swscale.c

ffmpeg的完整源代码和作者自己写的文档。不但有在Linux的工程哦

    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)
        {
            int a=0;
            int b=((xpos+xInc)>>16) - xx;
            int c=((xpos+xInc*2)>>16) - xx;
            int d=((xpos+xInc*3)>>16) - xx;

            filter[i  ] = (( xpos         & 0xFFFF) ^ 0xFFFF)>>9;
            filter[i+1] = (((xpos+xInc  ) & 0xFFFF) ^ 0xFFFF)>>9;
            filter[i+2] = (((xpos+xInc*2) & 0xFFFF) ^ 0xFFFF)>>9;
            filter[i+3] = (((xpos+xInc*3) & 0xFFFF) ^ 0xFFFF)>>9;
            filterPos[i/2]= xx;

            if (d+1<4)
            {
                int maxShift= 3-(d+1);
                int shift=0;

                memcpy(funnyCode + fragmentPos, fragmentB, fragmentLengthB);

                funnyCode[fragmentPos + imm8OfPShufW1B]=
                    (a+1) | ((b+1)<<2) | ((c+1)<<4) | ((d+1)<<6);
                funnyCode[fragmentPos + imm8OfPShufW2B]=
                    a | (b<<2) | (c<<4) | (d<<6);

                if (i+3>=dstW) shift=maxShift; //avoid overread
                else if ((filterPos[i/2]&3) <= maxShift) shift=filterPos[i/2]&3; //Align

                if (shift && i>=shift)
                {
                    funnyCode[fragmentPos + imm8OfPShufW1B]+= 0x55*shift;
                    funnyCode[fragmentPos + imm8OfPShufW2B]+= 0x55*shift;
                    filterPos[i/2]-=shift;
                }

                fragmentPos+= fragmentLengthB;
            }
            else
            {
                int maxShift= 3-d;
                int shift=0;

                memcpy(funnyCode + fragmentPos, fragmentA, fragmentLengthA);

                funnyCode[fragmentPos + imm8OfPShufW1A]=
                funnyCode[fragmentPos + imm8OfPShufW2A]=
                    a | (b<<2) | (c<<4) | (d<<6);

                if (i+4>=dstW) shift=maxShift; //avoid overread
                else if ((filterPos[i/2]&3) <= maxShift) shift=filterPos[i/2]&3; //partial align

                if (shift && i>=shift)
                {
                    funnyCode[fragmentPos + imm8OfPShufW1A]+= 0x55*shift;
                    funnyCode[fragmentPos + imm8OfPShufW2A]+= 0x55*shift;
                    filterPos[i/2]-=shift;
                }

                fragmentPos+= fragmentLengthA;
            }

            funnyCode[fragmentPos]= RET;
        }
        xpos+=xInc;
    }
    filterPos[i/2]= xpos>>16; // needed to jump to the next part
}
#endif /* COMPILE_MMX2 */

static void globalInit(void){
    // generating tables:
    int i;
    for (i=0; i<768; i++){
        int c= av_clip_uint8(i-256);
        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;
    else if (flags & SWS_CPU_CAPS_3DNOW)
        return swScale_3DNow;
    else if (flags & SWS_CPU_CAPS_MMX)
        return swScale_MMX;
    else
        return swScale_C;

#else
#ifdef ARCH_POWERPC
    if (flags & SWS_CPU_CAPS_ALTIVEC)
        return swScale_altivec;
    else
        return swScale_C;
#endif
    return swScale_C;
#endif /* defined(ARCH_X86) */
#else //RUNTIME_CPUDETECT
#ifdef HAVE_MMX2
    return swScale_MMX2;
#elif defined (HAVE_3DNOW)
    return swScale_3DNow;
#elif defined (HAVE_MMX)
    return swScale_MMX;
#elif defined (HAVE_ALTIVEC)
    return swScale_altivec;
#else
    return swScale_C;
#endif
#endif //!RUNTIME_CPUDETECT
}

static int PlanarToNV12Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
                               int srcSliceH, uint8_t* dstParam[], int dstStride[]){
    uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;
    /* Copy Y plane */
    if (dstStride[0]==srcStride[0] && srcStride[0] > 0)
        memcpy(dst, src[0], srcSliceH*dstStride[0]);
    else
    {
        int i;
        uint8_t *srcPtr= src[0];
        uint8_t *dstPtr= dst;
        for (i=0; i<srcSliceH; i++)
        {
            memcpy(dstPtr, srcPtr, c->srcW);
            srcPtr+= srcStride[0];
            dstPtr+= dstStride[0];
        }
    }
    dst = dstParam[1] + dstStride[1]*srcSliceY/2;
    if (c->dstFormat == PIX_FMT_NV12)
        interleaveBytes( src[1],src[2],dst,c->srcW/2,srcSliceH/2,srcStride[1],srcStride[2],dstStride[0] );
    else
        interleaveBytes( src[2],src[1],dst,c->srcW/2,srcSliceH/2,srcStride[2],srcStride[1],dstStride[0] );

    return srcSliceH;
}

static int PlanarToYuy2Wrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
                               int srcSliceH, uint8_t* dstParam[], int dstStride[]){
    uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;

    yv12toyuy2( src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0] );

    return srcSliceH;
}

static int PlanarToUyvyWrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
                               int srcSliceH, uint8_t* dstParam[], int dstStride[]){
    uint8_t *dst=dstParam[0] + dstStride[0]*srcSliceY;

    yv12touyvy( src[0],src[1],src[2],dst,c->srcW,srcSliceH,srcStride[0],srcStride[1],dstStride[0] );

    return srcSliceH;
}

/* {RGB,BGR}{15,16,24,32} -> {RGB,BGR}{15,16,24,32} */
static int rgb2rgbWrapper(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
                          int srcSliceH, uint8_t* dst[], int dstStride[]){
    const int srcFormat= c->srcFormat;
    const int dstFormat= c->dstFormat;
    const int srcBpp= (fmt_depth(srcFormat) + 7) >> 3;
    const int dstBpp= (fmt_depth(dstFormat) + 7) >> 3;
    const int srcId= fmt_depth(srcFormat) >> 2; /* 1:0, 4:1, 8:2, 15:3, 16:4, 24:6, 32:8 */
    const int dstId= fmt_depth(dstFormat) >> 2;
    void (*conv)(const uint8_t *src, uint8_t *dst, long src_size)=NULL;

    /* BGR -> BGR */
    if (  (isBGR(srcFormat) && isBGR(dstFormat))
       || (isRGB(srcFormat) && isRGB(dstFormat))){
        switch(srcId | (dstId<<4)){
        case 0x34: conv= rgb16to15; break;
        case 0x36: conv= rgb24to15; break;
        case 0x38: conv= rgb32to15; break;
        case 0x43: conv= rgb15to16; break;
        case 0x46: conv= rgb24to16; break;
        case 0x48: conv= rgb32to16; break;
        case 0x63: conv= rgb15to24; break;
        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 0x33: conv= rgb15tobgr15; break;
        case 0x34: conv= rgb16tobgr15; break;
        case 0x36: conv= rgb24tobgr15; break;
        case 0x38: conv= rgb32tobgr15; break;
        case 0x43: conv= rgb15tobgr16; break;
        case 0x44: conv= rgb16tobgr16; break;
        case 0x46: conv= rgb24tobgr16; break;
        case 0x48: conv= rgb32tobgr16; break;
        case 0x63: conv= rgb15tobgr24; break;
        case 0x64: conv= rgb16tobgr24; break;
        case 0x66: conv= rgb24tobgr24; break;
        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(conv)
    {
        if (dstStride[0]*srcBpp == srcStride[0]*dstBpp)
            conv(src[0], dst[0] + dstStride[0]*srcSliceY, srcSliceH*srcStride[0]);
        else
        {
            int i;
            uint8_t *srcPtr= src[0];
            uint8_t *dstPtr= dst[0] + dstStride[0]*srcSliceY;

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

        for (i=0; i<srcSliceH; i++)
        {
            memcpy(dstPtr, srcPtr, c->srcW);
            srcPtr+= srcStride[0];
            dstPtr+= dstStride[0];
        }
    }

    if (c->dstFormat==PIX_FMT_YUV420P){
        planar2x(src[1], dst[1], c->chrSrcW, c->chrSrcH, srcStride[1], dstStride[1]);
        planar2x(src[2], dst[2], c->chrSrcW, c->chrSrcH, srcStride[2], dstStride[2]);
    }else{
        planar2x(src[1], dst[2], c->chrSrcW, c->chrSrcH, srcStride[1], dstStride[2]);
        planar2x(src[2], dst[1], c->chrSrcW, c->chrSrcH, srcStride[2], dstStride[1]);
    }
    return srcSliceH;
}

/* unscaled copy like stuff (assumes nearly identical formats) */
static int simpleCopy(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
                      int srcSliceH, uint8_t* dst[], int dstStride[]){

    if (isPacked(c->srcFormat))
    {
        if (dstStride[0]==srcStride[0] && srcStride[0] > 0)
            memcpy(dst[0] + dstStride[0]*srcSliceY, src[0], srcSliceH*dstStride[0]);
        else
        {
            int i;
            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++)
            {
                memcpy(dstPtr, srcPtr, length);
                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);
            int y=      plane==0 ? srcSliceY: -((-srcSliceY)>>c->chrDstVSubSample);
            int height= plane==0 ? srcSliceH: -((-srcSliceH)>>c->chrDstVSubSample);

            if ((isGray(c->srcFormat) || isGray(c->dstFormat)) && plane>0)
            {
                if (!isGray(c->dstFormat))
                    memset(dst[plane], 128, dstStride[plane]*height);
            }
            else
            {
                if (dstStride[plane]==srcStride[plane] && srcStride[plane] > 0)
                    memcpy(dst[plane] + dstStride[plane]*y, src[plane], height*dstStride[plane]);
                else
                {
                    int i;
                    uint8_t *srcPtr= src[plane];
                    uint8_t *dstPtr= dst[plane] + dstStride[plane]*y;
                    for (i=0; i<height; i++)
                    {
                        memcpy(dstPtr, srcPtr, length);
                        srcPtr+= srcStride[plane];
                        dstPtr+= dstStride[plane];
                    }
                }
            }
        }
    }
    return srcSliceH;
}

static int gray16togray(SwsContext *c, uint8_t* src[], int srcStride[], int srcSliceY,
                        int srcSliceH, uint8_t* dst[], int dstStride[]){

    int length= c->srcW;
    int y=      srcSliceY;
    int height= srcSliceH;
    int i, j;
    uint8_t *srcPtr= src[0];
    uint8_t *dstPtr= dst[0] + dst