comparison vc1.c @ 3367:8c7b8ffc2485 libavcodec

Some optimization and fixes - mostly reworked MC and bitplanes.

3366:c59aa4cdf042

    CS_MID_RATE_INTRA,
    CS_MID_RATE_INTER,
    CS_HIGH_RATE_INTRA,
    CS_HIGH_RATE_INTER
};

/** Bitplane struct
 * We mainly need data and is_raw, so this struct could be avoided
 * to save a level of indirection; feel free to modify
 * @fixme For now, stride=width
 * @warning Data are bits, either 1 or 0
 */
typedef struct BitPlane {
    uint8_t *data;      ///< Data buffer
    int width;          ///< Width of the buffer
    int stride;         ///< Stride of the buffer
    int height;         ///< Plane height
    uint8_t is_raw;     ///< Bit values must be read at MB level
} BitPlane;


/** Block data for DC/AC prediction
*/
typedef struct Block {
    uint16_t dc;
    int16_t hor_ac[7];
    int16_t vert_ac[7];
    int16_t dcstep, step;
} Block;

/** The VC1 Context
 * @fixme Change size wherever another size is more efficient
 * Many members are only used for Advanced Profile
 */

    uint8_t mvrange;
    uint8_t pquantizer;           ///< Uniform (over sequence) quantizer in use
    uint8_t *previous_line_cbpcy; ///< To use for predicted CBPCY
    VLC *cbpcy_vlc;               ///< CBPCY VLC table
    int tt_index;                 ///< Index for Transform Type tables
    BitPlane mv_type_mb_plane;    ///< bitplane for mv_type == (4MV)
    BitPlane skip_mb_plane;       ///< bitplane for skipped MBs
    BitPlane direct_mb_plane;     ///< bitplane for "direct" MBs

    /** Frame decoding info for S/M profiles only */
    //@{
    uint8_t rangeredfrm; ///< out_sample = CLIP((in_sample-128)*2+128)
    uint8_t interpfrm;

    uint8_t uvsamp;
    uint8_t postproc;
    int hrd_num_leaky_buckets;
    uint8_t bit_rate_exponent;
    uint8_t buffer_size_exponent;
    BitPlane ac_pred_plane;       ///< AC prediction flags bitplane
    BitPlane over_flags_plane;    ///< Overflags bitplane
    uint8_t condover;
    uint16_t *hrd_rate, *hrd_buffer;
    uint8_t *hrd_fullness;
    uint8_t range_mapy_flag;
    uint8_t range_mapuv_flag;

static int vc1_init_common(VC1Context *v)
{
    static int done = 0;
    int i = 0;

    /* Set the bit planes */
    v->mv_type_mb_plane = (struct BitPlane) { NULL, 0, 0, 0 };
    v->direct_mb_plane = (struct BitPlane) { NULL, 0, 0, 0 };
    v->skip_mb_plane = (struct BitPlane) { NULL, 0, 0, 0 };
    v->ac_pred_plane = v->over_flags_plane = (struct BitPlane) { NULL, 0, 0, 0 };
    v->hrd_rate = v->hrd_buffer = NULL;

    /* VLC tables */
    if(!done)
    {

    IMODE_ROWSKIP,
    IMODE_COLSKIP
};
/** @} */ //imode defines

/** Allocate the buffer from a bitplane, given its dimensions
 * @param bp Bitplane which buffer is to allocate
 * @param[in] width Width of the buffer
 * @param[in] height Height of the buffer
 * @return Status
 * @todo TODO: Take into account stride
 * @todo TODO: Allow use of external buffers ?
 */
static int alloc_bitplane(BitPlane *bp, int width, int height)
{
    if (!bp || bp->width<0 || bp->height<0) return -1;
    bp->data = (uint8_t*)av_malloc(width*height);
    if (!bp->data) return -1;
    bp->width = bp->stride = width;
    bp->height = height;
    return 0;
}

/** Free the bitplane's buffer
 * @param bp Bitplane which buffer is to free
 */
static void free_bitplane(BitPlane *bp)
{
    bp->width = bp->stride = bp->height = 0;
    if (bp->data) av_freep(&bp->data);
}

/** Decode rows by checking if they are skipped
 * @param plane Buffer to store decoded bits
 * @param[in] width Width of this buffer
 * @param[in] height Height of this buffer
 * @param[in] stride of this buffer

 * @param v VC-1 context for bit reading and logging
 * @return Status
 * @fixme FIXME: Optimize
 * @todo TODO: Decide if a struct is needed
 */
static int bitplane_decoding(BitPlane *bp, VC1Context *v)
{
    GetBitContext *gb = &v->s.gb;

    int imode, x, y, code, offset;
    uint8_t invert, *planep = bp->data;

    invert = get_bits(gb, 1);
    imode = get_vlc2(gb, vc1_imode_vlc.table, VC1_IMODE_VLC_BITS, 1);

    bp->is_raw = 0;
    switch (imode)
    {
    case IMODE_RAW:
        //Data is actually read in the MB layer (same for all tests == "raw")
        bp->is_raw = 1; //invert ignored
        return invert;
    case IMODE_DIFF2:
    case IMODE_NORM2:
        if ((bp->height * bp->width) & 1)
        {
            *planep++ = get_bits(gb, 1);
            offset = 1;
        }
        else offset = 0;
        // decode bitplane as one long line
        for (y = offset; y < bp->height * bp->width; y += 2) {
            code = get_vlc2(gb, vc1_norm2_vlc.table, VC1_NORM2_VLC_BITS, 1);
            *planep++ = code & 1;
            offset++;
            if(offset == bp->width) {
                offset = 0;
                planep += bp->stride - bp->width;
            }
            *planep++ = code >> 1;
            offset++;
            if(offset == bp->width) {
                offset = 0;
                planep += bp->stride - bp->width;
            }
        }
        break;
    case IMODE_DIFF6:
    case IMODE_NORM6:
        if(!(bp->height % 3) && (bp->width % 3)) { // use 2x3 decoding
            for(y = 0; y < bp->height; y+= 3) {
                for(x = bp->width & 1; x < bp->width; x += 2) {
                    code = get_vlc2(gb, vc1_norm6_vlc.table, VC1_NORM6_VLC_BITS, 2);
                    if(code < 0){
                        av_log(v->s.avctx, AV_LOG_DEBUG, "invalid NORM-6 VLC\n");
                        return -1;
                    }
                    planep[x + 0] = (code >> 0) & 1;
                    planep[x + 1] = (code >> 1) & 1;
                    planep[x + 0 + bp->stride] = (code >> 2) & 1;
                    planep[x + 1 + bp->stride] = (code >> 3) & 1;
                    planep[x + 0 + bp->stride * 2] = (code >> 4) & 1;
                    planep[x + 1 + bp->stride * 2] = (code >> 5) & 1;
                }
                planep += bp->stride * 3;
            }
            if(bp->width & 1) decode_colskip(bp->data, 1, bp->height, bp->stride, &v->s.gb);
        } else { // 3x2
            for(y = bp->height & 1; y < bp->height; y += 2) {
                for(x = bp->width % 3; x < bp->width; x += 3) {
                    code = get_vlc2(gb, vc1_norm6_vlc.table, VC1_NORM6_VLC_BITS, 2);
                    if(code < 0){
                        av_log(v->s.avctx, AV_LOG_DEBUG, "invalid NORM-6 VLC\n");
                        return -1;
                    }
                    planep[x + 0] = (code >> 0) & 1;
                    planep[x + 1] = (code >> 1) & 1;
                    planep[x + 2] = (code >> 2) & 1;
                    planep[x + 0 + bp->stride] = (code >> 3) & 1;
                    planep[x + 1 + bp->stride] = (code >> 4) & 1;
                    planep[x + 2 + bp->stride] = (code >> 5) & 1;
                }
                planep += bp->stride * 2;
            }
            x = bp->width % 3;
            if(x) decode_colskip(bp->data  ,             x, bp->height    , bp->stride, &v->s.gb);
            if(bp->height & 1) decode_rowskip(bp->data+x, bp->width - x, bp->height & 1, bp->stride, &v->s.gb);
        }
        break;
    case IMODE_ROWSKIP:
        decode_rowskip(bp->data, bp->width, bp->height, bp->stride, &v->s.gb);
        break;
    case IMODE_COLSKIP:
        decode_colskip(bp->data, bp->width, bp->height, bp->stride, &v->s.gb);
        break;
    default: break;
    }

    /* Applying diff operator */
    if (imode == IMODE_DIFF2 || imode == IMODE_DIFF6)
    {
        planep = bp->data;
        planep[0] ^= invert;
        for (x=1; x<bp->width; x++)
            planep[x] ^= planep[x-1];
        for (y=1; y<bp->height; y++)
        {
            planep += bp->stride;
            planep[0] ^= planep[-bp->stride];
            for (x=1; x<bp->width; x++)
            {
                if (planep[x-1] != planep[x-bp->stride]) planep[x] ^= invert;
                else                                     planep[x] ^= planep[x-1];
            }
        }
    }
    else if (invert)
    {
        planep = bp->data;
        for (x=0; x<bp->width*bp->height; x++) planep[x] = !planep[x]; //FIXME stride
    }
    return (imode<<1) + invert;
}
/** @} */ //Bitplane group

/***********************************************************************/
/** VOP Dquant decoding
 * @param v VC-1 Context

    for(i = 0; i < 64; i++)
        block[i] += 128;
}


/** Put block onto picture
 * @todo move to DSPContext
 */
static void vc1_put_block(VC1Context *v, DCTELEM block[6][64])
{

    dsp->put_pixels_clamped(block[4], v->s.dest[1], us);
    dsp->put_pixels_clamped(block[5], v->s.dest[2], vs);
}

/** Do motion compensation over 1 macroblock
 * Mostly adapted hpel_motion and qpel_motion from mpegvideo.c
 */
static void vc1_mc_1mv(VC1Context *v)
{
    MpegEncContext *s = &v->s;
    DSPContext *dsp = &v->s.dsp;
    uint8_t *srcY, *srcU, *srcV;
    int dxy, mx, my, src_x, src_y;
    int width = s->mb_width * 16, height = s->mb_height * 16;

    if(!v->s.last_picture.data[0])return;

    mx = s->mv[0][0][0] >> s->mspel;
    my = s->mv[0][0][1] >> s->mspel;
    srcY = s->last_picture.data[0];
    srcU = s->last_picture.data[1];
    srcV = s->last_picture.data[2];

    if(s->mspel) { // hpel mc
        dxy = ((my & 1) << 1) | (mx & 1);
        src_x = s->mb_x * 16 + (mx >> 1);
        src_y = s->mb_y * 16 + (my >> 1);
/*        src_x = clip(src_x, -16, width); //FIXME unneeded for emu?
        if (src_x == width)
            dxy &= ~1;
        src_y = clip(src_y, -16, height);
        if (src_y == height)
            dxy &= ~2;*/
        srcY += src_y * s->linesize + src_x;
        srcU += (src_y >> 1) * s->uvlinesize + (src_x >> 1);
        srcV += (src_y >> 1) * s->uvlinesize + (src_x >> 1);

        if((unsigned)src_x > s->h_edge_pos - (mx&1) - 16
           || (unsigned)src_y > s->v_edge_pos - (my&1) - 16){
            uint8_t *uvbuf= s->edge_emu_buffer + 18 * s->linesize;

            ff_emulated_edge_mc(s->edge_emu_buffer, srcY, s->linesize, 16+1, 16+1,
                             src_x, src_y, s->h_edge_pos, s->v_edge_pos);
            srcY = s->edge_emu_buffer;
            ff_emulated_edge_mc(uvbuf, srcU, s->uvlinesize, 8+1, 8+1,
                             src_x >> 1, src_y >> 1, s->h_edge_pos >> 1, s->v_edge_pos >> 1);
            ff_emulated_edge_mc(uvbuf + 16, srcV, s->uvlinesize, 8+1, 8+1,
                             src_x >> 1, src_y >> 1, s->h_edge_pos >> 1, s->v_edge_pos >> 1);
            srcU = uvbuf;
            srcV = uvbuf + 16;
        }
        dsp->put_no_rnd_pixels_tab[0][dxy](s->dest[0], srcY, s->linesize, 16);
        dsp->put_no_rnd_pixels_tab[1][0](s->dest[1], srcU, s->uvlinesize, 8);
        dsp->put_no_rnd_pixels_tab[1][0](s->dest[2], srcV, s->uvlinesize, 8);
    } else {
        int motion_x = mx, motion_y = my, uvdxy, uvsrc_x, uvsrc_y;
        dxy = ((motion_y & 3) << 2) | (motion_x & 3);
        src_x = s->mb_x * 16 + (mx >> 2);
        src_y = s->mb_y * 16 + (my >> 2);

        mx= motion_x/2;
        my= motion_y/2;

        mx= (mx>>1)|(mx&1);
        my= (my>>1)|(my&1);

        uvdxy= (mx&1) | ((my&1)<<1);
        mx>>=1;
        my>>=1;

        uvsrc_x = s->mb_x * 8 + mx;
        uvsrc_y = s->mb_y * 8 + my;

        srcY = s->last_picture.data[0] +   src_y *   s->linesize +   src_x;
        srcU = s->last_picture.data[1] + uvsrc_y * s->uvlinesize + uvsrc_x;
        srcV = s->last_picture.data[2] + uvsrc_y * s->uvlinesize + uvsrc_x;

        if(   (unsigned)src_x > s->h_edge_pos - (motion_x&3) - 16
              || (unsigned)src_y > s->v_edge_pos - (motion_y&3) - 16  ){
            uint8_t *uvbuf= s->edge_emu_buffer + 18*s->linesize;
            ff_emulated_edge_mc(s->edge_emu_buffer, srcY, s->linesize, 17, 17,
                                src_x, src_y, s->h_edge_pos, s->v_edge_pos);
            srcY = s->edge_emu_buffer;
            ff_emulated_edge_mc(uvbuf, srcU, s->uvlinesize, 9, 9,
                                    uvsrc_x, uvsrc_y, s->h_edge_pos>>1, s->v_edge_pos>>1);
            ff_emulated_edge_mc(uvbuf + 16, srcV, s->uvlinesize, 9, 9,
                                    uvsrc_x, uvsrc_y, s->h_edge_pos>>1, s->v_edge_pos>>1);
            srcU = uvbuf;
            srcV = uvbuf + 16;
        }

        dsp->put_no_rnd_qpel_pixels_tab[0][dxy](s->dest[0], srcY, s->linesize);
        dsp->put_no_rnd_pixels_tab[1][uvdxy](s->dest[1], srcU, s->uvlinesize, 8);
        dsp->put_no_rnd_pixels_tab[1][uvdxy](s->dest[2], srcV, s->uvlinesize, 8);
    }
}

/**
 * Decode Simple/Main Profiles sequence header
 * @see Figure 7-8, p16-17

            v->mv_mode2 = mv_pmode_table[lowquant][get_prefix(gb, 1, 3)];
            v->lumscale = get_bits(gb, 6);
            v->lumshift = get_bits(gb, 6);
        }
        if(v->mv_mode == MV_PMODE_1MV_HPEL || v->mv_mode == MV_PMODE_1MV_HPEL_BILIN)
            v->s.mspel = 1;
        else
            v->s.mspel = 0;

if(v->mv_mode != MV_PMODE_1MV && v->mv_mode != MV_PMODE_1MV_HPEL && v->mv_mode != MV_PMODE_1MV_HPEL_BILIN) {
    av_log(v->s.avctx, AV_LOG_ERROR, "Only 1MV P-frames are supported by now\n");
    return -1;
}
        if ((v->mv_mode == MV_PMODE_INTENSITY_COMP &&
                 v->mv_mode2 == MV_PMODE_MIXED_MV)
                || v->mv_mode == MV_PMODE_MIXED_MV)
        {
            status = bitplane_decoding(&v->mv_type_mb_plane, v);
            if (status < 0) return -1;
            av_log(v->s.avctx, AV_LOG_DEBUG, "MB MV Type plane encoding: "
                   "Imode: %i, Invert: %i\n", status>>1, status&1);
        }
        status = bitplane_decoding(&v->skip_mb_plane, v);
        if (status < 0) return -1;
        av_log(v->s.avctx, AV_LOG_DEBUG, "MB Skip plane encoding: "
               "Imode: %i, Invert: %i\n", status>>1, status&1);

        /* Hopefully this is correct for P frames */

  else mb_has_coeffs = 0;                                           \
  s->mb_intra = 0;                                                  \
  if (!index) { _dmv_x = _dmv_y = 0; }                              \
  else if (index == 35)                                             \
  {                                                                 \
    _dmv_x = get_bits(gb, v->k_x - s->mspel);                       \
    _dmv_y = get_bits(gb, v->k_y - s->mspel);                       \
  }                                                                 \
  else if (index == 36)                                             \
  {                                                                 \
    _dmv_x = 0;                                                     \
    _dmv_y = 0;                                                     \
    s->mb_intra = 1;                                                \
  }                                                                 \
  else                                                              \
  {                                                                 \
    index1 = index%6;                                               \
    if (s->mspel && index1 == 5) val = 1;                           \
    else                         val = 0;                           \
    if(size_table[index1] - val > 0)                                \
        val = get_bits(gb, size_table[index1] - val);               \
    else                                   val = 0;                 \
    sign = 0 - (val&1);                                             \
    _dmv_x = (sign ^ ((val>>1) + offset_table[index1])) - sign;     \
                                                                    \
    index1 = index/6;                                               \
    if (s->mspel && index1 == 5) val = 1;                           \
    else                         val = 0;                           \
    if(size_table[index1] - val > 0)                                \
        val = get_bits(gb, size_table[index1] - val);               \
    else                                   val = 0;                 \
    sign = 0 - (val&1);                                             \
    _dmv_y = (sign ^ ((val>>1) + offset_table[index1])) - sign;     \

    int px, py;
    int sum;
    int mb_pos = s->mb_x + s->mb_y * s->mb_stride;

    /* scale MV difference to be quad-pel */
    dmv_x <<= s->mspel;
    dmv_y <<= s->mspel;

    wrap = s->b8_stride;
    xy = s->block_index[0];

    C = s->current_picture.motion_val[0][xy - (1 << mv1)];

        lst = index >= vc1_last_decode_table[codingset];
        if(get_bits(gb, 1))
            level = -level;
    } else {
        escape = decode210(gb);
        if (escape == 0) {
            index = get_vlc2(gb, vc1_ac_coeff_table[codingset].table, AC_VLC_BITS, 3);
            run = vc1_index_decode_table[codingset][index][0];
            level = vc1_index_decode_table[codingset][index][1];
            lst = index >= vc1_last_decode_table[codingset];
            if(lst)
                level += vc1_last_delta_level_table[codingset][run];
            else
                level += vc1_delta_level_table[codingset][run];
            if(get_bits(gb, 1))
                level = -level;
        } else if (escape == 1) {
            index = get_vlc2(gb, vc1_ac_coeff_table[codingset].table, AC_VLC_BITS, 3);
            run = vc1_index_decode_table[codingset][index][0];
            level = vc1_index_decode_table[codingset][index][1];
            lst = index >= vc1_last_decode_table[codingset];
            if(lst)
                run += vc1_last_delta_run_table[codingset][level] + 1;
            else
                run += vc1_delta_run_table[codingset][level] + 1;
            if(get_bits(gb, 1))
                level = -level;
        } else {
            int sign;
            lst = get_bits(gb, 1);

                break;
            block[zz_table[i++]] = value;
        }

        /* apply AC prediction if needed */
        if(s->ac_pred) {
            /* scale predictors if needed*/
            int mb_pos2, q1, q2;

            mb_pos2 = mb_pos - dc_pred_dir - (1 - dc_pred_dir) * s->mb_stride;
            q1 = s->current_picture.qscale_table[mb_pos];

            }
            if(!a_avail) {
                memset(ac_val + 8, 0, 8 * sizeof(ac_val[0]));
                dc_pred_dir = 1;
            }
            if(!q1 && q1 && q2 && q1 != q2) {
                q1 = q1 * 2 - 1;
                q2 = q2 * 2 - 1;

                if(dc_pred_dir) { //left
                    for(k = 1; k < 8; k++)

        scale = mquant * 2;
        memset(ac_val2, 0, 16 * 2);
        if(dc_pred_dir) {//left
            ac_val -= 16;
            if(s->ac_pred)
                memcpy(ac_val2, ac_val, 8 * 2);
        } else {//top
            ac_val -= 16 * s->block_wrap[n];
            if(s->ac_pred)
                memcpy(ac_val2 + 8, ac_val + 8, 8 * 2);
        }

        /* apply AC prediction if needed */
        if(s->ac_pred) {
            if(dc_pred_dir) { //left
                for(k = 1; k < 8; k++) {
                    block[k << 3] = ac_val[k] * scale;
                    if(!v->pquantizer)
                        block[k << 3] += (block[k << 3] < 0) ? -mquant : mquant;

    }
    if(ttblk == TT_4X8_RIGHT || ttblk == TT_4X8_LEFT) {
        ttblk = TT_4X8;
        subblkpat = 2 - (ttblk == TT_4X8_LEFT);
    }

    switch(ttblk) {
    case TT_8X8:
        i = 0;
        last = 0;
        while (!last) {

        break;
    case TT_4X4:
        for(j = 0; j < 4; j++) {
            last = subblkpat & (1 << (3 - j));
            i = 0;
            off = (j & 1) * 4 + (j & 2) * 32;
            while (!last) {
                vc1_decode_ac_coeff(v, &last, &skip, &value, v->codingset2);
                i += skip;
                if(i > 15)
                    break;
                idx = vc1_simple_progressive_4x4_zz[i++];
                block[idx + off] = value * scale;
            }
            vc1_inv_trans(block + off, 4, 4);
        }
        break;
    case TT_8X4:
        for(j = 0; j < 2; j++) {
            last = subblkpat & (1 << (1 - j));

                if(i > 31)
                    break;
                idx = vc1_simple_progressive_8x4_zz[i++];
                block[idx + off] = value * scale;
            }
            if(!(subblkpat & (1 << (1 - j)))) vc1_inv_trans(block + off, 8, 4);
        }
        break;
    case TT_4X8:
        for(j = 0; j < 2; j++) {
            last = subblkpat & (1 << (1 - j));

                if(i > 31)
                    break;
                idx = vc1_simple_progressive_8x4_zz[i++];
                block[idx + off] = value * scale;
            }
            vc1_inv_trans(block + off, 4, 8);
        }
        break;
    }
    return 0;
}

 */
static int vc1_decode_p_mb(VC1Context *v, DCTELEM block[6][64])
{
    MpegEncContext *s = &v->s;
    GetBitContext *gb = &s->gb;
    int i, j, mb_offset = s->mb_x + s->mb_y*s->mb_width; /* XXX: mb_stride */
    int mb_pos = s->mb_x + s->mb_y * s->mb_stride;
    int cbp; /* cbp decoding stuff */
    int hybrid_pred; /* Prediction types */
    int mqdiff, mquant; /* MB quantization */
    int ttmb = v->ttmb; /* MB Transform type */

    int dmv_x, dmv_y; /* Differential MV components */
    int index, index1; /* LUT indices */
    int val, sign; /* temp values */
    int first_block = 1;
    int dst_idx, off;

    mquant = v->pq; /* Loosy initialization */

    if (v->mv_type_mb_plane.is_raw)
        v->mv_type_mb_plane.data[mb_offset] = get_bits(gb, 1);
    if (v->skip_mb_plane.is_raw)
        v->skip_mb_plane.data[mb_offset] = get_bits(gb, 1);
    s->current_picture.mbskip_table[mb_pos] = v->skip_mb_plane.data[mb_offset];
    if (!v->mv_type_mb_plane.data[mb_offset]) /* 1MV mode */
    {
        if (!v->skip_mb_plane.data[mb_offset])
        {
            GET_MVDATA(dmv_x, dmv_y);

            s->current_picture.mb_type[mb_pos] = s->mb_intra ? MB_TYPE_INTRA : MB_TYPE_16x16;
            vc1_pred_mv(s, dmv_x, dmv_y, 1, v->range_x, v->range_y);

                        v->a_avail = 1;
                    if((i == 1 || i == 3) || (s->mb_x && IS_INTRA(s->current_picture.mb_type[mb_pos - 1])))
                        v->c_avail = 1;

                    vc1_decode_intra_block(v, block[i], i, val, mquant, (i&4)?v->codingset2:v->codingset);
                    vc1_inv_trans(s->block[i], 8, 8);
                    for(j = 0; j < 64; j++) s->block[i][j] += 128;
                    s->dsp.put_pixels_clamped(s->block[i], s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2));
                    /* TODO: proper loop filtering */
                    if(v->a_avail)
                        s->dsp.h263_v_loop_filter(s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2), s->y_dc_scale);
                    if(v->c_avail)
                        s->dsp.h263_h_loop_filter(s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2), s->y_dc_scale);
                } else if(val) {
                    vc1_decode_p_block(v, block[i], i, mquant, ttmb, first_block);
                    if(!v->ttmbf && ttmb < 8) ttmb = -1;
                    first_block = 0;
                    s->dsp.add_pixels_clamped(s->block[i], s->dest[dst_idx] + off, (i&4)?s->uvlinesize:s->linesize);
                }
            }
        }
        else //Skipped
        {

            return 0;
        }
    } //1MV mode
    else //4MV mode
    {//FIXME: looks not conforming to standard and is not even theoretically complete
        if (!v->skip_mb_plane.data[mb_offset] /* unskipped MB */)
        {
            int blk_intra[4], blk_coded[4];
            /* Get CBPCY */
            cbp = get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);
            for (i=0; i<4; i++)

    GetBitContext gb;

    if (!avctx->extradata_size || !avctx->extradata) return -1;
    avctx->pix_fmt = PIX_FMT_YUV420P;
    v->s.avctx = avctx;

    if(ff_h263_decode_init(avctx) < 0)
        return -1;
    if (vc1_init_common(v) < 0) return -1;

    av_log(avctx, AV_LOG_INFO, "This decoder is not supposed to produce picture. Dont report this as a bug!\n");
    av_log(avctx, AV_LOG_INFO, "If you see a picture, don't believe your eyes.\n");

    avctx->flags |= CODEC_FLAG_EMU_EDGE;
    avctx->coded_width = avctx->width;
    avctx->coded_height = avctx->height;
    if (avctx->codec_id == CODEC_ID_WMV3)
    {
        int count = 0;

    s->mb_width = (avctx->coded_width+15)>>4;
    s->mb_height = (avctx->coded_height+15)>>4;

    /* Allocate mb bitplanes */
    if (alloc_bitplane(&v->mv_type_mb_plane, s->mb_width, s->mb_height) < 0)
        return -1;
    if (alloc_bitplane(&v->mv_type_mb_plane, s->mb_width, s->mb_height) < 0)
        return -1;
    if (alloc_bitplane(&v->skip_mb_plane, s->mb_width, s->mb_height) < 0)
        return -1;
    if (alloc_bitplane(&v->direct_mb_plane, s->mb_width, s->mb_height) < 0)
        return -1;

    /* For predictors */
    v->previous_line_cbpcy = (uint8_t *)av_malloc(s->mb_stride*4);
    if (!v->previous_line_cbpcy) return -1;

    /* Init coded blocks info */
    if (v->profile == PROFILE_ADVANCED)
    {
        if (alloc_bitplane(&v->over_flags_plane, s->mb_width, s->mb_height) < 0)
            return -1;
        if (alloc_bitplane(&v->ac_pred_plane, s->mb_width, s->mb_height) < 0)
            return -1;
    }

    return 0;
}

    VC1Context *v = avctx->priv_data;

    av_freep(&v->hrd_rate);
    av_freep(&v->hrd_buffer);
    MPV_common_end(&v->s);
    free_bitplane(&v->mv_type_mb_plane);
    free_bitplane(&v->skip_mb_plane);
    free_bitplane(&v->direct_mb_plane);
    return 0;
}


AVCodec vc1_decoder = {

3367:8c7b8ffc2485

    CS_MID_RATE_INTRA,
    CS_MID_RATE_INTER,
    CS_HIGH_RATE_INTRA,
    CS_HIGH_RATE_INTER
};

/** The VC1 Context
 * @fixme Change size wherever another size is more efficient
 * Many members are only used for Advanced Profile
 */

    uint8_t mvrange;
    uint8_t pquantizer;           ///< Uniform (over sequence) quantizer in use
    uint8_t *previous_line_cbpcy; ///< To use for predicted CBPCY
    VLC *cbpcy_vlc;               ///< CBPCY VLC table
    int tt_index;                 ///< Index for Transform Type tables
    uint8_t* mv_type_mb_plane;    ///< bitplane for mv_type == (4MV)
    uint8_t* skip_mb_plane;       ///< bitplane for skipped MBs
//    BitPlane direct_mb_plane;     ///< bitplane for "direct" MBs
    int mv_type_is_raw;           ///< mv type mb plane is not coded
    int skip_is_raw;              ///< skip mb plane is not coded

    /** Frame decoding info for S/M profiles only */
    //@{
    uint8_t rangeredfrm; ///< out_sample = CLIP((in_sample-128)*2+128)
    uint8_t interpfrm;

    uint8_t uvsamp;
    uint8_t postproc;
    int hrd_num_leaky_buckets;
    uint8_t bit_rate_exponent;
    uint8_t buffer_size_exponent;
//    BitPlane ac_pred_plane;       ///< AC prediction flags bitplane
//    BitPlane over_flags_plane;    ///< Overflags bitplane
    uint8_t condover;
    uint16_t *hrd_rate, *hrd_buffer;
    uint8_t *hrd_fullness;
    uint8_t range_mapy_flag;
    uint8_t range_mapuv_flag;

static int vc1_init_common(VC1Context *v)
{
    static int done = 0;
    int i = 0;

    v->hrd_rate = v->hrd_buffer = NULL;

    /* VLC tables */
    if(!done)
    {

    IMODE_ROWSKIP,
    IMODE_COLSKIP
};
/** @} */ //imode defines

/** Decode rows by checking if they are skipped
 * @param plane Buffer to store decoded bits
 * @param[in] width Width of this buffer
 * @param[in] height Height of this buffer
 * @param[in] stride of this buffer

 * @param v VC-1 context for bit reading and logging
 * @return Status
 * @fixme FIXME: Optimize
 * @todo TODO: Decide if a struct is needed
 */
static int bitplane_decoding(uint8_t* data, int *raw_flag, VC1Context *v)
{
    GetBitContext *gb = &v->s.gb;

    int imode, x, y, code, offset;
    uint8_t invert, *planep = data;
    int width, height, stride;

    width = v->s.mb_width;
    height = v->s.mb_height;
    stride = v->s.mb_stride;
    invert = get_bits(gb, 1);
    imode = get_vlc2(gb, vc1_imode_vlc.table, VC1_IMODE_VLC_BITS, 1);

    *raw_flag = 0;
    switch (imode)
    {
    case IMODE_RAW:
        //Data is actually read in the MB layer (same for all tests == "raw")
        *raw_flag = 1; //invert ignored
        return invert;
    case IMODE_DIFF2:
    case IMODE_NORM2:
        if ((height * width) & 1)
        {
            *planep++ = get_bits(gb, 1);
            offset = 1;
        }
        else offset = 0;
        // decode bitplane as one long line
        for (y = offset; y < height * width; y += 2) {
            code = get_vlc2(gb, vc1_norm2_vlc.table, VC1_NORM2_VLC_BITS, 1);
            *planep++ = code & 1;
            offset++;
            if(offset == width) {
                offset = 0;
                planep += stride - width;
            }
            *planep++ = code >> 1;
            offset++;
            if(offset == width) {
                offset = 0;
                planep += stride - width;
            }
        }
        break;
    case IMODE_DIFF6:
    case IMODE_NORM6:
        if(!(height % 3) && (width % 3)) { // use 2x3 decoding
            for(y = 0; y < height; y+= 3) {
                for(x = width & 1; x < width; x += 2) {
                    code = get_vlc2(gb, vc1_norm6_vlc.table, VC1_NORM6_VLC_BITS, 2);
                    if(code < 0){
                        av_log(v->s.avctx, AV_LOG_DEBUG, "invalid NORM-6 VLC\n");
                        return -1;
                    }
                    planep[x + 0] = (code >> 0) & 1;
                    planep[x + 1] = (code >> 1) & 1;
                    planep[x + 0 + stride] = (code >> 2) & 1;
                    planep[x + 1 + stride] = (code >> 3) & 1;
                    planep[x + 0 + stride * 2] = (code >> 4) & 1;
                    planep[x + 1 + stride * 2] = (code >> 5) & 1;
                }
                planep += stride * 3;
            }
            if(width & 1) decode_colskip(data, 1, height, stride, &v->s.gb);
        } else { // 3x2
            for(y = height & 1; y < height; y += 2) {
                for(x = width % 3; x < width; x += 3) {
                    code = get_vlc2(gb, vc1_norm6_vlc.table, VC1_NORM6_VLC_BITS, 2);
                    if(code < 0){
                        av_log(v->s.avctx, AV_LOG_DEBUG, "invalid NORM-6 VLC\n");
                        return -1;
                    }
                    planep[x + 0] = (code >> 0) & 1;
                    planep[x + 1] = (code >> 1) & 1;
                    planep[x + 2] = (code >> 2) & 1;
                    planep[x + 0 + stride] = (code >> 3) & 1;
                    planep[x + 1 + stride] = (code >> 4) & 1;
                    planep[x + 2 + stride] = (code >> 5) & 1;
                }
                planep += stride * 2;
            }
            x = width % 3;
            if(x) decode_colskip(data  ,             x, height    , stride, &v->s.gb);
            if(height & 1) decode_rowskip(data+x, width - x, 1, stride, &v->s.gb);
        }
        break;
    case IMODE_ROWSKIP:
        decode_rowskip(data, width, height, stride, &v->s.gb);
        break;
    case IMODE_COLSKIP:
        decode_colskip(data, width, height, stride, &v->s.gb);
        break;
    default: break;
    }

    /* Applying diff operator */
    if (imode == IMODE_DIFF2 || imode == IMODE_DIFF6)
    {
        planep = data;
        planep[0] ^= invert;
        for (x=1; x<width; x++)
            planep[x] ^= planep[x-1];
        for (y=1; y<height; y++)
        {
            planep += stride;
            planep[0] ^= planep[-stride];
            for (x=1; x<width; x++)
            {
                if (planep[x-1] != planep[x-stride]) planep[x] ^= invert;
                else                                 planep[x] ^= planep[x-1];
            }
        }
    }
    else if (invert)
    {
        planep = data;
        for (x=0; x<stride*height; x++) planep[x] = !planep[x]; //FIXME stride
    }
    return (imode<<1) + invert;
}

/** @} */ //Bitplane group

/***********************************************************************/
/** VOP Dquant decoding
 * @param v VC-1 Context

    for(i = 0; i < 64; i++)
        block[i] += 128;
}


static void vc1_v_overlap(uint8_t* src, int stride)
{
    int i;
    int a, b, c, d;
    for(i = 0; i < 8; i++) {
        a = src[-2*stride];
        b = src[-stride];
        c = src[0];
        d = src[stride];

        src[-2*stride] = (7*a + d) >> 3;
        src[-stride] = (-a + 7*b + c + d) >> 3;
        src[0] = (a + b + 7*c - d) >> 3;
        src[stride] = (a + 7*d) >> 3;
        src++;
    }
}

static void vc1_h_overlap(uint8_t* src, int stride)
{
    int i;
    int a, b, c, d;
    for(i = 0; i < 8; i++) {
        a = src[-2];
        b = src[-1];
        c = src[0];
        d = src[1];

        src[-2] = (7*a + d) >> 3;
        src[-1] = (-a + 7*b + c + d) >> 3;
        src[0] = (a + b + 7*c - d) >> 3;
        src[1] = (a + 7*d) >> 3;
        src += stride;
    }
}

/** Put block onto picture
 * @todo move to DSPContext
 */
static void vc1_put_block(VC1Context *v, DCTELEM block[6][64])
{

    dsp->put_pixels_clamped(block[4], v->s.dest[1], us);
    dsp->put_pixels_clamped(block[5], v->s.dest[2], vs);
}

/* clip motion vector as specified in 8.3.6.5 */
#define CLIP_RANGE(mv, src, lim, bs)      \
    if(mv < -bs) mv = -bs - src * bs; \
    if(mv > lim) mv = lim - src * bs;

/** Do motion compensation over 1 macroblock
 * Mostly adapted hpel_motion and qpel_motion from mpegvideo.c
 */
static void vc1_mc_1mv(VC1Context *v)
{
    MpegEncContext *s = &v->s;
    DSPContext *dsp = &v->s.dsp;
    uint8_t *srcY, *srcU, *srcV;
    int dxy, uvdxy, mx, my, uvmx, uvmy, src_x, src_y, uvsrc_x, uvsrc_y;

    if(!v->s.last_picture.data[0])return;

    mx = s->mv[0][0][0];
    my = s->mv[0][0][1];
    uvmx = (mx + ((mx & 3) == 3)) >> 1;
    uvmy = (my + ((my & 3) == 3)) >> 1;
    srcY = s->last_picture.data[0];
    srcU = s->last_picture.data[1];
    srcV = s->last_picture.data[2];

    if(v->fastuvmc) { // XXX: 8.3.5.4.5 specifies something different
        uvmx = (uvmx + 1) >> 1;
        uvmy = (uvmy + 1) >> 1;
    }

    src_x = s->mb_x * 16 + (mx >> 2);
    src_y = s->mb_y * 16 + (my >> 2);
    uvsrc_x = s->mb_x * 8 + (uvmx >> 2);
    uvsrc_y = s->mb_y * 8 + (uvmy >> 2);

    CLIP_RANGE(  src_x, s->mb_x, s->mb_width  * 16, 16);
    CLIP_RANGE(  src_y, s->mb_y, s->mb_height * 16, 16);
    CLIP_RANGE(uvsrc_x, s->mb_x, s->mb_width  *  8,  8);
    CLIP_RANGE(uvsrc_y, s->mb_y, s->mb_height *  8,  8);

    srcY += src_y * s->linesize + src_x;
    srcU += uvsrc_y * s->uvlinesize + uvsrc_x;
    srcV += uvsrc_y * s->uvlinesize + uvsrc_x;

    if((unsigned)src_x > s->h_edge_pos - (mx&3) - 16
       || (unsigned)src_y > s->v_edge_pos - (my&3) - 16){
        uint8_t *uvbuf= s->edge_emu_buffer + 18 * s->linesize;

        ff_emulated_edge_mc(s->edge_emu_buffer, srcY, s->linesize, 16+1, 16+1,
                            src_x, src_y, s->h_edge_pos, s->v_edge_pos);
        srcY = s->edge_emu_buffer;
        ff_emulated_edge_mc(uvbuf     , srcU, s->uvlinesize, 8+1, 8+1,
                            uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, s->v_edge_pos >> 1);
        ff_emulated_edge_mc(uvbuf + 16, srcV, s->uvlinesize, 8+1, 8+1,
                            uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, s->v_edge_pos >> 1);
        srcU = uvbuf;
        srcV = uvbuf + 16;
    }

    if(!s->quarter_sample) { // hpel mc
        mx >>= 1;
        my >>= 1;
        uvmx >>= 1;
        uvmy >>= 1;
        dxy = ((my & 1) << 1) | (mx & 1);
        uvdxy = 0;

        dsp->put_no_rnd_pixels_tab[0][dxy](s->dest[0], srcY, s->linesize, 16);
    } else {
        dxy = ((my & 3) << 2) | (mx & 3);
        uvdxy = ((uvmy & 1) << 1) | (uvmx & 1);

        dsp->put_no_rnd_qpel_pixels_tab[0][dxy](s->dest[0], srcY, s->linesize);
    }
    dsp->put_mspel_pixels_tab[uvdxy](s->dest[1], srcU, s->uvlinesize);
    dsp->put_mspel_pixels_tab[uvdxy](s->dest[2], srcV, s->uvlinesize);
}

/**
 * Decode Simple/Main Profiles sequence header
 * @see Figure 7-8, p16-17

            v->mv_mode2 = mv_pmode_table[lowquant][get_prefix(gb, 1, 3)];
            v->lumscale = get_bits(gb, 6);
            v->lumshift = get_bits(gb, 6);
        }
        if(v->mv_mode == MV_PMODE_1MV_HPEL || v->mv_mode == MV_PMODE_1MV_HPEL_BILIN)
            v->s.quarter_sample = 0;
        else
            v->s.quarter_sample = 1;

if(v->mv_mode != MV_PMODE_1MV && v->mv_mode != MV_PMODE_1MV_HPEL && v->mv_mode != MV_PMODE_1MV_HPEL_BILIN) {
    av_log(v->s.avctx, AV_LOG_ERROR, "Only 1MV P-frames are supported by now\n");
    return -1;
}
        if ((v->mv_mode == MV_PMODE_INTENSITY_COMP &&
                 v->mv_mode2 == MV_PMODE_MIXED_MV)
                || v->mv_mode == MV_PMODE_MIXED_MV)
        {
            status = bitplane_decoding(v->mv_type_mb_plane, &v->mv_type_is_raw, v);
            if (status < 0) return -1;
            av_log(v->s.avctx, AV_LOG_DEBUG, "MB MV Type plane encoding: "
                   "Imode: %i, Invert: %i\n", status>>1, status&1);
        } else {
            v->mv_type_is_raw = 0;
            memset(v->mv_type_mb_plane, 0, v->s.mb_stride * v->s.mb_height);
        }
        status = bitplane_decoding(v->skip_mb_plane, &v->skip_is_raw, v);
        if (status < 0) return -1;
        av_log(v->s.avctx, AV_LOG_DEBUG, "MB Skip plane encoding: "
               "Imode: %i, Invert: %i\n", status>>1, status&1);

        /* Hopefully this is correct for P frames */

  else mb_has_coeffs = 0;                                           \
  s->mb_intra = 0;                                                  \
  if (!index) { _dmv_x = _dmv_y = 0; }                              \
  else if (index == 35)                                             \
  {                                                                 \
    _dmv_x = get_bits(gb, v->k_x - 1 + s->quarter_sample);          \
    _dmv_y = get_bits(gb, v->k_y - 1 + s->quarter_sample);          \
  }                                                                 \
  else if (index == 36)                                             \
  {                                                                 \
    _dmv_x = 0;                                                     \
    _dmv_y = 0;                                                     \
    s->mb_intra = 1;                                                \
  }                                                                 \
  else                                                              \
  {                                                                 \
    index1 = index%6;                                               \
    if (!s->quarter_sample && index1 == 5) val = 1;                 \
    else                                   val = 0;                 \
    if(size_table[index1] - val > 0)                                \
        val = get_bits(gb, size_table[index1] - val);               \
    else                                   val = 0;                 \
    sign = 0 - (val&1);                                             \
    _dmv_x = (sign ^ ((val>>1) + offset_table[index1])) - sign;     \
                                                                    \
    index1 = index/6;                                               \
    if (!s->quarter_sample && index1 == 5) val = 1;                 \
    else                                   val = 0;                 \
    if(size_table[index1] - val > 0)                                \
        val = get_bits(gb, size_table[index1] - val);               \
    else                                   val = 0;                 \
    sign = 0 - (val&1);                                             \
    _dmv_y = (sign ^ ((val>>1) + offset_table[index1])) - sign;     \

    int px, py;
    int sum;
    int mb_pos = s->mb_x + s->mb_y * s->mb_stride;

    /* scale MV difference to be quad-pel */
    dmv_x <<= 1 - s->quarter_sample;
    dmv_y <<= 1 - s->quarter_sample;

    wrap = s->b8_stride;
    xy = s->block_index[0];

    C = s->current_picture.motion_val[0][xy - (1 << mv1)];

        lst = index >= vc1_last_decode_table[codingset];
        if(get_bits(gb, 1))
            level = -level;
    } else {
        escape = decode210(gb);
        if (escape != 2) {
            index = get_vlc2(gb, vc1_ac_coeff_table[codingset].table, AC_VLC_BITS, 3);
            run = vc1_index_decode_table[codingset][index][0];
            level = vc1_index_decode_table[codingset][index][1];
            lst = index >= vc1_last_decode_table[codingset];
            if(escape == 0) {
                if(lst)
                    level += vc1_last_delta_level_table[codingset][run];
                else
                    level += vc1_delta_level_table[codingset][run];
            } else {
                if(lst)
                    run += vc1_last_delta_run_table[codingset][level] + 1;
                else
                    run += vc1_delta_run_table[codingset][level] + 1;
            }
            if(get_bits(gb, 1))
                level = -level;
        } else {
            int sign;
            lst = get_bits(gb, 1);

                break;
            block[zz_table[i++]] = value;
        }

        /* apply AC prediction if needed */
        if(s->ac_pred && (v->a_avail || v->c_avail)) {
            /* scale predictors if needed*/
            int mb_pos2, q1, q2;

            mb_pos2 = mb_pos - dc_pred_dir - (1 - dc_pred_dir) * s->mb_stride;
            q1 = s->current_picture.qscale_table[mb_pos];

            }
            if(!a_avail) {
                memset(ac_val + 8, 0, 8 * sizeof(ac_val[0]));
                dc_pred_dir = 1;
            }
            if(q2 && q1 != q2) {
                q1 = q1 * 2 - 1;
                q2 = q2 * 2 - 1;

                if(dc_pred_dir) { //left
                    for(k = 1; k < 8; k++)

        scale = mquant * 2;
        memset(ac_val2, 0, 16 * 2);
        if(dc_pred_dir) {//left
            ac_val -= 16;
            if(s->ac_pred && (v->a_avail || v->c_avail))
                memcpy(ac_val2, ac_val, 8 * 2);
        } else {//top
            ac_val -= 16 * s->block_wrap[n];
            if(s->ac_pred && (v->a_avail || v->c_avail))
                memcpy(ac_val2 + 8, ac_val + 8, 8 * 2);
        }

        /* apply AC prediction if needed */
        if(s->ac_pred && (v->a_avail || v->c_avail)) {
            if(dc_pred_dir) { //left
                for(k = 1; k < 8; k++) {
                    block[k << 3] = ac_val[k] * scale;
                    if(!v->pquantizer)
                        block[k << 3] += (block[k << 3] < 0) ? -mquant : mquant;

    }
    if(ttblk == TT_4X8_RIGHT || ttblk == TT_4X8_LEFT) {
        ttblk = TT_4X8;
        subblkpat = 2 - (ttblk == TT_4X8_LEFT);
    }
    switch(ttblk) {
    case TT_8X8:
        i = 0;
        last = 0;
        while (!last) {

        break;
    case TT_4X4:
        for(j = 0; j < 4; j++) {
            last = subblkpat & (1 << (3 - j));
            i = 0;
            off = (j & 1) * 4 + (j & 2) * 16;
            while (!last) {
                vc1_decode_ac_coeff(v, &last, &skip, &value, v->codingset2);
                i += skip;
                if(i > 15)
                    break;
                idx = vc1_simple_progressive_4x4_zz[i++];
                block[idx + off] = value * scale;
            }
            if(!(subblkpat & (1 << (3 - j))))
                vc1_inv_trans(block + off, 4, 4);
        }
        break;
    case TT_8X4:
        for(j = 0; j < 2; j++) {
            last = subblkpat & (1 << (1 - j));

                if(i > 31)
                    break;
                idx = vc1_simple_progressive_8x4_zz[i++];
                block[idx + off] = value * scale;
            }
            if(!(subblkpat & (1 << (1 - j))))
                vc1_inv_trans(block + off, 8, 4);
        }
        break;
    case TT_4X8:
        for(j = 0; j < 2; j++) {
            last = subblkpat & (1 << (1 - j));

                if(i > 31)
                    break;
                idx = vc1_simple_progressive_8x4_zz[i++];
                block[idx + off] = value * scale;
            }
            if(!(subblkpat & (1 << (1 - j))))
                vc1_inv_trans(block + off, 4, 8);
        }
        break;
    }
    return 0;
}

 */
static int vc1_decode_p_mb(VC1Context *v, DCTELEM block[6][64])
{
    MpegEncContext *s = &v->s;
    GetBitContext *gb = &s->gb;
    int i, j;
    int mb_pos = s->mb_x + s->mb_y * s->mb_stride;
    int cbp; /* cbp decoding stuff */
    int hybrid_pred; /* Prediction types */
    int mqdiff, mquant; /* MB quantization */
    int ttmb = v->ttmb; /* MB Transform type */

    int dmv_x, dmv_y; /* Differential MV components */
    int index, index1; /* LUT indices */
    int val, sign; /* temp values */
    int first_block = 1;
    int dst_idx, off;
    int skipped, fourmv;

    mquant = v->pq; /* Loosy initialization */

    if (v->mv_type_is_raw)
        fourmv = get_bits1(gb);
    else
        fourmv = v->mv_type_mb_plane[mb_pos];
    if (v->skip_is_raw)
        skipped = get_bits1(gb);
    else
        skipped = v->skip_mb_plane[mb_pos];

    if (!fourmv) /* 1MV mode */
    {
        if (!skipped)
        {
            GET_MVDATA(dmv_x, dmv_y);

            s->current_picture.mb_type[mb_pos] = s->mb_intra ? MB_TYPE_INTRA : MB_TYPE_16x16;
            vc1_pred_mv(s, dmv_x, dmv_y, 1, v->range_x, v->range_y);

                        v->a_avail = 1;
                    if((i == 1 || i == 3) || (s->mb_x && IS_INTRA(s->current_picture.mb_type[mb_pos - 1])))
                        v->c_avail = 1;

                    vc1_decode_intra_block(v, block[i], i, val, mquant, (i&4)?v->codingset2:v->codingset);
                    vc1_inv_trans(block[i], 8, 8);
                    for(j = 0; j < 64; j++) block[i][j] += 128;
                    s->dsp.put_pixels_clamped(block[i], s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2));
                    /* TODO: proper loop filtering */
                    if(v->pq >= 9 && v->overlap) {
                        if(v->a_avail)
                            s->dsp.h263_v_loop_filter(s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2), s->y_dc_scale);
                        if(v->c_avail)
                            s->dsp.h263_h_loop_filter(s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2), s->y_dc_scale);
                    }
                } else if(val) {
                    vc1_decode_p_block(v, block[i], i, mquant, ttmb, first_block);
                    if(!v->ttmbf && ttmb < 8) ttmb = -1;
                    first_block = 0;
                    s->dsp.add_pixels_clamped(block[i], s->dest[dst_idx] + off, (i&4)?s->uvlinesize:s->linesize);
                }
            }
        }
        else //Skipped
        {

            return 0;
        }
    } //1MV mode
    else //4MV mode
    {//FIXME: looks not conforming to standard and is not even theoretically complete
        if (!skipped /* unskipped MB */)
        {
            int blk_intra[4], blk_coded[4];
            /* Get CBPCY */
            cbp = get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);
            for (i=0; i<4; i++)

    GetBitContext gb;

    if (!avctx->extradata_size || !avctx->extradata) return -1;
    avctx->pix_fmt = PIX_FMT_YUV420P;
    v->s.avctx = avctx;
    avctx->flags |= CODEC_FLAG_EMU_EDGE;
    v->s.flags |= CODEC_FLAG_EMU_EDGE;

    if(ff_h263_decode_init(avctx) < 0)
        return -1;
    if (vc1_init_common(v) < 0) return -1;

    av_log(avctx, AV_LOG_INFO, "This decoder is not supposed to produce picture. Dont report this as a bug!\n");
    av_log(avctx, AV_LOG_INFO, "If you see a picture, don't believe your eyes.\n");

    avctx->coded_width = avctx->width;
    avctx->coded_height = avctx->height;
    if (avctx->codec_id == CODEC_ID_WMV3)
    {
        int count = 0;

    s->mb_width = (avctx->coded_width+15)>>4;
    s->mb_height = (avctx->coded_height+15)>>4;

    /* Allocate mb bitplanes */
    v->mv_type_mb_plane = av_malloc(s->mb_stride * s->mb_height);
    v->skip_mb_plane = av_malloc(s->mb_stride * s->mb_height);

    /* For predictors */
    v->previous_line_cbpcy = (uint8_t *)av_malloc(s->mb_stride*4);
    if (!v->previous_line_cbpcy) return -1;

    /* Init coded blocks info */
    if (v->profile == PROFILE_ADVANCED)
    {
//        if (alloc_bitplane(&v->over_flags_plane, s->mb_width, s->mb_height) < 0)
//            return -1;
//        if (alloc_bitplane(&v->ac_pred_plane, s->mb_width, s->mb_height) < 0)
//            return -1;
    }

    return 0;
}

    VC1Context *v = avctx->priv_data;

    av_freep(&v->hrd_rate);
    av_freep(&v->hrd_buffer);
    MPV_common_end(&v->s);
    av_freep(&v->mv_type_mb_plane);
    av_freep(&v->skip_mb_plane);
    return 0;
}


AVCodec vc1_decoder = {