diff options
Diffstat (limited to 'lib/Target/X86/X86InstrSSE.td')
| -rw-r--r-- | lib/Target/X86/X86InstrSSE.td | 63 |
1 files changed, 36 insertions, 27 deletions
diff --git a/lib/Target/X86/X86InstrSSE.td b/lib/Target/X86/X86InstrSSE.td index c66240f..6be366b 100644 --- a/lib/Target/X86/X86InstrSSE.td +++ b/lib/Target/X86/X86InstrSSE.td @@ -3824,51 +3824,51 @@ let ExeDomain = SSEPackedInt in { let Predicates = [HasAVX] in { def : Pat<(int_x86_sse2_psll_dq VR128:$src1, imm:$src2), - (v2i64 (VPSLLDQri VR128:$src1, (BYTE_imm imm:$src2)))>; + (VPSLLDQri VR128:$src1, (BYTE_imm imm:$src2))>; def : Pat<(int_x86_sse2_psrl_dq VR128:$src1, imm:$src2), - (v2i64 (VPSRLDQri VR128:$src1, (BYTE_imm imm:$src2)))>; + (VPSRLDQri VR128:$src1, (BYTE_imm imm:$src2))>; def : Pat<(int_x86_sse2_psll_dq_bs VR128:$src1, imm:$src2), - (v2i64 (VPSLLDQri VR128:$src1, imm:$src2))>; + (VPSLLDQri VR128:$src1, imm:$src2)>; def : Pat<(int_x86_sse2_psrl_dq_bs VR128:$src1, imm:$src2), - (v2i64 (VPSRLDQri VR128:$src1, imm:$src2))>; + (VPSRLDQri VR128:$src1, imm:$src2)>; def : Pat<(v2f64 (X86fsrl VR128:$src1, i32immSExt8:$src2)), - (v2f64 (VPSRLDQri VR128:$src1, (BYTE_imm imm:$src2)))>; + (VPSRLDQri VR128:$src1, (BYTE_imm imm:$src2))>; // Shift up / down and insert zero's. def : Pat<(v2i64 (X86vshl VR128:$src, (i8 imm:$amt))), - (v2i64 (VPSLLDQri VR128:$src, (BYTE_imm imm:$amt)))>; + (VPSLLDQri VR128:$src, (BYTE_imm imm:$amt))>; def : Pat<(v2i64 (X86vshr VR128:$src, (i8 imm:$amt))), - (v2i64 (VPSRLDQri VR128:$src, (BYTE_imm imm:$amt)))>; + (VPSRLDQri VR128:$src, (BYTE_imm imm:$amt))>; } let Predicates = [HasAVX2] in { def : Pat<(int_x86_avx2_psll_dq VR256:$src1, imm:$src2), - (v4i64 (VPSLLDQYri VR256:$src1, (BYTE_imm imm:$src2)))>; + (VPSLLDQYri VR256:$src1, (BYTE_imm imm:$src2))>; def : Pat<(int_x86_avx2_psrl_dq VR256:$src1, imm:$src2), - (v4i64 (VPSRLDQYri VR256:$src1, (BYTE_imm imm:$src2)))>; + (VPSRLDQYri VR256:$src1, (BYTE_imm imm:$src2))>; def : Pat<(int_x86_avx2_psll_dq_bs VR256:$src1, imm:$src2), - (v4i64 (VPSLLDQYri VR256:$src1, imm:$src2))>; + (VPSLLDQYri VR256:$src1, imm:$src2)>; def : Pat<(int_x86_avx2_psrl_dq_bs VR256:$src1, imm:$src2), - (v4i64 (VPSRLDQYri VR256:$src1, imm:$src2))>; + (VPSRLDQYri VR256:$src1, imm:$src2)>; } let Predicates = [HasSSE2] in { def : Pat<(int_x86_sse2_psll_dq VR128:$src1, imm:$src2), - (v2i64 (PSLLDQri VR128:$src1, (BYTE_imm imm:$src2)))>; + (PSLLDQri VR128:$src1, (BYTE_imm imm:$src2))>; def : Pat<(int_x86_sse2_psrl_dq VR128:$src1, imm:$src2), - (v2i64 (PSRLDQri VR128:$src1, (BYTE_imm imm:$src2)))>; + (PSRLDQri VR128:$src1, (BYTE_imm imm:$src2))>; def : Pat<(int_x86_sse2_psll_dq_bs VR128:$src1, imm:$src2), - (v2i64 (PSLLDQri VR128:$src1, imm:$src2))>; + (PSLLDQri VR128:$src1, imm:$src2)>; def : Pat<(int_x86_sse2_psrl_dq_bs VR128:$src1, imm:$src2), - (v2i64 (PSRLDQri VR128:$src1, imm:$src2))>; + (PSRLDQri VR128:$src1, imm:$src2)>; def : Pat<(v2f64 (X86fsrl VR128:$src1, i32immSExt8:$src2)), - (v2f64 (PSRLDQri VR128:$src1, (BYTE_imm imm:$src2)))>; + (PSRLDQri VR128:$src1, (BYTE_imm imm:$src2))>; // Shift up / down and insert zero's. def : Pat<(v2i64 (X86vshl VR128:$src, (i8 imm:$amt))), - (v2i64 (PSLLDQri VR128:$src, (BYTE_imm imm:$amt)))>; + (PSLLDQri VR128:$src, (BYTE_imm imm:$amt))>; def : Pat<(v2i64 (X86vshr VR128:$src, (i8 imm:$amt))), - (v2i64 (PSRLDQri VR128:$src, (BYTE_imm imm:$amt)))>; + (PSRLDQri VR128:$src, (BYTE_imm imm:$amt))>; } //===---------------------------------------------------------------------===// @@ -5316,11 +5316,11 @@ let isCommutable = 0 in { int_x86_avx2_pmadd_ub_sw>, VEX_4V; defm VPSHUFB : SS3I_binop_rm_int_y<0x00, "vpshufb", memopv32i8, int_x86_avx2_pshuf_b>, VEX_4V; - defm VPSIGNB : SS3I_binop_rm_int_y<0x08, "vpsignb", memopv16i8, + defm VPSIGNB : SS3I_binop_rm_int_y<0x08, "vpsignb", memopv32i8, int_x86_avx2_psign_b>, VEX_4V; - defm VPSIGNW : SS3I_binop_rm_int_y<0x09, "vpsignw", memopv8i16, + defm VPSIGNW : SS3I_binop_rm_int_y<0x09, "vpsignw", memopv16i16, int_x86_avx2_psign_w>, VEX_4V; - defm VPSIGND : SS3I_binop_rm_int_y<0x0A, "vpsignd", memopv4i32, + defm VPSIGND : SS3I_binop_rm_int_y<0x0A, "vpsignd", memopv8i32, int_x86_avx2_psign_d>, VEX_4V; } defm VPMULHRSW : SS3I_binop_rm_int_y<0x0B, "vpmulhrsw", memopv16i16, @@ -5363,11 +5363,11 @@ let Predicates = [HasSSSE3] in { def : Pat<(X86pshufb VR128:$src, (bc_v16i8 (memopv2i64 addr:$mask))), (PSHUFBrm128 VR128:$src, addr:$mask)>; - def : Pat<(X86psignb VR128:$src1, VR128:$src2), + def : Pat<(v16i8 (X86psignb VR128:$src1, VR128:$src2)), (PSIGNBrr128 VR128:$src1, VR128:$src2)>; - def : Pat<(X86psignw VR128:$src1, VR128:$src2), + def : Pat<(v8i16 (X86psignw VR128:$src1, VR128:$src2)), (PSIGNWrr128 VR128:$src1, VR128:$src2)>; - def : Pat<(X86psignd VR128:$src1, VR128:$src2), + def : Pat<(v4i32 (X86psignd VR128:$src1, VR128:$src2)), (PSIGNDrr128 VR128:$src1, VR128:$src2)>; } @@ -5377,14 +5377,23 @@ let Predicates = [HasAVX] in { def : Pat<(X86pshufb VR128:$src, (bc_v16i8 (memopv2i64 addr:$mask))), (VPSHUFBrm128 VR128:$src, addr:$mask)>; - def : Pat<(X86psignb VR128:$src1, VR128:$src2), + def : Pat<(v16i8 (X86psignb VR128:$src1, VR128:$src2)), (VPSIGNBrr128 VR128:$src1, VR128:$src2)>; - def : Pat<(X86psignw VR128:$src1, VR128:$src2), + def : Pat<(v8i16 (X86psignw VR128:$src1, VR128:$src2)), (VPSIGNWrr128 VR128:$src1, VR128:$src2)>; - def : Pat<(X86psignd VR128:$src1, VR128:$src2), + def : Pat<(v4i32 (X86psignd VR128:$src1, VR128:$src2)), (VPSIGNDrr128 VR128:$src1, VR128:$src2)>; } +let Predicates = [HasAVX2] in { + def : Pat<(v32i8 (X86psignb VR256:$src1, VR256:$src2)), + (VPSIGNBrr256 VR256:$src1, VR256:$src2)>; + def : Pat<(v16i16 (X86psignw VR256:$src1, VR256:$src2)), + (VPSIGNWrr256 VR256:$src1, VR256:$src2)>; + def : Pat<(v8i32 (X86psignd VR256:$src1, VR256:$src2)), + (VPSIGNDrr256 VR256:$src1, VR256:$src2)>; +} + //===---------------------------------------------------------------------===// // SSSE3 - Packed Align Instruction Patterns //===---------------------------------------------------------------------===// |