vc1.c

/*
 * VC-1 and WMV3 decoder
 * Copyright (c) 2006-2007 Konstantin Shishkov
 * Partly based on vc9.c (c) 2005 Anonymous, Alex Beregszaszi, Michael Niedermayer
 *
 * This file is part of FFmpeg.
 *
 * FFmpeg is free software; you can redistribute it and/or
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
 * version 2.1 of the License, or (at your option) any later version.
 *
 * FFmpeg is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * Lesser General Public License for more details.
 *
 * You should have received a copy of the GNU Lesser General Public
 * License along with FFmpeg; if not, write to the Free Software
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
 */

/**
 * @file vc1.c
 * VC-1 and WMV3 decoder
 *
 */
#include "dsputil.h"
#include "avcodec.h"
#include "mpegvideo.h"
#include "vc1.h"
#include "vc1data.h"
#include "vc1acdata.h"
#include "msmpeg4data.h"

#undef NDEBUG
#include <assert.h>

#define MB_INTRA_VLC_BITS 9
#define DC_VLC_BITS 9
#define AC_VLC_BITS 9
static const uint16_t table_mb_intra[64][2];


/**
 * Get unary code of limited length
 * @fixme FIXME Slow and ugly
 * @param gb GetBitContext
 * @param[in] stop The bitstop value (unary code of 1's or 0's)
 * @param[in] len Maximum length
 * @return Unary length/index
 */
static int get_prefix(GetBitContext *gb, int stop, int len)
{
#if 1
    int i;

    for(i = 0; i < len && get_bits1(gb) != stop; i++);
    return i;
/*  int i = 0, tmp = !stop;

  while (i != len && tmp != stop)
  {
    tmp = get_bits(gb, 1);
    i++;
  }
  if (i == len && tmp != stop) return len+1;
  return i;*/
#else
  unsigned int buf;
  int log;

  OPEN_READER(re, gb);
  UPDATE_CACHE(re, gb);
  buf=GET_CACHE(re, gb); //Still not sure
  if (stop) buf = ~buf;

  log= av_log2(-buf); //FIXME: -?
  if (log < limit){
    LAST_SKIP_BITS(re, gb, log+1);
    CLOSE_READER(re, gb);
    return log;
  }

  LAST_SKIP_BITS(re, gb, limit);
  CLOSE_READER(re, gb);
  return limit;
#endif
}

static inline int decode210(GetBitContext *gb){
    if (get_bits1(gb))
        return 0;
    else
        return 2 - get_bits1(gb);
}

/**
 * Init VC-1 specific tables and VC1Context members
 * @param v The VC1Context to initialize
 * @return Status
 */
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)
    {
        done = 1;
        init_vlc(&ff_vc1_bfraction_vlc, VC1_BFRACTION_VLC_BITS, 23,
                 ff_vc1_bfraction_bits, 1, 1,
                 ff_vc1_bfraction_codes, 1, 1, 1);
        init_vlc(&ff_vc1_norm2_vlc, VC1_NORM2_VLC_BITS, 4,
                 ff_vc1_norm2_bits, 1, 1,
                 ff_vc1_norm2_codes, 1, 1, 1);
        init_vlc(&ff_vc1_norm6_vlc, VC1_NORM6_VLC_BITS, 64,
                 ff_vc1_norm6_bits, 1, 1,
                 ff_vc1_norm6_codes, 2, 2, 1);
        init_vlc(&ff_vc1_imode_vlc, VC1_IMODE_VLC_BITS, 7,
                 ff_vc1_imode_bits, 1, 1,
                 ff_vc1_imode_codes, 1, 1, 1);
        for (i=0; i<3; i++)
        {
            init_vlc(&ff_vc1_ttmb_vlc[i], VC1_TTMB_VLC_BITS, 16,
                     ff_vc1_ttmb_bits[i], 1, 1,
                     ff_vc1_ttmb_codes[i], 2, 2, 1);
            init_vlc(&ff_vc1_ttblk_vlc[i], VC1_TTBLK_VLC_BITS, 8,
                     ff_vc1_ttblk_bits[i], 1, 1,
                     ff_vc1_ttblk_codes[i], 1, 1, 1);
            init_vlc(&ff_vc1_subblkpat_vlc[i], VC1_SUBBLKPAT_VLC_BITS, 15,
                     ff_vc1_subblkpat_bits[i], 1, 1,
                     ff_vc1_subblkpat_codes[i], 1, 1, 1);
        }
        for(i=0; i<4; i++)
        {
            init_vlc(&ff_vc1_4mv_block_pattern_vlc[i], VC1_4MV_BLOCK_PATTERN_VLC_BITS, 16,
                     ff_vc1_4mv_block_pattern_bits[i], 1, 1,
                     ff_vc1_4mv_block_pattern_codes[i], 1, 1, 1);
            init_vlc(&ff_vc1_cbpcy_p_vlc[i], VC1_CBPCY_P_VLC_BITS, 64,
                     ff_vc1_cbpcy_p_bits[i], 1, 1,
                     ff_vc1_cbpcy_p_codes[i], 2, 2, 1);
            init_vlc(&ff_vc1_mv_diff_vlc[i], VC1_MV_DIFF_VLC_BITS, 73,
                     ff_vc1_mv_diff_bits[i], 1, 1,
                     ff_vc1_mv_diff_codes[i], 2, 2, 1);
        }
        for(i=0; i<8; i++)
            init_vlc(&ff_vc1_ac_coeff_table[i], AC_VLC_BITS, vc1_ac_sizes[i],
                     &vc1_ac_tables[i][0][1], 8, 4,
                     &vc1_ac_tables[i][0][0], 8, 4, 1);
        init_vlc(&ff_msmp4_mb_i_vlc, MB_INTRA_VLC_BITS, 64,
                 &ff_msmp4_mb_i_table[0][1], 4, 2,
                 &ff_msmp4_mb_i_table[0][0], 4, 2, 1);
    }

    /* Other defaults */
    v->pq = -1;
    v->mvrange = 0; /* 7.1.1.18, p80 */

    return 0;
}

/***********************************************************************/
/**
 * @defgroup bitplane VC9 Bitplane decoding
 * @see 8.7, p56
 * @{
 */

/** @addtogroup bitplane
 * Imode types
 * @{
 */
enum Imode {
    IMODE_RAW,
    IMODE_NORM2,
    IMODE_DIFF2,
    IMODE_NORM6,
    IMODE_DIFF6,
    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
 */
static void decode_rowskip(uint8_t* plane, int width, int height, int stride, GetBitContext *gb){
    int x, y;

    for (y=0; y<height; y++){
        if (!get_bits(gb, 1)) //rowskip
            memset(plane, 0, width);
        else
            for (x=0; x<width; x++)
                plane[x] = get_bits(gb, 1);
        plane += stride;
    }
}

/** Decode columns 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
 * @fixme FIXME: Optimize
 */
static void decode_colskip(uint8_t* plane, int width, int height, int stride, GetBitContext *gb){
    int x, y;

    for (x=0; x<width; x++){
        if (!get_bits(gb, 1)) //colskip
            for (y=0; y<height; y++)
                plane[y*stride] = 0;
        else
            for (y=0; y<height; y++)
                plane[y*stride] = get_bits(gb, 1);
        plane ++;
    }
}

/** Decode a bitplane's bits
 * @param bp Bitplane where to store the decode bits
 * @param v VC-1 context for bit reading and logging
 * @return Status
 * @fixme FIXME: Optimize
 */
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, ff_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, ff_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, ff_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
            planep += (height & 1) * stride;
            for(y = height & 1; y < height; y += 2) {
                for(x = width % 3; x < width; x += 3) {
                    code = get_vlc2(gb, ff_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
 */
static int vop_dquant_decoding(VC1Context *v)
{
    GetBitContext *gb = &v->s.gb;
    int pqdiff;

    //variable size
    if (v->dquant == 2)
    {
        pqdiff = get_bits(gb, 3);
        if (pqdiff == 7) v->altpq = get_bits(gb, 5);
        else v->altpq = v->pq + pqdiff + 1;
    }
    else
    {
        v->dquantfrm = get_bits(gb, 1);
        if ( v->dquantfrm )
        {
            v->dqprofile = get_bits(gb, 2);
            switch (v->dqprofile)
            {
            case DQPROFILE_SINGLE_EDGE:
            case DQPROFILE_DOUBLE_EDGES:
                v->dqsbedge = get_bits(gb, 2);
                break;
            case DQPROFILE_ALL_MBS:
                v->dqbilevel = get_bits(gb, 1);
            default: break; //Forbidden ?
            }
            if (v->dqbilevel || v->dqprofile != DQPROFILE_ALL_MBS)
            {
                pqdiff = get_bits(gb, 3);
                if (pqdiff == 7) v->altpq = get_bits(gb, 5);
                else v->altpq = v->pq + pqdiff + 1;
            }
        }
    }
    return 0;
}

/** Put block onto picture
 */
static void vc1_put_block(VC1Context *v, DCTELEM block[6][64])
{
    uint8_t *Y;
    int ys, us, vs;
    DSPContext *dsp = &v->s.dsp;

    if(v->rangeredfrm) {
        int i, j, k;
        for(k = 0; k < 6; k++)
            for(j = 0; j < 8; j++)
                for(i = 0; i < 8; i++)
                    block[k][i + j*8] = ((block[k][i + j*8] - 128) << 1) + 128;

    }
    ys = v->s.current_picture.linesize[0];
    us = v->s.current_picture.linesize[1];
    vs = v->s.current_picture.linesize[2];
    Y = v->s.dest[0];

    dsp->put_pixels_clamped(block[0], Y, ys);
    dsp->put_pixels_clamped(block[1], Y + 8, ys);
    Y += ys * 8;
    dsp->put_pixels_clamped(block[2], Y, ys);
    dsp->put_pixels_clamped(block[3], Y + 8, ys);

    if(!(v->s.flags & CODEC_FLAG_GRAY)) {
        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, int dir)
{
    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[dir][0][0];
    my = s->mv[dir][0][1];

    // store motion vectors for further use in B frames
    if(s->pict_type == P_TYPE) {
        s->current_picture.motion_val[1][s->block_index[0]][0] = mx;
        s->current_picture.motion_val[1][s->block_index[0]][1] = my;
    }
    uvmx = (mx + ((mx & 3) == 3)) >> 1;
    uvmy = (my + ((my & 3) == 3)) >> 1;
    if(v->fastuvmc) {
        uvmx = uvmx + ((uvmx<0)?(uvmx&1):-(uvmx&1));
        uvmy = uvmy + ((uvmy<0)?(uvmy&1):-(uvmy&1));
    }
    if(!dir) {
        srcY = s->last_picture.data[0];
        srcU = s->last_picture.data[1];
        srcV = s->last_picture.data[2];
    } else {
        srcY = s->next_picture.data[0];
        srcU = s->next_picture.data[1];
        srcV = s->next_picture.data[2];
    }

    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);

    if(v->profile != PROFILE_ADVANCED){
        src_x   = av_clip(  src_x, -16, s->mb_width  * 16);
        src_y   = av_clip(  src_y, -16, s->mb_height * 16);
        uvsrc_x = av_clip(uvsrc_x,  -8, s->mb_width  *  8);
        uvsrc_y = av_clip(uvsrc_y,  -8, s->mb_height *  8);
    }else{
        src_x   = av_clip(  src_x, -17, s->avctx->coded_width);
        src_y   = av_clip(  src_y, -18, s->avctx->coded_height + 1);
        uvsrc_x = av_clip(uvsrc_x,  -8, s->avctx->coded_width  >> 1);
        uvsrc_y = av_clip(uvsrc_y,  -8, s->avctx->coded_height >> 1);
    }

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

    /* for grayscale we should not try to read from unknown area */
    if(s->flags & CODEC_FLAG_GRAY) {
        srcU = s->edge_emu_buffer + 18 * s->linesize;
        srcV = s->edge_emu_buffer + 18 * s->linesize;
    }

    if(v->rangeredfrm || (v->mv_mode == MV_PMODE_INTENSITY_COMP)
       || (unsigned)(src_x - s->mspel) > s->h_edge_pos - (mx&3) - 16 - s->mspel*3
       || (unsigned)(src_y - s->mspel) > s->v_edge_pos - (my&3) - 16 - s->mspel*3){
        uint8_t *uvbuf= s->edge_emu_buffer + 19 * s->linesize;

        srcY -= s->mspel * (1 + s->linesize);
        ff_emulated_edge_mc(s->edge_emu_buffer, srcY, s->linesize, 17+s->mspel*2, 17+s->mspel*2,
                            src_x - s->mspel, src_y - s->mspel, 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 we deal with range reduction we need to scale source blocks */
        if(v->rangeredfrm) {
            int i, j;
            uint8_t *src, *src2;

            src = srcY;
            for(j = 0; j < 17 + s->mspel*2; j++) {
                for(i = 0; i < 17 + s->mspel*2; i++) src[i] = ((src[i] - 128) >> 1) + 128;
                src += s->linesize;
            }
            src = srcU; src2 = srcV;
            for(j = 0; j < 9; j++) {
                for(i = 0; i < 9; i++) {
                    src[i] = ((src[i] - 128) >> 1) + 128;
                    src2[i] = ((src2[i] - 128) >> 1) + 128;
                }
                src += s->uvlinesize;
                src2 += s->uvlinesize;
            }
        }
        /* if we deal with intensity compensation we need to scale source blocks */
        if(v->mv_mode == MV_PMODE_INTENSITY_COMP) {
            int i, j;
            uint8_t *src, *src2;

            src = srcY;
            for(j = 0; j < 17 + s->mspel*2; j++) {
                for(i = 0; i < 17 + s->mspel*2; i++) src[i] = v->luty[src[i]];
                src += s->linesize;
            }
            src = srcU; src2 = srcV;
            for(j = 0; j < 9; j++) {
                for(i = 0; i < 9; i++) {
                    src[i] = v->lutuv[src[i]];
                    src2[i] = v->lutuv[src2[i]];
                }
                src += s->uvlinesize;
                src2 += s->uvlinesize;
            }
        }
        srcY += s->mspel * (1 + s->linesize);
    }

    if(s->mspel) {
        dxy = ((my & 3) << 2) | (mx & 3);
        dsp->put_vc1_mspel_pixels_tab[dxy](s->dest[0]    , srcY    , s->linesize, v->rnd);
        dsp->put_vc1_mspel_pixels_tab[dxy](s->dest[0] + 8, srcY + 8, s->linesize, v->rnd);
        srcY += s->linesize * 8;
        dsp->put_vc1_mspel_pixels_tab[dxy](s->dest[0] + 8 * s->linesize    , srcY    , s->linesize, v->rnd);
        dsp->put_vc1_mspel_pixels_tab[dxy](s->dest[0] + 8 * s->linesize + 8, srcY + 8, s->linesize, v->rnd);
    } else { // hpel mc - always used for luma
        dxy = (my & 2) | ((mx & 2) >> 1);

        if(!v->rnd)
            dsp->put_pixels_tab[0][dxy](s->dest[0], srcY, s->linesize, 16);
        else
            dsp->put_no_rnd_pixels_tab[0][dxy](s->dest[0], srcY, s->linesize, 16);
    }

    if(s->flags & CODEC_FLAG_GRAY) return;
    /* Chroma MC always uses qpel bilinear */
    uvdxy = ((uvmy & 3) << 2) | (uvmx & 3);
    uvmx = (uvmx&3)<<1;
    uvmy = (uvmy&3)<<1;
    if(!v->rnd){
        dsp->put_h264_chroma_pixels_tab[0](s->dest[1], srcU, s->uvlinesize, 8, uvmx, uvmy);
        dsp->put_h264_chroma_pixels_tab[0](s->dest[2], srcV, s->uvlinesize, 8, uvmx, uvmy);
    }else{
        dsp->put_no_rnd_h264_chroma_pixels_tab[0](s->dest[1], srcU, s->uvlinesize, 8, uvmx, uvmy);
        dsp->put_no_rnd_h264_chroma_pixels_tab[0](s->dest[2], srcV, s->uvlinesize, 8, uvmx, uvmy);
    }
}

/** Do motion compensation for 4-MV macroblock - luminance block
 */
static void vc1_mc_4mv_luma(VC1Context *v, int n)
{
    MpegEncContext *s = &v->s;
    DSPContext *dsp = &v->s.dsp;
    uint8_t *srcY;
    int dxy, mx, my, src_x, src_y;
    int off;

    if(!v->s.last_picture.data[0])return;
    mx = s->mv[0][n][0];
    my = s->mv[0][n][1];
    srcY = s->last_picture.data[0];

    off = s->linesize * 4 * (n&2) + (n&1) * 8;

    src_x = s->mb_x * 16 + (n&1) * 8 + (mx >> 2);
    src_y = s->mb_y * 16 + (n&2) * 4 + (my >> 2);

    if(v->profile != PROFILE_ADVANCED){
        src_x   = av_clip(  src_x, -16, s->mb_width  * 16);
        src_y   = av_clip(  src_y, -16, s->mb_height * 16);
    }else{
        src_x   = av_clip(  src_x, -17, s->avctx->coded_width);
        src_y   = av_clip(  src_y, -18, s->avctx->coded_height + 1);
    }

    srcY += src_y * s->linesize + src_x;

    if(v->rangeredfrm || (v->mv_mode == MV_PMODE_INTENSITY_COMP)
       || (unsigned)(src_x - s->mspel) > s->h_edge_pos - (mx&3) - 8 - s->mspel*2
       || (unsigned)(src_y - s->mspel) > s->v_edge_pos - (my&3) - 8 - s->mspel*2){
        srcY -= s->mspel * (1 + s->linesize);
        ff_emulated_edge_mc(s->edge_emu_buffer, srcY, s->linesize, 9+s->mspel*2, 9+s->mspel*2,
                            src_x - s->mspel, src_y - s->mspel, s->h_edge_pos, s->v_edge_pos);
        srcY = s->edge_emu_buffer;
        /* if we deal with range reduction we need to scale source blocks */
        if(v->rangeredfrm) {
            int i, j;
            uint8_t *src;

            src = srcY;
            for(j = 0; j < 9 + s->mspel*2; j++) {
                for(i = 0; i < 9 + s->mspel*2; i++) src[i] = ((src[i] - 128) >> 1) + 128;
                src += s->linesize;
            }
        }
        /* if we deal with intensity compensation we need to scale source blocks */
        if(v->mv_mode == MV_PMODE_INTENSITY_COMP) {
            int i, j;
            uint8_t *src;

            src = srcY;
            for(j = 0; j < 9 + s->mspel*2; j++) {
                for(i = 0; i < 9 + s->mspel*2; i++) src[i] = v->luty[src[i]];
                src += s->linesize;
            }
        }
        srcY += s->mspel * (1 + s->linesize);
    }

    if(s->mspel) {
        dxy = ((my & 3) << 2) | (mx & 3);
        dsp->put_vc1_mspel_pixels_tab[dxy](s->dest[0] + off, srcY, s->linesize, v->rnd);
    } else { // hpel mc - always used for luma
        dxy = (my & 2) | ((mx & 2) >> 1);
        if(!v->rnd)
            dsp->put_pixels_tab[1][dxy](s->dest[0] + off, srcY, s->linesize, 8);
        else
            dsp->put_no_rnd_pixels_tab[1][dxy](s->dest[0] + off, srcY, s->linesize, 8);
    }
}

static inline int median4(int a, int b, int c, int d)
{
    if(a < b) {
        if(c < d) return (FFMIN(b, d) + FFMAX(a, c)) / 2;
        else      return (FFMIN(b, c) + FFMAX(a, d)) / 2;
    } else {
        if(c < d) return (FFMIN(a, d) + FFMAX(b, c)) / 2;
        else      return (FFMIN(a, c) + FFMAX(b, d)) / 2;
    }
}


/** Do motion compensation for 4-MV macroblock - both chroma blocks
 */
static void vc1_mc_4mv_chroma(VC1Context *v)
{
    MpegEncContext *s = &v->s;
    DSPContext *dsp = &v->s.dsp;
    uint8_t *srcU, *srcV;
    int uvdxy, uvmx, uvmy, uvsrc_x, uvsrc_y;
    int i, idx, tx = 0, ty = 0;
    int mvx[4], mvy[4], intra[4];
    static const int count[16] = { 0, 1, 1, 2, 1, 2, 2, 3, 1, 2, 2, 3, 2, 3, 3, 4};

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

    for(i = 0; i < 4; i++) {
        mvx[i] = s->mv[0][i][0];
        mvy[i] = s->mv[0][i][1];
        intra[i] = v->mb_type[0][s->block_index[i]];
    }

    /* calculate chroma MV vector from four luma MVs */
    idx = (intra[3] << 3) | (intra[2] << 2) | (intra[1] << 1) | intra[0];
    if(!idx) { // all blocks are inter
        tx = median4(mvx[0], mvx[1], mvx[2], mvx[3]);
        ty = median4(mvy[0], mvy[1], mvy[2], mvy[3]);
    } else if(count[idx] == 1) { // 3 inter blocks
        switch(idx) {
        case 0x1:
            tx = mid_pred(mvx[1], mvx[2], mvx[3]);
            ty = mid_pred(mvy[1], mvy[2], mvy[3]);
            break;
        case 0x2:
            tx = mid_pred(mvx[0], mvx[2], mvx[3]);
            ty = mid_pred(mvy[0], mvy[2], mvy[3]);
            break;
        case 0x4:
            tx = mid_pred(mvx[0], mvx[1], mvx[3]);
            ty = mid_pred(mvy[0], mvy[1], mvy[3]);
            break;
        case 0x8:
            tx = mid_pred(mvx[0], mvx[1], mvx[2]);
            ty = mid_pred(mvy[0], mvy[1], mvy[2]);
            break;
        }
    } else if(count[idx] == 2) {
        int t1 = 0, t2 = 0;
        for(i=0; i<3;i++) if(!intra[i]) {t1 = i; break;}
        for(i= t1+1; i<4; i++)if(!intra[i]) {t2 = i; break;}
        tx = (mvx[t1] + mvx[t2]) / 2;
        ty = (mvy[t1] + mvy[t2]) / 2;
    } else {
        s->current_picture.motion_val[1][s->block_index[0]][0] = 0;
        s->current_picture.motion_val[1][s->block_index[0]][1] = 0;
        return; //no need to do MC for inter blocks
    }

    s->current_picture.motion_val[1][s->block_index[0]][0] = tx;
    s->current_picture.motion_val[1][s->block_index[0]][1] = ty;
    uvmx = (tx + ((tx&3) == 3)) >> 1;
    uvmy = (ty + ((ty&3) == 3)) >> 1;
    if(v->fastuvmc) {
        uvmx = uvmx + ((uvmx<0)?(uvmx&1):-(uvmx&1));
        uvmy = uvmy + ((uvmy<0)?(uvmy&1):-(uvmy&1));
    }

    uvsrc_x = s->mb_x * 8 + (uvmx >> 2);
    uvsrc_y = s->mb_y * 8 + (uvmy >> 2);

    if(v->profile != PROFILE_ADVANCED){
        uvsrc_x = av_clip(uvsrc_x,  -8, s->mb_width  *  8);
        uvsrc_y = av_clip(uvsrc_y,  -8, s->mb_height *  8);
    }else{
        uvsrc_x = av_clip(uvsrc_x,  -8, s->avctx->coded_width  >> 1);
        uvsrc_y = av_clip(uvsrc_y,  -8, s->avctx->coded_height >> 1);
    }

    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(v->rangeredfrm || (v->mv_mode == MV_PMODE_INTENSITY_COMP)
       || (unsigned)uvsrc_x > (s->h_edge_pos >> 1) - 9
       || (unsigned)uvsrc_y > (s->v_edge_pos >> 1) - 9){
        ff_emulated_edge_mc(s->edge_emu_buffer     , 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(s->edge_emu_buffer + 16, srcV, s->uvlinesize, 8+1, 8+1,
                            uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, s->v_edge_pos >> 1);
        srcU = s->edge_emu_buffer;
        srcV = s->edge_emu_buffer + 16;

        /* if we deal with range reduction we need to scale source blocks */
        if(v->rangeredfrm) {
            int i, j;
            uint8_t *src, *src2;

            src = srcU; src2 = srcV;
            for(j = 0; j < 9; j++) {
                for(i = 0; i < 9; i++) {
                    src[i] = ((src[i] - 128) >> 1) + 128;
                    src2[i] = ((src2[i] - 128) >> 1) + 128;
                }
                src += s->uvlinesize;
                src2 += s->uvlinesize;
            }
        }
        /* if we deal with intensity compensation we need to scale source blocks */
        if(v->mv_mode == MV_PMODE_INTENSITY_COMP) {
            int i, j;
            uint8_t *src, *src2;

            src = srcU; src2 = srcV;
            for(j = 0; j < 9; j++) {
                for(i = 0; i < 9; i++) {
                    src[i] = v->lutuv[src[i]];
                    src2[i] = v->lutuv[src2[i]];
                }
                src += s->uvlinesize;
                src2 += s->uvlinesize;
            }
        }
    }

    /* Chroma MC always uses qpel bilinear */
    uvdxy = ((uvmy & 3) << 2) | (uvmx & 3);
    uvmx = (uvmx&3)<<1;
    uvmy = (uvmy&3)<<1;
    if(!v->rnd){
        dsp->put_h264_chroma_pixels_tab[0](s->dest[1], srcU, s->uvlinesize, 8, uvmx, uvmy);
        dsp->put_h264_chroma_pixels_tab[0](s->dest[2], srcV, s->uvlinesize, 8, uvmx, uvmy);
    }else{
        dsp->put_no_rnd_h264_chroma_pixels_tab[0](s->dest[1], srcU, s->uvlinesize, 8, uvmx, uvmy);
        dsp->put_no_rnd_h264_chroma_pixels_tab[0](s->dest[2], srcV, s->uvlinesize, 8, uvmx, uvmy);
    }
}

static int decode_sequence_header_adv(VC1Context *v, GetBitContext *gb);

/**
 * Decode Simple/Main Profiles sequence header
 * @see Figure 7-8, p16-17
 * @param avctx Codec context
 * @param gb GetBit context initialized from Codec context extra_data
 * @return Status
 */
static int decode_sequence_header(AVCodecContext *avctx, GetBitContext *gb)
{
    VC1Context *v = avctx->priv_data;

    av_log(avctx, AV_LOG_DEBUG, "Header: %0X\n", show_bits(gb, 32));
    v->profile = get_bits(gb, 2);
    if (v->profile == PROFILE_COMPLEX)
    {
        av_log(avctx, AV_LOG_ERROR, "WMV3 Complex Profile is not fully supported\n");
    }

    if (v->profile == PROFILE_ADVANCED)
    {
        return decode_sequence_header_adv(v, gb);
    }
    else
    {
        v->res_sm = get_bits(gb, 2); //reserved
        if (v->res_sm)
        {
            av_log(avctx, AV_LOG_ERROR,
                   "Reserved RES_SM=%i is forbidden\n", v->res_sm);
            return -1;
        }
    }

    // (fps-2)/4 (->30)
    v->frmrtq_postproc = get_bits(gb, 3); //common
    // (bitrate-32kbps)/64kbps
    v->bitrtq_postproc = get_bits(gb, 5); //common
    v->s.loop_filter = get_bits(gb, 1); //common
    if(v->s.loop_filter == 1 && v->profile == PROFILE_SIMPLE)
    {
        av_log(avctx, AV_LOG_ERROR,
               "LOOPFILTER shell not be enabled in simple profile\n");
    }

    v->res_x8 = get_bits(gb, 1); //reserved
    if (v->res_x8)
    {
        av_log(avctx, AV_LOG_ERROR,
               "1 for reserved RES_X8 is forbidden\n");
        //return -1;
    }
    v->multires = get_bits(gb, 1);
    v->res_fasttx = get_bits(gb, 1);
    if (!v->res_fasttx)
    {
        av_log(avctx, AV_LOG_ERROR,
               "0 for reserved RES_FASTTX is forbidden\n");
        //return -1;
    }

    v->fastuvmc =  get_bits(gb, 1); //common
    if (!v->profile && !v->fastuvmc)
    {
        av_log(avctx, AV_LOG_ERROR,
               "FASTUVMC unavailable in Simple Profile\n");
        return -1;
    }
    v->extended_mv =  get_bits(gb, 1); //common
    if (!v->profile && v->extended_mv)
    {
        av_log(avctx, AV_LOG_ERROR,
               "Extended MVs unavailable in Simple Profile\n");
        return -1;
    }
    v->dquant =  get_bits(gb, 2); //common
    v->vstransform =  get_bits(gb, 1); //common

    v->res_transtab = get_bits(gb, 1);
    if (v->res_transtab)
    {
        av_log(avctx, AV_LOG_ERROR,
               "1 for reserved RES_TRANSTAB is forbidden\n");
        return -1;
    }

    v->overlap = get_bits(gb, 1); //common

    v->s.resync_marker = get_bits(gb, 1);
    v->rangered = get_bits(gb, 1);
    if (v->rangered && v->profile == PROFILE_SIMPLE)
    {
        av_log(avctx, AV_LOG_INFO,
               "RANGERED should be set to 0 in simple profile\n");
    }

    v->s.max_b_frames = avctx->max_b_frames = get_bits(gb, 3); //common
    v->quantizer_mode = get_bits(gb, 2); //common

    v->finterpflag = get_bits(gb, 1); //common
    v->res_rtm_flag = get_bits(gb, 1); //reserved
    if (!v->res_rtm_flag)
    {
//            av_log(avctx, AV_LOG_ERROR,
//                   "0 for reserved RES_RTM_FLAG is forbidden\n");
        av_log(avctx, AV_LOG_ERROR,
               "Old WMV3 version detected, only I-frames will be decoded\n");
        //return -1;
    }
    //TODO: figure out what they mean (always 0x402F)
    if(!v->res_fasttx) skip_bits(gb, 16);
    av_log(avctx, AV_LOG_DEBUG,
               "Profile %i:\nfrmrtq_postproc=%i, bitrtq_postproc=%i\n"
               "LoopFilter=%i, MultiRes=%i, FastUVMC=%i, Extended MV=%i\n"
               "Rangered=%i, VSTransform=%i, Overlap=%i, SyncMarker=%i\n"
               "DQuant=%i, Quantizer mode=%i, Max B frames=%i\n",
               v->profile, v->frmrtq_postproc, v->bitrtq_postproc,
               v->s.loop_filter, v->multires, v->fastuvmc, v->extended_mv,
               v->rangered, v->vstransform, v->overlap, v->s.resync_marker,
               v->dquant, v->quantizer_mode, avctx->max_b_frames
               );
    return 0;
}

static int decode_sequence_header_adv(VC1Context *v, GetBitContext *gb)
{
    v->res_rtm_flag = 1;
    v->level = get_bits(gb, 3);
    if(v->level >= 5)
    {
        av_log(v->s.avctx, AV_LOG_ERROR, "Reserved LEVEL %i\n",v->level);
    }
    v->chromaformat = get_bits(gb, 2);
    if (v->chromaformat != 1)
    {
        av_log(v->s.avctx, AV_LOG_ERROR,
               "Only 4:2:0 chroma format supported\n");
        return -1;
    }

    // (fps-2)/4 (->30)
    v->frmrtq_postproc = get_bits(gb, 3); //common
    // (bitrate-32kbps)/64kbps
    v->bitrtq_postproc = get_bits(gb, 5); //common
    v->postprocflag = get_bits(gb, 1); //common

    v->s.avctx->coded_width = (get_bits(gb, 12) + 1) << 1;
    v->s.avctx->coded_height = (get_bits(gb, 12) + 1) << 1;
    v->s.avctx->width = v->s.avctx->coded_width;
    v->s.avctx->height = v->s.avctx->coded_height;
    v->broadcast = get_bits1(gb);
    v->interlace = get_bits1(gb);
    v->tfcntrflag = get_bits1(gb);
    v->finterpflag = get_bits1(gb);
    get_bits1(gb); // reserved

    v->s.h_edge_pos = v->s.avctx->coded_width;
    v->s.v_edge_pos = v->s.avctx->coded_height;

    av_log(v->s.avctx, AV_LOG_DEBUG,
               "Advanced Profile level %i:\nfrmrtq_postproc=%i, bitrtq_postproc=%i\n"
               "LoopFilter=%i, ChromaFormat=%i, Pulldown=%i, Interlace: %i\n"
               "TFCTRflag=%i, FINTERPflag=%i\n",
               v->level, v->frmrtq_postproc, v->bitrtq_postproc,
               v->s.loop_filter, v->chromaformat, v->broadcast, v->interlace,
               v->tfcntrflag, v->finterpflag
               );

    v->psf = get_bits1(gb);
    if(v->psf) { //PsF, 6.1.13
        av_log(v->s.avctx, AV_LOG_ERROR, "Progressive Segmented Frame mode: not supported (yet)\n");
        return -1;
    }
    v->s.max_b_frames = v->s.avctx->max_b_frames = 7;
    if(get_bits1(gb)) { //Display Info - decoding is not affected by it
        int w, h, ar = 0;
        av_log(v->s.avctx, AV_LOG_DEBUG, "Display extended info:\n");
        v->s.avctx->width  = v->s.width  = w = get_bits(gb, 14) + 1;
        v->s.avctx->height = v->s.height = h = get_bits(gb, 14) + 1;
        av_log(v->s.avctx, AV_LOG_DEBUG, "Display dimensions: %ix%i\n", w, h);
        if(get_bits1(gb))
            ar = get_bits(gb, 4);
        if(ar && ar < 14){
            v->s.avctx->sample_aspect_ratio = ff_vc1_pixel_aspect[ar];
        }else if(ar == 15){
            w = get_bits(gb, 8);
            h = get_bits(gb, 8);
            v->s.avctx->sample_aspect_ratio = (AVRational){w, h};
        }

        if(get_bits1(gb)){ //framerate stuff
            if(get_bits1(gb)) {
                v->s.avctx->time_base.num = 32;