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

AVX: Add additional vbroadcast replacement sequences for integers. 

Remove the v2f64 patterns because it does not match any vbroadcast
instruction.

llvm-svn: 155461
This commit is contained in:
Nadav Rotem 2012-04-24 18:09:59 +00:00
parent 26bdff9b82
commit 810734b7f4
1 changed files with 30 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

33
llvm/lib/Target/X86/X86InstrSSE.td
View File

@ -7736,6 +7736,16 @@ let Predicates = [HasAVX2] in {
def : Pat<(v4f64 (X86VBroadcast FR64:$src)),
(VBROADCASTSDrr
(INSERT_SUBREG (v2f64 (IMPLICIT_DEF)), FR64:$src, sub_sd))>;
def : Pat<(v4i32 (X86VBroadcast GR32:$src)),
(VBROADCASTSSrr
(INSERT_SUBREG (v4i32 (IMPLICIT_DEF)), GR32:$src, sub_ss))>;
def : Pat<(v8i32 (X86VBroadcast GR32:$src)),
(VBROADCASTSSYrr
(INSERT_SUBREG (v4i32 (IMPLICIT_DEF)), GR32:$src, sub_ss))>;
def : Pat<(v4i64 (X86VBroadcast GR64:$src)),
(VBROADCASTSDrr
(INSERT_SUBREG (v2i64 (IMPLICIT_DEF)), GR64:$src, sub_sd))>;
}
}
@ -7758,9 +7768,6 @@ def : Pat<(v4i32 (X86VBroadcast (loadi32 addr:$src))),
// is used by additional users, which prevents the pattern selection.
let AddedComplexity = 20 in {
// 128bit broadcasts:
def : Pat<(v2f64 (X86VBroadcast FR64:$src)),
(VPSHUFDri
(INSERT_SUBREG (v2f64 (IMPLICIT_DEF)), FR64:$src, sub_sd), 0)>;
def : Pat<(v4f32 (X86VBroadcast FR32:$src)),
(VPSHUFDri
(INSERT_SUBREG (v4f32 (IMPLICIT_DEF)), FR32:$src, sub_ss), 0)>;
@ -7780,6 +7787,26 @@ def : Pat<(v4i32 (X86VBroadcast (loadi32 addr:$src))),
(VPSHUFDri
(INSERT_SUBREG (v2f64 (IMPLICIT_DEF)), FR64:$src, sub_sd),
0), 1)>;
def : Pat<(v4i32 (X86VBroadcast GR32:$src)),
(VPSHUFDri
(INSERT_SUBREG (v4i32 (IMPLICIT_DEF)), GR32:$src, sub_ss), 0)>;
def : Pat<(v8i32 (X86VBroadcast GR32:$src)),
(VINSERTF128rr (INSERT_SUBREG (v8i32 (IMPLICIT_DEF)),
(VPSHUFDri
(INSERT_SUBREG (v4i32 (IMPLICIT_DEF)), GR32:$src, sub_ss), 0),
sub_xmm),
(VPSHUFDri
(INSERT_SUBREG (v4i32 (IMPLICIT_DEF)), GR32:$src, sub_ss),
0), 1)>;
def : Pat<(v4i64 (X86VBroadcast GR64:$src)),
(VINSERTF128rr (INSERT_SUBREG (v4i64 (IMPLICIT_DEF)),
(VPSHUFDri
(INSERT_SUBREG (v2i64 (IMPLICIT_DEF)), GR64:$src, sub_sd), 0),
sub_xmm),
(VPSHUFDri
(INSERT_SUBREG (v2i64 (IMPLICIT_DEF)), GR64:$src, sub_sd),
0), 1)>;
}
}