virtualenv/FFmpeg
1
0
Fork 0
mirror of https://github.com/FFmpeg/FFmpeg.git synced 2025-08-10 06:10:52 +02:00
Code Issues Releases Activity

bugfixes

Browse Source 
small speedup for the non x86 stuff

Originally committed as revision 2567 to svn://svn.mplayerhq.hu/mplayer/trunk/postproc
This commit is contained in:
Michael Niedermayer
2001-10-30 20:30:24 +00:00
parent 51da31f16b
commit 1faf086709
2 changed files with 18 additions and 14 deletions
Download Patch File Download Diff File Expand all files Collapse all files

16
postproc/swscale.c
View File

@@ -33,6 +33,7 @@ Special versions: fast Y 1:1 scaling (no interpolation in y direction)
TODO TODO
more intelligent missalignment avoidance for the horizontal scaler more intelligent missalignment avoidance for the horizontal scaler
bicubic scaler
*/ */
#define ABS(a) ((a) > 0 ? (a) : (-(a))) #define ABS(a) ((a) > 0 ? (a) : (-(a)))
@@ -912,7 +913,6 @@ static inline void hyscale(uint16_t *dst, int dstWidth, uint8_t *src, int srcWid
unsigned int xpos=0; unsigned int xpos=0;
// *** horizontal scale Y line to temp buffer // *** horizontal scale Y line to temp buffer
#ifdef ARCH_X86 #ifdef ARCH_X86
#ifdef HAVE_MMX2 #ifdef HAVE_MMX2
if(canMMX2BeUsed) if(canMMX2BeUsed)
{ {
@@ -1012,7 +1012,7 @@ FUNNY_Y_CODE
for(i=0;i<dstWidth;i++){ for(i=0;i<dstWidth;i++){
register unsigned int xx=xpos>>16; register unsigned int xx=xpos>>16;
register unsigned int xalpha=(xpos&0xFFFF)>>9; register unsigned int xalpha=(xpos&0xFFFF)>>9;
dst[i]=(src[xx]*(xalpha^127)+src[xx+1]*xalpha); dst[i]= (src[xx]<<7) + (src[xx+1] - src[xx])*xalpha;
xpos+=xInc; xpos+=xInc;
} }
#endif #endif
@@ -1068,8 +1068,6 @@ FUNNYUVCODE
FUNNYUVCODE FUNNYUVCODE
FUNNYUVCODE FUNNYUVCODE
FUNNYUVCODE FUNNYUVCODE
"xorl %%eax, %%eax \n\t" // i "xorl %%eax, %%eax \n\t" // i
"movl %6, %%esi \n\t" // src "movl %6, %%esi \n\t" // src
"movl %1, %%edi \n\t" // buf1 "movl %1, %%edi \n\t" // buf1
@@ -1143,6 +1141,10 @@ FUNNYUVCODE
register unsigned int xalpha=(xpos&0xFFFF)>>9; register unsigned int xalpha=(xpos&0xFFFF)>>9;
dst[i]=(src1[xx]*(xalpha^127)+src1[xx+1]*xalpha); dst[i]=(src1[xx]*(xalpha^127)+src1[xx+1]*xalpha);
dst[i+2048]=(src2[xx]*(xalpha^127)+src2[xx+1]*xalpha); dst[i+2048]=(src2[xx]*(xalpha^127)+src2[xx+1]*xalpha);
/* slower
dst[i]= (src1[xx]<<7) + (src1[xx+1] - src1[xx])*xalpha;
dst[i+2048]=(src2[xx]<<7) + (src2[xx+1] - src2[xx])*xalpha;
*/
xpos+=xInc; xpos+=xInc;
} }
#endif #endif
@@ -1195,8 +1197,8 @@ canMMX2BeUsed= (s_xinc <= 0x10000 && (dstw&31)==0 && (srcWidth&15)==0) ? 1 : 0;
if(canMMX2BeUsed) s_xinc+= 20; if(canMMX2BeUsed) s_xinc+= 20;
else s_xinc = ((srcWidth-2)<<16)/(dstw-2) - 20; else s_xinc = ((srcWidth-2)<<16)/(dstw-2) - 20;
if(fullUVIpol && !dstbpp==12) s_xinc2= s_xinc>>1; if(fullUVIpol && !(dstbpp==12)) s_xinc2= s_xinc>>1;
else s_xinc2= s_xinc; else s_xinc2= s_xinc;
// force calculation of the horizontal interpolation of the first line // force calculation of the horizontal interpolation of the first line
if(y==0){ if(y==0){
@@ -1440,7 +1442,7 @@ else s_xinc2= s_xinc;
#ifdef HAVE_MMX #ifdef HAVE_MMX
__asm __volatile(SFENCE:::"memory"); __asm __volatile(SFENCE:::"memory");
__asm __volatile(EMMS); __asm __volatile(EMMS:::"memory");
#endif #endif
} }

16
postproc/swscale_template.c
View File

@@ -33,6 +33,7 @@ Special versions: fast Y 1:1 scaling (no interpolation in y direction)
TODO TODO
more intelligent missalignment avoidance for the horizontal scaler more intelligent missalignment avoidance for the horizontal scaler
bicubic scaler
*/ */
#define ABS(a) ((a) > 0 ? (a) : (-(a))) #define ABS(a) ((a) > 0 ? (a) : (-(a)))
@@ -912,7 +913,6 @@ static inline void hyscale(uint16_t *dst, int dstWidth, uint8_t *src, int srcWid
unsigned int xpos=0; unsigned int xpos=0;
// *** horizontal scale Y line to temp buffer // *** horizontal scale Y line to temp buffer
#ifdef ARCH_X86 #ifdef ARCH_X86
#ifdef HAVE_MMX2 #ifdef HAVE_MMX2
if(canMMX2BeUsed) if(canMMX2BeUsed)
{ {
@@ -1012,7 +1012,7 @@ FUNNY_Y_CODE
for(i=0;i<dstWidth;i++){ for(i=0;i<dstWidth;i++){
register unsigned int xx=xpos>>16; register unsigned int xx=xpos>>16;
register unsigned int xalpha=(xpos&0xFFFF)>>9; register unsigned int xalpha=(xpos&0xFFFF)>>9;
dst[i]=(src[xx]*(xalpha^127)+src[xx+1]*xalpha); dst[i]= (src[xx]<<7) + (src[xx+1] - src[xx])*xalpha;
xpos+=xInc; xpos+=xInc;
} }
#endif #endif
@@ -1068,8 +1068,6 @@ FUNNYUVCODE
FUNNYUVCODE FUNNYUVCODE
FUNNYUVCODE FUNNYUVCODE
FUNNYUVCODE FUNNYUVCODE
"xorl %%eax, %%eax \n\t" // i "xorl %%eax, %%eax \n\t" // i
"movl %6, %%esi \n\t" // src "movl %6, %%esi \n\t" // src
"movl %1, %%edi \n\t" // buf1 "movl %1, %%edi \n\t" // buf1
@@ -1143,6 +1141,10 @@ FUNNYUVCODE
register unsigned int xalpha=(xpos&0xFFFF)>>9; register unsigned int xalpha=(xpos&0xFFFF)>>9;
dst[i]=(src1[xx]*(xalpha^127)+src1[xx+1]*xalpha); dst[i]=(src1[xx]*(xalpha^127)+src1[xx+1]*xalpha);
dst[i+2048]=(src2[xx]*(xalpha^127)+src2[xx+1]*xalpha); dst[i+2048]=(src2[xx]*(xalpha^127)+src2[xx+1]*xalpha);
/* slower
dst[i]= (src1[xx]<<7) + (src1[xx+1] - src1[xx])*xalpha;
dst[i+2048]=(src2[xx]<<7) + (src2[xx+1] - src2[xx])*xalpha;
*/
xpos+=xInc; xpos+=xInc;
} }
#endif #endif
@@ -1195,8 +1197,8 @@ canMMX2BeUsed= (s_xinc <= 0x10000 && (dstw&31)==0 && (srcWidth&15)==0) ? 1 : 0;
if(canMMX2BeUsed) s_xinc+= 20; if(canMMX2BeUsed) s_xinc+= 20;
else s_xinc = ((srcWidth-2)<<16)/(dstw-2) - 20; else s_xinc = ((srcWidth-2)<<16)/(dstw-2) - 20;
if(fullUVIpol && !dstbpp==12) s_xinc2= s_xinc>>1; if(fullUVIpol && !(dstbpp==12)) s_xinc2= s_xinc>>1;
else s_xinc2= s_xinc; else s_xinc2= s_xinc;
// force calculation of the horizontal interpolation of the first line // force calculation of the horizontal interpolation of the first line
if(y==0){ if(y==0){
@@ -1440,7 +1442,7 @@ else s_xinc2= s_xinc;
#ifdef HAVE_MMX #ifdef HAVE_MMX
__asm __volatile(SFENCE:::"memory"); __asm __volatile(SFENCE:::"memory");
__asm __volatile(EMMS); __asm __volatile(EMMS:::"memory");
#endif #endif
} }