dsputil.c

64-bits H.264 from ffmpeg 2008 version Build in VC++ 2008 no error warning, Jesse Stone, Taiwan

      {
        {
            // 9/7 8x8 dec=3
            {268, 239, 239, 213},
            {  0, 224, 224, 152},
            {  0, 135, 135, 110},
        },{
            // 9/7 16x16 or 32x32 dec=4
            {344, 310, 310, 280},
            {  0, 320, 320, 228},
            {  0, 175, 175, 136},
            {  0, 129, 129, 102},
        }
      },{
        {
            // 5/3 8x8 dec=3
            {275, 245, 245, 218},
            {  0, 230, 230, 156},
            {  0, 138, 138, 113},
        },{
            // 5/3 16x16 or 32x32 dec=4
            {352, 317, 317, 286},
            {  0, 328, 328, 233},
            {  0, 180, 180, 140},
            {  0, 132, 132, 105},
        }
      }
    };

    for (i = 0; i < h; i++) {
        for (j = 0; j < w; j+=4) {
            tmp[32*i+j+0] = (pix1[j+0] - pix2[j+0])<<4;
            tmp[32*i+j+1] = (pix1[j+1] - pix2[j+1])<<4;
            tmp[32*i+j+2] = (pix1[j+2] - pix2[j+2])<<4;
            tmp[32*i+j+3] = (pix1[j+3] - pix2[j+3])<<4;
        }
        pix1 += line_size;
        pix2 += line_size;
    }

    ff_spatial_dwt(tmp, w, h, 32, type, dec_count);

    s=0;
    assert(w==h);
    for(level=0; level<dec_count; level++){
        for(ori= level ? 1 : 0; ori<4; ori++){
            int size= w>>(dec_count-level);
            int sx= (ori&1) ? size : 0;
            int stride= 32<<(dec_count-level);
            int sy= (ori&2) ? stride>>1 : 0;

            for(i=0; i<size; i++){
                for(j=0; j<size; j++){
                    int v= tmp[sx + sy + i*stride + j] * scale[type][dec_count-3][level][ori];
                    s += FFABS(v);
                }
            }
        }
    }
    assert(s>=0);
    return s>>9;
}

static int w53_8_c(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int h){
    return w_c(v, pix1, pix2, line_size,  8, h, 1);
}

static int w97_8_c(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int h){
    return w_c(v, pix1, pix2, line_size,  8, h, 0);
}

static int w53_16_c(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int h){
    return w_c(v, pix1, pix2, line_size, 16, h, 1);
}

static int w97_16_c(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int h){
    return w_c(v, pix1, pix2, line_size, 16, h, 0);
}

int w53_32_c(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int h){
    return w_c(v, pix1, pix2, line_size, 32, h, 1);
}

int w97_32_c(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int h){
    return w_c(v, pix1, pix2, line_size, 32, h, 0);
}
#endif

/* draw the edges of width 'w' of an image of size width, height */
//FIXME check that this is ok for mpeg4 interlaced
static void draw_edges_c(uint8_t *buf, int wrap, int width, int height, int w)
{
    uint8_t *ptr, *last_line;
    int i;

    last_line = buf + (height - 1) * wrap;
    for(i=0;i<w;i++) {
        /* top and bottom */
        memcpy(buf - (i + 1) * wrap, buf, width);
        memcpy(last_line + (i + 1) * wrap, last_line, width);
    }
    /* left and right */
    ptr = buf;
    for(i=0;i<height;i++) {
        memset(ptr - w, ptr[0], w);
        memset(ptr + width, ptr[width-1], w);
        ptr += wrap;
    }
    /* corners */
    for(i=0;i<w;i++) {
        memset(buf - (i + 1) * wrap - w, buf[0], w); /* top left */
        memset(buf - (i + 1) * wrap + width, buf[width-1], w); /* top right */
        memset(last_line + (i + 1) * wrap - w, last_line[0], w); /* top left */
        memset(last_line + (i + 1) * wrap + width, last_line[width-1], w); /* top right */
    }
}

/**
 * Copies a rectangular area of samples to a temporary buffer and replicates the boarder samples.
 * @param buf destination buffer
 * @param src source buffer
 * @param linesize number of bytes between 2 vertically adjacent samples in both the source and destination buffers
 * @param block_w width of block
 * @param block_h height of block
 * @param src_x x coordinate of the top left sample of the block in the source buffer
 * @param src_y y coordinate of the top left sample of the block in the source buffer
 * @param w width of the source buffer
 * @param h height of the source buffer
 */
void ff_emulated_edge_mc(uint8_t *buf, uint8_t *src, int linesize, int block_w, int block_h,
                                    int src_x, int src_y, int w, int h){
    int x, y;
    int start_y, start_x, end_y, end_x;

    if(src_y>= h){
        src+= (h-1-src_y)*linesize;
        src_y=h-1;
    }else if(src_y<=-block_h){
        src+= (1-block_h-src_y)*linesize;
        src_y=1-block_h;
    }
    if(src_x>= w){
        src+= (w-1-src_x);
        src_x=w-1;
    }else if(src_x<=-block_w){
        src+= (1-block_w-src_x);
        src_x=1-block_w;
    }

    start_y= FFMAX(0, -src_y);
    start_x= FFMAX(0, -src_x);
    end_y= FFMIN(block_h, h-src_y);
    end_x= FFMIN(block_w, w-src_x);

    // copy existing part
    for(y=start_y; y<end_y; y++){
        for(x=start_x; x<end_x; x++){
            buf[x + y*linesize]= src[x + y*linesize];
        }
    }

    //top
    for(y=0; y<start_y; y++){
        for(x=start_x; x<end_x; x++){
            buf[x + y*linesize]= buf[x + start_y*linesize];
        }
    }

    //bottom
    for(y=end_y; y<block_h; y++){
        for(x=start_x; x<end_x; x++){
            buf[x + y*linesize]= buf[x + (end_y-1)*linesize];
        }
    }

    for(y=0; y<block_h; y++){
       //left
        for(x=0; x<start_x; x++){
            buf[x + y*linesize]= buf[start_x + y*linesize];
        }

       //right
        for(x=end_x; x<block_w; x++){
            buf[x + y*linesize]= buf[end_x - 1 + y*linesize];
        }
    }
}

static void get_pixels_c(DCTELEM *restrict block, const uint8_t *pixels, int line_size)
{
    int i;

    /* read the pixels */
    for(i=0;i<8;i++) {
        block[0] = pixels[0];
        block[1] = pixels[1];
        block[2] = pixels[2];
        block[3] = pixels[3];
        block[4] = pixels[4];
        block[5] = pixels[5];
        block[6] = pixels[6];
        block[7] = pixels[7];
        pixels += line_size;
        block += 8;
    }
}

static void diff_pixels_c(DCTELEM *restrict block, const uint8_t *s1,
                          const uint8_t *s2, int stride){
    int i;

    /* read the pixels */
    for(i=0;i<8;i++) {
        block[0] = s1[0] - s2[0];
        block[1] = s1[1] - s2[1];
        block[2] = s1[2] - s2[2];
        block[3] = s1[3] - s2[3];
        block[4] = s1[4] - s2[4];
        block[5] = s1[5] - s2[5];
        block[6] = s1[6] - s2[6];
        block[7] = s1[7] - s2[7];
        s1 += stride;
        s2 += stride;
        block += 8;
    }
}


static void put_pixels_clamped_c(const DCTELEM *block, uint8_t *restrict pixels,
                                 int line_size)
{
    int i;
    uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;

    /* read the pixels */
    for(i=0;i<8;i++) {
        pixels[0] = cm[block[0]];
        pixels[1] = cm[block[1]];
        pixels[2] = cm[block[2]];
        pixels[3] = cm[block[3]];
        pixels[4] = cm[block[4]];
        pixels[5] = cm[block[5]];
        pixels[6] = cm[block[6]];
        pixels[7] = cm[block[7]];

        pixels += line_size;
        block += 8;
    }
}

/*static void put_pixels_clamped4_c(const DCTELEM *block, uint8_t *restrict pixels,
                                 int line_size)
{
    int i;
    uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;

    for(i=0;i<4;i++) {
        pixels[0] = cm[block[0]];
        pixels[1] = cm[block[1]];
        pixels[2] = cm[block[2]];
        pixels[3] = cm[block[3]];

        pixels += line_size;
        block += 8;
    }
}*/

/*static void put_pixels_clamped2_c(const DCTELEM *block, uint8_t *restrict pixels,
                                 int line_size)
{
    int i;
    uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;

    for(i=0;i<2;i++) {
        pixels[0] = cm[block[0]];
        pixels[1] = cm[block[1]];

        pixels += line_size;
        block += 8;
    }
}*/

static void put_signed_pixels_clamped_c(const DCTELEM *block,
                                        uint8_t *restrict pixels,
                                        int line_size)
{
    int i, j;

    for (i = 0; i < 8; i++) {
        for (j = 0; j < 8; j++) {
            if (*block < -128)
                *pixels = 0;
            else if (*block > 127)
                *pixels = 255;
            else
                *pixels = (uint8_t)(*block + 128);
            block++;
            pixels++;
        }
        pixels += (line_size - 8);
    }
}

static void add_pixels_clamped_c(const DCTELEM *block, uint8_t *restrict pixels,
                          int line_size)
{
    int i;
    uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;

    /* read the pixels */
    for(i=0;i<8;i++) {
        pixels[0] = cm[pixels[0] + block[0]];
        pixels[1] = cm[pixels[1] + block[1]];
        pixels[2] = cm[pixels[2] + block[2]];
        pixels[3] = cm[pixels[3] + block[3]];
        pixels[4] = cm[pixels[4] + block[4]];
        pixels[5] = cm[pixels[5] + block[5]];
        pixels[6] = cm[pixels[6] + block[6]];
        pixels[7] = cm[pixels[7] + block[7]];
        pixels += line_size;
        block += 8;
    }
}

/*static void add_pixels_clamped4_c(const DCTELEM *block, uint8_t *restrict pixels,
                          int line_size)
{
    int i;
    uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;

    for(i=0;i<4;i++) {
        pixels[0] = cm[pixels[0] + block[0]];
        pixels[1] = cm[pixels[1] + block[1]];
        pixels[2] = cm[pixels[2] + block[2]];
        pixels[3] = cm[pixels[3] + block[3]];
        pixels += line_size;
        block += 8;
    }
}*/

/*static void add_pixels_clamped2_c(const DCTELEM *block, uint8_t *restrict pixels,
                          int line_size)
{
    int i;
    uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;

    for(i=0;i<2;i++) {
        pixels[0] = cm[pixels[0] + block[0]];
        pixels[1] = cm[pixels[1] + block[1]];
        pixels += line_size;
        block += 8;
    }
}*/

static void add_pixels8_c(uint8_t *restrict pixels, DCTELEM *block, int line_size)