svq3.c 39.7 KB
Newer Older
1
/*
2
 * Copyright (c) 2003 The FFmpeg Project
3
 *
4 5 6
 * This file is part of FFmpeg.
 *
 * FFmpeg is free software; you can redistribute it and/or
7 8
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
9
 * version 2.1 of the License, or (at your option) any later version.
10
 *
11
 * FFmpeg is distributed in the hope that it will be useful,
12 13 14 15 16
 * 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
17
 * License along with FFmpeg; if not, write to the Free Software
18
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19 20 21
 */

/*
22 23
 * How to use this decoder:
 * SVQ3 data is transported within Apple Quicktime files. Quicktime files
24 25 26 27 28 29
 * have stsd atoms to describe media trak properties. A stsd atom for a
 * video trak contains 1 or more ImageDescription atoms. These atoms begin
 * with the 4-byte length of the atom followed by the codec fourcc. Some
 * decoders need information in this atom to operate correctly. Such
 * is the case with SVQ3. In order to get the best use out of this decoder,
 * the calling app must make the SVQ3 ImageDescription atom available
30 31
 * via the AVCodecContext's extradata[_size] field:
 *
32
 * AVCodecContext.extradata = pointer to ImageDescription, first characters
33
 * are expected to be 'S', 'V', 'Q', and '3', NOT the 4-byte atom length
34 35
 * AVCodecContext.extradata_size = size of ImageDescription atom memory
 * buffer (which will be the same as the ImageDescription atom size field
36 37 38 39
 * from the QT file, minus 4 bytes since the length is missing)
 *
 * You will know you have these parameters passed correctly when the decoder
 * correctly decodes this file:
Diego Biurrun's avatar
Diego Biurrun committed
40
 *  http://samples.mplayerhq.hu/V-codecs/SVQ3/Vertical400kbit.sorenson3.mov
41
 */
42 43 44 45 46 47 48 49
#include "internal.h"
#include "dsputil.h"
#include "avcodec.h"
#include "mpegvideo.h"
#include "h264.h"

#include "h264data.h" //FIXME FIXME FIXME

50
#include "h264_mvpred.h"
51 52 53 54
#include "golomb.h"
#include "rectangle.h"
#include "vdpau_internal.h"

55
#if CONFIG_ZLIB
56 57 58
#include <zlib.h>
#endif

59 60
#include "svq1.h"

61
/**
62
 * @file
63 64 65
 * svq3 decoder.
 */

66 67 68 69 70 71 72
typedef struct {
    H264Context h;
    int halfpel_flag;
    int thirdpel_flag;
    int unknown_flag;
    int next_slice_index;
    uint32_t watermark_key;
73 74
    uint8_t *buf;
    int buf_size;
75 76
} SVQ3Context;

77 78
#define FULLPEL_MODE  1
#define HALFPEL_MODE  2
Michael Niedermayer's avatar
Michael Niedermayer committed
79
#define THIRDPEL_MODE 3
80
#define PREDICT_MODE  4
81

Michael Niedermayer's avatar
Michael Niedermayer committed
82 83 84 85 86 87
/* dual scan (from some older h264 draft)
 o-->o-->o   o
         |  /|
 o   o   o / o
 | / |   |/  |
 o   o   o   o
88
   /
Michael Niedermayer's avatar
Michael Niedermayer committed
89 90
 o-->o-->o-->o
*/
91 92 93 94 95
static const uint8_t svq3_scan[16] = {
    0+0*4, 1+0*4, 2+0*4, 2+1*4,
    2+2*4, 3+0*4, 3+1*4, 3+2*4,
    0+1*4, 0+2*4, 1+1*4, 1+2*4,
    0+3*4, 1+3*4, 2+3*4, 3+3*4,
96 97 98
};

static const uint8_t svq3_pred_0[25][2] = {
99 100 101 102 103 104 105 106 107
    { 0, 0 },
    { 1, 0 }, { 0, 1 },
    { 0, 2 }, { 1, 1 }, { 2, 0 },
    { 3, 0 }, { 2, 1 }, { 1, 2 }, { 0, 3 },
    { 0, 4 }, { 1, 3 }, { 2, 2 }, { 3, 1 }, { 4, 0 },
    { 4, 1 }, { 3, 2 }, { 2, 3 }, { 1, 4 },
    { 2, 4 }, { 3, 3 }, { 4, 2 },
    { 4, 3 }, { 3, 4 },
    { 4, 4 }
108 109 110
};

static const int8_t svq3_pred_1[6][6][5] = {
111 112 113 114 115 116 117 118 119 120 121 122
    { { 2,-1,-1,-1,-1 }, { 2, 1,-1,-1,-1 }, { 1, 2,-1,-1,-1 },
      { 2, 1,-1,-1,-1 }, { 1, 2,-1,-1,-1 }, { 1, 2,-1,-1,-1 } },
    { { 0, 2,-1,-1,-1 }, { 0, 2, 1, 4, 3 }, { 0, 1, 2, 4, 3 },
      { 0, 2, 1, 4, 3 }, { 2, 0, 1, 3, 4 }, { 0, 4, 2, 1, 3 } },
    { { 2, 0,-1,-1,-1 }, { 2, 1, 0, 4, 3 }, { 1, 2, 4, 0, 3 },
      { 2, 1, 0, 4, 3 }, { 2, 1, 4, 3, 0 }, { 1, 2, 4, 0, 3 } },
    { { 2, 0,-1,-1,-1 }, { 2, 0, 1, 4, 3 }, { 1, 2, 0, 4, 3 },
      { 2, 1, 0, 4, 3 }, { 2, 1, 3, 4, 0 }, { 2, 4, 1, 0, 3 } },
    { { 0, 2,-1,-1,-1 }, { 0, 2, 1, 3, 4 }, { 1, 2, 3, 0, 4 },
      { 2, 0, 1, 3, 4 }, { 2, 1, 3, 0, 4 }, { 2, 0, 4, 3, 1 } },
    { { 0, 2,-1,-1,-1 }, { 0, 2, 4, 1, 3 }, { 1, 4, 2, 0, 3 },
      { 4, 2, 0, 1, 3 }, { 2, 0, 1, 4, 3 }, { 4, 2, 1, 0, 3 } },
123 124 125
};

static const struct { uint8_t run; uint8_t level; } svq3_dct_tables[2][16] = {
126 127 128 129
    { { 0, 0 }, { 0, 1 }, { 1, 1 }, { 2, 1 }, { 0, 2 }, { 3, 1 }, { 4, 1 }, { 5, 1 },
      { 0, 3 }, { 1, 2 }, { 2, 2 }, { 6, 1 }, { 7, 1 }, { 8, 1 }, { 9, 1 }, { 0, 4 } },
    { { 0, 0 }, { 0, 1 }, { 1, 1 }, { 0, 2 }, { 2, 1 }, { 0, 3 }, { 0, 4 }, { 0, 5 },
      { 3, 1 }, { 4, 1 }, { 1, 2 }, { 1, 3 }, { 0, 6 }, { 0, 7 }, { 0, 8 }, { 0, 9 } }
130 131 132
};

static const uint32_t svq3_dequant_coeff[32] = {
133 134 135 136
     3881,  4351,  4890,  5481,  6154,  6914,  7761,  8718,
     9781, 10987, 12339, 13828, 15523, 17435, 19561, 21873,
    24552, 27656, 30847, 34870, 38807, 43747, 49103, 54683,
    61694, 68745, 77615, 89113,100253,109366,126635,141533
137 138
};

139
void ff_svq3_luma_dc_dequant_idct_c(DCTELEM *output, DCTELEM *input, int qp){
140
    const int qmul = svq3_dequant_coeff[qp];
141 142 143
#define stride 16
    int i;
    int temp[16];
144
    static const uint8_t x_offset[4]={0, 1*stride, 4*stride, 5*stride};
145

146 147 148 149 150
    for(i=0; i<4; i++){
        const int z0 = 13*(input[4*i+0] +    input[4*i+2]);
        const int z1 = 13*(input[4*i+0] -    input[4*i+2]);
        const int z2 =  7* input[4*i+1] - 17*input[4*i+3];
        const int z3 = 17* input[4*i+1] +  7*input[4*i+3];
151 152 153 154 155

        temp[4*i+0] = z0+z3;
        temp[4*i+1] = z1+z2;
        temp[4*i+2] = z1-z2;
        temp[4*i+3] = z0-z3;
156 157
    }

158 159 160 161 162 163 164 165 166 167 168
    for(i=0; i<4; i++){
        const int offset= x_offset[i];
        const int z0= 13*(temp[4*0+i] +    temp[4*2+i]);
        const int z1= 13*(temp[4*0+i] -    temp[4*2+i]);
        const int z2=  7* temp[4*1+i] - 17*temp[4*3+i];
        const int z3= 17* temp[4*1+i] +  7*temp[4*3+i];

        output[stride* 0+offset] = ((z0 + z3)*qmul + 0x80000) >> 20;
        output[stride* 2+offset] = ((z1 + z2)*qmul + 0x80000) >> 20;
        output[stride* 8+offset] = ((z1 - z2)*qmul + 0x80000) >> 20;
        output[stride*10+offset] = ((z0 - z3)*qmul + 0x80000) >> 20;
169 170 171 172
    }
}
#undef stride

173
void ff_svq3_add_idct_c(uint8_t *dst, DCTELEM *block, int stride, int qp,
174
                            int dc)
175
{
176
    const int qmul = svq3_dequant_coeff[qp];
177
    int i;
178
    uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
179 180 181 182 183 184

    if (dc) {
        dc = 13*13*((dc == 1) ? 1538*block[0] : ((qmul*(block[0] >> 3)) / 2));
        block[0] = 0;
    }

185 186 187 188 189
    for (i = 0; i < 4; i++) {
        const int z0 = 13*(block[0 + 4*i] +    block[2 + 4*i]);
        const int z1 = 13*(block[0 + 4*i] -    block[2 + 4*i]);
        const int z2 =  7* block[1 + 4*i] - 17*block[3 + 4*i];
        const int z3 = 17* block[1 + 4*i] +  7*block[3 + 4*i];
190

191 192 193 194
        block[0 + 4*i] = z0 + z3;
        block[1 + 4*i] = z1 + z2;
        block[2 + 4*i] = z1 - z2;
        block[3 + 4*i] = z0 - z3;
195 196
    }

197 198 199 200 201 202 203 204 205 206 207
    for (i = 0; i < 4; i++) {
        const int z0 = 13*(block[i + 4*0] +    block[i + 4*2]);
        const int z1 = 13*(block[i + 4*0] -    block[i + 4*2]);
        const int z2 =  7* block[i + 4*1] - 17*block[i + 4*3];
        const int z3 = 17* block[i + 4*1] +  7*block[i + 4*3];
        const int rr = (dc + 0x80000);

        dst[i + stride*0] = cm[ dst[i + stride*0] + (((z0 + z3)*qmul + rr) >> 20) ];
        dst[i + stride*1] = cm[ dst[i + stride*1] + (((z1 + z2)*qmul + rr) >> 20) ];
        dst[i + stride*2] = cm[ dst[i + stride*2] + (((z1 - z2)*qmul + rr) >> 20) ];
        dst[i + stride*3] = cm[ dst[i + stride*3] + (((z0 - z3)*qmul + rr) >> 20) ];
208 209 210
    }
}

211 212 213
static inline int svq3_decode_block(GetBitContext *gb, DCTELEM *block,
                                    int index, const int type)
{
214 215
    static const uint8_t *const scan_patterns[4] =
    { luma_dc_zigzag_scan, zigzag_scan, svq3_scan, chroma_dc_scan };
216

217 218 219
    int run, level, sign, vlc, limit;
    const int intra = (3 * type) >> 2;
    const uint8_t *const scan = scan_patterns[type];
220

221 222
    for (limit = (16 >> intra); index < 16; index = limit, limit += 8) {
        for (; (vlc = svq3_get_ue_golomb(gb)) != 0; index++) {
223

224 225
          if (vlc == INVALID_VLC)
              return -1;
226

227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252
          sign = (vlc & 0x1) - 1;
          vlc  = (vlc + 1) >> 1;

          if (type == 3) {
              if (vlc < 3) {
                  run   = 0;
                  level = vlc;
              } else if (vlc < 4) {
                  run   = 1;
                  level = 1;
              } else {
                  run   = (vlc & 0x3);
                  level = ((vlc + 9) >> 2) - run;
              }
          } else {
              if (vlc < 16) {
                  run   = svq3_dct_tables[intra][vlc].run;
                  level = svq3_dct_tables[intra][vlc].level;
              } else if (intra) {
                  run   = (vlc & 0x7);
                  level = (vlc >> 3) + ((run == 0) ? 8 : ((run < 2) ? 2 : ((run < 5) ? 0 : -1)));
              } else {
                  run   = (vlc & 0xF);
                  level = (vlc >> 4) + ((run == 0) ? 4 : ((run < 3) ? 2 : ((run < 10) ? 1 : 0)));
              }
          }
253

254 255
          if ((index += run) >= limit)
              return -1;
256

257 258
          block[scan[index]] = (level ^ sign) - sign;
        }
259

260 261 262
        if (type != 2) {
            break;
        }
263 264
    }

265
    return 0;
266 267
}

268 269 270 271 272
static inline void svq3_mc_dir_part(MpegEncContext *s,
                                    int x, int y, int width, int height,
                                    int mx, int my, int dxy,
                                    int thirdpel, int dir, int avg)
{
273 274 275 276 277 278 279
    const Picture *pic = (dir == 0) ? &s->last_picture : &s->next_picture;
    uint8_t *src, *dest;
    int i, emu = 0;
    int blocksize = 2 - (width>>3); //16->0, 8->1, 4->2

    mx += x;
    my += y;
280

281 282
    if (mx < 0 || mx >= (s->h_edge_pos - width  - 1) ||
        my < 0 || my >= (s->v_edge_pos - height - 1)) {
283

284 285 286
        if ((s->flags & CODEC_FLAG_EMU_EDGE)) {
            emu = 1;
        }
287

288 289
        mx = av_clip (mx, -16, (s->h_edge_pos - width  + 15));
        my = av_clip (my, -16, (s->v_edge_pos - height + 15));
290 291
    }

292 293 294 295 296
    /* form component predictions */
    dest = s->current_picture.data[0] + x + y*s->linesize;
    src  = pic->data[0] + mx + my*s->linesize;

    if (emu) {
297
        s->dsp.emulated_edge_mc(s->edge_emu_buffer, src, s->linesize, (width + 1), (height + 1),
298
                            mx, my, s->h_edge_pos, s->v_edge_pos);
299
        src = s->edge_emu_buffer;
300
    }
301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317
    if (thirdpel)
        (avg ? s->dsp.avg_tpel_pixels_tab : s->dsp.put_tpel_pixels_tab)[dxy](dest, src, s->linesize, width, height);
    else
        (avg ? s->dsp.avg_pixels_tab : s->dsp.put_pixels_tab)[blocksize][dxy](dest, src, s->linesize, height);

    if (!(s->flags & CODEC_FLAG_GRAY)) {
        mx     = (mx + (mx < (int) x)) >> 1;
        my     = (my + (my < (int) y)) >> 1;
        width  = (width  >> 1);
        height = (height >> 1);
        blocksize++;

        for (i = 1; i < 3; i++) {
            dest = s->current_picture.data[i] + (x >> 1) + (y >> 1)*s->uvlinesize;
            src  = pic->data[i] + mx + my*s->uvlinesize;

            if (emu) {
318
                s->dsp.emulated_edge_mc(s->edge_emu_buffer, src, s->uvlinesize, (width + 1), (height + 1),
319 320 321 322 323 324 325 326 327
                                    mx, my, (s->h_edge_pos >> 1), (s->v_edge_pos >> 1));
                src = s->edge_emu_buffer;
            }
            if (thirdpel)
                (avg ? s->dsp.avg_tpel_pixels_tab : s->dsp.put_tpel_pixels_tab)[dxy](dest, src, s->uvlinesize, width, height);
            else
                (avg ? s->dsp.avg_pixels_tab : s->dsp.put_pixels_tab)[blocksize][dxy](dest, src, s->uvlinesize, height);
        }
    }
328 329
}

330 331 332
static inline int svq3_mc_dir(H264Context *h, int size, int mode, int dir,
                              int avg)
{
333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430
    int i, j, k, mx, my, dx, dy, x, y;
    MpegEncContext *const s = (MpegEncContext *) h;
    const int part_width  = ((size & 5) == 4) ? 4 : 16 >> (size & 1);
    const int part_height = 16 >> ((unsigned) (size + 1) / 3);
    const int extra_width = (mode == PREDICT_MODE) ? -16*6 : 0;
    const int h_edge_pos  = 6*(s->h_edge_pos - part_width ) - extra_width;
    const int v_edge_pos  = 6*(s->v_edge_pos - part_height) - extra_width;

    for (i = 0; i < 16; i += part_height) {
        for (j = 0; j < 16; j += part_width) {
            const int b_xy = (4*s->mb_x + (j >> 2)) + (4*s->mb_y + (i >> 2))*h->b_stride;
            int dxy;
            x = 16*s->mb_x + j;
            y = 16*s->mb_y + i;
            k = ((j >> 2) & 1) + ((i >> 1) & 2) + ((j >> 1) & 4) + (i & 8);

            if (mode != PREDICT_MODE) {
                pred_motion(h, k, (part_width >> 2), dir, 1, &mx, &my);
            } else {
                mx = s->next_picture.motion_val[0][b_xy][0]<<1;
                my = s->next_picture.motion_val[0][b_xy][1]<<1;

                if (dir == 0) {
                    mx = ((mx * h->frame_num_offset) / h->prev_frame_num_offset + 1) >> 1;
                    my = ((my * h->frame_num_offset) / h->prev_frame_num_offset + 1) >> 1;
                } else {
                    mx = ((mx * (h->frame_num_offset - h->prev_frame_num_offset)) / h->prev_frame_num_offset + 1) >> 1;
                    my = ((my * (h->frame_num_offset - h->prev_frame_num_offset)) / h->prev_frame_num_offset + 1) >> 1;
                }
            }

            /* clip motion vector prediction to frame border */
            mx = av_clip(mx, extra_width - 6*x, h_edge_pos - 6*x);
            my = av_clip(my, extra_width - 6*y, v_edge_pos - 6*y);

            /* get (optional) motion vector differential */
            if (mode == PREDICT_MODE) {
                dx = dy = 0;
            } else {
                dy = svq3_get_se_golomb(&s->gb);
                dx = svq3_get_se_golomb(&s->gb);

                if (dx == INVALID_VLC || dy == INVALID_VLC) {
                    av_log(h->s.avctx, AV_LOG_ERROR, "invalid MV vlc\n");
                    return -1;
                }
            }

            /* compute motion vector */
            if (mode == THIRDPEL_MODE) {
                int fx, fy;
                mx  = ((mx + 1)>>1) + dx;
                my  = ((my + 1)>>1) + dy;
                fx  = ((unsigned)(mx + 0x3000))/3 - 0x1000;
                fy  = ((unsigned)(my + 0x3000))/3 - 0x1000;
                dxy = (mx - 3*fx) + 4*(my - 3*fy);

                svq3_mc_dir_part(s, x, y, part_width, part_height, fx, fy, dxy, 1, dir, avg);
                mx += mx;
                my += my;
            } else if (mode == HALFPEL_MODE || mode == PREDICT_MODE) {
                mx  = ((unsigned)(mx + 1 + 0x3000))/3 + dx - 0x1000;
                my  = ((unsigned)(my + 1 + 0x3000))/3 + dy - 0x1000;
                dxy = (mx&1) + 2*(my&1);

                svq3_mc_dir_part(s, x, y, part_width, part_height, mx>>1, my>>1, dxy, 0, dir, avg);
                mx *= 3;
                my *= 3;
            } else {
                mx = ((unsigned)(mx + 3 + 0x6000))/6 + dx - 0x1000;
                my = ((unsigned)(my + 3 + 0x6000))/6 + dy - 0x1000;

                svq3_mc_dir_part(s, x, y, part_width, part_height, mx, my, 0, 0, dir, avg);
                mx *= 6;
                my *= 6;
            }

            /* update mv_cache */
            if (mode != PREDICT_MODE) {
                int32_t mv = pack16to32(mx,my);

                if (part_height == 8 && i < 8) {
                    *(int32_t *) h->mv_cache[dir][scan8[k] + 1*8] = mv;

                    if (part_width == 8 && j < 8) {
                        *(int32_t *) h->mv_cache[dir][scan8[k] + 1 + 1*8] = mv;
                    }
                }
                if (part_width == 8 && j < 8) {
                    *(int32_t *) h->mv_cache[dir][scan8[k] + 1] = mv;
                }
                if (part_width == 4 || part_height == 4) {
                    *(int32_t *) h->mv_cache[dir][scan8[k]] = mv;
                }
            }

            /* write back motion vectors */
            fill_rectangle(s->current_picture.motion_val[dir][b_xy], part_width>>2, part_height>>2, h->b_stride, pack16to32(mx,my), 4);
431
        }
432 433
    }

434
    return 0;
435 436
}

437
static int svq3_decode_mb(SVQ3Context *svq3, unsigned int mb_type)
438
{
439
    H264Context *h = &svq3->h;
440 441 442 443 444 445 446 447 448 449 450 451 452
    int i, j, k, m, dir, mode;
    int cbp = 0;
    uint32_t vlc;
    int8_t *top, *left;
    MpegEncContext *const s = (MpegEncContext *) h;
    const int mb_xy = h->mb_xy;
    const int b_xy  = 4*s->mb_x + 4*s->mb_y*h->b_stride;

    h->top_samples_available      = (s->mb_y == 0) ? 0x33FF : 0xFFFF;
    h->left_samples_available     = (s->mb_x == 0) ? 0x5F5F : 0xFFFF;
    h->topright_samples_available = 0xFFFF;

    if (mb_type == 0) {           /* SKIP */
453
        if (s->pict_type == AV_PICTURE_TYPE_P || s->next_picture.mb_type[mb_xy] == -1) {
454 455
            svq3_mc_dir_part(s, 16*s->mb_x, 16*s->mb_y, 16, 16, 0, 0, 0, 0, 0, 0);

456
            if (s->pict_type == AV_PICTURE_TYPE_B) {
457 458 459 460 461 462 463 464 465 466
                svq3_mc_dir_part(s, 16*s->mb_x, 16*s->mb_y, 16, 16, 0, 0, 0, 0, 1, 1);
            }

            mb_type = MB_TYPE_SKIP;
        } else {
            mb_type = FFMIN(s->next_picture.mb_type[mb_xy], 6);
            if (svq3_mc_dir(h, mb_type, PREDICT_MODE, 0, 0) < 0)
                return -1;
            if (svq3_mc_dir(h, mb_type, PREDICT_MODE, 1, 1) < 0)
                return -1;
467

468
            mb_type = MB_TYPE_16x16;
469
        }
470
    } else if (mb_type < 8) {     /* INTER */
471
        if (svq3->thirdpel_flag && svq3->halfpel_flag == !get_bits1 (&s->gb)) {
472
            mode = THIRDPEL_MODE;
473
        } else if (svq3->halfpel_flag && svq3->thirdpel_flag == !get_bits1 (&s->gb)) {
474 475 476
            mode = HALFPEL_MODE;
        } else {
            mode = FULLPEL_MODE;
477
        }
478

479 480 481 482 483 484 485 486 487 488
        /* fill caches */
        /* note ref_cache should contain here:
            ????????
            ???11111
            N??11111
            N??11111
            N??11111
        */

        for (m = 0; m < 2; m++) {
489
            if (s->mb_x > 0 && h->intra4x4_pred_mode[h->mb2br_xy[mb_xy - 1]+6] != -1) {
490 491 492 493 494 495 496 497 498 499
                for (i = 0; i < 4; i++) {
                    *(uint32_t *) h->mv_cache[m][scan8[0] - 1 + i*8] = *(uint32_t *) s->current_picture.motion_val[m][b_xy - 1 + i*h->b_stride];
                }
            } else {
                for (i = 0; i < 4; i++) {
                    *(uint32_t *) h->mv_cache[m][scan8[0] - 1 + i*8] = 0;
                }
            }
            if (s->mb_y > 0) {
                memcpy(h->mv_cache[m][scan8[0] - 1*8], s->current_picture.motion_val[m][b_xy - h->b_stride], 4*2*sizeof(int16_t));
500
                memset(&h->ref_cache[m][scan8[0] - 1*8], (h->intra4x4_pred_mode[h->mb2br_xy[mb_xy - s->mb_stride]] == -1) ? PART_NOT_AVAILABLE : 1, 4);
501 502 503 504

                if (s->mb_x < (s->mb_width - 1)) {
                    *(uint32_t *) h->mv_cache[m][scan8[0] + 4 - 1*8] = *(uint32_t *) s->current_picture.motion_val[m][b_xy - h->b_stride + 4];
                    h->ref_cache[m][scan8[0] + 4 - 1*8] =
505 506
                        (h->intra4x4_pred_mode[h->mb2br_xy[mb_xy - s->mb_stride + 1]+6] == -1 ||
                         h->intra4x4_pred_mode[h->mb2br_xy[mb_xy - s->mb_stride    ]  ] == -1) ? PART_NOT_AVAILABLE : 1;
507 508 509 510
                }else
                    h->ref_cache[m][scan8[0] + 4 - 1*8] = PART_NOT_AVAILABLE;
                if (s->mb_x > 0) {
                    *(uint32_t *) h->mv_cache[m][scan8[0] - 1 - 1*8] = *(uint32_t *) s->current_picture.motion_val[m][b_xy - h->b_stride - 1];
511
                    h->ref_cache[m][scan8[0] - 1 - 1*8] = (h->intra4x4_pred_mode[h->mb2br_xy[mb_xy - s->mb_stride - 1]+3] == -1) ? PART_NOT_AVAILABLE : 1;
512 513 514 515 516
                }else
                    h->ref_cache[m][scan8[0] - 1 - 1*8] = PART_NOT_AVAILABLE;
            }else
                memset(&h->ref_cache[m][scan8[0] - 1*8 - 1], PART_NOT_AVAILABLE, 8);

517
            if (s->pict_type != AV_PICTURE_TYPE_B)
518
                break;
519
        }
520

521
        /* decode motion vector(s) and form prediction(s) */
522
        if (s->pict_type == AV_PICTURE_TYPE_P) {
523 524
            if (svq3_mc_dir(h, (mb_type - 1), mode, 0, 0) < 0)
                return -1;
525
        } else {        /* AV_PICTURE_TYPE_B */
526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541
            if (mb_type != 2) {
                if (svq3_mc_dir(h, 0, mode, 0, 0) < 0)
                    return -1;
            } else {
                for (i = 0; i < 4; i++) {
                    memset(s->current_picture.motion_val[0][b_xy + i*h->b_stride], 0, 4*2*sizeof(int16_t));
                }
            }
            if (mb_type != 1) {
                if (svq3_mc_dir(h, 0, mode, 1, (mb_type == 3)) < 0)
                    return -1;
            } else {
                for (i = 0; i < 4; i++) {
                    memset(s->current_picture.motion_val[1][b_xy + i*h->b_stride], 0, 4*2*sizeof(int16_t));
                }
            }
542
        }
543

544 545 546 547 548 549 550
        mb_type = MB_TYPE_16x16;
    } else if (mb_type == 8 || mb_type == 33) {   /* INTRA4x4 */
        memset(h->intra4x4_pred_mode_cache, -1, 8*5*sizeof(int8_t));

        if (mb_type == 8) {
            if (s->mb_x > 0) {
                for (i = 0; i < 4; i++) {
551
                    h->intra4x4_pred_mode_cache[scan8[0] - 1 + i*8] = h->intra4x4_pred_mode[h->mb2br_xy[mb_xy - 1]+6-i];
552 553 554 555 556 557
                }
                if (h->intra4x4_pred_mode_cache[scan8[0] - 1] == -1) {
                    h->left_samples_available = 0x5F5F;
                }
            }
            if (s->mb_y > 0) {
558 559 560 561
                h->intra4x4_pred_mode_cache[4+8*0] = h->intra4x4_pred_mode[h->mb2br_xy[mb_xy - s->mb_stride]+0];
                h->intra4x4_pred_mode_cache[5+8*0] = h->intra4x4_pred_mode[h->mb2br_xy[mb_xy - s->mb_stride]+1];
                h->intra4x4_pred_mode_cache[6+8*0] = h->intra4x4_pred_mode[h->mb2br_xy[mb_xy - s->mb_stride]+2];
                h->intra4x4_pred_mode_cache[7+8*0] = h->intra4x4_pred_mode[h->mb2br_xy[mb_xy - s->mb_stride]+3];
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

                if (h->intra4x4_pred_mode_cache[4+8*0] == -1) {
                    h->top_samples_available = 0x33FF;
                }
            }

            /* decode prediction codes for luma blocks */
            for (i = 0; i < 16; i+=2) {
                vlc = svq3_get_ue_golomb(&s->gb);

                if (vlc >= 25){
                    av_log(h->s.avctx, AV_LOG_ERROR, "luma prediction:%d\n", vlc);
                    return -1;
                }

                left    = &h->intra4x4_pred_mode_cache[scan8[i] - 1];
                top     = &h->intra4x4_pred_mode_cache[scan8[i] - 8];

                left[1] = svq3_pred_1[top[0] + 1][left[0] + 1][svq3_pred_0[vlc][0]];
                left[2] = svq3_pred_1[top[1] + 1][left[1] + 1][svq3_pred_0[vlc][1]];

                if (left[1] == -1 || left[2] == -1){
                    av_log(h->s.avctx, AV_LOG_ERROR, "weird prediction\n");
                    return -1;
                }
            }
        } else {    /* mb_type == 33, DC_128_PRED block type */
            for (i = 0; i < 4; i++) {
                memset(&h->intra4x4_pred_mode_cache[scan8[0] + 8*i], DC_PRED, 4);
            }
592
        }
593

594
        ff_h264_write_back_intra_pred_mode(h);
595

596
        if (mb_type == 8) {
597
            ff_h264_check_intra4x4_pred_mode(h);
598

599 600 601 602 603 604
            h->top_samples_available  = (s->mb_y == 0) ? 0x33FF : 0xFFFF;
            h->left_samples_available = (s->mb_x == 0) ? 0x5F5F : 0xFFFF;
        } else {
            for (i = 0; i < 4; i++) {
                memset(&h->intra4x4_pred_mode_cache[scan8[0] + 8*i], DC_128_PRED, 4);
            }
605

606 607 608
            h->top_samples_available  = 0x33FF;
            h->left_samples_available = 0x5F5F;
        }
609

610 611 612 613
        mb_type = MB_TYPE_INTRA4x4;
    } else {                      /* INTRA16x16 */
        dir = i_mb_type_info[mb_type - 8].pred_mode;
        dir = (dir >> 1) ^ 3*(dir & 1) ^ 1;
614

615
        if ((h->intra16x16_pred_mode = ff_h264_check_intra_pred_mode(h, dir)) == -1){
616 617 618
            av_log(h->s.avctx, AV_LOG_ERROR, "check_intra_pred_mode = -1\n");
            return -1;
        }
619

620 621
        cbp = i_mb_type_info[mb_type - 8].cbp;
        mb_type = MB_TYPE_INTRA16x16;
622
    }
623

624
    if (!IS_INTER(mb_type) && s->pict_type != AV_PICTURE_TYPE_I) {
625 626 627
        for (i = 0; i < 4; i++) {
            memset(s->current_picture.motion_val[0][b_xy + i*h->b_stride], 0, 4*2*sizeof(int16_t));
        }
628
        if (s->pict_type == AV_PICTURE_TYPE_B) {
629 630 631 632
            for (i = 0; i < 4; i++) {
                memset(s->current_picture.motion_val[1][b_xy + i*h->b_stride], 0, 4*2*sizeof(int16_t));
            }
        }
633
    }
634
    if (!IS_INTRA4x4(mb_type)) {
635
        memset(h->intra4x4_pred_mode+h->mb2br_xy[mb_xy], DC_PRED, 8);
636
    }
637
    if (!IS_SKIP(mb_type) || s->pict_type == AV_PICTURE_TYPE_B) {
638 639
        memset(h->non_zero_count_cache + 8, 0, 4*9*sizeof(uint8_t));
        s->dsp.clear_blocks(h->mb);
640
    }
641

642
    if (!IS_INTRA16x16(mb_type) && (!IS_SKIP(mb_type) || s->pict_type == AV_PICTURE_TYPE_B)) {
643 644 645 646
        if ((vlc = svq3_get_ue_golomb(&s->gb)) >= 48){
            av_log(h->s.avctx, AV_LOG_ERROR, "cbp_vlc=%d\n", vlc);
            return -1;
        }
647

648
        cbp = IS_INTRA(mb_type) ? golomb_to_intra4x4_cbp[vlc] : golomb_to_inter_cbp[vlc];
649
    }
650
    if (IS_INTRA16x16(mb_type) || (s->pict_type != AV_PICTURE_TYPE_I && s->adaptive_quant && cbp)) {
651
        s->qscale += svq3_get_se_golomb(&s->gb);
652

653 654
        if (s->qscale > 31){
            av_log(h->s.avctx, AV_LOG_ERROR, "qscale:%d\n", s->qscale);
655 656
            return -1;
        }
657
    }
658
    if (IS_INTRA16x16(mb_type)) {
659 660 661
        AV_ZERO128(h->mb_luma_dc+0);
        AV_ZERO128(h->mb_luma_dc+8);
        if (svq3_decode_block(&s->gb, h->mb_luma_dc, 0, 1)){
662 663
            av_log(h->s.avctx, AV_LOG_ERROR, "error while decoding intra luma dc\n");
            return -1;
664
        }
665
    }
666

667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683
    if (cbp) {
        const int index = IS_INTRA16x16(mb_type) ? 1 : 0;
        const int type = ((s->qscale < 24 && IS_INTRA4x4(mb_type)) ? 2 : 1);

        for (i = 0; i < 4; i++) {
            if ((cbp & (1 << i))) {
                for (j = 0; j < 4; j++) {
                    k = index ? ((j&1) + 2*(i&1) + 2*(j&2) + 4*(i&2)) : (4*i + j);
                    h->non_zero_count_cache[ scan8[k] ] = 1;

                    if (svq3_decode_block(&s->gb, &h->mb[16*k], index, type)){
                        av_log(h->s.avctx, AV_LOG_ERROR, "error while decoding block\n");
                        return -1;
                    }
                }
            }
        }
684

685 686
        if ((cbp & 0x30)) {
            for (i = 0; i < 2; ++i) {
687 688 689 690
              if (svq3_decode_block(&s->gb, &h->mb[16*(16 + 4*i)], 0, 3)){
                av_log(h->s.avctx, AV_LOG_ERROR, "error while decoding chroma dc block\n");
                return -1;
              }
691 692 693 694 695 696 697 698 699 700 701 702
            }

            if ((cbp & 0x20)) {
                for (i = 0; i < 8; i++) {
                    h->non_zero_count_cache[ scan8[16+i] ] = 1;

                    if (svq3_decode_block(&s->gb, &h->mb[16*(16 + i)], 1, 1)){
                        av_log(h->s.avctx, AV_LOG_ERROR, "error while decoding chroma ac block\n");
                        return -1;
                    }
                }
            }
703
        }
704 705
    }

706
    h->cbp= cbp;
707
    s->current_picture.mb_type[mb_xy] = mb_type;
708

709
    if (IS_INTRA(mb_type)) {
710
        h->chroma_pred_mode = ff_h264_check_intra_pred_mode(h, DC_PRED8x8);
711
    }
712

713
    return 0;
714 715
}

716
static int svq3_decode_slice_header(AVCodecContext *avctx)
717
{
718 719 720
    SVQ3Context *svq3 = avctx->priv_data;
    H264Context *h = &svq3->h;
    MpegEncContext *s = &h->s;
721 722
    const int mb_xy = h->mb_xy;
    int i, header;
723

724
    header = get_bits(&s->gb, 8);
725

726 727
    if (((header & 0x9F) != 1 && (header & 0x9F) != 2) || (header & 0x60) == 0) {
        /* TODO: what? */
728
        av_log(avctx, AV_LOG_ERROR, "unsupported slice header (%02X)\n", header);
729 730 731
        return -1;
    } else {
        int length = (header >> 5) & 3;
732

733
        svq3->next_slice_index = get_bits_count(&s->gb) + 8*show_bits(&s->gb, 8*length) + 8*length;
734

735 736
        if (svq3->next_slice_index > s->gb.size_in_bits) {
            av_log(avctx, AV_LOG_ERROR, "slice after bitstream end\n");
737
            return -1;
738
    }
739

740
        s->gb.size_in_bits = svq3->next_slice_index - 8*(length - 1);
741
        skip_bits(&s->gb, 8);
742

743
        if (svq3->watermark_key) {
744
            uint32_t header = AV_RL32(&s->gb.buffer[(get_bits_count(&s->gb)>>3)+1]);
745
            AV_WL32(&s->gb.buffer[(get_bits_count(&s->gb)>>3)+1], header ^ svq3->watermark_key);
746 747 748 749 750
        }
        if (length > 0) {
            memcpy((uint8_t *) &s->gb.buffer[get_bits_count(&s->gb) >> 3],
                   &s->gb.buffer[s->gb.size_in_bits >> 3], (length - 1));
        }
751
        skip_bits_long(&s->gb, 0);
752 753
    }

754 755 756 757
    if ((i = svq3_get_ue_golomb(&s->gb)) == INVALID_VLC || i >= 3){
        av_log(h->s.avctx, AV_LOG_ERROR, "illegal slice type %d \n", i);
        return -1;
    }
758

759
    h->slice_type = golomb_to_pict_type[i];
760

761 762 763 764 765 766 767
    if ((header & 0x9F) == 2) {
        i = (s->mb_num < 64) ? 6 : (1 + av_log2 (s->mb_num - 1));
        s->mb_skip_run = get_bits(&s->gb, i) - (s->mb_x + (s->mb_y * s->mb_width));
    } else {
        skip_bits1(&s->gb);
        s->mb_skip_run = 0;
    }
768

769 770 771
    h->slice_num = get_bits(&s->gb, 8);
    s->qscale = get_bits(&s->gb, 5);
    s->adaptive_quant = get_bits1(&s->gb);
772

773 774
    /* unknown fields */
    skip_bits1(&s->gb);
775

776
    if (svq3->unknown_flag) {
777 778
        skip_bits1(&s->gb);
    }
779

780 781
    skip_bits1(&s->gb);
    skip_bits(&s->gb, 2);
782

783 784 785
    while (get_bits1(&s->gb)) {
        skip_bits(&s->gb, 8);
    }
786

787
    /* reset intra predictors and invalidate motion vector references */
788
    if (s->mb_x > 0) {
789 790
        memset(h->intra4x4_pred_mode+h->mb2br_xy[mb_xy - 1      ]+3, -1, 4*sizeof(int8_t));
        memset(h->intra4x4_pred_mode+h->mb2br_xy[mb_xy - s->mb_x]  , -1, 8*sizeof(int8_t)*s->mb_x);
791 792
    }
    if (s->mb_y > 0) {
793
        memset(h->intra4x4_pred_mode+h->mb2br_xy[mb_xy - s->mb_stride], -1, 8*sizeof(int8_t)*(s->mb_width - s->mb_x));
794 795

        if (s->mb_x > 0) {
796
            h->intra4x4_pred_mode[h->mb2br_xy[mb_xy - s->mb_stride - 1]+3] = -1;
797
        }
798 799
    }

800
    return 0;
801 802
}

803
static av_cold int svq3_decode_init(AVCodecContext *avctx)
804
{
805 806 807
    SVQ3Context *svq3 = avctx->priv_data;
    H264Context *h = &svq3->h;
    MpegEncContext *s = &h->s;
808
    int m;
809 810 811
    unsigned char *extradata;
    unsigned int size;

812
    if (ff_h264_decode_init(avctx) < 0)
813 814
        return -1;

815 816 817
    s->flags  = avctx->flags;
    s->flags2 = avctx->flags2;
    s->unrestricted_mv = 1;
818
    h->is_complex=1;
819
    avctx->pix_fmt = avctx->codec->pix_fmts[0];
820 821

    if (!s->context_initialized) {
822 823
        h->chroma_qp[0] = h->chroma_qp[1] = 4;

824
        svq3->halfpel_flag  = 1;
825
        svq3->thirdpel_flag = 1;
826
        svq3->unknown_flag  = 0;
827 828 829 830 831 832 833 834 835


        /* prowl for the "SEQH" marker in the extradata */
        extradata = (unsigned char *)avctx->extradata;
        for (m = 0; m < avctx->extradata_size; m++) {
            if (!memcmp(extradata, "SEQH", 4))
                break;
            extradata++;
        }
836

837 838
        /* if a match was found, parse the extra data */
        if (extradata && !memcmp(extradata, "SEQH", 4)) {
839

840
            GetBitContext gb;
841
            int frame_size_code;
842

843 844
            size = AV_RB32(&extradata[4]);
            init_get_bits(&gb, extradata + 8, size*8);
845

846
            /* 'frame size code' and optional 'width, height' */
847 848 849 850 851 852 853 854 855 856 857 858 859
            frame_size_code = get_bits(&gb, 3);
            switch (frame_size_code) {
                case 0: avctx->width = 160; avctx->height = 120; break;
                case 1: avctx->width = 128; avctx->height =  96; break;
                case 2: avctx->width = 176; avctx->height = 144; break;
                case 3: avctx->width = 352; avctx->height = 288; break;
                case 4: avctx->width = 704; avctx->height = 576; break;
                case 5: avctx->width = 240; avctx->height = 180; break;
                case 6: avctx->width = 320; avctx->height = 240; break;
                case 7:
                    avctx->width  = get_bits(&gb, 12);
                    avctx->height = get_bits(&gb, 12);
                    break;
860
            }
861

862 863
            svq3->halfpel_flag  = get_bits1(&gb);
            svq3->thirdpel_flag = get_bits1(&gb);
864

865 866 867 868 869
            /* unknown fields */
            skip_bits1(&gb);
            skip_bits1(&gb);
            skip_bits1(&gb);
            skip_bits1(&gb);
870

871
            s->low_delay = get_bits1(&gb);
872

873 874 875 876 877 878 879
            /* unknown field */
            skip_bits1(&gb);

            while (get_bits1(&gb)) {
                skip_bits(&gb, 8);
            }

880
            svq3->unknown_flag = get_bits1(&gb);
881
            avctx->has_b_frames = !s->low_delay;
882
            if (svq3->unknown_flag) {
883
#if CONFIG_ZLIB
884 885 886 887 888 889
                unsigned watermark_width  = svq3_get_ue_golomb(&gb);
                unsigned watermark_height = svq3_get_ue_golomb(&gb);
                int u1 = svq3_get_ue_golomb(&gb);
                int u2 = get_bits(&gb, 8);
                int u3 = get_bits(&gb, 2);
                int u4 = svq3_get_ue_golomb(&gb);
890
                unsigned long buf_len = watermark_width*watermark_height*4;
891 892 893 894 895 896 897 898 899
                int offset = (get_bits_count(&gb)+7)>>3;
                uint8_t *buf;

                if ((uint64_t)watermark_width*4 > UINT_MAX/watermark_height)
                    return -1;

                buf = av_malloc(buf_len);
                av_log(avctx, AV_LOG_DEBUG, "watermark size: %dx%d\n", watermark_width, watermark_height);
                av_log(avctx, AV_LOG_DEBUG, "u1: %x u2: %x u3: %x compressed data size: %d offset: %d\n", u1, u2, u3, u4, offset);
900
                if (uncompress(buf, &buf_len, extradata + 8 + offset, size - offset) != Z_OK) {
901 902 903 904
                    av_log(avctx, AV_LOG_ERROR, "could not uncompress watermark logo\n");
                    av_free(buf);
                    return -1;
                }
905 906 907
                svq3->watermark_key = ff_svq1_packet_checksum(buf, buf_len, 0);
                svq3->watermark_key = svq3->watermark_key << 16 | svq3->watermark_key;
                av_log(avctx, AV_LOG_DEBUG, "watermark key %#x\n", svq3->watermark_key);
908
                av_free(buf);
909
#else
910 911
                av_log(avctx, AV_LOG_ERROR, "this svq3 file contains watermark which need zlib support compiled in\n");
                return -1;
912
#endif
913 914
            }
        }
915 916 917 918 919 920 921 922 923 924

        s->width  = avctx->width;
        s->height = avctx->height;

        if (MPV_common_init(s) < 0)
            return -1;

        h->b_stride = 4*s->mb_width;

        ff_h264_alloc_tables(h);
925
    }
926

927 928 929 930 931
    return 0;
}

static int svq3_decode_frame(AVCodecContext *avctx,
                             void *data, int *data_size,
932
                             AVPacket *avpkt)
933
{
934 935 936
    SVQ3Context *svq3 = avctx->priv_data;
    H264Context *h = &svq3->h;
    MpegEncContext *s = &h->s;
937
    int buf_size = avpkt->size;
938
    int m, mb_type, left;
939
    uint8_t *buf;
940

941 942 943 944 945 946 947 948
    /* special case for last picture */
    if (buf_size == 0) {
        if (s->next_picture_ptr && !s->low_delay) {
            *(AVFrame *) data = *(AVFrame *) &s->next_picture;
            s->next_picture_ptr = NULL;
            *data_size = sizeof(AVFrame);
        }
        return 0;
949
    }
950

951
    s->mb_x = s->mb_y = h->mb_xy = 0;
952

953
    if (svq3->watermark_key) {
954 955
        av_fast_malloc(&svq3->buf, &svq3->buf_size,
                       buf_size+FF_INPUT_BUFFER_PADDING_SIZE);
956 957 958 959 960 961 962 963 964 965
        if (!svq3->buf)
            return AVERROR(ENOMEM);
        memcpy(svq3->buf, avpkt->data, buf_size);
        buf = svq3->buf;
    } else {
        buf = avpkt->data;
    }

    init_get_bits(&s->gb, buf, 8*buf_size);

966
    if (svq3_decode_slice_header(avctx))
967
        return -1;
968

969 970
    s->pict_type = h->slice_type;
    s->picture_number = h->slice_num;
971

972 973
    if (avctx->debug&FF_DEBUG_PICT_INFO){
        av_log(h->s.avctx, AV_LOG_DEBUG, "%c hpel:%d, tpel:%d aqp:%d qp:%d, slice_num:%02X\n",
974
               av_get_picture_type_char(s->pict_type), svq3->halfpel_flag, svq3->thirdpel_flag,
975 976
               s->adaptive_quant, s->qscale, h->slice_num);
    }
977

978
    /* for skipping the frame */
979
    s->current_picture.pict_type = s->pict_type;
980
    s->current_picture.key_frame = (s->pict_type == AV_PICTURE_TYPE_I);
981 982

    /* Skip B-frames if we do not have reference frames. */
983
    if (s->last_picture_ptr == NULL && s->pict_type == AV_PICTURE_TYPE_B)
984
        return 0;
985 986
    if (  (avctx->skip_frame >= AVDISCARD_NONREF && s->pict_type == AV_PICTURE_TYPE_B)
        ||(avctx->skip_frame >= AVDISCARD_NONKEY && s->pict_type != AV_PICTURE_TYPE_I)
987 988 989 990
        || avctx->skip_frame >= AVDISCARD_ALL)
        return 0;

    if (s->next_p_frame_damaged) {
991
        if (s->pict_type == AV_PICTURE_TYPE_B)
992 993 994 995
            return 0;
        else
            s->next_p_frame_damaged = 0;
    }
996

997
    if (ff_h264_frame_start(h) < 0)
998
        return -1;
999

1000
    if (s->pict_type == AV_PICTURE_TYPE_B) {
1001
        h->frame_num_offset = (h->slice_num - h->prev_frame_num);
1002

1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013
        if (h->frame_num_offset < 0) {
            h->frame_num_offset += 256;
        }
        if (h->frame_num_offset == 0 || h->frame_num_offset >= h->prev_frame_num_offset) {
            av_log(h->s.avctx, AV_LOG_ERROR, "error in B-frame picture id\n");
            return -1;
        }
    } else {
        h->prev_frame_num = h->frame_num;
        h->frame_num = h->slice_num;
        h->prev_frame_num_offset = (h->frame_num - h->prev_frame_num);
1014

1015 1016 1017
        if (h->prev_frame_num_offset < 0) {
            h->prev_frame_num_offset += 256;
        }
1018 1019
    }

1020 1021 1022 1023 1024 1025 1026 1027 1028
    for (m = 0; m < 2; m++){
        int i;
        for (i = 0; i < 4; i++){
            int j;
            for (j = -1; j < 4; j++)
                h->ref_cache[m][scan8[0] + 8*i + j]= 1;
            if (i < 3)
                h->ref_cache[m][scan8[0] + 8*i + j]= PART_NOT_AVAILABLE;
        }
1029 1030
    }

1031 1032 1033
    for (s->mb_y = 0; s->mb_y < s->mb_height; s->mb_y++) {
        for (s->mb_x = 0; s->mb_x < s->mb_width; s->mb_x++) {
            h->mb_xy = s->mb_x + s->mb_y*s->mb_stride;
1034

1035 1036
            if ( (get_bits_count(&s->gb) + 7) >= s->gb.size_in_bits &&
                ((get_bits_count(&s->gb) & 7) == 0 || show_bits(&s->gb, (-get_bits_count(&s->gb) & 7)) == 0)) {
1037

1038
                skip_bits(&s->gb, svq3->next_slice_index - get_bits_count(&s->gb));
1039
                s->gb.size_in_bits = 8*buf_size;
1040

1041
                if (svq3_decode_slice_header(avctx))
1042
                    return -1;
1043

1044 1045
                /* TODO: support s->mb_skip_run */
            }
1046

1047
            mb_type = svq3_get_ue_golomb(&s->gb);
1048

1049
            if (s->pict_type == AV_PICTURE_TYPE_I) {
1050
                mb_type += 8;
1051
            } else if (s->pict_type == AV_PICTURE_TYPE_B && mb_type >= 4) {
1052 1053
                mb_type += 4;
            }
1054
            if ((unsigned)mb_type > 33 || svq3_decode_mb(svq3, mb_type)) {
1055 1056 1057
                av_log(h->s.avctx, AV_LOG_ERROR, "error while decoding MB %d %d\n", s->mb_x, s->mb_y);
                return -1;
            }
1058

1059
            if (mb_type != 0) {
1060
                ff_h264_hl_decode_mb (h);
1061
            }
1062

1063
            if (s->pict_type != AV_PICTURE_TYPE_B && !s->low_delay) {
1064
                s->current_picture.mb_type[s->mb_x + s->mb_y*s->mb_stride] =
1065
                    (s->pict_type == AV_PICTURE_TYPE_P && mb_type < 8) ? (mb_type - 1) : -1;
1066 1067
            }
        }
1068

1069
        ff_draw_horiz_band(s, 16*s->mb_y, 16);
1070
    }
1071

1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083
    left = buf_size*8 - get_bits_count(&s->gb);

    if (s->mb_y != s->mb_height || s->mb_x != s->mb_width) {
        av_log(avctx, AV_LOG_INFO, "frame num %d incomplete pic x %d y %d left %d\n", avctx->frame_number, s->mb_y, s->mb_x, left);
        //av_hex_dump(stderr, buf+buf_size-8, 8);
    }

    if (left < 0) {
        av_log(avctx, AV_LOG_ERROR, "frame num %d left %d\n", avctx->frame_number, left);
        return -1;
    }

1084
    MPV_frame_end(s);
1085

1086
    if (s->pict_type == AV_PICTURE_TYPE_B || s->low_delay) {
1087 1088 1089 1090
        *(AVFrame *) data = *(AVFrame *) &s->current_picture;
    } else {
        *(AVFrame *) data = *(AVFrame *) &s->last_picture;
    }
1091

1092 1093 1094 1095
    /* Do not output the last pic after seeking. */
    if (s->last_picture_ptr || s->low_delay) {
        *data_size = sizeof(AVFrame);
    }
1096

1097
    return buf_size;
1098 1099
}

1100 1101 1102 1103 1104 1105 1106 1107 1108 1109
static int svq3_decode_end(AVCodecContext *avctx)
{
    SVQ3Context *svq3 = avctx->priv_data;
    H264Context *h = &svq3->h;
    MpegEncContext *s = &h->s;

    ff_h264_free_context(h);

    MPV_common_end(s);

1110 1111 1112
    av_freep(&svq3->buf);
    svq3->buf_size = 0;

1113 1114
    return 0;
}
1115

1116
AVCodec ff_svq3_decoder = {
1117
    "svq3",
1118
    AVMEDIA_TYPE_VIDEO,
1119
    CODEC_ID_SVQ3,
1120
    sizeof(SVQ3Context),
1121
    svq3_decode_init,
1122
    NULL,
1123
    svq3_decode_end,
1124
    svq3_decode_frame,
1125
    CODEC_CAP_DRAW_HORIZ_BAND | CODEC_CAP_DR1 | CODEC_CAP_DELAY,
1126
    .long_name = NULL_IF_CONFIG_SMALL("Sorenson Vector Quantizer 3 / Sorenson Video 3 / SVQ3"),
1127
    .pix_fmts= (const enum PixelFormat[]){PIX_FMT_YUVJ420P, PIX_FMT_NONE},
1128
};