highperfocused/ffmpeg
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

x86/dsputil: use HADDD where applicable

Browse Source 
Signed-off-by: James Almer <jamrial@gmail.com>
Reviewed-by: "Ronald S. Bultje" <rsbultje@gmail.com>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
This commit is contained in:
James Almer
2014-04-11 20:00:21 -03:00
committed by Michael Niedermayer
parent 76ed71a72b
commit 197fe392db
1 changed files with 3 additions and 20 deletions
Download Patch File Download Diff File Expand all files Collapse all files

23
libavcodec/x86/dsputil.asm
View File

@@ -50,14 +50,7 @@ cglobal scalarproduct_int16, 3,3,3, v1, v2, order
paddd m2, m1 paddd m2, m1
add orderq, mmsize*2 add orderq, mmsize*2
jl .loop jl .loop
%if mmsize == 16 HADDD m2, m0
movhlps m0, m2
paddd m2, m0
pshuflw m0, m2, 0x4e
%else
pshufw m0, m2, 0x4e
%endif
paddd m2, m0
movd eax, m2 movd eax, m2
%if mmsize == 8 %if mmsize == 8
emms emms
@@ -99,14 +92,7 @@ cglobal scalarproduct_and_madd_int16, 4,4,8, v1, v2, v3, order, mul
mova [v1q + orderq + mmsize], m3 mova [v1q + orderq + mmsize], m3
add orderq, mmsize*2 add orderq, mmsize*2
jl .loop jl .loop
%if mmsize == 16 HADDD m6, m0
movhlps m0, m6
paddd m6, m0
pshuflw m0, m6, 0x4e
%else
pshufw m0, m6, 0x4e
%endif
paddd m6, m0
movd eax, m6 movd eax, m6
RET RET
%endmacro %endmacro
@@ -200,10 +186,7 @@ SCALARPRODUCT_LOOP 4
SCALARPRODUCT_LOOP 2 SCALARPRODUCT_LOOP 2
SCALARPRODUCT_LOOP 0 SCALARPRODUCT_LOOP 0
.end: .end:
movhlps m0, m6 HADDD m6, m0
paddd m6, m0
pshuflw m0, m6, 0x4e
paddd m6, m0
movd eax, m6 movd eax, m6
RET RET