wangtao
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

[X86] AVX512: Generate Pat<>'s for the vpermt2* intrinsics via multiclass 

This new multiclass, avx512_perm_table_3src derives from the current one and
provides the Pat<>.  The next patch will add another Pat<> that uses the
writemask.

Note that I dropped the type annotation from the intrinsic call, i.e.: (v16f32
VR512:$src1) -> R512:$src1.  I think that this should be fine (at least many
intrinsic calls don't provide them) and it greatly reduces the number of
template arguments.

llvm-svn: 212222
This commit is contained in:
Adam Nemet 2014-07-02 21:25:58 +00:00
parent 2415a497b5
commit efe9c98a16
1 changed files with 14 additions and 19 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

33
llvm/lib/Target/X86/X86InstrAVX512.td
View File

@ -691,30 +691,25 @@ defm VPERMI2PD : avx512_perm_3src<0x77, "vpermi2pd", VR512, memopv8f64,
i512mem, X86VPermiv3, v8f64, VK8WM>,
EVEX_V512, VEX_W, EVEX_CD8<64, CD8VF>;
defm VPERMT2D : avx512_perm_3src<0x7E, "vpermt2d", VR512, memopv16i32, i512mem,
multiclass avx512_perm_table_3src<bits<8> opc, string Suffix, RegisterClass RC,
PatFrag mem_frag, X86MemOperand x86memop,
SDNode OpNode, ValueType OpVT, RegisterClass KRC> :
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)>;
}
defm VPERMT2D : avx512_perm_table_3src<0x7E, "d", VR512, memopv16i32, i512mem,
X86VPermv3, v16i32, VK16WM>, EVEX_V512, EVEX_CD8<32, CD8VF>;
defm VPERMT2Q : avx512_perm_3src<0x7E, "vpermt2q", VR512, memopv8i64, i512mem,
defm VPERMT2Q : avx512_perm_table_3src<0x7E, "q", VR512, memopv8i64, i512mem,
X86VPermv3, v8i64, VK8WM>, EVEX_V512, VEX_W, EVEX_CD8<64, CD8VF>;
defm VPERMT2PS : avx512_perm_3src<0x7F, "vpermt2ps", VR512, memopv16f32, i512mem,
defm VPERMT2PS : avx512_perm_table_3src<0x7F, "ps", VR512, memopv16f32, i512mem,
X86VPermv3, v16f32, VK16WM>, EVEX_V512, EVEX_CD8<32, CD8VF>;
defm VPERMT2PD : avx512_perm_3src<0x7F, "vpermt2pd", VR512, memopv8f64, i512mem,
defm VPERMT2PD : avx512_perm_table_3src<0x7F, "pd", VR512, memopv8f64, i512mem,
X86VPermv3, v8f64, VK8WM>, 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)>;
//===----------------------------------------------------------------------===//
// AVX-512 - BLEND using mask
//