diff options
Diffstat (limited to 'llvm/lib/Target/X86/X86InstrSSE.td')
| -rw-r--r-- | llvm/lib/Target/X86/X86InstrSSE.td | 32 |
1 files changed, 19 insertions, 13 deletions
diff --git a/llvm/lib/Target/X86/X86InstrSSE.td b/llvm/lib/Target/X86/X86InstrSSE.td index 6f6afe74cdb..8f88b9b333b 100644 --- a/llvm/lib/Target/X86/X86InstrSSE.td +++ b/llvm/lib/Target/X86/X86InstrSSE.td @@ -6049,7 +6049,7 @@ let Predicates = [HasAVX] in { VEX_4V, VEX_WIG; defm VBLENDPSY : SS41I_blend_rmi<0x0C, "vblendps", X86Blendi, v8f32, VR256, loadv8f32, f256mem, 0, SSEPackedSingle, - WriteFBlend, BlendCommuteImm8>, + WriteFBlendY, BlendCommuteImm8>, VEX_4V, VEX_L, VEX_WIG; defm VBLENDPD : SS41I_blend_rmi<0x0D, "vblendpd", X86Blendi, v2f64, VR128, loadv2f64, f128mem, 0, SSEPackedDouble, @@ -6057,7 +6057,7 @@ let Predicates = [HasAVX] in { VEX_4V, VEX_WIG; defm VBLENDPDY : SS41I_blend_rmi<0x0D, "vblendpd", X86Blendi, v4f64, VR256, loadv4f64, f256mem, 0, SSEPackedDouble, - WriteFBlend, BlendCommuteImm4>, + WriteFBlendY, BlendCommuteImm4>, VEX_4V, VEX_L, VEX_WIG; defm VPBLENDW : SS41I_blend_rmi<0x0E, "vpblendw", X86Blendi, v8i16, VR128, loadv2i64, i128mem, 0, SSEPackedInt, @@ -6130,7 +6130,7 @@ defm VBLENDVPD : SS41I_quaternary_int_avx<0x4B, "vblendvpd", VR128, f128mem, WriteFVarBlend>; defm VBLENDVPDY : SS41I_quaternary_int_avx<0x4B, "vblendvpd", VR256, f256mem, loadv4f64, int_x86_avx_blendv_pd_256, - WriteFVarBlend>, VEX_L; + WriteFVarBlendY>, VEX_L; } // ExeDomain = SSEPackedDouble let ExeDomain = SSEPackedSingle in { defm VBLENDVPS : SS41I_quaternary_int_avx<0x4A, "vblendvps", VR128, f128mem, @@ -6138,7 +6138,7 @@ defm VBLENDVPS : SS41I_quaternary_int_avx<0x4A, "vblendvps", VR128, f128mem, WriteFVarBlend>; defm VBLENDVPSY : SS41I_quaternary_int_avx<0x4A, "vblendvps", VR256, f256mem, loadv8f32, int_x86_avx_blendv_ps_256, - WriteFVarBlend>, VEX_L; + WriteFVarBlendY>, VEX_L; } // ExeDomain = SSEPackedSingle defm VPBLENDVB : SS41I_quaternary_int_avx<0x4C, "vpblendvb", VR128, i128mem, loadv2i64, int_x86_sse41_pblendvb, @@ -7156,45 +7156,51 @@ defm VMASKMOVPD : avx_movmask_rm<0x2D, 0x2F, "vmaskmovpd", multiclass avx_permil<bits<8> opc_rm, bits<8> opc_rmi, string OpcodeStr, RegisterClass RC, X86MemOperand x86memop_f, X86MemOperand x86memop_i, PatFrag i_frag, - ValueType f_vt, ValueType i_vt> { + ValueType f_vt, ValueType i_vt, + X86FoldableSchedWrite sched, + X86FoldableSchedWrite varsched> { let Predicates = [HasAVX, NoVLX] in { def rr : AVX8I<opc_rm, MRMSrcReg, (outs RC:$dst), (ins RC:$src1, RC:$src2), !strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}"), [(set RC:$dst, (f_vt (X86VPermilpv RC:$src1, (i_vt RC:$src2))))]>, VEX_4V, - Sched<[WriteFVarShuffle]>; + Sched<[varsched]>; def rm : AVX8I<opc_rm, MRMSrcMem, (outs RC:$dst), (ins RC:$src1, x86memop_i:$src2), !strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}"), [(set RC:$dst, (f_vt (X86VPermilpv RC:$src1, (i_vt (bitconvert (i_frag addr:$src2))))))]>, VEX_4V, - Sched<[WriteFVarShuffleLd, ReadAfterLd]>; + Sched<[varsched.Folded, ReadAfterLd]>; def ri : AVXAIi8<opc_rmi, MRMSrcReg, (outs RC:$dst), (ins RC:$src1, u8imm:$src2), !strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}"), [(set RC:$dst, (f_vt (X86VPermilpi RC:$src1, (i8 imm:$src2))))]>, VEX, - Sched<[WriteFShuffle]>; + Sched<[sched]>; def mi : AVXAIi8<opc_rmi, MRMSrcMem, (outs RC:$dst), (ins x86memop_f:$src1, u8imm:$src2), !strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}"), [(set RC:$dst, (f_vt (X86VPermilpi (load addr:$src1), (i8 imm:$src2))))]>, VEX, - Sched<[WriteFShuffleLd]>; + Sched<[sched.Folded]>; }// Predicates = [HasAVX, NoVLX] } let ExeDomain = SSEPackedSingle in { defm VPERMILPS : avx_permil<0x0C, 0x04, "vpermilps", VR128, f128mem, i128mem, - loadv2i64, v4f32, v4i32>; + loadv2i64, v4f32, v4i32, WriteFShuffle, + WriteFVarShuffle>; defm VPERMILPSY : avx_permil<0x0C, 0x04, "vpermilps", VR256, f256mem, i256mem, - loadv4i64, v8f32, v8i32>, VEX_L; + loadv4i64, v8f32, v8i32, WriteFShuffle, + WriteFVarShuffleY>, VEX_L; } let ExeDomain = SSEPackedDouble in { defm VPERMILPD : avx_permil<0x0D, 0x05, "vpermilpd", VR128, f128mem, i128mem, - loadv2i64, v2f64, v2i64>; + loadv2i64, v2f64, v2i64, WriteFShuffle, + WriteFVarShuffle>; defm VPERMILPDY : avx_permil<0x0D, 0x05, "vpermilpd", VR256, f256mem, i256mem, - loadv4i64, v4f64, v4i64>, VEX_L; + loadv4i64, v4f64, v4i64, WriteFShuffle, + WriteFVarShuffleY>, VEX_L; } //===----------------------------------------------------------------------===// |
