RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-11-23 19:23:23 +01:00
Code Issues Projects Releases Wiki Activity

Eliminate sub_ss, sub_sd from broadcast patterns.

Browse Source 
The (COPY_TO_REGCLASS GR32:$src, VR128) pattern looks odd, but
copyPhysReg does the right thing with it. (The old pattern would
eventually produce the same cross-class copy).

llvm-svn: 160830
This commit is contained in:
Jakob Stoklund Olesen 2012-07-26 22:59:06 +00:00
parent 8d971d19cb
commit c1dd7a213d
1 changed files with 16 additions and 42 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

58
lib/Target/X86/X86InstrSSE.td
View File

@ -7704,24 +7704,18 @@ let Predicates = [HasAVX2] in {
// is used by additional users, which prevents the pattern selection.
let AddedComplexity = 20 in {
def : Pat<(v4f32 (X86VBroadcast FR32:$src)),
(VBROADCASTSSrr
(INSERT_SUBREG (v4f32 (IMPLICIT_DEF)), FR32:$src, sub_ss))>;
(VBROADCASTSSrr (COPY_TO_REGCLASS FR32:$src, VR128))>;
def : Pat<(v8f32 (X86VBroadcast FR32:$src)),
(VBROADCASTSSYrr
(INSERT_SUBREG (v4f32 (IMPLICIT_DEF)), FR32:$src, sub_ss))>;
(VBROADCASTSSYrr (COPY_TO_REGCLASS FR32:$src, VR128))>;
def : Pat<(v4f64 (X86VBroadcast FR64:$src)),
(VBROADCASTSDYrr
(INSERT_SUBREG (v2f64 (IMPLICIT_DEF)), FR64:$src, sub_sd))>;
(VBROADCASTSDYrr (COPY_TO_REGCLASS FR64:$src, VR128))>;
def : Pat<(v4i32 (X86VBroadcast GR32:$src)),
(VBROADCASTSSrr
(INSERT_SUBREG (v4i32 (IMPLICIT_DEF)), GR32:$src, sub_ss))>;
(VBROADCASTSSrr (COPY_TO_REGCLASS GR32:$src, VR128))>;
def : Pat<(v8i32 (X86VBroadcast GR32:$src)),
(VBROADCASTSSYrr
(INSERT_SUBREG (v4i32 (IMPLICIT_DEF)), GR32:$src, sub_ss))>;
(VBROADCASTSSYrr (COPY_TO_REGCLASS GR32:$src, VR128))>;
def : Pat<(v4i64 (X86VBroadcast GR64:$src)),
(VBROADCASTSDYrr
(INSERT_SUBREG (v2i64 (IMPLICIT_DEF)), GR64:$src, sub_sd))>;
(VBROADCASTSDYrr (COPY_TO_REGCLASS GR64:$src, VR128))>;
}
}
@ -7745,46 +7739,26 @@ def : Pat<(v4i32 (X86VBroadcast (loadi32 addr:$src))),
let AddedComplexity = 20 in {
// 128bit broadcasts:
def : Pat<(v4f32 (X86VBroadcast FR32:$src)),
(VPSHUFDri
(INSERT_SUBREG (v4f32 (IMPLICIT_DEF)), FR32:$src, sub_ss), 0)>;
(VPSHUFDri (COPY_TO_REGCLASS FR32:$src, VR128), 0)>;
def : Pat<(v8f32 (X86VBroadcast FR32:$src)),
(VINSERTF128rr (INSERT_SUBREG (v8f32 (IMPLICIT_DEF)),
(VPSHUFDri
(INSERT_SUBREG (v4f32 (IMPLICIT_DEF)), FR32:$src, sub_ss), 0),
sub_xmm),
(VPSHUFDri
(INSERT_SUBREG (v4f32 (IMPLICIT_DEF)), FR32:$src, sub_ss),
0), 1)>;
(VPSHUFDri (COPY_TO_REGCLASS FR32:$src, VR128), 0), sub_xmm),
(VPSHUFDri (COPY_TO_REGCLASS FR32:$src, VR128), 0), 1)>;
def : Pat<(v4f64 (X86VBroadcast FR64:$src)),
(VINSERTF128rr (INSERT_SUBREG (v4f64 (IMPLICIT_DEF)),
(VPSHUFDri
(INSERT_SUBREG (v2f64 (IMPLICIT_DEF)), FR64:$src, sub_sd),
0x44),
sub_xmm),
(VPSHUFDri
(INSERT_SUBREG (v2f64 (IMPLICIT_DEF)), FR64:$src, sub_sd),
0x44), 1)>;
(VPSHUFDri (COPY_TO_REGCLASS FR64:$src, VR128), 0x44), sub_xmm),
(VPSHUFDri (COPY_TO_REGCLASS FR64:$src, VR128), 0x44), 1)>;
def : Pat<(v4i32 (X86VBroadcast GR32:$src)),
(VPSHUFDri
(INSERT_SUBREG (v4i32 (IMPLICIT_DEF)), GR32:$src, sub_ss), 0)>;
(VPSHUFDri (COPY_TO_REGCLASS GR32:$src, VR128), 0)>;
def : Pat<(v8i32 (X86VBroadcast GR32:$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)>;
(VPSHUFDri (COPY_TO_REGCLASS GR32:$src, VR128), 0), sub_xmm),
(VPSHUFDri (COPY_TO_REGCLASS GR32:$src, VR128), 0), 1)>;
def : Pat<(v4i64 (X86VBroadcast GR64:$src)),
(VINSERTF128rr (INSERT_SUBREG (v4i64 (IMPLICIT_DEF)),
(VPSHUFDri
(INSERT_SUBREG (v2i64 (IMPLICIT_DEF)), GR64:$src, sub_sd),
0x44),
sub_xmm),
(VPSHUFDri
(INSERT_SUBREG (v2i64 (IMPLICIT_DEF)), GR64:$src, sub_sd),
0x44), 1)>;
(VPSHUFDri (COPY_TO_REGCLASS GR64:$src, VR128), 0x44), sub_xmm),
(VPSHUFDri (COPY_TO_REGCLASS GR64:$src, VR128), 0x44), 1)>;
}
}