:$src),
- !strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"),
- [(set VR256:$dst, (IntId (load addr:$src)))]>,
- Sched<[Sched.Folded]>;
-}
-
-let Predicates = [HasAVX] in {
-defm VPMOVSXBW : SS41I_binop_rm_int8<0x20, "vpmovsxbw",
- int_x86_sse41_pmovsxbw,
- DEFAULT_ITINS_SHUFFLESCHED>, VEX;
-defm VPMOVSXWD : SS41I_binop_rm_int8<0x23, "vpmovsxwd",
- int_x86_sse41_pmovsxwd,
- DEFAULT_ITINS_SHUFFLESCHED>, VEX;
-defm VPMOVSXDQ : SS41I_binop_rm_int8<0x25, "vpmovsxdq",
- int_x86_sse41_pmovsxdq,
- DEFAULT_ITINS_SHUFFLESCHED>, VEX;
-defm VPMOVZXBW : SS41I_binop_rm_int8<0x30, "vpmovzxbw",
- int_x86_sse41_pmovzxbw,
- DEFAULT_ITINS_SHUFFLESCHED>, VEX;
-defm VPMOVZXWD : SS41I_binop_rm_int8<0x33, "vpmovzxwd",
- int_x86_sse41_pmovzxwd,
- DEFAULT_ITINS_SHUFFLESCHED>, VEX;
-defm VPMOVZXDQ : SS41I_binop_rm_int8<0x35, "vpmovzxdq",
- int_x86_sse41_pmovzxdq,
- DEFAULT_ITINS_SHUFFLESCHED>, VEX;
-}
-
-let Predicates = [HasAVX2] in {
-defm VPMOVSXBW : SS41I_binop_rm_int16_y<0x20, "vpmovsxbw",
- int_x86_avx2_pmovsxbw,
- WriteShuffle>, VEX, VEX_L;
-defm VPMOVSXWD : SS41I_binop_rm_int16_y<0x23, "vpmovsxwd",
- int_x86_avx2_pmovsxwd,
- WriteShuffle>, VEX, VEX_L;
-defm VPMOVSXDQ : SS41I_binop_rm_int16_y<0x25, "vpmovsxdq",
- int_x86_avx2_pmovsxdq,
- WriteShuffle>, VEX, VEX_L;
-defm VPMOVZXBW : SS41I_binop_rm_int16_y<0x30, "vpmovzxbw",
- int_x86_avx2_pmovzxbw,
- WriteShuffle>, VEX, VEX_L;
-defm VPMOVZXWD : SS41I_binop_rm_int16_y<0x33, "vpmovzxwd",
- int_x86_avx2_pmovzxwd,
- WriteShuffle>, VEX, VEX_L;
-defm VPMOVZXDQ : SS41I_binop_rm_int16_y<0x35, "vpmovzxdq",
- int_x86_avx2_pmovzxdq,
- WriteShuffle>, VEX, VEX_L;
-}
-
-defm PMOVSXBW : SS41I_binop_rm_int8<0x20, "pmovsxbw", int_x86_sse41_pmovsxbw,
- SSE_INTALU_ITINS_SHUFF_P>;
-defm PMOVSXWD : SS41I_binop_rm_int8<0x23, "pmovsxwd", int_x86_sse41_pmovsxwd,
- SSE_INTALU_ITINS_SHUFF_P>;
-defm PMOVSXDQ : SS41I_binop_rm_int8<0x25, "pmovsxdq", int_x86_sse41_pmovsxdq,
- SSE_INTALU_ITINS_SHUFF_P>;
-defm PMOVZXBW : SS41I_binop_rm_int8<0x30, "pmovzxbw", int_x86_sse41_pmovzxbw,
- SSE_INTALU_ITINS_SHUFF_P>;
-defm PMOVZXWD : SS41I_binop_rm_int8<0x33, "pmovzxwd", int_x86_sse41_pmovzxwd,
- SSE_INTALU_ITINS_SHUFF_P>;
-defm PMOVZXDQ : SS41I_binop_rm_int8<0x35, "pmovzxdq", int_x86_sse41_pmovzxdq,
- SSE_INTALU_ITINS_SHUFF_P>;
-
-let Predicates = [HasAVX] in {
- // Common patterns involving scalar load.
- def : Pat<(int_x86_sse41_pmovsxbw (vzmovl_v2i64 addr:$src)),
- (VPMOVSXBWrm addr:$src)>;
- def : Pat<(int_x86_sse41_pmovsxbw (vzload_v2i64 addr:$src)),
- (VPMOVSXBWrm addr:$src)>;
- def : Pat<(int_x86_sse41_pmovsxbw (bc_v16i8 (loadv2i64 addr:$src))),
- (VPMOVSXBWrm addr:$src)>;
-
- def : Pat<(int_x86_sse41_pmovsxwd (vzmovl_v2i64 addr:$src)),
- (VPMOVSXWDrm addr:$src)>;
- def : Pat<(int_x86_sse41_pmovsxwd (vzload_v2i64 addr:$src)),
- (VPMOVSXWDrm addr:$src)>;
- def : Pat<(int_x86_sse41_pmovsxwd (bc_v8i16 (loadv2i64 addr:$src))),
- (VPMOVSXWDrm addr:$src)>;
-
- def : Pat<(int_x86_sse41_pmovsxdq (vzmovl_v2i64 addr:$src)),
- (VPMOVSXDQrm addr:$src)>;
- def : Pat<(int_x86_sse41_pmovsxdq (vzload_v2i64 addr:$src)),
- (VPMOVSXDQrm addr:$src)>;
- def : Pat<(int_x86_sse41_pmovsxdq (bc_v4i32 (loadv2i64 addr:$src))),
- (VPMOVSXDQrm addr:$src)>;
-
- def : Pat<(int_x86_sse41_pmovzxbw (vzmovl_v2i64 addr:$src)),
- (VPMOVZXBWrm addr:$src)>;
- def : Pat<(int_x86_sse41_pmovzxbw (vzload_v2i64 addr:$src)),
- (VPMOVZXBWrm addr:$src)>;
- def : Pat<(int_x86_sse41_pmovzxbw (bc_v16i8 (loadv2i64 addr:$src))),
- (VPMOVZXBWrm addr:$src)>;
-
- def : Pat<(int_x86_sse41_pmovzxwd (vzmovl_v2i64 addr:$src)),
- (VPMOVZXWDrm addr:$src)>;
- def : Pat<(int_x86_sse41_pmovzxwd (vzload_v2i64 addr:$src)),
- (VPMOVZXWDrm addr:$src)>;
- def : Pat<(int_x86_sse41_pmovzxwd (bc_v8i16 (loadv2i64 addr:$src))),
- (VPMOVZXWDrm addr:$src)>;
-
- def : Pat<(int_x86_sse41_pmovzxdq (vzmovl_v2i64 addr:$src)),
- (VPMOVZXDQrm addr:$src)>;
- def : Pat<(int_x86_sse41_pmovzxdq (vzload_v2i64 addr:$src)),
- (VPMOVZXDQrm addr:$src)>;
- def : Pat<(int_x86_sse41_pmovzxdq (bc_v4i32 (loadv2i64 addr:$src))),
- (VPMOVZXDQrm addr:$src)>;
-}
-
-let Predicates = [UseSSE41] in {
- // Common patterns involving scalar load.
- def : Pat<(int_x86_sse41_pmovsxbw (vzmovl_v2i64 addr:$src)),
- (PMOVSXBWrm addr:$src)>;
- def : Pat<(int_x86_sse41_pmovsxbw (vzload_v2i64 addr:$src)),
- (PMOVSXBWrm addr:$src)>;
- def : Pat<(int_x86_sse41_pmovsxbw (bc_v16i8 (loadv2i64 addr:$src))),
- (PMOVSXBWrm addr:$src)>;
-
- def : Pat<(int_x86_sse41_pmovsxwd (vzmovl_v2i64 addr:$src)),
- (PMOVSXWDrm addr:$src)>;
- def : Pat<(int_x86_sse41_pmovsxwd (vzload_v2i64 addr:$src)),
- (PMOVSXWDrm addr:$src)>;
- def : Pat<(int_x86_sse41_pmovsxwd (bc_v8i16 (loadv2i64 addr:$src))),
- (PMOVSXWDrm addr:$src)>;
-
- def : Pat<(int_x86_sse41_pmovsxdq (vzmovl_v2i64 addr:$src)),
- (PMOVSXDQrm addr:$src)>;
- def : Pat<(int_x86_sse41_pmovsxdq (vzload_v2i64 addr:$src)),
- (PMOVSXDQrm addr:$src)>;
- def : Pat<(int_x86_sse41_pmovsxdq (bc_v4i32 (loadv2i64 addr:$src))),
- (PMOVSXDQrm addr:$src)>;
-
- def : Pat<(int_x86_sse41_pmovzxbw (vzmovl_v2i64 addr:$src)),
- (PMOVZXBWrm addr:$src)>;
- def : Pat<(int_x86_sse41_pmovzxbw (vzload_v2i64 addr:$src)),
- (PMOVZXBWrm addr:$src)>;
- def : Pat<(int_x86_sse41_pmovzxbw (bc_v16i8 (loadv2i64 addr:$src))),
- (PMOVZXBWrm addr:$src)>;
-
- def : Pat<(int_x86_sse41_pmovzxwd (vzmovl_v2i64 addr:$src)),
- (PMOVZXWDrm addr:$src)>;
- def : Pat<(int_x86_sse41_pmovzxwd (vzload_v2i64 addr:$src)),
- (PMOVZXWDrm addr:$src)>;
- def : Pat<(int_x86_sse41_pmovzxwd (bc_v8i16 (loadv2i64 addr:$src))),
- (PMOVZXWDrm addr:$src)>;
-
- def : Pat<(int_x86_sse41_pmovzxdq (vzmovl_v2i64 addr:$src)),
- (PMOVZXDQrm addr:$src)>;
- def : Pat<(int_x86_sse41_pmovzxdq (vzload_v2i64 addr:$src)),
- (PMOVZXDQrm addr:$src)>;
- def : Pat<(int_x86_sse41_pmovzxdq (bc_v4i32 (loadv2i64 addr:$src))),
- (PMOVZXDQrm addr:$src)>;
-}
-
-multiclass SS41I_binop_rm_int4<bits<8> opc, string OpcodeStr, Intrinsic IntId,
- OpndItins itins = DEFAULT_ITINS> {
- def rr : SS48I<opc, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
+multiclass SS41I_pmovx_rrrm<bits<8> opc, string OpcodeStr, X86MemOperand MemOp,
+ RegisterClass OutRC, RegisterClass InRC,
+ OpndItins itins> {
+ def rr : SS48I<opc, MRMSrcReg, (outs OutRC:$dst), (ins InRC:$src),
!strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"),
- [(set VR128:$dst, (IntId VR128:$src))], itins.rr>,
+ [], itins.rr>,
Sched<[itins.Sched]>;
- def rm : SS48I<opc, MRMSrcMem, (outs VR128:$dst), (ins i32mem:$src),
- !strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"),
- [(set VR128:$dst,
- (IntId (bitconvert (v4i32 (scalar_to_vector (loadi32 addr:$src))))))],
- itins.rm>, Sched<[itins.Sched.Folded]>;
-}
-
-multiclass SS41I_binop_rm_int8_y<bits<8> opc, string OpcodeStr,
- Intrinsic IntId, X86FoldableSchedWrite Sched> {
- def Yrr : SS48I<opc, MRMSrcReg, (outs VR256:$dst), (ins VR128:$src),
- !strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"),
- [(set VR256:$dst, (IntId VR128:$src))]>, Sched<[Sched]>;
-
- def Yrm : SS48I<opc, MRMSrcMem, (outs VR256:$dst), (ins i32mem:$src),
- !strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"),
- [(set VR256:$dst,
- (IntId (bitconvert (v2i64 (scalar_to_vector (loadi64 addr:$src))))))]>,
- Sched<[Sched.Folded]>;
-}
-
-let Predicates = [HasAVX] in {
-defm VPMOVSXBD : SS41I_binop_rm_int4<0x21, "vpmovsxbd", int_x86_sse41_pmovsxbd,
- DEFAULT_ITINS_SHUFFLESCHED>, VEX;
-defm VPMOVSXWQ : SS41I_binop_rm_int4<0x24, "vpmovsxwq", int_x86_sse41_pmovsxwq,
- DEFAULT_ITINS_SHUFFLESCHED>, VEX;
-defm VPMOVZXBD : SS41I_binop_rm_int4<0x31, "vpmovzxbd", int_x86_sse41_pmovzxbd,
- DEFAULT_ITINS_SHUFFLESCHED>, VEX;
-defm VPMOVZXWQ : SS41I_binop_rm_int4<0x34, "vpmovzxwq", int_x86_sse41_pmovzxwq,
- DEFAULT_ITINS_SHUFFLESCHED>, VEX;
-}
-
-let Predicates = [HasAVX2] in {
-defm VPMOVSXBD : SS41I_binop_rm_int8_y<0x21, "vpmovsxbd",
- int_x86_avx2_pmovsxbd, WriteShuffle>,
- VEX, VEX_L;
-defm VPMOVSXWQ : SS41I_binop_rm_int8_y<0x24, "vpmovsxwq",
- int_x86_avx2_pmovsxwq, WriteShuffle>,
- VEX, VEX_L;
-defm VPMOVZXBD : SS41I_binop_rm_int8_y<0x31, "vpmovzxbd",
- int_x86_avx2_pmovzxbd, WriteShuffle>,
- VEX, VEX_L;
-defm VPMOVZXWQ : SS41I_binop_rm_int8_y<0x34, "vpmovzxwq",
- int_x86_avx2_pmovzxwq, WriteShuffle>,
- VEX, VEX_L;
-}
-
-defm PMOVSXBD : SS41I_binop_rm_int4<0x21, "pmovsxbd", int_x86_sse41_pmovsxbd,
- SSE_INTALU_ITINS_SHUFF_P>;
-defm PMOVSXWQ : SS41I_binop_rm_int4<0x24, "pmovsxwq", int_x86_sse41_pmovsxwq,
- SSE_INTALU_ITINS_SHUFF_P>;
-defm PMOVZXBD : SS41I_binop_rm_int4<0x31, "pmovzxbd", int_x86_sse41_pmovzxbd,
- SSE_INTALU_ITINS_SHUFF_P>;
-defm PMOVZXWQ : SS41I_binop_rm_int4<0x34, "pmovzxwq", int_x86_sse41_pmovzxwq,
- SSE_INTALU_ITINS_SHUFF_P>;
-
-let Predicates = [HasAVX] in {
- // Common patterns involving scalar load
- def : Pat<(int_x86_sse41_pmovsxbd (vzmovl_v4i32 addr:$src)),
- (VPMOVSXBDrm addr:$src)>;
- def : Pat<(int_x86_sse41_pmovsxwq (vzmovl_v4i32 addr:$src)),
- (VPMOVSXWQrm addr:$src)>;
-
- def : Pat<(int_x86_sse41_pmovzxbd (vzmovl_v4i32 addr:$src)),
- (VPMOVZXBDrm addr:$src)>;
- def : Pat<(int_x86_sse41_pmovzxwq (vzmovl_v4i32 addr:$src)),
- (VPMOVZXWQrm addr:$src)>;
-}
-
-let Predicates = [UseSSE41] in {
- // Common patterns involving scalar load
- def : Pat<(int_x86_sse41_pmovsxbd (vzmovl_v4i32 addr:$src)),
- (PMOVSXBDrm addr:$src)>;
- def : Pat<(int_x86_sse41_pmovsxwq (vzmovl_v4i32 addr:$src)),
- (PMOVSXWQrm addr:$src)>;
-
- def : Pat<(int_x86_sse41_pmovzxbd (vzmovl_v4i32 addr:$src)),
- (PMOVZXBDrm addr:$src)>;
- def : Pat<(int_x86_sse41_pmovzxwq (vzmovl_v4i32 addr:$src)),
- (PMOVZXWQrm addr:$src)>;
-}
-
-multiclass SS41I_binop_rm_int2<bits<8> opc, string OpcodeStr, Intrinsic IntId,
- X86FoldableSchedWrite Sched> {
- def rr : SS48I<opc, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
- !strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"),
- [(set VR128:$dst, (IntId VR128:$src))]>, Sched<[Sched]>;
-
- // Expecting a i16 load any extended to i32 value.
- def rm : SS48I<opc, MRMSrcMem, (outs VR128:$dst), (ins i16mem:$src),
- !strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"),
- [(set VR128:$dst, (IntId (bitconvert
- (v4i32 (scalar_to_vector (loadi16_anyext addr:$src))))))]>,
- Sched<[Sched.Folded]>;
-}
-
-multiclass SS41I_binop_rm_int4_y<bits<8> opc, string OpcodeStr,
- Intrinsic IntId, X86FoldableSchedWrite Sched> {
- def Yrr : SS48I<opc, MRMSrcReg, (outs VR256:$dst), (ins VR128:$src),
+ def rm : SS48I<opc, MRMSrcMem, (outs OutRC:$dst), (ins MemOp:$src),
!strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"),
- [(set VR256:$dst, (IntId VR128:$src))]>, Sched<[Sched]>;
-
- // Expecting a i16 load any extended to i32 value.
- def Yrm : SS48I<opc, MRMSrcMem, (outs VR256:$dst), (ins i16mem:$src),
- !strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"),
- [(set VR256:$dst, (IntId (bitconvert
- (v4i32 (scalar_to_vector (loadi32 addr:$src))))))]>,
- Sched<[Sched.Folded]>;
-}
-
-let Predicates = [HasAVX] in {
-defm VPMOVSXBQ : SS41I_binop_rm_int2<0x22, "vpmovsxbq", int_x86_sse41_pmovsxbq,
- WriteShuffle>, VEX;
-defm VPMOVZXBQ : SS41I_binop_rm_int2<0x32, "vpmovzxbq", int_x86_sse41_pmovzxbq,
- WriteShuffle>, VEX;
-}
-let Predicates = [HasAVX2] in {
-defm VPMOVSXBQ : SS41I_binop_rm_int4_y<0x22, "vpmovsxbq", int_x86_avx2_pmovsxbq,
- WriteShuffle>, VEX, VEX_L;
-defm VPMOVZXBQ : SS41I_binop_rm_int4_y<0x32, "vpmovzxbq", int_x86_avx2_pmovzxbq,
- WriteShuffle>, VEX, VEX_L;
+ [],
+ itins.rm>, Sched<[itins.Sched.Folded]>;
}
-defm PMOVSXBQ : SS41I_binop_rm_int2<0x22, "pmovsxbq", int_x86_sse41_pmovsxbq,
- WriteShuffle>;
-defm PMOVZXBQ : SS41I_binop_rm_int2<0x32, "pmovzxbq", int_x86_sse41_pmovzxbq,
- WriteShuffle>;
-let Predicates = [HasAVX2] in {
- def : Pat<(v16i16 (X86vsext (v16i8 VR128:$src))), (VPMOVSXBWYrr VR128:$src)>;
- def : Pat<(v8i32 (X86vsext (v16i8 VR128:$src))), (VPMOVSXBDYrr VR128:$src)>;
- def : Pat<(v4i64 (X86vsext (v16i8 VR128:$src))), (VPMOVSXBQYrr VR128:$src)>;
-
- def : Pat<(v8i32 (X86vsext (v8i16 VR128:$src))), (VPMOVSXWDYrr VR128:$src)>;
- def : Pat<(v4i64 (X86vsext (v8i16 VR128:$src))), (VPMOVSXWQYrr VR128:$src)>;
-
- def : Pat<(v4i64 (X86vsext (v4i32 VR128:$src))), (VPMOVSXDQYrr VR128:$src)>;
-
- def : Pat<(v16i16 (X86vsext (v32i8 VR256:$src))),
- (VPMOVSXBWYrr (EXTRACT_SUBREG VR256:$src, sub_xmm))>;
- def : Pat<(v8i32 (X86vsext (v32i8 VR256:$src))),
- (VPMOVSXBDYrr (EXTRACT_SUBREG VR256:$src, sub_xmm))>;
- def : Pat<(v4i64 (X86vsext (v32i8 VR256:$src))),
- (VPMOVSXBQYrr (EXTRACT_SUBREG VR256:$src, sub_xmm))>;
-
- def : Pat<(v8i32 (X86vsext (v16i16 VR256:$src))),
- (VPMOVSXWDYrr (EXTRACT_SUBREG VR256:$src, sub_xmm))>;
- def : Pat<(v4i64 (X86vsext (v16i16 VR256:$src))),
- (VPMOVSXWQYrr (EXTRACT_SUBREG VR256:$src, sub_xmm))>;
-
- def : Pat<(v4i64 (X86vsext (v8i32 VR256:$src))),
- (VPMOVSXDQYrr (EXTRACT_SUBREG VR256:$src, sub_xmm))>;
-
- def : Pat<(v8i32 (X86vsext (v8i16 (bitconvert (v2i64 (load addr:$src)))))),
- (VPMOVSXWDYrm addr:$src)>;
- def : Pat<(v4i64 (X86vsext (v4i32 (bitconvert (v2i64 (load addr:$src)))))),
- (VPMOVSXDQYrm addr:$src)>;
-
- def : Pat<(v8i32 (X86vsext (v16i8 (bitconvert (v2i64
- (scalar_to_vector (loadi64 addr:$src))))))),
- (VPMOVSXBDYrm addr:$src)>;
- def : Pat<(v8i32 (X86vsext (v16i8 (bitconvert (v2f64
- (scalar_to_vector (loadf64 addr:$src))))))),
- (VPMOVSXBDYrm addr:$src)>;
-
- def : Pat<(v4i64 (X86vsext (v8i16 (bitconvert (v2i64
- (scalar_to_vector (loadi64 addr:$src))))))),
- (VPMOVSXWQYrm addr:$src)>;
- def : Pat<(v4i64 (X86vsext (v8i16 (bitconvert (v2f64
- (scalar_to_vector (loadf64 addr:$src))))))),
- (VPMOVSXWQYrm addr:$src)>;
-
- def : Pat<(v4i64 (X86vsext (v16i8 (bitconvert (v4i32
- (scalar_to_vector (loadi32 addr:$src))))))),
- (VPMOVSXBQYrm addr:$src)>;
-}
-
-let Predicates = [HasAVX] in {
- // Common patterns involving scalar load
- def : Pat<(int_x86_sse41_pmovsxbq
- (bitconvert (v4i32 (X86vzmovl
- (v4i32 (scalar_to_vector (loadi32 addr:$src))))))),
- (VPMOVSXBQrm addr:$src)>;
-
- def : Pat<(int_x86_sse41_pmovzxbq
- (bitconvert (v4i32 (X86vzmovl
- (v4i32 (scalar_to_vector (loadi32 addr:$src))))))),
- (VPMOVZXBQrm addr:$src)>;
-}
-
-let Predicates = [UseSSE41] in {
- def : Pat<(v8i16 (X86vsext (v16i8 VR128:$src))), (PMOVSXBWrr VR128:$src)>;
- def : Pat<(v4i32 (X86vsext (v16i8 VR128:$src))), (PMOVSXBDrr VR128:$src)>;
- def : Pat<(v2i64 (X86vsext (v16i8 VR128:$src))), (PMOVSXBQrr VR128:$src)>;
-
- def : Pat<(v4i32 (X86vsext (v8i16 VR128:$src))), (PMOVSXWDrr VR128:$src)>;
- def : Pat<(v2i64 (X86vsext (v8i16 VR128:$src))), (PMOVSXWQrr VR128:$src)>;
-
- def : Pat<(v2i64 (X86vsext (v4i32 VR128:$src))), (PMOVSXDQrr VR128:$src)>;
-
- // Common patterns involving scalar load
- def : Pat<(int_x86_sse41_pmovsxbq
- (bitconvert (v4i32 (X86vzmovl
- (v4i32 (scalar_to_vector (loadi32 addr:$src))))))),
- (PMOVSXBQrm addr:$src)>;
-
- def : Pat<(int_x86_sse41_pmovzxbq
- (bitconvert (v4i32 (X86vzmovl
- (v4i32 (scalar_to_vector (loadi32 addr:$src))))))),
- (PMOVZXBQrm addr:$src)>;
-
- def : Pat<(v4i32 (X86vsext (v8i16 (bitconvert (v2i64
- (scalar_to_vector (loadi64 addr:$src))))))),
- (PMOVSXWDrm addr:$src)>;
- def : Pat<(v4i32 (X86vsext (v8i16 (bitconvert (v2f64
- (scalar_to_vector (loadf64 addr:$src))))))),
- (PMOVSXWDrm addr:$src)>;
- def : Pat<(v4i32 (X86vsext (v16i8 (bitconvert (v4i32
- (scalar_to_vector (loadi32 addr:$src))))))),
- (PMOVSXBDrm addr:$src)>;
- def : Pat<(v2i64 (X86vsext (v8i16 (bitconvert (v4i32
- (scalar_to_vector (loadi32 addr:$src))))))),
- (PMOVSXWQrm addr:$src)>;
- def : Pat<(v2i64 (X86vsext (v16i8 (bitconvert (v4i32
- (scalar_to_vector (extloadi32i16 addr:$src))))))),
- (PMOVSXBQrm addr:$src)>;
- def : Pat<(v2i64 (X86vsext (v4i32 (bitconvert (v2i64
- (scalar_to_vector (loadi64 addr:$src))))))),
- (PMOVSXDQrm addr:$src)>;
- def : Pat<(v2i64 (X86vsext (v4i32 (bitconvert (v2f64
- (scalar_to_vector (loadf64 addr:$src))))))),
- (PMOVSXDQrm addr:$src)>;
- def : Pat<(v8i16 (X86vsext (v16i8 (bitconvert (v2i64
- (scalar_to_vector (loadi64 addr:$src))))))),
- (PMOVSXBWrm addr:$src)>;
- def : Pat<(v8i16 (X86vsext (v16i8 (bitconvert (v2f64
- (scalar_to_vector (loadf64 addr:$src))))))),
- (PMOVSXBWrm addr:$src)>;
+multiclass SS41I_pmovx_rm_all<bits<8> opc, string OpcodeStr,
+ X86MemOperand MemOp, X86MemOperand MemYOp,
+ OpndItins SSEItins, OpndItins AVXItins,
+ OpndItins AVX2Itins> {
+ defm NAME : SS41I_pmovx_rrrm<opc, OpcodeStr, MemOp, VR128, VR128, SSEItins>;
+ let Predicates = [HasAVX] in
+ defm V#NAME : SS41I_pmovx_rrrm<opc, !strconcat("v", OpcodeStr), MemOp,
+ VR128, VR128, AVXItins>, VEX;
+ let Predicates = [HasAVX2] in
+ defm V#NAME#Y : SS41I_pmovx_rrrm<opc, !strconcat("v", OpcodeStr), MemYOp,
+ VR256, VR128, AVX2Itins>, VEX, VEX_L;
+}
+
+multiclass SS41I_pmovx_rm<bits<8> opc, string OpcodeStr,
+ X86MemOperand MemOp, X86MemOperand MemYOp> {
+ defm PMOVSX#NAME : SS41I_pmovx_rm_all<opc, !strconcat("pmovsx", OpcodeStr),
+ MemOp, MemYOp,
+ SSE_INTALU_ITINS_SHUFF_P,
+ DEFAULT_ITINS_SHUFFLESCHED,
+ DEFAULT_ITINS_SHUFFLESCHED>;
+ defm PMOVZX#NAME : SS41I_pmovx_rm_all<!add(opc, 0x10),
+ !strconcat("pmovzx", OpcodeStr),
+ MemOp, MemYOp,
+ SSE_INTALU_ITINS_SHUFF_P,
+ DEFAULT_ITINS_SHUFFLESCHED,
+ DEFAULT_ITINS_SHUFFLESCHED>;
+}
+
+defm BW : SS41I_pmovx_rm<0x20, "bw", i64mem, i128mem>;
+defm WD : SS41I_pmovx_rm<0x23, "wd", i64mem, i128mem>;
+defm DQ : SS41I_pmovx_rm<0x25, "dq", i64mem, i128mem>;
+
+defm BD : SS41I_pmovx_rm<0x21, "bd", i32mem, i64mem>;
+defm WQ : SS41I_pmovx_rm<0x24, "wq", i32mem, i64mem>;
+
+defm BQ : SS41I_pmovx_rm<0x22, "bq", i16mem, i32mem>;
+
+// AVX2 Patterns
+multiclass SS41I_pmovx_avx2_patterns<string OpcPrefix, SDNode ExtOp> {
+ // Register-Register patterns
+ def : Pat<(v16i16 (ExtOp (v16i8 VR128:$src))),
+ (!cast<I>(OpcPrefix#BWYrr) VR128:$src)>;
+ def : Pat<(v8i32 (ExtOp (v16i8 VR128:$src))),
+ (!cast<I>(OpcPrefix#BDYrr) VR128:$src)>;
+ def : Pat<(v4i64 (ExtOp (v16i8 VR128:$src))),
+ (!cast<I>(OpcPrefix#BQYrr) VR128:$src)>;
+
+ def : Pat<(v8i32 (ExtOp (v8i16 VR128:$src))),
+ (!cast<I>(OpcPrefix#WDYrr) VR128:$src)>;
+ def : Pat<(v4i64 (ExtOp (v8i16 VR128:$src))),
+ (!cast<I>(OpcPrefix#WQYrr) VR128:$src)>;
+
+ def : Pat<(v4i64 (ExtOp (v4i32 VR128:$src))),
+ (!cast<I>(OpcPrefix#DQYrr) VR128:$src)>;
+
+ // On AVX2, we also support 256bit inputs.
+ // FIXME: remove these patterns when the old shuffle lowering goes away.
+ def : Pat<(v16i16 (ExtOp (v32i8 VR256:$src))),
+ (!cast<I>(OpcPrefix#BWYrr) (EXTRACT_SUBREG VR256:$src, sub_xmm))>;
+ def : Pat<(v8i32 (ExtOp (v32i8 VR256:$src))),
+ (!cast<I>(OpcPrefix#BDYrr) (EXTRACT_SUBREG VR256:$src, sub_xmm))>;
+ def : Pat<(v4i64 (ExtOp (v32i8 VR256:$src))),
+ (!cast<I>(OpcPrefix#BQYrr) (EXTRACT_SUBREG VR256:$src, sub_xmm))>;
+
+ def : Pat<(v8i32 (ExtOp (v16i16 VR256:$src))),
+ (!cast<I>(OpcPrefix#WDYrr) (EXTRACT_SUBREG VR256:$src, sub_xmm))>;
+ def : Pat<(v4i64 (ExtOp (v16i16 VR256:$src))),
+ (!cast<I>(OpcPrefix#WQYrr) (EXTRACT_SUBREG VR256:$src, sub_xmm))>;
+
+ def : Pat<(v4i64 (ExtOp (v8i32 VR256:$src))),
+ (!cast<I>(OpcPrefix#DQYrr) (EXTRACT_SUBREG VR256:$src, sub_xmm))>;
+
+ // AVX2 Register-Memory patterns
+ def : Pat<(v16i16 (ExtOp (bc_v16i8 (loadv2i64 addr:$src)))),
+ (!cast<I>(OpcPrefix#BWYrm) addr:$src)>;
+ def : Pat<(v16i16 (ExtOp (v16i8 (vzmovl_v2i64 addr:$src)))),
+ (!cast<I>(OpcPrefix#BWYrm) addr:$src)>;
+ def : Pat<(v16i16 (ExtOp (v16i8 (vzload_v2i64 addr:$src)))),
+ (!cast<I>(OpcPrefix#BWYrm) addr:$src)>;
+ def : Pat<(v16i16 (ExtOp (bc_v16i8 (loadv2i64 addr:$src)))),
+ (!cast<I>(OpcPrefix#BWYrm) addr:$src)>;
+
+ def : Pat<(v8i32 (ExtOp (bc_v16i8 (v2i64 (scalar_to_vector (loadi64 addr:$src)))))),
+ (!cast<I>(OpcPrefix#BDYrm) addr:$src)>;
+ def : Pat<(v8i32 (ExtOp (v16i8 (vzmovl_v2i64 addr:$src)))),
+ (!cast<I>(OpcPrefix#BDYrm) addr:$src)>;
+ def : Pat<(v8i32 (ExtOp (v16i8 (vzload_v2i64 addr:$src)))),
+ (!cast<I>(OpcPrefix#BDYrm) addr:$src)>;
+ def : Pat<(v8i32 (ExtOp (bc_v16i8 (loadv2i64 addr:$src)))),
+ (!cast<I>(OpcPrefix#BDYrm) addr:$src)>;
+
+ def : Pat<(v4i64 (ExtOp (bc_v16i8 (v4i32 (scalar_to_vector (loadi32 addr:$src)))))),
+ (!cast<I>(OpcPrefix#BQYrm) addr:$src)>;
+ def : Pat<(v4i64 (ExtOp (v16i8 (vzmovl_v4i32 addr:$src)))),
+ (!cast<I>(OpcPrefix#BQYrm) addr:$src)>;
+ def : Pat<(v4i64 (ExtOp (v16i8 (vzload_v2i64 addr:$src)))),
+ (!cast<I>(OpcPrefix#BQYrm) addr:$src)>;
+ def : Pat<(v4i64 (ExtOp (bc_v16i8 (loadv2i64 addr:$src)))),
+ (!cast<I>(OpcPrefix#BQYrm) addr:$src)>;
+
+ def : Pat<(v8i32 (ExtOp (bc_v8i16 (loadv2i64 addr:$src)))),
+ (!cast<I>(OpcPrefix#WDYrm) addr:$src)>;
+ def : Pat<(v8i32 (ExtOp (v8i16 (vzmovl_v2i64 addr:$src)))),
+ (!cast<I>(OpcPrefix#WDYrm) addr:$src)>;
+ def : Pat<(v8i32 (ExtOp (v8i16 (vzload_v2i64 addr:$src)))),
+ (!cast<I>(OpcPrefix#WDYrm) addr:$src)>;
+ def : Pat<(v8i32 (ExtOp (bc_v8i16 (loadv2i64 addr:$src)))),
+ (!cast<I>(OpcPrefix#WDYrm) addr:$src)>;
+
+ def : Pat<(v4i64 (ExtOp (bc_v8i16 (v2i64 (scalar_to_vector (loadi64 addr:$src)))))),
+ (!cast<I>(OpcPrefix#WQYrm) addr:$src)>;
+ def : Pat<(v4i64 (ExtOp (v8i16 (vzmovl_v2i64 addr:$src)))),
+ (!cast<I>(OpcPrefix#WQYrm) addr:$src)>;
+ def : Pat<(v4i64 (ExtOp (v8i16 (vzload_v2i64 addr:$src)))),
+ (!cast<I>(OpcPrefix#WQYrm) addr:$src)>;
+ def : Pat<(v4i64 (ExtOp (bc_v8i16 (loadv2i64 addr:$src)))),
+ (!cast<I>(OpcPrefix#WQYrm) addr:$src)>;
+
+ def : Pat<(v4i64 (ExtOp (bc_v4i32 (loadv2i64 addr:$src)))),
+ (!cast<I>(OpcPrefix#DQYrm) addr:$src)>;
+ def : Pat<(v4i64 (ExtOp (v4i32 (vzmovl_v2i64 addr:$src)))),
+ (!cast<I>(OpcPrefix#DQYrm) addr:$src)>;
+ def : Pat<(v4i64 (ExtOp (v4i32 (vzload_v2i64 addr:$src)))),
+ (!cast<I>(OpcPrefix#DQYrm) addr:$src)>;
+ def : Pat<(v4i64 (ExtOp (bc_v4i32 (loadv2i64 addr:$src)))),
+ (!cast<I>(OpcPrefix#DQYrm) addr:$src)>;
}
let Predicates = [HasAVX2] in {
- def : Pat<(v16i16 (X86vzext (v16i8 VR128:$src))), (VPMOVZXBWYrr VR128:$src)>;
- def : Pat<(v8i32 (X86vzext (v16i8 VR128:$src))), (VPMOVZXBDYrr VR128:$src)>;
- def : Pat<(v4i64 (X86vzext (v16i8 VR128:$src))), (VPMOVZXBQYrr VR128:$src)>;
-
- def : Pat<(v8i32 (X86vzext (v8i16 VR128:$src))), (VPMOVZXWDYrr VR128:$src)>;
- def : Pat<(v4i64 (X86vzext (v8i16 VR128:$src))), (VPMOVZXWQYrr VR128:$src)>;
-
- def : Pat<(v4i64 (X86vzext (v4i32 VR128:$src))), (VPMOVZXDQYrr VR128:$src)>;
-
- def : Pat<(v16i16 (X86vzext (v32i8 VR256:$src))),
- (VPMOVZXBWYrr (EXTRACT_SUBREG VR256:$src, sub_xmm))>;
- def : Pat<(v8i32 (X86vzext (v32i8 VR256:$src))),
- (VPMOVZXBDYrr (EXTRACT_SUBREG VR256:$src, sub_xmm))>;
- def : Pat<(v4i64 (X86vzext (v32i8 VR256:$src))),
- (VPMOVZXBQYrr (EXTRACT_SUBREG VR256:$src, sub_xmm))>;
-
- def : Pat<(v8i32 (X86vzext (v16i16 VR256:$src))),
- (VPMOVZXWDYrr (EXTRACT_SUBREG VR256:$src, sub_xmm))>;
- def : Pat<(v4i64 (X86vzext (v16i16 VR256:$src))),
- (VPMOVZXWQYrr (EXTRACT_SUBREG VR256:$src, sub_xmm))>;
-
- def : Pat<(v4i64 (X86vzext (v8i32 VR256:$src))),
- (VPMOVZXDQYrr (EXTRACT_SUBREG VR256:$src, sub_xmm))>;
+ defm : SS41I_pmovx_avx2_patterns<"VPMOVSX", X86vsext>;
+ defm : SS41I_pmovx_avx2_patterns<"VPMOVZX", X86vzext>;
+}
+
+// SSE4.1/AVX patterns.
+multiclass SS41I_pmovx_patterns<string OpcPrefix, SDNode ExtOp,
+ PatFrag ExtLoad16> {
+ def : Pat<(v8i16 (ExtOp (v16i8 VR128:$src))),
+ (!cast<I>(OpcPrefix#BWrr) VR128:$src)>;
+ def : Pat<(v4i32 (ExtOp (v16i8 VR128:$src))),
+ (!cast<I>(OpcPrefix#BDrr) VR128:$src)>;
+ def : Pat<(v2i64 (ExtOp (v16i8 VR128:$src))),
+ (!cast<I>(OpcPrefix#BQrr) VR128:$src)>;
+
+ def : Pat<(v4i32 (ExtOp (v8i16 VR128:$src))),
+ (!cast<I>(OpcPrefix#WDrr) VR128:$src)>;
+ def : Pat<(v2i64 (ExtOp (v8i16 VR128:$src))),
+ (!cast<I>(OpcPrefix#WQrr) VR128:$src)>;
+
+ def : Pat<(v2i64 (ExtOp (v4i32 VR128:$src))),
+ (!cast<I>(OpcPrefix#DQrr) VR128:$src)>;
+
+ def : Pat<(v8i16 (ExtOp (bc_v16i8 (v2i64 (scalar_to_vector (loadi64 addr:$src)))))),
+ (!cast<I>(OpcPrefix#BWrm) addr:$src)>;
+ def : Pat<(v8i16 (ExtOp (bc_v16i8 (v2f64 (scalar_to_vector (loadf64 addr:$src)))))),
+ (!cast<I>(OpcPrefix#BWrm) addr:$src)>;
+ def : Pat<(v8i16 (ExtOp (v16i8 (vzmovl_v2i64 addr:$src)))),
+ (!cast<I>(OpcPrefix#BWrm) addr:$src)>;
+ def : Pat<(v8i16 (ExtOp (v16i8 (vzload_v2i64 addr:$src)))),
+ (!cast<I>(OpcPrefix#BWrm) addr:$src)>;
+ def : Pat<(v8i16 (ExtOp (bc_v16i8 (loadv2i64 addr:$src)))),
+ (!cast<I>(OpcPrefix#BWrm) addr:$src)>;
+
+ def : Pat<(v4i32 (ExtOp (bc_v16i8 (v4i32 (scalar_to_vector (loadi32 addr:$src)))))),
+ (!cast<I>(OpcPrefix#BDrm) addr:$src)>;
+ def : Pat<(v4i32 (ExtOp (v16i8 (vzmovl_v4i32 addr:$src)))),
+ (!cast<I>(OpcPrefix#BDrm) addr:$src)>;
+ def : Pat<(v4i32 (ExtOp (v16i8 (vzload_v2i64 addr:$src)))),
+ (!cast<I>(OpcPrefix#BDrm) addr:$src)>;
+ def : Pat<(v4i32 (ExtOp (bc_v16i8 (loadv2i64 addr:$src)))),
+ (!cast<I>(OpcPrefix#BDrm) addr:$src)>;
+
+ def : Pat<(v2i64 (ExtOp (bc_v16i8 (v4i32 (scalar_to_vector (ExtLoad16 addr:$src)))))),
+ (!cast<I>(OpcPrefix#BQrm) addr:$src)>;
+ def : Pat<(v2i64 (ExtOp (v16i8 (vzmovl_v4i32 addr:$src)))),
+ (!cast<I>(OpcPrefix#BQrm) addr:$src)>;
+ def : Pat<(v2i64 (ExtOp (v16i8 (vzload_v2i64 addr:$src)))),
+ (!cast<I>(OpcPrefix#BQrm) addr:$src)>;
+ def : Pat<(v2i64 (ExtOp (bc_v16i8 (loadv2i64 addr:$src)))),
+ (!cast<I>(OpcPrefix#BQrm) addr:$src)>;
+
+ def : Pat<(v4i32 (ExtOp (bc_v8i16 (v2i64 (scalar_to_vector (loadi64 addr:$src)))))),
+ (!cast<I>(OpcPrefix#WDrm) addr:$src)>;
+ def : Pat<(v4i32 (ExtOp (bc_v8i16 (v2f64 (scalar_to_vector (loadf64 addr:$src)))))),
+ (!cast<I>(OpcPrefix#WDrm) addr:$src)>;
+ def : Pat<(v4i32 (ExtOp (v8i16 (vzmovl_v2i64 addr:$src)))),
+ (!cast<I>(OpcPrefix#WDrm) addr:$src)>;
+ def : Pat<(v4i32 (ExtOp (v8i16 (vzload_v2i64 addr:$src)))),
+ (!cast<I>(OpcPrefix#WDrm) addr:$src)>;
+ def : Pat<(v4i32 (ExtOp (bc_v8i16 (loadv2i64 addr:$src)))),
+ (!cast<I>(OpcPrefix#WDrm) addr:$src)>;
+
+ def : Pat<(v2i64 (ExtOp (bc_v8i16 (v4i32 (scalar_to_vector (loadi32 addr:$src)))))),
+ (!cast<I>(OpcPrefix#WQrm) addr:$src)>;
+ def : Pat<(v2i64 (ExtOp (v8i16 (vzmovl_v4i32 addr:$src)))),
+ (!cast<I>(OpcPrefix#WQrm) addr:$src)>;
+ def : Pat<(v2i64 (ExtOp (v8i16 (vzload_v2i64 addr:$src)))),
+ (!cast<I>(OpcPrefix#WQrm) addr:$src)>;
+ def : Pat<(v2i64 (ExtOp (bc_v8i16 (loadv2i64 addr:$src)))),
+ (!cast<I>(OpcPrefix#WQrm) addr:$src)>;
+
+ def : Pat<(v2i64 (ExtOp (bc_v4i32 (v2i64 (scalar_to_vector (loadi64 addr:$src)))))),
+ (!cast<I>(OpcPrefix#DQrm) addr:$src)>;
+ def : Pat<(v2i64 (ExtOp (bc_v4i32 (v2f64 (scalar_to_vector (loadf64 addr:$src)))))),
+ (!cast<I>(OpcPrefix#DQrm) addr:$src)>;
+ def : Pat<(v2i64 (ExtOp (v4i32 (vzmovl_v2i64 addr:$src)))),
+ (!cast<I>(OpcPrefix#DQrm) addr:$src)>;
+ def : Pat<(v2i64 (ExtOp (v4i32 (vzload_v2i64 addr:$src)))),
+ (!cast<I>(OpcPrefix#DQrm) addr:$src)>;
+ def : Pat<(v2i64 (ExtOp (bc_v4i32 (loadv2i64 addr:$src)))),
+ (!cast<I>(OpcPrefix#DQrm) addr:$src)>;
}
let Predicates = [HasAVX] in {
- def : Pat<(v8i16 (X86vzext (v16i8 VR128:$src))), (VPMOVZXBWrr VR128:$src)>;
- def : Pat<(v4i32 (X86vzext (v16i8 VR128:$src))), (VPMOVZXBDrr VR128:$src)>;
- def : Pat<(v2i64 (X86vzext (v16i8 VR128:$src))), (VPMOVZXBQrr VR128:$src)>;
-
- def : Pat<(v4i32 (X86vzext (v8i16 VR128:$src))), (VPMOVZXWDrr VR128:$src)>;
- def : Pat<(v2i64 (X86vzext (v8i16 VR128:$src))), (VPMOVZXWQrr VR128:$src)>;
-
- def : Pat<(v2i64 (X86vzext (v4i32 VR128:$src))), (VPMOVZXDQrr VR128:$src)>;
-
- def : Pat<(v8i16 (X86vzext (v16i8 (bitconvert (v2i64 (scalar_to_vector (loadi64 addr:$src))))))),
- (VPMOVZXBWrm addr:$src)>;
- def : Pat<(v8i16 (X86vzext (v16i8 (bitconvert (v2f64 (scalar_to_vector (loadf64 addr:$src))))))),
- (VPMOVZXBWrm addr:$src)>;
- def : Pat<(v4i32 (X86vzext (v16i8 (bitconvert (v4i32 (scalar_to_vector (loadi32 addr:$src))))))),
- (VPMOVZXBDrm addr:$src)>;
- def : Pat<(v2i64 (X86vzext (v16i8 (bitconvert (v4i32 (scalar_to_vector (loadi16_anyext addr:$src))))))),
- (VPMOVZXBQrm addr:$src)>;
-
- def : Pat<(v4i32 (X86vzext (v8i16 (bitconvert (v2i64 (scalar_to_vector (loadi64 addr:$src))))))),
- (VPMOVZXWDrm addr:$src)>;
- def : Pat<(v4i32 (X86vzext (v8i16 (bitconvert (v2f64 (scalar_to_vector (loadf64 addr:$src))))))),
- (VPMOVZXWDrm addr:$src)>;
- def : Pat<(v2i64 (X86vzext (v8i16 (bitconvert (v4i32 (scalar_to_vector (loadi32 addr:$src))))))),
- (VPMOVZXWQrm addr:$src)>;
-
- def : Pat<(v2i64 (X86vzext (v4i32 (bitconvert (v2i64 (scalar_to_vector (loadi64 addr:$src))))))),
- (VPMOVZXDQrm addr:$src)>;
- def : Pat<(v2i64 (X86vzext (v4i32 (bitconvert (v2f64 (scalar_to_vector (loadf64 addr:$src))))))),
- (VPMOVZXDQrm addr:$src)>;
- def : Pat<(v2i64 (X86vzext (v4i32 (bitconvert (v2i64 (X86vzload addr:$src)))))),
- (VPMOVZXDQrm addr:$src)>;
-
- def : Pat<(v8i16 (X86vsext (v16i8 VR128:$src))), (VPMOVSXBWrr VR128:$src)>;
- def : Pat<(v4i32 (X86vsext (v16i8 VR128:$src))), (VPMOVSXBDrr VR128:$src)>;
- def : Pat<(v2i64 (X86vsext (v16i8 VR128:$src))), (VPMOVSXBQrr VR128:$src)>;
-
- def : Pat<(v4i32 (X86vsext (v8i16 VR128:$src))), (VPMOVSXWDrr VR128:$src)>;
- def : Pat<(v2i64 (X86vsext (v8i16 VR128:$src))), (VPMOVSXWQrr VR128:$src)>;
-
- def : Pat<(v2i64 (X86vsext (v4i32 VR128:$src))), (VPMOVSXDQrr VR128:$src)>;
-
- def : Pat<(v4i32 (X86vsext (v8i16 (bitconvert (v2i64
- (scalar_to_vector (loadi64 addr:$src))))))),
- (VPMOVSXWDrm addr:$src)>;
- def : Pat<(v2i64 (X86vsext (v4i32 (bitconvert (v2i64
- (scalar_to_vector (loadi64 addr:$src))))))),
- (VPMOVSXDQrm addr:$src)>;
- def : Pat<(v4i32 (X86vsext (v8i16 (bitconvert (v2f64
- (scalar_to_vector (loadf64 addr:$src))))))),
- (VPMOVSXWDrm addr:$src)>;
- def : Pat<(v2i64 (X86vsext (v4i32 (bitconvert (v2f64
- (scalar_to_vector (loadf64 addr:$src))))))),
- (VPMOVSXDQrm addr:$src)>;
- def : Pat<(v8i16 (X86vsext (v16i8 (bitconvert (v2i64
- (scalar_to_vector (loadi64 addr:$src))))))),
- (VPMOVSXBWrm addr:$src)>;
- def : Pat<(v8i16 (X86vsext (v16i8 (bitconvert (v2f64
- (scalar_to_vector (loadf64 addr:$src))))))),
- (VPMOVSXBWrm addr:$src)>;
-
- def : Pat<(v4i32 (X86vsext (v16i8 (bitconvert (v4i32
- (scalar_to_vector (loadi32 addr:$src))))))),
- (VPMOVSXBDrm addr:$src)>;
- def : Pat<(v2i64 (X86vsext (v8i16 (bitconvert (v4i32
- (scalar_to_vector (loadi32 addr:$src))))))),
- (VPMOVSXWQrm addr:$src)>;
- def : Pat<(v2i64 (X86vsext (v16i8 (bitconvert (v4i32
- (scalar_to_vector (extloadi32i16 addr:$src))))))),
- (VPMOVSXBQrm addr:$src)>;
+ defm : SS41I_pmovx_patterns<"VPMOVSX", X86vsext, extloadi32i16>;
+ defm : SS41I_pmovx_patterns<"VPMOVZX", X86vzext, loadi16_anyext>;
}
let Predicates = [UseSSE41] in {
- def : Pat<(v8i16 (X86vzext (v16i8 VR128:$src))), (PMOVZXBWrr VR128:$src)>;
- def : Pat<(v4i32 (X86vzext (v16i8 VR128:$src))), (PMOVZXBDrr VR128:$src)>;
- def : Pat<(v2i64 (X86vzext (v16i8 VR128:$src))), (PMOVZXBQrr VR128:$src)>;
-
- def : Pat<(v4i32 (X86vzext (v8i16 VR128:$src))), (PMOVZXWDrr VR128:$src)>;
- def : Pat<(v2i64 (X86vzext (v8i16 VR128:$src))), (PMOVZXWQrr VR128:$src)>;
-
- def : Pat<(v2i64 (X86vzext (v4i32 VR128:$src))), (PMOVZXDQrr VR128:$src)>;
-
- def : Pat<(v8i16 (X86vzext (v16i8 (bitconvert (v2i64 (scalar_to_vector (loadi64 addr:$src))))))),
- (PMOVZXBWrm addr:$src)>;
- def : Pat<(v8i16 (X86vzext (v16i8 (bitconvert (v2f64 (scalar_to_vector (loadf64 addr:$src))))))),
- (PMOVZXBWrm addr:$src)>;
- def : Pat<(v4i32 (X86vzext (v16i8 (bitconvert (v4i32 (scalar_to_vector (loadi32 addr:$src))))))),
- (PMOVZXBDrm addr:$src)>;
- def : Pat<(v2i64 (X86vzext (v16i8 (bitconvert (v4i32 (scalar_to_vector (loadi16_anyext addr:$src))))))),
- (PMOVZXBQrm addr:$src)>;
-
- def : Pat<(v4i32 (X86vzext (v8i16 (bitconvert (v2i64 (scalar_to_vector (loadi64 addr:$src))))))),
- (PMOVZXWDrm addr:$src)>;
- def : Pat<(v4i32 (X86vzext (v8i16 (bitconvert (v2f64 (scalar_to_vector (loadf64 addr:$src))))))),
- (PMOVZXWDrm addr:$src)>;
- def : Pat<(v2i64 (X86vzext (v8i16 (bitconvert (v4i32 (scalar_to_vector (loadi32 addr:$src))))))),
- (PMOVZXWQrm addr:$src)>;
-
- def : Pat<(v2i64 (X86vzext (v4i32 (bitconvert (v2i64 (scalar_to_vector (loadi64 addr:$src))))))),
- (PMOVZXDQrm addr:$src)>;
- def : Pat<(v2i64 (X86vzext (v4i32 (bitconvert (v2f64 (scalar_to_vector (loadf64 addr:$src))))))),
- (PMOVZXDQrm addr:$src)>;
- def : Pat<(v2i64 (X86vzext (v4i32 (bitconvert (v2i64 (X86vzload addr:$src)))))),
- (PMOVZXDQrm addr:$src)>;
+ defm : SS41I_pmovx_patterns<"PMOVSX", X86vsext, extloadi32i16>;
+ defm : SS41I_pmovx_patterns<"PMOVZX", X86vzext, loadi16_anyext>;
}
//===----------------------------------------------------------------------===//
[(set GR32orGR64:$dst, (X86pextrb (v16i8 VR128:$src1),
imm:$src2))]>,
Sched<[WriteShuffle]>;
- let neverHasSideEffects = 1, mayStore = 1,
+ let hasSideEffects = 0, mayStore = 1,
SchedRW = [WriteShuffleLd, WriteRMW] in
def mr : SS4AIi8<opc, MRMDestMem, (outs),
(ins i8mem:$dst, VR128:$src1, i32i8imm:$src2),
"\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
[]>, Sched<[WriteShuffle]>;
- let neverHasSideEffects = 1, mayStore = 1,
+ let hasSideEffects = 0, mayStore = 1,
SchedRW = [WriteShuffleLd, WriteRMW] in
def mr : SS4AIi8<opc, MRMDestMem, (outs),
(ins i16mem:$dst, VR128:$src1, i32i8imm:$src2),
Sched<[itins.Sched.Folded, ReadAfterLd]>;
}
-let Predicates = [HasAVX] in {
+let Predicates = [HasAVX, NoVLX] in {
let isCommutable = 0 in
defm VPMINSB : SS48I_binop_rm<0x38, "vpminsb", X86smin, v16i8, VR128,
loadv2i64, i128mem, 0, SSE_INTALU_ITINS_P>,
SSE_INTMUL_ITINS_P, 1, 0>, VEX_4V;
}
-let Predicates = [HasAVX2] in {
+let Predicates = [HasAVX2, NoVLX] in {
let isCommutable = 0 in
defm VPMINSBY : SS48I_binop_rm<0x38, "vpminsb", X86smin, v32i8, VR256,
loadv4i64, i256mem, 0, SSE_INTALU_ITINS_P>,
SSE_INTMUL_ITINS_P, 1>;
}
-let Predicates = [HasAVX] in {
+let Predicates = [HasAVX, NoVLX] in {
defm VPMULLD : SS48I_binop_rm<0x40, "vpmulld", mul, v4i32, VR128,
memopv2i64, i128mem, 0, SSE_PMULLD_ITINS>,
VEX_4V;
let Predicates = [HasAVX] in {
let isCommutable = 0 in {
- let ExeDomain = SSEPackedSingle in {
- defm VBLENDPS : SS41I_binop_rmi_int<0x0C, "vblendps", int_x86_sse41_blendps,
- VR128, loadv4f32, f128mem, 0,
- DEFAULT_ITINS_FBLENDSCHED>, VEX_4V;
- defm VBLENDPSY : SS41I_binop_rmi_int<0x0C, "vblendps",
- int_x86_avx_blend_ps_256, VR256, loadv8f32,
- f256mem, 0, DEFAULT_ITINS_FBLENDSCHED>,
- VEX_4V, VEX_L;
- }
- let ExeDomain = SSEPackedDouble in {
- defm VBLENDPD : SS41I_binop_rmi_int<0x0D, "vblendpd", int_x86_sse41_blendpd,
- VR128, loadv2f64, f128mem, 0,
- DEFAULT_ITINS_FBLENDSCHED>, VEX_4V;
- defm VBLENDPDY : SS41I_binop_rmi_int<0x0D, "vblendpd",
- int_x86_avx_blend_pd_256,VR256, loadv4f64,
- f256mem, 0, DEFAULT_ITINS_FBLENDSCHED>,
- VEX_4V, VEX_L;
- }
+ defm VMPSADBW : SS41I_binop_rmi_int<0x42, "vmpsadbw", int_x86_sse41_mpsadbw,
+ VR128, loadv2i64, i128mem, 0,
+ DEFAULT_ITINS_MPSADSCHED>, VEX_4V;
+ }
+
+ let ExeDomain = SSEPackedSingle in {
+ defm VBLENDPS : SS41I_binop_rmi_int<0x0C, "vblendps", int_x86_sse41_blendps,
+ VR128, loadv4f32, f128mem, 0,
+ DEFAULT_ITINS_FBLENDSCHED>, VEX_4V;
+ defm VBLENDPSY : SS41I_binop_rmi_int<0x0C, "vblendps",
+ int_x86_avx_blend_ps_256, VR256, loadv8f32,
+ f256mem, 0, DEFAULT_ITINS_FBLENDSCHED>,
+ VEX_4V, VEX_L;
+ }
+ let ExeDomain = SSEPackedDouble in {
+ defm VBLENDPD : SS41I_binop_rmi_int<0x0D, "vblendpd", int_x86_sse41_blendpd,
+ VR128, loadv2f64, f128mem, 0,
+ DEFAULT_ITINS_FBLENDSCHED>, VEX_4V;
+ defm VBLENDPDY : SS41I_binop_rmi_int<0x0D, "vblendpd",
+ int_x86_avx_blend_pd_256,VR256, loadv4f64,
+ f256mem, 0, DEFAULT_ITINS_FBLENDSCHED>,
+ VEX_4V, VEX_L;
+ }
defm VPBLENDW : SS41I_binop_rmi_int<0x0E, "vpblendw", int_x86_sse41_pblendw,
VR128, loadv2i64, i128mem, 0,
DEFAULT_ITINS_BLENDSCHED>, VEX_4V;
- defm VMPSADBW : SS41I_binop_rmi_int<0x42, "vmpsadbw", int_x86_sse41_mpsadbw,
- VR128, loadv2i64, i128mem, 0,
- DEFAULT_ITINS_MPSADSCHED>, VEX_4V;
- }
+
let ExeDomain = SSEPackedSingle in
defm VDPPS : SS41I_binop_rmi_int<0x40, "vdpps", int_x86_sse41_dpps,
VR128, loadv4f32, f128mem, 0,
let Constraints = "$src1 = $dst" in {
let isCommutable = 0 in {
+ defm MPSADBW : SS41I_binop_rmi_int<0x42, "mpsadbw", int_x86_sse41_mpsadbw,
+ VR128, memopv2i64, i128mem,
+ 1, SSE_MPSADBW_ITINS>;
+ }
let ExeDomain = SSEPackedSingle in
defm BLENDPS : SS41I_binop_rmi_int<0x0C, "blendps", int_x86_sse41_blendps,
VR128, memopv4f32, f128mem,
defm PBLENDW : SS41I_binop_rmi_int<0x0E, "pblendw", int_x86_sse41_pblendw,
VR128, memopv2i64, i128mem,
1, SSE_INTALU_ITINS_BLEND_P>;
- defm MPSADBW : SS41I_binop_rmi_int<0x42, "mpsadbw", int_x86_sse41_mpsadbw,
- VR128, memopv2i64, i128mem,
- 1, SSE_MPSADBW_ITINS>;
- }
let ExeDomain = SSEPackedSingle in
defm DPPS : SS41I_binop_rmi_int<0x40, "dpps", int_x86_sse41_dpps,
VR128, memopv4f32, f128mem, 1,
(VPBLENDWYrri VR256:$src1, VR256:$src2, imm:$mask)>;
}
+// Patterns
+let Predicates = [UseAVX] in {
+ let AddedComplexity = 15 in {
+ // Move scalar to XMM zero-extended, zeroing a VR128 then do a
+ // MOVS{S,D} to the lower bits.
+ def : Pat<(v4f32 (X86vzmovl (v4f32 (scalar_to_vector FR32:$src)))),
+ (VMOVSSrr (v4f32 (V_SET0)), FR32:$src)>;
+ def : Pat<(v4f32 (X86vzmovl (v4f32 VR128:$src))),
+ (VBLENDPSrri (v4f32 (V_SET0)), VR128:$src, (i8 1))>;
+ def : Pat<(v4i32 (X86vzmovl (v4i32 VR128:$src))),
+ (VPBLENDWrri (v4i32 (V_SET0)), VR128:$src, (i8 3))>;
+ 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:$src))),
+ (VBLENDPSYrri (v8f32 (AVX_SET0)), VR256:$src, (i8 1))>;
+ def : Pat<(v8i32 (X86vzmovl (v8i32 VR256:$src))),
+ (VBLENDPSYrri (v8i32 (AVX_SET0)), VR256:$src, (i8 1))>;
+ }
+
+ def : Pat<(v8f32 (X86vzmovl (insert_subvector undef,
+ (v4f32 (scalar_to_vector FR32:$src)), (iPTR 0)))),
+ (SUBREG_TO_REG (i32 0),
+ (v4f32 (VMOVSSrr (v4f32 (V_SET0)), FR32:$src)),
+ sub_xmm)>;
+ def : Pat<(v4f64 (X86vzmovl (insert_subvector undef,
+ (v2f64 (scalar_to_vector FR64:$src)), (iPTR 0)))),
+ (SUBREG_TO_REG (i64 0),
+ (v2f64 (VMOVSDrr (v2f64 (V_SET0)), FR64:$src)),
+ sub_xmm)>;
+
+ // Move low f64 and clear high bits.
+ def : Pat<(v4f64 (X86vzmovl (v4f64 VR256:$src))),
+ (VBLENDPDYrri (v4f64 (AVX_SET0)), VR256:$src, (i8 1))>;
+
+ def : Pat<(v4i64 (X86vzmovl (v4i64 VR256:$src))),
+ (VBLENDPDYrri (v4i64 (AVX_SET0)), VR256:$src, (i8 1))>;
+}
+
+let Predicates = [UseSSE41] in {
+ // With SSE41 we can use blends for these patterns.
+ def : Pat<(v4f32 (X86vzmovl (v4f32 VR128:$src))),
+ (BLENDPSrri (v4f32 (V_SET0)), VR128:$src, (i8 1))>;
+ def : Pat<(v4i32 (X86vzmovl (v4i32 VR128:$src))),
+ (PBLENDWrri (v4i32 (V_SET0)), VR128:$src, (i8 3))>;
+ def : Pat<(v2f64 (X86vzmovl (v2f64 VR128:$src))),
+ (BLENDPDrri (v2f64 (V_SET0)), VR128:$src, (i8 1))>;
+}
+
+
/// SS41I_ternary_int - SSE 4.1 ternary operator
let Uses = [XMM0], Constraints = "$src1 = $dst" in {
multiclass SS41I_ternary_int<bits<8> opc, string OpcodeStr, PatFrag mem_frag,
[]>, Sched<[WritePCmpIStrMLd, ReadAfterLd]>;
}
-let Defs = [XMM0, EFLAGS], neverHasSideEffects = 1 in {
+let Defs = [XMM0, EFLAGS], hasSideEffects = 0 in {
let Predicates = [HasAVX] in
defm VPCMPISTRM128 : pcmpistrm_SS42AI<"vpcmpistrm">, VEX;
defm PCMPISTRM128 : pcmpistrm_SS42AI<"pcmpistrm"> ;
[]>, Sched<[WritePCmpEStrMLd, ReadAfterLd]>;
}
-let Defs = [XMM0, EFLAGS], Uses = [EAX, EDX], neverHasSideEffects = 1 in {
+let Defs = [XMM0, EFLAGS], Uses = [EAX, EDX], hasSideEffects = 0 in {
let Predicates = [HasAVX] in
defm VPCMPESTRM128 : SS42AI_pcmpestrm<"vpcmpestrm">, VEX;
defm PCMPESTRM128 : SS42AI_pcmpestrm<"pcmpestrm">;
[]>, Sched<[WritePCmpIStrILd, ReadAfterLd]>;
}
-let Defs = [ECX, EFLAGS], neverHasSideEffects = 1 in {
+let Defs = [ECX, EFLAGS], hasSideEffects = 0 in {
let Predicates = [HasAVX] in
defm VPCMPISTRI : SS42AI_pcmpistri<"vpcmpistri">, VEX;
defm PCMPISTRI : SS42AI_pcmpistri<"pcmpistri">;
[]>, Sched<[WritePCmpEStrILd, ReadAfterLd]>;
}
-let Defs = [ECX, EFLAGS], Uses = [EAX, EDX], neverHasSideEffects = 1 in {
+let Defs = [ECX, EFLAGS], Uses = [EAX, EDX], hasSideEffects = 0 in {
let Predicates = [HasAVX] in
defm VPCMPESTRI : SS42AI_pcmpestri<"vpcmpestri">, VEX;
defm PCMPESTRI : SS42AI_pcmpestri<"pcmpestri">;
//===----------------------------------------------------------------------===//
// VINSERTF128 - Insert packed floating-point values
//
-let neverHasSideEffects = 1, ExeDomain = SSEPackedSingle in {
+let hasSideEffects = 0, ExeDomain = SSEPackedSingle in {
def VINSERTF128rr : AVXAIi8<0x18, MRMSrcReg, (outs VR256:$dst),
(ins VR256:$src1, VR128:$src2, i8imm:$src3),
"vinsertf128\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}",
(INSERT_get_vinsert128_imm VR256:$ins))>;
}
+// Combine two consecutive 16-byte loads with a common destination register into
+// one 32-byte load to that register.
+let Predicates = [HasAVX, HasFastMem32] in {
+ def : Pat<(insert_subvector
+ (v8f32 (insert_subvector undef, (loadv4f32 addr:$src), (iPTR 0))),
+ (loadv4f32 (add addr:$src, (iPTR 16))),
+ (iPTR 4)),
+ (VMOVUPSYrm addr:$src)>;
+
+ def : Pat<(insert_subvector
+ (v4f64 (insert_subvector undef, (loadv2f64 addr:$src), (iPTR 0))),
+ (loadv2f64 (add addr:$src, (iPTR 16))),
+ (iPTR 2)),
+ (VMOVUPDYrm addr:$src)>;
+
+ def : Pat<(insert_subvector
+ (v32i8 (insert_subvector
+ undef, (bc_v16i8 (loadv2i64 addr:$src)), (iPTR 0))),
+ (bc_v16i8 (loadv2i64 (add addr:$src, (iPTR 16)))),
+ (iPTR 16)),
+ (VMOVDQUYrm addr:$src)>;
+
+ def : Pat<(insert_subvector
+ (v16i16 (insert_subvector
+ undef, (bc_v8i16 (loadv2i64 addr:$src)), (iPTR 0))),
+ (bc_v8i16 (loadv2i64 (add addr:$src, (iPTR 16)))),
+ (iPTR 8)),
+ (VMOVDQUYrm addr:$src)>;
+
+ def : Pat<(insert_subvector
+ (v8i32 (insert_subvector
+ undef, (bc_v4i32 (loadv2i64 addr:$src)), (iPTR 0))),
+ (bc_v4i32 (loadv2i64 (add addr:$src, (iPTR 16)))),
+ (iPTR 4)),
+ (VMOVDQUYrm addr:$src)>;
+
+ def : Pat<(insert_subvector
+ (v4i64 (insert_subvector undef, (loadv2i64 addr:$src), (iPTR 0))),
+ (loadv2i64 (add addr:$src, (iPTR 16))),
+ (iPTR 2)),
+ (VMOVDQUYrm addr:$src)>;
+}
+
let Predicates = [HasAVX1Only] in {
def : Pat<(vinsert128_insert:$ins (v4i64 VR256:$src1), (v2i64 VR128:$src2),
(iPTR imm)),
//===----------------------------------------------------------------------===//
// VEXTRACTF128 - Extract packed floating-point values
//
-let neverHasSideEffects = 1, ExeDomain = SSEPackedSingle in {
+let hasSideEffects = 0, ExeDomain = SSEPackedSingle in {
def VEXTRACTF128rr : AVXAIi8<0x19, MRMDestReg, (outs VR128:$dst),
(ins VR256:$src1, i8imm:$src2),
"vextractf128\t{$src2, $src1, $dst|$dst, $src1, $src2}",
"vcvtph2ps\t{$src, $dst|$dst, $src}",
[(set RC:$dst, (Int VR128:$src))]>,
T8PD, VEX, Sched<[WriteCvtF2F]>;
- let neverHasSideEffects = 1, mayLoad = 1 in
+ let hasSideEffects = 0, mayLoad = 1 in
def rm : I<0x13, MRMSrcMem, (outs RC:$dst), (ins x86memop:$src),
"vcvtph2ps\t{$src, $dst|$dst, $src}", []>, T8PD, VEX,
Sched<[WriteCvtF2FLd]>;
"vcvtps2ph\t{$src2, $src1, $dst|$dst, $src1, $src2}",
[(set VR128:$dst, (Int RC:$src1, imm:$src2))]>,
TAPD, VEX, Sched<[WriteCvtF2F]>;
- let neverHasSideEffects = 1, mayStore = 1,
+ let hasSideEffects = 0, mayStore = 1,
SchedRW = [WriteCvtF2FLd, WriteRMW] in
def mr : Ii8<0x1D, MRMDestMem, (outs),
(ins x86memop:$dst, RC:$src1, i32i8imm:$src2),
Sched<[WriteBlendLd, ReadAfterLd]>, VEX_4V;
}
-let isCommutable = 0 in {
defm VPBLENDD : AVX2_binop_rmi_int<0x02, "vpblendd", int_x86_avx2_pblendd_128,
VR128, loadv2i64, i128mem>;
defm VPBLENDDY : AVX2_binop_rmi_int<0x02, "vpblendd", int_x86_avx2_pblendd_256,
VR256, loadv4i64, i256mem>, VEX_L;
-}
def : Pat<(v4i32 (X86Blendi (v4i32 VR128:$src1), (v4i32 VR128:$src2),
imm:$mask)),
def : Pat<(v4f64 (X86VBroadcast (v2f64 VR128:$src))),
(VBROADCASTSDYrr VR128:$src)>;
+ // Provide aliases for broadcast from the same regitser class that
+ // automatically does the extract.
+ def : Pat<(v32i8 (X86VBroadcast (v32i8 VR256:$src))),
+ (VPBROADCASTBYrr (v16i8 (EXTRACT_SUBREG (v32i8 VR256:$src),
+ sub_xmm)))>;
+ def : Pat<(v16i16 (X86VBroadcast (v16i16 VR256:$src))),
+ (VPBROADCASTWYrr (v8i16 (EXTRACT_SUBREG (v16i16 VR256:$src),
+ sub_xmm)))>;
+ def : Pat<(v8i32 (X86VBroadcast (v8i32 VR256:$src))),
+ (VPBROADCASTDYrr (v4i32 (EXTRACT_SUBREG (v8i32 VR256:$src),
+ sub_xmm)))>;
+ def : Pat<(v4i64 (X86VBroadcast (v4i64 VR256:$src))),
+ (VPBROADCASTQYrr (v2i64 (EXTRACT_SUBREG (v4i64 VR256:$src),
+ sub_xmm)))>;
+ def : Pat<(v8f32 (X86VBroadcast (v8f32 VR256:$src))),
+ (VBROADCASTSSYrr (v4f32 (EXTRACT_SUBREG (v8f32 VR256:$src),
+ sub_xmm)))>;
+ def : Pat<(v4f64 (X86VBroadcast (v4f64 VR256:$src))),
+ (VBROADCASTSDYrr (v2f64 (EXTRACT_SUBREG (v4f64 VR256:$src),
+ sub_xmm)))>;
+
// 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 {
//===----------------------------------------------------------------------===//
// VINSERTI128 - Insert packed integer values
//
-let neverHasSideEffects = 1 in {
+let hasSideEffects = 0 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, $src3}",
[(set VR128:$dst,
(int_x86_avx2_vextracti128 VR256:$src1, imm:$src2))]>,
Sched<[WriteShuffle256]>, VEX, VEX_L;
-let neverHasSideEffects = 1, mayStore = 1 in
+let hasSideEffects = 0, mayStore = 1 in
def VEXTRACTI128mr : AVX2AIi8<0x39, MRMDestMem, (outs),
(ins i128mem:$dst, VR256:$src1, i8imm:$src2),
"vextracti128\t{$src2, $src1, $dst|$dst, $src1, $src2}", []>,
int_x86_avx2_maskstore_q,
int_x86_avx2_maskstore_q_256>, VEX_W;
+def: Pat<(masked_store addr:$ptr, (v8i32 VR256:$mask), (v8f32 VR256:$src)),
+ (VMASKMOVPSYmr addr:$ptr, VR256:$mask, VR256:$src)>;
+
+def: Pat<(masked_store addr:$ptr, (v8i32 VR256:$mask), (v8i32 VR256:$src)),
+ (VPMASKMOVDYmr addr:$ptr, VR256:$mask, VR256:$src)>;
+
+def: Pat<(masked_store addr:$ptr, (v4i32 VR128:$mask), (v4f32 VR128:$src)),
+ (VMASKMOVPSmr addr:$ptr, VR128:$mask, VR128:$src)>;
+
+def: Pat<(masked_store addr:$ptr, (v4i32 VR128:$mask), (v4i32 VR128:$src)),
+ (VPMASKMOVDmr addr:$ptr, VR128:$mask, VR128:$src)>;
+
+def: Pat<(v8f32 (masked_load addr:$ptr, (v8i32 VR256:$mask), undef)),
+ (VMASKMOVPSYrm VR256:$mask, addr:$ptr)>;
+
+def: Pat<(v8f32 (masked_load addr:$ptr, (v8i32 VR256:$mask),
+ (bc_v8f32 (v8i32 immAllZerosV)))),
+ (VMASKMOVPSYrm VR256:$mask, addr:$ptr)>;
+
+def: Pat<(v8f32 (masked_load addr:$ptr, (v8i32 VR256:$mask), (v8f32 VR256:$src0))),
+ (VBLENDVPSYrr VR256:$src0, (VMASKMOVPSYrm VR256:$mask, addr:$ptr),
+ VR256:$mask)>;
+
+def: Pat<(v8i32 (masked_load addr:$ptr, (v8i32 VR256:$mask), undef)),
+ (VPMASKMOVDYrm VR256:$mask, addr:$ptr)>;
+
+def: Pat<(v8i32 (masked_load addr:$ptr, (v8i32 VR256:$mask), (v8i32 immAllZerosV))),
+ (VPMASKMOVDYrm VR256:$mask, addr:$ptr)>;
+
+def: Pat<(v8i32 (masked_load addr:$ptr, (v8i32 VR256:$mask), (v8i32 VR256:$src0))),
+ (VBLENDVPSYrr VR256:$src0, (VPMASKMOVDYrm VR256:$mask, addr:$ptr),
+ VR256:$mask)>;
+
+def: Pat<(v4f32 (masked_load addr:$ptr, (v4i32 VR128:$mask), undef)),
+ (VMASKMOVPSrm VR128:$mask, addr:$ptr)>;
+
+def: Pat<(v4f32 (masked_load addr:$ptr, (v4i32 VR128:$mask),
+ (bc_v4f32 (v4i32 immAllZerosV)))),
+ (VMASKMOVPSrm VR128:$mask, addr:$ptr)>;
+
+def: Pat<(v4f32 (masked_load addr:$ptr, (v4i32 VR128:$mask), (v4f32 VR128:$src0))),
+ (VBLENDVPSrr VR128:$src0, (VMASKMOVPSrm VR128:$mask, addr:$ptr),
+ VR128:$mask)>;
+
+def: Pat<(v4i32 (masked_load addr:$ptr, (v4i32 VR128:$mask), undef)),
+ (VPMASKMOVDrm VR128:$mask, addr:$ptr)>;
+
+def: Pat<(v4i32 (masked_load addr:$ptr, (v4i32 VR128:$mask), (v4i32 immAllZerosV))),
+ (VPMASKMOVDrm VR128:$mask, addr:$ptr)>;
+
+def: Pat<(v4i32 (masked_load addr:$ptr, (v4i32 VR128:$mask), (v4i32 VR128:$src0))),
+ (VBLENDVPSrr VR128:$src0, (VPMASKMOVDrm VR128:$mask, addr:$ptr),
+ VR128:$mask)>;
+
+def: Pat<(masked_store addr:$ptr, (v4i64 VR256:$mask), (v4f64 VR256:$src)),
+ (VMASKMOVPDYmr addr:$ptr, VR256:$mask, VR256:$src)>;
+
+def: Pat<(masked_store addr:$ptr, (v4i64 VR256:$mask), (v4i64 VR256:$src)),
+ (VPMASKMOVQYmr addr:$ptr, VR256:$mask, VR256:$src)>;
+
+def: Pat<(v4f64 (masked_load addr:$ptr, (v4i64 VR256:$mask), undef)),
+ (VMASKMOVPDYrm VR256:$mask, addr:$ptr)>;
+
+def: Pat<(v4f64 (masked_load addr:$ptr, (v4i64 VR256:$mask),
+ (v4f64 immAllZerosV))),
+ (VMASKMOVPDYrm VR256:$mask, addr:$ptr)>;
+
+def: Pat<(v4f64 (masked_load addr:$ptr, (v4i64 VR256:$mask), (v4f64 VR256:$src0))),
+ (VBLENDVPDYrr VR256:$src0, (VMASKMOVPDYrm VR256:$mask, addr:$ptr),
+ VR256:$mask)>;
+
+def: Pat<(v4i64 (masked_load addr:$ptr, (v4i64 VR256:$mask), undef)),
+ (VPMASKMOVQYrm VR256:$mask, addr:$ptr)>;
+
+def: Pat<(v4i64 (masked_load addr:$ptr, (v4i64 VR256:$mask),
+ (bc_v4i64 (v8i32 immAllZerosV)))),
+ (VPMASKMOVQYrm VR256:$mask, addr:$ptr)>;
+
+def: Pat<(v4i64 (masked_load addr:$ptr, (v4i64 VR256:$mask), (v4i64 VR256:$src0))),
+ (VBLENDVPDYrr VR256:$src0, (VPMASKMOVQYrm VR256:$mask, addr:$ptr),
+ VR256:$mask)>;
+
+def: Pat<(masked_store addr:$ptr, (v2i64 VR128:$mask), (v2f64 VR128:$src)),
+ (VMASKMOVPDmr addr:$ptr, VR128:$mask, VR128:$src)>;
+
+def: Pat<(masked_store addr:$ptr, (v2i64 VR128:$mask), (v2i64 VR128:$src)),
+ (VPMASKMOVQmr addr:$ptr, VR128:$mask, VR128:$src)>;
+
+def: Pat<(v2f64 (masked_load addr:$ptr, (v2i64 VR128:$mask), undef)),
+ (VMASKMOVPDrm VR128:$mask, addr:$ptr)>;
+
+def: Pat<(v2f64 (masked_load addr:$ptr, (v2i64 VR128:$mask),
+ (v2f64 immAllZerosV))),
+ (VMASKMOVPDrm VR128:$mask, addr:$ptr)>;
+
+def: Pat<(v2f64 (masked_load addr:$ptr, (v2i64 VR128:$mask), (v2f64 VR128:$src0))),
+ (VBLENDVPDrr VR128:$src0, (VMASKMOVPDrm VR128:$mask, addr:$ptr),
+ VR128:$mask)>;
+
+def: Pat<(v2i64 (masked_load addr:$ptr, (v2i64 VR128:$mask), undef)),
+ (VPMASKMOVQrm VR128:$mask, addr:$ptr)>;
+
+def: Pat<(v2i64 (masked_load addr:$ptr, (v2i64 VR128:$mask),
+ (bc_v2i64 (v4i32 immAllZerosV)))),
+ (VPMASKMOVQrm VR128:$mask, addr:$ptr)>;
+
+def: Pat<(v2i64 (masked_load addr:$ptr, (v2i64 VR128:$mask), (v2i64 VR128:$src0))),
+ (VBLENDVPDrr VR128:$src0, (VPMASKMOVQrm VR128:$mask, addr:$ptr),
+ VR128:$mask)>;
//===----------------------------------------------------------------------===//
// Variable Bit Shifts
projects / oota-llvm.git / blobdiff