diff options
| -rw-r--r-- | llvm/lib/Target/X86/X86InstrSSE.td | 24 |
1 files changed, 14 insertions, 10 deletions
diff --git a/llvm/lib/Target/X86/X86InstrSSE.td b/llvm/lib/Target/X86/X86InstrSSE.td index 918c1845aa3..0f9a4818967 100644 --- a/llvm/lib/Target/X86/X86InstrSSE.td +++ b/llvm/lib/Target/X86/X86InstrSSE.td @@ -3589,7 +3589,7 @@ def VMOVNTPDmr : VPDI<0x2B, MRMDestMem, (outs), let ExeDomain = SSEPackedInt in def VMOVNTDQmr : VPDI<0xE7, MRMDestMem, (outs), - (ins f128mem:$dst, VR128:$src), + (ins i128mem:$dst, VR128:$src), "movntdq\t{$src, $dst|$dst, $src}", [(alignednontemporalstore (v2i64 VR128:$src), addr:$dst)], @@ -3609,7 +3609,7 @@ def VMOVNTPDYmr : VPDI<0x2B, MRMDestMem, (outs), IIC_SSE_MOVNT>, VEX, VEX_L, VEX_WIG; let ExeDomain = SSEPackedInt in def VMOVNTDQYmr : VPDI<0xE7, MRMDestMem, (outs), - (ins f256mem:$dst, VR256:$src), + (ins i256mem:$dst, VR256:$src), "movntdq\t{$src, $dst|$dst, $src}", [(alignednontemporalstore (v4i64 VR256:$src), addr:$dst)], @@ -8383,7 +8383,8 @@ let Predicates = [HasAVX1Only] in { // multiclass avx2_perm<bits<8> opc, string OpcodeStr, PatFrag mem_frag, - ValueType OpVT, X86FoldableSchedWrite Sched> { + ValueType OpVT, X86FoldableSchedWrite Sched, + X86MemOperand memOp> { let Predicates = [HasAVX2, NoVLX] in { def Yrr : AVX28I<opc, MRMSrcReg, (outs VR256:$dst), (ins VR256:$src1, VR256:$src2), @@ -8393,7 +8394,7 @@ multiclass avx2_perm<bits<8> opc, string OpcodeStr, PatFrag mem_frag, (OpVT (X86VPermv VR256:$src1, VR256:$src2)))]>, Sched<[Sched]>, VEX_4V, VEX_L; def Yrm : AVX28I<opc, MRMSrcMem, (outs VR256:$dst), - (ins VR256:$src1, i256mem:$src2), + (ins VR256:$src1, memOp:$src2), !strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}"), [(set VR256:$dst, @@ -8403,12 +8404,15 @@ multiclass avx2_perm<bits<8> opc, string OpcodeStr, PatFrag mem_frag, } } -defm VPERMD : avx2_perm<0x36, "vpermd", loadv4i64, v8i32, WriteShuffle256>; +defm VPERMD : avx2_perm<0x36, "vpermd", loadv4i64, v8i32, WriteShuffle256, + i256mem>; let ExeDomain = SSEPackedSingle in -defm VPERMPS : avx2_perm<0x16, "vpermps", loadv8f32, v8f32, WriteFShuffle256>; +defm VPERMPS : avx2_perm<0x16, "vpermps", loadv8f32, v8f32, WriteFShuffle256, + f256mem>; multiclass avx2_perm_imm<bits<8> opc, string OpcodeStr, PatFrag mem_frag, - ValueType OpVT, X86FoldableSchedWrite Sched> { + ValueType OpVT, X86FoldableSchedWrite Sched, + X86MemOperand memOp> { let Predicates = [HasAVX2, NoVLX] in { def Yri : AVX2AIi8<opc, MRMSrcReg, (outs VR256:$dst), (ins VR256:$src1, u8imm:$src2), @@ -8418,7 +8422,7 @@ multiclass avx2_perm_imm<bits<8> opc, string OpcodeStr, PatFrag mem_frag, (OpVT (X86VPermi VR256:$src1, (i8 imm:$src2))))]>, Sched<[Sched]>, VEX, VEX_L; def Ymi : AVX2AIi8<opc, MRMSrcMem, (outs VR256:$dst), - (ins i256mem:$src1, u8imm:$src2), + (ins memOp:$src1, u8imm:$src2), !strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}"), [(set VR256:$dst, @@ -8429,10 +8433,10 @@ multiclass avx2_perm_imm<bits<8> opc, string OpcodeStr, PatFrag mem_frag, } defm VPERMQ : avx2_perm_imm<0x00, "vpermq", loadv4i64, v4i64, - WriteShuffle256>, VEX_W; + WriteShuffle256, i256mem>, VEX_W; let ExeDomain = SSEPackedDouble in defm VPERMPD : avx2_perm_imm<0x01, "vpermpd", loadv4f64, v4f64, - WriteFShuffle256>, VEX_W; + WriteFShuffle256, f256mem>, VEX_W; //===----------------------------------------------------------------------===// // VPERM2I128 - Permute Floating-Point Values in 128-bit chunks |
