generic-library/ffmpeg
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

gmc_mmx tweaks

Browse Source 
Originally committed as revision 5269 to svn://svn.ffmpeg.org/ffmpeg/trunk
This commit is contained in:
Loren Merritt 2006-04-05 04:13:41 +00:00
parent 87b041e0f0
commit 75ca1a5f70
1 changed files with 11 additions and 13 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

24
libavcodec/i386/dsputil_mmx.c
View File

@ -2406,7 +2406,6 @@ static void just_return() { return; }
static void gmc_mmx(uint8_t *dst, uint8_t *src, int stride, int h, int ox, int oy, static void gmc_mmx(uint8_t *dst, uint8_t *src, int stride, int h, int ox, int oy,
int dxx, int dxy, int dyx, int dyy, int shift, int r, int width, int height){ int dxx, int dxy, int dyx, int dyy, int shift, int r, int width, int height){
const int w = 8; const int w = 8;
const int s = 1<<shift;
const int ix = ox>>(16+shift); const int ix = ox>>(16+shift);
const int iy = oy>>(16+shift); const int iy = oy>>(16+shift);
const int oxs = ox>>4; const int oxs = ox>>4;
@ -2437,14 +2436,21 @@ static void gmc_mmx(uint8_t *dst, uint8_t *src, int stride, int h, int ox, int o
return; return;
} }
src += ix + iy*stride;
if( (unsigned)ix >= width-w || if( (unsigned)ix >= width-w ||
(unsigned)iy >= height-h ) (unsigned)iy >= height-h )
{ {
ff_emulated_edge_mc(edge_buf, src+ix+iy*stride, stride, w+1, h+1, ix, iy, width, height); ff_emulated_edge_mc(edge_buf, src, stride, w+1, h+1, ix, iy, width, height);
src = edge_buf; src = edge_buf;
} }
else
src += ix + iy*stride; asm volatile(
"movd %0, %%mm6 \n\t"
"pxor %%mm7, %%mm7 \n\t"
"punpcklwd %%mm6, %%mm6 \n\t"
"punpcklwd %%mm6, %%mm6 \n\t"
:: "r"(1<<shift)
);
for(x=0; x<w; x+=4){ for(x=0; x<w; x+=4){
uint16_t dx4[4] = { oxs - dxys + dxxs*(x+0), uint16_t dx4[4] = { oxs - dxys + dxxs*(x+0),
@ -2456,14 +2462,6 @@ static void gmc_mmx(uint8_t *dst, uint8_t *src, int stride, int h, int ox, int o
oys - dyys + dyxs*(x+2), oys - dyys + dyxs*(x+2),
oys - dyys + dyxs*(x+3) }; oys - dyys + dyxs*(x+3) };
asm volatile(
"movd %0, %%mm6 \n\t"
"pxor %%mm7, %%mm7 \n\t"
"punpcklwd %%mm6, %%mm6 \n\t"
"punpcklwd %%mm6, %%mm6 \n\t"
:: "g"(s)
);
for(y=0; y<h; y++){ for(y=0; y<h; y++){
asm volatile( asm volatile(
"movq %0, %%mm4 \n\t" "movq %0, %%mm4 \n\t"
@ -2503,10 +2501,10 @@ static void gmc_mmx(uint8_t *dst, uint8_t *src, int stride, int h, int ox, int o
"punpcklbw %%mm7, %%mm4 \n\t" "punpcklbw %%mm7, %%mm4 \n\t"
"pmullw %%mm5, %%mm1 \n\t" // src[1,0] * dx*(s-dy) "pmullw %%mm5, %%mm1 \n\t" // src[1,0] * dx*(s-dy)
"pmullw %%mm4, %%mm0 \n\t" // src[0,0] * (s-dx)*(s-dy) "pmullw %%mm4, %%mm0 \n\t" // src[0,0] * (s-dx)*(s-dy)
"paddw %5, %%mm1 \n\t"
"paddw %%mm3, %%mm2 \n\t" "paddw %%mm3, %%mm2 \n\t"
"paddw %%mm1, %%mm0 \n\t" "paddw %%mm1, %%mm0 \n\t"
"paddw %%mm2, %%mm0 \n\t" "paddw %%mm2, %%mm0 \n\t"
"paddw %5, %%mm0 \n\t"
"psrlw %6, %%mm0 \n\t" "psrlw %6, %%mm0 \n\t"
"packuswb %%mm0, %%mm0 \n\t" "packuswb %%mm0, %%mm0 \n\t"