vp3.c 102 KB
Newer Older
1
/*
2
 * Copyright (C) 2003-2004 the ffmpeg project
3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22
 *
 * 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
 *
 */

/**
 * @file vp3.c
 * On2 VP3 Video Decoder
23 24 25 26 27 28
 *
 * VP3 Video Decoder by Mike Melanson (mike at multimedia.cx)
 * For more information about the VP3 coding process, visit:
 *   http://multimedia.cx/
 *
 * Theora decoder by Alex Beregszaszi
29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50
 */

#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 "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.
 *
51
 * KEYFRAMES_ONLY: set this to 1 to only see keyframes (VP3 slideshow mode)
52 53 54 55 56 57 58 59 60 61 62 63
 * 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
 */

64 65
#define KEYFRAMES_ONLY 0

66 67 68 69 70 71 72 73 74 75 76 77
#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
78
#define debug_vp3(args...) av_log(NULL, AV_LOG_DEBUG, ## args)
79 80 81 82 83
#else
static inline void debug_vp3(const char *format, ...) { }
#endif

#if DEBUG_INIT
84
#define debug_init(args...) av_log(NULL, AV_LOG_DEBUG, ## args)
85 86 87 88 89
#else
static inline void debug_init(const char *format, ...) { }
#endif

#if DEBUG_DEQUANTIZERS
90
#define debug_dequantizers(args...) av_log(NULL, AV_LOG_DEBUG, ## args)
91 92 93 94 95
#else
static inline void debug_dequantizers(const char *format, ...) { } 
#endif

#if DEBUG_BLOCK_CODING
96
#define debug_block_coding(args...) av_log(NULL, AV_LOG_DEBUG, ## args)
97 98 99 100 101
#else
static inline void debug_block_coding(const char *format, ...) { } 
#endif

#if DEBUG_MODES
102
#define debug_modes(args...) av_log(NULL, AV_LOG_DEBUG, ## args) 
103 104 105 106 107
#else
static inline void debug_modes(const char *format, ...) { } 
#endif

#if DEBUG_VECTORS
108
#define debug_vectors(args...) av_log(NULL, AV_LOG_DEBUG, ## args)
109 110 111 112 113
#else
static inline void debug_vectors(const char *format, ...) { } 
#endif

#if DEBUG_TOKEN 
114
#define debug_token(args...) av_log(NULL, AV_LOG_DEBUG, ## args)
115 116 117 118 119
#else
static inline void debug_token(const char *format, ...) { } 
#endif

#if DEBUG_VLC
120
#define debug_vlc(args...) av_log(NULL, AV_LOG_DEBUG, ## args)
121 122 123 124 125
#else
static inline void debug_vlc(const char *format, ...) { } 
#endif

#if DEBUG_DC_PRED
126
#define debug_dc_pred(args...) av_log(NULL, AV_LOG_DEBUG, ## args)
127 128 129 130 131
#else
static inline void debug_dc_pred(const char *format, ...) { } 
#endif

#if DEBUG_IDCT
132
#define debug_idct(args...) av_log(NULL, AV_LOG_DEBUG, ## args)
133 134 135 136
#else
static inline void debug_idct(const char *format, ...) { } 
#endif

137 138 139 140 141 142 143
typedef struct Coeff {
    struct Coeff *next;
    DCTELEM coeff;
    uint8_t index;
} Coeff;

//FIXME split things out into their own arrays
144
typedef struct Vp3Fragment {
145
    Coeff *next_coeff;
146 147 148 149
    /* 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 */
150 151 152 153 154
    uint16_t macroblock;
    uint8_t coding_method;
    uint8_t coeff_count;
    int8_t motion_x;
    int8_t motion_y;
155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221
} 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;
222
    int theora, theora_tables;
223
    int version;
224 225 226 227 228 229
    int width, height;
    AVFrame golden_frame;
    AVFrame last_frame;
    AVFrame current_frame;
    int keyframe;
    DSPContext dsp;
230
    int flipped_image;
231 232 233 234 235 236 237

    int quality_index;
    int last_quality_index;

    int superblock_count;
    int superblock_width;
    int superblock_height;
238 239 240 241
    int y_superblock_width;
    int y_superblock_height;
    int c_superblock_width;
    int c_superblock_height;
242 243 244 245 246 247 248 249 250 251 252 253 254
    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;
255 256
    Coeff *coeffs;
    Coeff *next_coeff;
257 258
    int u_fragment_start;
    int v_fragment_start;
259
    
260 261
    ScanTable scantable;
    
262 263
    /* tables */
    uint16_t coded_dc_scale_factor[64];
264
    uint32_t coded_ac_scale_factor[64];
265 266 267
    uint16_t coded_intra_y_dequant[64];
    uint16_t coded_intra_c_dequant[64];
    uint16_t coded_inter_dequant[64];
268 269 270 271 272 273 274 275 276 277 278 279 280

    /* 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];

281 282 283 284 285
    VLC superblock_run_length_vlc;
    VLC fragment_run_length_vlc;
    VLC mode_code_vlc;
    VLC motion_vector_vlc;

286 287 288 289 290
    /* these arrays need to be on 16-byte boundaries since SSE2 operations
     * index into them */
    int16_t __align16 intra_y_dequant[64];
    int16_t __align16 intra_c_dequant[64];
    int16_t __align16 inter_dequant[64];
291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307

    /* 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;
308 309
    /* This is an array that indicates how a particular macroblock 
     * is coded. */
310
    unsigned char *macroblock_coding;
311

312 313 314 315 316
    int first_coded_y_fragment;
    int first_coded_c_fragment;
    int last_coded_y_fragment;
    int last_coded_c_fragment;

Michael Niedermayer's avatar
Michael Niedermayer committed
317
    uint8_t edge_emu_buffer[9*2048]; //FIXME dynamic alloc
Michael Niedermayer's avatar
Michael Niedermayer committed
318
    uint8_t qscale_table[2048]; //FIXME dynamic alloc (width+15)/16
319

320 321 322 323 324 325 326 327 328
    /* Huffman decode */
    int hti;
    unsigned int hbits;
    int entries;
    int huff_code_size;
    uint16_t huffman_table[80][32][2];

    uint32_t filter_limit_values[64];
    int bounding_values_array[256];
329 330
} Vp3DecodeContext;

331 332 333
static int theora_decode_comments(AVCodecContext *avctx, GetBitContext gb);
static int theora_decode_tables(AVCodecContext *avctx, GetBitContext gb);

334 335 336 337 338 339 340 341
/************************************************************************
 * VP3 specific functions
 ************************************************************************/

/*
 * This function sets up all of the various blocks mappings:
 * superblocks <-> fragments, macroblocks <-> fragments,
 * superblocks <-> macroblocks
342 343
 *
 * Returns 0 is successful; returns 1 if *anything* went wrong.
344
 */
345
static int init_block_mapping(Vp3DecodeContext *s) 
346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438
{
    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;
439
            current_width = -1;
440
            current_height = 0;
441 442
            superblock_row_inc = 3 * s->fragment_width - 
                (s->y_superblock_width * 4 - s->fragment_width);
443 444 445 446 447 448 449 450 451 452
            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;
453
            current_width = -1;
454
            current_height = 0;
455 456
            superblock_row_inc = 3 * (s->fragment_width / 2) - 
                (s->c_superblock_width * 4 - s->fragment_width / 2);
457 458 459 460 461 462 463 464 465 466
            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;
467
            current_width = -1;
468
            current_height = 0;
469 470
            superblock_row_inc = 3 * (s->fragment_width / 2) - 
                (s->c_superblock_width * 4 - s->fragment_width / 2);
471 472 473 474 475 476 477
            hilbert = hilbert_walk_c;

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

        }

478
        if (current_width >= right_edge - 1) {
479
            /* reset width and move to next superblock row */
480
            current_width = -1;
481 482 483 484 485 486 487 488 489
            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];
490
            current_width += travel_width[j];
491 492 493
            current_height += travel_height[j];

            /* check if the fragment is in bounds */
494
            if ((current_width < right_edge) &&
495 496
                (current_height < bottom_edge)) {
                s->superblock_fragments[mapping_index] = current_fragment;
497 498 499
                debug_init("    mapping fragment %d to superblock %d, position %d (%d/%d x %d/%d)\n", 
                    s->superblock_fragments[mapping_index], i, j,
                    current_width, right_edge, current_height, bottom_edge);
500 501
            } else {
                s->superblock_fragments[mapping_index] = -1;
502 503 504
                debug_init("    superblock %d, position %d has no fragment (%d/%d x %d/%d)\n", 
                    i, j,
                    current_width, right_edge, current_height, bottom_edge);
505 506 507 508 509 510 511 512 513 514
            }

            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;
515
    current_width = -1;
516
    current_height = 0;
517 518
    superblock_row_inc = s->macroblock_width -
        (s->y_superblock_width * 2 - s->macroblock_width);;
519 520 521 522 523
    hilbert = hilbert_walk_mb;
    mapping_index = 0;
    current_macroblock = -1;
    for (i = 0; i < s->u_superblock_start; i++) {

524
        if (current_width >= right_edge - 1) {
525
            /* reset width and move to next superblock row */
526
            current_width = -1;
527 528 529 530 531 532 533 534 535
            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];
536
            current_width += travel_width_mb[j];
537 538 539
            current_height += travel_height_mb[j];

            /* check if the macroblock is in bounds */
540
            if ((current_width < right_edge) &&
541 542
                (current_height < bottom_edge)) {
                s->superblock_macroblocks[mapping_index] = current_macroblock;
543 544 545
                debug_init("    mapping macroblock %d to superblock %d, position %d (%d/%d x %d/%d)\n",
                    s->superblock_macroblocks[mapping_index], i, j,
                    current_width, right_edge, current_height, bottom_edge);
546 547
            } else {
                s->superblock_macroblocks[mapping_index] = -1;
548 549 550
                debug_init("    superblock %d, position %d has no macroblock (%d/%d x %d/%d)\n",
                    i, j,
                    current_width, right_edge, current_height, bottom_edge);
551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597
            }

            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);
598
            s->all_fragments[c_fragment].macroblock = s->macroblock_count;
599 600 601 602 603
            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);
604
            s->all_fragments[c_fragment].macroblock = s->macroblock_count;
605 606 607 608 609 610 611 612 613 614 615 616 617 618
            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;
    }
619 620

    return 0;  /* successful path out */
621 622 623 624 625 626 627 628 629 630 631 632 633
}

/*
 * 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++) {
        s->all_fragments[i].coeff_count = 0;
634 635 636
        s->all_fragments[i].motion_x = 127;
        s->all_fragments[i].motion_y = 127;
        s->all_fragments[i].next_coeff= NULL;
637 638 639
        s->coeffs[i].index=
        s->coeffs[i].coeff=0;
        s->coeffs[i].next= NULL;
640 641 642 643
    }
}

/*
644
 * This function sets up the dequantization tables used for a particular
645 646 647 648 649
 * frame.
 */
static void init_dequantizer(Vp3DecodeContext *s)
{

650
    int ac_scale_factor = s->coded_ac_scale_factor[s->quality_index];
651
    int dc_scale_factor = s->coded_dc_scale_factor[s->quality_index];
652 653 654 655 656 657 658 659 660 661 662 663
    int i, j;

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

    /* 
     * Scale dequantizers:
     *
     *   quantizer * sf
     *   --------------
     *        100
     *
     * where sf = dc_scale_factor for DC quantizer
664
     *         or ac_scale_factor for AC quantizer
665 666 667
     *
     * Then, saturate the result to a lower limit of MIN_DEQUANT_VAL.
     */
668
#define SCALER 4
669 670

    /* scale DC quantizers */
671
    s->intra_y_dequant[0] = s->coded_intra_y_dequant[0] * dc_scale_factor / 100;
672 673 674 675
    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;

676
    s->intra_c_dequant[0] = s->coded_intra_c_dequant[0] * dc_scale_factor / 100;
677 678 679 680
    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;

681
    s->inter_dequant[0] = s->coded_inter_dequant[0] * dc_scale_factor / 100;
682 683 684 685 686 687 688
    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++) {
689 690
        int k= s->scantable.scantable[i];
        j = s->scantable.permutated[i];
691

692
        s->intra_y_dequant[j] = s->coded_intra_y_dequant[k] * ac_scale_factor / 100;
693 694 695 696
        if (s->intra_y_dequant[j] < MIN_DEQUANT_VAL)
            s->intra_y_dequant[j] = MIN_DEQUANT_VAL;
        s->intra_y_dequant[j] *= SCALER;

697
        s->intra_c_dequant[j] = s->coded_intra_c_dequant[k] * ac_scale_factor / 100;
698 699 700 701
        if (s->intra_c_dequant[j] < MIN_DEQUANT_VAL)
            s->intra_c_dequant[j] = MIN_DEQUANT_VAL;
        s->intra_c_dequant[j] *= SCALER;

702
        s->inter_dequant[j] = s->coded_inter_dequant[k] * ac_scale_factor / 100;
703 704 705 706
        if (s->inter_dequant[j] < MIN_DEQUANT_VAL * 2)
            s->inter_dequant[j] = MIN_DEQUANT_VAL * 2;
        s->inter_dequant[j] *= SCALER;
    }
Michael Niedermayer's avatar
Michael Niedermayer committed
707 708
    
    memset(s->qscale_table, (FFMAX(s->intra_y_dequant[1], s->intra_c_dequant[1])+8)/16, 512); //FIXME finetune
709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738

    /* 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");
}

739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760
/*
 * This function initializes the loop filter boundary limits if the frame's
 * quality index is different from the previous frame's.
 */
static void init_loop_filter(Vp3DecodeContext *s)
{
    int *bounding_values= s->bounding_values_array+127;
    int filter_limit;
    int x;

    filter_limit = s->filter_limit_values[s->quality_index];

    /* set up the bounding values */
    memset(s->bounding_values_array, 0, 256 * sizeof(int));
    for (x = 0; x < filter_limit; x++) {
        bounding_values[-x - filter_limit] = -filter_limit + x;
        bounding_values[-x] = -x;
        bounding_values[x] = x;
        bounding_values[x + filter_limit] = filter_limit - x;
    }
}

761 762 763 764
/*
 * This function unpacks all of the superblock/macroblock/fragment coding 
 * information from the bitstream.
 */
765
static int unpack_superblocks(Vp3DecodeContext *s, GetBitContext *gb)
766 767 768 769 770 771
{
    int bit = 0;
    int current_superblock = 0;
    int current_run = 0;
    int decode_fully_flags = 0;
    int decode_partial_blocks = 0;
772
    int first_c_fragment_seen;
773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791

    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) {
792
            if (current_run-- == 0) {
793
                bit ^= 1;
794
                current_run = get_vlc2(gb, 
795 796
                    s->superblock_run_length_vlc.table, 6, 2);
                if (current_run == 33)
797
                    current_run += get_bits(gb, 12);
798 799 800 801 802 803 804
                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 */
805
                if (bit == 0) {
806
                    decode_fully_flags = 1;
807
                } else {
808

809 810 811 812
                    /* make a note of the fact that there are partially coded
                     * superblocks */
                    decode_partial_blocks = 1;
                }
813
            }
814
            s->superblock_coding[current_superblock++] = bit;
815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831
        }

        /* 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) {

832
                    if (current_run-- == 0) {
833
                        bit ^= 1;
834
                        current_run = get_vlc2(gb, 
835 836
                            s->superblock_run_length_vlc.table, 6, 2);
                        if (current_run == 33)
837
                            current_run += get_bits(gb, 12);
838 839 840 841 842
                    }

                    debug_block_coding("      setting superblock %d to %s\n",
                        current_superblock,
                        (bit) ? "fully coded" : "not coded");
843
                    s->superblock_coding[current_superblock] = 2*bit;
844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863
                }
                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;
864
    s->next_coeff= s->coeffs + s->fragment_count;
865 866
    s->first_coded_y_fragment = s->first_coded_c_fragment = 0;
    s->last_coded_y_fragment = s->last_coded_c_fragment = -1;
867
    first_c_fragment_seen = 0;
868
    memset(s->macroblock_coding, MODE_COPY, s->macroblock_count);
869 870 871 872 873 874 875
    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];
876
            if (current_fragment >= s->fragment_count) {
877
                av_log(s->avctx, AV_LOG_ERROR, "  vp3:unpack_superblocks(): bad fragment number (%d >= %d)\n",
878 879 880
                    current_fragment, s->fragment_count);
                return 1;
            }
881 882 883 884 885 886 887 888 889 890 891
            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 */
892
                    if (current_run-- == 0) {
893
                        bit ^= 1;
894
                        current_run = get_vlc2(gb, 
895
                            s->fragment_run_length_vlc.table, 5, 2);
896 897 898
                    }

                    if (bit) {
899 900
                        /* default mode; actual mode will be decoded in 
                         * the next phase */
901 902
                        s->all_fragments[current_fragment].coding_method = 
                            MODE_INTER_NO_MV;
903
                        s->all_fragments[current_fragment].next_coeff= s->coeffs + current_fragment;
904
                        s->coded_fragment_list[s->coded_fragment_list_index] = 
905
                            current_fragment;
906
                        if ((current_fragment >= s->u_fragment_start) &&
907 908
                            (s->last_coded_y_fragment == -1) &&
                            (!first_c_fragment_seen)) {
909 910
                            s->first_coded_c_fragment = s->coded_fragment_list_index;
                            s->last_coded_y_fragment = s->first_coded_c_fragment - 1;
911
                            first_c_fragment_seen = 1;
912 913
                        }
                        s->coded_fragment_list_index++;
914
                        s->macroblock_coding[s->all_fragments[current_fragment].macroblock] = MODE_INTER_NO_MV;
915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930
                        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);
                    }

                } 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;
931
                    s->all_fragments[current_fragment].next_coeff= s->coeffs + current_fragment;
932
                    s->coded_fragment_list[s->coded_fragment_list_index] = 
933
                        current_fragment;
934
                    if ((current_fragment >= s->u_fragment_start) &&
935 936
                        (s->last_coded_y_fragment == -1) &&
                        (!first_c_fragment_seen)) {
937 938
                        s->first_coded_c_fragment = s->coded_fragment_list_index;
                        s->last_coded_y_fragment = s->first_coded_c_fragment - 1;
939
                        first_c_fragment_seen = 1;
940 941
                    }
                    s->coded_fragment_list_index++;
942
                    s->macroblock_coding[s->all_fragments[current_fragment].macroblock] = MODE_INTER_NO_MV;
943 944 945 946 947 948
                    debug_block_coding("      superblock %d is fully coded, fragment %d is coded\n",
                        i, current_fragment);
                }
            }
        }
    }
949

950 951
    if (!first_c_fragment_seen)
        /* only Y fragments coded in this frame */
952
        s->last_coded_y_fragment = s->coded_fragment_list_index - 1;
953
    else 
954
        /* end the list of coded C fragments */
955
        s->last_coded_c_fragment = s->coded_fragment_list_index - 1;
956

957 958 959 960 961 962
    debug_block_coding("    %d total coded fragments, y: %d -> %d, c: %d -> %d\n",
        s->coded_fragment_list_index,
        s->first_coded_y_fragment,
        s->last_coded_y_fragment,
        s->first_coded_c_fragment,
        s->last_coded_c_fragment);
963 964

    return 0;
965 966 967 968 969 970
}

/*
 * This function unpacks all the coding mode data for individual macroblocks
 * from the bitstream.
 */
971
static int unpack_modes(Vp3DecodeContext *s, GetBitContext *gb)
972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996
{
    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++)
997
                ModeAlphabet[scheme][get_bits(gb, 3)] = i;
998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010
        }

        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) ||
1011
                    (s->macroblock_coding[current_macroblock] == MODE_COPY))
1012
                    continue;
1013
                if (current_macroblock >= s->macroblock_count) {
1014
                    av_log(s->avctx, AV_LOG_ERROR, "  vp3:unpack_modes(): bad macroblock number (%d >= %d)\n",
1015 1016 1017
                        current_macroblock, s->macroblock_count);
                    return 1;
                }
1018 1019 1020 1021 1022

                /* mode 7 means get 3 bits for each coding mode */
                if (scheme == 7)
                    coding_mode = get_bits(gb, 3);
                else
1023 1024
                    coding_mode = ModeAlphabet[scheme]
                        [get_vlc2(gb, s->mode_code_vlc.table, 3, 3)];
1025

1026
                s->macroblock_coding[current_macroblock] = coding_mode;
1027 1028 1029
                for (k = 0; k < 6; k++) {
                    current_fragment = 
                        s->macroblock_fragments[current_macroblock * 6 + k];
1030 1031 1032
                    if (current_fragment == -1)
                        continue;
                    if (current_fragment >= s->fragment_count) {
1033
                        av_log(s->avctx, AV_LOG_ERROR, "  vp3:unpack_modes(): bad fragment number (%d >= %d)\n",
1034 1035 1036
                            current_fragment, s->fragment_count);
                        return 1;
                    }
1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047
                    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);
            }
        }
    }
1048 1049

    return 0;
1050 1051
}

1052 1053 1054 1055
/*
 * This function unpacks all the motion vectors for the individual
 * macroblocks from the bitstream.
 */
1056
static int unpack_vectors(Vp3DecodeContext *s, GetBitContext *gb)
1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090
{
    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) ||
1091
                    (s->macroblock_coding[current_macroblock] == MODE_COPY))
1092
                    continue;
1093
                if (current_macroblock >= s->macroblock_count) {
1094
                    av_log(s->avctx, AV_LOG_ERROR, "  vp3:unpack_vectors(): bad macroblock number (%d >= %d)\n",
1095 1096 1097
                        current_macroblock, s->macroblock_count);
                    return 1;
                }
1098 1099

                current_fragment = s->macroblock_fragments[current_macroblock * 6];
1100
                if (current_fragment >= s->fragment_count) {
1101
                    av_log(s->avctx, AV_LOG_ERROR, "  vp3:unpack_vectors(): bad fragment number (%d >= %d\n",
1102 1103 1104
                        current_fragment, s->fragment_count);
                    return 1;
                }
1105
                switch (s->macroblock_coding[current_macroblock]) {
1106 1107 1108 1109 1110

                case MODE_INTER_PLUS_MV:
                case MODE_GOLDEN_MV:
                    /* all 6 fragments use the same motion vector */
                    if (coding_mode == 0) {
1111 1112
                        motion_x[0] = motion_vector_table[get_vlc2(gb, s->motion_vector_vlc.table, 6, 2)];
                        motion_y[0] = motion_vector_table[get_vlc2(gb, s->motion_vector_vlc.table, 6, 2)];
1113
                    } else {
1114 1115
                        motion_x[0] = fixed_motion_vector_table[get_bits(gb, 6)];
                        motion_y[0] = fixed_motion_vector_table[get_bits(gb, 6)];
1116
                    }
1117

1118 1119 1120 1121 1122 1123
                    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 */
1124
                    if (s->macroblock_coding[current_macroblock] ==
1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138
                        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) {
1139 1140
                            motion_x[k] = motion_vector_table[get_vlc2(gb, s->motion_vector_vlc.table, 6, 2)];
                            motion_y[k] = motion_vector_table[get_vlc2(gb, s->motion_vector_vlc.table, 6, 2)];
1141
                        } else {
1142 1143
                            motion_x[k] = fixed_motion_vector_table[get_bits(gb, 6)];
                            motion_y[k] = fixed_motion_vector_table[get_bits(gb, 6)];
1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197
                        }
                        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;
1198 1199 1200 1201 1202 1203 1204 1205

                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;
1206 1207 1208 1209 1210
                }

                /* assign the motion vectors to the correct fragments */
                debug_vectors("    vectors for macroblock starting @ fragment %d (coding method %d):\n",
                    current_fragment,
1211
                    s->macroblock_coding[current_macroblock]);
1212 1213 1214
                for (k = 0; k < 6; k++) {
                    current_fragment = 
                        s->macroblock_fragments[current_macroblock * 6 + k];
1215 1216 1217
                    if (current_fragment == -1)
                        continue;
                    if (current_fragment >= s->fragment_count) {
1218
                        av_log(s->avctx, AV_LOG_ERROR, "  vp3:unpack_vectors(): bad fragment number (%d >= %d)\n",
1219 1220 1221
                            current_fragment, s->fragment_count);
                        return 1;
                    }
1222
                    s->all_fragments[current_fragment].motion_x = motion_x[k];
1223
                    s->all_fragments[current_fragment].motion_y = motion_y[k];
1224 1225
                    debug_vectors("    vector %d: fragment %d = (%d, %d)\n",
                        k, current_fragment, motion_x[k], motion_y[k]);
1226 1227 1228 1229
                }
            }
        }
    }
1230 1231

    return 0;
1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252
}

/* 
 * 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;
1253 1254
    int zero_run = 0;
    DCTELEM coeff = 0;
1255
    Vp3Fragment *fragment;
1256
    uint8_t *perm= s->scantable.permutated;
1257
    int bits_to_get;
1258

1259
    if ((first_fragment >= s->fragment_count) ||
1260 1261
        (last_fragment >= s->fragment_count)) {

1262
        av_log(s->avctx, AV_LOG_ERROR, "  vp3:unpack_vlcs(): bad fragment number (%d -> %d ?)\n",
1263
            first_fragment, last_fragment);
1264
        return 0;
1265 1266
    }

1267
    for (i = first_fragment; i <= last_fragment; i++) {
1268 1269 1270 1271 1272 1273 1274 1275 1276 1277

        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 */
1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293
            if (token <= 6) {
                eob_run = eob_run_base[token];
                if (eob_run_get_bits[token])
                    eob_run += get_bits(gb, eob_run_get_bits[token]);
                coeff = zero_run = 0;
            } else {
                bits_to_get = coeff_get_bits[token];
                if (!bits_to_get)
                    coeff = coeff_tables[token][0];
                else
                    coeff = coeff_tables[token][get_bits(gb, bits_to_get)];

                zero_run = zero_run_base[token];
                if (zero_run_get_bits[token])
                    zero_run += get_bits(gb, zero_run_get_bits[token]);
            }
1294 1295 1296 1297
        }

        if (!eob_run) {
            fragment->coeff_count += zero_run;
1298 1299 1300 1301 1302 1303 1304
            if (fragment->coeff_count < 64){
                fragment->next_coeff->coeff= coeff;
                fragment->next_coeff->index= perm[fragment->coeff_count++]; //FIXME perm here already?
                fragment->next_coeff->next= s->next_coeff;
                s->next_coeff->next=NULL;
                fragment->next_coeff= s->next_coeff++;
            }
1305
            debug_vlc(" fragment %d coeff = %d\n",
1306
                s->coded_fragment_list[i], fragment->next_coeff[coeff_index]);
1307
        } else {
1308
            fragment->coeff_count |= 128;
1309
            debug_vlc(" fragment %d eob with %d coefficients\n", 
1310
                s->coded_fragment_list[i], fragment->coeff_count&127);
1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321
            eob_run--;
        }
    }

    return eob_run;
}

/*
 * This function unpacks all of the DCT coefficient data from the
 * bitstream.
 */
1322
static int unpack_dct_coeffs(Vp3DecodeContext *s, GetBitContext *gb)
1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338
{
    int i;
    int dc_y_table;
    int dc_c_table;
    int ac_y_table;
    int ac_c_table;
    int residual_eob_run = 0;

    /* 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, 
1339
        s->first_coded_y_fragment, s->last_coded_y_fragment, residual_eob_run);
1340 1341 1342 1343 1344

    /* 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,
1345
        s->first_coded_c_fragment, s->last_coded_c_fragment, residual_eob_run);
1346

1347
    /* fetch the AC table indices */
1348 1349 1350
    ac_y_table = get_bits(gb, 4);
    ac_c_table = get_bits(gb, 4);

1351
    /* unpack the group 1 AC coefficients (coeffs 1-5) */
1352 1353 1354 1355 1356
    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, 
1357
            s->first_coded_y_fragment, s->last_coded_y_fragment, residual_eob_run);
1358 1359 1360 1361

        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, 
1362
            s->first_coded_c_fragment, s->last_coded_c_fragment, residual_eob_run);
1363 1364
    }

1365
    /* unpack the group 2 AC coefficients (coeffs 6-14) */
1366 1367 1368 1369 1370
    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, 
1371
            s->first_coded_y_fragment, s->last_coded_y_fragment, residual_eob_run);
1372 1373 1374 1375

        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, 
1376
            s->first_coded_c_fragment, s->last_coded_c_fragment, residual_eob_run);
1377 1378
    }

1379
    /* unpack the group 3 AC coefficients (coeffs 15-27) */
1380 1381 1382 1383 1384
    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, 
1385
            s->first_coded_y_fragment, s->last_coded_y_fragment, residual_eob_run);
1386 1387 1388 1389

        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, 
1390
            s->first_coded_c_fragment, s->last_coded_c_fragment, residual_eob_run);
1391 1392
    }

1393
    /* unpack the group 4 AC coefficients (coeffs 28-63) */
1394 1395 1396 1397 1398
    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, 
1399
            s->first_coded_y_fragment, s->last_coded_y_fragment, residual_eob_run);
1400 1401 1402 1403

        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, 
1404
            s->first_coded_c_fragment, s->last_coded_c_fragment, residual_eob_run);
1405
    }
1406 1407

    return 0;
1408 1409 1410 1411 1412 1413 1414 1415 1416 1417
}

/*
 * 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)
1418
#define DC_COEFF(u) (s->coeffs[u].index ? 0 : s->coeffs[u].coeff) //FIXME do somethin to simplify this
1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528
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, ",
1529
                    i, predictor_group, DC_COEFF(i));
1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543

                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 */
1544 1545 1546 1547
                    vul = DC_COEFF(ul);
                    vu = DC_COEFF(u);
                    vur = DC_COEFF(ur);
                    vl = DC_COEFF(l);
1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568

                    /* 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 */
1569 1570
                    vu = DC_COEFF(u);
                    vur = DC_COEFF(ur);
1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589

                    /* 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 */
1590
                    vl = DC_COEFF(l);
1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618

                    /* 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 */
1619 1620 1621
                    vul = DC_COEFF(ul);
                    vu = DC_COEFF(u);
                    vl = DC_COEFF(l);
1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640

                    /* 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 */
1641
                    predicted_dc = last_dc[current_frame_type];
1642
                    debug_dc_pred("from last DC (%d) = %d\n", 
1643
                        current_frame_type, DC_COEFF(i));
1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673

                } 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;
                    }

                    debug_dc_pred("from pred DC = %d\n", 
1674
                    DC_COEFF(i));
1675 1676
                }

1677 1678 1679 1680 1681 1682 1683 1684
                /* at long last, apply the predictor */
                if(s->coeffs[i].index){
                    *s->next_coeff= s->coeffs[i];
                    s->coeffs[i].index=0;
                    s->coeffs[i].coeff=0;
                    s->coeffs[i].next= s->next_coeff++;
                }
                s->coeffs[i].coeff += predicted_dc;
1685
                /* save the DC */
1686 1687 1688 1689 1690 1691 1692
                last_dc[current_frame_type] = DC_COEFF(i);
                if(DC_COEFF(i) && !(s->all_fragments[i].coeff_count&127)){
                    s->all_fragments[i].coeff_count= 129;
//                    s->all_fragments[i].next_coeff= s->next_coeff;
                    s->coeffs[i].next= s->next_coeff;
                    (s->next_coeff++)->next=NULL;
                }
1693 1694 1695 1696 1697
            }
        }
    }
}

1698 1699 1700 1701 1702 1703

static void horizontal_filter(unsigned char *first_pixel, int stride,
    int *bounding_values);
static void vertical_filter(unsigned char *first_pixel, int stride,
    int *bounding_values);

1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727
/*
 * Perform the final rendering for a particular slice of data.
 * The slice number ranges from 0..(macroblock_height - 1).
 */
static void render_slice(Vp3DecodeContext *s, int slice)
{
    int x, y;
    int m, n;
    int i;  /* indicates current fragment */
    int16_t *dequantizer;
    DCTELEM __align16 block[64];
    unsigned char *output_plane;
    unsigned char *last_plane;
    unsigned char *golden_plane;
    int stride;
    int motion_x = 0xdeadbeef, motion_y = 0xdeadbeef;
    int upper_motion_limit, lower_motion_limit;
    int motion_halfpel_index;
    uint8_t *motion_source;
    int plane;
    int plane_width;
    int plane_height;
    int slice_height;
    int current_macroblock_entry = slice * s->macroblock_width * 6;
1728
    int fragment_width;
1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774 1775

    if (slice >= s->macroblock_height)
        return;

    for (plane = 0; plane < 3; plane++) {

        /* set up plane-specific parameters */
        if (plane == 0) {
            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];
            if (!s->flipped_image) stride = -stride;
            upper_motion_limit = 7 * s->current_frame.linesize[0];
            lower_motion_limit = s->height * s->current_frame.linesize[0] + s->width - 8;
            y = slice * FRAGMENT_PIXELS * 2;
            plane_width = s->width;
            plane_height = s->height;
            slice_height = y + FRAGMENT_PIXELS * 2;
            i = s->macroblock_fragments[current_macroblock_entry + 0];
        } else if (plane == 1) {
            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];
            if (!s->flipped_image) stride = -stride;
            upper_motion_limit = 7 * s->current_frame.linesize[1];
            lower_motion_limit = (s->height / 2) * s->current_frame.linesize[1] + (s->width / 2) - 8;
            y = slice * FRAGMENT_PIXELS;
            plane_width = s->width / 2;
            plane_height = s->height / 2;
            slice_height = y + FRAGMENT_PIXELS;
            i = s->macroblock_fragments[current_macroblock_entry + 4];
        } else {
            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];
            if (!s->flipped_image) stride = -stride;
            upper_motion_limit = 7 * s->current_frame.linesize[2];
            lower_motion_limit = (s->height / 2) * s->current_frame.linesize[2] + (s->width / 2) - 8;
            y = slice * FRAGMENT_PIXELS;
            plane_width = s->width / 2;
            plane_height = s->height / 2;
            slice_height = y + FRAGMENT_PIXELS;
            i = s->macroblock_fragments[current_macroblock_entry + 5];
        }
1776
        fragment_width = plane_width / FRAGMENT_PIXELS;
1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 1789 1790 1791 1792 1793 1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923
    
        if(ABS(stride) > 2048)
            return; //various tables are fixed size

        /* for each fragment row in the slice (both of them)... */
        for (; y < slice_height; y += 8) {

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

                if ((i < 0) || (i >= s->fragment_count)) {
                    av_log(s->avctx, AV_LOG_ERROR, "  vp3:render_slice(): bad fragment number (%d)\n", i);
                    return;
                }

                /* transform if this block was coded */
                if ((s->all_fragments[i].coding_method != MODE_COPY) &&
                    !((s->avctx->flags & CODEC_FLAG_GRAY) && plane)) {

                    if ((s->all_fragments[i].coding_method == MODE_USING_GOLDEN) ||
                        (s->all_fragments[i].coding_method == MODE_GOLDEN_MV))
                        motion_source= golden_plane;
                    else 
                        motion_source= last_plane;

                    motion_source += s->all_fragments[i].first_pixel;
                    motion_halfpel_index = 0;

                    /* sort out the motion vector if this fragment is coded
                     * using a motion vector method */
                    if ((s->all_fragments[i].coding_method > MODE_INTRA) &&
                        (s->all_fragments[i].coding_method != MODE_USING_GOLDEN)) {
                        int src_x, src_y;
                        motion_x = s->all_fragments[i].motion_x;
                        motion_y = s->all_fragments[i].motion_y;
                        if(plane){
                            motion_x= (motion_x>>1) | (motion_x&1);
                            motion_y= (motion_y>>1) | (motion_y&1);
                        }

                        src_x= (motion_x>>1) + x;
                        src_y= (motion_y>>1) + y;
                        if ((motion_x == 127) || (motion_y == 127))
                            av_log(s->avctx, AV_LOG_ERROR, " help! got invalid motion vector! (%X, %X)\n", motion_x, motion_y);

                        motion_halfpel_index = motion_x & 0x01;
                        motion_source += (motion_x >> 1);

                        motion_halfpel_index |= (motion_y & 0x01) << 1;
                        motion_source += ((motion_y >> 1) * stride);

                        if(src_x<0 || src_y<0 || src_x + 9 >= plane_width || src_y + 9 >= plane_height){
                            uint8_t *temp= s->edge_emu_buffer;
                            if(stride<0) temp -= 9*stride;
                            else temp += 9*stride;

                            ff_emulated_edge_mc(temp, motion_source, stride, 9, 9, src_x, src_y, plane_width, plane_height);
                            motion_source= temp;
                        }
                    }
                

                    /* first, take care of copying a block from either the
                     * previous or the golden frame */
                    if (s->all_fragments[i].coding_method != MODE_INTRA) {
                        /* Note, it is possible to implement all MC cases with 
                           put_no_rnd_pixels_l2 which would look more like the 
                           VP3 source but this would be slower as 
                           put_no_rnd_pixels_tab is better optimzed */
                        if(motion_halfpel_index != 3){
                            s->dsp.put_no_rnd_pixels_tab[1][motion_halfpel_index](
                                output_plane + s->all_fragments[i].first_pixel,
                                motion_source, stride, 8);
                        }else{
                            int d= (motion_x ^ motion_y)>>31; // d is 0 if motion_x and _y have the same sign, else -1
                            s->dsp.put_no_rnd_pixels_l2[1](
                                output_plane + s->all_fragments[i].first_pixel,
                                motion_source - d, 
                                motion_source + stride + 1 + d, 
                                stride, 8);
                        }
                        dequantizer = s->inter_dequant;
                    }else{
                        if (plane == 0)
                            dequantizer = s->intra_y_dequant;
                        else
                            dequantizer = s->intra_c_dequant;
                    }

                    /* dequantize the DCT coefficients */
                    debug_idct("fragment %d, coding mode %d, DC = %d, dequant = %d:\n", 
                        i, s->all_fragments[i].coding_method, 
                        DC_COEFF(i), dequantizer[0]);

                    if(s->avctx->idct_algo==FF_IDCT_VP3){
                        Coeff *coeff= s->coeffs + i;
                        memset(block, 0, sizeof(block));
                        while(coeff->next){
                            block[coeff->index]= coeff->coeff * dequantizer[coeff->index];
                            coeff= coeff->next;
                        }
                    }else{
                        Coeff *coeff= s->coeffs + i;
                        memset(block, 0, sizeof(block));
                        while(coeff->next){
                            block[coeff->index]= (coeff->coeff * dequantizer[coeff->index] + 2)>>2;
                            coeff= coeff->next;
                        }
                    }

                    /* invert DCT and place (or add) in final output */
                
                    if (s->all_fragments[i].coding_method == MODE_INTRA) {
                        if(s->avctx->idct_algo!=FF_IDCT_VP3)
                            block[0] += 128<<3;
                        s->dsp.idct_put(
                            output_plane + s->all_fragments[i].first_pixel,
                            stride,
                            block);
                    } else {
                        s->dsp.idct_add(
                            output_plane + s->all_fragments[i].first_pixel,
                            stride,
                            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);

                }
1924
#if 0
1925 1926 1927 1928 1929 1930 1931
                /* perform the left edge filter if:
                 *   - the fragment is not on the left column
                 *   - the fragment is coded in this frame
                 *   - the fragment is not coded in this frame but the left
                 *     fragment is coded in this frame (this is done instead
                 *     of a right edge filter when rendering the left fragment
                 *     since this fragment is not available yet) */
1932
                if ((x > 0) &&
1933 1934 1935
                    ((s->all_fragments[i].coding_method != MODE_COPY) ||
                     ((s->all_fragments[i].coding_method == MODE_COPY) &&
                      (s->all_fragments[i - 1].coding_method != MODE_COPY)) )) {
1936
                    horizontal_filter(
1937 1938
                        output_plane + s->all_fragments[i].first_pixel + 7*stride,
                        -stride, bounding_values);
1939 1940
                }

1941 1942 1943 1944 1945 1946 1947
                /* perform the top edge filter if:
                 *   - the fragment is not on the top row
                 *   - the fragment is coded in this frame
                 *   - the fragment is not coded in this frame but the above
                 *     fragment is coded in this frame (this is done instead
                 *     of a bottom edge filter when rendering the above
                 *     fragment since this fragment is not available yet) */
1948
                if ((y > 0) &&
1949 1950 1951
                    ((s->all_fragments[i].coding_method != MODE_COPY) ||
                     ((s->all_fragments[i].coding_method == MODE_COPY) &&
                      (s->all_fragments[i - fragment_width].coding_method != MODE_COPY)) )) {
1952
                    vertical_filter(
1953 1954
                        output_plane + s->all_fragments[i].first_pixel - stride,
                        -stride, bounding_values);
1955
                }
1956
#endif
1957 1958 1959 1960 1961 1962 1963
            }
        }
    }

     /* this looks like a good place for slice dispatch... */
     /* algorithm:
      *   if (slice == s->macroblock_height - 1)
1964 1965 1966
      *     dispatch (both last slice & 2nd-to-last slice);
      *   else if (slice > 0)
      *     dispatch (slice - 1);
1967 1968 1969 1970 1971
      */

    emms_c();
}

1972 1973 1974
static void horizontal_filter(unsigned char *first_pixel, int stride,
    int *bounding_values)
{
1975
    unsigned char *end;
1976 1977
    int filter_value;

1978
    for (end= first_pixel + 8*stride; first_pixel < end; first_pixel += stride) {
1979
        filter_value = 
1980 1981
            (first_pixel[-2] - first_pixel[ 1])
         +3*(first_pixel[ 0] - first_pixel[-1]);
1982
        filter_value = bounding_values[(filter_value + 4) >> 3];
1983 1984
        first_pixel[-1] = clip_uint8(first_pixel[-1] + filter_value);
        first_pixel[ 0] = clip_uint8(first_pixel[ 0] - filter_value);
1985 1986 1987 1988 1989 1990
    }
}

static void vertical_filter(unsigned char *first_pixel, int stride,
    int *bounding_values)
{
1991
    unsigned char *end;
1992
    int filter_value;
1993
    const int nstride= -stride;
1994

1995
    for (end= first_pixel + 8; first_pixel < end; first_pixel++) {
1996
        filter_value = 
1997 1998
            (first_pixel[2 * nstride] - first_pixel[ stride])
         +3*(first_pixel[0          ] - first_pixel[nstride]);
1999
        filter_value = bounding_values[(filter_value + 4) >> 3];
2000
        first_pixel[nstride] = clip_uint8(first_pixel[nstride] + filter_value);
2001
        first_pixel[0] = clip_uint8(first_pixel[0] - filter_value);
2002 2003 2004 2005 2006 2007 2008 2009 2010 2011
    }
}

static void apply_loop_filter(Vp3DecodeContext *s)
{
    int x, y, plane;
    int width, height;
    int fragment;
    int stride;
    unsigned char *plane_data;
2012
    int *bounding_values= s->bounding_values_array+127;
2013

2014
#if 0
2015
    int bounding_values_array[256];
2016 2017 2018 2019 2020 2021 2022
    int filter_limit;

    /* find the right loop limit value */
    for (x = 63; x >= 0; x--) {
        if (vp31_ac_scale_factor[x] >= s->quality_index)
            break;
    }
2023
    filter_limit = vp31_filter_limit_values[s->quality_index];
2024 2025

    /* set up the bounding values */
2026
    memset(bounding_values_array, 0, 256 * sizeof(int));
2027 2028 2029 2030 2031 2032
    for (x = 0; x < filter_limit; x++) {
        bounding_values[-x - filter_limit] = -filter_limit + x;
        bounding_values[-x] = -x;
        bounding_values[x] = x;
        bounding_values[x + filter_limit] = filter_limit - x;
    }
2033
#endif
2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060

    for (plane = 0; plane < 3; plane++) {

        if (plane == 0) {
            /* Y plane parameters */
            fragment = 0;
            width = s->fragment_width;
            height = s->fragment_height;
            stride = s->current_frame.linesize[0];
            plane_data = s->current_frame.data[0];
        } else if (plane == 1) {
            /* U plane parameters */
            fragment = s->u_fragment_start;
            width = s->fragment_width / 2;
            height = s->fragment_height / 2;
            stride = s->current_frame.linesize[1];
            plane_data = s->current_frame.data[1];
        } else {
            /* V plane parameters */
            fragment = s->v_fragment_start;
            width = s->fragment_width / 2;
            height = s->fragment_height / 2;
            stride = s->current_frame.linesize[2];
            plane_data = s->current_frame.data[2];
        }

        for (y = 0; y < height; y++) {
2061

2062
            for (x = 0; x < width; x++) {
2063
START_TIMER
2064 2065 2066 2067
                /* do not perform left edge filter for left columns frags */
                if ((x > 0) &&
                    (s->all_fragments[fragment].coding_method != MODE_COPY)) {
                    horizontal_filter(
2068
                        plane_data + s->all_fragments[fragment].first_pixel - 7*stride, 
2069 2070 2071 2072 2073 2074 2075
                        stride, bounding_values);
                }

                /* do not perform top edge filter for top row fragments */
                if ((y > 0) &&
                    (s->all_fragments[fragment].coding_method != MODE_COPY)) {
                    vertical_filter(
2076
                        plane_data + s->all_fragments[fragment].first_pixel + stride, 
2077 2078 2079 2080 2081 2082 2083 2084 2085 2086
                        stride, bounding_values);
                }

                /* do not perform right edge filter for right column
                 * fragments or if right fragment neighbor is also coded
                 * in this frame (it will be filtered in next iteration) */
                if ((x < width - 1) &&
                    (s->all_fragments[fragment].coding_method != MODE_COPY) &&
                    (s->all_fragments[fragment + 1].coding_method == MODE_COPY)) {
                    horizontal_filter(
2087
                        plane_data + s->all_fragments[fragment + 1].first_pixel - 7*stride, 
2088 2089 2090 2091 2092 2093 2094 2095 2096 2097
                        stride, bounding_values);
                }

                /* do not perform bottom edge filter for bottom row
                 * fragments or if bottom fragment neighbor is also coded
                 * in this frame (it will be filtered in the next row) */
                if ((y < height - 1) &&
                    (s->all_fragments[fragment].coding_method != MODE_COPY) &&
                    (s->all_fragments[fragment + width].coding_method == MODE_COPY)) {
                    vertical_filter(
2098
                        plane_data + s->all_fragments[fragment + width].first_pixel + stride, 
2099 2100 2101 2102
                        stride, bounding_values);
                }

                fragment++;
2103
STOP_TIMER("loop filter")
2104 2105 2106
            }
        }
    }
2107 2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120 2121 2122 2123 2124 2125 2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136 2137 2138 2139 2140 2141 2142 2143 2144 2145 2146 2147 2148 2149 2150 2151 2152 2153 2154 2155 2156 2157 2158 2159
}

/* 
 * 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);
        }
    }
}

2160 2161 2162 2163 2164 2165 2166 2167 2168 2169 2170 2171 2172 2173 2174 2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189 2190 2191 2192 2193 2194 2195 2196 2197 2198 2199 2200 2201 2202 2203 2204 2205 2206
/* FIXME: this should be merged with the above! */
static void theora_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 = 1; y <= s->fragment_height; 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 = 1; y <= s->fragment_height / 2; 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 = 1; y <= s->fragment_height / 2; 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);
        }
    }
}

2207 2208 2209 2210 2211 2212 2213
/*
 * This is the ffmpeg/libavcodec API init function.
 */
static int vp3_decode_init(AVCodecContext *avctx)
{
    Vp3DecodeContext *s = avctx->priv_data;
    int i;
2214 2215 2216 2217
    int c_width;
    int c_height;
    int y_superblock_count;
    int c_superblock_count;
2218

2219 2220 2221 2222 2223
    if (avctx->codec_tag == MKTAG('V','P','3','0'))
	s->version = 0;
    else
	s->version = 1;

2224
    s->avctx = avctx;
2225 2226
    s->width = (avctx->width + 15) & 0xFFFFFFF0;
    s->height = (avctx->height + 15) & 0xFFFFFFF0;
2227 2228
    avctx->pix_fmt = PIX_FMT_YUV420P;
    avctx->has_b_frames = 0;
2229 2230
    if(avctx->idct_algo==FF_IDCT_AUTO)
        avctx->idct_algo=FF_IDCT_VP3;
2231
    dsputil_init(&s->dsp, avctx);
2232 2233
    
    ff_init_scantable(s->dsp.idct_permutation, &s->scantable, ff_zigzag_direct);
2234 2235 2236 2237 2238

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

2239 2240 2241 2242 2243 2244 2245 2246 2247 2248 2249 2250 2251 2252
    s->y_superblock_width = (s->width + 31) / 32;
    s->y_superblock_height = (s->height + 31) / 32;
    y_superblock_count = s->y_superblock_width * s->y_superblock_height;

    /* work out the dimensions for the C planes */
    c_width = s->width / 2;
    c_height = s->height / 2;
    s->c_superblock_width = (c_width + 31) / 32;
    s->c_superblock_height = (c_height + 31) / 32;
    c_superblock_count = s->c_superblock_width * s->c_superblock_height;

    s->superblock_count = y_superblock_count + (c_superblock_count * 2);
    s->u_superblock_start = y_superblock_count;
    s->v_superblock_start = s->u_superblock_start + c_superblock_count;
2253 2254 2255 2256 2257 2258 2259 2260 2261 2262 2263 2264 2265 2266
    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;

2267 2268 2269 2270 2271 2272 2273 2274
    debug_init("  Y plane: %d x %d\n", s->width, s->height);
    debug_init("  C plane: %d x %d\n", c_width, c_height);
    debug_init("  Y superblocks: %d x %d, %d total\n",
        s->y_superblock_width, s->y_superblock_height, y_superblock_count);
    debug_init("  C superblocks: %d x %d, %d total\n",
        s->c_superblock_width, s->c_superblock_height, c_superblock_count);
    debug_init("  total superblocks = %d, U starts @ %d, V starts @ %d\n", 
        s->superblock_count, s->u_superblock_start, s->v_superblock_start);
2275 2276 2277 2278 2279 2280 2281 2282 2283 2284
    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));
2285
    s->coeffs = av_malloc(s->fragment_count * sizeof(Coeff) * 65);
2286 2287 2288
    s->coded_fragment_list = av_malloc(s->fragment_count * sizeof(int));
    s->pixel_addresses_inited = 0;

2289 2290 2291 2292 2293
    if (!s->theora_tables)
    {
	for (i = 0; i < 64; i++)
	    s->coded_dc_scale_factor[i] = vp31_dc_scale_factor[i];
	for (i = 0; i < 64; i++)
2294
	    s->coded_ac_scale_factor[i] = vp31_ac_scale_factor[i];
2295 2296 2297 2298 2299 2300
	for (i = 0; i < 64; i++)
	    s->coded_intra_y_dequant[i] = vp31_intra_y_dequant[i];
	for (i = 0; i < 64; i++)
	    s->coded_intra_c_dequant[i] = vp31_intra_c_dequant[i];
	for (i = 0; i < 64; i++)
	    s->coded_inter_dequant[i] = vp31_inter_dequant[i];
2301 2302
	for (i = 0; i < 64; i++)
	    s->filter_limit_values[i] = vp31_filter_limit_values[i];
2303

2304 2305 2306 2307 2308 2309 2310 2311 2312 2313 2314 2315 2316 2317 2318 2319 2320 2321 2322 2323 2324 2325 2326 2327 2328 2329 2330 2331 2332 2333 2334 2335 2336 2337 2338 2339 2340 2341 2342 2343 2344 2345 2346 2347 2348 2349 2350 2351 2352 2353 2354 2355 2356 2357 2358 2359
        /* 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, 0);

            /* group 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, 0);

            /* group 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, 0);

            /* group 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, 0);

            /* group 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, 0);
        }
    } else {
        for (i = 0; i < 16; i++) {

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

            /* group 1 AC histograms */
            init_vlc(&s->ac_vlc_1[i], 5, 32,
                &s->huffman_table[i+16][0][1], 4, 2,
                &s->huffman_table[i+16][0][0], 4, 2, 0);

            /* group 2 AC histograms */
            init_vlc(&s->ac_vlc_2[i], 5, 32,
                &s->huffman_table[i+16*2][0][1], 4, 2,
                &s->huffman_table[i+16*2][0][0], 4, 2, 0);

            /* group 3 AC histograms */
            init_vlc(&s->ac_vlc_3[i], 5, 32,
                &s->huffman_table[i+16*3][0][1], 4, 2,
                &s->huffman_table[i+16*3][0][0], 4, 2, 0);

            /* group 4 AC histograms */
            init_vlc(&s->ac_vlc_4[i], 5, 32,
                &s->huffman_table[i+16*4][0][1], 4, 2,
                &s->huffman_table[i+16*4][0][0], 4, 2, 0);
        }
2360 2361
    }

2362 2363 2364 2365
    init_vlc(&s->superblock_run_length_vlc, 6, 34,
        &superblock_run_length_vlc_table[0][1], 4, 2,
        &superblock_run_length_vlc_table[0][0], 4, 2, 0);

2366 2367 2368 2369 2370 2371 2372 2373 2374 2375 2376 2377
    init_vlc(&s->fragment_run_length_vlc, 5, 31,
        &fragment_run_length_vlc_table[0][1], 4, 2,
        &fragment_run_length_vlc_table[0][0], 4, 2, 0);

    init_vlc(&s->mode_code_vlc, 3, 8,
        &mode_code_vlc_table[0][1], 2, 1,
        &mode_code_vlc_table[0][0], 2, 1, 0);

    init_vlc(&s->motion_vector_vlc, 6, 63,
        &motion_vector_vlc_table[0][1], 2, 1,
        &motion_vector_vlc_table[0][0], 2, 1, 0);

2378 2379 2380 2381
    /* 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));
2382
    s->macroblock_coding = av_malloc(s->macroblock_count + 1);
2383 2384
    init_block_mapping(s);

2385 2386 2387 2388
    for (i = 0; i < 3; i++) {
        s->current_frame.data[i] = NULL;
        s->last_frame.data[i] = NULL;
        s->golden_frame.data[i] = NULL;
2389 2390
    }

2391 2392 2393 2394 2395 2396 2397 2398 2399 2400 2401 2402 2403
    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;
2404
    int i;
2405 2406

    init_get_bits(&gb, buf, buf_size * 8);
2407 2408 2409
    
    if (s->theora && get_bits1(&gb))
    {
2410
	int ptype = get_bits(&gb, 7);
2411

2412 2413 2414
	skip_bits(&gb, 6*8); /* "theora" */
	
	switch(ptype)
2415
	{
2416 2417 2418 2419 2420 2421 2422 2423 2424
	    case 1:
		theora_decode_comments(avctx, gb);
		break;
	    case 2:
		theora_decode_tables(avctx, gb);
    		init_dequantizer(s);
		break;
	    default:
		av_log(avctx, AV_LOG_ERROR, "Unknown Theora config packet: %d\n", ptype);
2425
	}
2426
	return buf_size;
2427
    }
2428 2429 2430

    s->keyframe = !get_bits1(&gb);
    if (!s->theora)
2431
	skip_bits(&gb, 1);
2432 2433
    s->last_quality_index = s->quality_index;
    s->quality_index = get_bits(&gb, 6);
2434
    if (s->theora >= 0x030200)
2435
        skip_bits1(&gb);
2436

2437 2438 2439
    if (s->avctx->debug & FF_DEBUG_PICT_INFO)
	av_log(s->avctx, AV_LOG_INFO, " VP3 %sframe #%d: Q index = %d\n",
	    s->keyframe?"key":"", counter, s->quality_index);
2440 2441
    counter++;

2442
    if (s->quality_index != s->last_quality_index) {
2443
        init_dequantizer(s);
2444 2445
        init_loop_filter(s);
    }
2446

2447
    if (s->keyframe) {
2448 2449 2450 2451 2452 2453 2454 2455 2456 2457 2458 2459 2460 2461 2462 2463 2464 2465
	if (!s->theora)
	{
	    skip_bits(&gb, 4); /* width code */
	    skip_bits(&gb, 4); /* height code */
	    if (s->version)
	    {
		s->version = get_bits(&gb, 5);
		if (counter == 1)
		    av_log(s->avctx, AV_LOG_DEBUG, "VP version: %d\n", s->version);
	    }
	}
	if (s->version || s->theora)
	{
    	    if (get_bits1(&gb))
    	        av_log(s->avctx, AV_LOG_ERROR, "Warning, unsupported keyframe coding type?!\n");
	    skip_bits(&gb, 2); /* reserved? */
	}

2466 2467 2468
        if (s->last_frame.data[0] == s->golden_frame.data[0]) {
            if (s->golden_frame.data[0])
                avctx->release_buffer(avctx, &s->golden_frame);
2469
            s->last_frame= s->golden_frame; /* ensure that we catch any access to this released frame */
2470 2471 2472 2473 2474 2475
        } else {
            if (s->golden_frame.data[0])
                avctx->release_buffer(avctx, &s->golden_frame);
            if (s->last_frame.data[0])
                avctx->release_buffer(avctx, &s->last_frame);
        }
2476

2477
        s->golden_frame.reference = 3;
2478
        if(avctx->get_buffer(avctx, &s->golden_frame) < 0) {
2479
            av_log(s->avctx, AV_LOG_ERROR, "vp3: get_buffer() failed\n");
2480 2481 2482 2483
            return -1;
        }

        /* golden frame is also the current frame */
2484
        memcpy(&s->current_frame, &s->golden_frame, sizeof(AVFrame));
2485 2486 2487

        /* time to figure out pixel addresses? */
        if (!s->pixel_addresses_inited)
2488 2489 2490 2491 2492 2493
	{
	    if (!s->flipped_image)
        	vp3_calculate_pixel_addresses(s);
	    else
		theora_calculate_pixel_addresses(s);
	}
2494 2495
    } else {
        /* allocate a new current frame */
2496
        s->current_frame.reference = 3;
2497
        if(avctx->get_buffer(avctx, &s->current_frame) < 0) {
2498
            av_log(s->avctx, AV_LOG_ERROR, "vp3: get_buffer() failed\n");
2499 2500 2501 2502
            return -1;
        }
    }

Michael Niedermayer's avatar
Michael Niedermayer committed
2503 2504 2505
    s->current_frame.qscale_table= s->qscale_table; //FIXME allocate individual tables per AVFrame
    s->current_frame.qstride= 0;

2506
    {START_TIMER
2507
    init_frame(s, &gb);
2508
    STOP_TIMER("init_frame")}
2509

2510 2511 2512 2513 2514 2515 2516 2517 2518 2519 2520 2521 2522
#if KEYFRAMES_ONLY
if (!s->keyframe) {

    memcpy(s->current_frame.data[0], s->golden_frame.data[0],
        s->current_frame.linesize[0] * s->height);
    memcpy(s->current_frame.data[1], s->golden_frame.data[1],
        s->current_frame.linesize[1] * s->height / 2);
    memcpy(s->current_frame.data[2], s->golden_frame.data[2],
        s->current_frame.linesize[2] * s->height / 2);

} else {
#endif

2523 2524 2525 2526 2527 2528 2529 2530 2531 2532 2533 2534 2535 2536 2537 2538 2539 2540 2541 2542 2543
    {START_TIMER
    if (unpack_superblocks(s, &gb)){
        av_log(s->avctx, AV_LOG_ERROR, "error in unpack_superblocks\n");
        return -1;
    }
    STOP_TIMER("unpack_superblocks")}
    {START_TIMER
    if (unpack_modes(s, &gb)){
        av_log(s->avctx, AV_LOG_ERROR, "error in unpack_modes\n");
        return -1;
    }
    STOP_TIMER("unpack_modes")}
    {START_TIMER
    if (unpack_vectors(s, &gb)){
        av_log(s->avctx, AV_LOG_ERROR, "error in unpack_vectors\n");
        return -1;
    }
    STOP_TIMER("unpack_vectors")}
    {START_TIMER
    if (unpack_dct_coeffs(s, &gb)){
        av_log(s->avctx, AV_LOG_ERROR, "error in unpack_dct_coeffs\n");
2544 2545
        return -1;
    }
2546 2547
    STOP_TIMER("unpack_dct_coeffs")}
    {START_TIMER
2548 2549

    reverse_dc_prediction(s, 0, s->fragment_width, s->fragment_height);
2550 2551 2552 2553 2554
    if ((avctx->flags & CODEC_FLAG_GRAY) == 0) {
        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);
2555 2556 2557 2558 2559 2560 2561
    }
    STOP_TIMER("reverse_dc_prediction")}
    {START_TIMER

    for (i = 0; i < s->macroblock_height; i++)
        render_slice(s, i);
    STOP_TIMER("render_fragments")}
2562

2563
    {START_TIMER
2564
    apply_loop_filter(s);
2565
    STOP_TIMER("apply_loop_filter")}
2566 2567 2568 2569
#if KEYFRAMES_ONLY
}
#endif

2570 2571 2572
    *data_size=sizeof(AVFrame);
    *(AVFrame*)data= s->current_frame;

2573 2574 2575 2576 2577
    /* 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);
2578

2579 2580
    /* shuffle frames (last = current) */
    memcpy(&s->last_frame, &s->current_frame, sizeof(AVFrame));
2581
    s->current_frame.data[0]= NULL; /* ensure that we catch any access to this released frame */
2582 2583 2584 2585 2586 2587 2588 2589 2590 2591 2592 2593

    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);
2594
    av_free(s->coeffs);
2595 2596 2597 2598
    av_free(s->coded_fragment_list);
    av_free(s->superblock_fragments);
    av_free(s->superblock_macroblocks);
    av_free(s->macroblock_fragments);
2599
    av_free(s->macroblock_coding);
2600
    
2601
    /* release all frames */
2602
    if (s->golden_frame.data[0] && s->golden_frame.data[0] != s->last_frame.data[0])
2603 2604 2605 2606 2607
        avctx->release_buffer(avctx, &s->golden_frame);
    if (s->last_frame.data[0])
        avctx->release_buffer(avctx, &s->last_frame);
    /* no need to release the current_frame since it will always be pointing
     * to the same frame as either the golden or last frame */
2608 2609 2610 2611

    return 0;
}

2612 2613 2614 2615 2616 2617 2618 2619 2620 2621 2622 2623 2624 2625 2626 2627 2628 2629 2630 2631 2632 2633 2634 2635 2636 2637 2638 2639 2640 2641 2642 2643
static int read_huffman_tree(AVCodecContext *avctx, GetBitContext *gb)
{
    Vp3DecodeContext *s = avctx->priv_data;

    if (get_bits(gb, 1)) {
        int token;
        if (s->entries >= 32) { /* overflow */
            av_log(avctx, AV_LOG_ERROR, "huffman tree overflow\n");
            return -1;
        }
        token = get_bits(gb, 5);
        //av_log(avctx, AV_LOG_DEBUG, "hti %d hbits %x token %d entry : %d size %d\n", s->hti, s->hbits, token, s->entries, s->huff_code_size);
        s->huffman_table[s->hti][token][0] = s->hbits;
        s->huffman_table[s->hti][token][1] = s->huff_code_size;
        s->entries++;
    }
    else {
        if (s->huff_code_size >= 32) {/* overflow */
            av_log(avctx, AV_LOG_ERROR, "huffman tree overflow\n");
            return -1;
        }
        s->huff_code_size++;
        s->hbits <<= 1;
        read_huffman_tree(avctx, gb);
        s->hbits |= 1;
        read_huffman_tree(avctx, gb);
        s->hbits >>= 1;
        s->huff_code_size--;
    }
    return 0;
}

2644 2645 2646
static int theora_decode_header(AVCodecContext *avctx, GetBitContext gb)
{
    Vp3DecodeContext *s = avctx->priv_data;
2647 2648 2649 2650 2651 2652 2653 2654
    int major, minor, micro;

    major = get_bits(&gb, 8); /* version major */
    minor = get_bits(&gb, 8); /* version minor */
    micro = get_bits(&gb, 8); /* version micro */
    av_log(avctx, AV_LOG_INFO, "Theora bitstream version %d.%d.%d\n",
	major, minor, micro);

2655 2656 2657
    /* FIXME: endianess? */
    s->theora = (major << 16) | (minor << 8) | micro;

2658
    /* 3.2.0 aka alpha3 has the same frame orientation as original vp3 */
2659
    /* but previous versions have the image flipped relative to vp3 */
2660
    if (s->theora < 0x030200)
2661 2662 2663 2664
    {
	s->flipped_image = 1;
        av_log(avctx, AV_LOG_DEBUG, "Old (<alpha3) Theora bitstream, flipped image\n");
    }
2665 2666 2667 2668

    s->width = get_bits(&gb, 16) << 4;
    s->height = get_bits(&gb, 16) << 4;
    
2669 2670 2671 2672 2673
    if(avcodec_check_dimensions(avctx, s->width, s->height)){
        s->width= s->height= 0;
        return -1;
    }
    
2674 2675 2676 2677 2678 2679 2680 2681 2682 2683 2684
    skip_bits(&gb, 24); /* frame width */
    skip_bits(&gb, 24); /* frame height */

    skip_bits(&gb, 8); /* offset x */
    skip_bits(&gb, 8); /* offset y */

    skip_bits(&gb, 32); /* fps numerator */
    skip_bits(&gb, 32); /* fps denumerator */
    skip_bits(&gb, 24); /* aspect numerator */
    skip_bits(&gb, 24); /* aspect denumerator */
    
2685
    if (s->theora < 0x030200)
2686
	skip_bits(&gb, 5); /* keyframe frequency force */
2687 2688 2689 2690 2691
    skip_bits(&gb, 8); /* colorspace */
    skip_bits(&gb, 24); /* bitrate */

    skip_bits(&gb, 6); /* last(?) quality index */
    
2692
    if (s->theora >= 0x030200)
2693 2694 2695 2696 2697
    {
	skip_bits(&gb, 5); /* keyframe frequency force */
	skip_bits(&gb, 5); /* spare bits */
    }
    
2698 2699 2700 2701 2702 2703 2704 2705
//    align_get_bits(&gb);
    
    avctx->width = s->width;
    avctx->height = s->height;

    return 0;
}

2706 2707 2708 2709
static int theora_decode_comments(AVCodecContext *avctx, GetBitContext gb)
{
    int nb_comments, i, tmp;

Michael Niedermayer's avatar
Michael Niedermayer committed
2710
    tmp = get_bits_long(&gb, 32);
2711 2712 2713
    tmp = be2me_32(tmp);
    while(tmp--)
	    skip_bits(&gb, 8);
2714

Michael Niedermayer's avatar
Michael Niedermayer committed
2715
    nb_comments = get_bits_long(&gb, 32);
2716
    nb_comments = be2me_32(nb_comments);
2717 2718
    for (i = 0; i < nb_comments; i++)
    {
Michael Niedermayer's avatar
Michael Niedermayer committed
2719
	tmp = get_bits_long(&gb, 32);
2720 2721
	tmp = be2me_32(tmp);
	while(tmp--)
2722 2723 2724 2725 2726 2727
	    skip_bits(&gb, 8);
    }
    
    return 0;
}

2728 2729 2730
static int theora_decode_tables(AVCodecContext *avctx, GetBitContext gb)
{
    Vp3DecodeContext *s = avctx->priv_data;
2731 2732 2733 2734
    int i, n;

    if (s->theora >= 0x030200) {
        n = get_bits(&gb, 3);
2735
        /* loop filter limit values table */
2736
        for (i = 0; i < 64; i++)
2737
            s->filter_limit_values[i] = get_bits(&gb, n);
2738
    }
2739
    
2740 2741 2742 2743
    if (s->theora >= 0x030200)
        n = get_bits(&gb, 4) + 1;
    else
        n = 16;
2744 2745
    /* quality threshold table */
    for (i = 0; i < 64; i++)
2746
	s->coded_ac_scale_factor[i] = get_bits(&gb, n);
2747

2748 2749 2750 2751
    if (s->theora >= 0x030200)
        n = get_bits(&gb, 4) + 1;
    else
        n = 16;
2752 2753
    /* dc scale factor table */
    for (i = 0; i < 64; i++)
2754
	s->coded_dc_scale_factor[i] = get_bits(&gb, n);
2755

2756 2757 2758 2759 2760 2761 2762 2763
    if (s->theora >= 0x030200)
        n = get_bits(&gb, 9) + 1;
    else
        n = 3;
    if (n != 3) {
        av_log(NULL,AV_LOG_ERROR, "unsupported nbms : %d\n", n);
        return -1;
    }
2764 2765 2766 2767 2768 2769 2770 2771 2772 2773 2774
    /* y coeffs */
    for (i = 0; i < 64; i++)
	s->coded_intra_y_dequant[i] = get_bits(&gb, 8);

    /* uv coeffs */
    for (i = 0; i < 64; i++)
	s->coded_intra_c_dequant[i] = get_bits(&gb, 8);

    /* inter coeffs */
    for (i = 0; i < 64; i++)
	s->coded_inter_dequant[i] = get_bits(&gb, 8);
2775

2776 2777 2778 2779 2780 2781 2782 2783 2784 2785 2786 2787 2788 2789 2790 2791 2792 2793 2794 2795 2796 2797 2798 2799 2800 2801 2802 2803 2804 2805 2806 2807 2808 2809 2810
    /* Huffman tables */
    for (i = 0; i <= 1; i++) {
        for (n = 0; n <= 2; n++) {
            int newqr;
            if (i > 0 || n > 0)
                newqr = get_bits(&gb, 1);
            else
                newqr = 1;
            if (!newqr) {
                if (i > 0)
                    get_bits(&gb, 1);
            }
            else {
                int qi = 0;
                skip_bits(&gb, av_log2(2)+1);
                while (qi < 63) {
                    qi += get_bits(&gb, av_log2(63-qi)+1) + 1;
                    skip_bits(&gb, av_log2(2)+1);
                }
                if (qi > 63)
                    av_log(NULL, AV_LOG_ERROR, "error...\n");
            }
        }
    }

    for (s->hti = 0; s->hti < 80; s->hti++) {
        s->entries = 0;
        s->huff_code_size = 1;
        if (!get_bits(&gb, 1)) {
            s->hbits = 0;
            read_huffman_tree(avctx, &gb);
            s->hbits = 1;
            read_huffman_tree(avctx, &gb);
        }
    }
2811 2812 2813 2814 2815 2816 2817 2818 2819 2820 2821
    
    s->theora_tables = 1;
    
    return 0;
}

static int theora_decode_init(AVCodecContext *avctx)
{
    Vp3DecodeContext *s = avctx->priv_data;
    GetBitContext gb;
    int ptype;
2822 2823
    uint8_t *p= avctx->extradata;
    int op_bytes, i;
2824 2825 2826 2827 2828 2829
    
    s->theora = 1;

    if (!avctx->extradata_size)
	return -1;

2830 2831 2832 2833 2834 2835
  for(i=0;i<3;i++) {
    op_bytes = *(p++)<<8;
    op_bytes += *(p++);

    init_get_bits(&gb, p, op_bytes);
    p += op_bytes;
2836 2837 2838 2839 2840 2841 2842 2843 2844 2845 2846 2847 2848 2849 2850

    ptype = get_bits(&gb, 8);
    debug_vp3("Theora headerpacket type: %x\n", ptype);
	    
    if (!(ptype & 0x80))
	return -1;
	
    skip_bits(&gb, 6*8); /* "theora" */
	
    switch(ptype)
    {
        case 0x80:
            theora_decode_header(avctx, gb);
    	    break;
	case 0x81:
2851
	    theora_decode_comments(avctx, gb);
2852 2853 2854 2855 2856
	    break;
	case 0x82:
	    theora_decode_tables(avctx, gb);
	    break;
    }
2857
  }
2858

2859
    vp3_decode_init(avctx);
2860 2861 2862
    return 0;
}

2863 2864 2865 2866 2867 2868 2869 2870 2871 2872 2873 2874
AVCodec vp3_decoder = {
    "vp3",
    CODEC_TYPE_VIDEO,
    CODEC_ID_VP3,
    sizeof(Vp3DecodeContext),
    vp3_decode_init,
    NULL,
    vp3_decode_end,
    vp3_decode_frame,
    0,
    NULL
};
2875

2876
#ifndef CONFIG_LIBTHEORA
2877 2878 2879 2880 2881 2882 2883 2884 2885 2886 2887 2888
AVCodec theora_decoder = {
    "theora",
    CODEC_TYPE_VIDEO,
    CODEC_ID_THEORA,
    sizeof(Vp3DecodeContext),
    theora_decode_init,
    NULL,
    vp3_decode_end,
    vp3_decode_frame,
    0,
    NULL
};
2889
#endif