swscale.c

uclinux 下的vlc播放器源代码

				((uint8_t*)dest)[0]= acc;\
				dest++;\
			}\
\
/*\
((uint8_t*)dest)-= dstW>>4;\
{\
			int acc=0;\
			int left=0;\
			static int top[1024];\
			static int last_new[1024][1024];\
			static int last_in3[1024][1024];\
			static int drift[1024][1024];\
			int topLeft=0;\
			int shift=0;\
			int count=0;\
			const uint8_t * const d128=dither_8x8_220[y&7];\
			int error_new=0;\
			int error_in3=0;\
			int f=0;\
			\
			for(i=dstW>>1; i<dstW; i++){\
				int in= ((buf0[i  ]*yalpha1+buf1[i  ]*yalpha)>>19);\
				int in2 = (76309 * (in - 16) + 32768) >> 16;\
				int in3 = (in2 < 0) ? 0 : ((in2 > 255) ? 255 : in2);\
				int old= (left*7 + topLeft + top[i]*5 + top[i+1]*3)/20 + in3\
					+ (last_new[y][i] - in3)*f/256;\
				int new= old> 128 ? 255 : 0;\
\
				error_new+= ABS(last_new[y][i] - new);\
				error_in3+= ABS(last_in3[y][i] - in3);\
				f= error_new - error_in3*4;\
				if(f<0) f=0;\
				if(f>256) f=256;\
\
				topLeft= top[i];\
				left= top[i]= old - new;\
				last_new[y][i]= new;\
				last_in3[y][i]= in3;\
\
				acc+= acc + (new&1);\
				if((i&7)==6){\
					((uint8_t*)dest)[0]= acc;\
					((uint8_t*)dest)++;\
				}\
			}\
}\
*/\
		}\
		break;\
	case IMGFMT_YUY2:\
		func2\
			((uint8_t*)dest)[2*i2+0]= Y1;\
			((uint8_t*)dest)[2*i2+1]= U;\
			((uint8_t*)dest)[2*i2+2]= Y2;\
			((uint8_t*)dest)[2*i2+3]= V;\
		}		\
		break;\
	case IMGFMT_UYVY:\
		func2\
			((uint8_t*)dest)[2*i2+0]= U;\
			((uint8_t*)dest)[2*i2+1]= Y1;\
			((uint8_t*)dest)[2*i2+2]= V;\
			((uint8_t*)dest)[2*i2+3]= Y2;\
		}		\
		break;\
	}\


static inline void yuv2packedXinC(SwsContext *c, int16_t *lumFilter, int16_t **lumSrc, int lumFilterSize,
				    int16_t *chrFilter, int16_t **chrSrc, int chrFilterSize,
				    uint8_t *dest, int dstW, int y)
{
	int i;
	switch(c->dstFormat)
	{
	case IMGFMT_RGB32:
	case IMGFMT_BGR32:
		YSCALE_YUV_2_RGBX_C(uint32_t)
			((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1];
			((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2];
		}
		break;
	case IMGFMT_RGB24:
		YSCALE_YUV_2_RGBX_C(uint8_t)
			((uint8_t*)dest)[0]= r[Y1];
			((uint8_t*)dest)[1]= g[Y1];
			((uint8_t*)dest)[2]= b[Y1];
			((uint8_t*)dest)[3]= r[Y2];
			((uint8_t*)dest)[4]= g[Y2];
			((uint8_t*)dest)[5]= b[Y2];
			dest+=6;
		}
		break;
	case IMGFMT_BGR24:
		YSCALE_YUV_2_RGBX_C(uint8_t)
			((uint8_t*)dest)[0]= b[Y1];
			((uint8_t*)dest)[1]= g[Y1];
			((uint8_t*)dest)[2]= r[Y1];
			((uint8_t*)dest)[3]= b[Y2];
			((uint8_t*)dest)[4]= g[Y2];
			((uint8_t*)dest)[5]= r[Y2];
			dest+=6;
		}
		break;
	case IMGFMT_RGB16:
	case IMGFMT_BGR16:
		{
			const int dr1= dither_2x2_8[y&1    ][0];
			const int dg1= dither_2x2_4[y&1    ][0];
			const int db1= dither_2x2_8[(y&1)^1][0];
			const int dr2= dither_2x2_8[y&1    ][1];
			const int dg2= dither_2x2_4[y&1    ][1];
			const int db2= dither_2x2_8[(y&1)^1][1];
			YSCALE_YUV_2_RGBX_C(uint16_t)
				((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];
				((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];
			}
		}
		break;
	case IMGFMT_RGB15:
	case IMGFMT_BGR15:
		{
			const int dr1= dither_2x2_8[y&1    ][0];
			const int dg1= dither_2x2_8[y&1    ][1];
			const int db1= dither_2x2_8[(y&1)^1][0];
			const int dr2= dither_2x2_8[y&1    ][1];
			const int dg2= dither_2x2_8[y&1    ][0];
			const int db2= dither_2x2_8[(y&1)^1][1];
			YSCALE_YUV_2_RGBX_C(uint16_t)
				((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];
				((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];
			}
		}
		break;
	case IMGFMT_RGB8:
	case IMGFMT_BGR8:
		{
			const uint8_t * const d64= dither_8x8_73[y&7];
			const uint8_t * const d32= dither_8x8_32[y&7];
			YSCALE_YUV_2_RGBX_C(uint8_t)
				((uint8_t*)dest)[i2+0]= r[Y1+d32[(i2+0)&7]] + g[Y1+d32[(i2+0)&7]] + b[Y1+d64[(i2+0)&7]];
				((uint8_t*)dest)[i2+1]= r[Y2+d32[(i2+1)&7]] + g[Y2+d32[(i2+1)&7]] + b[Y2+d64[(i2+1)&7]];
			}
		}
		break;
	case IMGFMT_RGB4:
	case IMGFMT_BGR4:
		{
			const uint8_t * const d64= dither_8x8_73 [y&7];
			const uint8_t * const d128=dither_8x8_220[y&7];
			YSCALE_YUV_2_RGBX_C(uint8_t)
				((uint8_t*)dest)[i]= r[Y1+d128[(i2+0)&7]] + g[Y1+d64[(i2+0)&7]] + b[Y1+d128[(i2+0)&7]]
				                  +((r[Y2+d128[(i2+1)&7]] + g[Y2+d64[(i2+1)&7]] + b[Y2+d128[(i2+1)&7]])<<4);
			}
		}
		break;
	case IMGFMT_RG4B:
	case IMGFMT_BG4B:
		{
			const uint8_t * const d64= dither_8x8_73 [y&7];
			const uint8_t * const d128=dither_8x8_220[y&7];
			YSCALE_YUV_2_RGBX_C(uint8_t)
				((uint8_t*)dest)[i2+0]= r[Y1+d128[(i2+0)&7]] + g[Y1+d64[(i2+0)&7]] + b[Y1+d128[(i2+0)&7]];
				((uint8_t*)dest)[i2+1]= r[Y2+d128[(i2+1)&7]] + g[Y2+d64[(i2+1)&7]] + b[Y2+d128[(i2+1)&7]];
			}
		}
		break;
	case IMGFMT_RGB1:
	case IMGFMT_BGR1:
		{
			const uint8_t * const d128=dither_8x8_220[y&7];
			uint8_t *g= c->table_gU[128] + c->table_gV[128];
			int acc=0;
			for(i=0; i<dstW-1; i+=2){
				int j;
				int Y1=1<<18;
				int Y2=1<<18;

				for(j=0; j<lumFilterSize; j++)
				{
					Y1 += lumSrc[j][i] * lumFilter[j];
					Y2 += lumSrc[j][i+1] * lumFilter[j];
				}
				Y1>>=19;
				Y2>>=19;
				if((Y1|Y2)&256)
				{
					if(Y1>255)   Y1=255;
					else if(Y1<0)Y1=0;
					if(Y2>255)   Y2=255;
					else if(Y2<0)Y2=0;
				}
				acc+= acc + g[Y1+d128[(i+0)&7]];
				acc+= acc + g[Y2+d128[(i+1)&7]];
				if((i&7)==6){
					((uint8_t*)dest)[0]= acc;
					dest++;
				}
			}
		}
		break;
	case IMGFMT_YUY2:
		YSCALE_YUV_2_PACKEDX_C(void)
			((uint8_t*)dest)[2*i2+0]= Y1;
			((uint8_t*)dest)[2*i2+1]= U;
			((uint8_t*)dest)[2*i2+2]= Y2;
			((uint8_t*)dest)[2*i2+3]= V;
		}
                break;
	case IMGFMT_UYVY:
		YSCALE_YUV_2_PACKEDX_C(void)
			((uint8_t*)dest)[2*i2+0]= U;
			((uint8_t*)dest)[2*i2+1]= Y1;
			((uint8_t*)dest)[2*i2+2]= V;
			((uint8_t*)dest)[2*i2+3]= Y2;
		}
                break;
	}
}


//Note: we have C, X86, MMX, MMX2, 3DNOW version therse no 3DNOW+MMX2 one
//Plain C versions
#if !defined (HAVE_MMX) || defined (RUNTIME_CPUDETECT)
#define COMPILE_C
#endif

#ifdef ARCH_POWERPC
#ifdef HAVE_ALTIVEC
#define COMPILE_ALTIVEC
#endif //HAVE_ALTIVEC
#endif //ARCH_POWERPC

#ifdef ARCH_X86

#if (defined (HAVE_MMX) && !defined (HAVE_3DNOW) && !defined (HAVE_MMX2)) || defined (RUNTIME_CPUDETECT)
#define COMPILE_MMX
#endif

#if defined (HAVE_MMX2) || defined (RUNTIME_CPUDETECT)
#define COMPILE_MMX2
#endif

#if (defined (HAVE_3DNOW) && !defined (HAVE_MMX2)) || defined (RUNTIME_CPUDETECT)
#define COMPILE_3DNOW
#endif
#endif //ARCH_X86

#undef HAVE_MMX
#undef HAVE_MMX2
#undef HAVE_3DNOW

#ifdef COMPILE_C
#undef HAVE_MMX
#undef HAVE_MMX2
#undef HAVE_3DNOW
#undef HAVE_ALTIVEC
#define RENAME(a) a ## _C
#include "swscale_template.c"
#endif

#ifdef ARCH_POWERPC
#ifdef COMPILE_ALTIVEC
#undef RENAME
#define HAVE_ALTIVEC
#define RENAME(a) a ## _altivec
#include "swscale_template.c"
#endif
#endif //ARCH_POWERPC

#ifdef ARCH_X86

//X86 versions
/*
#undef RENAME
#undef HAVE_MMX
#undef HAVE_MMX2
#undef HAVE_3DNOW
#define ARCH_X86
#define RENAME(a) a ## _X86
#include "swscale_template.c"
*/
//MMX versions
#ifdef COMPILE_MMX
#undef RENAME
#define HAVE_MMX
#undef HAVE_MMX2
#undef HAVE_3DNOW
#define RENAME(a) a ## _MMX
#include "swscale_template.c"
#endif

//MMX2 versions
#ifdef COMPILE_MMX2
#undef RENAME
#define HAVE_MMX
#define HAVE_MMX2
#undef HAVE_3DNOW
#define RENAME(a) a ## _MMX2
#include "swscale_template.c"
#endif

//3DNOW versions
#ifdef COMPILE_3DNOW
#undef RENAME
#define HAVE_MMX
#undef HAVE_MMX2
#define HAVE_3DNOW
#define RENAME(a) a ## _3DNow
#include "swscale_template.c"
#endif

#endif //ARCH_X86

// minor note: the HAVE_xyz is messed up after that line so don't use it

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);
}

static inline void initFilter(int16_t **outFilter, int16_t **filterPos, int *outFilterSize, int xInc,
			      int srcW, int dstW, int filterAlign, int one, int flags,
			      SwsVector *srcFilter, SwsVector *dstFilter)
{
	int i;
	int filterSize;
	int filter2Size;
	int minFilterSize;
	double *filter=NULL;
	double *filter2=NULL;
#ifdef ARCH_X86
	if(flags & SWS_CPU_CAPS_MMX)
		asm volatile("emms\n\t"::: "memory"); //FIXME this shouldnt be required but it IS (even for non mmx versions)
#endif

	// Note the +1 is for the MMXscaler which reads over the end
	*filterPos = (int16_t*)memalign(8, (dstW+1)*sizeof(int16_t));

	if(ABS(xInc - 0x10000) <10) // unscaled
	{
		int i;
		filterSize= 1;
		filter= (double*)memalign(8, dstW*sizeof(double)*filterSize);
		for(i=0; i<dstW*filterSize; i++) filter[i]=0;

		for(i=0; i<dstW; i++)
		{
			filter[i*filterSize]=1;
			(*filterPos)[i]=i;
		}

	}
	else if(flags&SWS_POINT) // lame looking point sampling mode
	{
		int i;
		int xDstInSrc;
		filterSize= 1;
		filter= (double*)memalign(8, dstW*sizeof(double)*filterSize);
		
		xDstInSrc= xInc/2 - 0x8000;
		for(i=0; i<dstW; i++)
		{
			int xx= (xDstInSrc - ((filterSize-1)<<15) + (1<<15))>>16;

			(*filterPos)[i]= xx;
			filter[i]= 1.0;
			xDstInSrc+= xInc;
		}
	}
	else if((xInc <= (1<<16) && (flags&SWS_AREA)) || (flags&SWS_FAST_BILINEAR)) // bilinear upscale
	{
		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= (double*)memalign(8, dstW*sizeof(double)*filterSize);

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

			(*filterPos)[i]= xx;
				//Bilinear upscale / linear interpolate / Area averaging
				for(j=0; j<filterSize; j++)
				{
					double d= ABS((xx<<16) - xDstInSrc)/(double)(1<<16);
					double coeff= 1.0 - d;
					if(coeff<0) coeff=0;
					filter[i*filterSize + j]= coeff;
					xx++;
				}
			xDstInSrc+= xInc;
		}
	}
	else
	{
		double xDstInSrc;
		double sizeFactor, filterSizeInSrc;
		const double xInc1= (double)xInc / (double)(1<<16);
		int param= (flags&SWS_PARAM_MASK)>>SWS_PARAM_SHIFT;

		if     (flags&SWS_BICUBIC)	sizeFactor= 4.0;
		else if(flags&SWS_X)		sizeFactor= 8.0;
		else if(flags&SWS_AREA)		sizeFactor= 1.0; //downscale only, for upscale it is bilinear
		else if(flags&SWS_GAUSS)	sizeFactor= 8.0;   // infinite ;)
		else if(flags&SWS_LANCZOS)	sizeFactor= param ? 2.0*param : 6.0;
		else if(flags&SWS_SINC)		sizeFactor= 20.0; // infinite ;)
		else if(flags&SWS_SPLINE)	sizeFactor= 20.0;  // infinite ;)
		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;

		filter= (double*)memalign(16, dstW*sizeof(double)*filterSize);

		xDstInSrc= xInc1 / 2.0 - 0.5;
		for(i=0; i<dstW; i++)
		{
			int xx= (int)(xDstInSrc - (filterSize-1)*0.5 + 0.5);
			int j;
			(*filterPos)[i]= xx;
			for(j=0; j<filterSize; j++)
			{
				double d= ABS(xx - xDstInSrc)/filterSizeInSrc*sizeFactor;
				double coeff;
				if(flags & SWS_BICUBIC)