FFmpeg/libavcodec/vp3.c

/*
 *
 * Copyright (C) 2003 the ffmpeg project
 *
 * This library 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 of the License, or (at your option) any later version.
 *
 * This library 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 this library; if not, write to the Free Software
 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
 *
 * VP3 Video Decoder by Mike Melanson (melanson@pcisys.net)
 *
 */

/**
 * @file vp3.c
 * On2 VP3 Video Decoder
 */

#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <unistd.h>

#include "common.h"
#include "avcodec.h"
#include "dsputil.h"
#include "mpegvideo.h"
#include "dsputil.h"
#include "bswap.h"

#include "vp3data.h"

#define FRAGMENT_PIXELS 8

/* 
 * Debugging Variables
 * 
 * Define one or more of the following compile-time variables to 1 to obtain
 * elaborate information about certain aspects of the decoding process.
 *
 * DEBUG_VP3: high-level decoding flow
 * DEBUG_INIT: initialization parameters
 * DEBUG_DEQUANTIZERS: display how the dequanization tables are built
 * DEBUG_BLOCK_CODING: unpacking the superblock/macroblock/fragment coding
 * DEBUG_MODES: unpacking the coding modes for individual fragments
 * DEBUG_VECTORS: display the motion vectors
 * DEBUG_TOKEN: display exhaustive information about each DCT token
 * DEBUG_VLC: display the VLCs as they are extracted from the stream
 * DEBUG_DC_PRED: display the process of reversing DC prediction
 * DEBUG_IDCT: show every detail of the IDCT process
 */

#define DEBUG_VP3 0
#define DEBUG_INIT 0
#define DEBUG_DEQUANTIZERS 0
#define DEBUG_BLOCK_CODING 0
#define DEBUG_MODES 0
#define DEBUG_VECTORS 0
#define DEBUG_TOKEN 0
#define DEBUG_VLC 0
#define DEBUG_DC_PRED 0
#define DEBUG_IDCT 0

#if DEBUG_VP3
#define debug_vp3 printf
#else
static inline void debug_vp3(const char *format, ...) { }
#endif

#if DEBUG_INIT
#define debug_init printf
#else
static inline void debug_init(const char *format, ...) { }
#endif

#if DEBUG_DEQUANTIZERS
#define debug_dequantizers printf 
#else
static inline void debug_dequantizers(const char *format, ...) { } 
#endif

#if DEBUG_BLOCK_CODING
#define debug_block_coding printf 
#else
static inline void debug_block_coding(const char *format, ...) { } 
#endif

#if DEBUG_MODES
#define debug_modes printf 
#else
static inline void debug_modes(const char *format, ...) { } 
#endif

#if DEBUG_VECTORS
#define debug_vectors printf 
#else
static inline void debug_vectors(const char *format, ...) { } 
#endif

#if DEBUG_TOKEN 
#define debug_token printf 
#else
static inline void debug_token(const char *format, ...) { } 
#endif

#if DEBUG_VLC
#define debug_vlc printf 
#else
static inline void debug_vlc(const char *format, ...) { } 
#endif

#if DEBUG_DC_PRED
#define debug_dc_pred printf 
#else
static inline void debug_dc_pred(const char *format, ...) { } 
#endif

#if DEBUG_IDCT
#define debug_idct printf 
#else
static inline void debug_idct(const char *format, ...) { } 
#endif

typedef struct Vp3Fragment {
    DCTELEM coeffs[64];
    int coding_method;
    int coeff_count;
    int last_coeff;
    int motion_x;
    int motion_y;
    /* this indicates which ffmpeg put_pixels() function to use:
     * 00b = no halfpel, 01b = x halfpel, 10b = y halfpel, 11b = both halfpel */
    int motion_halfpel_index;
    /* address of first pixel taking into account which plane the fragment
     * lives on as well as the plane stride */
    int first_pixel;
    /* this is the macroblock that the fragment belongs to */
    int macroblock;
} Vp3Fragment;

#define SB_NOT_CODED        0
#define SB_PARTIALLY_CODED  1
#define SB_FULLY_CODED      2

#define MODE_INTER_NO_MV      0
#define MODE_INTRA            1
#define MODE_INTER_PLUS_MV    2
#define MODE_INTER_LAST_MV    3
#define MODE_INTER_PRIOR_LAST 4
#define MODE_USING_GOLDEN     5
#define MODE_GOLDEN_MV        6
#define MODE_INTER_FOURMV     7
#define CODING_MODE_COUNT     8

/* special internal mode */
#define MODE_COPY             8

/* There are 6 preset schemes, plus a free-form scheme */
static int ModeAlphabet[7][CODING_MODE_COUNT] =
{
    /* this is the custom scheme */
    { 0, 0, 0, 0, 0, 0, 0, 0 },

    /* scheme 1: Last motion vector dominates */
    {    MODE_INTER_LAST_MV,    MODE_INTER_PRIOR_LAST,  
         MODE_INTER_PLUS_MV,    MODE_INTER_NO_MV,
         MODE_INTRA,            MODE_USING_GOLDEN,      
         MODE_GOLDEN_MV,        MODE_INTER_FOURMV },

    /* scheme 2 */
    {    MODE_INTER_LAST_MV,    MODE_INTER_PRIOR_LAST,  
         MODE_INTER_NO_MV,      MODE_INTER_PLUS_MV,
         MODE_INTRA,            MODE_USING_GOLDEN,      
         MODE_GOLDEN_MV,        MODE_INTER_FOURMV },

    /* scheme 3 */
    {    MODE_INTER_LAST_MV,    MODE_INTER_PLUS_MV,     
         MODE_INTER_PRIOR_LAST, MODE_INTER_NO_MV,
         MODE_INTRA,            MODE_USING_GOLDEN,      
         MODE_GOLDEN_MV,        MODE_INTER_FOURMV },

    /* scheme 4 */
    {    MODE_INTER_LAST_MV,    MODE_INTER_PLUS_MV,     
         MODE_INTER_NO_MV,      MODE_INTER_PRIOR_LAST,
         MODE_INTRA,            MODE_USING_GOLDEN,      
         MODE_GOLDEN_MV,        MODE_INTER_FOURMV },

    /* scheme 5: No motion vector dominates */
    {    MODE_INTER_NO_MV,      MODE_INTER_LAST_MV,     
         MODE_INTER_PRIOR_LAST, MODE_INTER_PLUS_MV,
         MODE_INTRA,            MODE_USING_GOLDEN,      
         MODE_GOLDEN_MV,        MODE_INTER_FOURMV },

    /* scheme 6 */
    {    MODE_INTER_NO_MV,      MODE_USING_GOLDEN,      
         MODE_INTER_LAST_MV,    MODE_INTER_PRIOR_LAST,
         MODE_INTER_PLUS_MV,    MODE_INTRA,             
         MODE_GOLDEN_MV,        MODE_INTER_FOURMV },

};

#define MIN_DEQUANT_VAL 2

typedef struct Vp3DecodeContext {
    AVCodecContext *avctx;
    int width, height;
    AVFrame golden_frame;
    AVFrame last_frame;
    AVFrame current_frame;
    int keyframe;
    DSPContext dsp;

    int quality_index;
    int last_quality_index;

    int superblock_count;
    int superblock_width;
    int superblock_height;
    int u_superblock_start;
    int v_superblock_start;
    unsigned char *superblock_coding;

    int macroblock_count;
    int macroblock_width;
    int macroblock_height;

    int fragment_count;
    int fragment_width;
    int fragment_height;

    Vp3Fragment *all_fragments;
    int u_fragment_start;
    int v_fragment_start;

    /* this is a list of indices into the all_fragments array indicating
     * which of the fragments are coded */
    int *coded_fragment_list;
    int coded_fragment_list_index;
    int pixel_addresses_inited;

    VLC dc_vlc[16];
    VLC ac_vlc_1[16];
    VLC ac_vlc_2[16];
    VLC ac_vlc_3[16];
    VLC ac_vlc_4[16];

    int16_t intra_y_dequant[64];
    int16_t intra_c_dequant[64];
    int16_t inter_dequant[64];

    /* This table contains superblock_count * 16 entries. Each set of 16
     * numbers corresponds to the fragment indices 0..15 of the superblock.
     * An entry will be -1 to indicate that no entry corresponds to that
     * index. */
    int *superblock_fragments;

    /* This table contains superblock_count * 4 entries. Each set of 4
     * numbers corresponds to the macroblock indices 0..3 of the superblock.
     * An entry will be -1 to indicate that no entry corresponds to that
     * index. */
    int *superblock_macroblocks;

    /* This table contains macroblock_count * 6 entries. Each set of 6
     * numbers corresponds to the fragment indices 0..5 which comprise
     * the macroblock (4 Y fragments and 2 C fragments). */
    int *macroblock_fragments;
    /* This is an array of flags indicating whether a particular 
     * macroblock is coded. */
    unsigned char *macroblock_coded;

} Vp3DecodeContext;

/************************************************************************
 * VP3 specific functions
 ************************************************************************/

/*
 * This function sets up all of the various blocks mappings:
 * superblocks <-> fragments, macroblocks <-> fragments,
 * superblocks <-> macroblocks
 */
static void init_block_mapping(Vp3DecodeContext *s) 
{
    int i, j;
    signed int hilbert_walk_y[16];
    signed int hilbert_walk_c[16];
    signed int hilbert_walk_mb[4];

    int current_fragment = 0;
    int current_width = 0;
    int current_height = 0;
    int right_edge = 0;
    int bottom_edge = 0;
    int superblock_row_inc = 0;
    int *hilbert = NULL;
    int mapping_index = 0;

    int current_macroblock;
    int c_fragment;

    signed char travel_width[16] = {
         1,  1,  0, -1, 
         0,  0,  1,  0,
         1,  0,  1,  0,
         0, -1,  0,  1
    };

    signed char travel_height[16] = {
         0,  0,  1,  0,
         1,  1,  0, -1,
         0,  1,  0, -1,
        -1,  0, -1,  0
    };

    signed char travel_width_mb[4] = {
         1,  0,  1,  0
    };

    signed char travel_height_mb[4] = {
         0,  1,  0, -1
    };

    debug_vp3("  vp3: initialize block mapping tables\n");

    /* figure out hilbert pattern per these frame dimensions */
    hilbert_walk_y[0]  = 1;
    hilbert_walk_y[1]  = 1;
    hilbert_walk_y[2]  = s->fragment_width;
    hilbert_walk_y[3]  = -1;
    hilbert_walk_y[4]  = s->fragment_width;
    hilbert_walk_y[5]  = s->fragment_width;
    hilbert_walk_y[6]  = 1;
    hilbert_walk_y[7]  = -s->fragment_width;
    hilbert_walk_y[8]  = 1;
    hilbert_walk_y[9]  = s->fragment_width;
    hilbert_walk_y[10]  = 1;
    hilbert_walk_y[11] = -s->fragment_width;
    hilbert_walk_y[12] = -s->fragment_width;
    hilbert_walk_y[13] = -1;
    hilbert_walk_y[14] = -s->fragment_width;
    hilbert_walk_y[15] = 1;

    hilbert_walk_c[0]  = 1;
    hilbert_walk_c[1]  = 1;
    hilbert_walk_c[2]  = s->fragment_width / 2;
    hilbert_walk_c[3]  = -1;
    hilbert_walk_c[4]  = s->fragment_width / 2;
    hilbert_walk_c[5]  = s->fragment_width / 2;
    hilbert_walk_c[6]  = 1;
    hilbert_walk_c[7]  = -s->fragment_width / 2;
    hilbert_walk_c[8]  = 1;
    hilbert_walk_c[9]  = s->fragment_width / 2;
    hilbert_walk_c[10]  = 1;
    hilbert_walk_c[11] = -s->fragment_width / 2;
    hilbert_walk_c[12] = -s->fragment_width / 2;
    hilbert_walk_c[13] = -1;
    hilbert_walk_c[14] = -s->fragment_width / 2;
    hilbert_walk_c[15] = 1;

    hilbert_walk_mb[0] = 1;
    hilbert_walk_mb[1] = s->macroblock_width;
    hilbert_walk_mb[2] = 1;
    hilbert_walk_mb[3] = -s->macroblock_width;

    /* iterate through each superblock (all planes) and map the fragments */
    for (i = 0; i < s->superblock_count; i++) {
        debug_init("    superblock %d (u starts @ %d, v starts @ %d)\n",
            i, s->u_superblock_start, s->v_superblock_start);

        /* time to re-assign the limits? */
        if (i == 0) {

            /* start of Y superblocks */
            right_edge = s->fragment_width;
            bottom_edge = s->fragment_height;
            current_width = 0;
            current_height = 0;
            superblock_row_inc = 3 * s->fragment_width;
            hilbert = hilbert_walk_y;

            /* the first operation for this variable is to advance by 1 */
            current_fragment = -1;

        } else if (i == s->u_superblock_start) {

            /* start of U superblocks */
            right_edge = s->fragment_width / 2;
            bottom_edge = s->fragment_height / 2;
            current_width = 0;
            current_height = 0;
            superblock_row_inc = 3 * (s->fragment_width / 2);
            hilbert = hilbert_walk_c;

            /* the first operation for this variable is to advance by 1 */
            current_fragment = s->u_fragment_start - 1;

        } else if (i == s->v_superblock_start) {

            /* start of V superblocks */
            right_edge = s->fragment_width / 2;
            bottom_edge = s->fragment_height / 2;
            current_width = 0;
            current_height = 0;
            superblock_row_inc = 3 * (s->fragment_width / 2);
            hilbert = hilbert_walk_c;

            /* the first operation for this variable is to advance by 1 */
            current_fragment = s->v_fragment_start - 1;

        }

        if (current_width >= right_edge) {
            /* reset width and move to next superblock row */
            current_width = 0;
            current_height += 4;

            /* fragment is now at the start of a new superblock row */
            current_fragment += superblock_row_inc;
        }

        /* iterate through all 16 fragments in a superblock */
        for (j = 0; j < 16; j++) {
            current_fragment += hilbert[j];
            current_height += travel_height[j];

            /* check if the fragment is in bounds */
            if ((current_width <= right_edge) &&
                (current_height < bottom_edge)) {
                s->superblock_fragments[mapping_index] = current_fragment;
                debug_init("    mapping fragment %d to superblock %d, position %d\n", 
                    s->superblock_fragments[mapping_index], i, j);
            } else {
                s->superblock_fragments[mapping_index] = -1;
                debug_init("    superblock %d, position %d has no fragment\n", 
                    i, j);
            }

            current_width += travel_width[j];
            mapping_index++;
        }
    }

    /* initialize the superblock <-> macroblock mapping; iterate through
     * all of the Y plane superblocks to build this mapping */
    right_edge = s->macroblock_width;
    bottom_edge = s->macroblock_height;
    current_width = 0;
    current_height = 0;
    superblock_row_inc = s->macroblock_width;
    hilbert = hilbert_walk_mb;
    mapping_index = 0;
    current_macroblock = -1;
    for (i = 0; i < s->u_superblock_start; i++) {

        if (current_width >= right_edge) {
            /* reset width and move to next superblock row */
            current_width = 0;
            current_height += 2;

            /* macroblock is now at the start of a new superblock row */
            current_macroblock += superblock_row_inc;
        }

        /* iterate through each potential macroblock in the superblock */
        for (j = 0; j < 4; j++) {
            current_macroblock += hilbert_walk_mb[j];
            current_height += travel_height_mb[j];

            /* check if the macroblock is in bounds */
            if ((current_width <= right_edge) &&
                (current_height < bottom_edge)) {
                s->superblock_macroblocks[mapping_index] = current_macroblock;
                debug_init("    mapping macroblock %d to superblock %d, position %d\n",
                    s->superblock_macroblocks[mapping_index], i, j);
            } else {
                s->superblock_macroblocks[mapping_index] = -1;
                debug_init("    superblock %d, position %d has no macroblock\n",
                    i, j);
            }

            current_width += travel_width_mb[j];
            mapping_index++;
        }
    }

    /* initialize the macroblock <-> fragment mapping */
    current_fragment = 0;
    current_macroblock = 0;
    mapping_index = 0;
    for (i = 0; i < s->fragment_height; i += 2) {

        for (j = 0; j < s->fragment_width; j += 2) {

            debug_init("    macroblock %d contains fragments: ", current_macroblock);
            s->all_fragments[current_fragment].macroblock = current_macroblock;
            s->macroblock_fragments[mapping_index++] = current_fragment;
            debug_init("%d ", current_fragment);

            if (j + 1 < s->fragment_width) {
                s->all_fragments[current_fragment + 1].macroblock = current_macroblock;
                s->macroblock_fragments[mapping_index++] = current_fragment + 1;
                debug_init("%d ", current_fragment + 1);
            } else
                s->macroblock_fragments[mapping_index++] = -1;

            if (i + 1 < s->fragment_height) {
                s->all_fragments[current_fragment + s->fragment_width].macroblock = 
                    current_macroblock;
                s->macroblock_fragments[mapping_index++] = 
                    current_fragment + s->fragment_width;
                debug_init("%d ", current_fragment + s->fragment_width);
            } else
                s->macroblock_fragments[mapping_index++] = -1;

            if ((j + 1 < s->fragment_width) && (i + 1 < s->fragment_height)) {
                s->all_fragments[current_fragment + s->fragment_width + 1].macroblock = 
                    current_macroblock;
                s->macroblock_fragments[mapping_index++] = 
                    current_fragment + s->fragment_width + 1;
                debug_init("%d ", current_fragment + s->fragment_width + 1);
            } else
                s->macroblock_fragments[mapping_index++] = -1;

            /* C planes */
            c_fragment = s->u_fragment_start + 
                (i * s->fragment_width / 4) + (j / 2);
        s->all_fragments[c_fragment].macroblock = s->macroblock_count;
            s->macroblock_fragments[mapping_index++] = c_fragment;
            debug_init("%d ", c_fragment);

            c_fragment = s->v_fragment_start + 
                (i * s->fragment_width / 4) + (j / 2);
        s->all_fragments[c_fragment].macroblock = s->macroblock_count;
            s->macroblock_fragments[mapping_index++] = c_fragment;
            debug_init("%d ", c_fragment);

            debug_init("\n");

            if (j + 2 <= s->fragment_width)
                current_fragment += 2;
            else 
                current_fragment++;
            current_macroblock++;
        }

        current_fragment += s->fragment_width;
    }
}

/*
 * This function unpacks a single token (which should be in the range 0..31)
 * and returns a zero run (number of zero coefficients in current DCT matrix
 * before next non-zero coefficient), the next DCT coefficient, and the
 * number of consecutive, non-EOB'd DCT blocks to EOB.
 */
static void unpack_token(GetBitContext *gb, int token, int *zero_run,
                         DCTELEM *coeff, int *eob_run) 
{
    int sign;

    *zero_run = 0;
    *eob_run = 0;
    *coeff = 0;

    debug_token("    vp3 token %d: ", token);
    switch (token) {

    case 0:
        debug_token("DCT_EOB_TOKEN, EOB next block\n");
        *eob_run = 1;
        break;

    case 1:
        debug_token("DCT_EOB_PAIR_TOKEN, EOB next 2 blocks\n");
        *eob_run = 2;
        break;

    case 2:
        debug_token("DCT_EOB_TRIPLE_TOKEN, EOB next 3 blocks\n");
        *eob_run = 3;
        break;

    case 3:
        debug_token("DCT_REPEAT_RUN_TOKEN, ");
        *eob_run = get_bits(gb, 2) + 4;
        debug_token("EOB the next %d blocks\n", *eob_run);
        break;

    case 4:
        debug_token("DCT_REPEAT_RUN2_TOKEN, ");
        *eob_run = get_bits(gb, 3) + 8;
        debug_token("EOB the next %d blocks\n", *eob_run);
        break;

    case 5:
        debug_token("DCT_REPEAT_RUN3_TOKEN, ");
        *eob_run = get_bits(gb, 4) + 16;
        debug_token("EOB the next %d blocks\n", *eob_run);
        break;

    case 6:
        debug_token("DCT_REPEAT_RUN4_TOKEN, ");
        *eob_run = get_bits(gb, 12);
        debug_token("EOB the next %d blocks\n", *eob_run);
        break;

    case 7:
        debug_token("DCT_SHORT_ZRL_TOKEN, ");
        /* note that this token actually indicates that (3 extra bits) + 1 0s
         * should be output; this case specifies a run of (3 EBs) 0s and a
         * coefficient of 0. */
        *zero_run = get_bits(gb, 3);
        *coeff = 0;
        debug_token("skip the next %d positions in output matrix\n", *zero_run + 1);
        break;

    case 8:
        debug_token("DCT_ZRL_TOKEN, ");
        /* note that this token actually indicates that (6 extra bits) + 1 0s
         * should be output; this case specifies a run of (6 EBs) 0s and a
         * coefficient of 0. */
        *zero_run = get_bits(gb, 6);
        *coeff = 0;
        debug_token("skip the next %d positions in output matrix\n", *zero_run + 1);
        break;

    case 9:
        debug_token("ONE_TOKEN, output 1\n");
        *coeff = 1;
        break;

    case 10:
        debug_token("MINUS_ONE_TOKEN, output -1\n");
        *coeff = -1;
        break;

    case 11:
        debug_token("TWO_TOKEN, output 2\n");
        *coeff = 2;
        break;

    case 12:
        debug_token("MINUS_TWO_TOKEN, output -2\n");
        *coeff = -2;
        break;

    case 13:
    case 14:
    case 15:
    case 16:
        debug_token("LOW_VAL_TOKENS, ");
        if (get_bits(gb, 1))
            *coeff = -(3 + (token - 13));
        else
            *coeff = 3 + (token - 13);
        debug_token("output %d\n", *coeff);
        break;

    case 17:
        debug_token("DCT_VAL_CATEGORY3, ");
        sign = get_bits(gb, 1);
        *coeff = 7 + get_bits(gb, 1);
        if (sign)
            *coeff = -(*coeff);
        debug_token("output %d\n", *coeff);
        break;

    case 18:
        debug_token("DCT_VAL_CATEGORY4, ");
        sign = get_bits(gb, 1);
        *coeff = 9 + get_bits(gb, 2);
        if (sign)
            *coeff = -(*coeff);
        debug_token("output %d\n", *coeff);
        break;

    case 19:
        debug_token("DCT_VAL_CATEGORY5, ");
        sign = get_bits(gb, 1);
        *coeff = 13 + get_bits(gb, 3);
        if (sign)
            *coeff = -(*coeff);
        debug_token("output %d\n", *coeff);
        break;

    case 20:
        debug_token("DCT_VAL_CATEGORY6, ");
        sign = get_bits(gb, 1);
        *coeff = 21 + get_bits(gb, 4);
        if (sign)
            *coeff = -(*coeff);
        debug_token("output %d\n", *coeff);
        break;

    case 21:
        debug_token("DCT_VAL_CATEGORY7, ");
        sign = get_bits(gb, 1);
        *coeff = 37 + get_bits(gb, 5);
        if (sign)
            *coeff = -(*coeff);
        debug_token("output %d\n", *coeff);
        break;

    case 22:
        debug_token("DCT_VAL_CATEGORY8, ");
        sign = get_bits(gb, 1);
        *coeff = 69 + get_bits(gb, 9);
        if (sign)
            *coeff = -(*coeff);
        debug_token("output %d\n", *coeff);
        break;

    case 23:
    case 24:
    case 25:
    case 26:
    case 27:
        debug_token("DCT_RUN_CATEGORY1, ");
        *zero_run = token - 22;
        if (get_bits(gb, 1))
            *coeff = -1;
        else
            *coeff = 1;
        debug_token("output %d 0s, then %d\n", *zero_run, *coeff);
        break;

    case 28:
        debug_token("DCT_RUN_CATEGORY1B, ");
        if (get_bits(gb, 1))
            *coeff = -1;
        else
            *coeff = 1;
        *zero_run = 6 + get_bits(gb, 2);
        debug_token("output %d 0s, then %d\n", *zero_run, *coeff);
        break;

    case 29:
        debug_token("DCT_RUN_CATEGORY1C, ");
        if (get_bits(gb, 1))
            *coeff = -1;
        else
            *coeff = 1;
        *zero_run = 10 + get_bits(gb, 3);
        debug_token("output %d 0s, then %d\n", *zero_run, *coeff);
        break;

    case 30:
        debug_token("DCT_RUN_CATEGORY2, ");
        sign = get_bits(gb, 1);
        *coeff = 2 + get_bits(gb, 1);
        if (sign)
            *coeff = -(*coeff);
        *zero_run = 1;
        debug_token("output %d 0s, then %d\n", *zero_run, *coeff);
        break;

    case 31:
        debug_token("DCT_RUN_CATEGORY2, ");
        sign = get_bits(gb, 1);
        *coeff = 2 + get_bits(gb, 1);
        if (sign)
            *coeff = -(*coeff);
        *zero_run = 2 + get_bits(gb, 1);
        debug_token("output %d 0s, then %d\n", *zero_run, *coeff);
        break;

    default:
        printf ("  vp3: help! Got a bad token: %d > 31\n", token);
        break;

  }
}

/*
 * This function wipes out all of the fragment data.
 */
static void init_frame(Vp3DecodeContext *s, GetBitContext *gb)
{
    int i;

    /* zero out all of the fragment information */
    s->coded_fragment_list_index = 0;
    for (i = 0; i < s->fragment_count; i++) {
        memset(s->all_fragments[i].coeffs, 0, 64 * sizeof(DCTELEM));
        s->all_fragments[i].coeff_count = 0;
        s->all_fragments[i].last_coeff = 0;
    }
}

/*
 * This function sets of the dequantization tables used for a particular
 * frame.
 */
static void init_dequantizer(Vp3DecodeContext *s)
{

    int quality_scale = vp31_quality_threshold[s->quality_index];
    int dc_scale_factor = vp31_dc_scale_factor[s->quality_index];
    int i, j;

    debug_vp3("  vp3: initializing dequantization tables\n");

    /* 
     * Scale dequantizers:
     *
     *   quantizer * sf
     *   --------------
     *        100
     *
     * where sf = dc_scale_factor for DC quantizer
     *           or quality_scale for AC quantizer
     *
     * Then, saturate the result to a lower limit of MIN_DEQUANT_VAL.
     */
#define SCALER 1

    /* scale DC quantizers */
    s->intra_y_dequant[0] = vp31_intra_y_dequant[0] * dc_scale_factor / 100;
    if (s->intra_y_dequant[0] < MIN_DEQUANT_VAL * 2)
        s->intra_y_dequant[0] = MIN_DEQUANT_VAL * 2;
    s->intra_y_dequant[0] *= SCALER;

    s->intra_c_dequant[0] = vp31_intra_c_dequant[0] * dc_scale_factor / 100;
    if (s->intra_c_dequant[0] < MIN_DEQUANT_VAL * 2)
        s->intra_c_dequant[0] = MIN_DEQUANT_VAL * 2;
    s->intra_c_dequant[0] *= SCALER;

    s->inter_dequant[0] = vp31_inter_dequant[0] * dc_scale_factor / 100;
    if (s->inter_dequant[0] < MIN_DEQUANT_VAL * 4)
        s->inter_dequant[0] = MIN_DEQUANT_VAL * 4;
    s->inter_dequant[0] *= SCALER;

    /* scale AC quantizers, zigzag at the same time in preparation for
     * the dequantization phase */
    for (i = 1; i < 64; i++) {

        j = quant_index[i];

        s->intra_y_dequant[j] = vp31_intra_y_dequant[i] * quality_scale / 100;
        if (s->intra_y_dequant[j] < MIN_DEQUANT_VAL)
            s->intra_y_dequant[j] = MIN_DEQUANT_VAL;
        s->intra_y_dequant[j] *= SCALER;

        s->intra_c_dequant[j] = vp31_intra_c_dequant[i] * quality_scale / 100;
        if (s->intra_c_dequant[j] < MIN_DEQUANT_VAL)
            s->intra_c_dequant[j] = MIN_DEQUANT_VAL;
        s->intra_c_dequant[j] *= SCALER;

        s->inter_dequant[j] = vp31_inter_dequant[i] * quality_scale / 100;
        if (s->inter_dequant[j] < MIN_DEQUANT_VAL * 2)
            s->inter_dequant[j] = MIN_DEQUANT_VAL * 2;
        s->inter_dequant[j] *= SCALER;
    }

    /* print debug information as requested */
    debug_dequantizers("intra Y dequantizers:\n");
    for (i = 0; i < 8; i++) {
      for (j = i * 8; j < i * 8 + 8; j++) {
        debug_dequantizers(" %4d,", s->intra_y_dequant[j]);
      }
      debug_dequantizers("\n");
    }
    debug_dequantizers("\n");

    debug_dequantizers("intra C dequantizers:\n");
    for (i = 0; i < 8; i++) {
      for (j = i * 8; j < i * 8 + 8; j++) {
        debug_dequantizers(" %4d,", s->intra_c_dequant[j]);
      }
      debug_dequantizers("\n");
    }
    debug_dequantizers("\n");

    debug_dequantizers("interframe dequantizers:\n");
    for (i = 0; i < 8; i++) {
      for (j = i * 8; j < i * 8 + 8; j++) {
        debug_dequantizers(" %4d,", s->inter_dequant[j]);
      }
      debug_dequantizers("\n");
    }
    debug_dequantizers("\n");
}

/*
 * This function is used to fetch runs of 1s or 0s from the bitstream for
 * use in determining which superblocks are fully and partially coded.
 *
 *  Codeword                RunLength
 *  0                       1
 *  10x                     2-3
 *  110x                    4-5
 *  1110xx                  6-9
 *  11110xxx                10-17
 *  111110xxxx              18-33
 *  111111xxxxxxxxxxxx      34-4129
 */
static int get_superblock_run_length(GetBitContext *gb)
{

    if (get_bits(gb, 1) == 0)
        return 1;

    else if (get_bits(gb, 1) == 0)
        return (2 + get_bits(gb, 1));

    else if (get_bits(gb, 1) == 0)
        return (4 + get_bits(gb, 1));

    else if (get_bits(gb, 1) == 0)
        return (6 + get_bits(gb, 2));

    else if (get_bits(gb, 1) == 0)
        return (10 + get_bits(gb, 3));

    else if (get_bits(gb, 1) == 0)
        return (18 + get_bits(gb, 4));

    else
        return (34 + get_bits(gb, 12));

}

/*
 * This function is used to fetch runs of 1s or 0s from the bitstream for
 * use in determining which particular fragments are coded.
 *
 * Codeword                RunLength
 * 0x                      1-2
 * 10x                     3-4
 * 110x                    5-6
 * 1110xx                  7-10
 * 11110xx                 11-14
 * 11111xxxx               15-30
 */
static int get_fragment_run_length(GetBitContext *gb)
{

    if (get_bits(gb, 1) == 0)
        return (1 + get_bits(gb, 1));

    else if (get_bits(gb, 1) == 0)
        return (3 + get_bits(gb, 1));

    else if (get_bits(gb, 1) == 0)
        return (5 + get_bits(gb, 1));

    else if (get_bits(gb, 1) == 0)
        return (7 + get_bits(gb, 2));

    else if (get_bits(gb, 1) == 0)
        return (11 + get_bits(gb, 2));

    else
        return (15 + get_bits(gb, 4));

}

/*
 * This function decodes a VLC from the bitstream and returns a number
 * that ranges from 0..7. The number indicates which of the 8 coding
 * modes to use.
 *
 *  VLC       Number
 *  0            0
 *  10           1
 *  110          2
 *  1110         3
 *  11110        4
 *  111110       5
 *  1111110      6
 *  1111111      7
 *
 */
static int get_mode_code(GetBitContext *gb)
{

    if (get_bits(gb, 1) == 0)
        return 0;

    else if (get_bits(gb, 1) == 0)
        return 1;

    else if (get_bits(gb, 1) == 0)
        return 2;

    else if (get_bits(gb, 1) == 0)
        return 3;

    else if (get_bits(gb, 1) == 0)
        return 4;

    else if (get_bits(gb, 1) == 0)
        return 5;

    else if (get_bits(gb, 1) == 0)
        return 6;

    else
        return 7;

}

/*
 * This function extracts a motion vector from the bitstream using a VLC
 * scheme. 3 bits are fetched from the bitstream and 1 of 8 actions is
 * taken depending on the value on those 3 bits:
 *
 *  0: return 0
 *  1: return 1
 *  2: return -1
 *  3: if (next bit is 1) return -2, else return 2
 *  4: if (next bit is 1) return -3, else return 3
 *  5: return 4 + (next 2 bits), next bit is sign
 *  6: return 8 + (next 3 bits), next bit is sign
 *  7: return 16 + (next 4 bits), next bit is sign
 */
static int get_motion_vector_vlc(GetBitContext *gb)
{
    int bits;

    bits = get_bits(gb, 3);

    switch(bits) {

    case 0:
        bits = 0;
        break;

    case 1:
        bits = 1;
        break;

    case 2:
        bits = -1;
        break;

    case 3:
        if (get_bits(gb, 1) == 0)
            bits = 2;
        else
            bits = -2;
        break;

    case 4:
        if (get_bits(gb, 1) == 0)
            bits = 3;
        else
            bits = -3;
        break;

    case 5:
        bits = 4 + get_bits(gb, 2);
        if (get_bits(gb, 1) == 1)
            bits = -bits;
        break;

    case 6:
        bits = 8 + get_bits(gb, 3);
        if (get_bits(gb, 1) == 1)
            bits = -bits;
        break;

    case 7:
        bits = 16 + get_bits(gb, 4);
        if (get_bits(gb, 1) == 1)
            bits = -bits;
        break;

    }

    return bits;
}

/*
 * This function fetches a 5-bit number from the stream followed by
 * a sign and calls it a motion vector.
 */
static int get_motion_vector_fixed(GetBitContext *gb)
{

    int bits;

    bits = get_bits(gb, 5);

    if (get_bits(gb, 1) == 1)
        bits = -bits;

    return bits;
}

/*
 * This function unpacks all of the superblock/macroblock/fragment coding 
 * information from the bitstream.
 */
static void unpack_superblocks(Vp3DecodeContext *s, GetBitContext *gb)
{
    int bit = 0;
    int current_superblock = 0;
    int current_run = 0;
    int decode_fully_flags = 0;
    int decode_partial_blocks = 0;

    int i, j;
    int current_fragment;

    debug_vp3("  vp3: unpacking superblock coding\n");

    if (s->keyframe) {

        debug_vp3("    keyframe-- all superblocks are fully coded\n");
        memset(s->superblock_coding, SB_FULLY_CODED, s->superblock_count);

    } else {

        /* unpack the list of partially-coded superblocks */
        bit = get_bits(gb, 1);
        /* toggle the bit because as soon as the first run length is 
         * fetched the bit will be toggled again */
        bit ^= 1;
        while (current_superblock < s->superblock_count) {
            if (current_run == 0) {
                bit ^= 1;
                current_run = get_superblock_run_length(gb);
                debug_block_coding("      setting superblocks %d..%d to %s\n",
                    current_superblock,
                    current_superblock + current_run - 1,
                    (bit) ? "partially coded" : "not coded");

                /* if any of the superblocks are not partially coded, flag
                 * a boolean to decode the list of fully-coded superblocks */
                if (bit == 0)
                    decode_fully_flags = 1;
            } else {

                /* make a note of the fact that there are partially coded
                 * superblocks */
                decode_partial_blocks = 1;

            }
            s->superblock_coding[current_superblock++] = 
                (bit) ? SB_PARTIALLY_CODED : SB_NOT_CODED;
            current_run--;
        }

        /* unpack the list of fully coded superblocks if any of the blocks were
         * not marked as partially coded in the previous step */
        if (decode_fully_flags) {

            current_superblock = 0;
            current_run = 0;
            bit = get_bits(gb, 1);
            /* toggle the bit because as soon as the first run length is 
             * fetched the bit will be toggled again */
            bit ^= 1;
            while (current_superblock < s->superblock_count) {

                /* skip any superblocks already marked as partially coded */
                if (s->superblock_coding[current_superblock] == SB_NOT_CODED) {

                    if (current_run == 0) {
                        bit ^= 1;
                        current_run = get_superblock_run_length(gb);
                    }

                    debug_block_coding("      setting superblock %d to %s\n",
                        current_superblock,
                        (bit) ? "fully coded" : "not coded");
                    s->superblock_coding[current_superblock] = 
                        (bit) ? SB_FULLY_CODED : SB_NOT_CODED;
                    current_run--;
                }
                current_superblock++;
            }
        }

        /* if there were partial blocks, initialize bitstream for
         * unpacking fragment codings */
        if (decode_partial_blocks) {

            current_run = 0;
            bit = get_bits(gb, 1);
            /* toggle the bit because as soon as the first run length is 
             * fetched the bit will be toggled again */
            bit ^= 1;
        }
    }

    /* figure out which fragments are coded; iterate through each
     * superblock (all planes) */
    s->coded_fragment_list_index = 0;
    memset(s->macroblock_coded, 0, s->macroblock_count);
    for (i = 0; i < s->superblock_count; i++) {

        /* iterate through all 16 fragments in a superblock */
        for (j = 0; j < 16; j++) {

            /* if the fragment is in bounds, check its coding status */
            current_fragment = s->superblock_fragments[i * 16 + j];
            if (current_fragment != -1) {
                if (s->superblock_coding[i] == SB_NOT_CODED) {

                    /* copy all the fragments from the prior frame */
                    s->all_fragments[current_fragment].coding_method = 
                        MODE_COPY;

                } else if (s->superblock_coding[i] == SB_PARTIALLY_CODED) {

                    /* fragment may or may not be coded; this is the case
                     * that cares about the fragment coding runs */
                    if (current_run == 0) {
                        bit ^= 1;
                        current_run = get_fragment_run_length(gb);
                    }

                    if (bit) {
                        /* mode will be decoded in the next phase */
                        s->all_fragments[current_fragment].coding_method = 
                            MODE_INTER_NO_MV;
                        s->coded_fragment_list[s->coded_fragment_list_index++] = 
                            current_fragment;
                        s->macroblock_coded[s->all_fragments[current_fragment].macroblock] = 1;
                        debug_block_coding("      superblock %d is partially coded, fragment %d is coded\n",
                            i, current_fragment);
                    } else {
                        /* not coded; copy this fragment from the prior frame */
                        s->all_fragments[current_fragment].coding_method =
                            MODE_COPY;
                        debug_block_coding("      superblock %d is partially coded, fragment %d is not coded\n",
                            i, current_fragment);
                    }

                    current_run--;

                } else {

                    /* fragments are fully coded in this superblock; actual
                     * coding will be determined in next step */
                    s->all_fragments[current_fragment].coding_method = 
                        MODE_INTER_NO_MV;
                    s->coded_fragment_list[s->coded_fragment_list_index++] = 
                        current_fragment;
                    s->macroblock_coded[s->all_fragments[current_fragment].macroblock] = 1;
                    debug_block_coding("      superblock %d is fully coded, fragment %d is coded\n",
                        i, current_fragment);
                }
            }
        }
    }
}

/*
 * This function unpacks all the coding mode data for individual macroblocks
 * from the bitstream.
 */
static void unpack_modes(Vp3DecodeContext *s, GetBitContext *gb)
{
    int i, j, k;
    int scheme;
    int current_macroblock;
    int current_fragment;
    int coding_mode;

    debug_vp3("  vp3: unpacking encoding modes\n");

    if (s->keyframe) {
        debug_vp3("    keyframe-- all blocks are coded as INTRA\n");

        for (i = 0; i < s->fragment_count; i++)
            s->all_fragments[i].coding_method = MODE_INTRA;

    } else {

        /* fetch the mode coding scheme for this frame */
        scheme = get_bits(gb, 3);
        debug_modes("    using mode alphabet %d\n", scheme);

        /* is it a custom coding scheme? */
        if (scheme == 0) {
            debug_modes("    custom mode alphabet ahead:\n");
            for (i = 0; i < 8; i++)
                ModeAlphabet[0][i] = get_bits(gb, 3);
        }

        for (i = 0; i < 8; i++)
            debug_modes("      mode[%d][%d] = %d\n", scheme, i, 
                ModeAlphabet[scheme][i]);

        /* iterate through all of the macroblocks that contain 1 or more
         * coded fragments */
        for (i = 0; i < s->u_superblock_start; i++) {

            for (j = 0; j < 4; j++) {
                current_macroblock = s->superblock_macroblocks[i * 4 + j];
                if ((current_macroblock == -1) ||
                    (!s->macroblock_coded[current_macroblock]))
                    continue;

                /* mode 7 means get 3 bits for each coding mode */
                if (scheme == 7)
                    coding_mode = get_bits(gb, 3);
                else
                    coding_mode = ModeAlphabet[scheme][get_mode_code(gb)];

                for (k = 0; k < 6; k++) {
                    current_fragment = 
                        s->macroblock_fragments[current_macroblock * 6 + k];
                    if (s->all_fragments[current_fragment].coding_method != 
                        MODE_COPY)
                        s->all_fragments[current_fragment].coding_method =
                            coding_mode;
                }

                debug_modes("    coding method for macroblock starting @ fragment %d = %d\n",
                    s->macroblock_fragments[current_macroblock * 6], coding_mode);
            }
        }
    }
}

/*
 * This function adjusts the components of a motion vector for the halfpel
 * motion grid. c_plane indicates whether the vector applies to the U or V
 * plane. The function returns the halfpel function index to be used in
 * ffmpeg's put_pixels[]() array of functions.
 */
static inline int adjust_vector(int *x, int *y, int c_plane)
{
    int motion_halfpel_index = 0;
    int x_halfpel;
    int y_halfpel;

    if (!c_plane) {

        x_halfpel = *x & 1;
        motion_halfpel_index |= x_halfpel;
        if (*x >= 0)
            *x >>= 1;
        else
            *x = -( (-(*x) >> 1) + x_halfpel);

        y_halfpel = *y & 1;
        motion_halfpel_index |= (y_halfpel << 1);
        if (*y >= 0)
            *y >>= 1;
        else
            *y = -( (-(*y) >> 1) + y_halfpel);

    } else {

        x_halfpel = ((*x & 0x03) != 0);
        motion_halfpel_index |= x_halfpel;
        if (*x >= 0)
            *x >>= 2;
        else
            *x = -( (-(*x) >> 2) + x_halfpel);

        y_halfpel = ((*y & 0x03) != 0);
        motion_halfpel_index |= (y_halfpel << 1);
        if (*y >= 0)
            *y >>= 2;
        else
            *y = -( (-(*y) >> 2) + y_halfpel);

    }

    return motion_halfpel_index;
}

/*
 * This function unpacks all the motion vectors for the individual
 * macroblocks from the bitstream.
 */
static void unpack_vectors(Vp3DecodeContext *s, GetBitContext *gb)
{
    int i, j, k;
    int coding_mode;
    int motion_x[6];
    int motion_y[6];
    int last_motion_x = 0;
    int last_motion_y = 0;
    int prior_last_motion_x = 0;
    int prior_last_motion_y = 0;
    int current_macroblock;
    int current_fragment;

    debug_vp3("  vp3: unpacking motion vectors\n");

    if (s->keyframe) {

        debug_vp3("    keyframe-- there are no motion vectors\n");

    } else {

        memset(motion_x, 0, 6 * sizeof(int));
        memset(motion_y, 0, 6 * sizeof(int));

        /* coding mode 0 is the VLC scheme; 1 is the fixed code scheme */
        coding_mode = get_bits(gb, 1);
        debug_vectors("    using %s scheme for unpacking motion vectors\n",
            (coding_mode == 0) ? "VLC" : "fixed-length");

        /* iterate through all of the macroblocks that contain 1 or more
         * coded fragments */
        for (i = 0; i < s->u_superblock_start; i++) {

            for (j = 0; j < 4; j++) {
                current_macroblock = s->superblock_macroblocks[i * 4 + j];
                if ((current_macroblock == -1) ||
                    (!s->macroblock_coded[current_macroblock]))
                    continue;

                current_fragment = s->macroblock_fragments[current_macroblock * 6];
                switch (s->all_fragments[current_fragment].coding_method) {

                case MODE_INTER_PLUS_MV:
                case MODE_GOLDEN_MV:
                    /* all 6 fragments use the same motion vector */
                    if (coding_mode == 0) {
                        motion_x[0] = get_motion_vector_vlc(gb);
                        motion_y[0] = get_motion_vector_vlc(gb);
                    } else {
                        motion_x[0] = get_motion_vector_fixed(gb);
                        motion_y[0] = get_motion_vector_fixed(gb);
                    }
                    for (k = 1; k < 6; k++) {
                        motion_x[k] = motion_x[0];
                        motion_y[k] = motion_y[0];
                    }

                    /* vector maintenance, only on MODE_INTER_PLUS_MV */
                    if (s->all_fragments[current_fragment].coding_method ==
                        MODE_INTER_PLUS_MV) {
                        prior_last_motion_x = last_motion_x;
                        prior_last_motion_y = last_motion_y;
                        last_motion_x = motion_x[0];
                        last_motion_y = motion_y[0];
                    }
                    break;

                case MODE_INTER_FOURMV:
                    /* fetch 4 vectors from the bitstream, one for each
                     * Y fragment, then average for the C fragment vectors */
                    motion_x[4] = motion_y[4] = 0;
                    for (k = 0; k < 4; k++) {
                        if (coding_mode == 0) {
                            motion_x[k] = get_motion_vector_vlc(gb);
                            motion_y[k] = get_motion_vector_vlc(gb);
                        } else {
                            motion_x[k] = get_motion_vector_fixed(gb);
                            motion_y[k] = get_motion_vector_fixed(gb);
                        }
                        motion_x[4] += motion_x[k];
                        motion_y[4] += motion_y[k];
                    }

                    if (motion_x[4] >= 0) 
                        motion_x[4] = (motion_x[4] + 2) / 4;
                    else
                        motion_x[4] = (motion_x[4] - 2) / 4;
                    motion_x[5] = motion_x[4];

                    if (motion_y[4] >= 0) 
                        motion_y[4] = (motion_y[4] + 2) / 4;
                    else
                        motion_y[4] = (motion_y[4] - 2) / 4;
                    motion_y[5] = motion_y[4];

                    /* vector maintenance; vector[3] is treated as the
                     * last vector in this case */
                    prior_last_motion_x = last_motion_x;
                    prior_last_motion_y = last_motion_y;
                    last_motion_x = motion_x[3];
                    last_motion_y = motion_y[3];
                    break;

                case MODE_INTER_LAST_MV:
                    /* all 6 fragments use the last motion vector */
                    motion_x[0] = last_motion_x;
                    motion_y[0] = last_motion_y;
                    for (k = 1; k < 6; k++) {
                        motion_x[k] = motion_x[0];
                        motion_y[k] = motion_y[0];
                    }

                    /* no vector maintenance (last vector remains the
                     * last vector) */
                    break;

                case MODE_INTER_PRIOR_LAST:
                    /* all 6 fragments use the motion vector prior to the
                     * last motion vector */
                    motion_x[0] = prior_last_motion_x;
                    motion_y[0] = prior_last_motion_y;
                    for (k = 1; k < 6; k++) {
                        motion_x[k] = motion_x[0];
                        motion_y[k] = motion_y[0];
                    }

                    /* vector maintenance */
                    prior_last_motion_x = last_motion_x;
                    prior_last_motion_y = last_motion_y;
                    last_motion_x = motion_x[0];
                    last_motion_y = motion_y[0];
                    break;

                default:
                    /* covers intra, inter without MV, golden without MV */
                    memset(motion_x, 0, 6 * sizeof(int));
                    memset(motion_y, 0, 6 * sizeof(int));

                    /* no vector maintenance */
                    break;
                }

                /* assign the motion vectors to the correct fragments */
                debug_vectors("    vectors for macroblock starting @ fragment %d (coding method %d):\n",
                    current_fragment,
                    s->all_fragments[current_fragment].coding_method);
                for (k = 0; k < 6; k++) {
                    current_fragment = 
                        s->macroblock_fragments[current_macroblock * 6 + k];
                    s->all_fragments[current_fragment].motion_halfpel_index =
                        adjust_vector(&motion_x[k], &motion_y[k],
                        ((k == 4) || (k == 5)));
                    s->all_fragments[current_fragment].motion_x = motion_x[k];
                    s->all_fragments[current_fragment].motion_y = motion_y[k];
                    debug_vectors("    vector %d: fragment %d = (%d, %d), index %d\n",
                        k, current_fragment, motion_x[k], motion_y[k],
                        s->all_fragments[current_fragment].motion_halfpel_index);
                }
            }
        }
    }
}

/* 
 * This function is called by unpack_dct_coeffs() to extract the VLCs from
 * the bitstream. The VLCs encode tokens which are used to unpack DCT
 * data. This function unpacks all the VLCs for either the Y plane or both
 * C planes, and is called for DC coefficients or different AC coefficient
 * levels (since different coefficient types require different VLC tables.
 *
 * This function returns a residual eob run. E.g, if a particular token gave
 * instructions to EOB the next 5 fragments and there were only 2 fragments
 * left in the current fragment range, 3 would be returned so that it could
 * be passed into the next call to this same function.
 */
static int unpack_vlcs(Vp3DecodeContext *s, GetBitContext *gb,
                        VLC *table, int coeff_index,
                        int first_fragment, int last_fragment,
                        int eob_run)
{
    int i;
    int token;
    int zero_run;
    DCTELEM coeff;
    Vp3Fragment *fragment;

    for (i = first_fragment; i < last_fragment; i++) {

        fragment = &s->all_fragments[s->coded_fragment_list[i]];
        if (fragment->coeff_count > coeff_index)
            continue;

        if (!eob_run) {
            /* decode a VLC into a token */
            token = get_vlc2(gb, table->table, 5, 3);
            debug_vlc(" token = %2d, ", token);
            /* use the token to get a zero run, a coefficient, and an eob run */
            unpack_token(gb, token, &zero_run, &coeff, &eob_run);
        }

        if (!eob_run) {
            fragment->coeff_count += zero_run;
            if (fragment->coeff_count < 64)
                fragment->coeffs[fragment->coeff_count++] = coeff;
            debug_vlc(" fragment %d coeff = %d\n",
                s->coded_fragment_list[i], fragment->coeffs[coeff_index]);
        } else {
            fragment->last_coeff = fragment->coeff_count;
            fragment->coeff_count = 64;
            debug_vlc(" fragment %d eob with %d coefficients\n", 
                s->coded_fragment_list[i], fragment->last_coeff);
            eob_run--;
        }
    }

    return eob_run;
}

/*
 * This function unpacks all of the DCT coefficient data from the
 * bitstream.
 */
static void unpack_dct_coeffs(Vp3DecodeContext *s, GetBitContext *gb)
{
    int i;
    int dc_y_table;
    int dc_c_table;
    int ac_y_table;
    int ac_c_table;
    int residual_eob_run = 0;

    /* for the binary search */
    int left, middle, right, found;
    /* this indicates the first fragment of the color plane data */
    int plane_split = 0;

    debug_vp3("  vp3: unpacking DCT coefficients\n");

    /* find the plane split (the first color plane fragment) using a binary 
     * search; test the boundaries first */
    if (s->coded_fragment_list_index == 0)
        return;
    if (s->u_fragment_start <= s->coded_fragment_list[0])
        plane_split = 0;  /* this means no Y fragments */
    else if (s->coded_fragment_list[s->coded_fragment_list_index - 1] >
        s->u_fragment_start) {

        left = 0;
        right = s->coded_fragment_list_index - 1;
        found = 0;
        do {
            middle = (left + right + 1) / 2;
            if ((s->coded_fragment_list[middle] >= s->u_fragment_start) &&
                (s->coded_fragment_list[middle - 1] < s->u_fragment_start))
                found = 1;
            else if (s->coded_fragment_list[middle] < s->u_fragment_start)
                left = middle;
            else
                right = middle;
        } while (!found);

        plane_split = middle;
    }

    debug_vp3("  plane split @ index %d (fragment %d)\n", plane_split,
        s->coded_fragment_list[plane_split]);

    /* fetch the DC table indices */
    dc_y_table = get_bits(gb, 4);
    dc_c_table = get_bits(gb, 4);

    /* unpack the Y plane DC coefficients */
    debug_vp3("  vp3: unpacking Y plane DC coefficients using table %d\n",
        dc_y_table);
    residual_eob_run = unpack_vlcs(s, gb, &s->dc_vlc[dc_y_table], 0, 
        0, plane_split, residual_eob_run);

    /* unpack the C plane DC coefficients */
    debug_vp3("  vp3: unpacking C plane DC coefficients using table %d\n",
        dc_c_table);
    residual_eob_run = unpack_vlcs(s, gb, &s->dc_vlc[dc_c_table], 0,
        plane_split, s->coded_fragment_list_index, residual_eob_run);

    /* fetch the level 1 AC table indices */
    ac_y_table = get_bits(gb, 4);
    ac_c_table = get_bits(gb, 4);

    /* unpack the level 1 AC coefficients (coeffs 1-5) */
    for (i = 1; i <= 5; i++) {

        debug_vp3("  vp3: unpacking level %d Y plane AC coefficients using table %d\n",
            i, ac_y_table);
        residual_eob_run = unpack_vlcs(s, gb, &s->ac_vlc_1[ac_y_table], i, 
            0, plane_split, residual_eob_run);

        debug_vp3("  vp3: unpacking level %d C plane AC coefficients using table %d\n",
            i, ac_c_table);
        residual_eob_run = unpack_vlcs(s, gb, &s->ac_vlc_1[ac_c_table], i, 
            plane_split, s->coded_fragment_list_index, residual_eob_run);
    }

    /* unpack the level 2 AC coefficients (coeffs 6-14) */
    for (i = 6; i <= 14; i++) {

        debug_vp3("  vp3: unpacking level %d Y plane AC coefficients using table %d\n",
            i, ac_y_table);
        residual_eob_run = unpack_vlcs(s, gb, &s->ac_vlc_2[ac_y_table], i, 
            0, plane_split, residual_eob_run);

        debug_vp3("  vp3: unpacking level %d C plane AC coefficients using table %d\n",
            i, ac_c_table);
        residual_eob_run = unpack_vlcs(s, gb, &s->ac_vlc_2[ac_c_table], i, 
            plane_split, s->coded_fragment_list_index, residual_eob_run);
    }

    /* unpack the level 3 AC coefficients (coeffs 15-27) */
    for (i = 15; i <= 27; i++) {

        debug_vp3("  vp3: unpacking level %d Y plane AC coefficients using table %d\n",
            i, ac_y_table);
        residual_eob_run = unpack_vlcs(s, gb, &s->ac_vlc_3[ac_y_table], i, 
            0, plane_split, residual_eob_run);

        debug_vp3("  vp3: unpacking level %d C plane AC coefficients using table %d\n",
            i, ac_c_table);
        residual_eob_run = unpack_vlcs(s, gb, &s->ac_vlc_3[ac_c_table], i, 
            plane_split, s->coded_fragment_list_index, residual_eob_run);
    }

    /* unpack the level 4 AC coefficients (coeffs 28-63) */
    for (i = 28; i <= 63; i++) {

        debug_vp3("  vp3: unpacking level %d Y plane AC coefficients using table %d\n",
            i, ac_y_table);
        residual_eob_run = unpack_vlcs(s, gb, &s->ac_vlc_4[ac_y_table], i, 
            0, plane_split, residual_eob_run);

        debug_vp3("  vp3: unpacking level %d C plane AC coefficients using table %d\n",
            i, ac_c_table);
        residual_eob_run = unpack_vlcs(s, gb, &s->ac_vlc_4[ac_c_table], i, 
            plane_split, s->coded_fragment_list_index, residual_eob_run);
    }
}

/*
 * This function reverses the DC prediction for each coded fragment in
 * the frame. Much of this function is adapted directly from the original 
 * VP3 source code.
 */
#define COMPATIBLE_FRAME(x) \
  (compatible_frame[s->all_fragments[x].coding_method] == current_frame_type)
#define FRAME_CODED(x) (s->all_fragments[x].coding_method != MODE_COPY)
#define HIGHBITDUPPED(X) (((signed short) X)  >> 15)
static inline int iabs (int x) { return ((x < 0) ? -x : x); }

static void reverse_dc_prediction(Vp3DecodeContext *s,
                                  int first_fragment,
                                  int fragment_width,
                                  int fragment_height) 
{

#define PUL 8
#define PU 4
#define PUR 2
#define PL 1

    int x, y;
    int i = first_fragment;

    /*
     * Fragment prediction groups:
     *
     * 32222222226
     * 10000000004
     * 10000000004
     * 10000000004
     * 10000000004
     *
     * Note: Groups 5 and 7 do not exist as it would mean that the 
     * fragment's x coordinate is both 0 and (width - 1) at the same time.
     */
    int predictor_group;
    short predicted_dc;

    /* validity flags for the left, up-left, up, and up-right fragments */
    int fl, ful, fu, fur;

    /* DC values for the left, up-left, up, and up-right fragments */
    int vl, vul, vu, vur;

    /* indices for the left, up-left, up, and up-right fragments */
    int l, ul, u, ur;

    /* 
     * The 6 fields mean:
     *   0: up-left multiplier
     *   1: up multiplier
     *   2: up-right multiplier
     *   3: left multiplier
     *   4: mask
     *   5: right bit shift divisor (e.g., 7 means >>=7, a.k.a. div by 128)
     */
    int predictor_transform[16][6] = {
        {  0,  0,  0,  0,   0,  0 },
        {  0,  0,  0,  1,   0,  0 },        // PL
        {  0,  0,  1,  0,   0,  0 },        // PUR
        {  0,  0, 53, 75, 127,  7 },        // PUR|PL
        {  0,  1,  0,  0,   0,  0 },        // PU
        {  0,  1,  0,  1,   1,  1 },        // PU|PL
        {  0,  1,  0,  0,   0,  0 },        // PU|PUR
        {  0,  0, 53, 75, 127,  7 },        // PU|PUR|PL
        {  1,  0,  0,  0,   0,  0 },        // PUL
        {  0,  0,  0,  1,   0,  0 },        // PUL|PL
        {  1,  0,  1,  0,   1,  1 },        // PUL|PUR
        {  0,  0, 53, 75, 127,  7 },        // PUL|PUR|PL
        {  0,  1,  0,  0,   0,  0 },        // PUL|PU
        {-26, 29,  0, 29,  31,  5 },        // PUL|PU|PL
        {  3, 10,  3,  0,  15,  4 },        // PUL|PU|PUR
        {-26, 29,  0, 29,  31,  5 }         // PUL|PU|PUR|PL
    };

    /* This table shows which types of blocks can use other blocks for
     * prediction. For example, INTRA is the only mode in this table to
     * have a frame number of 0. That means INTRA blocks can only predict
     * from other INTRA blocks. There are 2 golden frame coding types; 
     * blocks encoding in these modes can only predict from other blocks
     * that were encoded with these 1 of these 2 modes. */
    unsigned char compatible_frame[8] = {
        1,    /* MODE_INTER_NO_MV */
        0,    /* MODE_INTRA */
        1,    /* MODE_INTER_PLUS_MV */
        1,    /* MODE_INTER_LAST_MV */
        1,    /* MODE_INTER_PRIOR_MV */
        2,    /* MODE_USING_GOLDEN */
        2,    /* MODE_GOLDEN_MV */
        1     /* MODE_INTER_FOUR_MV */
    };
    int current_frame_type;

    /* there is a last DC predictor for each of the 3 frame types */
    short last_dc[3];

    int transform = 0;

    debug_vp3("  vp3: reversing DC prediction\n");

    vul = vu = vur = vl = 0;
    last_dc[0] = last_dc[1] = last_dc[2] = 0;

    /* for each fragment row... */
    for (y = 0; y < fragment_height; y++) {

        /* for each fragment in a row... */
        for (x = 0; x < fragment_width; x++, i++) {

            /* reverse prediction if this block was coded */
            if (s->all_fragments[i].coding_method != MODE_COPY) {

                current_frame_type = 
                    compatible_frame[s->all_fragments[i].coding_method];
                predictor_group = (x == 0) + ((y == 0) << 1) +
                    ((x + 1 == fragment_width) << 2);
                debug_dc_pred(" frag %d: group %d, orig DC = %d, ",
                    i, predictor_group, s->all_fragments[i].coeffs[0]);

                switch (predictor_group) {

                case 0:
                    /* main body of fragments; consider all 4 possible
                     * fragments for prediction */

                    /* calculate the indices of the predicting fragments */
                    ul = i - fragment_width - 1;
                    u = i - fragment_width;
                    ur = i - fragment_width + 1;
                    l = i - 1;

                    /* fetch the DC values for the predicting fragments */
                    vul = s->all_fragments[ul].coeffs[0];
                    vu = s->all_fragments[u].coeffs[0];
                    vur = s->all_fragments[ur].coeffs[0];
                    vl = s->all_fragments[l].coeffs[0];

                    /* figure out which fragments are valid */
                    ful = FRAME_CODED(ul) && COMPATIBLE_FRAME(ul);
                    fu = FRAME_CODED(u) && COMPATIBLE_FRAME(u);
                    fur = FRAME_CODED(ur) && COMPATIBLE_FRAME(ur);
                    fl = FRAME_CODED(l) && COMPATIBLE_FRAME(l);

                    /* decide which predictor transform to use */
                    transform = (fl*PL) | (fu*PU) | (ful*PUL) | (fur*PUR);

                    break;

                case 1:
                    /* left column of fragments, not including top corner;
                     * only consider up and up-right fragments */

                    /* calculate the indices of the predicting fragments */
                    u = i - fragment_width;
                    ur = i - fragment_width + 1;

                    /* fetch the DC values for the predicting fragments */
                    vu = s->all_fragments[u].coeffs[0];
                    vur = s->all_fragments[ur].coeffs[0];

                    /* figure out which fragments are valid */
                    fur = FRAME_CODED(ur) && COMPATIBLE_FRAME(ur);
                    fu = FRAME_CODED(u) && COMPATIBLE_FRAME(u);

                    /* decide which predictor transform to use */
                    transform = (fu*PU) | (fur*PUR);

                    break;

                case 2:
                case 6:
                    /* top row of fragments, not including top-left frag;
                     * only consider the left fragment for prediction */

                    /* calculate the indices of the predicting fragments */
                    l = i - 1;

                    /* fetch the DC values for the predicting fragments */
                    vl = s->all_fragments[l].coeffs[0];

                    /* figure out which fragments are valid */
                    fl = FRAME_CODED(l) && COMPATIBLE_FRAME(l);

                    /* decide which predictor transform to use */
                    transform = (fl*PL);

                    break;

                case 3:
                    /* top-left fragment */

                    /* nothing to predict from in this case */
                    transform = 0;

                    break;

                case 4:
                    /* right column of fragments, not including top corner;
                     * consider up-left, up, and left fragments for
                     * prediction */

                    /* calculate the indices of the predicting fragments */
                    ul = i - fragment_width - 1;
                    u = i - fragment_width;
                    l = i - 1;

                    /* fetch the DC values for the predicting fragments */
                    vul = s->all_fragments[ul].coeffs[0];
                    vu = s->all_fragments[u].coeffs[0];
                    vl = s->all_fragments[l].coeffs[0];

                    /* figure out which fragments are valid */
                    ful = FRAME_CODED(ul) && COMPATIBLE_FRAME(ul);
                    fu = FRAME_CODED(u) && COMPATIBLE_FRAME(u);
                    fl = FRAME_CODED(l) && COMPATIBLE_FRAME(l);

                    /* decide which predictor transform to use */
                    transform = (fl*PL) | (fu*PU) | (ful*PUL);

                    break;

                }

                debug_dc_pred("transform = %d, ", transform);

                if (transform == 0) {

                    /* if there were no fragments to predict from, use last
                     * DC saved */
                    s->all_fragments[i].coeffs[0] += last_dc[current_frame_type];
                    debug_dc_pred("from last DC (%d) = %d\n", 
                        current_frame_type, s->all_fragments[i].coeffs[0]);

                } else {

                    /* apply the appropriate predictor transform */
                    predicted_dc =
                        (predictor_transform[transform][0] * vul) +
                        (predictor_transform[transform][1] * vu) +
                        (predictor_transform[transform][2] * vur) +
                        (predictor_transform[transform][3] * vl);

                    /* if there is a shift value in the transform, add
                     * the sign bit before the shift */
                    if (predictor_transform[transform][5] != 0) {
                        predicted_dc += ((predicted_dc >> 15) & 
                            predictor_transform[transform][4]);
                        predicted_dc >>= predictor_transform[transform][5];
                    }

                    /* check for outranging on the [ul u l] and
                     * [ul u ur l] predictors */
                    if ((transform == 13) || (transform == 15)) {
                        if (iabs(predicted_dc - vu) > 128)
                            predicted_dc = vu;
                        else if (iabs(predicted_dc - vl) > 128)
                            predicted_dc = vl;
                        else if (iabs(predicted_dc - vul) > 128)
                            predicted_dc = vul;
                    }

                    /* at long last, apply the predictor */
                    s->all_fragments[i].coeffs[0] += predicted_dc;
                    debug_dc_pred("from pred DC = %d\n", 
                    s->all_fragments[i].coeffs[0]);
                }

                /* save the DC */
                last_dc[current_frame_type] = s->all_fragments[i].coeffs[0];
            }
        }
    }
}

/*
 * This function performs the final rendering of each fragment's data
 * onto the output frame.
 */
static void render_fragments(Vp3DecodeContext *s,
                             int first_fragment,
                             int width,
                             int height,
                             int plane /* 0 = Y, 1 = U, 2 = V */) 
{
    int x, y;
    int m, n;
    int i = first_fragment;
    int j;
    int16_t *dequantizer;
    DCTELEM dequant_block[64];
    unsigned char *output_plane;
    unsigned char *last_plane;
    unsigned char *golden_plane;
    int stride;
    int motion_x, motion_y;
    int motion_x_limit, motion_y_limit;
    int motion_halfpel_index;
    unsigned char *motion_source;

    debug_vp3("  vp3: rendering final fragments for %s\n",
        (plane == 0) ? "Y plane" : (plane == 1) ? "U plane" : "V plane");

    /* set up plane-specific parameters */
    if (plane == 0) {
        dequantizer = s->intra_y_dequant;
        output_plane = s->current_frame.data[0];
        last_plane = s->last_frame.data[0];
        golden_plane = s->golden_frame.data[0];
        stride = -s->current_frame.linesize[0];
    } else if (plane == 1) {
        dequantizer = s->intra_c_dequant;
        output_plane = s->current_frame.data[1];
        last_plane = s->last_frame.data[1];
        golden_plane = s->golden_frame.data[1];
        stride = -s->current_frame.linesize[1];
    } else {
        dequantizer = s->intra_c_dequant;
        output_plane = s->current_frame.data[2];
        last_plane = s->last_frame.data[2];
        golden_plane = s->golden_frame.data[2];
        stride = -s->current_frame.linesize[2];
    }

    motion_x_limit = width - 8;
    motion_y_limit = height - 8;

    /* for each fragment row... */
    for (y = 0; y < height; y += 8) {

        /* for each fragment in a row... */
        for (x = 0; x < width; x += 8, i++) {

            /* transform if this block was coded */
            if (s->all_fragments[i].coding_method != MODE_COPY) {
//            if (s->all_fragments[i].coding_method == MODE_INTRA) {

                /* sort out the motion vector */
                motion_x = x + s->all_fragments[i].motion_x;
                motion_y = y + s->all_fragments[i].motion_y;
                motion_halfpel_index = s->all_fragments[i].motion_halfpel_index;

                if (motion_x < 0)
                    motion_x = 0;
                if (motion_y < 0)
                    motion_y = 0;
                if (motion_x > motion_x_limit)
                    motion_x = motion_x_limit;
                if (motion_y > motion_y_limit)
                    motion_y = motion_y_limit;

                /* first, take care of copying a block from either the
                 * previous or the golden frame */
                if ((s->all_fragments[i].coding_method == MODE_USING_GOLDEN) ||
                    (s->all_fragments[i].coding_method == MODE_GOLDEN_MV)) {

                    motion_source = golden_plane;
                    motion_source += motion_x;
                    motion_source += (motion_y * -stride);

                    s->dsp.put_pixels_tab[1][motion_halfpel_index](
                        output_plane + s->all_fragments[i].first_pixel,
                        motion_source,
                        stride, 8);

                } else 
                if (s->all_fragments[i].coding_method != MODE_INTRA) {

                    motion_source = last_plane;
                    motion_source += motion_x;
                    motion_source += (motion_y * -stride);

                    s->dsp.put_pixels_tab[1][motion_halfpel_index](
                        output_plane + s->all_fragments[i].first_pixel,
                        motion_source,
                        stride, 8);
                }

                /* dequantize the DCT coefficients */
                debug_idct("fragment %d, coding mode %d, DC = %d, dequant = %d:\n", 
                    i, s->all_fragments[i].coding_method, 
                    s->all_fragments[i].coeffs[0], dequantizer[0]);
                for (j = 0; j < 64; j++)
                    dequant_block[dequant_index[j]] =
                        s->all_fragments[i].coeffs[j] *
                        dequantizer[j];
                dequant_block[0] += 1024;

                debug_idct("dequantized block:\n");
                for (m = 0; m < 8; m++) {
                    for (n = 0; n < 8; n++) {
                        debug_idct(" %5d", dequant_block[m * 8 + n]);
                    }
                    debug_idct("\n");
                }
                debug_idct("\n");

                /* invert DCT and place in final output */

                if (s->all_fragments[i].coding_method == MODE_INTRA)
                    s->dsp.idct_put(
                        output_plane + s->all_fragments[i].first_pixel,
                        stride, dequant_block);
                else
//                    s->dsp.idct_add(
                    s->dsp.idct_put(
                        output_plane + s->all_fragments[i].first_pixel,
                        stride, dequant_block);

                debug_idct("block after idct_%s():\n",
                    (s->all_fragments[i].coding_method == MODE_INTRA)?
                    "put" : "add");
                for (m = 0; m < 8; m++) {
                    for (n = 0; n < 8; n++) {
                        debug_idct(" %3d", *(output_plane + 
                            s->all_fragments[i].first_pixel + (m * stride + n)));
                    }
                    debug_idct("\n");
                }
                debug_idct("\n");

            } else {

                /* copy directly from the previous frame */
                s->dsp.put_pixels_tab[1][0](
                    output_plane + s->all_fragments[i].first_pixel,
                    last_plane + s->all_fragments[i].first_pixel,
                    stride, 8);

            }
        }
    }

    emms_c();

}

/* 
 * This function computes the first pixel addresses for each fragment.
 * This function needs to be invoked after the first frame is allocated
 * so that it has access to the plane strides.
 */
static void vp3_calculate_pixel_addresses(Vp3DecodeContext *s) 
{

    int i, x, y;

    /* figure out the first pixel addresses for each of the fragments */
    /* Y plane */
    i = 0;
    for (y = s->fragment_height; y > 0; y--) {
        for (x = 0; x < s->fragment_width; x++) {
            s->all_fragments[i++].first_pixel = 
                s->golden_frame.linesize[0] * y * FRAGMENT_PIXELS -
                    s->golden_frame.linesize[0] +
                    x * FRAGMENT_PIXELS;
            debug_init("  fragment %d, first pixel @ %d\n", 
                i-1, s->all_fragments[i-1].first_pixel);
        }
    }

    /* U plane */
    i = s->u_fragment_start;
    for (y = s->fragment_height / 2; y > 0; y--) {
        for (x = 0; x < s->fragment_width / 2; x++) {
            s->all_fragments[i++].first_pixel = 
                s->golden_frame.linesize[1] * y * FRAGMENT_PIXELS -
                    s->golden_frame.linesize[1] +
                    x * FRAGMENT_PIXELS;
            debug_init("  fragment %d, first pixel @ %d\n", 
                i-1, s->all_fragments[i-1].first_pixel);
        }
    }

    /* V plane */
    i = s->v_fragment_start;
    for (y = s->fragment_height / 2; y > 0; y--) {
        for (x = 0; x < s->fragment_width / 2; x++) {
            s->all_fragments[i++].first_pixel = 
                s->golden_frame.linesize[2] * y * FRAGMENT_PIXELS -
                    s->golden_frame.linesize[2] +
                    x * FRAGMENT_PIXELS;
            debug_init("  fragment %d, first pixel @ %d\n", 
                i-1, s->all_fragments[i-1].first_pixel);
        }
    }
}

/*
 * This is the ffmpeg/libavcodec API init function.
 */
static int vp3_decode_init(AVCodecContext *avctx)
{
    Vp3DecodeContext *s = avctx->priv_data;
    int i;

    s->avctx = avctx;
    s->width = avctx->width;
    s->height = avctx->height;
    avctx->pix_fmt = PIX_FMT_YUV420P;
    avctx->has_b_frames = 0;
    dsputil_init(&s->dsp, avctx);

    /* initialize to an impossible value which will force a recalculation
     * in the first frame decode */
    s->quality_index = -1;

    s->superblock_width = (s->width + 31) / 32;
    s->superblock_height = (s->height + 31) / 32;
    s->superblock_count = s->superblock_width * s->superblock_height * 3 / 2;
    s->u_superblock_start = s->superblock_width * s->superblock_height;
    s->v_superblock_start = s->superblock_width * s->superblock_height * 5 / 4;
    s->superblock_coding = av_malloc(s->superblock_count);

    s->macroblock_width = (s->width + 15) / 16;
    s->macroblock_height = (s->height + 15) / 16;
    s->macroblock_count = s->macroblock_width * s->macroblock_height;

    s->fragment_width = s->width / FRAGMENT_PIXELS;
    s->fragment_height = s->height / FRAGMENT_PIXELS;

    /* fragment count covers all 8x8 blocks for all 3 planes */
    s->fragment_count = s->fragment_width * s->fragment_height * 3 / 2;
    s->u_fragment_start = s->fragment_width * s->fragment_height;
    s->v_fragment_start = s->fragment_width * s->fragment_height * 5 / 4;

    debug_init("  width: %d x %d\n", s->width, s->height);
    debug_init("  superblocks: %d x %d, %d total\n",
        s->superblock_width, s->superblock_height, s->superblock_count);
    debug_init("  macroblocks: %d x %d, %d total\n",
        s->macroblock_width, s->macroblock_height, s->macroblock_count);
    debug_init("  %d fragments, %d x %d, u starts @ %d, v starts @ %d\n",
        s->fragment_count,
        s->fragment_width,
        s->fragment_height,
        s->u_fragment_start,
        s->v_fragment_start);

    s->all_fragments = av_malloc(s->fragment_count * sizeof(Vp3Fragment));
    s->coded_fragment_list = av_malloc(s->fragment_count * sizeof(int));
    s->pixel_addresses_inited = 0;

    /* init VLC tables */
    for (i = 0; i < 16; i++) {

        /* Dc histograms */
        init_vlc(&s->dc_vlc[i], 5, 32,
            &dc_bias[i][0][1], 4, 2,
            &dc_bias[i][0][0], 4, 2);

        /* level 1 AC histograms */
        init_vlc(&s->ac_vlc_1[i], 5, 32,
            &ac_bias_0[i][0][1], 4, 2,
            &ac_bias_0[i][0][0], 4, 2);

        /* level 2 AC histograms */
        init_vlc(&s->ac_vlc_2[i], 5, 32,
            &ac_bias_1[i][0][1], 4, 2,
            &ac_bias_1[i][0][0], 4, 2);

        /* level 3 AC histograms */
        init_vlc(&s->ac_vlc_3[i], 5, 32,
            &ac_bias_2[i][0][1], 4, 2,
            &ac_bias_2[i][0][0], 4, 2);

        /* level 4 AC histograms */
        init_vlc(&s->ac_vlc_4[i], 5, 32,
            &ac_bias_3[i][0][1], 4, 2,
            &ac_bias_3[i][0][0], 4, 2);
    }

    /* build quantization table */
    for (i = 0; i < 64; i++)
        quant_index[dequant_index[i]] = i;

    /* work out the block mapping tables */
    s->superblock_fragments = av_malloc(s->superblock_count * 16 * sizeof(int));
    s->superblock_macroblocks = av_malloc(s->superblock_count * 4 * sizeof(int));
    s->macroblock_fragments = av_malloc(s->macroblock_count * 6 * sizeof(int));
    s->macroblock_coded = av_malloc(s->macroblock_count + 1);
    init_block_mapping(s);

    for (i = 0; i < 3; i++) {
        s->current_frame.data[i] = NULL;
        s->last_frame.data[i] = NULL;
        s->golden_frame.data[i] = NULL;
    }

    return 0;
}

/*
 * This is the ffmpeg/libavcodec API frame decode function.
 */
static int vp3_decode_frame(AVCodecContext *avctx, 
                            void *data, int *data_size,
                            uint8_t *buf, int buf_size)
{
    Vp3DecodeContext *s = avctx->priv_data;
    GetBitContext gb;
    static int counter = 0;

    *data_size = 0;

    init_get_bits(&gb, buf, buf_size * 8);

    s->keyframe = get_bits(&gb, 1);
    s->keyframe ^= 1;
    skip_bits(&gb, 1);
    s->last_quality_index = s->quality_index;
    s->quality_index = get_bits(&gb, 6);
    if (s->quality_index != s->last_quality_index)
        init_dequantizer(s);

    debug_vp3(" VP3 frame #%d: Q index = %d", counter, s->quality_index);
    counter++;

    if (s->keyframe) {
        /* release the previous golden frame and get a new one */
        if (s->golden_frame.data[0])
            avctx->release_buffer(avctx, &s->golden_frame);

        /* last frame, if allocated, is hereby invalidated */
        if (s->last_frame.data[0])
            avctx->release_buffer(avctx, &s->last_frame);

        s->golden_frame.reference = 0;
        if(avctx->get_buffer(avctx, &s->golden_frame) < 0) {
            printf("vp3: get_buffer() failed\n");
            return -1;
        }

        /* golden frame is also the current frame */
        memcpy(&s->current_frame, &s->golden_frame, sizeof(AVFrame));

        /* time to figure out pixel addresses? */
        if (!s->pixel_addresses_inited)
            vp3_calculate_pixel_addresses(s);

    } else {

        /* allocate a new current frame */
        s->current_frame.reference = 0;
        if(avctx->get_buffer(avctx, &s->current_frame) < 0) {
            printf("vp3: get_buffer() failed\n");
            return -1;
        }

    }

    if (s->keyframe) {
      debug_vp3(", keyframe\n");
      /* skip the other 2 header bytes for now */
      skip_bits(&gb, 16);
    } else
      debug_vp3("\n");

    init_frame(s, &gb);

    unpack_superblocks(s, &gb);
    unpack_modes(s, &gb);
    unpack_vectors(s, &gb);
    unpack_dct_coeffs(s, &gb);

    reverse_dc_prediction(s, 0, s->fragment_width, s->fragment_height);
    reverse_dc_prediction(s, s->u_fragment_start,
        s->fragment_width / 2, s->fragment_height / 2);
    reverse_dc_prediction(s, s->v_fragment_start,
        s->fragment_width / 2, s->fragment_height / 2);

    render_fragments(s, 0, s->width, s->height, 0);
    render_fragments(s, s->u_fragment_start, s->width / 2, s->height / 2, 1);
    render_fragments(s, s->v_fragment_start, s->width / 2, s->height / 2, 2);

    *data_size=sizeof(AVFrame);
    *(AVFrame*)data= s->current_frame;

    /* release the last frame, if it is allocated and if it is not the
     * golden frame */
    if ((s->last_frame.data[0]) &&
        (s->last_frame.data[0] != s->golden_frame.data[0]))
        avctx->release_buffer(avctx, &s->last_frame);

    /* shuffle frames (last = current) */
    memcpy(&s->last_frame, &s->current_frame, sizeof(AVFrame));

    return buf_size;
}

/*
 * This is the ffmpeg/libavcodec API module cleanup function.
 */
static int vp3_decode_end(AVCodecContext *avctx)
{
    Vp3DecodeContext *s = avctx->priv_data;

    av_free(s->all_fragments);
    av_free(s->coded_fragment_list);
    av_free(s->superblock_fragments);
    av_free(s->superblock_macroblocks);
    av_free(s->macroblock_fragments);
    av_free(s->macroblock_coded);

    /* release all frames */
    avctx->release_buffer(avctx, &s->golden_frame);
    avctx->release_buffer(avctx, &s->last_frame);
    avctx->release_buffer(avctx, &s->current_frame);

    return 0;
}

AVCodec vp3_decoder = {
    "vp3",
    CODEC_TYPE_VIDEO,
    CODEC_ID_VP3,
    sizeof(Vp3DecodeContext),
    vp3_decode_init,
    NULL,
    vp3_decode_end,
    vp3_decode_frame,
    0,
    NULL
};