Skip to content

F
ffmpeg-mt
Commit f275de9e authored by michael's avatar michael
Browse files
Options

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

mmx code
this also makes the affected code 4% faster


git-svn-id: file:///var/local/repositories/ffmpeg/trunk@10156 9553f0bf-9b14-0410-a0b8-cfaf0461ba5b
parent e664f648
Show whitespace changes
Inline Side-by-side
Showing with 15 additions and 18 deletions
libavcodec/i386/snowdsp_mmx.c
View file @ f275de9e
...@@ -294,9 +294,10 @@ void ff_snow_horizontal_compose97i_mmx(DWTELEM *b, int width){ ...@@ -294,9 +294,10 @@ void ff_snow_horizontal_compose97i_mmx(DWTELEM *b, int width){
DWTELEM * const ref = b+w2 - 1; DWTELEM * const ref = b+w2 - 1;
i = 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( 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){ for(; i<w_l-3; i+=4){
asm volatile( asm volatile(
...@@ -304,22 +305,18 @@ void ff_snow_horizontal_compose97i_mmx(DWTELEM *b, int width){ ...@@ -304,22 +305,18 @@ void ff_snow_horizontal_compose97i_mmx(DWTELEM *b, int width){
"movq 8(%1), %%mm4 \n\t" "movq 8(%1), %%mm4 \n\t"
"paddd 4(%1), %%mm0 \n\t" "paddd 4(%1), %%mm0 \n\t"
"paddd 12(%1), %%mm4 \n\t" "paddd 12(%1), %%mm4 \n\t"
"movq %%mm7, %%mm1 \n\t" "paddd %%mm7, %%mm0 \n\t"
"movq %%mm7, %%mm5 \n\t" "paddd %%mm7, %%mm4 \n\t"
"psubd %%mm0, %%mm1 \n\t" "psrad $2, %%mm0 \n\t"
"psubd %%mm4, %%mm5 \n\t" "psrad $2, %%mm4 \n\t"
"movq (%0), %%mm0 \n\t" "movq (%0), %%mm1 \n\t"
"movq 8(%0), %%mm4 \n\t" "movq 8(%0), %%mm5 \n\t"
"pslld $2, %%mm0 \n\t" "paddd %%mm1, %%mm0 \n\t"
"pslld $2, %%mm4 \n\t" "paddd %%mm5, %%mm4 \n\t"
"psubd %%mm0, %%mm1 \n\t" "psrad $2, %%mm0 \n\t"
"psubd %%mm4, %%mm5 \n\t" "psrad $2, %%mm4 \n\t"
"psrad $4, %%mm1 \n\t" "paddd %%mm1, %%mm0 \n\t"
"psrad $4, %%mm5 \n\t" "paddd %%mm5, %%mm4 \n\t"
"movq (%0), %%mm0 \n\t"
"movq 8(%0), %%mm4 \n\t"
"psubd %%mm1, %%mm0 \n\t"
"psubd %%mm5, %%mm4 \n\t"
"movq %%mm0, (%0) \n\t" "movq %%mm0, (%0) \n\t"
"movq %%mm4, 8(%0) \n\t" "movq %%mm4, 8(%0) \n\t"
:: "r"(&b[i]), "r"(&ref[i]) :: "r"(&b[i]), "r"(&ref[i])
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment