diff options
Diffstat (limited to 'lib/Target/X86/X86InstrSSE.td')
| -rw-r--r-- | lib/Target/X86/X86InstrSSE.td | 57 |
1 files changed, 54 insertions, 3 deletions
diff --git a/lib/Target/X86/X86InstrSSE.td b/lib/Target/X86/X86InstrSSE.td index 384238741b188..cce938baafe1c 100644 --- a/lib/Target/X86/X86InstrSSE.td +++ b/lib/Target/X86/X86InstrSSE.td @@ -4462,12 +4462,12 @@ def MOVPDI2DImr : PDI<0x7E, MRMDestMem, (outs), (ins i32mem:$dst, VR128:$src), // Move Packed Doubleword Int first element to Doubleword Int // let SchedRW = [WriteMove] in { -def VMOVPQIto64rr : I<0x7E, MRMDestReg, (outs GR64:$dst), (ins VR128:$src), - "vmov{d|q}\t{$src, $dst|$dst, $src}", +def VMOVPQIto64rr : VRPDI<0x7E, MRMDestReg, (outs GR64:$dst), (ins VR128:$src), + "mov{d|q}\t{$src, $dst|$dst, $src}", [(set GR64:$dst, (vector_extract (v2i64 VR128:$src), (iPTR 0)))], IIC_SSE_MOVD_ToGP>, - TB, OpSize, VEX, VEX_W, Requires<[HasAVX, In64BitMode]>; + VEX; def MOVPQIto64rr : RPDI<0x7E, MRMDestReg, (outs GR64:$dst), (ins VR128:$src), "mov{d|q}\t{$src, $dst|$dst, $src}", @@ -5094,6 +5094,16 @@ multiclass SS3I_unop_rm_int_y<bits<8> opc, string OpcodeStr, Sched<[WriteVecALULd]>; } +// Helper fragments to match sext vXi1 to vXiY. +def v16i1sextv16i8 : PatLeaf<(v16i8 (X86pcmpgt (bc_v16i8 (v4i32 immAllZerosV)), + VR128:$src))>; +def v8i1sextv8i16 : PatLeaf<(v8i16 (X86vsrai VR128:$src, (i32 15)))>; +def v4i1sextv4i32 : PatLeaf<(v4i32 (X86vsrai VR128:$src, (i32 31)))>; +def v32i1sextv32i8 : PatLeaf<(v32i8 (X86pcmpgt (bc_v32i8 (v8i32 immAllZerosV)), + VR256:$src))>; +def v16i1sextv16i16: PatLeaf<(v16i16 (X86vsrai VR256:$src, (i32 15)))>; +def v8i1sextv8i32 : PatLeaf<(v8i32 (X86vsrai VR256:$src, (i32 31)))>; + let Predicates = [HasAVX] in { defm VPABSB : SS3I_unop_rm_int<0x1C, "vpabsb", int_x86_ssse3_pabs_b_128>, VEX; @@ -5101,6 +5111,19 @@ let Predicates = [HasAVX] in { int_x86_ssse3_pabs_w_128>, VEX; defm VPABSD : SS3I_unop_rm_int<0x1E, "vpabsd", int_x86_ssse3_pabs_d_128>, VEX; + + def : Pat<(xor + (bc_v2i64 (v16i1sextv16i8)), + (bc_v2i64 (add (v16i8 VR128:$src), (v16i1sextv16i8)))), + (VPABSBrr128 VR128:$src)>; + def : Pat<(xor + (bc_v2i64 (v8i1sextv8i16)), + (bc_v2i64 (add (v8i16 VR128:$src), (v8i1sextv8i16)))), + (VPABSWrr128 VR128:$src)>; + def : Pat<(xor + (bc_v2i64 (v4i1sextv4i32)), + (bc_v2i64 (add (v4i32 VR128:$src), (v4i1sextv4i32)))), + (VPABSDrr128 VR128:$src)>; } let Predicates = [HasAVX2] in { @@ -5110,6 +5133,19 @@ let Predicates = [HasAVX2] in { int_x86_avx2_pabs_w>, VEX, VEX_L; defm VPABSD : SS3I_unop_rm_int_y<0x1E, "vpabsd", int_x86_avx2_pabs_d>, VEX, VEX_L; + + def : Pat<(xor + (bc_v4i64 (v32i1sextv32i8)), + (bc_v4i64 (add (v32i8 VR256:$src), (v32i1sextv32i8)))), + (VPABSBrr256 VR256:$src)>; + def : Pat<(xor + (bc_v4i64 (v16i1sextv16i16)), + (bc_v4i64 (add (v16i16 VR256:$src), (v16i1sextv16i16)))), + (VPABSWrr256 VR256:$src)>; + def : Pat<(xor + (bc_v4i64 (v8i1sextv8i32)), + (bc_v4i64 (add (v8i32 VR256:$src), (v8i1sextv8i32)))), + (VPABSDrr256 VR256:$src)>; } defm PABSB : SS3I_unop_rm_int<0x1C, "pabsb", @@ -5119,6 +5155,21 @@ defm PABSW : SS3I_unop_rm_int<0x1D, "pabsw", defm PABSD : SS3I_unop_rm_int<0x1E, "pabsd", int_x86_ssse3_pabs_d_128>; +let Predicates = [HasSSSE3] in { + def : Pat<(xor + (bc_v2i64 (v16i1sextv16i8)), + (bc_v2i64 (add (v16i8 VR128:$src), (v16i1sextv16i8)))), + (PABSBrr128 VR128:$src)>; + def : Pat<(xor + (bc_v2i64 (v8i1sextv8i16)), + (bc_v2i64 (add (v8i16 VR128:$src), (v8i1sextv8i16)))), + (PABSWrr128 VR128:$src)>; + def : Pat<(xor + (bc_v2i64 (v4i1sextv4i32)), + (bc_v2i64 (add (v4i32 VR128:$src), (v4i1sextv4i32)))), + (PABSDrr128 VR128:$src)>; +} + //===---------------------------------------------------------------------===// // SSSE3 - Packed Binary Operator Instructions //===---------------------------------------------------------------------===// |