mpegvideo.c 86.5 KB
Newer Older
Fabrice Bellard's avatar
Fabrice Bellard committed
1 2
/*
 * The simplest mpeg encoder (well, it was the simplest!)
3
 * Copyright (c) 2000,2001 Fabrice Bellard
4
 * Copyright (c) 2002-2004 Michael Niedermayer <michaelni@gmx.at>
Fabrice Bellard's avatar
Fabrice Bellard committed
5
 *
6 7
 * 4MV & hq & B-frame encoding stuff by Michael Niedermayer <michaelni@gmx.at>
 *
8 9 10
 * This file is part of FFmpeg.
 *
 * FFmpeg is free software; you can redistribute it and/or
11 12
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
13
 * version 2.1 of the License, or (at your option) any later version.
Fabrice Bellard's avatar
Fabrice Bellard committed
14
 *
15
 * FFmpeg is distributed in the hope that it will be useful,
Fabrice Bellard's avatar
Fabrice Bellard committed
16
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
17 18
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * Lesser General Public License for more details.
Fabrice Bellard's avatar
Fabrice Bellard committed
19
 *
20
 * You should have received a copy of the GNU Lesser General Public
21
 * License along with FFmpeg; if not, write to the Free Software
22
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
Fabrice Bellard's avatar
Fabrice Bellard committed
23
 */
24

Michael Niedermayer's avatar
Michael Niedermayer committed
25
/**
26
 * @file
Michael Niedermayer's avatar
Michael Niedermayer committed
27
 * The simplest mpeg encoder (well, it was the simplest!).
28 29
 */

30
#include "libavutil/attributes.h"
31
#include "libavutil/avassert.h"
32
#include "libavutil/imgutils.h"
33
#include "libavutil/internal.h"
34
#include "libavutil/motion_vector.h"
35
#include "libavutil/timer.h"
Fabrice Bellard's avatar
Fabrice Bellard committed
36
#include "avcodec.h"
37
#include "blockdsp.h"
38
#include "h264chroma.h"
39
#include "idctdsp.h"
40
#include "internal.h"
41
#include "mathops.h"
42
#include "mpeg_er.h"
43
#include "mpegutils.h"
Fabrice Bellard's avatar
Fabrice Bellard committed
44
#include "mpegvideo.h"
45
#include "mpegvideodata.h"
46
#include "mjpegenc.h"
47
#include "msmpeg4.h"
48
#include "qpeldsp.h"
49
#include "thread.h"
50
#include "wmv2.h"
51
#include <limits.h>
Fabrice Bellard's avatar
Fabrice Bellard committed
52

53 54 55 56 57 58 59 60
static void dct_unquantize_mpeg1_intra_c(MpegEncContext *s,
                                   int16_t *block, int n, int qscale)
{
    int i, level, nCoeffs;
    const uint16_t *quant_matrix;

    nCoeffs= s->block_last_index[n];

61
    block[0] *= n < 4 ? s->y_dc_scale : s->c_dc_scale;
62
    /* XXX: only MPEG-1 */
63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116
    quant_matrix = s->intra_matrix;
    for(i=1;i<=nCoeffs;i++) {
        int j= s->intra_scantable.permutated[i];
        level = block[j];
        if (level) {
            if (level < 0) {
                level = -level;
                level = (int)(level * qscale * quant_matrix[j]) >> 3;
                level = (level - 1) | 1;
                level = -level;
            } else {
                level = (int)(level * qscale * quant_matrix[j]) >> 3;
                level = (level - 1) | 1;
            }
            block[j] = level;
        }
    }
}

static void dct_unquantize_mpeg1_inter_c(MpegEncContext *s,
                                   int16_t *block, int n, int qscale)
{
    int i, level, nCoeffs;
    const uint16_t *quant_matrix;

    nCoeffs= s->block_last_index[n];

    quant_matrix = s->inter_matrix;
    for(i=0; i<=nCoeffs; i++) {
        int j= s->intra_scantable.permutated[i];
        level = block[j];
        if (level) {
            if (level < 0) {
                level = -level;
                level = (((level << 1) + 1) * qscale *
                         ((int) (quant_matrix[j]))) >> 4;
                level = (level - 1) | 1;
                level = -level;
            } else {
                level = (((level << 1) + 1) * qscale *
                         ((int) (quant_matrix[j]))) >> 4;
                level = (level - 1) | 1;
            }
            block[j] = level;
        }
    }
}

static void dct_unquantize_mpeg2_intra_c(MpegEncContext *s,
                                   int16_t *block, int n, int qscale)
{
    int i, level, nCoeffs;
    const uint16_t *quant_matrix;

117 118
    if (s->q_scale_type) qscale = ff_mpeg2_non_linear_qscale[qscale];
    else                 qscale <<= 1;
119

120 121 122
    if(s->alternate_scan) nCoeffs= 63;
    else nCoeffs= s->block_last_index[n];

123
    block[0] *= n < 4 ? s->y_dc_scale : s->c_dc_scale;
124 125 126 127 128 129 130
    quant_matrix = s->intra_matrix;
    for(i=1;i<=nCoeffs;i++) {
        int j= s->intra_scantable.permutated[i];
        level = block[j];
        if (level) {
            if (level < 0) {
                level = -level;
131
                level = (int)(level * qscale * quant_matrix[j]) >> 4;
132 133
                level = -level;
            } else {
134
                level = (int)(level * qscale * quant_matrix[j]) >> 4;
135 136 137 138 139 140 141 142 143 144 145 146 147
            }
            block[j] = level;
        }
    }
}

static void dct_unquantize_mpeg2_intra_bitexact(MpegEncContext *s,
                                   int16_t *block, int n, int qscale)
{
    int i, level, nCoeffs;
    const uint16_t *quant_matrix;
    int sum=-1;

148 149
    if (s->q_scale_type) qscale = ff_mpeg2_non_linear_qscale[qscale];
    else                 qscale <<= 1;
150

151 152 153
    if(s->alternate_scan) nCoeffs= 63;
    else nCoeffs= s->block_last_index[n];

154 155
    block[0] *= n < 4 ? s->y_dc_scale : s->c_dc_scale;
    sum += block[0];
156 157 158 159 160 161 162
    quant_matrix = s->intra_matrix;
    for(i=1;i<=nCoeffs;i++) {
        int j= s->intra_scantable.permutated[i];
        level = block[j];
        if (level) {
            if (level < 0) {
                level = -level;
163
                level = (int)(level * qscale * quant_matrix[j]) >> 4;
164 165
                level = -level;
            } else {
166
                level = (int)(level * qscale * quant_matrix[j]) >> 4;
167 168 169 170 171 172 173 174 175 176 177 178 179 180 181
            }
            block[j] = level;
            sum+=level;
        }
    }
    block[63]^=sum&1;
}

static void dct_unquantize_mpeg2_inter_c(MpegEncContext *s,
                                   int16_t *block, int n, int qscale)
{
    int i, level, nCoeffs;
    const uint16_t *quant_matrix;
    int sum=-1;

182 183
    if (s->q_scale_type) qscale = ff_mpeg2_non_linear_qscale[qscale];
    else                 qscale <<= 1;
184

185 186 187 188 189 190 191 192 193 194 195
    if(s->alternate_scan) nCoeffs= 63;
    else nCoeffs= s->block_last_index[n];

    quant_matrix = s->inter_matrix;
    for(i=0; i<=nCoeffs; i++) {
        int j= s->intra_scantable.permutated[i];
        level = block[j];
        if (level) {
            if (level < 0) {
                level = -level;
                level = (((level << 1) + 1) * qscale *
196
                         ((int) (quant_matrix[j]))) >> 5;
197 198 199
                level = -level;
            } else {
                level = (((level << 1) + 1) * qscale *
200
                         ((int) (quant_matrix[j]))) >> 5;
201 202 203 204 205 206 207 208 209 210 211 212 213 214
            }
            block[j] = level;
            sum+=level;
        }
    }
    block[63]^=sum&1;
}

static void dct_unquantize_h263_intra_c(MpegEncContext *s,
                                  int16_t *block, int n, int qscale)
{
    int i, level, qmul, qadd;
    int nCoeffs;

215
    av_assert2(s->block_last_index[n]>=0 || s->h263_aic);
216 217 218 219

    qmul = qscale << 1;

    if (!s->h263_aic) {
220
        block[0] *= n < 4 ? s->y_dc_scale : s->c_dc_scale;
221 222 223 224 225 226 227
        qadd = (qscale - 1) | 1;
    }else{
        qadd = 0;
    }
    if(s->ac_pred)
        nCoeffs=63;
    else
228
        nCoeffs= s->intra_scantable.raster_end[ s->block_last_index[n] ];
229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248

    for(i=1; i<=nCoeffs; i++) {
        level = block[i];
        if (level) {
            if (level < 0) {
                level = level * qmul - qadd;
            } else {
                level = level * qmul + qadd;
            }
            block[i] = level;
        }
    }
}

static void dct_unquantize_h263_inter_c(MpegEncContext *s,
                                  int16_t *block, int n, int qscale)
{
    int i, level, qmul, qadd;
    int nCoeffs;

249
    av_assert2(s->block_last_index[n]>=0);
250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268

    qadd = (qscale - 1) | 1;
    qmul = qscale << 1;

    nCoeffs= s->inter_scantable.raster_end[ s->block_last_index[n] ];

    for(i=0; i<=nCoeffs; i++) {
        level = block[i];
        if (level) {
            if (level < 0) {
                level = level * qmul - qadd;
            } else {
                level = level * qmul + qadd;
            }
            block[i] = level;
        }
    }
}

269

270
static void gray16(uint8_t *dst, const uint8_t *src, ptrdiff_t linesize, int h)
271 272 273 274 275
{
    while(h--)
        memset(dst + h*linesize, 128, 16);
}

276
static void gray8(uint8_t *dst, const uint8_t *src, ptrdiff_t linesize, int h)
277 278 279 280 281
{
    while(h--)
        memset(dst + h*linesize, 128, 8);
}

282
/* init common dct for both encoder and decoder */
283
static av_cold int dct_init(MpegEncContext *s)
Fabrice Bellard's avatar
Fabrice Bellard committed
284
{
285
    ff_blockdsp_init(&s->bdsp, s->avctx);
286
    ff_h264chroma_init(&s->h264chroma, 8); //for lowres
287
    ff_hpeldsp_init(&s->hdsp, s->avctx->flags);
288
    ff_mpegvideodsp_init(&s->mdsp);
289
    ff_videodsp_init(&s->vdsp, s->avctx->bits_per_raw_sample);
290

291 292 293 294 295 296 297 298 299 300 301 302 303
    if (s->avctx->debug & FF_DEBUG_NOMC) {
        int i;
        for (i=0; i<4; i++) {
            s->hdsp.avg_pixels_tab[0][i] = gray16;
            s->hdsp.put_pixels_tab[0][i] = gray16;
            s->hdsp.put_no_rnd_pixels_tab[0][i] = gray16;

            s->hdsp.avg_pixels_tab[1][i] = gray8;
            s->hdsp.put_pixels_tab[1][i] = gray8;
            s->hdsp.put_no_rnd_pixels_tab[1][i] = gray8;
        }
    }

304 305 306 307 308
    s->dct_unquantize_h263_intra = dct_unquantize_h263_intra_c;
    s->dct_unquantize_h263_inter = dct_unquantize_h263_inter_c;
    s->dct_unquantize_mpeg1_intra = dct_unquantize_mpeg1_intra_c;
    s->dct_unquantize_mpeg1_inter = dct_unquantize_mpeg1_inter_c;
    s->dct_unquantize_mpeg2_intra = dct_unquantize_mpeg2_intra_c;
309
    if (s->avctx->flags & AV_CODEC_FLAG_BITEXACT)
310
        s->dct_unquantize_mpeg2_intra = dct_unquantize_mpeg2_intra_bitexact;
311
    s->dct_unquantize_mpeg2_inter = dct_unquantize_mpeg2_inter_c;
312

313
    if (HAVE_INTRINSICS_NEON)
314
        ff_mpv_common_init_neon(s);
315

316
    if (ARCH_ALPHA)
317
        ff_mpv_common_init_axp(s);
318
    if (ARCH_ARM)
319
        ff_mpv_common_init_arm(s);
320
    if (ARCH_PPC)
321
        ff_mpv_common_init_ppc(s);
322
    if (ARCH_X86)
323
        ff_mpv_common_init_x86(s);
324 325
    if (ARCH_MIPS)
        ff_mpv_common_init_mips(s);
326

327 328 329 330 331
    return 0;
}

av_cold void ff_mpv_idct_init(MpegEncContext *s)
{
332 333
    if (s->codec_id == AV_CODEC_ID_MPEG4)
        s->idsp.mpeg4_studio_profile = s->studio_profile;
334 335
    ff_idctdsp_init(&s->idsp, s->avctx);

336
    /* load & permutate scantables
337 338 339
     * note: only wmv uses different ones
     */
    if (s->alternate_scan) {
340 341
        ff_init_scantable(s->idsp.idct_permutation, &s->inter_scantable, ff_alternate_vertical_scan);
        ff_init_scantable(s->idsp.idct_permutation, &s->intra_scantable, ff_alternate_vertical_scan);
342
    } else {
343 344
        ff_init_scantable(s->idsp.idct_permutation, &s->inter_scantable, ff_zigzag_direct);
        ff_init_scantable(s->idsp.idct_permutation, &s->intra_scantable, ff_zigzag_direct);
345
    }
346 347
    ff_init_scantable(s->idsp.idct_permutation, &s->intra_h_scantable, ff_alternate_horizontal_scan);
    ff_init_scantable(s->idsp.idct_permutation, &s->intra_v_scantable, ff_alternate_vertical_scan);
348 349
}

350 351 352 353
static int alloc_picture(MpegEncContext *s, Picture *pic, int shared)
{
    return ff_alloc_picture(s->avctx, pic, &s->me, &s->sc, shared, 0,
                            s->chroma_x_shift, s->chroma_y_shift, s->out_format,
354
                            s->mb_stride, s->mb_width, s->mb_height, s->b8_stride,
355 356 357
                            &s->linesize, &s->uvlinesize);
}

358
static int init_duplicate_context(MpegEncContext *s)
359
{
360 361 362
    int y_size = s->b8_stride * (2 * s->mb_height + 1);
    int c_size = s->mb_stride * (s->mb_height + 1);
    int yc_size = y_size + 2 * c_size;
363 364
    int i;

365 366 367
    if (s->mb_height & 1)
        yc_size += 2*s->b8_stride + 2*s->mb_stride;

368
    s->sc.edge_emu_buffer =
369 370
    s->me.scratchpad   =
    s->me.temp         =
371 372 373
    s->sc.rd_scratchpad   =
    s->sc.b_scratchpad    =
    s->sc.obmc_scratchpad = NULL;
374

375
    if (s->encoding) {
376 377 378 379
        FF_ALLOCZ_OR_GOTO(s->avctx, s->me.map,
                          ME_MAP_SIZE * sizeof(uint32_t), fail)
        FF_ALLOCZ_OR_GOTO(s->avctx, s->me.score_map,
                          ME_MAP_SIZE * sizeof(uint32_t), fail)
380
        if (s->noise_reduction) {
381 382
            FF_ALLOCZ_OR_GOTO(s->avctx, s->dct_error_sum,
                              2 * 64 * sizeof(int), fail)
383
        }
384
    }
Diego Biurrun's avatar
Diego Biurrun committed
385
    FF_ALLOCZ_OR_GOTO(s->avctx, s->blocks, 64 * 12 * 2 * sizeof(int16_t), fail)
386
    s->block = s->blocks[0];
387

388
    for (i = 0; i < 12; i++) {
389
        s->pblocks[i] = &s->block[i];
390
    }
391 392

    FF_ALLOCZ_OR_GOTO(s->avctx, s->block32, sizeof(*s->block32), fail)
393 394
    s->dpcm_direction = 0;
    FF_ALLOCZ_OR_GOTO(s->avctx, s->dpcm_macroblock, sizeof(*s->dpcm_macroblock), fail)
395

396 397
    if (s->avctx->codec_tag == AV_RL32("VCR2")) {
        // exchange uv
398
        FFSWAP(void *, s->pblocks[4], s->pblocks[5]);
399
    }
400

401 402
    if (s->out_format == FMT_H263) {
        /* ac values */
403 404
        FF_ALLOCZ_OR_GOTO(s->avctx, s->ac_val_base,
                          yc_size * sizeof(int16_t) * 16, fail);
405 406 407 408 409
        s->ac_val[0] = s->ac_val_base + s->b8_stride + 1;
        s->ac_val[1] = s->ac_val_base + y_size + s->mb_stride + 1;
        s->ac_val[2] = s->ac_val[1] + c_size;
    }

410 411
    return 0;
fail:
412
    return -1; // free() through ff_mpv_common_end()
413 414
}

415 416
static void free_duplicate_context(MpegEncContext *s)
{
417
    if (!s)
418
        return;
419

420
    av_freep(&s->sc.edge_emu_buffer);
421
    av_freep(&s->me.scratchpad);
422
    s->me.temp =
423 424 425
    s->sc.rd_scratchpad =
    s->sc.b_scratchpad =
    s->sc.obmc_scratchpad = NULL;
426

427 428 429 430
    av_freep(&s->dct_error_sum);
    av_freep(&s->me.map);
    av_freep(&s->me.score_map);
    av_freep(&s->blocks);
431
    av_freep(&s->block32);
432
    av_freep(&s->dpcm_macroblock);
433
    av_freep(&s->ac_val_base);
434
    s->block = NULL;
435 436
}

437 438 439
static void backup_duplicate_context(MpegEncContext *bak, MpegEncContext *src)
{
#define COPY(a) bak->a = src->a
440
    COPY(sc.edge_emu_buffer);
441
    COPY(me.scratchpad);
442
    COPY(me.temp);
443 444 445
    COPY(sc.rd_scratchpad);
    COPY(sc.b_scratchpad);
    COPY(sc.obmc_scratchpad);
446 447 448 449
    COPY(me.map);
    COPY(me.score_map);
    COPY(blocks);
    COPY(block);
450
    COPY(block32);
451 452
    COPY(dpcm_macroblock);
    COPY(dpcm_direction);
453 454 455 456 457
    COPY(start_mb_y);
    COPY(end_mb_y);
    COPY(me.map_generation);
    COPY(pb);
    COPY(dct_error_sum);
458 459
    COPY(dct_count[0]);
    COPY(dct_count[1]);
460 461 462 463
    COPY(ac_val_base);
    COPY(ac_val[0]);
    COPY(ac_val[1]);
    COPY(ac_val[2]);
464 465 466
#undef COPY
}

467
int ff_update_duplicate_context(MpegEncContext *dst, MpegEncContext *src)
468
{
469
    MpegEncContext bak;
470
    int i, ret;
471 472
    // FIXME copy only needed parts
    // START_TIMER
473 474 475
    backup_duplicate_context(&bak, dst);
    memcpy(dst, src, sizeof(MpegEncContext));
    backup_duplicate_context(dst, &bak);
476
    for (i = 0; i < 12; i++) {
477
        dst->pblocks[i] = &dst->block[i];
478
    }
479 480
    if (dst->avctx->codec_tag == AV_RL32("VCR2")) {
        // exchange uv
481
        FFSWAP(void *, dst->pblocks[4], dst->pblocks[5]);
482
    }
483
    if (!dst->sc.edge_emu_buffer &&
484 485
        (ret = ff_mpeg_framesize_alloc(dst->avctx, &dst->me,
                                       &dst->sc, dst->linesize)) < 0) {
486 487 488 489
        av_log(dst->avctx, AV_LOG_ERROR, "failed to allocate context "
               "scratch buffers.\n");
        return ret;
    }
490 491
    // STOP_TIMER("update_duplicate_context")
    // about 10k cycles / 0.01 sec for  1000frames on 1ghz with 2 threads
492
    return 0;
493 494
}

495 496
int ff_mpeg_update_thread_context(AVCodecContext *dst,
                                  const AVCodecContext *src)
497
{
498
    int i, ret;
499 500
    MpegEncContext *s = dst->priv_data, *s1 = src->priv_data;

501
    if (dst == src)
502
        return 0;
503

504 505
    av_assert0(s != s1);

506 507 508
    // FIXME can parameters change on I-frames?
    // in that case dst may need a reinit
    if (!s->context_initialized) {
509
        int err;
510 511 512 513 514 515
        memcpy(s, s1, sizeof(MpegEncContext));

        s->avctx                 = dst;
        s->bitstream_buffer      = NULL;
        s->bitstream_buffer_size = s->allocated_bitstream_buffer_size = 0;

516
        if (s1->context_initialized){
517 518
//             s->picture_range_start  += MAX_PICTURE_COUNT;
//             s->picture_range_end    += MAX_PICTURE_COUNT;
519
            ff_mpv_idct_init(s);
520
            if((err = ff_mpv_common_init(s)) < 0){
521 522
                memset(s, 0, sizeof(MpegEncContext));
                s->avctx = dst;
523
                return err;
524
            }
525
        }
526 527
    }

528 529 530 531
    if (s->height != s1->height || s->width != s1->width || s->context_reinit) {
        s->context_reinit = 0;
        s->height = s1->height;
        s->width  = s1->width;
532
        if ((ret = ff_mpv_common_frame_size_change(s)) < 0)
533
            return ret;
534 535
    }

536 537 538 539 540
    s->avctx->coded_height  = s1->avctx->coded_height;
    s->avctx->coded_width   = s1->avctx->coded_width;
    s->avctx->width         = s1->avctx->width;
    s->avctx->height        = s1->avctx->height;

541 542
    s->quarter_sample       = s1->quarter_sample;

543 544 545
    s->coded_picture_number = s1->coded_picture_number;
    s->picture_number       = s1->picture_number;

546
    av_assert0(!s->picture || s->picture != s1->picture);
547
    if(s->picture)
548
    for (i = 0; i < MAX_PICTURE_COUNT; i++) {
549
        ff_mpeg_unref_picture(s->avctx, &s->picture[i]);
550
        if (s1->picture && s1->picture[i].f->buf[0] &&
551
            (ret = ff_mpeg_ref_picture(s->avctx, &s->picture[i], &s1->picture[i])) < 0)
552
            return ret;
553
    }
554

555 556
#define UPDATE_PICTURE(pic)\
do {\
557
    ff_mpeg_unref_picture(s->avctx, &s->pic);\
558
    if (s1->pic.f && s1->pic.f->buf[0])\
559
        ret = ff_mpeg_ref_picture(s->avctx, &s->pic, &s1->pic);\
560
    else\
561
        ret = ff_update_picture_tables(&s->pic, &s1->pic);\
562 563 564 565 566 567 568
    if (ret < 0)\
        return ret;\
} while (0)

    UPDATE_PICTURE(current_picture);
    UPDATE_PICTURE(last_picture);
    UPDATE_PICTURE(next_picture);
569

570 571 572 573 574
#define REBASE_PICTURE(pic, new_ctx, old_ctx)                                 \
    ((pic && pic >= old_ctx->picture &&                                       \
      pic < old_ctx->picture + MAX_PICTURE_COUNT) ?                           \
        &new_ctx->picture[pic - old_ctx->picture] : NULL)

575 576 577
    s->last_picture_ptr    = REBASE_PICTURE(s1->last_picture_ptr,    s, s1);
    s->current_picture_ptr = REBASE_PICTURE(s1->current_picture_ptr, s, s1);
    s->next_picture_ptr    = REBASE_PICTURE(s1->next_picture_ptr,    s, s1);
578

579
    // Error/bug resilience
580 581
    s->next_p_frame_damaged = s1->next_p_frame_damaged;
    s->workaround_bugs      = s1->workaround_bugs;
582
    s->padding_bug_score    = s1->padding_bug_score;
583

584
    // MPEG-4 timing info
585
    memcpy(&s->last_time_base, &s1->last_time_base,
586
           (char *) &s1->pb_field_time + sizeof(s1->pb_field_time) -
587
           (char *) &s1->last_time_base);
588 589 590 591

    // B-frame info
    s->max_b_frames = s1->max_b_frames;
    s->low_delay    = s1->low_delay;
592
    s->droppable    = s1->droppable;
593 594 595 596 597 598

    // DivX handling (doesn't work)
    s->divx_packed  = s1->divx_packed;

    if (s1->bitstream_buffer) {
        if (s1->bitstream_buffer_size +
599
            AV_INPUT_BUFFER_PADDING_SIZE > s->allocated_bitstream_buffer_size) {
600 601 602
            av_fast_malloc(&s->bitstream_buffer,
                           &s->allocated_bitstream_buffer_size,
                           s1->allocated_bitstream_buffer_size);
603 604
            if (!s->bitstream_buffer) {
                s->bitstream_buffer_size = 0;
605
                return AVERROR(ENOMEM);
606
            }
607 608
        }
        s->bitstream_buffer_size = s1->bitstream_buffer_size;
609 610 611
        memcpy(s->bitstream_buffer, s1->bitstream_buffer,
               s1->bitstream_buffer_size);
        memset(s->bitstream_buffer + s->bitstream_buffer_size, 0,
612
               AV_INPUT_BUFFER_PADDING_SIZE);
613 614
    }

615
    // linesize-dependent scratch buffer allocation
616
    if (!s->sc.edge_emu_buffer)
617
        if (s1->linesize) {
618 619
            if (ff_mpeg_framesize_alloc(s->avctx, &s->me,
                                        &s->sc, s1->linesize) < 0) {
620 621 622 623 624 625 626 627 628
                av_log(s->avctx, AV_LOG_ERROR, "Failed to allocate context "
                       "scratch buffers.\n");
                return AVERROR(ENOMEM);
            }
        } else {
            av_log(s->avctx, AV_LOG_ERROR, "Context scratch buffers could not "
                   "be allocated due to unknown size.\n");
        }

629
    // MPEG-2/interlacing info
630 631
    memcpy(&s->progressive_sequence, &s1->progressive_sequence,
           (char *) &s1->rtp_mode - (char *) &s1->progressive_sequence);
632

633 634 635
    if (!s1->first_field) {
        s->last_pict_type = s1->pict_type;
        if (s1->current_picture_ptr)
636
            s->last_lambda_for[s1->pict_type] = s1->current_picture_ptr->f->quality;
637 638 639 640 641
    }

    return 0;
}

Michael Niedermayer's avatar
Michael Niedermayer committed
642
/**
643
 * Set the given MpegEncContext to common defaults
644
 * (same for encoding and decoding).
645
 * The changed fields will not depend upon the
646
 * prior state of the MpegEncContext.
Michael Niedermayer's avatar
Michael Niedermayer committed
647
 */
648
void ff_mpv_common_defaults(MpegEncContext *s)
649 650 651 652 653 654 655 656 657 658
{
    s->y_dc_scale_table      =
    s->c_dc_scale_table      = ff_mpeg1_dc_scale_table;
    s->chroma_qscale_table   = ff_default_chroma_qscale_table;
    s->progressive_frame     = 1;
    s->progressive_sequence  = 1;
    s->picture_structure     = PICT_FRAME;

    s->coded_picture_number  = 0;
    s->picture_number        = 0;
659

660 661
    s->f_code                = 1;
    s->b_code                = 1;
662

663
    s->slice_context_count   = 1;
Michael Niedermayer's avatar
Michael Niedermayer committed
664 665 666
}

/**
667
 * Set the given MpegEncContext to defaults for decoding.
668 669
 * the changed fields will not depend upon
 * the prior state of the MpegEncContext.
Michael Niedermayer's avatar
Michael Niedermayer committed
670
 */
671
void ff_mpv_decode_defaults(MpegEncContext *s)
672
{
673
    ff_mpv_common_defaults(s);
Michael Niedermayer's avatar
Michael Niedermayer committed
674 675
}

676 677 678 679 680 681 682 683
void ff_mpv_decode_init(MpegEncContext *s, AVCodecContext *avctx)
{
    s->avctx           = avctx;
    s->width           = avctx->coded_width;
    s->height          = avctx->coded_height;
    s->codec_id        = avctx->codec->id;
    s->workaround_bugs = avctx->workaround_bugs;

684 685
    /* convert fourcc to upper case */
    s->codec_tag          = avpriv_toupper4(avctx->codec_tag);
686 687
}

688 689 690 691 692 693 694 695 696 697 698 699 700
/**
 * Initialize and allocates MpegEncContext fields dependent on the resolution.
 */
static int init_context_frame(MpegEncContext *s)
{
    int y_size, c_size, yc_size, i, mb_array_size, mv_table_size, x, y;

    s->mb_width   = (s->width + 15) / 16;
    s->mb_stride  = s->mb_width + 1;
    s->b8_stride  = s->mb_width * 2 + 1;
    mb_array_size = s->mb_height * s->mb_stride;
    mv_table_size = (s->mb_height + 2) * s->mb_stride + 1;

701
    /* set default edge pos, will be overridden
702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718
     * in decode_header if needed */
    s->h_edge_pos = s->mb_width * 16;
    s->v_edge_pos = s->mb_height * 16;

    s->mb_num     = s->mb_width * s->mb_height;

    s->block_wrap[0] =
    s->block_wrap[1] =
    s->block_wrap[2] =
    s->block_wrap[3] = s->b8_stride;
    s->block_wrap[4] =
    s->block_wrap[5] = s->mb_stride;

    y_size  = s->b8_stride * (2 * s->mb_height + 1);
    c_size  = s->mb_stride * (s->mb_height + 1);
    yc_size = y_size + 2   * c_size;

719 720 721
    if (s->mb_height & 1)
        yc_size += 2*s->b8_stride + 2*s->mb_stride;

722
    FF_ALLOCZ_OR_GOTO(s->avctx, s->mb_index2xy, (s->mb_num + 1) * sizeof(int),
723
                      fail); // error resilience code looks cleaner with this
724 725 726 727
    for (y = 0; y < s->mb_height; y++)
        for (x = 0; x < s->mb_width; x++)
            s->mb_index2xy[x + y * s->mb_width] = x + y * s->mb_stride;

728
    s->mb_index2xy[s->mb_height * s->mb_width] = (s->mb_height - 1) * s->mb_stride + s->mb_width; // FIXME really needed?
729 730 731

    if (s->encoding) {
        /* Allocate MV tables */
732 733 734 735 736 737
        FF_ALLOCZ_OR_GOTO(s->avctx, s->p_mv_table_base,                 mv_table_size * 2 * sizeof(int16_t), fail)
        FF_ALLOCZ_OR_GOTO(s->avctx, s->b_forw_mv_table_base,            mv_table_size * 2 * sizeof(int16_t), fail)
        FF_ALLOCZ_OR_GOTO(s->avctx, s->b_back_mv_table_base,            mv_table_size * 2 * sizeof(int16_t), fail)
        FF_ALLOCZ_OR_GOTO(s->avctx, s->b_bidir_forw_mv_table_base,      mv_table_size * 2 * sizeof(int16_t), fail)
        FF_ALLOCZ_OR_GOTO(s->avctx, s->b_bidir_back_mv_table_base,      mv_table_size * 2 * sizeof(int16_t), fail)
        FF_ALLOCZ_OR_GOTO(s->avctx, s->b_direct_mv_table_base,          mv_table_size * 2 * sizeof(int16_t), fail)
738 739 740
        s->p_mv_table            = s->p_mv_table_base + s->mb_stride + 1;
        s->b_forw_mv_table       = s->b_forw_mv_table_base + s->mb_stride + 1;
        s->b_back_mv_table       = s->b_back_mv_table_base + s->mb_stride + 1;
741 742
        s->b_bidir_forw_mv_table = s->b_bidir_forw_mv_table_base + s->mb_stride + 1;
        s->b_bidir_back_mv_table = s->b_bidir_back_mv_table_base + s->mb_stride + 1;
743 744 745
        s->b_direct_mv_table     = s->b_direct_mv_table_base + s->mb_stride + 1;

        /* Allocate MB type table */
746
        FF_ALLOCZ_OR_GOTO(s->avctx, s->mb_type, mb_array_size * sizeof(uint16_t), fail) // needed for encoding
747

748
        FF_ALLOCZ_OR_GOTO(s->avctx, s->lambda_table, mb_array_size * sizeof(int), fail)
749 750 751 752 753 754 755 756 757

        FF_ALLOC_OR_GOTO(s->avctx, s->cplx_tab,
                         mb_array_size * sizeof(float), fail);
        FF_ALLOC_OR_GOTO(s->avctx, s->bits_tab,
                         mb_array_size * sizeof(float), fail);

    }

    if (s->codec_id == AV_CODEC_ID_MPEG4 ||
758
        (s->avctx->flags & AV_CODEC_FLAG_INTERLACED_ME)) {
759 760 761 762 763 764 765 766 767 768 769 770
        /* interlaced direct mode decoding tables */
        for (i = 0; i < 2; i++) {
            int j, k;
            for (j = 0; j < 2; j++) {
                for (k = 0; k < 2; k++) {
                    FF_ALLOCZ_OR_GOTO(s->avctx,
                                      s->b_field_mv_table_base[i][j][k],
                                      mv_table_size * 2 * sizeof(int16_t),
                                      fail);
                    s->b_field_mv_table[i][j][k] = s->b_field_mv_table_base[i][j][k] +
                                                   s->mb_stride + 1;
                }
771 772 773
                FF_ALLOCZ_OR_GOTO(s->avctx, s->b_field_select_table [i][j], mb_array_size * 2 * sizeof(uint8_t), fail)
                FF_ALLOCZ_OR_GOTO(s->avctx, s->p_field_mv_table_base[i][j], mv_table_size * 2 * sizeof(int16_t), fail)
                s->p_field_mv_table[i][j] = s->p_field_mv_table_base[i][j] + s->mb_stride + 1;
774
            }
775
            FF_ALLOCZ_OR_GOTO(s->avctx, s->p_field_select_table[i], mb_array_size * 2 * sizeof(uint8_t), fail)
776 777 778 779
        }
    }
    if (s->out_format == FMT_H263) {
        /* cbp values */
780
        FF_ALLOCZ_OR_GOTO(s->avctx, s->coded_block_base, y_size + (s->mb_height&1)*2*s->b8_stride, fail);
781 782 783
        s->coded_block = s->coded_block_base + s->b8_stride + 1;

        /* cbp, ac_pred, pred_dir */
784 785
        FF_ALLOCZ_OR_GOTO(s->avctx, s->cbp_table     , mb_array_size * sizeof(uint8_t), fail);
        FF_ALLOCZ_OR_GOTO(s->avctx, s->pred_dir_table, mb_array_size * sizeof(uint8_t), fail);
786 787 788 789
    }

    if (s->h263_pred || s->h263_plus || !s->encoding) {
        /* dc values */
Lou Logan's avatar
Lou Logan committed
790
        // MN: we need these for error resilience of intra-frames
791
        FF_ALLOCZ_OR_GOTO(s->avctx, s->dc_val_base, yc_size * sizeof(int16_t), fail);
792 793 794 795 796 797 798
        s->dc_val[0] = s->dc_val_base + s->b8_stride + 1;
        s->dc_val[1] = s->dc_val_base + y_size + s->mb_stride + 1;
        s->dc_val[2] = s->dc_val[1] + c_size;
        for (i = 0; i < yc_size; i++)
            s->dc_val_base[i] = 1024;
    }

Lou Logan's avatar
Lou Logan committed
799
    /* which mb is an intra block */
800 801 802 803 804
    FF_ALLOCZ_OR_GOTO(s->avctx, s->mbintra_table, mb_array_size, fail);
    memset(s->mbintra_table, 1, mb_array_size);

    /* init macroblock skip table */
    FF_ALLOCZ_OR_GOTO(s->avctx, s->mbskip_table, mb_array_size + 2, fail);
805
    // Note the + 1 is for a quicker MPEG-4 slice_end detection
806

807
    return ff_mpeg_er_init(s);
808 809 810 811
fail:
    return AVERROR(ENOMEM);
}

812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827
static void clear_context(MpegEncContext *s)
{
    int i, j, k;

    memset(&s->next_picture, 0, sizeof(s->next_picture));
    memset(&s->last_picture, 0, sizeof(s->last_picture));
    memset(&s->current_picture, 0, sizeof(s->current_picture));
    memset(&s->new_picture, 0, sizeof(s->new_picture));

    memset(s->thread_context, 0, sizeof(s->thread_context));

    s->me.map = NULL;
    s->me.score_map = NULL;
    s->dct_error_sum = NULL;
    s->block = NULL;
    s->blocks = NULL;
828
    s->block32 = NULL;
829
    memset(s->pblocks, 0, sizeof(s->pblocks));
830 831
    s->dpcm_direction = 0;
    s->dpcm_macroblock = NULL;
832 833 834 835 836 837 838 839 840 841 842
    s->ac_val_base = NULL;
    s->ac_val[0] =
    s->ac_val[1] =
    s->ac_val[2] =NULL;
    s->sc.edge_emu_buffer = NULL;
    s->me.scratchpad = NULL;
    s->me.temp =
    s->sc.rd_scratchpad =
    s->sc.b_scratchpad =
    s->sc.obmc_scratchpad = NULL;

843

844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889
    s->bitstream_buffer = NULL;
    s->allocated_bitstream_buffer_size = 0;
    s->picture          = NULL;
    s->mb_type          = NULL;
    s->p_mv_table_base  = NULL;
    s->b_forw_mv_table_base = NULL;
    s->b_back_mv_table_base = NULL;
    s->b_bidir_forw_mv_table_base = NULL;
    s->b_bidir_back_mv_table_base = NULL;
    s->b_direct_mv_table_base = NULL;
    s->p_mv_table            = NULL;
    s->b_forw_mv_table       = NULL;
    s->b_back_mv_table       = NULL;
    s->b_bidir_forw_mv_table = NULL;
    s->b_bidir_back_mv_table = NULL;
    s->b_direct_mv_table     = NULL;
    for (i = 0; i < 2; i++) {
        for (j = 0; j < 2; j++) {
            for (k = 0; k < 2; k++) {
                s->b_field_mv_table_base[i][j][k] = NULL;
                s->b_field_mv_table[i][j][k] = NULL;
            }
            s->b_field_select_table[i][j] = NULL;
            s->p_field_mv_table_base[i][j] = NULL;
            s->p_field_mv_table[i][j] = NULL;
        }
        s->p_field_select_table[i] = NULL;
    }

    s->dc_val_base = NULL;
    s->coded_block_base = NULL;
    s->mbintra_table = NULL;
    s->cbp_table = NULL;
    s->pred_dir_table = NULL;

    s->mbskip_table = NULL;

    s->er.error_status_table = NULL;
    s->er.er_temp_buffer = NULL;
    s->mb_index2xy = NULL;
    s->lambda_table = NULL;

    s->cplx_tab = NULL;
    s->bits_tab = NULL;
}

890
/**
Michael Niedermayer's avatar
Michael Niedermayer committed
891 892 893
 * init common structure for both encoder and decoder.
 * this assumes that some variables like width/height are already set
 */
894
av_cold int ff_mpv_common_init(MpegEncContext *s)
895
{
896
    int i, ret;
897 898 899 900
    int nb_slices = (HAVE_THREADS &&
                     s->avctx->active_thread_type & FF_THREAD_SLICE) ?
                    s->avctx->thread_count : 1;

901 902
    clear_context(s);

903 904
    if (s->encoding && s->avctx->slices)
        nb_slices = s->avctx->slices;
905

906
    if (s->codec_id == AV_CODEC_ID_MPEG2VIDEO && !s->progressive_sequence)
907
        s->mb_height = (s->height + 31) / 32 * 2;
908
    else
Baptiste Coudurier's avatar
Baptiste Coudurier committed
909
        s->mb_height = (s->height + 15) / 16;
910

911
    if (s->avctx->pix_fmt == AV_PIX_FMT_NONE) {
912
        av_log(s->avctx, AV_LOG_ERROR,
913
               "decoding to AV_PIX_FMT_NONE is not supported.\n");
914 915 916
        return -1;
    }

917 918 919 920 921 922 923 924 925
    if (nb_slices > MAX_THREADS || (nb_slices > s->mb_height && s->mb_height)) {
        int max_slices;
        if (s->mb_height)
            max_slices = FFMIN(MAX_THREADS, s->mb_height);
        else
            max_slices = MAX_THREADS;
        av_log(s->avctx, AV_LOG_WARNING, "too many threads/slices (%d),"
               " reducing to %d\n", nb_slices, max_slices);
        nb_slices = max_slices;
Michael Niedermayer's avatar
Michael Niedermayer committed
926 927
    }

928 929
    if ((s->width || s->height) &&
        av_image_check_size(s->width, s->height, 0, s->avctx))
930 931
        return -1;

932
    dct_init(s);
933

934
    /* set chroma shifts */
935 936 937 938 939
    ret = av_pix_fmt_get_chroma_sub_sample(s->avctx->pix_fmt,
                                           &s->chroma_x_shift,
                                           &s->chroma_y_shift);
    if (ret)
        return ret;
940

941
    FF_ALLOCZ_OR_GOTO(s->avctx, s->picture,
942 943
                      MAX_PICTURE_COUNT * sizeof(Picture), fail);
    for (i = 0; i < MAX_PICTURE_COUNT; i++) {
944 945 946
        s->picture[i].f = av_frame_alloc();
        if (!s->picture[i].f)
            goto fail;
947
    }
948 949 950 951 952 953 954 955 956 957 958 959
    s->next_picture.f = av_frame_alloc();
    if (!s->next_picture.f)
        goto fail;
    s->last_picture.f = av_frame_alloc();
    if (!s->last_picture.f)
        goto fail;
    s->current_picture.f = av_frame_alloc();
    if (!s->current_picture.f)
        goto fail;
    s->new_picture.f = av_frame_alloc();
    if (!s->new_picture.f)
        goto fail;
960

961 962
    if (init_context_frame(s))
        goto fail;
963

964
    s->parse_context.state = -1;
965

966 967 968
    s->context_initialized = 1;
    memset(s->thread_context, 0, sizeof(s->thread_context));
    s->thread_context[0]   = s;
969

970
//     if (s->width && s->height) {
971 972 973 974 975
    if (nb_slices > 1) {
        for (i = 0; i < nb_slices; i++) {
            if (i) {
                s->thread_context[i] = av_memdup(s, sizeof(MpegEncContext));
                if (!s->thread_context[i])
976 977
                    goto fail;
            }
978
            if (init_duplicate_context(s->thread_context[i]) < 0)
979
                goto fail;
980 981 982 983
            s->thread_context[i]->start_mb_y =
                (s->mb_height * (i) + nb_slices / 2) / nb_slices;
            s->thread_context[i]->end_mb_y   =
                (s->mb_height * (i + 1) + nb_slices / 2) / nb_slices;
984
        }
985 986 987 988 989 990 991
    } else {
        if (init_duplicate_context(s) < 0)
            goto fail;
        s->start_mb_y = 0;
        s->end_mb_y   = s->mb_height;
    }
    s->slice_context_count = nb_slices;
992
//     }
993

Fabrice Bellard's avatar
Fabrice Bellard committed
994 995
    return 0;
 fail:
996
    ff_mpv_common_end(s);
Fabrice Bellard's avatar
Fabrice Bellard committed
997 998 999
    return -1;
}

1000 1001 1002 1003 1004
/**
 * Frees and resets MpegEncContext fields depending on the resolution.
 * Is used during resolution changes to avoid a full reinitialization of the
 * codec.
 */
1005
static void free_context_frame(MpegEncContext *s)
Fabrice Bellard's avatar
Fabrice Bellard committed
1006
{
1007
    int i, j, k;
Fabrice Bellard's avatar
Fabrice Bellard committed
1008

1009
    av_freep(&s->mb_type);
1010 1011 1012 1013 1014 1015
    av_freep(&s->p_mv_table_base);
    av_freep(&s->b_forw_mv_table_base);
    av_freep(&s->b_back_mv_table_base);
    av_freep(&s->b_bidir_forw_mv_table_base);
    av_freep(&s->b_bidir_back_mv_table_base);
    av_freep(&s->b_direct_mv_table_base);
1016 1017 1018 1019 1020 1021 1022 1023 1024
    s->p_mv_table            = NULL;
    s->b_forw_mv_table       = NULL;
    s->b_back_mv_table       = NULL;
    s->b_bidir_forw_mv_table = NULL;
    s->b_bidir_back_mv_table = NULL;
    s->b_direct_mv_table     = NULL;
    for (i = 0; i < 2; i++) {
        for (j = 0; j < 2; j++) {
            for (k = 0; k < 2; k++) {
1025
                av_freep(&s->b_field_mv_table_base[i][j][k]);
1026
                s->b_field_mv_table[i][j][k] = NULL;
1027 1028 1029
            }
            av_freep(&s->b_field_select_table[i][j]);
            av_freep(&s->p_field_mv_table_base[i][j]);
1030
            s->p_field_mv_table[i][j] = NULL;
1031 1032 1033
        }
        av_freep(&s->p_field_select_table[i]);
    }
1034

1035 1036
    av_freep(&s->dc_val_base);
    av_freep(&s->coded_block_base);
1037
    av_freep(&s->mbintra_table);
1038 1039
    av_freep(&s->cbp_table);
    av_freep(&s->pred_dir_table);
1040

1041
    av_freep(&s->mbskip_table);
1042

1043 1044
    av_freep(&s->er.error_status_table);
    av_freep(&s->er.er_temp_buffer);
1045
    av_freep(&s->mb_index2xy);
1046
    av_freep(&s->lambda_table);
1047

1048 1049 1050 1051 1052 1053
    av_freep(&s->cplx_tab);
    av_freep(&s->bits_tab);

    s->linesize = s->uvlinesize = 0;
}

1054
int ff_mpv_common_frame_size_change(MpegEncContext *s)
1055 1056 1057
{
    int i, err = 0;

1058 1059 1060
    if (!s->context_initialized)
        return AVERROR(EINVAL);

1061 1062 1063 1064 1065 1066 1067 1068 1069 1070
    if (s->slice_context_count > 1) {
        for (i = 0; i < s->slice_context_count; i++) {
            free_duplicate_context(s->thread_context[i]);
        }
        for (i = 1; i < s->slice_context_count; i++) {
            av_freep(&s->thread_context[i]);
        }
    } else
        free_duplicate_context(s);

1071
    free_context_frame(s);
1072 1073

    if (s->picture)
1074
        for (i = 0; i < MAX_PICTURE_COUNT; i++) {
1075 1076 1077 1078 1079 1080 1081 1082 1083 1084
                s->picture[i].needs_realloc = 1;
        }

    s->last_picture_ptr         =
    s->next_picture_ptr         =
    s->current_picture_ptr      = NULL;

    // init
    if (s->codec_id == AV_CODEC_ID_MPEG2VIDEO && !s->progressive_sequence)
        s->mb_height = (s->height + 31) / 32 * 2;
1085
    else
1086 1087 1088
        s->mb_height = (s->height + 15) / 16;

    if ((s->width || s->height) &&
1089
        (err = av_image_check_size(s->width, s->height, 0, s->avctx)) < 0)
1090
        goto fail;
1091 1092 1093 1094

    if ((err = init_context_frame(s)))
        goto fail;

1095
    memset(s->thread_context, 0, sizeof(s->thread_context));
1096 1097 1098 1099 1100 1101
    s->thread_context[0]   = s;

    if (s->width && s->height) {
        int nb_slices = s->slice_context_count;
        if (nb_slices > 1) {
            for (i = 0; i < nb_slices; i++) {
1102
                if (i) {
1103 1104 1105 1106 1107
                    s->thread_context[i] = av_memdup(s, sizeof(MpegEncContext));
                    if (!s->thread_context[i]) {
                        err = AVERROR(ENOMEM);
                        goto fail;
                    }
1108
                }
1109
                if ((err = init_duplicate_context(s->thread_context[i])) < 0)
1110
                    goto fail;
1111 1112 1113 1114
                s->thread_context[i]->start_mb_y =
                    (s->mb_height * (i) + nb_slices / 2) / nb_slices;
                s->thread_context[i]->end_mb_y   =
                    (s->mb_height * (i + 1) + nb_slices / 2) / nb_slices;
1115 1116
            }
        } else {
1117 1118
            err = init_duplicate_context(s);
            if (err < 0)
1119 1120 1121 1122 1123 1124 1125 1126 1127
                goto fail;
            s->start_mb_y = 0;
            s->end_mb_y   = s->mb_height;
        }
        s->slice_context_count = nb_slices;
    }

    return 0;
 fail:
1128
    ff_mpv_common_end(s);
1129 1130 1131
    return err;
}

1132
/* init common structure for both encoder and decoder */
1133
void ff_mpv_common_end(MpegEncContext *s)
1134 1135 1136
{
    int i;

1137 1138 1139
    if (!s)
        return ;

1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155
    if (s->slice_context_count > 1) {
        for (i = 0; i < s->slice_context_count; i++) {
            free_duplicate_context(s->thread_context[i]);
        }
        for (i = 1; i < s->slice_context_count; i++) {
            av_freep(&s->thread_context[i]);
        }
        s->slice_context_count = 1;
    } else free_duplicate_context(s);

    av_freep(&s->parse_context.buffer);
    s->parse_context.buffer_size = 0;

    av_freep(&s->bitstream_buffer);
    s->allocated_bitstream_buffer_size = 0;

1156 1157
    if (s->picture) {
        for (i = 0; i < MAX_PICTURE_COUNT; i++) {
1158
            ff_free_picture_tables(&s->picture[i]);
1159
            ff_mpeg_unref_picture(s->avctx, &s->picture[i]);
1160
            av_frame_free(&s->picture[i].f);
Michael Niedermayer's avatar
Michael Niedermayer committed
1161
        }
Fabrice Bellard's avatar
Fabrice Bellard committed
1162
    }
1163
    av_freep(&s->picture);
1164
    ff_free_picture_tables(&s->last_picture);
1165
    ff_mpeg_unref_picture(s->avctx, &s->last_picture);
1166
    av_frame_free(&s->last_picture.f);
1167
    ff_free_picture_tables(&s->current_picture);
1168
    ff_mpeg_unref_picture(s->avctx, &s->current_picture);
1169
    av_frame_free(&s->current_picture.f);
1170
    ff_free_picture_tables(&s->next_picture);
1171
    ff_mpeg_unref_picture(s->avctx, &s->next_picture);
1172 1173
    av_frame_free(&s->next_picture.f);
    ff_free_picture_tables(&s->new_picture);
1174
    ff_mpeg_unref_picture(s->avctx, &s->new_picture);
1175
    av_frame_free(&s->new_picture.f);
1176 1177 1178

    free_context_frame(s);

1179 1180 1181 1182 1183
    s->context_initialized      = 0;
    s->last_picture_ptr         =
    s->next_picture_ptr         =
    s->current_picture_ptr      = NULL;
    s->linesize = s->uvlinesize = 0;
Fabrice Bellard's avatar
Fabrice Bellard committed
1184 1185
}

1186

1187 1188 1189 1190 1191 1192 1193 1194
static void gray_frame(AVFrame *frame)
{
    int i, h_chroma_shift, v_chroma_shift;

    av_pix_fmt_get_chroma_sub_sample(frame->format, &h_chroma_shift, &v_chroma_shift);

    for(i=0; i<frame->height; i++)
        memset(frame->data[0] + frame->linesize[0]*i, 0x80, frame->width);
1195
    for(i=0; i<AV_CEIL_RSHIFT(frame->height, v_chroma_shift); i++) {
1196
        memset(frame->data[1] + frame->linesize[1]*i,
1197
               0x80, AV_CEIL_RSHIFT(frame->width, h_chroma_shift));
1198
        memset(frame->data[2] + frame->linesize[2]*i,
1199
               0x80, AV_CEIL_RSHIFT(frame->width, h_chroma_shift));
1200 1201 1202
    }
}

1203
/**
1204 1205
 * generic function called after decoding
 * the header and before a frame is decoded.
1206
 */
1207
int ff_mpv_frame_start(MpegEncContext *s, AVCodecContext *avctx)
Fabrice Bellard's avatar
Fabrice Bellard committed
1208
{
1209
    int i, ret;
1210
    Picture *pic;
1211
    s->mb_skipped = 0;
1212

1213 1214 1215 1216 1217
    if (!ff_thread_can_start_frame(avctx)) {
        av_log(avctx, AV_LOG_ERROR, "Attempt to start a frame outside SETUP state\n");
        return -1;
    }

1218
    /* mark & release old frames */
Anton Khirnov's avatar
Anton Khirnov committed
1219 1220
    if (s->pict_type != AV_PICTURE_TYPE_B && s->last_picture_ptr &&
        s->last_picture_ptr != s->next_picture_ptr &&
1221
        s->last_picture_ptr->f->buf[0]) {
1222
        ff_mpeg_unref_picture(s->avctx, s->last_picture_ptr);
Anton Khirnov's avatar
Anton Khirnov committed
1223
    }
1224

Anton Khirnov's avatar
Anton Khirnov committed
1225
    /* release forgotten pictures */
1226
    /* if (MPEG-124 / H.263) */
1227 1228 1229 1230
    for (i = 0; i < MAX_PICTURE_COUNT; i++) {
        if (&s->picture[i] != s->last_picture_ptr &&
            &s->picture[i] != s->next_picture_ptr &&
            s->picture[i].reference && !s->picture[i].needs_realloc) {
1231
            ff_mpeg_unref_picture(s->avctx, &s->picture[i]);
1232
        }
Anton Khirnov's avatar
Anton Khirnov committed
1233
    }
1234

1235
    ff_mpeg_unref_picture(s->avctx, &s->current_picture);
1236 1237
    ff_mpeg_unref_picture(s->avctx, &s->last_picture);
    ff_mpeg_unref_picture(s->avctx, &s->next_picture);
1238

1239 1240 1241 1242 1243
    /* release non reference frames */
    for (i = 0; i < MAX_PICTURE_COUNT; i++) {
        if (!s->picture[i].reference)
            ff_mpeg_unref_picture(s->avctx, &s->picture[i]);
    }
1244

1245
    if (s->current_picture_ptr && !s->current_picture_ptr->f->buf[0]) {
Lou Logan's avatar
Lou Logan committed
1246
        // we already have an unused image
1247 1248 1249
        // (maybe it was set before reading the header)
        pic = s->current_picture_ptr;
    } else {
1250
        i   = ff_find_unused_picture(s->avctx, s->picture, 0);
1251 1252 1253
        if (i < 0) {
            av_log(s->avctx, AV_LOG_ERROR, "no frame buffer available\n");
            return i;
1254
        }
1255 1256
        pic = &s->picture[i];
    }
1257

1258 1259 1260 1261 1262
    pic->reference = 0;
    if (!s->droppable) {
        if (s->pict_type != AV_PICTURE_TYPE_B)
            pic->reference = 3;
    }
1263

1264
    pic->f->coded_picture_number = s->coded_picture_number++;
1265

1266
    if (alloc_picture(s, pic, 0) < 0)
1267
        return -1;
1268

1269 1270
    s->current_picture_ptr = pic;
    // FIXME use only the vars from current_pic
1271
    s->current_picture_ptr->f->top_field_first = s->top_field_first;
1272 1273 1274
    if (s->codec_id == AV_CODEC_ID_MPEG1VIDEO ||
        s->codec_id == AV_CODEC_ID_MPEG2VIDEO) {
        if (s->picture_structure != PICT_FRAME)
1275
            s->current_picture_ptr->f->top_field_first =
1276
                (s->picture_structure == PICT_TOP_FIELD) == s->first_field;
Michael Niedermayer's avatar
Michael Niedermayer committed
1277
    }
1278
    s->current_picture_ptr->f->interlaced_frame = !s->progressive_frame &&
1279 1280
                                                 !s->progressive_sequence;
    s->current_picture_ptr->field_picture      =  s->picture_structure != PICT_FRAME;
1281

1282
    s->current_picture_ptr->f->pict_type = s->pict_type;
1283
    // if (s->avctx->flags && AV_CODEC_FLAG_QSCALE)
1284
    //     s->current_picture_ptr->quality = s->new_picture_ptr->quality;
1285
    s->current_picture_ptr->f->key_frame = s->pict_type == AV_PICTURE_TYPE_I;
1286

1287
    if ((ret = ff_mpeg_ref_picture(s->avctx, &s->current_picture,
1288 1289
                                   s->current_picture_ptr)) < 0)
        return ret;
1290

1291
    if (s->pict_type != AV_PICTURE_TYPE_B) {
1292
        s->last_picture_ptr = s->next_picture_ptr;
1293
        if (!s->droppable)
1294
            s->next_picture_ptr = s->current_picture_ptr;
Fabrice Bellard's avatar
Fabrice Bellard committed
1295
    }
1296
    ff_dlog(s->avctx, "L%p N%p C%p L%p N%p C%p type:%d drop:%d\n",
1297
            s->last_picture_ptr, s->next_picture_ptr,s->current_picture_ptr,
1298 1299 1300
            s->last_picture_ptr    ? s->last_picture_ptr->f->data[0]    : NULL,
            s->next_picture_ptr    ? s->next_picture_ptr->f->data[0]    : NULL,
            s->current_picture_ptr ? s->current_picture_ptr->f->data[0] : NULL,
1301
            s->pict_type, s->droppable);
1302

1303
    if ((!s->last_picture_ptr || !s->last_picture_ptr->f->buf[0]) &&
1304
        (s->pict_type != AV_PICTURE_TYPE_I)) {
Anton Khirnov's avatar
Anton Khirnov committed
1305 1306 1307
        int h_chroma_shift, v_chroma_shift;
        av_pix_fmt_get_chroma_sub_sample(s->avctx->pix_fmt,
                                         &h_chroma_shift, &v_chroma_shift);
1308
        if (s->pict_type == AV_PICTURE_TYPE_B && s->next_picture_ptr && s->next_picture_ptr->f->buf[0])
1309
            av_log(avctx, AV_LOG_DEBUG,
1310 1311
                   "allocating dummy last picture for B frame\n");
        else if (s->pict_type != AV_PICTURE_TYPE_I)
Anton Khirnov's avatar
Anton Khirnov committed
1312 1313 1314 1315
            av_log(avctx, AV_LOG_ERROR,
                   "warning: first frame is no keyframe\n");

        /* Allocate a dummy frame */
1316
        i = ff_find_unused_picture(s->avctx, s->picture, 0);
Anton Khirnov's avatar
Anton Khirnov committed
1317 1318 1319
        if (i < 0) {
            av_log(s->avctx, AV_LOG_ERROR, "no frame buffer available\n");
            return i;
1320
        }
Anton Khirnov's avatar
Anton Khirnov committed
1321
        s->last_picture_ptr = &s->picture[i];
1322 1323

        s->last_picture_ptr->reference   = 3;
1324 1325
        s->last_picture_ptr->f->key_frame = 0;
        s->last_picture_ptr->f->pict_type = AV_PICTURE_TYPE_P;
1326

1327
        if (alloc_picture(s, s->last_picture_ptr, 0) < 0) {
Anton Khirnov's avatar
Anton Khirnov committed
1328 1329
            s->last_picture_ptr = NULL;
            return -1;
1330
        }
1331

James Almer's avatar
James Almer committed
1332
        if (!avctx->hwaccel) {
1333
            for(i=0; i<avctx->height; i++)
1334
                memset(s->last_picture_ptr->f->data[0] + s->last_picture_ptr->f->linesize[0]*i,
1335
                       0x80, avctx->width);
1336
            if (s->last_picture_ptr->f->data[2]) {
1337
                for(i=0; i<AV_CEIL_RSHIFT(avctx->height, v_chroma_shift); i++) {
1338
                    memset(s->last_picture_ptr->f->data[1] + s->last_picture_ptr->f->linesize[1]*i,
1339
                        0x80, AV_CEIL_RSHIFT(avctx->width, h_chroma_shift));
1340
                    memset(s->last_picture_ptr->f->data[2] + s->last_picture_ptr->f->linesize[2]*i,
1341
                        0x80, AV_CEIL_RSHIFT(avctx->width, h_chroma_shift));
1342
                }
1343 1344 1345 1346
            }

            if(s->codec_id == AV_CODEC_ID_FLV1 || s->codec_id == AV_CODEC_ID_H263){
                for(i=0; i<avctx->height; i++)
1347
                memset(s->last_picture_ptr->f->data[0] + s->last_picture_ptr->f->linesize[0]*i, 16, avctx->width);
1348
            }
1349
        }
1350

Anton Khirnov's avatar
Anton Khirnov committed
1351 1352 1353
        ff_thread_report_progress(&s->last_picture_ptr->tf, INT_MAX, 0);
        ff_thread_report_progress(&s->last_picture_ptr->tf, INT_MAX, 1);
    }
1354
    if ((!s->next_picture_ptr || !s->next_picture_ptr->f->buf[0]) &&
Anton Khirnov's avatar
Anton Khirnov committed
1355 1356
        s->pict_type == AV_PICTURE_TYPE_B) {
        /* Allocate a dummy frame */
1357
        i = ff_find_unused_picture(s->avctx, s->picture, 0);
Anton Khirnov's avatar
Anton Khirnov committed
1358 1359 1360
        if (i < 0) {
            av_log(s->avctx, AV_LOG_ERROR, "no frame buffer available\n");
            return i;
1361
        }
Anton Khirnov's avatar
Anton Khirnov committed
1362
        s->next_picture_ptr = &s->picture[i];
1363 1364

        s->next_picture_ptr->reference   = 3;
1365 1366
        s->next_picture_ptr->f->key_frame = 0;
        s->next_picture_ptr->f->pict_type = AV_PICTURE_TYPE_P;
1367

1368
        if (alloc_picture(s, s->next_picture_ptr, 0) < 0) {
Anton Khirnov's avatar
Anton Khirnov committed
1369 1370
            s->next_picture_ptr = NULL;
            return -1;
1371
        }
Anton Khirnov's avatar
Anton Khirnov committed
1372 1373 1374
        ff_thread_report_progress(&s->next_picture_ptr->tf, INT_MAX, 0);
        ff_thread_report_progress(&s->next_picture_ptr->tf, INT_MAX, 1);
    }
1375

1376
#if 0 // BUFREF-FIXME
1377 1378
    memset(s->last_picture.f->data, 0, sizeof(s->last_picture.f->data));
    memset(s->next_picture.f->data, 0, sizeof(s->next_picture.f->data));
1379
#endif
Anton Khirnov's avatar
Anton Khirnov committed
1380
    if (s->last_picture_ptr) {
1381
        if (s->last_picture_ptr->f->buf[0] &&
1382
            (ret = ff_mpeg_ref_picture(s->avctx, &s->last_picture,
Anton Khirnov's avatar
Anton Khirnov committed
1383 1384 1385 1386
                                       s->last_picture_ptr)) < 0)
            return ret;
    }
    if (s->next_picture_ptr) {
1387
        if (s->next_picture_ptr->f->buf[0] &&
1388
            (ret = ff_mpeg_ref_picture(s->avctx, &s->next_picture,
Anton Khirnov's avatar
Anton Khirnov committed
1389 1390 1391
                                       s->next_picture_ptr)) < 0)
            return ret;
    }
1392

1393
    av_assert0(s->pict_type == AV_PICTURE_TYPE_I || (s->last_picture_ptr &&
1394
                                                 s->last_picture_ptr->f->buf[0]));
1395

1396
    if (s->picture_structure!= PICT_FRAME) {
1397
        int i;
1398 1399
        for (i = 0; i < 4; i++) {
            if (s->picture_structure == PICT_BOTTOM_FIELD) {
1400 1401
                s->current_picture.f->data[i] +=
                    s->current_picture.f->linesize[i];
1402
            }
1403 1404 1405
            s->current_picture.f->linesize[i] *= 2;
            s->last_picture.f->linesize[i]    *= 2;
            s->next_picture.f->linesize[i]    *= 2;
1406 1407
        }
    }
1408

1409
    /* set dequantizer, we can't do it during init as
1410 1411
     * it might change for MPEG-4 and we can't do it in the header
     * decode as init is not called for MPEG-4 there yet */
1412
    if (s->mpeg_quant || s->codec_id == AV_CODEC_ID_MPEG2VIDEO) {
1413 1414
        s->dct_unquantize_intra = s->dct_unquantize_mpeg2_intra;
        s->dct_unquantize_inter = s->dct_unquantize_mpeg2_inter;
1415
    } else if (s->out_format == FMT_H263 || s->out_format == FMT_H261) {
1416 1417
        s->dct_unquantize_intra = s->dct_unquantize_h263_intra;
        s->dct_unquantize_inter = s->dct_unquantize_h263_inter;
1418
    } else {
1419 1420 1421
        s->dct_unquantize_intra = s->dct_unquantize_mpeg1_intra;
        s->dct_unquantize_inter = s->dct_unquantize_mpeg1_inter;
    }
1422

1423 1424 1425 1426
    if (s->avctx->debug & FF_DEBUG_NOMC) {
        gray_frame(s->current_picture_ptr->f);
    }

1427
    return 0;
Fabrice Bellard's avatar
Fabrice Bellard committed
1428
}
1429

1430
/* called after a frame has been decoded. */
1431
void ff_mpv_frame_end(MpegEncContext *s)
Fabrice Bellard's avatar
Fabrice Bellard committed
1432
{
1433
    emms_c();
1434

1435
    if (s->current_picture.reference)
1436
        ff_thread_report_progress(&s->current_picture_ptr->tf, INT_MAX, 0);
Fabrice Bellard's avatar
Fabrice Bellard committed
1437 1438
}

1439
void ff_print_debug_info(MpegEncContext *s, Picture *p, AVFrame *pict)
1440
{
1441 1442
    ff_print_debug_info2(s->avctx, pict, s->mbskip_table, p->mb_type,
                         p->qscale_table, p->motion_val, &s->low_delay,
1443
                         s->mb_width, s->mb_height, s->mb_stride, s->quarter_sample);
1444 1445
}

1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457
int ff_mpv_export_qp_table(MpegEncContext *s, AVFrame *f, Picture *p, int qp_type)
{
    AVBufferRef *ref = av_buffer_ref(p->qscale_table_buf);
    int offset = 2*s->mb_stride + 1;
    if(!ref)
        return AVERROR(ENOMEM);
    av_assert0(ref->size >= offset + s->mb_stride * ((f->height+15)/16));
    ref->size -= offset;
    ref->data += offset;
    return av_frame_set_qp_table(f, ref, s->mb_stride, qp_type);
}

1458 1459 1460 1461
static inline int hpel_motion_lowres(MpegEncContext *s,
                                     uint8_t *dest, uint8_t *src,
                                     int field_based, int field_select,
                                     int src_x, int src_y,
1462
                                     int width, int height, ptrdiff_t stride,
1463 1464 1465 1466 1467
                                     int h_edge_pos, int v_edge_pos,
                                     int w, int h, h264_chroma_mc_func *pix_op,
                                     int motion_x, int motion_y)
{
    const int lowres   = s->avctx->lowres;
1468
    const int op_index = FFMIN(lowres, 3);
1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486
    const int s_mask   = (2 << lowres) - 1;
    int emu = 0;
    int sx, sy;

    if (s->quarter_sample) {
        motion_x /= 2;
        motion_y /= 2;
    }

    sx = motion_x & s_mask;
    sy = motion_y & s_mask;
    src_x += motion_x >> lowres + 1;
    src_y += motion_y >> lowres + 1;

    src   += src_y * stride + src_x;

    if ((unsigned)src_x > FFMAX( h_edge_pos - (!!sx) - w,                 0) ||
        (unsigned)src_y > FFMAX((v_edge_pos >> field_based) - (!!sy) - h, 0)) {
1487
        s->vdsp.emulated_edge_mc(s->sc.edge_emu_buffer, src,
1488 1489 1490 1491
                                 s->linesize, s->linesize,
                                 w + 1, (h + 1) << field_based,
                                 src_x, src_y   << field_based,
                                 h_edge_pos, v_edge_pos);
1492
        src = s->sc.edge_emu_buffer;
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
        emu = 1;
    }

    sx = (sx << 2) >> lowres;
    sy = (sy << 2) >> lowres;
    if (field_select)
        src += s->linesize;
    pix_op[op_index](dest, src, stride, h, sx, sy);
    return emu;
}

/* apply one mpeg motion vector to the three components */
static av_always_inline void mpeg_motion_lowres(MpegEncContext *s,
                                                uint8_t *dest_y,
                                                uint8_t *dest_cb,
                                                uint8_t *dest_cr,
                                                int field_based,
                                                int bottom_field,
                                                int field_select,
                                                uint8_t **ref_picture,
                                                h264_chroma_mc_func *pix_op,
                                                int motion_x, int motion_y,
                                                int h, int mb_y)
{
    uint8_t *ptr_y, *ptr_cb, *ptr_cr;
1518 1519
    int mx, my, src_x, src_y, uvsrc_x, uvsrc_y, sx, sy, uvsx, uvsy;
    ptrdiff_t uvlinesize, linesize;
1520
    const int lowres     = s->avctx->lowres;
1521
    const int op_index   = FFMIN(lowres-1+s->chroma_x_shift, 3);
1522 1523 1524 1525
    const int block_s    = 8>>lowres;
    const int s_mask     = (2 << lowres) - 1;
    const int h_edge_pos = s->h_edge_pos >> lowres;
    const int v_edge_pos = s->v_edge_pos >> lowres;
1526 1527
    linesize   = s->current_picture.f->linesize[0] << field_based;
    uvlinesize = s->current_picture.f->linesize[1] << field_based;
1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586

    // FIXME obviously not perfect but qpel will not work in lowres anyway
    if (s->quarter_sample) {
        motion_x /= 2;
        motion_y /= 2;
    }

    if(field_based){
        motion_y += (bottom_field - field_select)*((1 << lowres)-1);
    }

    sx = motion_x & s_mask;
    sy = motion_y & s_mask;
    src_x = s->mb_x * 2 * block_s + (motion_x >> lowres + 1);
    src_y = (mb_y * 2 * block_s >> field_based) + (motion_y >> lowres + 1);

    if (s->out_format == FMT_H263) {
        uvsx    = ((motion_x >> 1) & s_mask) | (sx & 1);
        uvsy    = ((motion_y >> 1) & s_mask) | (sy & 1);
        uvsrc_x = src_x >> 1;
        uvsrc_y = src_y >> 1;
    } else if (s->out_format == FMT_H261) {
        // even chroma mv's are full pel in H261
        mx      = motion_x / 4;
        my      = motion_y / 4;
        uvsx    = (2 * mx) & s_mask;
        uvsy    = (2 * my) & s_mask;
        uvsrc_x = s->mb_x * block_s + (mx >> lowres);
        uvsrc_y =    mb_y * block_s + (my >> lowres);
    } else {
        if(s->chroma_y_shift){
            mx      = motion_x / 2;
            my      = motion_y / 2;
            uvsx    = mx & s_mask;
            uvsy    = my & s_mask;
            uvsrc_x = s->mb_x * block_s                 + (mx >> lowres + 1);
            uvsrc_y =   (mb_y * block_s >> field_based) + (my >> lowres + 1);
        } else {
            if(s->chroma_x_shift){
            //Chroma422
                mx = motion_x / 2;
                uvsx = mx & s_mask;
                uvsy = motion_y & s_mask;
                uvsrc_y = src_y;
                uvsrc_x = s->mb_x*block_s               + (mx >> (lowres+1));
            } else {
            //Chroma444
                uvsx = motion_x & s_mask;
                uvsy = motion_y & s_mask;
                uvsrc_x = src_x;
                uvsrc_y = src_y;
            }
        }
    }

    ptr_y  = ref_picture[0] + src_y   * linesize   + src_x;
    ptr_cb = ref_picture[1] + uvsrc_y * uvlinesize + uvsrc_x;
    ptr_cr = ref_picture[2] + uvsrc_y * uvlinesize + uvsrc_x;

1587
    if ((unsigned) src_x > FFMAX( h_edge_pos - (!!sx) - 2 * block_s,       0) || uvsrc_y<0 ||
1588
        (unsigned) src_y > FFMAX((v_edge_pos >> field_based) - (!!sy) - h, 0)) {
1589
        s->vdsp.emulated_edge_mc(s->sc.edge_emu_buffer, ptr_y,
1590 1591
                                 linesize >> field_based, linesize >> field_based,
                                 17, 17 + field_based,
1592 1593
                                src_x, src_y << field_based, h_edge_pos,
                                v_edge_pos);
1594
        ptr_y = s->sc.edge_emu_buffer;
1595
        if (!CONFIG_GRAY || !(s->avctx->flags & AV_CODEC_FLAG_GRAY)) {
1596
            uint8_t *ubuf = s->sc.edge_emu_buffer + 18 * s->linesize;
1597
            uint8_t *vbuf =ubuf + 10 * s->uvlinesize;
1598 1599
            if (s->workaround_bugs & FF_BUG_IEDGE)
                vbuf -= s->uvlinesize;
1600
            s->vdsp.emulated_edge_mc(ubuf,  ptr_cb,
1601 1602
                                     uvlinesize >> field_based, uvlinesize >> field_based,
                                     9, 9 + field_based,
1603 1604
                                    uvsrc_x, uvsrc_y << field_based,
                                    h_edge_pos >> 1, v_edge_pos >> 1);
1605
            s->vdsp.emulated_edge_mc(vbuf,  ptr_cr,
1606 1607
                                     uvlinesize >> field_based,uvlinesize >> field_based,
                                     9, 9 + field_based,
1608 1609
                                    uvsrc_x, uvsrc_y << field_based,
                                    h_edge_pos >> 1, v_edge_pos >> 1);
1610 1611
            ptr_cb = ubuf;
            ptr_cr = vbuf;
1612 1613 1614
        }
    }

1615
    // FIXME use this for field pix too instead of the obnoxious hack which changes picture.f->data
1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631
    if (bottom_field) {
        dest_y  += s->linesize;
        dest_cb += s->uvlinesize;
        dest_cr += s->uvlinesize;
    }

    if (field_select) {
        ptr_y   += s->linesize;
        ptr_cb  += s->uvlinesize;
        ptr_cr  += s->uvlinesize;
    }

    sx = (sx << 2) >> lowres;
    sy = (sy << 2) >> lowres;
    pix_op[lowres - 1](dest_y, ptr_y, linesize, h, sx, sy);

1632
    if (!CONFIG_GRAY || !(s->avctx->flags & AV_CODEC_FLAG_GRAY)) {
1633
        int hc = s->chroma_y_shift ? (h+1-bottom_field)>>1 : h;
1634 1635
        uvsx = (uvsx << 2) >> lowres;
        uvsy = (uvsy << 2) >> lowres;
1636 1637 1638
        if (hc) {
            pix_op[op_index](dest_cb, ptr_cb, uvlinesize, hc, uvsx, uvsy);
            pix_op[op_index](dest_cr, ptr_cr, uvlinesize, hc, uvsx, uvsy);
1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650
        }
    }
    // FIXME h261 lowres loop filter
}

static inline void chroma_4mv_motion_lowres(MpegEncContext *s,
                                            uint8_t *dest_cb, uint8_t *dest_cr,
                                            uint8_t **ref_picture,
                                            h264_chroma_mc_func * pix_op,
                                            int mx, int my)
{
    const int lowres     = s->avctx->lowres;
1651
    const int op_index   = FFMIN(lowres, 3);
1652 1653 1654 1655
    const int block_s    = 8 >> lowres;
    const int s_mask     = (2 << lowres) - 1;
    const int h_edge_pos = s->h_edge_pos >> lowres + 1;
    const int v_edge_pos = s->v_edge_pos >> lowres + 1;
1656 1657
    int emu = 0, src_x, src_y, sx, sy;
    ptrdiff_t offset;
1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676
    uint8_t *ptr;

    if (s->quarter_sample) {
        mx /= 2;
        my /= 2;
    }

    /* In case of 8X8, we construct a single chroma motion vector
       with a special rounding */
    mx = ff_h263_round_chroma(mx);
    my = ff_h263_round_chroma(my);

    sx = mx & s_mask;
    sy = my & s_mask;
    src_x = s->mb_x * block_s + (mx >> lowres + 1);
    src_y = s->mb_y * block_s + (my >> lowres + 1);

    offset = src_y * s->uvlinesize + src_x;
    ptr = ref_picture[1] + offset;
1677 1678
    if ((unsigned) src_x > FFMAX(h_edge_pos - (!!sx) - block_s, 0) ||
        (unsigned) src_y > FFMAX(v_edge_pos - (!!sy) - block_s, 0)) {
1679
        s->vdsp.emulated_edge_mc(s->sc.edge_emu_buffer, ptr,
1680 1681 1682
                                 s->uvlinesize, s->uvlinesize,
                                 9, 9,
                                 src_x, src_y, h_edge_pos, v_edge_pos);
1683
        ptr = s->sc.edge_emu_buffer;
1684
        emu = 1;
1685 1686 1687 1688 1689 1690 1691
    }
    sx = (sx << 2) >> lowres;
    sy = (sy << 2) >> lowres;
    pix_op[op_index](dest_cb, ptr, s->uvlinesize, block_s, sx, sy);

    ptr = ref_picture[2] + offset;
    if (emu) {
1692
        s->vdsp.emulated_edge_mc(s->sc.edge_emu_buffer, ptr,
1693 1694 1695
                                 s->uvlinesize, s->uvlinesize,
                                 9, 9,
                                 src_x, src_y, h_edge_pos, v_edge_pos);
1696
        ptr = s->sc.edge_emu_buffer;
1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750 1751
    }
    pix_op[op_index](dest_cr, ptr, s->uvlinesize, block_s, sx, sy);
}

/**
 * motion compensation of a single macroblock
 * @param s context
 * @param dest_y luma destination pointer
 * @param dest_cb chroma cb/u destination pointer
 * @param dest_cr chroma cr/v destination pointer
 * @param dir direction (0->forward, 1->backward)
 * @param ref_picture array[3] of pointers to the 3 planes of the reference picture
 * @param pix_op halfpel motion compensation function (average or put normally)
 * the motion vectors are taken from s->mv and the MV type from s->mv_type
 */
static inline void MPV_motion_lowres(MpegEncContext *s,
                                     uint8_t *dest_y, uint8_t *dest_cb,
                                     uint8_t *dest_cr,
                                     int dir, uint8_t **ref_picture,
                                     h264_chroma_mc_func *pix_op)
{
    int mx, my;
    int mb_x, mb_y, i;
    const int lowres  = s->avctx->lowres;
    const int block_s = 8 >>lowres;

    mb_x = s->mb_x;
    mb_y = s->mb_y;

    switch (s->mv_type) {
    case MV_TYPE_16X16:
        mpeg_motion_lowres(s, dest_y, dest_cb, dest_cr,
                           0, 0, 0,
                           ref_picture, pix_op,
                           s->mv[dir][0][0], s->mv[dir][0][1],
                           2 * block_s, mb_y);
        break;
    case MV_TYPE_8X8:
        mx = 0;
        my = 0;
        for (i = 0; i < 4; i++) {
            hpel_motion_lowres(s, dest_y + ((i & 1) + (i >> 1) *
                               s->linesize) * block_s,
                               ref_picture[0], 0, 0,
                               (2 * mb_x + (i & 1)) * block_s,
                               (2 * mb_y + (i >> 1)) * block_s,
                               s->width, s->height, s->linesize,
                               s->h_edge_pos >> lowres, s->v_edge_pos >> lowres,
                               block_s, block_s, pix_op,
                               s->mv[dir][i][0], s->mv[dir][i][1]);

            mx += s->mv[dir][i][0];
            my += s->mv[dir][i][1];
        }

1752
        if (!CONFIG_GRAY || !(s->avctx->flags & AV_CODEC_FLAG_GRAY))
1753 1754 1755 1756 1757 1758 1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772
            chroma_4mv_motion_lowres(s, dest_cb, dest_cr, ref_picture,
                                     pix_op, mx, my);
        break;
    case MV_TYPE_FIELD:
        if (s->picture_structure == PICT_FRAME) {
            /* top field */
            mpeg_motion_lowres(s, dest_y, dest_cb, dest_cr,
                               1, 0, s->field_select[dir][0],
                               ref_picture, pix_op,
                               s->mv[dir][0][0], s->mv[dir][0][1],
                               block_s, mb_y);
            /* bottom field */
            mpeg_motion_lowres(s, dest_y, dest_cb, dest_cr,
                               1, 1, s->field_select[dir][1],
                               ref_picture, pix_op,
                               s->mv[dir][1][0], s->mv[dir][1][1],
                               block_s, mb_y);
        } else {
            if (s->picture_structure != s->field_select[dir][0] + 1 &&
                s->pict_type != AV_PICTURE_TYPE_B && !s->first_field) {
1773
                ref_picture = s->current_picture_ptr->f->data;
1774 1775 1776 1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 1789 1790

            }
            mpeg_motion_lowres(s, dest_y, dest_cb, dest_cr,
                               0, 0, s->field_select[dir][0],
                               ref_picture, pix_op,
                               s->mv[dir][0][0],
                               s->mv[dir][0][1], 2 * block_s, mb_y >> 1);
            }
        break;
    case MV_TYPE_16X8:
        for (i = 0; i < 2; i++) {
            uint8_t **ref2picture;

            if (s->picture_structure == s->field_select[dir][i] + 1 ||
                s->pict_type == AV_PICTURE_TYPE_B || s->first_field) {
                ref2picture = ref_picture;
            } else {
1791
                ref2picture = s->current_picture_ptr->f->data;
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
            }

            mpeg_motion_lowres(s, dest_y, dest_cb, dest_cr,
                               0, 0, s->field_select[dir][i],
                               ref2picture, pix_op,
                               s->mv[dir][i][0], s->mv[dir][i][1] +
                               2 * block_s * i, block_s, mb_y >> 1);

            dest_y  +=  2 * block_s *  s->linesize;
            dest_cb += (2 * block_s >> s->chroma_y_shift) * s->uvlinesize;
            dest_cr += (2 * block_s >> s->chroma_y_shift) * s->uvlinesize;
        }
        break;
    case MV_TYPE_DMV:
        if (s->picture_structure == PICT_FRAME) {
            for (i = 0; i < 2; i++) {
                int j;
                for (j = 0; j < 2; j++) {
                    mpeg_motion_lowres(s, dest_y, dest_cb, dest_cr,
                                       1, j, j ^ i,
                                       ref_picture, pix_op,
                                       s->mv[dir][2 * i + j][0],
                                       s->mv[dir][2 * i + j][1],
                                       block_s, mb_y);
                }
1817
                pix_op = s->h264chroma.avg_h264_chroma_pixels_tab;
1818 1819 1820 1821 1822 1823 1824 1825 1826 1827
            }
        } else {
            for (i = 0; i < 2; i++) {
                mpeg_motion_lowres(s, dest_y, dest_cb, dest_cr,
                                   0, 0, s->picture_structure != i + 1,
                                   ref_picture, pix_op,
                                   s->mv[dir][2 * i][0],s->mv[dir][2 * i][1],
                                   2 * block_s, mb_y >> 1);

                // after put we make avg of the same block
1828
                pix_op = s->h264chroma.avg_h264_chroma_pixels_tab;
1829 1830 1831 1832

                // opposite parity is always in the same
                // frame if this is second field
                if (!s->first_field) {
1833
                    ref_picture = s->current_picture_ptr->f->data;
1834 1835 1836 1837 1838
                }
            }
        }
        break;
    default:
1839
        av_assert2(0);
1840 1841 1842
    }
}

1843 1844 1845
/**
 * find the lowest MB row referenced in the MVs
 */
1846
static int lowest_referenced_row(MpegEncContext *s, int dir)
1847 1848 1849 1850
{
    int my_max = INT_MIN, my_min = INT_MAX, qpel_shift = !s->quarter_sample;
    int my, off, i, mvs;

1851 1852
    if (s->picture_structure != PICT_FRAME || s->mcsel)
        goto unhandled;
1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868

    switch (s->mv_type) {
        case MV_TYPE_16X16:
            mvs = 1;
            break;
        case MV_TYPE_16X8:
            mvs = 2;
            break;
        case MV_TYPE_8X8:
            mvs = 4;
            break;
        default:
            goto unhandled;
    }

    for (i = 0; i < mvs; i++) {
1869
        my = s->mv[dir][i][1];
1870 1871 1872 1873
        my_max = FFMAX(my_max, my);
        my_min = FFMIN(my_min, my);
    }

1874
    off = ((FFMAX(-my_min, my_max)<<qpel_shift) + 63) >> 6;
1875

1876
    return av_clip(s->mb_y + off, 0, s->mb_height - 1);
1877 1878 1879 1880
unhandled:
    return s->mb_height-1;
}

1881 1882
/* put block[] to dest[] */
static inline void put_dct(MpegEncContext *s,
Diego Biurrun's avatar
Diego Biurrun committed
1883
                           int16_t *block, int i, uint8_t *dest, int line_size, int qscale)
1884
{
1885
    s->dct_unquantize_intra(s, block, i, qscale);
1886
    s->idsp.idct_put(dest, line_size, block);
1887
}
1888

1889 1890
/* add block[] to dest[] */
static inline void add_dct(MpegEncContext *s,
Diego Biurrun's avatar
Diego Biurrun committed
1891
                           int16_t *block, int i, uint8_t *dest, int line_size)
1892 1893
{
    if (s->block_last_index[i] >= 0) {
1894
        s->idsp.idct_add(dest, line_size, block);
1895 1896
    }
}
1897

1898
static inline void add_dequant_dct(MpegEncContext *s,
Diego Biurrun's avatar
Diego Biurrun committed
1899
                           int16_t *block, int i, uint8_t *dest, int line_size, int qscale)
1900
{
Fabrice Bellard's avatar
Fabrice Bellard committed
1901
    if (s->block_last_index[i] >= 0) {
1902
        s->dct_unquantize_inter(s, block, i, qscale);
1903

1904
        s->idsp.idct_add(dest, line_size, block);
Fabrice Bellard's avatar
Fabrice Bellard committed
1905 1906 1907
    }
}

1908
/**
1909
 * Clean dc, ac, coded_block for the current non-intra MB.
1910 1911 1912
 */
void ff_clean_intra_table_entries(MpegEncContext *s)
{
1913
    int wrap = s->b8_stride;
1914
    int xy = s->block_index[0];
1915 1916 1917

    s->dc_val[0][xy           ] =
    s->dc_val[0][xy + 1       ] =
1918 1919 1920
    s->dc_val[0][xy     + wrap] =
    s->dc_val[0][xy + 1 + wrap] = 1024;
    /* ac pred */
1921 1922
    memset(s->ac_val[0][xy       ], 0, 32 * sizeof(int16_t));
    memset(s->ac_val[0][xy + wrap], 0, 32 * sizeof(int16_t));
1923 1924 1925 1926 1927 1928 1929
    if (s->msmpeg4_version>=3) {
        s->coded_block[xy           ] =
        s->coded_block[xy + 1       ] =
        s->coded_block[xy     + wrap] =
        s->coded_block[xy + 1 + wrap] = 0;
    }
    /* chroma */
1930 1931
    wrap = s->mb_stride;
    xy = s->mb_x + s->mb_y * wrap;
1932 1933 1934
    s->dc_val[1][xy] =
    s->dc_val[2][xy] = 1024;
    /* ac pred */
1935 1936
    memset(s->ac_val[1][xy], 0, 16 * sizeof(int16_t));
    memset(s->ac_val[2][xy], 0, 16 * sizeof(int16_t));
1937

1938
    s->mbintra_table[xy]= 0;
1939 1940
}

Fabrice Bellard's avatar
Fabrice Bellard committed
1941 1942 1943 1944 1945 1946 1947 1948 1949 1950
/* generic function called after a macroblock has been parsed by the
   decoder or after it has been encoded by the encoder.

   Important variables used:
   s->mb_intra : true if intra macroblock
   s->mv_dir   : motion vector direction
   s->mv_type  : motion vector type
   s->mv       : motion vector
   s->interlaced_dct : true if interlaced dct used (mpeg2)
 */
1951
static av_always_inline
1952
void mpv_reconstruct_mb_internal(MpegEncContext *s, int16_t block[12][64],
1953
                            int lowres_flag, int is_mpeg12)
Fabrice Bellard's avatar
Fabrice Bellard committed
1954
{
1955
    const int mb_xy = s->mb_y * s->mb_stride + s->mb_x;
1956

1957 1958 1959
    if (CONFIG_XVMC &&
        s->avctx->hwaccel && s->avctx->hwaccel->decode_mb) {
        s->avctx->hwaccel->decode_mb(s);//xvmc uses pblocks
Ivan Kalvachev's avatar
Ivan Kalvachev committed
1960 1961
        return;
    }
Fabrice Bellard's avatar
Fabrice Bellard committed
1962

1963
    if(s->avctx->debug&FF_DEBUG_DCT_COEFF) {
1964
       /* print DCT coefficients */
1965
       int i,j;
1966 1967 1968
       av_log(s->avctx, AV_LOG_DEBUG, "DCT coeffs of MB at %dx%d:\n", s->mb_x, s->mb_y);
       for(i=0; i<6; i++){
           for(j=0; j<64; j++){
1969 1970
               av_log(s->avctx, AV_LOG_DEBUG, "%5d",
                      block[i][s->idsp.idct_permutation[j]]);
1971 1972 1973
           }
           av_log(s->avctx, AV_LOG_DEBUG, "\n");
       }
1974 1975
    }

1976
    s->current_picture.qscale_table[mb_xy] = s->qscale;
1977

Fabrice Bellard's avatar
Fabrice Bellard committed
1978 1979
    /* update DC predictors for P macroblocks */
    if (!s->mb_intra) {
Keiji Costantini's avatar
Keiji Costantini committed
1980
        if (!is_mpeg12 && (s->h263_pred || s->h263_aic)) {
1981
            if(s->mbintra_table[mb_xy])
1982
                ff_clean_intra_table_entries(s);
Fabrice Bellard's avatar
Fabrice Bellard committed
1983
        } else {
1984 1985
            s->last_dc[0] =
            s->last_dc[1] =
Fabrice Bellard's avatar
Fabrice Bellard committed
1986 1987 1988
            s->last_dc[2] = 128 << s->intra_dc_precision;
        }
    }
Keiji Costantini's avatar
Keiji Costantini committed
1989
    else if (!is_mpeg12 && (s->h263_pred || s->h263_aic))
1990
        s->mbintra_table[mb_xy]=1;
1991

1992
    if ((s->avctx->flags & AV_CODEC_FLAG_PSNR) || s->frame_skip_threshold || s->frame_skip_factor ||
1993 1994
        !(s->encoding && (s->intra_only || s->pict_type == AV_PICTURE_TYPE_B) &&
          s->avctx->mb_decision != FF_MB_DECISION_RD)) { // FIXME precalc
1995
        uint8_t *dest_y, *dest_cb, *dest_cr;
1996
        int dct_linesize, dct_offset;
Michael Niedermayer's avatar
Michael Niedermayer committed
1997 1998
        op_pixels_func (*op_pix)[4];
        qpel_mc_func (*op_qpix)[16];
1999 2000
        const int linesize   = s->current_picture.f->linesize[0]; //not s->linesize as this would be wrong for field pics
        const int uvlinesize = s->current_picture.f->linesize[1];
2001 2002
        const int readable= s->pict_type != AV_PICTURE_TYPE_B || s->encoding || s->avctx->draw_horiz_band || lowres_flag;
        const int block_size= lowres_flag ? 8>>s->avctx->lowres : 8;
2003

Michael Niedermayer's avatar
Michael Niedermayer committed
2004 2005 2006
        /* avoid copy if macroblock skipped in last frame too */
        /* skip only during decoding as we might trash the buffers during encoding a bit */
        if(!s->encoding){
2007
            uint8_t *mbskip_ptr = &s->mbskip_table[mb_xy];
Michael Niedermayer's avatar
Michael Niedermayer committed
2008

2009 2010
            if (s->mb_skipped) {
                s->mb_skipped= 0;
2011
                av_assert2(s->pict_type!=AV_PICTURE_TYPE_I);
2012
                *mbskip_ptr = 1;
2013
            } else if(!s->current_picture.reference) {
2014
                *mbskip_ptr = 1;
2015
            } else{
2016 2017
                *mbskip_ptr = 0; /* not skipped */
            }
Michael Niedermayer's avatar
Michael Niedermayer committed
2018
        }
2019

2020
        dct_linesize = linesize << s->interlaced_dct;
2021
        dct_offset   = s->interlaced_dct ? linesize : linesize * block_size;
2022

2023 2024 2025 2026 2027
        if(readable){
            dest_y=  s->dest[0];
            dest_cb= s->dest[1];
            dest_cr= s->dest[2];
        }else{
2028 2029 2030
            dest_y = s->sc.b_scratchpad;
            dest_cb= s->sc.b_scratchpad+16*linesize;
            dest_cr= s->sc.b_scratchpad+32*linesize;
2031
        }
2032

Fabrice Bellard's avatar
Fabrice Bellard committed
2033 2034
        if (!s->mb_intra) {
            /* motion handling */
2035
            /* decoding or more than one mb_type (MC was already done otherwise) */
2036
            if(!s->encoding){
2037

2038
                if(HAVE_THREADS && s->avctx->active_thread_type&FF_THREAD_FRAME) {
2039
                    if (s->mv_dir & MV_DIR_FORWARD) {
2040
                        ff_thread_await_progress(&s->last_picture_ptr->tf,
2041
                                                 lowest_referenced_row(s, 0),
2042
                                                 0);
2043 2044
                    }
                    if (s->mv_dir & MV_DIR_BACKWARD) {
2045
                        ff_thread_await_progress(&s->next_picture_ptr->tf,
2046
                                                 lowest_referenced_row(s, 1),
2047
                                                 0);
2048 2049 2050
                    }
                }

2051
                if(lowres_flag){
2052
                    h264_chroma_mc_func *op_pix = s->h264chroma.put_h264_chroma_pixels_tab;
2053 2054

                    if (s->mv_dir & MV_DIR_FORWARD) {
2055
                        MPV_motion_lowres(s, dest_y, dest_cb, dest_cr, 0, s->last_picture.f->data, op_pix);
2056
                        op_pix = s->h264chroma.avg_h264_chroma_pixels_tab;
2057 2058
                    }
                    if (s->mv_dir & MV_DIR_BACKWARD) {
2059
                        MPV_motion_lowres(s, dest_y, dest_cb, dest_cr, 1, s->next_picture.f->data, op_pix);
2060
                    }
2061
                }else{
2062
                    op_qpix = s->me.qpel_put;
2063
                    if ((!s->no_rounding) || s->pict_type==AV_PICTURE_TYPE_B){
2064
                        op_pix = s->hdsp.put_pixels_tab;
2065
                    }else{
2066
                        op_pix = s->hdsp.put_no_rnd_pixels_tab;
2067 2068
                    }
                    if (s->mv_dir & MV_DIR_FORWARD) {
2069
                        ff_mpv_motion(s, dest_y, dest_cb, dest_cr, 0, s->last_picture.f->data, op_pix, op_qpix);
2070
                        op_pix = s->hdsp.avg_pixels_tab;
2071 2072 2073
                        op_qpix= s->me.qpel_avg;
                    }
                    if (s->mv_dir & MV_DIR_BACKWARD) {
2074
                        ff_mpv_motion(s, dest_y, dest_cb, dest_cr, 1, s->next_picture.f->data, op_pix, op_qpix);
2075
                    }
2076
                }
Fabrice Bellard's avatar
Fabrice Bellard committed
2077 2078
            }

2079
            /* skip dequant / idct if we are really late ;) */
Michael Niedermayer's avatar
Michael Niedermayer committed
2080
            if(s->avctx->skip_idct){
2081 2082
                if(  (s->avctx->skip_idct >= AVDISCARD_NONREF && s->pict_type == AV_PICTURE_TYPE_B)
                   ||(s->avctx->skip_idct >= AVDISCARD_NONKEY && s->pict_type != AV_PICTURE_TYPE_I)
Michael Niedermayer's avatar
Michael Niedermayer committed
2083 2084 2085
                   || s->avctx->skip_idct >= AVDISCARD_ALL)
                    goto skip_idct;
            }
2086

Fabrice Bellard's avatar
Fabrice Bellard committed
2087
            /* add dct residue */
2088 2089
            if(s->encoding || !(   s->msmpeg4_version || s->codec_id==AV_CODEC_ID_MPEG1VIDEO || s->codec_id==AV_CODEC_ID_MPEG2VIDEO
                                || (s->codec_id==AV_CODEC_ID_MPEG4 && !s->mpeg_quant))){
2090
                add_dequant_dct(s, block[0], 0, dest_y                          , dct_linesize, s->qscale);
2091
                add_dequant_dct(s, block[1], 1, dest_y              + block_size, dct_linesize, s->qscale);
2092
                add_dequant_dct(s, block[2], 2, dest_y + dct_offset             , dct_linesize, s->qscale);
2093
                add_dequant_dct(s, block[3], 3, dest_y + dct_offset + block_size, dct_linesize, s->qscale);
2094

2095
                if (!CONFIG_GRAY || !(s->avctx->flags & AV_CODEC_FLAG_GRAY)) {
2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106
                    if (s->chroma_y_shift){
                        add_dequant_dct(s, block[4], 4, dest_cb, uvlinesize, s->chroma_qscale);
                        add_dequant_dct(s, block[5], 5, dest_cr, uvlinesize, s->chroma_qscale);
                    }else{
                        dct_linesize >>= 1;
                        dct_offset >>=1;
                        add_dequant_dct(s, block[4], 4, dest_cb,              dct_linesize, s->chroma_qscale);
                        add_dequant_dct(s, block[5], 5, dest_cr,              dct_linesize, s->chroma_qscale);
                        add_dequant_dct(s, block[6], 6, dest_cb + dct_offset, dct_linesize, s->chroma_qscale);
                        add_dequant_dct(s, block[7], 7, dest_cr + dct_offset, dct_linesize, s->chroma_qscale);
                    }
2107
                }
2108
            } else if(is_mpeg12 || (s->codec_id != AV_CODEC_ID_WMV2)){
2109
                add_dct(s, block[0], 0, dest_y                          , dct_linesize);
2110
                add_dct(s, block[1], 1, dest_y              + block_size, dct_linesize);
2111
                add_dct(s, block[2], 2, dest_y + dct_offset             , dct_linesize);
2112
                add_dct(s, block[3], 3, dest_y + dct_offset + block_size, dct_linesize);
Fabrice Bellard's avatar
Fabrice Bellard committed
2113

2114
                if (!CONFIG_GRAY || !(s->avctx->flags & AV_CODEC_FLAG_GRAY)) {
2115 2116 2117 2118 2119 2120
                    if(s->chroma_y_shift){//Chroma420
                        add_dct(s, block[4], 4, dest_cb, uvlinesize);
                        add_dct(s, block[5], 5, dest_cr, uvlinesize);
                    }else{
                        //chroma422
                        dct_linesize = uvlinesize << s->interlaced_dct;
2121
                        dct_offset   = s->interlaced_dct ? uvlinesize : uvlinesize*block_size;
2122 2123 2124 2125 2126 2127

                        add_dct(s, block[4], 4, dest_cb, dct_linesize);
                        add_dct(s, block[5], 5, dest_cr, dct_linesize);
                        add_dct(s, block[6], 6, dest_cb+dct_offset, dct_linesize);
                        add_dct(s, block[7], 7, dest_cr+dct_offset, dct_linesize);
                        if(!s->chroma_x_shift){//Chroma444
2128 2129 2130 2131
                            add_dct(s, block[8], 8, dest_cb+block_size, dct_linesize);
                            add_dct(s, block[9], 9, dest_cr+block_size, dct_linesize);
                            add_dct(s, block[10], 10, dest_cb+block_size+dct_offset, dct_linesize);
                            add_dct(s, block[11], 11, dest_cr+block_size+dct_offset, dct_linesize);
2132 2133 2134 2135
                        }
                    }
                }//fi gray
            }
2136
            else if (CONFIG_WMV2_DECODER || CONFIG_WMV2_ENCODER) {
Michael Niedermayer's avatar
Michael Niedermayer committed
2137
                ff_wmv2_add_mb(s, block, dest_y, dest_cb, dest_cr);
2138
            }
Fabrice Bellard's avatar
Fabrice Bellard committed
2139
        } else {
2140 2141 2142 2143
            /* Only MPEG-4 Simple Studio Profile is supported in > 8-bit mode.
               TODO: Integrate 10-bit properly into mpegvideo.c so that ER works properly */
            if (s->avctx->bits_per_raw_sample > 8){
                const int act_block_size = block_size * 2;
2144 2145 2146 2147 2148 2149 2150 2151 2152 2153 2154 2155 2156 2157 2158 2159 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

                if(s->dpcm_direction == 0) {
                    s->idsp.idct_put(dest_y,                           dct_linesize, (int16_t*)(*s->block32)[0]);
                    s->idsp.idct_put(dest_y              + act_block_size, dct_linesize, (int16_t*)(*s->block32)[1]);
                    s->idsp.idct_put(dest_y + dct_offset,              dct_linesize, (int16_t*)(*s->block32)[2]);
                    s->idsp.idct_put(dest_y + dct_offset + act_block_size, dct_linesize, (int16_t*)(*s->block32)[3]);

                    dct_linesize = uvlinesize << s->interlaced_dct;
                    dct_offset   = s->interlaced_dct ? uvlinesize : uvlinesize*block_size;

                    s->idsp.idct_put(dest_cb,              dct_linesize, (int16_t*)(*s->block32)[4]);
                    s->idsp.idct_put(dest_cr,              dct_linesize, (int16_t*)(*s->block32)[5]);
                    s->idsp.idct_put(dest_cb + dct_offset, dct_linesize, (int16_t*)(*s->block32)[6]);
                    s->idsp.idct_put(dest_cr + dct_offset, dct_linesize, (int16_t*)(*s->block32)[7]);
                    if(!s->chroma_x_shift){//Chroma444
                        s->idsp.idct_put(dest_cb + act_block_size,              dct_linesize, (int16_t*)(*s->block32)[8]);
                        s->idsp.idct_put(dest_cr + act_block_size,              dct_linesize, (int16_t*)(*s->block32)[9]);
                        s->idsp.idct_put(dest_cb + act_block_size + dct_offset, dct_linesize, (int16_t*)(*s->block32)[10]);
                        s->idsp.idct_put(dest_cr + act_block_size + dct_offset, dct_linesize, (int16_t*)(*s->block32)[11]);
                    }
                } else if(s->dpcm_direction == 1) {
                    int i, w, h;
                    uint16_t *dest_pcm[3] = {(uint16_t*)dest_y, (uint16_t*)dest_cb, (uint16_t*)dest_cr};
                    int linesize[3] = {dct_linesize, uvlinesize, uvlinesize};
                    for(i = 0; i < 3; i++) {
                        int idx = 0;
                        int vsub = i ? s->chroma_y_shift : 0;
                        int hsub = i ? s->chroma_x_shift : 0;
                        for(h = 0; h < (16 >> vsub); h++){
                            for(w = 0; w < (16 >> hsub); w++)
                                dest_pcm[i][w] = (*s->dpcm_macroblock)[i][idx++];
                            dest_pcm[i] += linesize[i] / 2;
                        }
                    }
                } else if(s->dpcm_direction == -1) {
                    int i, w, h;
                    uint16_t *dest_pcm[3] = {(uint16_t*)dest_y, (uint16_t*)dest_cb, (uint16_t*)dest_cr};
                    int linesize[3] = {dct_linesize, uvlinesize, uvlinesize};
                    for(i = 0; i < 3; i++) {
                        int idx = 0;
                        int vsub = i ? s->chroma_y_shift : 0;
                        int hsub = i ? s->chroma_x_shift : 0;
                        dest_pcm[i] += (linesize[i] / 2) * ((16 >> vsub) - 1);
                        for(h = (16 >> vsub)-1; h >= 1; h--){
                            for(w = (16 >> hsub)-1; w >= 1; w--)
                                dest_pcm[i][w] = (*s->dpcm_macroblock)[i][idx++];
                            dest_pcm[i] -= linesize[i] / 2;
                        }
                    }
2193 2194
                }
            }
Fabrice Bellard's avatar
Fabrice Bellard committed
2195
            /* dct only in intra block */
2196
            else if(s->encoding || !(s->codec_id==AV_CODEC_ID_MPEG1VIDEO || s->codec_id==AV_CODEC_ID_MPEG2VIDEO)){
2197 2198 2199 2200
                put_dct(s, block[0], 0, dest_y                          , dct_linesize, s->qscale);
                put_dct(s, block[1], 1, dest_y              + block_size, dct_linesize, s->qscale);
                put_dct(s, block[2], 2, dest_y + dct_offset             , dct_linesize, s->qscale);
                put_dct(s, block[3], 3, dest_y + dct_offset + block_size, dct_linesize, s->qscale);
2201

2202
                if (!CONFIG_GRAY || !(s->avctx->flags & AV_CODEC_FLAG_GRAY)) {
2203 2204 2205 2206 2207 2208 2209 2210 2211
                    if(s->chroma_y_shift){
                        put_dct(s, block[4], 4, dest_cb, uvlinesize, s->chroma_qscale);
                        put_dct(s, block[5], 5, dest_cr, uvlinesize, s->chroma_qscale);
                    }else{
                        dct_offset >>=1;
                        dct_linesize >>=1;
                        put_dct(s, block[4], 4, dest_cb,              dct_linesize, s->chroma_qscale);
                        put_dct(s, block[5], 5, dest_cr,              dct_linesize, s->chroma_qscale);
                        put_dct(s, block[6], 6, dest_cb + dct_offset, dct_linesize, s->chroma_qscale);
2212
                        put_dct(s, block[7], 7, dest_cr + dct_offset, dct_linesize, s->chroma_qscale);
2213 2214 2215
                    }
                }
            }else{
2216 2217 2218 2219
                s->idsp.idct_put(dest_y,                           dct_linesize, block[0]);
                s->idsp.idct_put(dest_y              + block_size, dct_linesize, block[1]);
                s->idsp.idct_put(dest_y + dct_offset,              dct_linesize, block[2]);
                s->idsp.idct_put(dest_y + dct_offset + block_size, dct_linesize, block[3]);
2220

2221
                if (!CONFIG_GRAY || !(s->avctx->flags & AV_CODEC_FLAG_GRAY)) {
2222
                    if(s->chroma_y_shift){
2223 2224
                        s->idsp.idct_put(dest_cb, uvlinesize, block[4]);
                        s->idsp.idct_put(dest_cr, uvlinesize, block[5]);
2225
                    }else{
2226

2227
                        dct_linesize = uvlinesize << s->interlaced_dct;
2228
                        dct_offset   = s->interlaced_dct ? uvlinesize : uvlinesize*block_size;
2229

2230 2231 2232 2233
                        s->idsp.idct_put(dest_cb,              dct_linesize, block[4]);
                        s->idsp.idct_put(dest_cr,              dct_linesize, block[5]);
                        s->idsp.idct_put(dest_cb + dct_offset, dct_linesize, block[6]);
                        s->idsp.idct_put(dest_cr + dct_offset, dct_linesize, block[7]);
2234
                        if(!s->chroma_x_shift){//Chroma444
2235 2236 2237 2238
                            s->idsp.idct_put(dest_cb + block_size,              dct_linesize, block[8]);
                            s->idsp.idct_put(dest_cr + block_size,              dct_linesize, block[9]);
                            s->idsp.idct_put(dest_cb + block_size + dct_offset, dct_linesize, block[10]);
                            s->idsp.idct_put(dest_cr + block_size + dct_offset, dct_linesize, block[11]);
2239 2240 2241
                        }
                    }
                }//gray
2242
            }
2243
        }
2244 2245
skip_idct:
        if(!readable){
2246
            s->hdsp.put_pixels_tab[0][0](s->dest[0], dest_y ,   linesize,16);
2247
            if (!CONFIG_GRAY || !(s->avctx->flags & AV_CODEC_FLAG_GRAY)) {
2248 2249
                s->hdsp.put_pixels_tab[s->chroma_x_shift][0](s->dest[1], dest_cb, uvlinesize,16 >> s->chroma_y_shift);
                s->hdsp.put_pixels_tab[s->chroma_x_shift][0](s->dest[2], dest_cr, uvlinesize,16 >> s->chroma_y_shift);
2250
            }
2251
        }
2252 2253 2254
    }
}

2255
void ff_mpv_reconstruct_mb(MpegEncContext *s, int16_t block[12][64])
2256
{
2257
#if !CONFIG_SMALL
Keiji Costantini's avatar
Keiji Costantini committed
2258
    if(s->out_format == FMT_MPEG1) {
2259 2260
        if(s->avctx->lowres) mpv_reconstruct_mb_internal(s, block, 1, 1);
        else                 mpv_reconstruct_mb_internal(s, block, 0, 1);
Keiji Costantini's avatar
Keiji Costantini committed
2261 2262
    } else
#endif
2263 2264
    if(s->avctx->lowres) mpv_reconstruct_mb_internal(s, block, 1, 0);
    else                  mpv_reconstruct_mb_internal(s, block, 0, 0);
2265 2266
}

2267 2268
void ff_mpeg_draw_horiz_band(MpegEncContext *s, int y, int h)
{
2269 2270
    ff_draw_horiz_band(s->avctx, s->current_picture_ptr->f,
                       s->last_picture_ptr ? s->last_picture_ptr->f : NULL, y, h, s->picture_structure,
2271
                       s->first_field, s->low_delay);
2272 2273
}

2274
void ff_init_block_index(MpegEncContext *s){ //FIXME maybe rename
2275 2276
    const int linesize   = s->current_picture.f->linesize[0]; //not s->linesize as this would be wrong for field pics
    const int uvlinesize = s->current_picture.f->linesize[1];
2277 2278
    const int width_of_mb = (4 + (s->avctx->bits_per_raw_sample > 8)) - s->avctx->lowres;
    const int height_of_mb = 4 - s->avctx->lowres;
2279

2280 2281 2282 2283 2284 2285 2286
    s->block_index[0]= s->b8_stride*(s->mb_y*2    ) - 2 + s->mb_x*2;
    s->block_index[1]= s->b8_stride*(s->mb_y*2    ) - 1 + s->mb_x*2;
    s->block_index[2]= s->b8_stride*(s->mb_y*2 + 1) - 2 + s->mb_x*2;
    s->block_index[3]= s->b8_stride*(s->mb_y*2 + 1) - 1 + s->mb_x*2;
    s->block_index[4]= s->mb_stride*(s->mb_y + 1)                + s->b8_stride*s->mb_height*2 + s->mb_x - 1;
    s->block_index[5]= s->mb_stride*(s->mb_y + s->mb_height + 2) + s->b8_stride*s->mb_height*2 + s->mb_x - 1;
    //block_index is not used by mpeg2, so it is not affected by chroma_format
2287

2288 2289 2290
    s->dest[0] = s->current_picture.f->data[0] + (int)((s->mb_x - 1U) <<  width_of_mb);
    s->dest[1] = s->current_picture.f->data[1] + (int)((s->mb_x - 1U) << (width_of_mb - s->chroma_x_shift));
    s->dest[2] = s->current_picture.f->data[2] + (int)((s->mb_x - 1U) << (width_of_mb - s->chroma_x_shift));
2291

2292
    if(!(s->pict_type==AV_PICTURE_TYPE_B && s->avctx->draw_horiz_band && s->picture_structure==PICT_FRAME))
2293
    {
2294
        if(s->picture_structure==PICT_FRAME){
2295 2296 2297
        s->dest[0] += s->mb_y *   linesize << height_of_mb;
        s->dest[1] += s->mb_y * uvlinesize << (height_of_mb - s->chroma_y_shift);
        s->dest[2] += s->mb_y * uvlinesize << (height_of_mb - s->chroma_y_shift);
2298
        }else{
2299 2300 2301
            s->dest[0] += (s->mb_y>>1) *   linesize << height_of_mb;
            s->dest[1] += (s->mb_y>>1) * uvlinesize << (height_of_mb - s->chroma_y_shift);
            s->dest[2] += (s->mb_y>>1) * uvlinesize << (height_of_mb - s->chroma_y_shift);
2302
            av_assert1((s->mb_y&1) == (s->picture_structure == PICT_BOTTOM_FIELD));
2303
        }
2304 2305
    }
}
2306

2307 2308 2309
void ff_mpeg_flush(AVCodecContext *avctx){
    int i;
    MpegEncContext *s = avctx->priv_data;
2310

2311
    if (!s || !s->picture)
2312
        return;
2313

2314
    for (i = 0; i < MAX_PICTURE_COUNT; i++)
2315
        ff_mpeg_unref_picture(s->avctx, &s->picture[i]);
2316
    s->current_picture_ptr = s->last_picture_ptr = s->next_picture_ptr = NULL;
2317

2318 2319 2320
    ff_mpeg_unref_picture(s->avctx, &s->current_picture);
    ff_mpeg_unref_picture(s->avctx, &s->last_picture);
    ff_mpeg_unref_picture(s->avctx, &s->next_picture);
2321

2322
    s->mb_x= s->mb_y= 0;
2323
    s->closed_gop= 0;
2324

2325 2326 2327 2328 2329 2330 2331 2332
    s->parse_context.state= -1;
    s->parse_context.frame_start_found= 0;
    s->parse_context.overread= 0;
    s->parse_context.overread_index= 0;
    s->parse_context.index= 0;
    s->parse_context.last_index= 0;
    s->bitstream_buffer_size=0;
    s->pp_time=0;
Fabrice Bellard's avatar
Fabrice Bellard committed
2333 2334
}

2335 2336 2337 2338 2339 2340 2341 2342 2343 2344 2345 2346 2347 2348 2349 2350
/**
 * set qscale and update qscale dependent variables.
 */
void ff_set_qscale(MpegEncContext * s, int qscale)
{
    if (qscale < 1)
        qscale = 1;
    else if (qscale > 31)
        qscale = 31;

    s->qscale = qscale;
    s->chroma_qscale= s->chroma_qscale_table[qscale];

    s->y_dc_scale= s->y_dc_scale_table[ qscale ];
    s->c_dc_scale= s->c_dc_scale_table[ s->chroma_qscale ];
}
2351

2352
void ff_mpv_report_decode_progress(MpegEncContext *s)
2353
{
2354
    if (s->pict_type != AV_PICTURE_TYPE_B && !s->partitioned_frame && !s->er.error_occurred)
2355
        ff_thread_report_progress(&s->current_picture_ptr->tf, s->mb_y, 0);
2356
}