virtualenv/FFmpeg
1
0
Fork 0
mirror of https://github.com/FFmpeg/FFmpeg.git synced 2024-11-21 10:55:51 +02:00
Code Issues Releases Activity

and of course the unneeded double subtractions were blindly put in the

Browse Source 
mmx code
this also makes the affected code 4% faster

Originally committed as revision 10156 to svn://svn.ffmpeg.org/ffmpeg/trunk
This commit is contained in:
Michael Niedermayer 2007-08-20 22:29:21 +00:00
parent f29bd6fa71
commit 3a9f44d5d5
1 changed files with 15 additions and 18 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

33
libavcodec/i386/snowdsp_mmx.c
View File

@ -294,9 +294,10 @@ void ff_snow_horizontal_compose97i_mmx(DWTELEM *b, int width){
DWTELEM * const ref = b+w2 - 1;
i = 1;
b[0] = b[0] - (((-2 * ref[1] + W_BO) - 4 * b[0]) >> W_BS);
b[0] = b[0] + (((2 * ref[1] + W_BO-1) + 4 * b[0]) >> W_BS);
asm volatile(
"pslld $1, %%mm7 \n\t" /* xmm7 already holds a '4' from 2 lifts ago. */
"pcmpeqd %%mm7, %%mm7 \n\t"
"psrld $29, %%mm7 \n\t"
::);
for(; i<w_l-3; i+=4){
asm volatile(
@ -304,22 +305,18 @@ void ff_snow_horizontal_compose97i_mmx(DWTELEM *b, int width){
"movq 8(%1), %%mm4 \n\t"
"paddd 4(%1), %%mm0 \n\t"
"paddd 12(%1), %%mm4 \n\t"
"movq %%mm7, %%mm1 \n\t"
"movq %%mm7, %%mm5 \n\t"
"psubd %%mm0, %%mm1 \n\t"
"psubd %%mm4, %%mm5 \n\t"
"movq (%0), %%mm0 \n\t"
"movq 8(%0), %%mm4 \n\t"
"pslld $2, %%mm0 \n\t"
"pslld $2, %%mm4 \n\t"
"psubd %%mm0, %%mm1 \n\t"
"psubd %%mm4, %%mm5 \n\t"
"psrad $4, %%mm1 \n\t"
"psrad $4, %%mm5 \n\t"
"movq (%0), %%mm0 \n\t"
"movq 8(%0), %%mm4 \n\t"
"psubd %%mm1, %%mm0 \n\t"
"psubd %%mm5, %%mm4 \n\t"
"paddd %%mm7, %%mm0 \n\t"
"paddd %%mm7, %%mm4 \n\t"
"psrad $2, %%mm0 \n\t"
"psrad $2, %%mm4 \n\t"
"movq (%0), %%mm1 \n\t"
"movq 8(%0), %%mm5 \n\t"
"paddd %%mm1, %%mm0 \n\t"
"paddd %%mm5, %%mm4 \n\t"
"psrad $2, %%mm0 \n\t"
"psrad $2, %%mm4 \n\t"
"paddd %%mm1, %%mm0 \n\t"
"paddd %%mm5, %%mm4 \n\t"
"movq %%mm0, (%0) \n\t"
"movq %%mm4, 8(%0) \n\t"
:: "r"(&b[i]), "r"(&ref[i])