comparison postproc/swscale_template.c @ 2566:a350d8bed636

bugfixes small speedup for the non x86 stuff

2565:4bc54a0f775f

BGR15 & BGR16 MMX verions support dithering
Special versions: fast Y 1:1 scaling (no interpolation in y direction)

TODO
more intelligent missalignment avoidance for the horizontal scaler
*/

#define ABS(a) ((a) > 0 ? (a) : (-(a)))
#define MIN(a,b) ((a) > (b) ? (b) : (a))
#define MAX(a,b) ((a) < (b) ? (b) : (a))

{
	int i;
      unsigned int xpos=0;
      // *** horizontal scale Y line to temp buffer
#ifdef ARCH_X86

#ifdef HAVE_MMX2
	if(canMMX2BeUsed)
	{
		asm volatile(
			"pxor %%mm7, %%mm7		\n\t"

#endif
#else
      for(i=0;i<dstWidth;i++){
	register unsigned int xx=xpos>>16;
        register unsigned int xalpha=(xpos&0xFFFF)>>9;
	dst[i]=(src[xx]*(xalpha^127)+src[xx+1]*xalpha);
	xpos+=xInc;
      }
#endif
}

FUNNYUVCODE
FUNNYUVCODE
FUNNYUVCODE
FUNNYUVCODE


		"xorl %%eax, %%eax		\n\t" // i
		"movl %6, %%esi			\n\t" // src
		"movl %1, %%edi			\n\t" // buf1
		"addl $4096, %%edi		\n\t"

      for(i=0;i<dstWidth;i++){
	  register unsigned int xx=xpos>>16;
          register unsigned int xalpha=(xpos&0xFFFF)>>9;
	  dst[i]=(src1[xx]*(xalpha^127)+src1[xx+1]*xalpha);
	  dst[i+2048]=(src2[xx]*(xalpha^127)+src2[xx+1]*xalpha);
	  xpos+=xInc;
      }
#endif
}

// would be like the vertical one, but that would require some special code for the
// first and last pixel
if(canMMX2BeUsed) 	s_xinc+= 20;
else			s_xinc = ((srcWidth-2)<<16)/(dstw-2) - 20;

if(fullUVIpol && !dstbpp==12) 	s_xinc2= s_xinc>>1;
else				s_xinc2= s_xinc;
  // force calculation of the horizontal interpolation of the first line

  if(y==0){
	s_last_ypos=-99;
	s_last_y1pos=-99;

#endif
  }

#ifdef HAVE_MMX
	__asm __volatile(SFENCE:::"memory");
	__asm __volatile(EMMS);
#endif
}


void SwScale_Init(){

2566:a350d8bed636

BGR15 & BGR16 MMX verions support dithering
Special versions: fast Y 1:1 scaling (no interpolation in y direction)

TODO
more intelligent missalignment avoidance for the horizontal scaler
bicubic scaler
*/

#define ABS(a) ((a) > 0 ? (a) : (-(a)))
#define MIN(a,b) ((a) > (b) ? (b) : (a))
#define MAX(a,b) ((a) < (b) ? (b) : (a))

{
	int i;
      unsigned int xpos=0;
      // *** horizontal scale Y line to temp buffer
#ifdef ARCH_X86
#ifdef HAVE_MMX2
	if(canMMX2BeUsed)
	{
		asm volatile(
			"pxor %%mm7, %%mm7		\n\t"

#endif
#else
      for(i=0;i<dstWidth;i++){
	register unsigned int xx=xpos>>16;
        register unsigned int xalpha=(xpos&0xFFFF)>>9;
	dst[i]= (src[xx]<<7) + (src[xx+1] - src[xx])*xalpha;
	xpos+=xInc;
      }
#endif
}

FUNNYUVCODE
FUNNYUVCODE
FUNNYUVCODE
FUNNYUVCODE
		"xorl %%eax, %%eax		\n\t" // i
		"movl %6, %%esi			\n\t" // src
		"movl %1, %%edi			\n\t" // buf1
		"addl $4096, %%edi		\n\t"

      for(i=0;i<dstWidth;i++){
	  register unsigned int xx=xpos>>16;
          register unsigned int xalpha=(xpos&0xFFFF)>>9;
	  dst[i]=(src1[xx]*(xalpha^127)+src1[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;
      }
#endif
}

// would be like the vertical one, but that would require some special code for the
// first and last pixel
if(canMMX2BeUsed) 	s_xinc+= 20;
else			s_xinc = ((srcWidth-2)<<16)/(dstw-2) - 20;

if(fullUVIpol && !(dstbpp==12)) 	s_xinc2= s_xinc>>1;
else					s_xinc2= s_xinc;
  // force calculation of the horizontal interpolation of the first line

  if(y==0){
	s_last_ypos=-99;
	s_last_y1pos=-99;

#endif
  }

#ifdef HAVE_MMX
	__asm __volatile(SFENCE:::"memory");
	__asm __volatile(EMMS:::"memory");
#endif
}


void SwScale_Init(){