comparison dsputil.c @ 1708:dea5b2946999 libavcodec

interlaced motion estimation interlaced mpeg2 encoding P & B frames rate distored interlaced mb decission alternate scantable support 4mv encoding fixes (thats also why the regression tests change) passing height to most dsp functions interlaced mpeg4 encoding (no direct mode MBs yet) various related cleanups disabled old motion estimaton algorithms (log, full, ...) they will either be fixed or removed

1707:027545a2fdbe

    for(;i<w; i++){
        dst[i+0]= bswap_32(src[i+0]);
    }
}

static int sse8_c(void *v, uint8_t * pix1, uint8_t * pix2, int line_size)
{
    int s, i;
    uint32_t *sq = squareTbl + 256;

    s = 0;
    for (i = 0; i < 8; i++) {
        s += sq[pix1[0] - pix2[0]];
        s += sq[pix1[1] - pix2[1]];
        s += sq[pix1[2] - pix2[2]];
        s += sq[pix1[3] - pix2[3]];
        s += sq[pix1[4] - pix2[4]];

        pix2 += line_size;
    }
    return s;
}

static int sse16_c(void *v, uint8_t *pix1, uint8_t *pix2, int line_size)
{
    int s, i;
    uint32_t *sq = squareTbl + 256;

    s = 0;
    for (i = 0; i < 16; i++) {
        s += sq[pix1[ 0] - pix2[ 0]];
        s += sq[pix1[ 1] - pix2[ 1]];
        s += sq[pix1[ 2] - pix2[ 2]];
        s += sq[pix1[ 3] - pix2[ 3]];
        s += sq[pix1[ 4] - pix2[ 4]];

        src[y*stride-2] = p0 - d2;
        src[y*stride+1] = p3 + d2;
    }
}

static inline int pix_abs16x16_c(uint8_t *pix1, uint8_t *pix2, int line_size)
{
    int s, i;

    s = 0;
    for(i=0;i<16;i++) {
        s += abs(pix1[0] - pix2[0]);
        s += abs(pix1[1] - pix2[1]);
        s += abs(pix1[2] - pix2[2]);
        s += abs(pix1[3] - pix2[3]);
        s += abs(pix1[4] - pix2[4]);

        pix2 += line_size;
    }
    return s;
}

static int pix_abs16x16_x2_c(uint8_t *pix1, uint8_t *pix2, int line_size)
{
    int s, i;

    s = 0;
    for(i=0;i<16;i++) {
        s += abs(pix1[0] - avg2(pix2[0], pix2[1]));
        s += abs(pix1[1] - avg2(pix2[1], pix2[2]));
        s += abs(pix1[2] - avg2(pix2[2], pix2[3]));
        s += abs(pix1[3] - avg2(pix2[3], pix2[4]));
        s += abs(pix1[4] - avg2(pix2[4], pix2[5]));

        pix2 += line_size;
    }
    return s;
}

static int pix_abs16x16_y2_c(uint8_t *pix1, uint8_t *pix2, int line_size)
{
    int s, i;
    uint8_t *pix3 = pix2 + line_size;

    s = 0;
    for(i=0;i<16;i++) {
        s += abs(pix1[0] - avg2(pix2[0], pix3[0]));
        s += abs(pix1[1] - avg2(pix2[1], pix3[1]));
        s += abs(pix1[2] - avg2(pix2[2], pix3[2]));
        s += abs(pix1[3] - avg2(pix2[3], pix3[3]));
        s += abs(pix1[4] - avg2(pix2[4], pix3[4]));

        pix3 += line_size;
    }
    return s;
}

static int pix_abs16x16_xy2_c(uint8_t *pix1, uint8_t *pix2, int line_size)
{
    int s, i;
    uint8_t *pix3 = pix2 + line_size;

    s = 0;
    for(i=0;i<16;i++) {
        s += abs(pix1[0] - avg4(pix2[0], pix2[1], pix3[0], pix3[1]));
        s += abs(pix1[1] - avg4(pix2[1], pix2[2], pix3[1], pix3[2]));
        s += abs(pix1[2] - avg4(pix2[2], pix2[3], pix3[2], pix3[3]));
        s += abs(pix1[3] - avg4(pix2[3], pix2[4], pix3[3], pix3[4]));
        s += abs(pix1[4] - avg4(pix2[4], pix2[5], pix3[4], pix3[5]));

        pix3 += line_size;
    }
    return s;
}

static inline int pix_abs8x8_c(uint8_t *pix1, uint8_t *pix2, int line_size)
{
    int s, i;

    s = 0;
    for(i=0;i<8;i++) {
        s += abs(pix1[0] - pix2[0]);
        s += abs(pix1[1] - pix2[1]);
        s += abs(pix1[2] - pix2[2]);
        s += abs(pix1[3] - pix2[3]);
        s += abs(pix1[4] - pix2[4]);

        pix2 += line_size;
    }
    return s;
}

static int pix_abs8x8_x2_c(uint8_t *pix1, uint8_t *pix2, int line_size)
{
    int s, i;

    s = 0;
    for(i=0;i<8;i++) {
        s += abs(pix1[0] - avg2(pix2[0], pix2[1]));
        s += abs(pix1[1] - avg2(pix2[1], pix2[2]));
        s += abs(pix1[2] - avg2(pix2[2], pix2[3]));
        s += abs(pix1[3] - avg2(pix2[3], pix2[4]));
        s += abs(pix1[4] - avg2(pix2[4], pix2[5]));

        pix2 += line_size;
    }
    return s;
}

static int pix_abs8x8_y2_c(uint8_t *pix1, uint8_t *pix2, int line_size)
{
    int s, i;
    uint8_t *pix3 = pix2 + line_size;

    s = 0;
    for(i=0;i<8;i++) {
        s += abs(pix1[0] - avg2(pix2[0], pix3[0]));
        s += abs(pix1[1] - avg2(pix2[1], pix3[1]));
        s += abs(pix1[2] - avg2(pix2[2], pix3[2]));
        s += abs(pix1[3] - avg2(pix2[3], pix3[3]));
        s += abs(pix1[4] - avg2(pix2[4], pix3[4]));

        pix3 += line_size;
    }
    return s;
}

static int pix_abs8x8_xy2_c(uint8_t *pix1, uint8_t *pix2, int line_size)
{
    int s, i;
    uint8_t *pix3 = pix2 + line_size;

    s = 0;
    for(i=0;i<8;i++) {
        s += abs(pix1[0] - avg4(pix2[0], pix2[1], pix3[0], pix3[1]));
        s += abs(pix1[1] - avg4(pix2[1], pix2[2], pix3[1], pix3[2]));
        s += abs(pix1[2] - avg4(pix2[2], pix2[3], pix3[2], pix3[3]));
        s += abs(pix1[3] - avg4(pix2[3], pix2[4], pix3[3], pix3[4]));
        s += abs(pix1[4] - avg4(pix2[4], pix2[5], pix3[4], pix3[5]));

        pix1 += line_size;
        pix2 += line_size;
        pix3 += line_size;
    }
    return s;
}

static int sad16x16_c(void *s, uint8_t *a, uint8_t *b, int stride){
    return pix_abs16x16_c(a,b,stride);
}

static int sad8x8_c(void *s, uint8_t *a, uint8_t *b, int stride){
    return pix_abs8x8_c(a,b,stride);
}

/**
 * permutes an 8x8 block.
 * @param block the block which will be permuted according to the given permutation vector

    y= a-b;\
}

#define BUTTERFLYA(x,y) (ABS((x)+(y)) + ABS((x)-(y)))

static int hadamard8_diff_c(/*MpegEncContext*/ void *s, uint8_t *dst, uint8_t *src, int stride){
    int i;
    int temp[64];
    int sum=0;

    for(i=0; i<8; i++){
        //FIXME try pointer walks
        BUTTERFLY2(temp[8*i+0], temp[8*i+1], src[stride*i+0]-dst[stride*i+0],src[stride*i+1]-dst[stride*i+1]);
        BUTTERFLY2(temp[8*i+2], temp[8*i+3], src[stride*i+2]-dst[stride*i+2],src[stride*i+3]-dst[stride*i+3]);

    }
    
    return sum;
}

static int dct_sad8x8_c(/*MpegEncContext*/ void *c, uint8_t *src1, uint8_t *src2, int stride){
    MpegEncContext * const s= (MpegEncContext *)c;
    uint64_t __align8 aligned_temp[sizeof(DCTELEM)*64/8];
    DCTELEM * const temp= (DCTELEM*)aligned_temp;
    int sum=0, i;

    s->dsp.diff_pixels(temp, src1, src2, stride);
    s->dsp.fdct(temp);

    for(i=0; i<64; i++)

    return sum;
}

void simple_idct(DCTELEM *block); //FIXME

static int quant_psnr8x8_c(/*MpegEncContext*/ void *c, uint8_t *src1, uint8_t *src2, int stride){
    MpegEncContext * const s= (MpegEncContext *)c;
    uint64_t __align8 aligned_temp[sizeof(DCTELEM)*64*2/8];
    DCTELEM * const temp= (DCTELEM*)aligned_temp;
    DCTELEM * const bak = ((DCTELEM*)aligned_temp)+64;
    int sum=0, i;

    s->mb_intra=0;
    
    s->dsp.diff_pixels(temp, src1, src2, stride);
    
    memcpy(bak, temp, 64*sizeof(DCTELEM));

        sum+= (temp[i]-bak[i])*(temp[i]-bak[i]);
        
    return sum;
}

static int rd8x8_c(/*MpegEncContext*/ void *c, uint8_t *src1, uint8_t *src2, int stride){
    MpegEncContext * const s= (MpegEncContext *)c;
    const uint8_t *scantable= s->intra_scantable.permutated;
    uint64_t __align8 aligned_temp[sizeof(DCTELEM)*64/8];
    uint64_t __align8 aligned_bak[stride];
    DCTELEM * const temp= (DCTELEM*)aligned_temp;

    int i, last, run, bits, level, distoration, start_i;
    const int esc_length= s->ac_esc_length;
    uint8_t * length;
    uint8_t * last_length;
    
    for(i=0; i<8; i++){
        ((uint32_t*)(bak + i*stride))[0]= ((uint32_t*)(src2 + i*stride))[0];
        ((uint32_t*)(bak + i*stride))[1]= ((uint32_t*)(src2 + i*stride))[1];
    }

            s->dct_unquantize_inter(s, temp, 0, s->qscale);
    }
    
    s->dsp.idct_add(bak, stride, temp);
    
    distoration= s->dsp.sse[1](NULL, bak, src1, stride);

    return distoration + ((bits*s->qscale*s->qscale*109 + 64)>>7);
}

static int bit8x8_c(/*MpegEncContext*/ void *c, uint8_t *src1, uint8_t *src2, int stride){
    MpegEncContext * const s= (MpegEncContext *)c;
    const uint8_t *scantable= s->intra_scantable.permutated;
    uint64_t __align8 aligned_temp[sizeof(DCTELEM)*64/8];
    DCTELEM * const temp= (DCTELEM*)aligned_temp;
    int i, last, run, bits, level, start_i;
    const int esc_length= s->ac_esc_length;
    uint8_t * length;
    uint8_t * last_length;
    
    s->dsp.diff_pixels(temp, src1, src2, stride);

    s->block_last_index[0/*FIXME*/]= last= s->fast_dct_quantize(s, temp, 0/*FIXME*/, s->qscale, &i);

    }

    return bits;
}


WARPER88_1616(hadamard8_diff_c, hadamard8_diff16_c)
WARPER88_1616(dct_sad8x8_c, dct_sad16x16_c)
WARPER88_1616(quant_psnr8x8_c, quant_psnr16x16_c)
WARPER88_1616(rd8x8_c, rd16x16_c)
WARPER88_1616(bit8x8_c, bit16x16_c)

/* XXX: those functions should be suppressed ASAP when all IDCTs are
 converted */
static void ff_jref_idct_put(uint8_t *dest, int line_size, DCTELEM *block)
{

    c->gmc1 = gmc1_c;
    c->gmc = gmc_c;
    c->clear_blocks = clear_blocks_c;
    c->pix_sum = pix_sum_c;
    c->pix_norm1 = pix_norm1_c;
    c->sse[0]= sse16_c;
    c->sse[1]= sse8_c;

    /* TODO [0] 16  [1] 8 */
    c->pix_abs16x16     = pix_abs16x16_c;
    c->pix_abs16x16_x2  = pix_abs16x16_x2_c;
    c->pix_abs16x16_y2  = pix_abs16x16_y2_c;
    c->pix_abs16x16_xy2 = pix_abs16x16_xy2_c;
    c->pix_abs8x8     = pix_abs8x8_c;
    c->pix_abs8x8_x2  = pix_abs8x8_x2_c;
    c->pix_abs8x8_y2  = pix_abs8x8_y2_c;
    c->pix_abs8x8_xy2 = pix_abs8x8_xy2_c;

#define dspfunc(PFX, IDX, NUM) \
    c->PFX ## _pixels_tab[IDX][0] = PFX ## _pixels ## NUM ## _c;     \
    c->PFX ## _pixels_tab[IDX][1] = PFX ## _pixels ## NUM ## _x2_c;  \
    c->PFX ## _pixels_tab[IDX][2] = PFX ## _pixels ## NUM ## _y2_c;  \

    c->put_mspel_pixels_tab[4]= put_mspel8_mc02_c;
    c->put_mspel_pixels_tab[5]= put_mspel8_mc12_c;
    c->put_mspel_pixels_tab[6]= put_mspel8_mc22_c;
    c->put_mspel_pixels_tab[7]= put_mspel8_mc32_c;
        
    c->hadamard8_diff[0]= hadamard8_diff16_c;
    c->hadamard8_diff[1]= hadamard8_diff_c;
    c->hadamard8_abs = hadamard8_abs_c;
    
    c->dct_sad[0]= dct_sad16x16_c;
    c->dct_sad[1]= dct_sad8x8_c;
    
    c->sad[0]= sad16x16_c;
    c->sad[1]= sad8x8_c;
    
    c->quant_psnr[0]= quant_psnr16x16_c;
    c->quant_psnr[1]= quant_psnr8x8_c;

    c->rd[0]= rd16x16_c;
    c->rd[1]= rd8x8_c;

    c->bit[0]= bit16x16_c;
    c->bit[1]= bit8x8_c;
        
    c->add_bytes= add_bytes_c;
    c->diff_bytes= diff_bytes_c;
    c->sub_hfyu_median_prediction= sub_hfyu_median_prediction_c;
    c->bswap_buf= bswap_buf;

1708:dea5b2946999

    for(;i<w; i++){
        dst[i+0]= bswap_32(src[i+0]);
    }
}

static int sse8_c(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int h)
{
    int s, i;
    uint32_t *sq = squareTbl + 256;

    s = 0;
    for (i = 0; i < h; i++) {
        s += sq[pix1[0] - pix2[0]];
        s += sq[pix1[1] - pix2[1]];
        s += sq[pix1[2] - pix2[2]];
        s += sq[pix1[3] - pix2[3]];
        s += sq[pix1[4] - pix2[4]];

        pix2 += line_size;
    }
    return s;
}

static int sse16_c(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h)
{
    int s, i;
    uint32_t *sq = squareTbl + 256;

    s = 0;
    for (i = 0; i < h; i++) {
        s += sq[pix1[ 0] - pix2[ 0]];
        s += sq[pix1[ 1] - pix2[ 1]];
        s += sq[pix1[ 2] - pix2[ 2]];
        s += sq[pix1[ 3] - pix2[ 3]];
        s += sq[pix1[ 4] - pix2[ 4]];

        src[y*stride-2] = p0 - d2;
        src[y*stride+1] = p3 + d2;
    }
}

static inline int pix_abs16_c(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h)
{
    int s, i;

    s = 0;
    for(i=0;i<h;i++) {
        s += abs(pix1[0] - pix2[0]);
        s += abs(pix1[1] - pix2[1]);
        s += abs(pix1[2] - pix2[2]);
        s += abs(pix1[3] - pix2[3]);
        s += abs(pix1[4] - pix2[4]);

        pix2 += line_size;
    }
    return s;
}

static int pix_abs16_x2_c(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h)
{
    int s, i;

    s = 0;
    for(i=0;i<h;i++) {
        s += abs(pix1[0] - avg2(pix2[0], pix2[1]));
        s += abs(pix1[1] - avg2(pix2[1], pix2[2]));
        s += abs(pix1[2] - avg2(pix2[2], pix2[3]));
        s += abs(pix1[3] - avg2(pix2[3], pix2[4]));
        s += abs(pix1[4] - avg2(pix2[4], pix2[5]));

        pix2 += line_size;
    }
    return s;
}

static int pix_abs16_y2_c(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h)
{
    int s, i;
    uint8_t *pix3 = pix2 + line_size;

    s = 0;
    for(i=0;i<h;i++) {
        s += abs(pix1[0] - avg2(pix2[0], pix3[0]));
        s += abs(pix1[1] - avg2(pix2[1], pix3[1]));
        s += abs(pix1[2] - avg2(pix2[2], pix3[2]));
        s += abs(pix1[3] - avg2(pix2[3], pix3[3]));
        s += abs(pix1[4] - avg2(pix2[4], pix3[4]));

        pix3 += line_size;
    }
    return s;
}

static int pix_abs16_xy2_c(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h)
{
    int s, i;
    uint8_t *pix3 = pix2 + line_size;

    s = 0;
    for(i=0;i<h;i++) {
        s += abs(pix1[0] - avg4(pix2[0], pix2[1], pix3[0], pix3[1]));
        s += abs(pix1[1] - avg4(pix2[1], pix2[2], pix3[1], pix3[2]));
        s += abs(pix1[2] - avg4(pix2[2], pix2[3], pix3[2], pix3[3]));
        s += abs(pix1[3] - avg4(pix2[3], pix2[4], pix3[3], pix3[4]));
        s += abs(pix1[4] - avg4(pix2[4], pix2[5], pix3[4], pix3[5]));

        pix3 += line_size;
    }
    return s;
}

static inline int pix_abs8_c(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h)
{
    int s, i;

    s = 0;
    for(i=0;i<h;i++) {
        s += abs(pix1[0] - pix2[0]);
        s += abs(pix1[1] - pix2[1]);
        s += abs(pix1[2] - pix2[2]);
        s += abs(pix1[3] - pix2[3]);
        s += abs(pix1[4] - pix2[4]);

        pix2 += line_size;
    }
    return s;
}

static int pix_abs8_x2_c(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h)
{
    int s, i;

    s = 0;
    for(i=0;i<h;i++) {
        s += abs(pix1[0] - avg2(pix2[0], pix2[1]));
        s += abs(pix1[1] - avg2(pix2[1], pix2[2]));
        s += abs(pix1[2] - avg2(pix2[2], pix2[3]));
        s += abs(pix1[3] - avg2(pix2[3], pix2[4]));
        s += abs(pix1[4] - avg2(pix2[4], pix2[5]));

        pix2 += line_size;
    }
    return s;
}

static int pix_abs8_y2_c(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h)
{
    int s, i;
    uint8_t *pix3 = pix2 + line_size;

    s = 0;
    for(i=0;i<h;i++) {
        s += abs(pix1[0] - avg2(pix2[0], pix3[0]));
        s += abs(pix1[1] - avg2(pix2[1], pix3[1]));
        s += abs(pix1[2] - avg2(pix2[2], pix3[2]));
        s += abs(pix1[3] - avg2(pix2[3], pix3[3]));
        s += abs(pix1[4] - avg2(pix2[4], pix3[4]));

        pix3 += line_size;
    }
    return s;
}

static int pix_abs8_xy2_c(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h)
{
    int s, i;
    uint8_t *pix3 = pix2 + line_size;

    s = 0;
    for(i=0;i<h;i++) {
        s += abs(pix1[0] - avg4(pix2[0], pix2[1], pix3[0], pix3[1]));
        s += abs(pix1[1] - avg4(pix2[1], pix2[2], pix3[1], pix3[2]));
        s += abs(pix1[2] - avg4(pix2[2], pix2[3], pix3[2], pix3[3]));
        s += abs(pix1[3] - avg4(pix2[3], pix2[4], pix3[3], pix3[4]));
        s += abs(pix1[4] - avg4(pix2[4], pix2[5], pix3[4], pix3[5]));

        pix1 += line_size;
        pix2 += line_size;
        pix3 += line_size;
    }
    return s;
}

/**
 * permutes an 8x8 block.
 * @param block the block which will be permuted according to the given permutation vector

    y= a-b;\
}

#define BUTTERFLYA(x,y) (ABS((x)+(y)) + ABS((x)-(y)))

static int hadamard8_diff8x8_c(/*MpegEncContext*/ void *s, uint8_t *dst, uint8_t *src, int stride, int h){
    int i;
    int temp[64];
    int sum=0;
    
    assert(h==8);

    for(i=0; i<8; i++){
        //FIXME try pointer walks
        BUTTERFLY2(temp[8*i+0], temp[8*i+1], src[stride*i+0]-dst[stride*i+0],src[stride*i+1]-dst[stride*i+1]);
        BUTTERFLY2(temp[8*i+2], temp[8*i+3], src[stride*i+2]-dst[stride*i+2],src[stride*i+3]-dst[stride*i+3]);

    }
    
    return sum;
}

static int dct_sad8x8_c(/*MpegEncContext*/ void *c, uint8_t *src1, uint8_t *src2, int stride, int h){
    MpegEncContext * const s= (MpegEncContext *)c;
    uint64_t __align8 aligned_temp[sizeof(DCTELEM)*64/8];
    DCTELEM * const temp= (DCTELEM*)aligned_temp;
    int sum=0, i;
    
    assert(h==8);

    s->dsp.diff_pixels(temp, src1, src2, stride);
    s->dsp.fdct(temp);

    for(i=0; i<64; i++)

    return sum;
}

void simple_idct(DCTELEM *block); //FIXME

static int quant_psnr8x8_c(/*MpegEncContext*/ void *c, uint8_t *src1, uint8_t *src2, int stride, int h){
    MpegEncContext * const s= (MpegEncContext *)c;
    uint64_t __align8 aligned_temp[sizeof(DCTELEM)*64*2/8];
    DCTELEM * const temp= (DCTELEM*)aligned_temp;
    DCTELEM * const bak = ((DCTELEM*)aligned_temp)+64;
    int sum=0, i;

    assert(h==8);
    s->mb_intra=0;
    
    s->dsp.diff_pixels(temp, src1, src2, stride);
    
    memcpy(bak, temp, 64*sizeof(DCTELEM));

        sum+= (temp[i]-bak[i])*(temp[i]-bak[i]);
        
    return sum;
}

static int rd8x8_c(/*MpegEncContext*/ void *c, uint8_t *src1, uint8_t *src2, int stride, int h){
    MpegEncContext * const s= (MpegEncContext *)c;
    const uint8_t *scantable= s->intra_scantable.permutated;
    uint64_t __align8 aligned_temp[sizeof(DCTELEM)*64/8];
    uint64_t __align8 aligned_bak[stride];
    DCTELEM * const temp= (DCTELEM*)aligned_temp;

    int i, last, run, bits, level, distoration, start_i;
    const int esc_length= s->ac_esc_length;
    uint8_t * length;
    uint8_t * last_length;
    
    assert(h==8);

    for(i=0; i<8; i++){
        ((uint32_t*)(bak + i*stride))[0]= ((uint32_t*)(src2 + i*stride))[0];
        ((uint32_t*)(bak + i*stride))[1]= ((uint32_t*)(src2 + i*stride))[1];
    }

            s->dct_unquantize_inter(s, temp, 0, s->qscale);
    }
    
    s->dsp.idct_add(bak, stride, temp);
    
    distoration= s->dsp.sse[1](NULL, bak, src1, stride, 8);

    return distoration + ((bits*s->qscale*s->qscale*109 + 64)>>7);
}

static int bit8x8_c(/*MpegEncContext*/ void *c, uint8_t *src1, uint8_t *src2, int stride, int h){
    MpegEncContext * const s= (MpegEncContext *)c;
    const uint8_t *scantable= s->intra_scantable.permutated;
    uint64_t __align8 aligned_temp[sizeof(DCTELEM)*64/8];
    DCTELEM * const temp= (DCTELEM*)aligned_temp;
    int i, last, run, bits, level, start_i;
    const int esc_length= s->ac_esc_length;
    uint8_t * length;
    uint8_t * last_length;

    assert(h==8);
    
    s->dsp.diff_pixels(temp, src1, src2, stride);

    s->block_last_index[0/*FIXME*/]= last= s->fast_dct_quantize(s, temp, 0/*FIXME*/, s->qscale, &i);

    }

    return bits;
}

WARPER8_16_SQ(hadamard8_diff8x8_c, hadamard8_diff16_c)
WARPER8_16_SQ(dct_sad8x8_c, dct_sad16_c)
WARPER8_16_SQ(quant_psnr8x8_c, quant_psnr16_c)
WARPER8_16_SQ(rd8x8_c, rd16_c)
WARPER8_16_SQ(bit8x8_c, bit16_c)

/* XXX: those functions should be suppressed ASAP when all IDCTs are
 converted */
static void ff_jref_idct_put(uint8_t *dest, int line_size, DCTELEM *block)
{

    c->gmc1 = gmc1_c;
    c->gmc = gmc_c;
    c->clear_blocks = clear_blocks_c;
    c->pix_sum = pix_sum_c;
    c->pix_norm1 = pix_norm1_c;

    /* TODO [0] 16  [1] 8 */
    c->pix_abs[0][0] = pix_abs16_c;
    c->pix_abs[0][1] = pix_abs16_x2_c;
    c->pix_abs[0][2] = pix_abs16_y2_c;
    c->pix_abs[0][3] = pix_abs16_xy2_c;
    c->pix_abs[1][0] = pix_abs8_c;
    c->pix_abs[1][1] = pix_abs8_x2_c;
    c->pix_abs[1][2] = pix_abs8_y2_c;
    c->pix_abs[1][3] = pix_abs8_xy2_c;

#define dspfunc(PFX, IDX, NUM) \
    c->PFX ## _pixels_tab[IDX][0] = PFX ## _pixels ## NUM ## _c;     \
    c->PFX ## _pixels_tab[IDX][1] = PFX ## _pixels ## NUM ## _x2_c;  \
    c->PFX ## _pixels_tab[IDX][2] = PFX ## _pixels ## NUM ## _y2_c;  \

    c->put_mspel_pixels_tab[4]= put_mspel8_mc02_c;
    c->put_mspel_pixels_tab[5]= put_mspel8_mc12_c;
    c->put_mspel_pixels_tab[6]= put_mspel8_mc22_c;
    c->put_mspel_pixels_tab[7]= put_mspel8_mc32_c;
        
    c->hadamard8_abs = hadamard8_abs_c;

#define SET_CMP_FUNC(name) \
    c->name[0]= name ## 16_c;\
    c->name[1]= name ## 8x8_c;
    
    SET_CMP_FUNC(hadamard8_diff)
    SET_CMP_FUNC(dct_sad)
    c->sad[0]= pix_abs16_c;
    c->sad[1]= pix_abs8_c;
    c->sse[0]= sse16_c;
    c->sse[1]= sse8_c;
    SET_CMP_FUNC(quant_psnr)
    SET_CMP_FUNC(rd)
    SET_CMP_FUNC(bit)
        
    c->add_bytes= add_bytes_c;
    c->diff_bytes= diff_bytes_c;
    c->sub_hfyu_median_prediction= sub_hfyu_median_prediction_c;
    c->bswap_buf= bswap_buf;