comparison libswscale/swscale.c @ 23129:9528d1ebe68f

cosmetics: Remove trailing whitespace.

23128:0c4b9bc10494

/*
  supported Input formats: YV12, I420/IYUV, YUY2, UYVY, BGR32, BGR24, BGR16, BGR15, RGB32, RGB24, Y8/Y800, YVU9/IF09, PAL8
  supported output formats: YV12, I420/IYUV, YUY2, UYVY, {BGR,RGB}{1,4,8,15,16,24,32}, Y8/Y800, YVU9/IF09
  {BGR,RGB}{1,4,8,15,16} support dithering
  
  unscaled special converters (YV12=I420=IYUV, Y800=Y8)
  YV12 -> {BGR,RGB}{1,4,8,15,16,24,32}
  x -> x
  YUV9 -> YV12
  YUV9/YV12 -> Y800

  BGR24 -> BGR32 & RGB24 -> RGB32
  BGR32 -> BGR24 & RGB32 -> RGB24
  BGR15 -> BGR16
*/

/* 
tested special converters (most are tested actually but i didnt write it down ...)
 YV12 -> BGR16
 YV12 -> YV12
 BGR15 -> BGR16
 BGR16 -> BGR16

// clipping helper table for C implementations:
static unsigned char clip_table[768];

static SwsVector *sws_getConvVec(SwsVector *a, SwsVector *b);
		  
extern const uint8_t dither_2x2_4[2][8];
extern const uint8_t dither_2x2_8[2][8];
extern const uint8_t dither_8x8_32[8][8];
extern const uint8_t dither_8x8_73[8][8];
extern const uint8_t dither_8x8_220[8][8];

				if(U>255)    U=255;\
				else if(U<0) U=0;\
				if(V>255)    V=255;\
				else if(V<0) V=0;\
			}
                        
#define YSCALE_YUV_2_RGBX_C(type) \
			YSCALE_YUV_2_PACKEDX_C(type)\
			r = (type *)c->table_rV[V];\
			g = (type *)(c->table_gU[U] + c->table_gV[V]);\
			b = (type *)c->table_bU[U];\

static double getSplineCoeff(double a, double b, double c, double d, double dist)
{
//	printf("%f %f %f %f %f\n", a,b,c,d,dist);
	if(dist<=1.0) 	return ((d*dist + c)*dist + b)*dist +a;
	else		return getSplineCoeff(	0.0, 
						 b+ 2.0*c + 3.0*d,
						        c + 3.0*d,
						-b- 3.0*c - 6.0*d,
						dist-1.0);
}

	{
		int i;
		int xDstInSrc;
		filterSize= 1;
		filter= av_malloc(dstW*sizeof(double)*filterSize);
		
		xDstInSrc= xInc/2 - 0x8000;
		for(i=0; i<dstW; i++)
		{
			int xx= (xDstInSrc - ((filterSize-1)<<15) + (1<<15))>>16;

	{
		int i;
		int xDstInSrc;
		if     (flags&SWS_BICUBIC) filterSize= 4;
		else if(flags&SWS_X      ) filterSize= 4;
		else			   filterSize= 2; // SWS_BILINEAR / SWS_AREA 
		filter= av_malloc(dstW*sizeof(double)*filterSize);

		xDstInSrc= xInc/2 - 0x8000;
		for(i=0; i<dstW; i++)
		{

		else if(flags&SWS_BILINEAR)	sizeFactor= 2.0;
		else {
			sizeFactor= 0.0; //GCC warning killer
			ASSERT(0)
		}
		
		if(xInc1 <= 1.0)	filterSizeInSrc= sizeFactor; // upscale
		else			filterSizeInSrc= sizeFactor*srcW / (double)dstW;

		filterSize= (int)ceil(1 + filterSizeInSrc); // will be reduced later if possible
		if(filterSize > srcW-2) filterSize=srcW-2;

				if(flags & SWS_BICUBIC)
				{
					double B= param[0] != SWS_PARAM_DEFAULT ? param[0] : 0.0;
					double C= param[1] != SWS_PARAM_DEFAULT ? param[1] : 0.6;

					if(d<1.0) 
						coeff = (12-9*B-6*C)*d*d*d + (-18+12*B+6*C)*d*d + 6-2*B;
					else if(d<2.0)
						coeff = (-B-6*C)*d*d*d + (6*B+30*C)*d*d + (-12*B-48*C)*d +8*B+24*C;
					else
						coeff=0.0;

					coeff*= pow(2.0, - p*d*d);
				}*/
				else if(flags & SWS_X)
				{
					double A= param[0] != SWS_PARAM_DEFAULT ? param[0] : 1.0;
					
					if(d<1.0)
						coeff = cos(d*PI);
					else
						coeff=-1.0;
					if(coeff<0.0) 	coeff= -pow(-coeff, A);

				{
					coeff = d ? sin(d*PI)/(d*PI) : 1.0;
				}
				else if(flags & SWS_LANCZOS)
				{
					double p= param[0] != SWS_PARAM_DEFAULT ? param[0] : 3.0; 
					coeff = d ? sin(d*PI)*sin(d*PI/p)/(d*d*PI*PI/p) : 1.0;
					if(d>p) coeff=0;
				}
				else if(flags & SWS_BILINEAR)
				{

			if(j>=filter2Size) filter[i*filterSize + j]= 0.0;
			else		   filter[i*filterSize + j]= filter2[i*filter2Size + j];
		}
	}
	av_free(filter2); filter2=NULL;
	

	//FIXME try to align filterpos if possible

	//fix borders
	for(i=0; i<dstW; i++)

			int intV= floor(v + 0.5);
			(*outFilter)[i*(*outFilterSize) + j]= intV;
			error = v - intV;
		}
	}
	
	(*filterPos)[dstW]= (*filterPos)[dstW-1]; // the MMX scaler will read over the end
	for(i=0; i<*outFilterSize; i++)
	{
		int j= dstW*(*outFilterSize);
		(*outFilter)[j + i]= (*outFilter)[j + i - (*outFilterSize)];

	asm volatile(
		"jmp 9f				\n\t"
	// Begin
		"0:				\n\t"
		"movq (%%"REG_d", %%"REG_a"), %%mm3\n\t" 
		"movd (%%"REG_c", %%"REG_S"), %%mm0\n\t" 
		"movd 1(%%"REG_c", %%"REG_S"), %%mm1\n\t"
		"punpcklbw %%mm7, %%mm1		\n\t"
		"punpcklbw %%mm7, %%mm0		\n\t"
		"pshufw $0xFF, %%mm1, %%mm1	\n\t"
		"1:				\n\t"

	asm volatile(
		"jmp 9f				\n\t"
	// Begin
		"0:				\n\t"
		"movq (%%"REG_d", %%"REG_a"), %%mm3\n\t" 
		"movd (%%"REG_c", %%"REG_S"), %%mm0\n\t" 
		"punpcklbw %%mm7, %%mm0		\n\t"
		"pshufw $0xFF, %%mm0, %%mm1	\n\t"
		"1:				\n\t"
		"pshufw $0xFF, %%mm0, %%mm0	\n\t"
		"2:				\n\t"

		"=r" (fragmentLengthB)
	);

	xpos= 0; //lumXInc/2 - 0x8000; // difference between pixel centers
	fragmentPos=0;
	
	for(i=0; i<dstW/numSplits; i++)
	{
		int xx=xpos>>16;

		if((i&3) == 0)

	clip_table[i]=c;
    }
}

static SwsFunc getSwsFunc(int flags){
    
#if defined(RUNTIME_CPUDETECT) && defined (CONFIG_GPL)
#if defined(ARCH_X86)
	// ordered per speed fasterst first
	if(flags & SWS_CPU_CAPS_MMX2)
		return swScale_MMX2;

		case 0x64: conv= rgb16to24; break;
		case 0x68: conv= rgb32to24; break;
		case 0x83: conv= rgb15to32; break;
		case 0x84: conv= rgb16to32; break;
		case 0x86: conv= rgb24to32; break;
		default: av_log(c, AV_LOG_ERROR, "swScaler: internal error %s -> %s converter\n", 
				 sws_format_name(srcFormat), sws_format_name(dstFormat)); break;
		}
	}else if(   (isBGR(srcFormat) && isRGB(dstFormat))
		 || (isRGB(srcFormat) && isBGR(dstFormat))){
		switch(srcId | (dstId<<4)){

		case 0x68: conv= rgb32tobgr24; break;
		case 0x83: conv= rgb15tobgr32; break;
		case 0x84: conv= rgb16tobgr32; break;
		case 0x86: conv= rgb24tobgr32; break;
		case 0x88: conv= rgb32tobgr32; break;
		default: av_log(c, AV_LOG_ERROR, "swScaler: internal error %s -> %s converter\n", 
				 sws_format_name(srcFormat), sws_format_name(dstFormat)); break;
		}
	}else{
		av_log(c, AV_LOG_ERROR, "swScaler: internal error %s -> %s converter\n", 
			 sws_format_name(srcFormat), sws_format_name(dstFormat));
	}

	if(dstStride[0]*srcBpp == srcStride[0]*dstBpp)
		conv(src[0], dst[0] + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]);

		{
			conv(srcPtr, dstPtr, c->srcW*srcBpp);
			srcPtr+= srcStride[0];
			dstPtr+= dstStride[0];
		}
	}     
	return srcSliceH;
}

static int bgr24toyv12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
             int srcSliceH, uint8_t* dst[], int dstStride[]){

	rgb24toyv12(
		src[0], 
		dst[0]+ srcSliceY    *dstStride[0], 
		dst[1]+(srcSliceY>>1)*dstStride[1], 
		dst[2]+(srcSliceY>>1)*dstStride[2],
		c->srcW, srcSliceH, 
		dstStride[0], dstStride[1], srcStride[0]);
	return srcSliceH;
}

static int yvu9toyv12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
             int srcSliceH, uint8_t* dst[], int dstStride[]){
	int i;

	/* copy Y */
	if(srcStride[0]==dstStride[0] && srcStride[0] > 0) 
		memcpy(dst[0]+ srcSliceY*dstStride[0], src[0], srcStride[0]*srcSliceH);
	else{
		uint8_t *srcPtr= src[0];
		uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;

			uint8_t *srcPtr= src[0];
			uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
			int length=0;

			/* universal length finder */
			while(length+c->srcW <= FFABS(dstStride[0]) 
			   && length+c->srcW <= FFABS(srcStride[0])) length+= c->srcW;
			ASSERT(length!=0);

			for(i=0; i<srcSliceH; i++)
			{

				srcPtr+= srcStride[0];
				dstPtr+= dstStride[0];
			}
		}
	}
	else 
	{ /* Planar YUV or gray */
		int plane;
		for(plane=0; plane<3; plane++)
		{
			int length= plane==0 ? c->srcW  : -((-c->srcW  )>>c->chrDstHSubSample);

	//FIXME factorize

#ifdef COMPILE_ALTIVEC
	if (c->flags & SWS_CPU_CAPS_ALTIVEC)
	    yuv2rgb_altivec_init_tables (c, inv_table, brightness, contrast, saturation);
#endif	
	return 0;
}

/**
 * @return -1 if not supported

	*srcRange  = c->srcRange;
	*dstRange  = c->dstRange;
	*brightness= c->brightness;
	*contrast  = c->contrast;
	*saturation= c->saturation;
	
	return 0;	
}

static int handle_jpeg(int *format)
{
	switch (*format) {

#endif /* RUNTIME_CPUDETECT */
	if(clip_table[512] != 255) globalInit();
	if(rgb15to16 == NULL) sws_rgb2rgb_init(flags);

	unscaled = (srcW == dstW && srcH == dstH);
	needsDither= (isBGR(dstFormat) || isRGB(dstFormat)) 
		     && (fmt_depth(dstFormat))<24
		     && ((fmt_depth(dstFormat))<(fmt_depth(srcFormat)) || (!(isRGB(srcFormat) || isBGR(srcFormat))));

	srcRange = handle_jpeg(&srcFormat);
	dstRange = handle_jpeg(&dstFormat);

	if(!isSupportedIn(srcFormat)) 
	{
		av_log(NULL, AV_LOG_ERROR, "swScaler: %s is not supported as input format\n", sws_format_name(srcFormat));
		return NULL;
	}
	if(!isSupportedOut(dstFormat))

	}

	/* sanity check */
	if(srcW<4 || srcH<1 || dstW<8 || dstH<1) //FIXME check if these are enough and try to lowwer them after fixing the relevant parts of the code
	{
		 av_log(NULL, AV_LOG_ERROR, "swScaler: %dx%d -> %dx%d is invalid scaling dimension\n", 
			srcW, srcH, dstW, dstH);
		return NULL;
	}

	if(!dstFilter) dstFilter= &dummyFilter;

	c->chrSrcW= -((-srcW) >> c->chrSrcHSubSample);
	c->chrSrcH= -((-srcH) >> c->chrSrcVSubSample);
	c->chrDstW= -((-dstW) >> c->chrDstHSubSample);
	c->chrDstH= -((-dstH) >> c->chrDstVSubSample);

	sws_setColorspaceDetails(c, Inverse_Table_6_9[SWS_CS_DEFAULT], srcRange, Inverse_Table_6_9[SWS_CS_DEFAULT] /* FIXME*/, dstRange, 0, 1<<16, 1<<16); 

	/* unscaled special Cases */
	if(unscaled && !usesHFilter && !usesVFilter)
	{
		/* yv12_to_nv12 */

		if((srcFormat==PIX_FMT_YUV420P || srcFormat==PIX_FMT_YUV422P) && (isBGR(dstFormat) || isRGB(dstFormat)))
		{
			c->swScale= yuv2rgb_get_func_ptr(c);
		}
#endif
		
		if( srcFormat==PIX_FMT_YUV410P && dstFormat==PIX_FMT_YUV420P )
		{
			c->swScale= yvu9toyv12Wrapper;
		}

		/* bgr24toYV12 */
		if(srcFormat==PIX_FMT_BGR24 && dstFormat==PIX_FMT_YUV420P)
			c->swScale= bgr24toyv12Wrapper;
		
		/* rgb/bgr -> rgb/bgr (no dither needed forms) */
		if(   (isBGR(srcFormat) || isRGB(srcFormat))
		   && (isBGR(dstFormat) || isRGB(dstFormat)) 
		   && srcFormat != PIX_FMT_BGR8 && dstFormat != PIX_FMT_BGR8
		   && srcFormat != PIX_FMT_RGB8 && dstFormat != PIX_FMT_RGB8
		   && srcFormat != PIX_FMT_BGR4 && dstFormat != PIX_FMT_BGR4
		   && srcFormat != PIX_FMT_RGB4 && dstFormat != PIX_FMT_RGB4
		   && srcFormat != PIX_FMT_BGR4_BYTE && dstFormat != PIX_FMT_BGR4_BYTE

		/* LQ converters if -sws 0 or -sws 4*/
		if(c->flags&(SWS_FAST_BILINEAR|SWS_POINT)){
			/* rgb/bgr -> rgb/bgr (dither needed forms) */
			if(  (isBGR(srcFormat) || isRGB(srcFormat))
			  && (isBGR(dstFormat) || isRGB(dstFormat)) 
			  && needsDither)
				c->swScale= rgb2rgbWrapper;

			/* yv12_to_yuy2 */
			if(srcFormat == PIX_FMT_YUV420P && 
			    (dstFormat == PIX_FMT_YUYV422 || dstFormat == PIX_FMT_UYVY422))
			{
				if (dstFormat == PIX_FMT_YUYV422)
				    c->swScale= PlanarToYuy2Wrapper;
				else

			}
		}

#ifdef COMPILE_ALTIVEC
		if ((c->flags & SWS_CPU_CAPS_ALTIVEC) &&
		    ((srcFormat == PIX_FMT_YUV420P && 
		      (dstFormat == PIX_FMT_YUYV422 || dstFormat == PIX_FMT_UYVY422)))) {
		  // unscaled YV12 -> packed YUV, we want speed
		  if (dstFormat == PIX_FMT_YUYV422)
		    c->swScale= yv12toyuy2_unscaled_altivec;
		  else

			c->swScale= graytogray16;
		}
		if(srcFormat != dstFormat && isGray16(srcFormat) && isGray16(dstFormat))
		{
			c->swScale= gray16swap;
		}		

		if(c->swScale){
			if(flags&SWS_PRINT_INFO)
				av_log(c, AV_LOG_INFO, "SwScaler: using unscaled %s -> %s special converter\n", 
					sws_format_name(srcFormat), sws_format_name(dstFormat));
			return c;
		}
	}

			av_log(c, AV_LOG_INFO, "SwScaler: Bicubic spline scaler, ");
		else
			av_log(c, AV_LOG_INFO, "SwScaler: ehh flags invalid?! ");

		if(dstFormat==PIX_FMT_BGR555 || dstFormat==PIX_FMT_BGR565)
			av_log(c, AV_LOG_INFO, "from %s to%s %s ", 
				sws_format_name(srcFormat), dither, sws_format_name(dstFormat));
		else
			av_log(c, AV_LOG_INFO, "from %s to %s ", 
				sws_format_name(srcFormat), sws_format_name(dstFormat));

		if(flags & SWS_CPU_CAPS_MMX2)
			av_log(c, AV_LOG_INFO, "using MMX2\n");
		else if(flags & SWS_CPU_CAPS_3DNOW)
			av_log(c, AV_LOG_INFO, "using 3DNOW\n");
		else if(flags & SWS_CPU_CAPS_MMX)
			av_log(c, AV_LOG_INFO, "using MMX\n");
		else if(flags & SWS_CPU_CAPS_ALTIVEC)
			av_log(c, AV_LOG_INFO, "using AltiVec\n");
		else 
			av_log(c, AV_LOG_INFO, "using C\n");
	}

	if(flags & SWS_PRINT_INFO)
	{

            src2[0] += (srcSliceH-1)*srcStride[0];
            if(c->srcFormat != PIX_FMT_PAL8)
                    src2[1] += ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[1];
            src2[2] += ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[2];
	    
	    return c->swScale(c, src2, srcStride2, c->srcH-srcSliceY-srcSliceH, srcSliceH, dst2, dstStride2);
	}
}

/**

int sws_scale_ordered(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
                           int srcSliceH, uint8_t* dst[], int dstStride[]){
	return sws_scale(c, src, srcStride, srcSliceY, srcSliceH, dst, dstStride);
}

SwsFilter *sws_getDefaultFilter(float lumaGBlur, float chromaGBlur, 
				float lumaSharpen, float chromaSharpen,
				float chromaHShift, float chromaVShift,
				int verbose)
{
	SwsFilter *filter= av_malloc(sizeof(SwsFilter));

	av_free(diff);
}

void sws_convVec(SwsVector *a, SwsVector *b){
	SwsVector *conv= sws_getConvVec(a, b);
	av_free(a->coeff);  
	a->coeff= conv->coeff;
	a->length= conv->length;
	av_free(conv);
}

23129:9528d1ebe68f

/*
  supported Input formats: YV12, I420/IYUV, YUY2, UYVY, BGR32, BGR24, BGR16, BGR15, RGB32, RGB24, Y8/Y800, YVU9/IF09, PAL8
  supported output formats: YV12, I420/IYUV, YUY2, UYVY, {BGR,RGB}{1,4,8,15,16,24,32}, Y8/Y800, YVU9/IF09
  {BGR,RGB}{1,4,8,15,16} support dithering

  unscaled special converters (YV12=I420=IYUV, Y800=Y8)
  YV12 -> {BGR,RGB}{1,4,8,15,16,24,32}
  x -> x
  YUV9 -> YV12
  YUV9/YV12 -> Y800

  BGR24 -> BGR32 & RGB24 -> RGB32
  BGR32 -> BGR24 & RGB32 -> RGB24
  BGR15 -> BGR16
*/

/*
tested special converters (most are tested actually but i didnt write it down ...)
 YV12 -> BGR16
 YV12 -> YV12
 BGR15 -> BGR16
 BGR16 -> BGR16

// clipping helper table for C implementations:
static unsigned char clip_table[768];

static SwsVector *sws_getConvVec(SwsVector *a, SwsVector *b);

extern const uint8_t dither_2x2_4[2][8];
extern const uint8_t dither_2x2_8[2][8];
extern const uint8_t dither_8x8_32[8][8];
extern const uint8_t dither_8x8_73[8][8];
extern const uint8_t dither_8x8_220[8][8];

				if(U>255)    U=255;\
				else if(U<0) U=0;\
				if(V>255)    V=255;\
				else if(V<0) V=0;\
			}

#define YSCALE_YUV_2_RGBX_C(type) \
			YSCALE_YUV_2_PACKEDX_C(type)\
			r = (type *)c->table_rV[V];\
			g = (type *)(c->table_gU[U] + c->table_gV[V]);\
			b = (type *)c->table_bU[U];\

static double getSplineCoeff(double a, double b, double c, double d, double dist)
{
//	printf("%f %f %f %f %f\n", a,b,c,d,dist);
	if(dist<=1.0) 	return ((d*dist + c)*dist + b)*dist +a;
	else		return getSplineCoeff(	0.0,
						 b+ 2.0*c + 3.0*d,
						        c + 3.0*d,
						-b- 3.0*c - 6.0*d,
						dist-1.0);
}

	{
		int i;
		int xDstInSrc;
		filterSize= 1;
		filter= av_malloc(dstW*sizeof(double)*filterSize);

		xDstInSrc= xInc/2 - 0x8000;
		for(i=0; i<dstW; i++)
		{
			int xx= (xDstInSrc - ((filterSize-1)<<15) + (1<<15))>>16;

	{
		int i;
		int xDstInSrc;
		if     (flags&SWS_BICUBIC) filterSize= 4;
		else if(flags&SWS_X      ) filterSize= 4;
		else			   filterSize= 2; // SWS_BILINEAR / SWS_AREA
		filter= av_malloc(dstW*sizeof(double)*filterSize);

		xDstInSrc= xInc/2 - 0x8000;
		for(i=0; i<dstW; i++)
		{

		else if(flags&SWS_BILINEAR)	sizeFactor= 2.0;
		else {
			sizeFactor= 0.0; //GCC warning killer
			ASSERT(0)
		}

		if(xInc1 <= 1.0)	filterSizeInSrc= sizeFactor; // upscale
		else			filterSizeInSrc= sizeFactor*srcW / (double)dstW;

		filterSize= (int)ceil(1 + filterSizeInSrc); // will be reduced later if possible
		if(filterSize > srcW-2) filterSize=srcW-2;

				if(flags & SWS_BICUBIC)
				{
					double B= param[0] != SWS_PARAM_DEFAULT ? param[0] : 0.0;
					double C= param[1] != SWS_PARAM_DEFAULT ? param[1] : 0.6;

					if(d<1.0)
						coeff = (12-9*B-6*C)*d*d*d + (-18+12*B+6*C)*d*d + 6-2*B;
					else if(d<2.0)
						coeff = (-B-6*C)*d*d*d + (6*B+30*C)*d*d + (-12*B-48*C)*d +8*B+24*C;
					else
						coeff=0.0;

					coeff*= pow(2.0, - p*d*d);
				}*/
				else if(flags & SWS_X)
				{
					double A= param[0] != SWS_PARAM_DEFAULT ? param[0] : 1.0;

					if(d<1.0)
						coeff = cos(d*PI);
					else
						coeff=-1.0;
					if(coeff<0.0) 	coeff= -pow(-coeff, A);

				{
					coeff = d ? sin(d*PI)/(d*PI) : 1.0;
				}
				else if(flags & SWS_LANCZOS)
				{
					double p= param[0] != SWS_PARAM_DEFAULT ? param[0] : 3.0;
					coeff = d ? sin(d*PI)*sin(d*PI/p)/(d*d*PI*PI/p) : 1.0;
					if(d>p) coeff=0;
				}
				else if(flags & SWS_BILINEAR)
				{

			if(j>=filter2Size) filter[i*filterSize + j]= 0.0;
			else		   filter[i*filterSize + j]= filter2[i*filter2Size + j];
		}
	}
	av_free(filter2); filter2=NULL;


	//FIXME try to align filterpos if possible

	//fix borders
	for(i=0; i<dstW; i++)

			int intV= floor(v + 0.5);
			(*outFilter)[i*(*outFilterSize) + j]= intV;
			error = v - intV;
		}
	}

	(*filterPos)[dstW]= (*filterPos)[dstW-1]; // the MMX scaler will read over the end
	for(i=0; i<*outFilterSize; i++)
	{
		int j= dstW*(*outFilterSize);
		(*outFilter)[j + i]= (*outFilter)[j + i - (*outFilterSize)];

	asm volatile(
		"jmp 9f				\n\t"
	// Begin
		"0:				\n\t"
		"movq (%%"REG_d", %%"REG_a"), %%mm3\n\t"
		"movd (%%"REG_c", %%"REG_S"), %%mm0\n\t"
		"movd 1(%%"REG_c", %%"REG_S"), %%mm1\n\t"
		"punpcklbw %%mm7, %%mm1		\n\t"
		"punpcklbw %%mm7, %%mm0		\n\t"
		"pshufw $0xFF, %%mm1, %%mm1	\n\t"
		"1:				\n\t"

	asm volatile(
		"jmp 9f				\n\t"
	// Begin
		"0:				\n\t"
		"movq (%%"REG_d", %%"REG_a"), %%mm3\n\t"
		"movd (%%"REG_c", %%"REG_S"), %%mm0\n\t"
		"punpcklbw %%mm7, %%mm0		\n\t"
		"pshufw $0xFF, %%mm0, %%mm1	\n\t"
		"1:				\n\t"
		"pshufw $0xFF, %%mm0, %%mm0	\n\t"
		"2:				\n\t"

		"=r" (fragmentLengthB)
	);

	xpos= 0; //lumXInc/2 - 0x8000; // difference between pixel centers
	fragmentPos=0;

	for(i=0; i<dstW/numSplits; i++)
	{
		int xx=xpos>>16;

		if((i&3) == 0)

	clip_table[i]=c;
    }
}

static SwsFunc getSwsFunc(int flags){

#if defined(RUNTIME_CPUDETECT) && defined (CONFIG_GPL)
#if defined(ARCH_X86)
	// ordered per speed fasterst first
	if(flags & SWS_CPU_CAPS_MMX2)
		return swScale_MMX2;

		case 0x64: conv= rgb16to24; break;
		case 0x68: conv= rgb32to24; break;
		case 0x83: conv= rgb15to32; break;
		case 0x84: conv= rgb16to32; break;
		case 0x86: conv= rgb24to32; break;
		default: av_log(c, AV_LOG_ERROR, "swScaler: internal error %s -> %s converter\n",
				 sws_format_name(srcFormat), sws_format_name(dstFormat)); break;
		}
	}else if(   (isBGR(srcFormat) && isRGB(dstFormat))
		 || (isRGB(srcFormat) && isBGR(dstFormat))){
		switch(srcId | (dstId<<4)){

		case 0x68: conv= rgb32tobgr24; break;
		case 0x83: conv= rgb15tobgr32; break;
		case 0x84: conv= rgb16tobgr32; break;
		case 0x86: conv= rgb24tobgr32; break;
		case 0x88: conv= rgb32tobgr32; break;
		default: av_log(c, AV_LOG_ERROR, "swScaler: internal error %s -> %s converter\n",
				 sws_format_name(srcFormat), sws_format_name(dstFormat)); break;
		}
	}else{
		av_log(c, AV_LOG_ERROR, "swScaler: internal error %s -> %s converter\n",
			 sws_format_name(srcFormat), sws_format_name(dstFormat));
	}

	if(dstStride[0]*srcBpp == srcStride[0]*dstBpp)
		conv(src[0], dst[0] + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]);

		{
			conv(srcPtr, dstPtr, c->srcW*srcBpp);
			srcPtr+= srcStride[0];
			dstPtr+= dstStride[0];
		}
	}
	return srcSliceH;
}

static int bgr24toyv12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
             int srcSliceH, uint8_t* dst[], int dstStride[]){

	rgb24toyv12(
		src[0],
		dst[0]+ srcSliceY    *dstStride[0],
		dst[1]+(srcSliceY>>1)*dstStride[1],
		dst[2]+(srcSliceY>>1)*dstStride[2],
		c->srcW, srcSliceH,
		dstStride[0], dstStride[1], srcStride[0]);
	return srcSliceH;
}

static int yvu9toyv12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
             int srcSliceH, uint8_t* dst[], int dstStride[]){
	int i;

	/* copy Y */
	if(srcStride[0]==dstStride[0] && srcStride[0] > 0)
		memcpy(dst[0]+ srcSliceY*dstStride[0], src[0], srcStride[0]*srcSliceH);
	else{
		uint8_t *srcPtr= src[0];
		uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;

			uint8_t *srcPtr= src[0];
			uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;
			int length=0;

			/* universal length finder */
			while(length+c->srcW <= FFABS(dstStride[0])
			   && length+c->srcW <= FFABS(srcStride[0])) length+= c->srcW;
			ASSERT(length!=0);

			for(i=0; i<srcSliceH; i++)
			{

				srcPtr+= srcStride[0];
				dstPtr+= dstStride[0];
			}
		}
	}
	else
	{ /* Planar YUV or gray */
		int plane;
		for(plane=0; plane<3; plane++)
		{
			int length= plane==0 ? c->srcW  : -((-c->srcW  )>>c->chrDstHSubSample);

	//FIXME factorize

#ifdef COMPILE_ALTIVEC
	if (c->flags & SWS_CPU_CAPS_ALTIVEC)
	    yuv2rgb_altivec_init_tables (c, inv_table, brightness, contrast, saturation);
#endif
	return 0;
}

/**
 * @return -1 if not supported

	*srcRange  = c->srcRange;
	*dstRange  = c->dstRange;
	*brightness= c->brightness;
	*contrast  = c->contrast;
	*saturation= c->saturation;

	return 0;
}

static int handle_jpeg(int *format)
{
	switch (*format) {

#endif /* RUNTIME_CPUDETECT */
	if(clip_table[512] != 255) globalInit();
	if(rgb15to16 == NULL) sws_rgb2rgb_init(flags);

	unscaled = (srcW == dstW && srcH == dstH);
	needsDither= (isBGR(dstFormat) || isRGB(dstFormat))
		     && (fmt_depth(dstFormat))<24
		     && ((fmt_depth(dstFormat))<(fmt_depth(srcFormat)) || (!(isRGB(srcFormat) || isBGR(srcFormat))));

	srcRange = handle_jpeg(&srcFormat);
	dstRange = handle_jpeg(&dstFormat);

	if(!isSupportedIn(srcFormat))
	{
		av_log(NULL, AV_LOG_ERROR, "swScaler: %s is not supported as input format\n", sws_format_name(srcFormat));
		return NULL;
	}
	if(!isSupportedOut(dstFormat))

	}

	/* sanity check */
	if(srcW<4 || srcH<1 || dstW<8 || dstH<1) //FIXME check if these are enough and try to lowwer them after fixing the relevant parts of the code
	{
		 av_log(NULL, AV_LOG_ERROR, "swScaler: %dx%d -> %dx%d is invalid scaling dimension\n",
			srcW, srcH, dstW, dstH);
		return NULL;
	}

	if(!dstFilter) dstFilter= &dummyFilter;

	c->chrSrcW= -((-srcW) >> c->chrSrcHSubSample);
	c->chrSrcH= -((-srcH) >> c->chrSrcVSubSample);
	c->chrDstW= -((-dstW) >> c->chrDstHSubSample);
	c->chrDstH= -((-dstH) >> c->chrDstVSubSample);

	sws_setColorspaceDetails(c, Inverse_Table_6_9[SWS_CS_DEFAULT], srcRange, Inverse_Table_6_9[SWS_CS_DEFAULT] /* FIXME*/, dstRange, 0, 1<<16, 1<<16);

	/* unscaled special Cases */
	if(unscaled && !usesHFilter && !usesVFilter)
	{
		/* yv12_to_nv12 */

		if((srcFormat==PIX_FMT_YUV420P || srcFormat==PIX_FMT_YUV422P) && (isBGR(dstFormat) || isRGB(dstFormat)))
		{
			c->swScale= yuv2rgb_get_func_ptr(c);
		}
#endif

		if( srcFormat==PIX_FMT_YUV410P && dstFormat==PIX_FMT_YUV420P )
		{
			c->swScale= yvu9toyv12Wrapper;
		}

		/* bgr24toYV12 */
		if(srcFormat==PIX_FMT_BGR24 && dstFormat==PIX_FMT_YUV420P)
			c->swScale= bgr24toyv12Wrapper;

		/* rgb/bgr -> rgb/bgr (no dither needed forms) */
		if(   (isBGR(srcFormat) || isRGB(srcFormat))
		   && (isBGR(dstFormat) || isRGB(dstFormat))
		   && srcFormat != PIX_FMT_BGR8 && dstFormat != PIX_FMT_BGR8
		   && srcFormat != PIX_FMT_RGB8 && dstFormat != PIX_FMT_RGB8
		   && srcFormat != PIX_FMT_BGR4 && dstFormat != PIX_FMT_BGR4
		   && srcFormat != PIX_FMT_RGB4 && dstFormat != PIX_FMT_RGB4
		   && srcFormat != PIX_FMT_BGR4_BYTE && dstFormat != PIX_FMT_BGR4_BYTE

		/* LQ converters if -sws 0 or -sws 4*/
		if(c->flags&(SWS_FAST_BILINEAR|SWS_POINT)){
			/* rgb/bgr -> rgb/bgr (dither needed forms) */
			if(  (isBGR(srcFormat) || isRGB(srcFormat))
			  && (isBGR(dstFormat) || isRGB(dstFormat))
			  && needsDither)
				c->swScale= rgb2rgbWrapper;

			/* yv12_to_yuy2 */
			if(srcFormat == PIX_FMT_YUV420P &&
			    (dstFormat == PIX_FMT_YUYV422 || dstFormat == PIX_FMT_UYVY422))
			{
				if (dstFormat == PIX_FMT_YUYV422)
				    c->swScale= PlanarToYuy2Wrapper;
				else

			}
		}

#ifdef COMPILE_ALTIVEC
		if ((c->flags & SWS_CPU_CAPS_ALTIVEC) &&
		    ((srcFormat == PIX_FMT_YUV420P &&
		      (dstFormat == PIX_FMT_YUYV422 || dstFormat == PIX_FMT_UYVY422)))) {
		  // unscaled YV12 -> packed YUV, we want speed
		  if (dstFormat == PIX_FMT_YUYV422)
		    c->swScale= yv12toyuy2_unscaled_altivec;
		  else

			c->swScale= graytogray16;
		}
		if(srcFormat != dstFormat && isGray16(srcFormat) && isGray16(dstFormat))
		{
			c->swScale= gray16swap;
		}

		if(c->swScale){
			if(flags&SWS_PRINT_INFO)
				av_log(c, AV_LOG_INFO, "SwScaler: using unscaled %s -> %s special converter\n",
					sws_format_name(srcFormat), sws_format_name(dstFormat));
			return c;
		}
	}

			av_log(c, AV_LOG_INFO, "SwScaler: Bicubic spline scaler, ");
		else
			av_log(c, AV_LOG_INFO, "SwScaler: ehh flags invalid?! ");

		if(dstFormat==PIX_FMT_BGR555 || dstFormat==PIX_FMT_BGR565)
			av_log(c, AV_LOG_INFO, "from %s to%s %s ",
				sws_format_name(srcFormat), dither, sws_format_name(dstFormat));
		else
			av_log(c, AV_LOG_INFO, "from %s to %s ",
				sws_format_name(srcFormat), sws_format_name(dstFormat));

		if(flags & SWS_CPU_CAPS_MMX2)
			av_log(c, AV_LOG_INFO, "using MMX2\n");
		else if(flags & SWS_CPU_CAPS_3DNOW)
			av_log(c, AV_LOG_INFO, "using 3DNOW\n");
		else if(flags & SWS_CPU_CAPS_MMX)
			av_log(c, AV_LOG_INFO, "using MMX\n");
		else if(flags & SWS_CPU_CAPS_ALTIVEC)
			av_log(c, AV_LOG_INFO, "using AltiVec\n");
		else
			av_log(c, AV_LOG_INFO, "using C\n");
	}

	if(flags & SWS_PRINT_INFO)
	{

            src2[0] += (srcSliceH-1)*srcStride[0];
            if(c->srcFormat != PIX_FMT_PAL8)
                    src2[1] += ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[1];
            src2[2] += ((srcSliceH>>c->chrSrcVSubSample)-1)*srcStride[2];

	    return c->swScale(c, src2, srcStride2, c->srcH-srcSliceY-srcSliceH, srcSliceH, dst2, dstStride2);
	}
}

/**

int sws_scale_ordered(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
                           int srcSliceH, uint8_t* dst[], int dstStride[]){
	return sws_scale(c, src, srcStride, srcSliceY, srcSliceH, dst, dstStride);
}

SwsFilter *sws_getDefaultFilter(float lumaGBlur, float chromaGBlur,
				float lumaSharpen, float chromaSharpen,
				float chromaHShift, float chromaVShift,
				int verbose)
{
	SwsFilter *filter= av_malloc(sizeof(SwsFilter));

	av_free(diff);
}

void sws_convVec(SwsVector *a, SwsVector *b){
	SwsVector *conv= sws_getConvVec(a, b);
	av_free(a->coeff);
	a->coeff= conv->coeff;
	a->length= conv->length;
	av_free(conv);
}