swscale.c

uclinux 下的vlc播放器源代码

	srcFormat = remove_dup_fourcc(origSrcFormat);
	dstFormat = remove_dup_fourcc(origDstFormat);

	unscaled = (srcW == dstW && srcH == dstH);
	needsDither= (isBGR(dstFormat) || isRGB(dstFormat)) 
		     && (dstFormat&0xFF)<24
		     && ((dstFormat&0xFF)<(srcFormat&0xFF) || (!(isRGB(srcFormat) || isBGR(srcFormat))));

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

	/* 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
	{
		 MSG_ERR("swScaler: %dx%d -> %dx%d is invalid scaling dimension\n", 
			srcW, srcH, dstW, dstH);
		return NULL;
	}

	if(!dstFilter) dstFilter= &dummyFilter;
	if(!srcFilter) srcFilter= &dummyFilter;

	c= memalign(64, sizeof(SwsContext));
	memset(c, 0, sizeof(SwsContext));

	c->srcW= srcW;
	c->srcH= srcH;
	c->dstW= dstW;
	c->dstH= dstH;
	c->lumXInc= ((srcW<<16) + (dstW>>1))/dstW;
	c->lumYInc= ((srcH<<16) + (dstH>>1))/dstH;
	c->flags= flags;
	c->dstFormat= dstFormat;
	c->srcFormat= srcFormat;
	c->origDstFormat= origDstFormat;
	c->origSrcFormat= origSrcFormat;
        c->vRounder= 4* 0x0001000100010001ULL;

	usesHFilter= usesVFilter= 0;
	if(dstFilter->lumV!=NULL && dstFilter->lumV->length>1) usesVFilter=1;
	if(dstFilter->lumH!=NULL && dstFilter->lumH->length>1) usesHFilter=1;
	if(dstFilter->chrV!=NULL && dstFilter->chrV->length>1) usesVFilter=1;
	if(dstFilter->chrH!=NULL && dstFilter->chrH->length>1) usesHFilter=1;
	if(srcFilter->lumV!=NULL && srcFilter->lumV->length>1) usesVFilter=1;
	if(srcFilter->lumH!=NULL && srcFilter->lumH->length>1) usesHFilter=1;
	if(srcFilter->chrV!=NULL && srcFilter->chrV->length>1) usesVFilter=1;
	if(srcFilter->chrH!=NULL && srcFilter->chrH->length>1) usesHFilter=1;

	getSubSampleFactors(&c->chrSrcHSubSample, &c->chrSrcVSubSample, srcFormat);
	getSubSampleFactors(&c->chrDstHSubSample, &c->chrDstVSubSample, dstFormat);

	// reuse chroma for 2 pixles rgb/bgr unless user wants full chroma interpolation
	if((isBGR(dstFormat) || isRGB(dstFormat)) && !(flags&SWS_FULL_CHR_H_INT)) c->chrDstHSubSample=1;

	// drop some chroma lines if the user wants it
	c->vChrDrop= (flags&SWS_SRC_V_CHR_DROP_MASK)>>SWS_SRC_V_CHR_DROP_SHIFT;
	c->chrSrcVSubSample+= c->vChrDrop;

	// drop every 2. pixel for chroma calculation unless user wants full chroma
	if((isBGR(srcFormat) || isRGB(srcFormat)) && !(flags&SWS_FULL_CHR_H_INP)) 
		c->chrSrcHSubSample=1;

	c->chrIntHSubSample= c->chrDstHSubSample;
	c->chrIntVSubSample= c->chrSrcVSubSample;

	// note the -((-x)>>y) is so that we allways round toward +inf
	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], 0, Inverse_Table_6_9[SWS_CS_DEFAULT] /* FIXME*/, 0, 0, 1<<16, 1<<16); 

	/* unscaled special Cases */
	if(unscaled && !usesHFilter && !usesVFilter)
	{
		/* yv12_to_nv12 */
		if(srcFormat == IMGFMT_YV12 && dstFormat == IMGFMT_NV12)
		{
			c->swScale= PlanarToNV12Wrapper;
		}
		/* yuv2bgr */
		if((srcFormat==IMGFMT_YV12 || srcFormat==IMGFMT_422P) && (isBGR(dstFormat) || isRGB(dstFormat)))
		{
			c->swScale= yuv2rgb_get_func_ptr(c);
		}
		
		if( srcFormat==IMGFMT_YVU9 && dstFormat==IMGFMT_YV12 )
		{
			c->swScale= yvu9toyv12Wrapper;
		}

		/* bgr24toYV12 */
		if(srcFormat==IMGFMT_BGR24 && dstFormat==IMGFMT_YV12)
			c->swScale= bgr24toyv12Wrapper;
		
		/* rgb/bgr -> rgb/bgr (no dither needed forms) */
		if(   (isBGR(srcFormat) || isRGB(srcFormat))
		   && (isBGR(dstFormat) || isRGB(dstFormat)) 
		   && !needsDither)
			c->swScale= rgb2rgbWrapper;

		/* 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 == IMGFMT_YV12 && 
			    (dstFormat == IMGFMT_YUY2 || dstFormat == IMGFMT_UYVY))
			{
				if (dstFormat == IMGFMT_YUY2)
				    c->swScale= PlanarToYuy2Wrapper;
				else
				    c->swScale= PlanarToUyvyWrapper;
			}
		}

#ifdef HAVE_ALTIVEC
		if ((c->flags & SWS_CPU_CAPS_ALTIVEC) &&
		    ((srcFormat == IMGFMT_YV12 && 
		      (dstFormat == IMGFMT_YUY2 || dstFormat == IMGFMT_UYVY)))) {
		  // unscaled YV12 -> packed YUV, we want speed
		  if (dstFormat == IMGFMT_YUY2)
		    c->swScale= yv12toyuy2_unscaled_altivec;
		  else
		    c->swScale= yv12touyvy_unscaled_altivec;
		}
#endif

		/* simple copy */
		if(   srcFormat == dstFormat
		   || (isPlanarYUV(srcFormat) && isGray(dstFormat))
		   || (isPlanarYUV(dstFormat) && isGray(srcFormat))
		  )
		{
			c->swScale= simpleCopy;
		}

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

	if(flags & SWS_CPU_CAPS_MMX2)
	{
		c->canMMX2BeUsed= (dstW >=srcW && (dstW&31)==0 && (srcW&15)==0) ? 1 : 0;
		if(!c->canMMX2BeUsed && dstW >=srcW && (srcW&15)==0 && (flags&SWS_FAST_BILINEAR))
		{
			if(flags&SWS_PRINT_INFO)
				MSG_INFO("SwScaler: output Width is not a multiple of 32 -> no MMX2 scaler\n");
		}
		if(usesHFilter) c->canMMX2BeUsed=0;
	}
	else
		c->canMMX2BeUsed=0;

	c->chrXInc= ((c->chrSrcW<<16) + (c->chrDstW>>1))/c->chrDstW;
	c->chrYInc= ((c->chrSrcH<<16) + (c->chrDstH>>1))/c->chrDstH;

	// match pixel 0 of the src to pixel 0 of dst and match pixel n-2 of src to pixel n-2 of dst
	// but only for the FAST_BILINEAR mode otherwise do correct scaling
	// n-2 is the last chrominance sample available
	// this is not perfect, but noone shuld notice the difference, the more correct variant
	// would be like the vertical one, but that would require some special code for the
	// first and last pixel
	if(flags&SWS_FAST_BILINEAR)
	{
		if(c->canMMX2BeUsed)
		{
			c->lumXInc+= 20;
			c->chrXInc+= 20;
		}
		//we don't use the x86asm scaler if mmx is available
		else if(flags & SWS_CPU_CAPS_MMX)
		{
			c->lumXInc = ((srcW-2)<<16)/(dstW-2) - 20;
			c->chrXInc = ((c->chrSrcW-2)<<16)/(c->chrDstW-2) - 20;
		}
	}

	/* precalculate horizontal scaler filter coefficients */
	{
		const int filterAlign=
		  (flags & SWS_CPU_CAPS_MMX) ? 4 :
		  (flags & SWS_CPU_CAPS_ALTIVEC) ? 8 :
		  1;

		initFilter(&c->hLumFilter, &c->hLumFilterPos, &c->hLumFilterSize, c->lumXInc,
				 srcW      ,       dstW, filterAlign, 1<<14,
				 (flags&SWS_BICUBLIN) ? (flags|SWS_BICUBIC)  : flags,
				 srcFilter->lumH, dstFilter->lumH);
		initFilter(&c->hChrFilter, &c->hChrFilterPos, &c->hChrFilterSize, c->chrXInc,
				 c->chrSrcW, c->chrDstW, filterAlign, 1<<14,
				 (flags&SWS_BICUBLIN) ? (flags|SWS_BILINEAR) : flags,
				 srcFilter->chrH, dstFilter->chrH);

#ifdef ARCH_X86
// can't downscale !!!
		if(c->canMMX2BeUsed && (flags & SWS_FAST_BILINEAR))
		{
			c->lumMmx2Filter   = (int16_t*)memalign(8, (dstW        /8+8)*sizeof(int16_t));
			c->chrMmx2Filter   = (int16_t*)memalign(8, (c->chrDstW  /4+8)*sizeof(int16_t));
			c->lumMmx2FilterPos= (int32_t*)memalign(8, (dstW      /2/8+8)*sizeof(int32_t));
			c->chrMmx2FilterPos= (int32_t*)memalign(8, (c->chrDstW/2/4+8)*sizeof(int32_t));

			initMMX2HScaler(      dstW, c->lumXInc, c->funnyYCode , c->lumMmx2Filter, c->lumMmx2FilterPos, 8);
			initMMX2HScaler(c->chrDstW, c->chrXInc, c->funnyUVCode, c->chrMmx2Filter, c->chrMmx2FilterPos, 4);
		}
#endif
	} // Init Horizontal stuff



	/* precalculate vertical scaler filter coefficients */
	{
		const int filterAlign=
		  (flags & SWS_CPU_CAPS_ALTIVEC) ? 8 :
		  1;

		initFilter(&c->vLumFilter, &c->vLumFilterPos, &c->vLumFilterSize, c->lumYInc,
				srcH      ,        dstH, filterAlign, (1<<12)-4,
				(flags&SWS_BICUBLIN) ? (flags|SWS_BICUBIC)  : flags,
				srcFilter->lumV, dstFilter->lumV);
		initFilter(&c->vChrFilter, &c->vChrFilterPos, &c->vChrFilterSize, c->chrYInc,
				c->chrSrcH, c->chrDstH, filterAlign, (1<<12)-4,
				(flags&SWS_BICUBLIN) ? (flags|SWS_BILINEAR) : flags,
				srcFilter->chrV, dstFilter->chrV);
	}

	// Calculate Buffer Sizes so that they won't run out while handling these damn slices
	c->vLumBufSize= c->vLumFilterSize;
	c->vChrBufSize= c->vChrFilterSize;
	for(i=0; i<dstH; i++)
	{
		int chrI= i*c->chrDstH / dstH;
		int nextSlice= MAX(c->vLumFilterPos[i   ] + c->vLumFilterSize - 1,
				 ((c->vChrFilterPos[chrI] + c->vChrFilterSize - 1)<<c->chrSrcVSubSample));

		nextSlice>>= c->chrSrcVSubSample;
		nextSlice<<= c->chrSrcVSubSample;
		if(c->vLumFilterPos[i   ] + c->vLumBufSize < nextSlice)
			c->vLumBufSize= nextSlice - c->vLumFilterPos[i   ];
		if(c->vChrFilterPos[chrI] + c->vChrBufSize < (nextSlice>>c->chrSrcVSubSample))
			c->vChrBufSize= (nextSlice>>c->chrSrcVSubSample) - c->vChrFilterPos[chrI];
	}

	// allocate pixbufs (we use dynamic allocation because otherwise we would need to
	c->lumPixBuf= (int16_t**)memalign(4, c->vLumBufSize*2*sizeof(int16_t*));
	c->chrPixBuf= (int16_t**)memalign(4, c->vChrBufSize*2*sizeof(int16_t*));
	//Note we need at least one pixel more at the end because of the mmx code (just in case someone wanna replace the 4000/8000)
	for(i=0; i<c->vLumBufSize; i++)
		c->lumPixBuf[i]= c->lumPixBuf[i+c->vLumBufSize]= (uint16_t*)memalign(8, 4000);
	for(i=0; i<c->vChrBufSize; i++)
		c->chrPixBuf[i]= c->chrPixBuf[i+c->vChrBufSize]= (uint16_t*)memalign(8, 8000);

	//try to avoid drawing green stuff between the right end and the stride end
	for(i=0; i<c->vLumBufSize; i++) memset(c->lumPixBuf[i], 0, 4000);
	for(i=0; i<c->vChrBufSize; i++) memset(c->chrPixBuf[i], 64, 8000);

	ASSERT(c->chrDstH <= dstH)

	if(flags&SWS_PRINT_INFO)
	{
#ifdef DITHER1XBPP
		char *dither= " dithered";
#else
		char *dither= "";
#endif
		if(flags&SWS_FAST_BILINEAR)
			MSG_INFO("\nSwScaler: FAST_BILINEAR scaler, ");
		else if(flags&SWS_BILINEAR)
			MSG_INFO("\nSwScaler: BILINEAR scaler, ");
		else if(flags&SWS_BICUBIC)
			MSG_INFO("\nSwScaler: BICUBIC scaler, ");
		else if(flags&SWS_X)
			MSG_INFO("\nSwScaler: Experimental scaler, ");
		else if(flags&SWS_POINT)
			MSG_INFO("\nSwScaler: Nearest Neighbor / POINT scaler, ");
		else if(flags&SWS_AREA)
			MSG_INFO("\nSwScaler: Area Averageing scaler, ");
		else if(flags&SWS_BICUBLIN)
			MSG_INFO("\nSwScaler: luma BICUBIC / chroma BILINEAR scaler, ");
		else if(flags&SWS_GAUSS)
			MSG_INFO("\nSwScaler: Gaussian scaler, ");
		else if(flags&SWS_SINC)
			MSG_INFO("\nSwScaler: Sinc scaler, ");
		else if(flags&SWS_LANCZOS)
			MSG_INFO("\nSwScaler: Lanczos scaler, ");
		else if(flags&SWS_SPLINE)
			MSG_INFO("\nSwScaler: Bicubic spline scaler, ");
		else
			MSG_INFO("\nSwScaler: ehh flags invalid?! ");

		if(dstFormat==IMGFMT_BGR15 || dstFormat==IMGFMT_BGR16)
			MSG_INFO("from %s to%s %s ", 
				vo_format_name(srcFormat), dither, vo_format_name(dstFormat));
		else
			MSG_INFO("from %s to %s ", 
				vo_format_name(srcFormat), vo_format_name(dstFormat));

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

	if(flags & SWS_PRINT_INFO)
	{
		if(flags & SWS_CPU_CAPS_MMX)
		{
			if(c->canMMX2BeUsed && (flags&SWS_FAST_BILINEAR))
				MSG_V("SwScaler: using FAST_BILINEAR MMX2 scaler for horizontal scaling\n");
			else
			{
				if(c->hLumFilterSize==4)
					MSG_V("SwScaler: using 4-tap MMX scaler for horizontal luminance scaling\n");
				else if(c->hLumFilterSize==8)
					MSG_V("SwScaler: using 8-tap MMX scaler for horizontal luminance scaling\n");
				else
					MSG_V("SwScaler: using n-tap MMX scaler for horizontal luminance scaling\n");

				if(c->hChrFilterSize==4)
					MSG_V("SwScaler: using 4-tap MMX scaler for horizontal chrominance scaling\n");
				else if(c->hChrFilterSize==8)
					MSG_V("SwScaler: using 8-tap MMX scaler for horizontal chrominance scaling\n");
				else
					MSG_V("SwScaler: using n-tap MMX scaler for horizontal chrominance scaling\n");
			}
		}
		else
		{
#ifdef ARCH_X86
			MSG_V("SwScaler: using X86-Asm scaler for horizontal scaling\n");
#else
			if(flags & SWS_FAST_BILINEAR)
				MSG_V("SwScaler: using FAST_BILINEAR C scaler for horizontal scaling\n");
			else
				MSG_V("SwScaler: using C scaler for horizontal scaling\n");
#endif
		}
		if(isPlanarYUV(dstFormat))
		{
			if(c->vLumFilterSize==1)
				MSG_V("SwScaler: using 1-tap %s \"scaler\" for vertical scaling (YV12 like)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
			else
				MSG_V("SwScaler: using n-tap %s scaler for vertical scaling (YV12 like)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
		}
		else
		{
			if(c->vLumFilterSize==1 && c->vChrFilterSize==2)
				MSG_V("SwScaler: using 1-tap %s \"scaler\" for vertical luminance scaling (BGR)\n"
				       "SwScaler:       2-tap scaler for vertical chrominance scaling (BGR)\n",(flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
			else if(c->vLumFilterSize==2 && c->vChrFilterSize==2)
				MSG_V("SwScaler: using 2-tap linear %s scaler for vertical scaling (BGR)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
			else
				MSG_V("SwScaler: using n-tap %s scaler for vertical scaling (BGR)\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
		}

		if(dstFormat==IMGFMT_BGR24)
			MSG_V("SwScaler: using %s YV12->BGR24 Converter\n",
				(flags & SWS_CPU_CAPS_MMX2) ? "MMX2" : ((flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C"));
		else if(dstFormat==IMGFMT_BGR32)
			MSG_V("SwScaler: using %s YV12->BGR32 Converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
		else if(dstFormat==IMGFMT_BGR16)
			MSG_V("SwScaler: using %s YV12->BGR16 Converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");
		else if(dstFormat==IMGFMT_BGR15)
			MSG_V("SwScaler: using %s YV12->BGR15 Converter\n", (flags & SWS_CPU_CAPS_MMX) ? "MMX" : "C");

		MSG_V("SwScaler: %dx%d -> %dx%d\n", srcW, srcH, dstW, dstH);
	}
	if(flags & SWS_PRINT_INFO)
	{
		MSG_DBG2("SwScaler:Lum srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n",
			c->srcW, c->srcH, c->dstW, c->dstH, c->lumXInc, c->lumYInc);
		MSG_DBG2("SwScaler:Chr srcW=%d srcH=%d dstW=%d dstH=%d xInc=%d yInc=%d\n",
			c->chrSrcW, c->chrSrcH, c->chrDstW, c->chrDstH, c->chrXInc, c->chrYInc);
	}

	c->swScale= getSwsFunc(flags);
	return c;
}

/**
 * swscale warper, so we don't need to export the SwsContext.
 * assumes planar YUV to be in YUV order instead of YVU
 */
int sws_scale_ordered(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
                           int srcSliceH, uint8_t* dst[], int dstStride[]){
	//copy strides, so they can safely be modified
	int srcStride2[3]= {srcStride[0], srcStride[1], srcStride[2]};
	int dstStride2[3]= {dstStride[0], dstStride[1], dstStride[2]};
	return c->swScale(c, src, srcStride2, srcSliceY, srcSliceH, dst, dstStride2);
}

/**
 * swscale warper, so we don't need to export the SwsContext
 */
int sws_scale(SwsContext *c, uint8_t* srcParam[], int srcStrideParam[], int srcSliceY,
                           int srcSliceH, uint8_t* dstParam[], int dstStrideParam[]){
	int srcStride[3];
	int dstStride[3];
	uint8_t *src[3];
	uint8_t *dst[3];
	sws_orderYUV(c->origSrcFormat, src, srcStride, srcParam, srcStrideParam);
	sws_orderYUV(c->origDstFormat, dst, dstStride, dstParam, dstStrideParam);
//printf("sws: slice %d %d\n", srcSliceY, srcSliceH);

	return c->swScale(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= malloc(sizeof(SwsFilter));

	if(lumaGBlur!=0.0){
		filter->lumH= sws_getGaussianVec(lumaGBlur, 3.0);
		filter->lumV= sws_getGaussianVec(lumaGBlur, 3.0);
	}else{
		filter->lumH= sws_getIden