// Bitcasts between 512-bit vector types. Return the original type since
// no instruction is needed for the conversion
let Predicates = [HasAVX512] in {
- def : Pat<(v8f64 (bitconvert (v16f32 VR512:$src))), (v8f64 VR512:$src)>;
- def : Pat<(v8f64 (bitconvert (v16i32 VR512:$src))), (v8f64 VR512:$src)>;
def : Pat<(v8f64 (bitconvert (v8i64 VR512:$src))), (v8f64 VR512:$src)>;
- def : Pat<(v16f32 (bitconvert (v16i32 VR512:$src))), (v16f32 VR512:$src)>;
+ def : Pat<(v8f64 (bitconvert (v16i32 VR512:$src))), (v8f64 VR512:$src)>;
+ def : Pat<(v8f64 (bitconvert (v32i16 VR512:$src))), (v8f64 VR512:$src)>;
+ def : Pat<(v8f64 (bitconvert (v64i8 VR512:$src))), (v8f64 VR512:$src)>;
+ def : Pat<(v8f64 (bitconvert (v16f32 VR512:$src))), (v8f64 VR512:$src)>;
def : Pat<(v16f32 (bitconvert (v8i64 VR512:$src))), (v16f32 VR512:$src)>;
+ def : Pat<(v16f32 (bitconvert (v16i32 VR512:$src))), (v16f32 VR512:$src)>;
+ def : Pat<(v16f32 (bitconvert (v32i16 VR512:$src))), (v16f32 VR512:$src)>;
+ def : Pat<(v16f32 (bitconvert (v64i8 VR512:$src))), (v16f32 VR512:$src)>;
def : Pat<(v16f32 (bitconvert (v8f64 VR512:$src))), (v16f32 VR512:$src)>;
- def : Pat<(v8i64 (bitconvert (v16f32 VR512:$src))), (v8i64 VR512:$src)>;
def : Pat<(v8i64 (bitconvert (v16i32 VR512:$src))), (v8i64 VR512:$src)>;
+ def : Pat<(v8i64 (bitconvert (v32i16 VR512:$src))), (v8i64 VR512:$src)>;
+ def : Pat<(v8i64 (bitconvert (v64i8 VR512:$src))), (v8i64 VR512:$src)>;
def : Pat<(v8i64 (bitconvert (v8f64 VR512:$src))), (v8i64 VR512:$src)>;
- def : Pat<(v16i32 (bitconvert (v16f32 VR512:$src))), (v16i32 VR512:$src)>;
- def : Pat<(v16i32 (bitconvert (v8i64 VR512:$src))), (v16i32 VR512:$src)>;
+ def : Pat<(v8i64 (bitconvert (v16f32 VR512:$src))), (v8i64 VR512:$src)>;
+ def : Pat<(v16i32 (bitconvert (v8i64 VR512:$src))), (v16i32 VR512:$src)>;
+ def : Pat<(v16i32 (bitconvert (v32i16 VR512:$src))), (v16i32 VR512:$src)>;
+ def : Pat<(v16i32 (bitconvert (v64i8 VR512:$src))), (v16i32 VR512:$src)>;
def : Pat<(v16i32 (bitconvert (v8f64 VR512:$src))), (v16i32 VR512:$src)>;
- def : Pat<(v8f64 (bitconvert (v8i64 VR512:$src))), (v8f64 VR512:$src)>;
+ def : Pat<(v32i16 (bitconvert (v8i64 VR512:$src))), (v32i16 VR512:$src)>;
+ def : Pat<(v32i16 (bitconvert (v16i32 VR512:$src))), (v32i16 VR512:$src)>;
+ def : Pat<(v32i16 (bitconvert (v64i8 VR512:$src))), (v32i16 VR512:$src)>;
+ def : Pat<(v32i16 (bitconvert (v8f64 VR512:$src))), (v32i16 VR512:$src)>;
+ def : Pat<(v32i16 (bitconvert (v16f32 VR512:$src))), (v32i16 VR512:$src)>;
+ def : Pat<(v32i16 (bitconvert (v16f32 VR512:$src))), (v32i16 VR512:$src)>;
+ def : Pat<(v64i8 (bitconvert (v8i64 VR512:$src))), (v64i8 VR512:$src)>;
+ def : Pat<(v64i8 (bitconvert (v16i32 VR512:$src))), (v64i8 VR512:$src)>;
+ def : Pat<(v64i8 (bitconvert (v32i16 VR512:$src))), (v64i8 VR512:$src)>;
+ def : Pat<(v64i8 (bitconvert (v8f64 VR512:$src))), (v64i8 VR512:$src)>;
+ def : Pat<(v64i8 (bitconvert (v16f32 VR512:$src))), (v64i8 VR512:$src)>;
def : Pat<(v2i64 (bitconvert (v4i32 VR128X:$src))), (v2i64 VR128X:$src)>;
def : Pat<(v2i64 (bitconvert (v8i16 VR128X:$src))), (v2i64 VR128X:$src)>;
(ins VR512:$src1, i128mem:$src2, i8imm:$src3),
"vinserti32x4\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}",
[]>, EVEX_4V, EVEX_V512, EVEX_CD8<32, CD8VT4>;
-
}
let hasSideEffects = 0 in {
// -- VPERM2I - 3 source operands form --
multiclass avx512_perm_3src<bits<8> opc, string OpcodeStr, RegisterClass RC,
PatFrag mem_frag, X86MemOperand x86memop,
- SDNode OpNode, ValueType OpVT> {
+ SDNode OpNode, ValueType OpVT, RegisterClass KRC> {
let Constraints = "$src1 = $dst" in {
def rr : AVX5128I<opc, MRMSrcReg, (outs RC:$dst),
(ins RC:$src1, RC:$src2, RC:$src3),
(OpVT (OpNode RC:$src1, RC:$src2, RC:$src3)))]>,
EVEX_4V;
+ def rrk : AVX5128I<opc, MRMSrcReg, (outs RC:$dst),
+ (ins RC:$src1, KRC:$mask, RC:$src2, RC:$src3),
+ !strconcat(OpcodeStr,
+ " \t{$src3, $src2, $dst {${mask}}|"
+ "$dst {${mask}}, $src2, $src3}"),
+ [(set RC:$dst, (OpVT (vselect KRC:$mask,
+ (OpNode RC:$src1, RC:$src2,
+ RC:$src3),
+ RC:$src1)))]>,
+ EVEX_4V, EVEX_K;
+
+ let AddedComplexity = 30 in // Prefer over VMOV*rrkz Pat<>
+ def rrkz : AVX5128I<opc, MRMSrcReg, (outs RC:$dst),
+ (ins RC:$src1, KRC:$mask, RC:$src2, RC:$src3),
+ !strconcat(OpcodeStr,
+ " \t{$src3, $src2, $dst {${mask}} {z} |",
+ "$dst {${mask}} {z}, $src2, $src3}"),
+ [(set RC:$dst, (OpVT (vselect KRC:$mask,
+ (OpNode RC:$src1, RC:$src2,
+ RC:$src3),
+ (OpVT (bitconvert
+ (v16i32 immAllZerosV))))))]>,
+ EVEX_4V, EVEX_KZ;
+
def rm : AVX5128I<opc, MRMSrcMem, (outs RC:$dst),
(ins RC:$src1, RC:$src2, x86memop:$src3),
!strconcat(OpcodeStr,
" \t{$src3, $src2, $dst|$dst, $src2, $src3}"),
[(set RC:$dst,
- (OpVT (OpNode RC:$src1, RC:$src2,
+ (OpVT (OpNode RC:$src1, RC:$src2,
(mem_frag addr:$src3))))]>, EVEX_4V;
+
+ def rmk : AVX5128I<opc, MRMSrcMem, (outs RC:$dst),
+ (ins RC:$src1, KRC:$mask, RC:$src2, x86memop:$src3),
+ !strconcat(OpcodeStr,
+ " \t{$src3, $src2, $dst {${mask}}|"
+ "$dst {${mask}}, $src2, $src3}"),
+ [(set RC:$dst,
+ (OpVT (vselect KRC:$mask,
+ (OpNode RC:$src1, RC:$src2,
+ (mem_frag addr:$src3)),
+ RC:$src1)))]>,
+ EVEX_4V, EVEX_K;
+
+ let AddedComplexity = 10 in // Prefer over the rrkz variant
+ def rmkz : AVX5128I<opc, MRMSrcMem, (outs RC:$dst),
+ (ins RC:$src1, KRC:$mask, RC:$src2, x86memop:$src3),
+ !strconcat(OpcodeStr,
+ " \t{$src3, $src2, $dst {${mask}} {z}|"
+ "$dst {${mask}} {z}, $src2, $src3}"),
+ [(set RC:$dst,
+ (OpVT (vselect KRC:$mask,
+ (OpNode RC:$src1, RC:$src2,
+ (mem_frag addr:$src3)),
+ (OpVT (bitconvert
+ (v16i32 immAllZerosV))))))]>,
+ EVEX_4V, EVEX_KZ;
}
}
-defm VPERMI2D : avx512_perm_3src<0x76, "vpermi2d", VR512, memopv16i32, i512mem,
- X86VPermiv3, v16i32>, EVEX_V512, EVEX_CD8<32, CD8VF>;
-defm VPERMI2Q : avx512_perm_3src<0x76, "vpermi2q", VR512, memopv8i64, i512mem,
- X86VPermiv3, v8i64>, EVEX_V512, VEX_W, EVEX_CD8<64, CD8VF>;
-defm VPERMI2PS : avx512_perm_3src<0x77, "vpermi2ps", VR512, memopv16f32, i512mem,
- X86VPermiv3, v16f32>, EVEX_V512, EVEX_CD8<32, CD8VF>;
-defm VPERMI2PD : avx512_perm_3src<0x77, "vpermi2pd", VR512, memopv8f64, i512mem,
- X86VPermiv3, v8f64>, EVEX_V512, VEX_W, EVEX_CD8<64, CD8VF>;
-
-defm VPERMT2D : avx512_perm_3src<0x7E, "vpermt2d", VR512, memopv16i32, i512mem,
- X86VPermv3, v16i32>, EVEX_V512, EVEX_CD8<32, CD8VF>;
-defm VPERMT2Q : avx512_perm_3src<0x7E, "vpermt2q", VR512, memopv8i64, i512mem,
- X86VPermv3, v8i64>, EVEX_V512, VEX_W, EVEX_CD8<64, CD8VF>;
-defm VPERMT2PS : avx512_perm_3src<0x7F, "vpermt2ps", VR512, memopv16f32, i512mem,
- X86VPermv3, v16f32>, EVEX_V512, EVEX_CD8<32, CD8VF>;
-defm VPERMT2PD : avx512_perm_3src<0x7F, "vpermt2pd", VR512, memopv8f64, i512mem,
- X86VPermv3, v8f64>, EVEX_V512, VEX_W, EVEX_CD8<64, CD8VF>;
-
-def : Pat<(v16f32 (int_x86_avx512_mask_vpermt_ps_512 (v16i32 VR512:$idx),
- (v16f32 VR512:$src1), (v16f32 VR512:$src2), (i16 -1))),
- (VPERMT2PSrr VR512:$src1, VR512:$idx, VR512:$src2)>;
-
-def : Pat<(v16i32 (int_x86_avx512_mask_vpermt_d_512 (v16i32 VR512:$idx),
- (v16i32 VR512:$src1), (v16i32 VR512:$src2), (i16 -1))),
- (VPERMT2Drr VR512:$src1, VR512:$idx, VR512:$src2)>;
-
-def : Pat<(v8f64 (int_x86_avx512_mask_vpermt_pd_512 (v8i64 VR512:$idx),
- (v8f64 VR512:$src1), (v8f64 VR512:$src2), (i8 -1))),
- (VPERMT2PDrr VR512:$src1, VR512:$idx, VR512:$src2)>;
-
-def : Pat<(v8i64 (int_x86_avx512_mask_vpermt_q_512 (v8i64 VR512:$idx),
- (v8i64 VR512:$src1), (v8i64 VR512:$src2), (i8 -1))),
- (VPERMT2Qrr VR512:$src1, VR512:$idx, VR512:$src2)>;
+defm VPERMI2D : avx512_perm_3src<0x76, "vpermi2d", VR512, memopv16i32,
+ i512mem, X86VPermiv3, v16i32, VK16WM>,
+ EVEX_V512, EVEX_CD8<32, CD8VF>;
+defm VPERMI2Q : avx512_perm_3src<0x76, "vpermi2q", VR512, memopv8i64,
+ i512mem, X86VPermiv3, v8i64, VK8WM>,
+ EVEX_V512, VEX_W, EVEX_CD8<64, CD8VF>;
+defm VPERMI2PS : avx512_perm_3src<0x77, "vpermi2ps", VR512, memopv16f32,
+ i512mem, X86VPermiv3, v16f32, VK16WM>,
+ EVEX_V512, EVEX_CD8<32, CD8VF>;
+defm VPERMI2PD : avx512_perm_3src<0x77, "vpermi2pd", VR512, memopv8f64,
+ i512mem, X86VPermiv3, v8f64, VK8WM>,
+ EVEX_V512, VEX_W, EVEX_CD8<64, CD8VF>;
+
+multiclass avx512_perm_table_3src<bits<8> opc, string Suffix, RegisterClass RC,
+ PatFrag mem_frag, X86MemOperand x86memop,
+ SDNode OpNode, ValueType OpVT, RegisterClass KRC,
+ ValueType MaskVT, RegisterClass MRC> :
+ avx512_perm_3src<opc, "vpermt2"##Suffix, RC, mem_frag, x86memop, OpNode,
+ OpVT, KRC> {
+ def : Pat<(OpVT (!cast<Intrinsic>("int_x86_avx512_mask_vpermt_"##Suffix##"_512")
+ VR512:$idx, VR512:$src1, VR512:$src2, -1)),
+ (!cast<Instruction>(NAME#rr) VR512:$src1, VR512:$idx, VR512:$src2)>;
+
+ def : Pat<(OpVT (!cast<Intrinsic>("int_x86_avx512_mask_vpermt_"##Suffix##"_512")
+ VR512:$idx, VR512:$src1, VR512:$src2, MRC:$mask)),
+ (!cast<Instruction>(NAME#rrk) VR512:$src1,
+ (MaskVT (COPY_TO_REGCLASS MRC:$mask, KRC)), VR512:$idx, VR512:$src2)>;
+}
+
+defm VPERMT2D : avx512_perm_table_3src<0x7E, "d", VR512, memopv16i32, i512mem,
+ X86VPermv3, v16i32, VK16WM, v16i1, GR16>,
+ EVEX_V512, EVEX_CD8<32, CD8VF>;
+defm VPERMT2Q : avx512_perm_table_3src<0x7E, "q", VR512, memopv8i64, i512mem,
+ X86VPermv3, v8i64, VK8WM, v8i1, GR8>,
+ EVEX_V512, VEX_W, EVEX_CD8<64, CD8VF>;
+defm VPERMT2PS : avx512_perm_table_3src<0x7F, "ps", VR512, memopv16f32, i512mem,
+ X86VPermv3, v16f32, VK16WM, v16i1, GR16>,
+ EVEX_V512, EVEX_CD8<32, CD8VF>;
+defm VPERMT2PD : avx512_perm_table_3src<0x7F, "pd", VR512, memopv8f64, i512mem,
+ X86VPermv3, v8f64, VK8WM, v8i1, GR8>,
+ EVEX_V512, VEX_W, EVEX_CD8<64, CD8VF>;
+
//===----------------------------------------------------------------------===//
// AVX-512 - BLEND using mask
//
(v16i32 (SUBREG_TO_REG (i32 0), VR256X:$src1, sub_ymm)),
(v16i32 (SUBREG_TO_REG (i32 0), VR256X:$src2, sub_ymm))), VK8)>;
-multiclass avx512_icmp_cc<bits<8> opc, RegisterClass KRC,
+multiclass avx512_icmp_cc<bits<8> opc, RegisterClass WMRC, RegisterClass KRC,
RegisterClass RC, X86MemOperand x86memop, PatFrag memop_frag,
SDNode OpNode, ValueType vt, Operand CC, string Suffix> {
def rri : AVX512AIi8<opc, MRMSrcReg,
!strconcat("vpcmp", Suffix,
"\t{$cc, $src2, $src1, $dst|$dst, $src1, $src2, $cc}"),
[], IIC_SSE_ALU_F32P_RR>, EVEX_4V;
+ def rrik_alt : AVX512AIi8<opc, MRMSrcReg,
+ (outs KRC:$dst), (ins WMRC:$mask, RC:$src1, RC:$src2, i8imm:$cc),
+ !strconcat("vpcmp", Suffix,
+ "\t{$cc, $src2, $src1, $dst {${mask}}|$dst {${mask}}, $src1, $src2, $cc}"),
+ [], IIC_SSE_ALU_F32P_RR>, EVEX_4V, EVEX_K;
def rmi_alt : AVX512AIi8<opc, MRMSrcMem,
(outs KRC:$dst), (ins RC:$src1, x86memop:$src2, i8imm:$cc),
!strconcat("vpcmp", Suffix,
"\t{$cc, $src2, $src1, $dst|$dst, $src1, $src2, $cc}"),
[], IIC_SSE_ALU_F32P_RM>, EVEX_4V;
+ def rmik_alt : AVX512AIi8<opc, MRMSrcMem,
+ (outs KRC:$dst), (ins WMRC:$mask, RC:$src1, x86memop:$src2, i8imm:$cc),
+ !strconcat("vpcmp", Suffix,
+ "\t{$cc, $src2, $src1, $dst {${mask}}|$dst {${mask}}, $src1, $src2, $cc}"),
+ [], IIC_SSE_ALU_F32P_RM>, EVEX_4V, EVEX_K;
}
}
-defm VPCMPDZ : avx512_icmp_cc<0x1F, VK16, VR512, i512mem, memopv16i32,
+defm VPCMPDZ : avx512_icmp_cc<0x1F, VK16WM, VK16, VR512, i512mem, memopv16i32,
X86cmpm, v16i32, AVXCC, "d">,
EVEX_V512, EVEX_CD8<32, CD8VF>;
-defm VPCMPUDZ : avx512_icmp_cc<0x1E, VK16, VR512, i512mem, memopv16i32,
+defm VPCMPUDZ : avx512_icmp_cc<0x1E, VK16WM, VK16, VR512, i512mem, memopv16i32,
X86cmpmu, v16i32, AVXCC, "ud">,
EVEX_V512, EVEX_CD8<32, CD8VF>;
-defm VPCMPQZ : avx512_icmp_cc<0x1F, VK8, VR512, i512mem, memopv8i64,
+defm VPCMPQZ : avx512_icmp_cc<0x1F, VK8WM, VK8, VR512, i512mem, memopv8i64,
X86cmpm, v8i64, AVXCC, "q">,
VEX_W, EVEX_V512, EVEX_CD8<64, CD8VF>;
-defm VPCMPUQZ : avx512_icmp_cc<0x1E, VK8, VR512, i512mem, memopv8i64,
+defm VPCMPUQZ : avx512_icmp_cc<0x1E, VK8WM, VK8, VR512, i512mem, memopv8i64,
X86cmpmu, v8i64, AVXCC, "uq">,
VEX_W, EVEX_V512, EVEX_CD8<64, CD8VF>;
projects / oota-llvm.git / blobdiff