dvenc.c 28.9 KB
Newer Older
1 2 3 4
/*
 * DV encoder
 * Copyright (c) 2003 Roman Shaposhnik
 *
5
 * This file is part of FFmpeg.
6
 *
7
 * FFmpeg is free software; you can redistribute it and/or
8 9 10 11
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
 * version 2.1 of the License, or (at your option) any later version.
 *
12
 * FFmpeg is distributed in the hope that it will be useful,
13 14 15 16 17
 * 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
18
 * License along with FFmpeg; if not, write to the Free Software
19
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20 21
 *
 * quant_deadzone code and fixes sponsored by NOA GmbH
22 23 24 25 26 27 28
 */

/**
 * @file
 * DV encoder
 */

29 30
#include "config.h"

31
#include "libavutil/attributes.h"
32
#include "libavutil/internal.h"
33
#include "libavutil/opt.h"
34
#include "libavutil/pixdesc.h"
35

36
#include "avcodec.h"
37 38 39
#include "dv.h"
#include "dv_profile_internal.h"
#include "dv_tablegen.h"
40
#include "fdctdsp.h"
41
#include "internal.h"
42
#include "mathops.h"
43
#include "me_cmp.h"
44
#include "pixblockdsp.h"
45 46
#include "put_bits.h"

47
static av_cold int dvvideo_encode_init(AVCodecContext *avctx)
48
{
49
    DVVideoContext *s = avctx->priv_data;
50
    FDCTDSPContext fdsp;
51
    MECmpContext mecc;
52
    PixblockDSPContext pdsp;
53 54
    int ret;

55
    s->sys = av_dv_codec_profile2(avctx->width, avctx->height, avctx->pix_fmt, avctx->time_base);
56
    if (!s->sys) {
57
        av_log(avctx, AV_LOG_ERROR, "Found no DV profile for %ix%i %s video. "
58
                                    "Valid DV profiles are:\n",
59 60 61 62
               avctx->width, avctx->height, av_get_pix_fmt_name(avctx->pix_fmt));
        ff_dv_print_profiles(avctx, AV_LOG_ERROR);
        return AVERROR(EINVAL);
    }
63 64 65 66
    if (avctx->height > 576) {
        av_log(avctx, AV_LOG_ERROR, "DVCPRO HD encoding is not supported.\n");
        return AVERROR_PATCHWELCOME;
    }
67
    ret = ff_dv_init_dynamic_tables(s, s->sys);
68 69 70 71
    if (ret < 0) {
        av_log(avctx, AV_LOG_ERROR, "Error initializing work tables.\n");
        return ret;
    }
72 73 74

    dv_vlc_map_tableinit();

75
    memset(&fdsp,0, sizeof(fdsp));
76
    memset(&mecc,0, sizeof(mecc));
77
    memset(&pdsp,0, sizeof(pdsp));
78
    ff_fdctdsp_init(&fdsp, avctx);
79
    ff_me_cmp_init(&mecc, avctx);
80
    ff_pixblockdsp_init(&pdsp, avctx);
81
    ff_set_cmp(&mecc, mecc.ildct_cmp, avctx->ildct_cmp);
82

83
    s->get_pixels = pdsp.get_pixels;
84
    s->ildct_cmp  = mecc.ildct_cmp[5];
85

86 87
    s->fdct[0]    = fdsp.fdct;
    s->fdct[1]    = fdsp.fdct248;
88

89 90 91 92
    return ff_dvvideo_init(avctx);
}

/* bit budget for AC only in 5 MBs */
93
static const int vs_total_ac_bits = (100 * 4 + 68 * 2) * 5;
94 95 96
static const int mb_area_start[5] = { 1, 6, 21, 43, 64 };

#if CONFIG_SMALL
97 98 99
/* Convert run and level (where level != 0) pair into VLC, returning bit size */
static av_always_inline int dv_rl2vlc(int run, int level, int sign,
                                      uint32_t *vlc)
100 101 102 103 104
{
    int size;
    if (run < DV_VLC_MAP_RUN_SIZE && level < DV_VLC_MAP_LEV_SIZE) {
        *vlc = dv_vlc_map[run][level].vlc | sign;
        size = dv_vlc_map[run][level].size;
105
    } else {
106 107 108 109 110 111 112 113
        if (level < DV_VLC_MAP_LEV_SIZE) {
            *vlc = dv_vlc_map[0][level].vlc | sign;
            size = dv_vlc_map[0][level].size;
        } else {
            *vlc = 0xfe00 | (level << 1) | sign;
            size = 16;
        }
        if (run) {
114 115 116
            *vlc |= ((run < 16) ? dv_vlc_map[run - 1][0].vlc :
                     (0x1f80 | (run - 1))) << size;
            size +=  (run < 16) ? dv_vlc_map[run - 1][0].size : 13;
117 118 119 120 121 122 123 124 125 126 127 128
        }
    }

    return size;
}

static av_always_inline int dv_rl2vlc_size(int run, int level)
{
    int size;

    if (run < DV_VLC_MAP_RUN_SIZE && level < DV_VLC_MAP_LEV_SIZE) {
        size = dv_vlc_map[run][level].size;
129
    } else {
130
        size = (level < DV_VLC_MAP_LEV_SIZE) ? dv_vlc_map[0][level].size : 16;
131 132
        if (run)
            size += (run < 16) ? dv_vlc_map[run - 1][0].size : 13;
133 134 135 136
    }
    return size;
}
#else
137
static av_always_inline int dv_rl2vlc(int run, int l, int sign, uint32_t *vlc)
138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162
{
    *vlc = dv_vlc_map[run][l].vlc | sign;
    return dv_vlc_map[run][l].size;
}

static av_always_inline int dv_rl2vlc_size(int run, int l)
{
    return dv_vlc_map[run][l].size;
}
#endif

typedef struct EncBlockInfo {
    int      area_q[4];
    int      bit_size[4];
    int      prev[5];
    int      cur_ac;
    int      cno;
    int      dct_mode;
    int16_t  mb[64];
    uint8_t  next[64];
    uint8_t  sign[64];
    uint8_t  partial_bit_count;
    uint32_t partial_bit_buffer; /* we can't use uint16_t here */
} EncBlockInfo;

163 164 165
static av_always_inline PutBitContext *dv_encode_ac(EncBlockInfo *bi,
                                                    PutBitContext *pb_pool,
                                                    PutBitContext *pb_end)
166 167
{
    int prev, bits_left;
168 169 170 171 172 173 174 175 176 177 178 179
    PutBitContext *pb = pb_pool;
    int size          = bi->partial_bit_count;
    uint32_t vlc      = bi->partial_bit_buffer;

    bi->partial_bit_count  =
    bi->partial_bit_buffer = 0;
    for (;;) {
        /* Find suitable storage space */
        for (; size > (bits_left = put_bits_left(pb)); pb++) {
            if (bits_left) {
                size -= bits_left;
                put_bits(pb, bits_left, vlc >> size);
180
                vlc = av_mod_uintp2(vlc, size);
181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204
            }
            if (pb + 1 >= pb_end) {
                bi->partial_bit_count  = size;
                bi->partial_bit_buffer = vlc;
                return pb;
            }
        }

        /* Store VLC */
        put_bits(pb, size, vlc);

        if (bi->cur_ac >= 64)
            break;

        /* Construct the next VLC */
        prev       = bi->cur_ac;
        bi->cur_ac = bi->next[prev];
        if (bi->cur_ac < 64) {
            size = dv_rl2vlc(bi->cur_ac - prev - 1, bi->mb[bi->cur_ac],
                             bi->sign[bi->cur_ac], &vlc);
        } else {
            size = 4;
            vlc  = 6; /* End Of Block stamp */
        }
205 206 207 208
    }
    return pb;
}

209
static av_always_inline int dv_guess_dct_mode(DVVideoContext *s, uint8_t *data,
210
                                              ptrdiff_t linesize)
211
{
212
    if (s->avctx->flags & AV_CODEC_FLAG_INTERLACED_DCT) {
213 214
        int ps = s->ildct_cmp(NULL, data, NULL, linesize, 8) - 400;
        if (ps > 0) {
215 216
            int is = s->ildct_cmp(NULL, data,            NULL, linesize << 1, 4) +
                     s->ildct_cmp(NULL, data + linesize, NULL, linesize << 1, 4);
217 218 219 220 221 222 223 224 225
            return ps > is;
        }
    }

    return 0;
}

static const int dv_weight_bits = 18;
static const int dv_weight_88[64] = {
226 227 228 229 230 231 232 233
    131072, 257107, 257107, 242189, 252167, 242189, 235923, 237536,
    237536, 235923, 229376, 231390, 223754, 231390, 229376, 222935,
    224969, 217965, 217965, 224969, 222935, 200636, 218652, 211916,
    212325, 211916, 218652, 200636, 188995, 196781, 205965, 206433,
    206433, 205965, 196781, 188995, 185364, 185364, 200636, 200704,
    200636, 185364, 185364, 174609, 180568, 195068, 195068, 180568,
    174609, 170091, 175557, 189591, 175557, 170091, 165371, 170627,
    170627, 165371, 160727, 153560, 160727, 144651, 144651, 136258,
234 235
};
static const int dv_weight_248[64] = {
236 237 238 239 240 241 242 243
    131072, 262144, 257107, 257107, 242189, 242189, 242189, 242189,
    237536, 237536, 229376, 229376, 200636, 200636, 224973, 224973,
    223754, 223754, 235923, 235923, 229376, 229376, 217965, 217965,
    211916, 211916, 196781, 196781, 185364, 185364, 206433, 206433,
    211916, 211916, 222935, 222935, 200636, 200636, 205964, 205964,
    200704, 200704, 180568, 180568, 175557, 175557, 195068, 195068,
    185364, 185364, 188995, 188995, 174606, 174606, 175557, 175557,
    170627, 170627, 153560, 153560, 165371, 165371, 144651, 144651,
244 245
};

246
static av_always_inline int dv_init_enc_block(EncBlockInfo *bi, uint8_t *data,
247 248
                                              ptrdiff_t linesize,
                                              DVVideoContext *s, int bias)
249 250
{
    const int *weight;
251
    const uint8_t *zigzag_scan;
252 253 254
    LOCAL_ALIGNED_16(int16_t, blk, [64]);
    int i, area;
    /* We offer two different methods for class number assignment: the
255 256 257
     * method suggested in SMPTE 314M Table 22, and an improved
     * method. The SMPTE method is very conservative; it assigns class
     * 3 (i.e. severe quantization) to any block where the largest AC
258
     * component is greater than 36. FFmpeg's DV encoder tracks AC bit
259 260 261 262
     * consumption precisely, so there is no need to bias most blocks
     * towards strongly lossy compression. Instead, we assign class 2
     * to most blocks, and use class 3 only when strictly necessary
     * (for blocks whose largest AC component exceeds 255). */
263 264

#if 0 /* SMPTE spec method */
265
    static const int classes[] = { 12, 24, 36, 0xffff };
266
#else /* improved FFmpeg method */
267
    static const int classes[] = { -1, -1, 255, 0xffff };
268 269 270
#endif
    int max  = classes[0];
    int prev = 0;
271 272
    const unsigned deadzone = s->quant_deadzone;
    const unsigned threshold = 2 * deadzone;
273

274
    av_assert2((((int) blk) & 15) == 0);
275

276 277 278 279 280
    bi->area_q[0]          =
    bi->area_q[1]          =
    bi->area_q[2]          =
    bi->area_q[3]          = 0;
    bi->partial_bit_count  = 0;
281
    bi->partial_bit_buffer = 0;
282
    bi->cur_ac             = 0;
283 284 285 286 287
    if (data) {
        bi->dct_mode = dv_guess_dct_mode(s, data, linesize);
        s->get_pixels(blk, data, linesize);
        s->fdct[bi->dct_mode](blk);
    } else {
288 289 290 291
        /* We rely on the fact that encoding all zeros leads to an immediate
         * EOB, which is precisely what the spec calls for in the "dummy"
         * blocks. */
        memset(blk, 0, 64 * sizeof(*blk));
292 293 294 295
        bi->dct_mode = 0;
    }
    bi->mb[0] = blk[0];

296
    zigzag_scan = bi->dct_mode ? ff_dv_zigzag248_direct : ff_zigzag_direct;
297
    weight      = bi->dct_mode ? dv_weight_248 : dv_weight_88;
298 299

    for (area = 0; area < 4; area++) {
300 301 302 303 304
        bi->prev[area]     = prev;
        bi->bit_size[area] = 1; // 4 areas 4 bits for EOB :)
        for (i = mb_area_start[area]; i < mb_area_start[area + 1]; i++) {
            int level = blk[zigzag_scan[i]];

305
            if (level + deadzone > threshold) {
306
                bi->sign[i] = (level >> 31) & 1;
307
                /* Weight it and shift down into range, adding for rounding.
308 309 310 311
                 * The extra division by a factor of 2^4 reverses the 8x
                 * expansion of the DCT AND the 2x doubling of the weights. */
                level     = (FFABS(level) * weight[i] + (1 << (dv_weight_bits + 3))) >>
                            (dv_weight_bits + 4);
312 313
                if (!level)
                    continue;
314 315 316 317 318 319 320 321
                bi->mb[i] = level;
                if (level > max)
                    max = level;
                bi->bit_size[area] += dv_rl2vlc_size(i - prev - 1, level);
                bi->next[prev]      = i;
                prev                = i;
            }
        }
322
    }
323 324 325
    bi->next[prev] = i;
    for (bi->cno = 0; max > classes[bi->cno]; bi->cno++)
        ;
326 327 328 329 330 331 332 333 334 335

    bi->cno += bias;

    if (bi->cno >= 3) {
        bi->cno = 3;
        prev    = 0;
        i       = bi->next[prev];
        for (area = 0; area < 4; area++) {
            bi->prev[area]     = prev;
            bi->bit_size[area] = 1; // 4 areas 4 bits for EOB :)
336
            for (; i < mb_area_start[area + 1]; i = bi->next[i]) {
337 338 339 340
                bi->mb[i] >>= 1;

                if (bi->mb[i]) {
                    bi->bit_size[area] += dv_rl2vlc_size(i - prev - 1, bi->mb[i]);
341 342
                    bi->next[prev]      = i;
                    prev                = i;
343 344 345
                }
            }
        }
346
        bi->next[prev] = i;
347 348
    }

349 350
    return bi->bit_size[0] + bi->bit_size[1] +
           bi->bit_size[2] + bi->bit_size[3];
351 352
}

353
static inline void dv_guess_qnos(EncBlockInfo *blks, int *qnos)
354 355 356
{
    int size[5];
    int i, j, k, a, prev, a2;
357
    EncBlockInfo *b;
358

359 360 361 362 363
    size[0] =
    size[1] =
    size[2] =
    size[3] =
    size[4] = 1 << 24;
364
    do {
365 366 367 368 369 370 371 372 373 374 375 376 377
        b = blks;
        for (i = 0; i < 5; i++) {
            if (!qnos[i])
                continue;

            qnos[i]--;
            size[i] = 0;
            for (j = 0; j < 6; j++, b++) {
                for (a = 0; a < 4; a++) {
                    if (b->area_q[a] != ff_dv_quant_shifts[qnos[i] + ff_dv_quant_offset[b->cno]][a]) {
                        b->bit_size[a] = 1; // 4 areas 4 bits for EOB :)
                        b->area_q[a]++;
                        prev = b->prev[a];
378
                        av_assert2(b->next[prev] >= mb_area_start[a + 1] || b->mb[prev]);
379 380 381 382 383 384 385 386 387
                        for (k = b->next[prev]; k < mb_area_start[a + 1]; k = b->next[k]) {
                            b->mb[k] >>= 1;
                            if (b->mb[k]) {
                                b->bit_size[a] += dv_rl2vlc_size(k - prev - 1, b->mb[k]);
                                prev            = k;
                            } else {
                                if (b->next[k] >= mb_area_start[a + 1] && b->next[k] < 64) {
                                    for (a2 = a + 1; b->next[k] >= mb_area_start[a2 + 1]; a2++)
                                        b->prev[a2] = prev;
388 389
                                    av_assert2(a2 < 4);
                                    av_assert2(b->mb[b->next[k]]);
390 391
                                    b->bit_size[a2] += dv_rl2vlc_size(b->next[k] - prev - 1, b->mb[b->next[k]]) -
                                                       dv_rl2vlc_size(b->next[k] - k    - 1, b->mb[b->next[k]]);
392
                                    av_assert2(b->prev[a2] == k && (a2 + 1 >= 4 || b->prev[a2 + 1] != k));
393
                                    b->prev[a2] = prev;
394 395 396 397 398
                                }
                                b->next[prev] = b->next[k];
                            }
                        }
                        b->prev[a + 1] = prev;
399
                    }
400
                    size[i] += b->bit_size[a];
401
                }
402 403
            }
            if (vs_total_ac_bits >= size[0] + size[1] + size[2] + size[3] + size[4])
404
                return;
405 406
        }
    } while (qnos[0] | qnos[1] | qnos[2] | qnos[3] | qnos[4]);
407

408 409 410 411
    for (a = 2; a == 2 || vs_total_ac_bits < size[0]; a += a) {
        b       = blks;
        size[0] = 5 * 6 * 4; // EOB
        for (j = 0; j < 6 * 5; j++, b++) {
412 413
            prev = b->prev[0];
            for (k = b->next[prev]; k < 64; k = b->next[k]) {
414
                if (b->mb[k] < a && b->mb[k] > -a) {
415
                    b->next[prev] = b->next[k];
416
                } else {
417
                    size[0] += dv_rl2vlc_size(k - prev - 1, b->mb[k]);
418
                    prev     = k;
419 420 421 422 423 424 425 426 427 428 429
                }
            }
        }
    }
}

static int dv_encode_video_segment(AVCodecContext *avctx, void *arg)
{
    DVVideoContext *s = avctx->priv_data;
    DVwork_chunk *work_chunk = arg;
    int mb_index, i, j;
430 431
    int mb_x, mb_y, c_offset;
    ptrdiff_t linesize, y_stride;
432 433
    uint8_t *y_ptr;
    uint8_t *dif;
434
    LOCAL_ALIGNED_8(uint8_t, scratch, [128]);
435 436 437 438 439 440 441 442 443
    EncBlockInfo enc_blks[5 * DV_MAX_BPM];
    PutBitContext pbs[5 * DV_MAX_BPM];
    PutBitContext *pb;
    EncBlockInfo *enc_blk;
    int vs_bit_size = 0;
    int qnos[5] = { 15, 15, 15, 15, 15 }; /* No quantization */
    int *qnosp = &qnos[0];

    dif     = &s->buf[work_chunk->buf_offset * 80];
444 445 446 447 448
    enc_blk = &enc_blks[0];
    for (mb_index = 0; mb_index < 5; mb_index++) {
        dv_calculate_mb_xy(s, work_chunk, mb_index, &mb_x, &mb_y);

        /* initializing luminance blocks */
449
        if ((s->sys->pix_fmt == AV_PIX_FMT_YUV420P)                      ||
450 451
            (s->sys->pix_fmt == AV_PIX_FMT_YUV411P && mb_x >= (704 / 8)) ||
            (s->sys->height >= 720 && mb_y != 134)) {
452
            y_stride = s->frame->linesize[0] << 3;
453 454 455
        } else {
            y_stride = 16;
        }
456 457
        y_ptr    = s->frame->data[0] +
                   ((mb_y * s->frame->linesize[0] + mb_x) << 3);
458
        linesize = s->frame->linesize[0];
459 460 461

        if (s->sys->video_stype == 4) { /* SD 422 */
            vs_bit_size +=
462 463 464 465
                dv_init_enc_block(enc_blk + 0, y_ptr,                linesize, s, 0) +
                dv_init_enc_block(enc_blk + 1, NULL,                 linesize, s, 0) +
                dv_init_enc_block(enc_blk + 2, y_ptr + 8,            linesize, s, 0) +
                dv_init_enc_block(enc_blk + 3, NULL,                 linesize, s, 0);
466 467
        } else {
            vs_bit_size +=
468 469 470 471
                dv_init_enc_block(enc_blk + 0, y_ptr,                linesize, s, 0) +
                dv_init_enc_block(enc_blk + 1, y_ptr + 8,            linesize, s, 0) +
                dv_init_enc_block(enc_blk + 2, y_ptr +     y_stride, linesize, s, 0) +
                dv_init_enc_block(enc_blk + 3, y_ptr + 8 + y_stride, linesize, s, 0);
472 473 474 475
        }
        enc_blk += 4;

        /* initializing chrominance blocks */
476
        c_offset = (((mb_y >>  (s->sys->pix_fmt == AV_PIX_FMT_YUV420P)) * s->frame->linesize[1] +
477 478
                     (mb_x >> ((s->sys->pix_fmt == AV_PIX_FMT_YUV411P) ? 2 : 1))) << 3);
        for (j = 2; j; j--) {
479 480 481
            uint8_t *c_ptr = s->frame->data[j] + c_offset;
            linesize = s->frame->linesize[j];
            y_stride = (mb_y == 134) ? 8 : (s->frame->linesize[j] << 3);
482
            if (s->sys->pix_fmt == AV_PIX_FMT_YUV411P && mb_x >= (704 / 8)) {
483 484
                uint8_t *d;
                uint8_t *b = scratch;
485
                for (i = 0; i < 8; i++) {
486 487 488 489 490 491 492 493 494
                    d      = c_ptr + (linesize << 3);
                    b[0]   = c_ptr[0];
                    b[1]   = c_ptr[1];
                    b[2]   = c_ptr[2];
                    b[3]   = c_ptr[3];
                    b[4]   = d[0];
                    b[5]   = d[1];
                    b[6]   = d[2];
                    b[7]   = d[3];
495
                    c_ptr += linesize;
496
                    b     += 16;
497
                }
498
                c_ptr    = scratch;
499 500 501
                linesize = 16;
            }

502 503 504 505
            vs_bit_size += dv_init_enc_block(enc_blk++, c_ptr, linesize, s, 1);
            if (s->sys->bpm == 8)
                vs_bit_size += dv_init_enc_block(enc_blk++, c_ptr + y_stride,
                                                 linesize, s, 1);
506 507 508 509 510 511 512
        }
    }

    if (vs_total_ac_bits < vs_bit_size)
        dv_guess_qnos(&enc_blks[0], qnosp);

    /* DIF encoding process */
513
    for (j = 0; j < 5 * s->sys->bpm;) {
514 515 516
        int start_mb = j;

        dif[3] = *qnosp++;
517
        dif   += 4;
518 519

        /* First pass over individual cells only */
520 521
        for (i = 0; i < s->sys->bpm; i++, j++) {
            int sz = s->sys->block_sizes[i] >> 3;
522 523 524 525 526 527

            init_put_bits(&pbs[j], dif, sz);
            put_sbits(&pbs[j], 9, ((enc_blks[j].mb[0] >> 3) - 1024 + 2) >> 2);
            put_bits(&pbs[j], 1, enc_blks[j].dct_mode);
            put_bits(&pbs[j], 2, enc_blks[j].cno);

528
            dv_encode_ac(&enc_blks[j], &pbs[j], &pbs[j + 1]);
529 530 531 532 533
            dif += sz;
        }

        /* Second pass over each MB space */
        pb = &pbs[start_mb];
534 535 536 537
        for (i = 0; i < s->sys->bpm; i++)
            if (enc_blks[start_mb + i].partial_bit_count)
                pb = dv_encode_ac(&enc_blks[start_mb + i], pb,
                                  &pbs[start_mb + s->sys->bpm]);
538 539 540 541
    }

    /* Third and final pass over the whole video segment space */
    pb = &pbs[0];
542 543 544 545
    for (j = 0; j < 5 * s->sys->bpm; j++) {
        if (enc_blks[j].partial_bit_count)
            pb = dv_encode_ac(&enc_blks[j], pb, &pbs[s->sys->bpm * 5]);
        if (enc_blks[j].partial_bit_count)
546 547 548
            av_log(avctx, AV_LOG_ERROR, "ac bitstream overflow\n");
    }

549 550 551 552 553 554 555 556 557 558 559
    for (j = 0; j < 5 * s->sys->bpm; j++) {
        int pos;
        int size = pbs[j].size_in_bits >> 3;
        flush_put_bits(&pbs[j]);
        pos = put_bits_count(&pbs[j]) >> 3;
        if (pos > size) {
            av_log(avctx, AV_LOG_ERROR,
                   "bitstream written beyond buffer size\n");
            return -1;
        }
        memset(pbs[j].buf + pos, 0xff, size - pos);
560 561 562 563 564 565
    }

    return 0;
}

static inline int dv_write_pack(enum dv_pack_type pack_id, DVVideoContext *c,
566
                                uint8_t *buf)
567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585
{
    /*
     * Here's what SMPTE314M says about these two:
     *    (page 6) APTn, AP1n, AP2n, AP3n: These data shall be identical
     *             as track application IDs (APTn = 001, AP1n =
     *             001, AP2n = 001, AP3n = 001), if the source signal
     *             comes from a digital VCR. If the signal source is
     *             unknown, all bits for these data shall be set to 1.
     *    (page 12) STYPE: STYPE defines a signal type of video signal
     *                     00000b = 4:1:1 compression
     *                     00100b = 4:2:2 compression
     *                     XXXXXX = Reserved
     * Now, I've got two problems with these statements:
     *   1. it looks like APT == 111b should be a safe bet, but it isn't.
     *      It seems that for PAL as defined in IEC 61834 we have to set
     *      APT to 000 and for SMPTE314M to 001.
     *   2. It is not at all clear what STYPE is used for 4:2:0 PAL
     *      compression scheme (if any).
     */
586
    int apt = (c->sys->pix_fmt == AV_PIX_FMT_YUV420P ? 0 : 1);
587
    int fs  = c->frame->top_field_first ? 0x00 : 0x40;
588 589

    uint8_t aspect = 0;
590 591
    if ((int) (av_q2d(c->avctx->sample_aspect_ratio) *
               c->avctx->width / c->avctx->height * 10) >= 17) /* 16:9 */
592 593
        aspect = 0x02;

594
    buf[0] = (uint8_t) pack_id;
595 596 597
    switch (pack_id) {
    case dv_header525: /* I can't imagine why these two weren't defined as real */
    case dv_header625: /* packs in SMPTE314M -- they definitely look like ones */
598 599 600 601 602 603 604 605 606 607 608 609
        buf[1] =  0xf8       | /* reserved -- always 1 */
                 (apt & 0x07); /* APT: Track application ID */
        buf[2] = (0    << 7) | /* TF1: audio data is 0 - valid; 1 - invalid */
                 (0x0f << 3) | /* reserved -- always 1 */
                 (apt & 0x07); /* AP1: Audio application ID */
        buf[3] = (0    << 7) | /* TF2: video data is 0 - valid; 1 - invalid */
                 (0x0f << 3) | /* reserved -- always 1 */
                 (apt & 0x07); /* AP2: Video application ID */
        buf[4] = (0    << 7) | /* TF3: subcode(SSYB) is 0 - valid; 1 - invalid */
                 (0x0f << 3) | /* reserved -- always 1 */
                 (apt & 0x07); /* AP3: Subcode application ID */
        break;
610
    case dv_video_source:
611 612 613 614 615 616 617 618 619 620
        buf[1] = 0xff;         /* reserved -- always 1 */
        buf[2] = (1 << 7) |    /* B/W: 0 - b/w, 1 - color */
                 (1 << 6) |    /* following CLF is valid - 0, invalid - 1 */
                 (3 << 4) |    /* CLF: color frames ID (see ITU-R BT.470-4) */
                 0xf;          /* reserved -- always 1 */
        buf[3] = (3 << 6)           | /* reserved -- always 1 */
                 (c->sys->dsf << 5) | /*  system: 60fields/50fields */
                 c->sys->video_stype; /* signal type video compression */
        buf[4] = 0xff;         /* VISC: 0xff -- no information */
        break;
621
    case dv_video_control:
622 623 624 625 626
        buf[1] = (0 << 6) |    /* Copy generation management (CGMS) 0 -- free */
                 0x3f;         /* reserved -- always 1 */
        buf[2] = 0xc8 |        /* reserved -- always b11001xxx */
                 aspect;
        buf[3] = (1 << 7) |    /* frame/field flag 1 -- frame, 0 -- field */
627
                 fs       |    /* first/second field flag 0 -- field 2, 1 -- field 1 */
628 629 630 631 632
                 (1 << 5) |    /* frame change flag 0 -- same picture as before, 1 -- different */
                 (1 << 4) |    /* 1 - interlaced, 0 - noninterlaced */
                 0xc;          /* reserved -- always b1100 */
        buf[4] = 0xff;         /* reserved -- always 1 */
        break;
633
    default:
634 635 636 637
        buf[1] =
        buf[2] =
        buf[3] =
        buf[4] = 0xff;
638 639 640 641 642 643
    }
    return 5;
}

static inline int dv_write_dif_id(enum dv_section_type t, uint8_t chan_num,
                                  uint8_t seq_num, uint8_t dif_num,
644
                                  uint8_t *buf)
645
{
646
    buf[0] = (uint8_t) t;      /* Section type */
647 648 649 650 651 652 653
    buf[1] = (seq_num  << 4) | /* DIF seq number 0-9 for 525/60; 0-11 for 625/50 */
             (chan_num << 3) | /* FSC: for 50Mb/s 0 - first channel; 1 - second */
             7;                /* reserved -- always 1 */
    buf[2] = dif_num;          /* DIF block number Video: 0-134, Audio: 0-8 */
    return 3;
}

654
static inline int dv_write_ssyb_id(uint8_t syb_num, uint8_t fr, uint8_t *buf)
655 656 657 658 659
{
    if (syb_num == 0 || syb_num == 6) {
        buf[0] = (fr << 7) | /* FR ID 1 - first half of each channel; 0 - second */
                 (0  << 4) | /* AP3 (Subcode application ID) */
                 0x0f;       /* reserved -- always 1 */
660
    } else if (syb_num == 11) {
661 662
        buf[0] = (fr << 7) | /* FR ID 1 - first half of each channel; 0 - second */
                 0x7f;       /* reserved -- always 1 */
663
    } else {
664 665 666 667 668 669 670 671 672 673
        buf[0] = (fr << 7) | /* FR ID 1 - first half of each channel; 0 - second */
                 (0  << 4) | /* APT (Track application ID) */
                 0x0f;       /* reserved -- always 1 */
    }
    buf[1] = 0xf0 |            /* reserved -- always 1 */
             (syb_num & 0x0f); /* SSYB number 0 - 11   */
    buf[2] = 0xff;             /* reserved -- always 1 */
    return 3;
}

674
static void dv_format_frame(DVVideoContext *c, uint8_t *buf)
675 676 677 678 679 680 681 682 683
{
    int chan, i, j, k;

    for (chan = 0; chan < c->sys->n_difchan; chan++) {
        for (i = 0; i < c->sys->difseg_size; i++) {
            memset(buf, 0xff, 80 * 6); /* first 6 DIF blocks are for control data */

            /* DV header: 1DIF */
            buf += dv_write_dif_id(dv_sect_header, chan, i, 0, buf);
684 685
            buf += dv_write_pack((c->sys->dsf ? dv_header625 : dv_header525),
                                 c, buf);
686 687 688 689 690 691
            buf += 72; /* unused bytes */

            /* DV subcode: 2DIFs */
            for (j = 0; j < 2; j++) {
                buf += dv_write_dif_id(dv_sect_subcode, chan, i, j, buf);
                for (k = 0; k < 6; k++)
692
                    buf += dv_write_ssyb_id(k, (i < c->sys->difseg_size / 2), buf) + 5;
693 694 695 696 697 698 699 700
                buf += 29; /* unused bytes */
            }

            /* DV VAUX: 3DIFS */
            for (j = 0; j < 3; j++) {
                buf += dv_write_dif_id(dv_sect_vaux, chan, i, j, buf);
                buf += dv_write_pack(dv_video_source,  c, buf);
                buf += dv_write_pack(dv_video_control, c, buf);
701
                buf += 7 * 5;
702 703
                buf += dv_write_pack(dv_video_source,  c, buf);
                buf += dv_write_pack(dv_video_control, c, buf);
704
                buf += 4 * 5 + 2; /* unused bytes */
705 706 707 708
            }

            /* DV Audio/Video: 135 Video DIFs + 9 Audio DIFs */
            for (j = 0; j < 135; j++) {
709
                if (j % 15 == 0) {
710
                    memset(buf, 0xff, 80);
711
                    buf += dv_write_dif_id(dv_sect_audio, chan, i, j / 15, buf);
712 713 714 715
                    buf += 77; /* audio control & shuffled PCM audio */
                }
                buf += dv_write_dif_id(dv_sect_video, chan, i, j, buf);
                buf += 77; /* 1 video macroblock: 1 bytes control
716 717 718
                            * 4 * 14 bytes Y 8x8 data
                            * 10 bytes Cr 8x8 data
                            * 10 bytes Cb 8x8 data */
719 720 721 722 723 724 725 726 727 728 729
            }
        }
    }
}

static int dvvideo_encode_frame(AVCodecContext *c, AVPacket *pkt,
                                const AVFrame *frame, int *got_packet)
{
    DVVideoContext *s = c->priv_data;
    int ret;

730
    if ((ret = ff_alloc_packet2(c, pkt, s->sys->frame_size, 0)) < 0)
731 732
        return ret;

733 734
    c->pix_fmt                = s->sys->pix_fmt;
    s->frame                  = frame;
735 736
#if FF_API_CODED_FRAME
FF_DISABLE_DEPRECATION_WARNINGS
737 738
    c->coded_frame->key_frame = 1;
    c->coded_frame->pict_type = AV_PICTURE_TYPE_I;
739 740
FF_ENABLE_DEPRECATION_WARNINGS
#endif
741 742

    s->buf = pkt->data;
743
    c->execute(c, dv_encode_video_segment, s->work_chunks, NULL,
744 745 746 747 748 749 750 751 752 753 754 755
               dv_work_pool_size(s->sys), sizeof(DVwork_chunk));

    emms_c();

    dv_format_frame(s, pkt->data);

    pkt->flags |= AV_PKT_FLAG_KEY;
    *got_packet = 1;

    return 0;
}

756 757 758
#define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
#define OFFSET(x) offsetof(DVVideoContext, x)
static const AVOption dv_options[] = {
759
    { "quant_deadzone",        "Quantizer dead zone",    OFFSET(quant_deadzone),       AV_OPT_TYPE_INT, { .i64 = 7 }, 0, 1024, VE },
760 761 762 763 764 765 766 767 768 769
    { NULL },
};

static const AVClass dvvideo_encode_class = {
    .class_name = "dvvideo encoder",
    .item_name  = av_default_item_name,
    .option     = dv_options,
    .version    = LIBAVUTIL_VERSION_INT,
};

770 771 772 773 774 775
AVCodec ff_dvvideo_encoder = {
    .name           = "dvvideo",
    .long_name      = NULL_IF_CONFIG_SMALL("DV (Digital Video)"),
    .type           = AVMEDIA_TYPE_VIDEO,
    .id             = AV_CODEC_ID_DVVIDEO,
    .priv_data_size = sizeof(DVVideoContext),
776
    .init           = dvvideo_encode_init,
777
    .encode2        = dvvideo_encode_frame,
778
    .capabilities   = AV_CODEC_CAP_SLICE_THREADS | AV_CODEC_CAP_FRAME_THREADS | AV_CODEC_CAP_INTRA_ONLY,
779
    .pix_fmts       = (const enum AVPixelFormat[]) {
780 781
        AV_PIX_FMT_YUV411P, AV_PIX_FMT_YUV422P,
        AV_PIX_FMT_YUV420P, AV_PIX_FMT_NONE
782
    },
783
    .priv_class     = &dvvideo_encode_class,
784
};