mpegvideo.c 105 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;
221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248
        qadd = (qscale - 1) | 1;
    }else{
        qadd = 0;
    }
    if(s->ac_pred)
        nCoeffs=63;
    else
        nCoeffs= s->inter_scantable.raster_end[ s->block_last_index[n] ];

    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

327 328 329 330 331 332 333
    return 0;
}

av_cold void ff_mpv_idct_init(MpegEncContext *s)
{
    ff_idctdsp_init(&s->idsp, s->avctx);

334
    /* load & permutate scantables
335 336 337
     * note: only wmv uses different ones
     */
    if (s->alternate_scan) {
338 339
        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);
340
    } else {
341 342
        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);
343
    }
344 345
    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);
346 347
}

348 349 350 351
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,
352
                            s->mb_stride, s->mb_width, s->mb_height, s->b8_stride,
353 354 355
                            &s->linesize, &s->uvlinesize);
}

356
static int init_duplicate_context(MpegEncContext *s)
357
{
358 359 360
    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;
361 362
    int i;

363 364 365
    if (s->mb_height & 1)
        yc_size += 2*s->b8_stride + 2*s->mb_stride;

366
    s->sc.edge_emu_buffer =
367 368
    s->me.scratchpad   =
    s->me.temp         =
369 370 371
    s->sc.rd_scratchpad   =
    s->sc.b_scratchpad    =
    s->sc.obmc_scratchpad = NULL;
372

373
    if (s->encoding) {
374 375 376 377
        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)
378
        if (s->noise_reduction) {
379 380
            FF_ALLOCZ_OR_GOTO(s->avctx, s->dct_error_sum,
                              2 * 64 * sizeof(int), fail)
381
        }
382
    }
Diego Biurrun's avatar
Diego Biurrun committed
383
    FF_ALLOCZ_OR_GOTO(s->avctx, s->blocks, 64 * 12 * 2 * sizeof(int16_t), fail)
384
    s->block = s->blocks[0];
385

386
    for (i = 0; i < 12; i++) {
387
        s->pblocks[i] = &s->block[i];
388
    }
389 390
    if (s->avctx->codec_tag == AV_RL32("VCR2")) {
        // exchange uv
391
        FFSWAP(void *, s->pblocks[4], s->pblocks[5]);
392
    }
393

394 395
    if (s->out_format == FMT_H263) {
        /* ac values */
396 397
        FF_ALLOCZ_OR_GOTO(s->avctx, s->ac_val_base,
                          yc_size * sizeof(int16_t) * 16, fail);
398 399 400 401 402
        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;
    }

403 404
    return 0;
fail:
405
    return -1; // free() through ff_mpv_common_end()
406 407
}

408 409
static void free_duplicate_context(MpegEncContext *s)
{
410
    if (!s)
411
        return;
412

413
    av_freep(&s->sc.edge_emu_buffer);
414
    av_freep(&s->me.scratchpad);
415
    s->me.temp =
416 417 418
    s->sc.rd_scratchpad =
    s->sc.b_scratchpad =
    s->sc.obmc_scratchpad = NULL;
419

420 421 422 423
    av_freep(&s->dct_error_sum);
    av_freep(&s->me.map);
    av_freep(&s->me.score_map);
    av_freep(&s->blocks);
424
    av_freep(&s->ac_val_base);
425
    s->block = NULL;
426 427
}

428 429 430
static void backup_duplicate_context(MpegEncContext *bak, MpegEncContext *src)
{
#define COPY(a) bak->a = src->a
431
    COPY(sc.edge_emu_buffer);
432
    COPY(me.scratchpad);
433
    COPY(me.temp);
434 435 436
    COPY(sc.rd_scratchpad);
    COPY(sc.b_scratchpad);
    COPY(sc.obmc_scratchpad);
437 438 439 440 441 442 443 444 445
    COPY(me.map);
    COPY(me.score_map);
    COPY(blocks);
    COPY(block);
    COPY(start_mb_y);
    COPY(end_mb_y);
    COPY(me.map_generation);
    COPY(pb);
    COPY(dct_error_sum);
446 447
    COPY(dct_count[0]);
    COPY(dct_count[1]);
448 449 450 451
    COPY(ac_val_base);
    COPY(ac_val[0]);
    COPY(ac_val[1]);
    COPY(ac_val[2]);
452 453 454
#undef COPY
}

455
int ff_update_duplicate_context(MpegEncContext *dst, MpegEncContext *src)
456
{
457
    MpegEncContext bak;
458
    int i, ret;
459 460
    // FIXME copy only needed parts
    // START_TIMER
461 462 463
    backup_duplicate_context(&bak, dst);
    memcpy(dst, src, sizeof(MpegEncContext));
    backup_duplicate_context(dst, &bak);
464
    for (i = 0; i < 12; i++) {
465
        dst->pblocks[i] = &dst->block[i];
466
    }
467 468
    if (dst->avctx->codec_tag == AV_RL32("VCR2")) {
        // exchange uv
469
        FFSWAP(void *, dst->pblocks[4], dst->pblocks[5]);
470
    }
471
    if (!dst->sc.edge_emu_buffer &&
472 473
        (ret = ff_mpeg_framesize_alloc(dst->avctx, &dst->me,
                                       &dst->sc, dst->linesize)) < 0) {
474 475 476 477
        av_log(dst->avctx, AV_LOG_ERROR, "failed to allocate context "
               "scratch buffers.\n");
        return ret;
    }
478 479
    // STOP_TIMER("update_duplicate_context")
    // about 10k cycles / 0.01 sec for  1000frames on 1ghz with 2 threads
480
    return 0;
481 482
}

483 484
int ff_mpeg_update_thread_context(AVCodecContext *dst,
                                  const AVCodecContext *src)
485
{
486
    int i, ret;
487 488
    MpegEncContext *s = dst->priv_data, *s1 = src->priv_data;

489
    if (dst == src)
490
        return 0;
491

492 493
    av_assert0(s != s1);

494 495 496
    // FIXME can parameters change on I-frames?
    // in that case dst may need a reinit
    if (!s->context_initialized) {
497
        int err;
498 499 500 501 502 503
        memcpy(s, s1, sizeof(MpegEncContext));

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

504
        if (s1->context_initialized){
505 506
//             s->picture_range_start  += MAX_PICTURE_COUNT;
//             s->picture_range_end    += MAX_PICTURE_COUNT;
507
            ff_mpv_idct_init(s);
508
            if((err = ff_mpv_common_init(s)) < 0){
509 510
                memset(s, 0, sizeof(MpegEncContext));
                s->avctx = dst;
511
                return err;
512
            }
513
        }
514 515
    }

516 517 518 519
    if (s->height != s1->height || s->width != s1->width || s->context_reinit) {
        s->context_reinit = 0;
        s->height = s1->height;
        s->width  = s1->width;
520
        if ((ret = ff_mpv_common_frame_size_change(s)) < 0)
521
            return ret;
522 523
    }

524 525 526 527 528 529 530 531
    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;

    s->coded_picture_number = s1->coded_picture_number;
    s->picture_number       = s1->picture_number;

532
    av_assert0(!s->picture || s->picture != s1->picture);
533
    if(s->picture)
534
    for (i = 0; i < MAX_PICTURE_COUNT; i++) {
535
        ff_mpeg_unref_picture(s->avctx, &s->picture[i]);
536
        if (s1->picture && s1->picture[i].f->buf[0] &&
537
            (ret = ff_mpeg_ref_picture(s->avctx, &s->picture[i], &s1->picture[i])) < 0)
538
            return ret;
539
    }
540

541 542
#define UPDATE_PICTURE(pic)\
do {\
543
    ff_mpeg_unref_picture(s->avctx, &s->pic);\
544
    if (s1->pic.f && s1->pic.f->buf[0])\
545
        ret = ff_mpeg_ref_picture(s->avctx, &s->pic, &s1->pic);\
546
    else\
547
        ret = ff_update_picture_tables(&s->pic, &s1->pic);\
548 549 550 551 552 553 554
    if (ret < 0)\
        return ret;\
} while (0)

    UPDATE_PICTURE(current_picture);
    UPDATE_PICTURE(last_picture);
    UPDATE_PICTURE(next_picture);
555

556 557 558 559 560
#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)

561 562 563
    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);
564

565
    // Error/bug resilience
566 567
    s->next_p_frame_damaged = s1->next_p_frame_damaged;
    s->workaround_bugs      = s1->workaround_bugs;
568
    s->padding_bug_score    = s1->padding_bug_score;
569

570
    // MPEG-4 timing info
571
    memcpy(&s->last_time_base, &s1->last_time_base,
572
           (char *) &s1->pb_field_time + sizeof(s1->pb_field_time) -
573
           (char *) &s1->last_time_base);
574 575 576 577

    // B-frame info
    s->max_b_frames = s1->max_b_frames;
    s->low_delay    = s1->low_delay;
578
    s->droppable    = s1->droppable;
579 580 581 582 583 584

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

    if (s1->bitstream_buffer) {
        if (s1->bitstream_buffer_size +
585
            AV_INPUT_BUFFER_PADDING_SIZE > s->allocated_bitstream_buffer_size) {
586 587 588
            av_fast_malloc(&s->bitstream_buffer,
                           &s->allocated_bitstream_buffer_size,
                           s1->allocated_bitstream_buffer_size);
589 590
            if (!s->bitstream_buffer) {
                s->bitstream_buffer_size = 0;
591
                return AVERROR(ENOMEM);
592
            }
593 594
        }
        s->bitstream_buffer_size = s1->bitstream_buffer_size;
595 596 597
        memcpy(s->bitstream_buffer, s1->bitstream_buffer,
               s1->bitstream_buffer_size);
        memset(s->bitstream_buffer + s->bitstream_buffer_size, 0,
598
               AV_INPUT_BUFFER_PADDING_SIZE);
599 600
    }

601
    // linesize-dependent scratch buffer allocation
602
    if (!s->sc.edge_emu_buffer)
603
        if (s1->linesize) {
604 605
            if (ff_mpeg_framesize_alloc(s->avctx, &s->me,
                                        &s->sc, s1->linesize) < 0) {
606 607 608 609 610 611 612 613 614
                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");
        }

615
    // MPEG-2/interlacing info
616 617
    memcpy(&s->progressive_sequence, &s1->progressive_sequence,
           (char *) &s1->rtp_mode - (char *) &s1->progressive_sequence);
618

619 620 621
    if (!s1->first_field) {
        s->last_pict_type = s1->pict_type;
        if (s1->current_picture_ptr)
622
            s->last_lambda_for[s1->pict_type] = s1->current_picture_ptr->f->quality;
623 624 625 626 627
    }

    return 0;
}

Michael Niedermayer's avatar
Michael Niedermayer committed
628
/**
629
 * Set the given MpegEncContext to common defaults
630
 * (same for encoding and decoding).
631
 * The changed fields will not depend upon the
632
 * prior state of the MpegEncContext.
Michael Niedermayer's avatar
Michael Niedermayer committed
633
 */
634
void ff_mpv_common_defaults(MpegEncContext *s)
635 636 637 638 639 640 641 642 643 644
{
    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;
645

646 647
    s->f_code                = 1;
    s->b_code                = 1;
648

649
    s->slice_context_count   = 1;
Michael Niedermayer's avatar
Michael Niedermayer committed
650 651 652
}

/**
653
 * Set the given MpegEncContext to defaults for decoding.
654 655
 * the changed fields will not depend upon
 * the prior state of the MpegEncContext.
Michael Niedermayer's avatar
Michael Niedermayer committed
656
 */
657
void ff_mpv_decode_defaults(MpegEncContext *s)
658
{
659
    ff_mpv_common_defaults(s);
Michael Niedermayer's avatar
Michael Niedermayer committed
660 661
}

662 663 664 665 666 667 668 669
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;

670 671
    /* convert fourcc to upper case */
    s->codec_tag          = avpriv_toupper4(avctx->codec_tag);
672 673
}

674 675 676 677 678 679 680 681 682 683 684 685 686
/**
 * 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;

687
    /* set default edge pos, will be overridden
688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704
     * 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;

705 706 707
    if (s->mb_height & 1)
        yc_size += 2*s->b8_stride + 2*s->mb_stride;

708
    FF_ALLOCZ_OR_GOTO(s->avctx, s->mb_index2xy, (s->mb_num + 1) * sizeof(int),
709
                      fail); // error resilience code looks cleaner with this
710 711 712 713
    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;

714
    s->mb_index2xy[s->mb_height * s->mb_width] = (s->mb_height - 1) * s->mb_stride + s->mb_width; // FIXME really needed?
715 716 717

    if (s->encoding) {
        /* Allocate MV tables */
718 719 720 721 722 723
        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)
724 725 726
        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;
727 728
        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;
729 730 731
        s->b_direct_mv_table     = s->b_direct_mv_table_base + s->mb_stride + 1;

        /* Allocate MB type table */
732
        FF_ALLOCZ_OR_GOTO(s->avctx, s->mb_type, mb_array_size * sizeof(uint16_t), fail) // needed for encoding
733

734
        FF_ALLOCZ_OR_GOTO(s->avctx, s->lambda_table, mb_array_size * sizeof(int), fail)
735 736 737 738 739 740 741 742 743

        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 ||
744
        (s->avctx->flags & AV_CODEC_FLAG_INTERLACED_ME)) {
745 746 747 748 749 750 751 752 753 754 755 756
        /* 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;
                }
757 758 759
                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;
760
            }
761
            FF_ALLOCZ_OR_GOTO(s->avctx, s->p_field_select_table[i], mb_array_size * 2 * sizeof(uint8_t), fail)
762 763 764 765
        }
    }
    if (s->out_format == FMT_H263) {
        /* cbp values */
766
        FF_ALLOCZ_OR_GOTO(s->avctx, s->coded_block_base, y_size + (s->mb_height&1)*2*s->b8_stride, fail);
767 768 769
        s->coded_block = s->coded_block_base + s->b8_stride + 1;

        /* cbp, ac_pred, pred_dir */
770 771
        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);
772 773 774 775
    }

    if (s->h263_pred || s->h263_plus || !s->encoding) {
        /* dc values */
Lou Logan's avatar
Lou Logan committed
776
        // MN: we need these for error resilience of intra-frames
777
        FF_ALLOCZ_OR_GOTO(s->avctx, s->dc_val_base, yc_size * sizeof(int16_t), fail);
778 779 780 781 782 783 784
        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
785
    /* which mb is an intra block */
786 787 788 789 790
    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);
791
    // Note the + 1 is for a quicker MPEG-4 slice_end detection
792

793
    return ff_mpeg_er_init(s);
794 795 796 797
fail:
    return AVERROR(ENOMEM);
}

798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825
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;
    memset(s->pblocks, 0, sizeof(s->pblocks));
    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;

826

827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872
    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;
}

873
/**
Michael Niedermayer's avatar
Michael Niedermayer committed
874 875 876
 * init common structure for both encoder and decoder.
 * this assumes that some variables like width/height are already set
 */
877
av_cold int ff_mpv_common_init(MpegEncContext *s)
878
{
879
    int i;
880 881 882 883
    int nb_slices = (HAVE_THREADS &&
                     s->avctx->active_thread_type & FF_THREAD_SLICE) ?
                    s->avctx->thread_count : 1;

884 885
    clear_context(s);

886 887
    if (s->encoding && s->avctx->slices)
        nb_slices = s->avctx->slices;
888

889
    if (s->codec_id == AV_CODEC_ID_MPEG2VIDEO && !s->progressive_sequence)
890
        s->mb_height = (s->height + 31) / 32 * 2;
891
    else
Baptiste Coudurier's avatar
Baptiste Coudurier committed
892
        s->mb_height = (s->height + 15) / 16;
893

894
    if (s->avctx->pix_fmt == AV_PIX_FMT_NONE) {
895
        av_log(s->avctx, AV_LOG_ERROR,
896
               "decoding to AV_PIX_FMT_NONE is not supported.\n");
897 898 899
        return -1;
    }

900 901 902 903 904 905 906 907 908
    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
909 910
    }

911 912
    if ((s->width || s->height) &&
        av_image_check_size(s->width, s->height, 0, s->avctx))
913 914
        return -1;

915
    dct_init(s);
916

917
    /* set chroma shifts */
918 919 920
    avcodec_get_chroma_sub_sample(s->avctx->pix_fmt,
                                  &s->chroma_x_shift,
                                  &s->chroma_y_shift);
921

922

923
    FF_ALLOCZ_OR_GOTO(s->avctx, s->picture,
924 925
                      MAX_PICTURE_COUNT * sizeof(Picture), fail);
    for (i = 0; i < MAX_PICTURE_COUNT; i++) {
926 927 928
        s->picture[i].f = av_frame_alloc();
        if (!s->picture[i].f)
            goto fail;
929
    }
930 931 932 933 934 935 936 937 938 939 940 941
    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;
942

943
        if (init_context_frame(s))
944
            goto fail;
945

946
        s->parse_context.state = -1;
947

948
        s->context_initialized = 1;
949
        memset(s->thread_context, 0, sizeof(s->thread_context));
950
        s->thread_context[0]   = s;
951

952
//     if (s->width && s->height) {
953 954
        if (nb_slices > 1) {
            for (i = 0; i < nb_slices; i++) {
955
                if (i) {
956 957 958
                    s->thread_context[i] = av_memdup(s, sizeof(MpegEncContext));
                    if (!s->thread_context[i])
                        goto fail;
959
                }
960
                if (init_duplicate_context(s->thread_context[i]) < 0)
961 962
                    goto fail;
                    s->thread_context[i]->start_mb_y =
963
                        (s->mb_height * (i) + nb_slices / 2) / nb_slices;
964
                    s->thread_context[i]->end_mb_y   =
965
                        (s->mb_height * (i + 1) + nb_slices / 2) / nb_slices;
966 967
            }
        } else {
968
            if (init_duplicate_context(s) < 0)
969
                goto fail;
970 971
            s->start_mb_y = 0;
            s->end_mb_y   = s->mb_height;
972
        }
973
        s->slice_context_count = nb_slices;
974
//     }
975

Fabrice Bellard's avatar
Fabrice Bellard committed
976 977
    return 0;
 fail:
978
    ff_mpv_common_end(s);
Fabrice Bellard's avatar
Fabrice Bellard committed
979 980 981
    return -1;
}

982 983 984 985 986
/**
 * Frees and resets MpegEncContext fields depending on the resolution.
 * Is used during resolution changes to avoid a full reinitialization of the
 * codec.
 */
987
static void free_context_frame(MpegEncContext *s)
Fabrice Bellard's avatar
Fabrice Bellard committed
988
{
989
    int i, j, k;
Fabrice Bellard's avatar
Fabrice Bellard committed
990

991
    av_freep(&s->mb_type);
992 993 994 995 996 997
    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);
998 999 1000 1001 1002 1003 1004 1005 1006
    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++) {
1007
                av_freep(&s->b_field_mv_table_base[i][j][k]);
1008
                s->b_field_mv_table[i][j][k] = NULL;
1009 1010 1011
            }
            av_freep(&s->b_field_select_table[i][j]);
            av_freep(&s->p_field_mv_table_base[i][j]);
1012
            s->p_field_mv_table[i][j] = NULL;
1013 1014 1015
        }
        av_freep(&s->p_field_select_table[i]);
    }
1016

1017 1018
    av_freep(&s->dc_val_base);
    av_freep(&s->coded_block_base);
1019
    av_freep(&s->mbintra_table);
1020 1021
    av_freep(&s->cbp_table);
    av_freep(&s->pred_dir_table);
1022

1023
    av_freep(&s->mbskip_table);
1024

1025 1026
    av_freep(&s->er.error_status_table);
    av_freep(&s->er.er_temp_buffer);
1027
    av_freep(&s->mb_index2xy);
1028
    av_freep(&s->lambda_table);
1029

1030 1031 1032 1033 1034 1035
    av_freep(&s->cplx_tab);
    av_freep(&s->bits_tab);

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

1036
int ff_mpv_common_frame_size_change(MpegEncContext *s)
1037 1038 1039
{
    int i, err = 0;

1040 1041 1042
    if (!s->context_initialized)
        return AVERROR(EINVAL);

1043 1044 1045 1046 1047 1048 1049 1050 1051 1052
    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);

1053
    free_context_frame(s);
1054 1055

    if (s->picture)
1056
        for (i = 0; i < MAX_PICTURE_COUNT; i++) {
1057 1058 1059 1060 1061 1062 1063 1064 1065 1066
                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;
1067
    else
1068 1069 1070
        s->mb_height = (s->height + 15) / 16;

    if ((s->width || s->height) &&
1071
        (err = av_image_check_size(s->width, s->height, 0, s->avctx)) < 0)
1072
        goto fail;
1073 1074 1075 1076

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

1077
    memset(s->thread_context, 0, sizeof(s->thread_context));
1078 1079 1080 1081 1082 1083
    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++) {
1084
                if (i) {
1085 1086 1087 1088 1089
                    s->thread_context[i] = av_memdup(s, sizeof(MpegEncContext));
                    if (!s->thread_context[i]) {
                        err = AVERROR(ENOMEM);
                        goto fail;
                    }
1090
                }
1091
                if ((err = init_duplicate_context(s->thread_context[i])) < 0)
1092 1093 1094 1095 1096 1097 1098
                    goto fail;
                    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;
            }
        } else {
1099 1100
            err = init_duplicate_context(s);
            if (err < 0)
1101 1102 1103 1104 1105 1106 1107 1108 1109
                goto fail;
            s->start_mb_y = 0;
            s->end_mb_y   = s->mb_height;
        }
        s->slice_context_count = nb_slices;
    }

    return 0;
 fail:
1110
    ff_mpv_common_end(s);
1111 1112 1113
    return err;
}

1114
/* init common structure for both encoder and decoder */
1115
void ff_mpv_common_end(MpegEncContext *s)
1116 1117 1118
{
    int i;

1119 1120 1121
    if (!s)
        return ;

1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137
    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;

1138 1139
    if (s->picture) {
        for (i = 0; i < MAX_PICTURE_COUNT; i++) {
1140
            ff_free_picture_tables(&s->picture[i]);
1141
            ff_mpeg_unref_picture(s->avctx, &s->picture[i]);
1142
            av_frame_free(&s->picture[i].f);
Michael Niedermayer's avatar
Michael Niedermayer committed
1143
        }
Fabrice Bellard's avatar
Fabrice Bellard committed
1144
    }
1145
    av_freep(&s->picture);
1146
    ff_free_picture_tables(&s->last_picture);
1147
    ff_mpeg_unref_picture(s->avctx, &s->last_picture);
1148
    av_frame_free(&s->last_picture.f);
1149
    ff_free_picture_tables(&s->current_picture);
1150
    ff_mpeg_unref_picture(s->avctx, &s->current_picture);
1151
    av_frame_free(&s->current_picture.f);
1152
    ff_free_picture_tables(&s->next_picture);
1153
    ff_mpeg_unref_picture(s->avctx, &s->next_picture);
1154 1155
    av_frame_free(&s->next_picture.f);
    ff_free_picture_tables(&s->new_picture);
1156
    ff_mpeg_unref_picture(s->avctx, &s->new_picture);
1157
    av_frame_free(&s->new_picture.f);
1158 1159 1160

    free_context_frame(s);

1161 1162 1163 1164 1165
    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
1166 1167
}

1168

1169 1170 1171 1172 1173 1174 1175 1176
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);
1177
    for(i=0; i<AV_CEIL_RSHIFT(frame->height, v_chroma_shift); i++) {
1178
        memset(frame->data[1] + frame->linesize[1]*i,
1179
               0x80, AV_CEIL_RSHIFT(frame->width, h_chroma_shift));
1180
        memset(frame->data[2] + frame->linesize[2]*i,
1181
               0x80, AV_CEIL_RSHIFT(frame->width, h_chroma_shift));
1182 1183 1184
    }
}

1185
/**
1186 1187
 * generic function called after decoding
 * the header and before a frame is decoded.
1188
 */
1189
int ff_mpv_frame_start(MpegEncContext *s, AVCodecContext *avctx)
Fabrice Bellard's avatar
Fabrice Bellard committed
1190
{
1191
    int i, ret;
1192
    Picture *pic;
1193
    s->mb_skipped = 0;
1194

1195 1196 1197 1198 1199
    if (!ff_thread_can_start_frame(avctx)) {
        av_log(avctx, AV_LOG_ERROR, "Attempt to start a frame outside SETUP state\n");
        return -1;
    }

1200
    /* mark & release old frames */
Anton Khirnov's avatar
Anton Khirnov committed
1201 1202
    if (s->pict_type != AV_PICTURE_TYPE_B && s->last_picture_ptr &&
        s->last_picture_ptr != s->next_picture_ptr &&
1203
        s->last_picture_ptr->f->buf[0]) {
1204
        ff_mpeg_unref_picture(s->avctx, s->last_picture_ptr);
Anton Khirnov's avatar
Anton Khirnov committed
1205
    }
1206

Anton Khirnov's avatar
Anton Khirnov committed
1207
    /* release forgotten pictures */
1208
    /* if (MPEG-124 / H.263) */
1209 1210 1211 1212
    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) {
1213
            ff_mpeg_unref_picture(s->avctx, &s->picture[i]);
1214
        }
Anton Khirnov's avatar
Anton Khirnov committed
1215
    }
1216

1217
    ff_mpeg_unref_picture(s->avctx, &s->current_picture);
1218 1219
    ff_mpeg_unref_picture(s->avctx, &s->last_picture);
    ff_mpeg_unref_picture(s->avctx, &s->next_picture);
1220

1221 1222 1223 1224 1225
    /* 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]);
    }
1226

1227
    if (s->current_picture_ptr && !s->current_picture_ptr->f->buf[0]) {
Lou Logan's avatar
Lou Logan committed
1228
        // we already have an unused image
1229 1230 1231
        // (maybe it was set before reading the header)
        pic = s->current_picture_ptr;
    } else {
1232
        i   = ff_find_unused_picture(s->avctx, s->picture, 0);
1233 1234 1235
        if (i < 0) {
            av_log(s->avctx, AV_LOG_ERROR, "no frame buffer available\n");
            return i;
1236
        }
1237 1238
        pic = &s->picture[i];
    }
1239

1240 1241 1242 1243 1244
    pic->reference = 0;
    if (!s->droppable) {
        if (s->pict_type != AV_PICTURE_TYPE_B)
            pic->reference = 3;
    }
1245

1246
    pic->f->coded_picture_number = s->coded_picture_number++;
1247

1248
    if (alloc_picture(s, pic, 0) < 0)
1249
        return -1;
1250

1251 1252
    s->current_picture_ptr = pic;
    // FIXME use only the vars from current_pic
1253
    s->current_picture_ptr->f->top_field_first = s->top_field_first;
1254 1255 1256
    if (s->codec_id == AV_CODEC_ID_MPEG1VIDEO ||
        s->codec_id == AV_CODEC_ID_MPEG2VIDEO) {
        if (s->picture_structure != PICT_FRAME)
1257
            s->current_picture_ptr->f->top_field_first =
1258
                (s->picture_structure == PICT_TOP_FIELD) == s->first_field;
Michael Niedermayer's avatar
Michael Niedermayer committed
1259
    }
1260
    s->current_picture_ptr->f->interlaced_frame = !s->progressive_frame &&
1261 1262
                                                 !s->progressive_sequence;
    s->current_picture_ptr->field_picture      =  s->picture_structure != PICT_FRAME;
1263

1264
    s->current_picture_ptr->f->pict_type = s->pict_type;
1265
    // if (s->avctx->flags && AV_CODEC_FLAG_QSCALE)
1266
    //     s->current_picture_ptr->quality = s->new_picture_ptr->quality;
1267
    s->current_picture_ptr->f->key_frame = s->pict_type == AV_PICTURE_TYPE_I;
1268

1269
    if ((ret = ff_mpeg_ref_picture(s->avctx, &s->current_picture,
1270 1271
                                   s->current_picture_ptr)) < 0)
        return ret;
1272

1273
    if (s->pict_type != AV_PICTURE_TYPE_B) {
1274
        s->last_picture_ptr = s->next_picture_ptr;
1275
        if (!s->droppable)
1276
            s->next_picture_ptr = s->current_picture_ptr;
Fabrice Bellard's avatar
Fabrice Bellard committed
1277
    }
1278
    ff_dlog(s->avctx, "L%p N%p C%p L%p N%p C%p type:%d drop:%d\n",
1279
            s->last_picture_ptr, s->next_picture_ptr,s->current_picture_ptr,
1280 1281 1282
            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,
1283
            s->pict_type, s->droppable);
1284

1285
    if ((!s->last_picture_ptr || !s->last_picture_ptr->f->buf[0]) &&
Anton Khirnov's avatar
Anton Khirnov committed
1286 1287 1288 1289 1290
        (s->pict_type != AV_PICTURE_TYPE_I ||
         s->picture_structure != PICT_FRAME)) {
        int h_chroma_shift, v_chroma_shift;
        av_pix_fmt_get_chroma_sub_sample(s->avctx->pix_fmt,
                                         &h_chroma_shift, &v_chroma_shift);
1291
        if (s->pict_type == AV_PICTURE_TYPE_B && s->next_picture_ptr && s->next_picture_ptr->f->buf[0])
1292
            av_log(avctx, AV_LOG_DEBUG,
1293 1294
                   "allocating dummy last picture for B frame\n");
        else if (s->pict_type != AV_PICTURE_TYPE_I)
Anton Khirnov's avatar
Anton Khirnov committed
1295 1296 1297
            av_log(avctx, AV_LOG_ERROR,
                   "warning: first frame is no keyframe\n");
        else if (s->picture_structure != PICT_FRAME)
1298
            av_log(avctx, AV_LOG_DEBUG,
Anton Khirnov's avatar
Anton Khirnov committed
1299 1300 1301
                   "allocate dummy last picture for field based first keyframe\n");

        /* Allocate a dummy frame */
1302
        i = ff_find_unused_picture(s->avctx, s->picture, 0);
Anton Khirnov's avatar
Anton Khirnov committed
1303 1304 1305
        if (i < 0) {
            av_log(s->avctx, AV_LOG_ERROR, "no frame buffer available\n");
            return i;
1306
        }
Anton Khirnov's avatar
Anton Khirnov committed
1307
        s->last_picture_ptr = &s->picture[i];
1308 1309

        s->last_picture_ptr->reference   = 3;
1310 1311
        s->last_picture_ptr->f->key_frame = 0;
        s->last_picture_ptr->f->pict_type = AV_PICTURE_TYPE_P;
1312

1313
        if (alloc_picture(s, s->last_picture_ptr, 0) < 0) {
Anton Khirnov's avatar
Anton Khirnov committed
1314 1315
            s->last_picture_ptr = NULL;
            return -1;
1316
        }
1317

1318 1319 1320 1321 1322
        if (!avctx->hwaccel
#if FF_API_CAP_VDPAU
            && !(avctx->codec->capabilities&AV_CODEC_CAP_HWACCEL_VDPAU)
#endif
            ) {
1323
            for(i=0; i<avctx->height; i++)
1324
                memset(s->last_picture_ptr->f->data[0] + s->last_picture_ptr->f->linesize[0]*i,
1325
                       0x80, avctx->width);
1326
            if (s->last_picture_ptr->f->data[2]) {
1327
                for(i=0; i<AV_CEIL_RSHIFT(avctx->height, v_chroma_shift); i++) {
1328
                    memset(s->last_picture_ptr->f->data[1] + s->last_picture_ptr->f->linesize[1]*i,
1329
                        0x80, AV_CEIL_RSHIFT(avctx->width, h_chroma_shift));
1330
                    memset(s->last_picture_ptr->f->data[2] + s->last_picture_ptr->f->linesize[2]*i,
1331
                        0x80, AV_CEIL_RSHIFT(avctx->width, h_chroma_shift));
1332
                }
1333 1334 1335 1336
            }

            if(s->codec_id == AV_CODEC_ID_FLV1 || s->codec_id == AV_CODEC_ID_H263){
                for(i=0; i<avctx->height; i++)
1337
                memset(s->last_picture_ptr->f->data[0] + s->last_picture_ptr->f->linesize[0]*i, 16, avctx->width);
1338
            }
1339
        }
1340

Anton Khirnov's avatar
Anton Khirnov committed
1341 1342 1343
        ff_thread_report_progress(&s->last_picture_ptr->tf, INT_MAX, 0);
        ff_thread_report_progress(&s->last_picture_ptr->tf, INT_MAX, 1);
    }
1344
    if ((!s->next_picture_ptr || !s->next_picture_ptr->f->buf[0]) &&
Anton Khirnov's avatar
Anton Khirnov committed
1345 1346
        s->pict_type == AV_PICTURE_TYPE_B) {
        /* Allocate a dummy frame */
1347
        i = ff_find_unused_picture(s->avctx, s->picture, 0);
Anton Khirnov's avatar
Anton Khirnov committed
1348 1349 1350
        if (i < 0) {
            av_log(s->avctx, AV_LOG_ERROR, "no frame buffer available\n");
            return i;
1351
        }
Anton Khirnov's avatar
Anton Khirnov committed
1352
        s->next_picture_ptr = &s->picture[i];
1353 1354

        s->next_picture_ptr->reference   = 3;
1355 1356
        s->next_picture_ptr->f->key_frame = 0;
        s->next_picture_ptr->f->pict_type = AV_PICTURE_TYPE_P;
1357

1358
        if (alloc_picture(s, s->next_picture_ptr, 0) < 0) {
Anton Khirnov's avatar
Anton Khirnov committed
1359 1360
            s->next_picture_ptr = NULL;
            return -1;
1361
        }
Anton Khirnov's avatar
Anton Khirnov committed
1362 1363 1364
        ff_thread_report_progress(&s->next_picture_ptr->tf, INT_MAX, 0);
        ff_thread_report_progress(&s->next_picture_ptr->tf, INT_MAX, 1);
    }
1365

1366
#if 0 // BUFREF-FIXME
1367 1368
    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));
1369
#endif
Anton Khirnov's avatar
Anton Khirnov committed
1370
    if (s->last_picture_ptr) {
1371
        if (s->last_picture_ptr->f->buf[0] &&
1372
            (ret = ff_mpeg_ref_picture(s->avctx, &s->last_picture,
Anton Khirnov's avatar
Anton Khirnov committed
1373 1374 1375 1376
                                       s->last_picture_ptr)) < 0)
            return ret;
    }
    if (s->next_picture_ptr) {
1377
        if (s->next_picture_ptr->f->buf[0] &&
1378
            (ret = ff_mpeg_ref_picture(s->avctx, &s->next_picture,
Anton Khirnov's avatar
Anton Khirnov committed
1379 1380 1381
                                       s->next_picture_ptr)) < 0)
            return ret;
    }
1382

1383
    av_assert0(s->pict_type == AV_PICTURE_TYPE_I || (s->last_picture_ptr &&
1384
                                                 s->last_picture_ptr->f->buf[0]));
1385

1386
    if (s->picture_structure!= PICT_FRAME) {
1387
        int i;
1388 1389
        for (i = 0; i < 4; i++) {
            if (s->picture_structure == PICT_BOTTOM_FIELD) {
1390 1391
                s->current_picture.f->data[i] +=
                    s->current_picture.f->linesize[i];
1392
            }
1393 1394 1395
            s->current_picture.f->linesize[i] *= 2;
            s->last_picture.f->linesize[i]    *= 2;
            s->next_picture.f->linesize[i]    *= 2;
1396 1397
        }
    }
1398

1399
    /* set dequantizer, we can't do it during init as
1400 1401
     * 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 */
1402
    if (s->mpeg_quant || s->codec_id == AV_CODEC_ID_MPEG2VIDEO) {
1403 1404
        s->dct_unquantize_intra = s->dct_unquantize_mpeg2_intra;
        s->dct_unquantize_inter = s->dct_unquantize_mpeg2_inter;
1405
    } else if (s->out_format == FMT_H263 || s->out_format == FMT_H261) {
1406 1407
        s->dct_unquantize_intra = s->dct_unquantize_h263_intra;
        s->dct_unquantize_inter = s->dct_unquantize_h263_inter;
1408
    } else {
1409 1410 1411
        s->dct_unquantize_intra = s->dct_unquantize_mpeg1_intra;
        s->dct_unquantize_inter = s->dct_unquantize_mpeg1_inter;
    }
1412

1413 1414 1415 1416
    if (s->avctx->debug & FF_DEBUG_NOMC) {
        gray_frame(s->current_picture_ptr->f);
    }

1417
    return 0;
Fabrice Bellard's avatar
Fabrice Bellard committed
1418
}
1419

1420
/* called after a frame has been decoded. */
1421
void ff_mpv_frame_end(MpegEncContext *s)
Fabrice Bellard's avatar
Fabrice Bellard committed
1422
{
1423
    emms_c();
1424

1425
    if (s->current_picture.reference)
1426
        ff_thread_report_progress(&s->current_picture_ptr->tf, INT_MAX, 0);
Fabrice Bellard's avatar
Fabrice Bellard committed
1427 1428
}

1429

1430
#if FF_API_VISMV
1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452
static int clip_line(int *sx, int *sy, int *ex, int *ey, int maxx)
{
    if(*sx > *ex)
        return clip_line(ex, ey, sx, sy, maxx);

    if (*sx < 0) {
        if (*ex < 0)
            return 1;
        *sy = *ey + (*sy - *ey) * (int64_t)*ex / (*ex - *sx);
        *sx = 0;
    }

    if (*ex > maxx) {
        if (*sx > maxx)
            return 1;
        *ey = *sy + (*ey - *sy) * (int64_t)(maxx - *sx) / (*ex - *sx);
        *ex = maxx;
    }
    return 0;
}


1453
/**
1454
 * Draw a line from (ex, ey) -> (sx, sy).
1455 1456 1457 1458 1459
 * @param w width of the image
 * @param h height of the image
 * @param stride stride/linesize of the image
 * @param color color of the arrow
 */
1460 1461 1462
static void draw_line(uint8_t *buf, int sx, int sy, int ex, int ey,
                      int w, int h, int stride, int color)
{
Michael Niedermayer's avatar
Michael Niedermayer committed
1463
    int x, y, fr, f;
1464

1465 1466 1467 1468 1469
    if (clip_line(&sx, &sy, &ex, &ey, w - 1))
        return;
    if (clip_line(&sy, &sx, &ey, &ex, h - 1))
        return;

1470 1471 1472 1473
    sx = av_clip(sx, 0, w - 1);
    sy = av_clip(sy, 0, h - 1);
    ex = av_clip(ex, 0, w - 1);
    ey = av_clip(ey, 0, h - 1);
1474

1475
    buf[sy * stride + sx] += color;
1476

1477 1478
    if (FFABS(ex - sx) > FFABS(ey - sy)) {
        if (sx > ex) {
1479 1480
            FFSWAP(int, sx, ex);
            FFSWAP(int, sy, ey);
1481
        }
1482 1483 1484
        buf += sx + sy * stride;
        ex  -= sx;
        f    = ((ey - sy) << 16) / ex;
1485
        for (x = 0; x <= ex; x++) {
1486 1487 1488
            y  = (x * f) >> 16;
            fr = (x * f) & 0xFFFF;
            buf[y * stride + x]       += (color * (0x10000 - fr)) >> 16;
1489
            if(fr) buf[(y + 1) * stride + x] += (color *            fr ) >> 16;
1490
        }
1491 1492
    } else {
        if (sy > ey) {
1493 1494
            FFSWAP(int, sx, ex);
            FFSWAP(int, sy, ey);
1495
        }
1496 1497 1498
        buf += sx + sy * stride;
        ey  -= sy;
        if (ey)
1499
            f = ((ex - sx) << 16) / ey;
1500 1501
        else
            f = 0;
1502
        for(y= 0; y <= ey; y++){
1503 1504
            x  = (y*f) >> 16;
            fr = (y*f) & 0xFFFF;
1505
            buf[y * stride + x]     += (color * (0x10000 - fr)) >> 16;
1506
            if(fr) buf[y * stride + x + 1] += (color *            fr ) >> 16;
1507 1508 1509 1510 1511
        }
    }
}

/**
1512
 * Draw an arrow from (ex, ey) -> (sx, sy).
1513 1514 1515 1516 1517
 * @param w width of the image
 * @param h height of the image
 * @param stride stride/linesize of the image
 * @param color color of the arrow
 */
1518
static void draw_arrow(uint8_t *buf, int sx, int sy, int ex,
1519
                       int ey, int w, int h, int stride, int color, int tail, int direction)
1520
{
1521 1522
    int dx,dy;

1523 1524 1525 1526 1527
    if (direction) {
        FFSWAP(int, sx, ex);
        FFSWAP(int, sy, ey);
    }

1528 1529 1530 1531
    sx = av_clip(sx, -100, w + 100);
    sy = av_clip(sy, -100, h + 100);
    ex = av_clip(ex, -100, w + 100);
    ey = av_clip(ey, -100, h + 100);
1532

1533 1534
    dx = ex - sx;
    dy = ey - sy;
1535

1536 1537 1538 1539
    if (dx * dx + dy * dy > 3 * 3) {
        int rx =  dx + dy;
        int ry = -dx + dy;
        int length = ff_sqrt((rx * rx + ry * ry) << 8);
1540

1541 1542 1543
        // FIXME subpixel accuracy
        rx = ROUNDED_DIV(rx * 3 << 4, length);
        ry = ROUNDED_DIV(ry * 3 << 4, length);
1544

1545 1546 1547 1548 1549
        if (tail) {
            rx = -rx;
            ry = -ry;
        }

1550 1551 1552 1553 1554
        draw_line(buf, sx, sy, sx + rx, sy + ry, w, h, stride, color);
        draw_line(buf, sx, sy, sx - ry, sy + rx, w, h, stride, color);
    }
    draw_line(buf, sx, sy, ex, ey, w, h, stride, color);
}
1555
#endif
1556

1557 1558
static int add_mb(AVMotionVector *mb, uint32_t mb_type,
                  int dst_x, int dst_y,
1559
                  int motion_x, int motion_y, int motion_scale,
1560 1561 1562 1563
                  int direction)
{
    mb->w = IS_8X8(mb_type) || IS_8X16(mb_type) ? 8 : 16;
    mb->h = IS_8X8(mb_type) || IS_16X8(mb_type) ? 8 : 16;
1564 1565 1566
    mb->motion_x = motion_x;
    mb->motion_y = motion_y;
    mb->motion_scale = motion_scale;
1567 1568
    mb->dst_x = dst_x;
    mb->dst_y = dst_y;
1569 1570
    mb->src_x = dst_x + motion_x / motion_scale;
    mb->src_y = dst_y + motion_y / motion_scale;
1571 1572 1573 1574 1575
    mb->source = direction ? 1 : -1;
    mb->flags = 0; // XXX: does mb_type contain extra information that could be exported here?
    return 1;
}

1576
/**
1577
 * Print debugging info for the given picture.
1578
 */
1579 1580
void ff_print_debug_info2(AVCodecContext *avctx, AVFrame *pict, uint8_t *mbskip_table,
                         uint32_t *mbtype_table, int8_t *qscale_table, int16_t (*motion_val[2])[2],
1581
                         int *low_delay,
1582
                         int mb_width, int mb_height, int mb_stride, int quarter_sample)
1583
{
1584
    if ((avctx->flags2 & AV_CODEC_FLAG2_EXPORT_MVS) && mbtype_table && motion_val[0]) {
1585
        const int shift = 1 + quarter_sample;
1586
        const int scale = 1 << shift;
1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609
        const int mv_sample_log2 = avctx->codec_id == AV_CODEC_ID_H264 || avctx->codec_id == AV_CODEC_ID_SVQ3 ? 2 : 1;
        const int mv_stride      = (mb_width << mv_sample_log2) +
                                   (avctx->codec->id == AV_CODEC_ID_H264 ? 0 : 1);
        int mb_x, mb_y, mbcount = 0;

        /* size is width * height * 2 * 4 where 2 is for directions and 4 is
         * for the maximum number of MB (4 MB in case of IS_8x8) */
        AVMotionVector *mvs = av_malloc_array(mb_width * mb_height, 2 * 4 * sizeof(AVMotionVector));
        if (!mvs)
            return;

        for (mb_y = 0; mb_y < mb_height; mb_y++) {
            for (mb_x = 0; mb_x < mb_width; mb_x++) {
                int i, direction, mb_type = mbtype_table[mb_x + mb_y * mb_stride];
                for (direction = 0; direction < 2; direction++) {
                    if (!USES_LIST(mb_type, direction))
                        continue;
                    if (IS_8X8(mb_type)) {
                        for (i = 0; i < 4; i++) {
                            int sx = mb_x * 16 + 4 + 8 * (i & 1);
                            int sy = mb_y * 16 + 4 + 8 * (i >> 1);
                            int xy = (mb_x * 2 + (i & 1) +
                                      (mb_y * 2 + (i >> 1)) * mv_stride) << (mv_sample_log2 - 1);
1610 1611 1612
                            int mx = motion_val[direction][xy][0];
                            int my = motion_val[direction][xy][1];
                            mbcount += add_mb(mvs + mbcount, mb_type, sx, sy, mx, my, scale, direction);
1613 1614 1615 1616 1617 1618
                        }
                    } else if (IS_16X8(mb_type)) {
                        for (i = 0; i < 2; i++) {
                            int sx = mb_x * 16 + 8;
                            int sy = mb_y * 16 + 4 + 8 * i;
                            int xy = (mb_x * 2 + (mb_y * 2 + i) * mv_stride) << (mv_sample_log2 - 1);
1619 1620
                            int mx = motion_val[direction][xy][0];
                            int my = motion_val[direction][xy][1];
1621 1622 1623 1624

                            if (IS_INTERLACED(mb_type))
                                my *= 2;

1625
                            mbcount += add_mb(mvs + mbcount, mb_type, sx, sy, mx, my, scale, direction);
1626 1627 1628 1629 1630 1631
                        }
                    } else if (IS_8X16(mb_type)) {
                        for (i = 0; i < 2; i++) {
                            int sx = mb_x * 16 + 4 + 8 * i;
                            int sy = mb_y * 16 + 8;
                            int xy = (mb_x * 2 + i + mb_y * 2 * mv_stride) << (mv_sample_log2 - 1);
1632 1633
                            int mx = motion_val[direction][xy][0];
                            int my = motion_val[direction][xy][1];
1634 1635 1636 1637

                            if (IS_INTERLACED(mb_type))
                                my *= 2;

1638
                            mbcount += add_mb(mvs + mbcount, mb_type, sx, sy, mx, my, scale, direction);
1639 1640 1641 1642 1643
                        }
                    } else {
                          int sx = mb_x * 16 + 8;
                          int sy = mb_y * 16 + 8;
                          int xy = (mb_x + mb_y * mv_stride) << mv_sample_log2;
1644 1645 1646
                          int mx = motion_val[direction][xy][0];
                          int my = motion_val[direction][xy][1];
                          mbcount += add_mb(mvs + mbcount, mb_type, sx, sy, mx, my, scale, direction);
1647 1648 1649 1650 1651 1652 1653 1654 1655 1656
                    }
                }
            }
        }

        if (mbcount) {
            AVFrameSideData *sd;

            av_log(avctx, AV_LOG_DEBUG, "Adding %d MVs info to frame %d\n", mbcount, avctx->frame_number);
            sd = av_frame_new_side_data(pict, AV_FRAME_DATA_MOTION_VECTORS, mbcount * sizeof(AVMotionVector));
1657 1658
            if (!sd) {
                av_freep(&mvs);
1659
                return;
1660
            }
1661 1662 1663 1664 1665 1666 1667
            memcpy(sd->data, mvs, mbcount * sizeof(AVMotionVector));
        }

        av_freep(&mvs);
    }

    /* TODO: export all the following to make them accessible for users (and filters) */
1668
    if (avctx->hwaccel || !mbtype_table
1669 1670 1671 1672
#if FF_API_CAP_VDPAU
        || (avctx->codec->capabilities&AV_CODEC_CAP_HWACCEL_VDPAU)
#endif
        )
1673
        return;
1674

1675

1676
    if (avctx->debug & (FF_DEBUG_SKIP | FF_DEBUG_QP | FF_DEBUG_MB_TYPE)) {
1677
        int x,y;
1678

1679
        av_log(avctx, AV_LOG_DEBUG, "New frame, type: %c\n",
1680
               av_get_picture_type_char(pict->pict_type));
1681 1682 1683
        for (y = 0; y < mb_height; y++) {
            for (x = 0; x < mb_width; x++) {
                if (avctx->debug & FF_DEBUG_SKIP) {
1684
                    int count = mbskip_table ? mbskip_table[x + y * mb_stride] : 0;
1685 1686
                    if (count > 9)
                        count = 9;
1687
                    av_log(avctx, AV_LOG_DEBUG, "%1d", count);
1688
                }
1689 1690
                if (avctx->debug & FF_DEBUG_QP) {
                    av_log(avctx, AV_LOG_DEBUG, "%2d",
1691
                           qscale_table[x + y * mb_stride]);
1692
                }
1693
                if (avctx->debug & FF_DEBUG_MB_TYPE) {
1694
                    int mb_type = mbtype_table[x + y * mb_stride];
1695 1696
                    // Type & MV direction
                    if (IS_PCM(mb_type))
1697
                        av_log(avctx, AV_LOG_DEBUG, "P");
1698
                    else if (IS_INTRA(mb_type) && IS_ACPRED(mb_type))
1699
                        av_log(avctx, AV_LOG_DEBUG, "A");
1700
                    else if (IS_INTRA4x4(mb_type))
1701
                        av_log(avctx, AV_LOG_DEBUG, "i");
1702
                    else if (IS_INTRA16x16(mb_type))
1703
                        av_log(avctx, AV_LOG_DEBUG, "I");
1704
                    else if (IS_DIRECT(mb_type) && IS_SKIP(mb_type))
1705
                        av_log(avctx, AV_LOG_DEBUG, "d");
1706
                    else if (IS_DIRECT(mb_type))
1707
                        av_log(avctx, AV_LOG_DEBUG, "D");
1708
                    else if (IS_GMC(mb_type) && IS_SKIP(mb_type))
1709
                        av_log(avctx, AV_LOG_DEBUG, "g");
1710
                    else if (IS_GMC(mb_type))
1711
                        av_log(avctx, AV_LOG_DEBUG, "G");
1712
                    else if (IS_SKIP(mb_type))
1713
                        av_log(avctx, AV_LOG_DEBUG, "S");
1714
                    else if (!USES_LIST(mb_type, 1))
1715
                        av_log(avctx, AV_LOG_DEBUG, ">");
1716
                    else if (!USES_LIST(mb_type, 0))
1717
                        av_log(avctx, AV_LOG_DEBUG, "<");
1718
                    else {
1719
                        av_assert2(USES_LIST(mb_type, 0) && USES_LIST(mb_type, 1));
1720
                        av_log(avctx, AV_LOG_DEBUG, "X");
1721
                    }
1722

1723 1724
                    // segmentation
                    if (IS_8X8(mb_type))
1725
                        av_log(avctx, AV_LOG_DEBUG, "+");
1726
                    else if (IS_16X8(mb_type))
1727
                        av_log(avctx, AV_LOG_DEBUG, "-");
1728
                    else if (IS_8X16(mb_type))
1729
                        av_log(avctx, AV_LOG_DEBUG, "|");
1730
                    else if (IS_INTRA(mb_type) || IS_16X16(mb_type))
1731
                        av_log(avctx, AV_LOG_DEBUG, " ");
1732
                    else
1733
                        av_log(avctx, AV_LOG_DEBUG, "?");
1734 1735


1736
                    if (IS_INTERLACED(mb_type))
1737
                        av_log(avctx, AV_LOG_DEBUG, "=");
1738
                    else
1739
                        av_log(avctx, AV_LOG_DEBUG, " ");
1740 1741
                }
            }
1742
            av_log(avctx, AV_LOG_DEBUG, "\n");
1743 1744
        }
    }
1745

1746 1747
    if ((avctx->debug & (FF_DEBUG_VIS_QP | FF_DEBUG_VIS_MB_TYPE)) ||
        (avctx->debug_mv)) {
1748
        int mb_y;
1749
        int i;
1750
        int h_chroma_shift, v_chroma_shift, block_height;
1751 1752 1753
#if FF_API_VISMV
        const int shift = 1 + quarter_sample;
        uint8_t *ptr;
1754 1755
        const int width          = avctx->width;
        const int height         = avctx->height;
1756
#endif
1757
        const int mv_sample_log2 = avctx->codec_id == AV_CODEC_ID_H264 || avctx->codec_id == AV_CODEC_ID_SVQ3 ? 2 : 1;
1758 1759
        const int mv_stride      = (mb_width << mv_sample_log2) +
                                   (avctx->codec->id == AV_CODEC_ID_H264 ? 0 : 1);
1760

1761 1762
        if (low_delay)
            *low_delay = 0; // needed to see the vectors without trashing the buffers
1763

1764
        avcodec_get_chroma_sub_sample(avctx->pix_fmt, &h_chroma_shift, &v_chroma_shift);
1765

1766 1767
        av_frame_make_writable(pict);

1768
        pict->opaque = NULL;
1769
#if FF_API_VISMV
1770
        ptr          = pict->data[0];
1771
#endif
1772
        block_height = 16 >> v_chroma_shift;
1773

1774
        for (mb_y = 0; mb_y < mb_height; mb_y++) {
1775
            int mb_x;
1776 1777
            for (mb_x = 0; mb_x < mb_width; mb_x++) {
                const int mb_index = mb_x + mb_y * mb_stride;
1778
#if FF_API_VISMV
1779
                if ((avctx->debug_mv) && motion_val[0]) {
1780 1781 1782 1783 1784
                    int type;
                    for (type = 0; type < 3; type++) {
                        int direction = 0;
                        switch (type) {
                        case 0:
1785
                            if ((!(avctx->debug_mv & FF_DEBUG_VIS_MV_P_FOR)) ||
1786
                                (pict->pict_type!= AV_PICTURE_TYPE_P))
1787
                                continue;
1788 1789 1790
                            direction = 0;
                            break;
                        case 1:
1791
                            if ((!(avctx->debug_mv & FF_DEBUG_VIS_MV_B_FOR)) ||
1792
                                (pict->pict_type!= AV_PICTURE_TYPE_B))
1793
                                continue;
1794 1795 1796
                            direction = 0;
                            break;
                        case 2:
1797
                            if ((!(avctx->debug_mv & FF_DEBUG_VIS_MV_B_BACK)) ||
1798
                                (pict->pict_type!= AV_PICTURE_TYPE_B))
1799
                                continue;
1800 1801 1802
                            direction = 1;
                            break;
                        }
1803
                        if (!USES_LIST(mbtype_table[mb_index], direction))
1804 1805
                            continue;

1806
                        if (IS_8X8(mbtype_table[mb_index])) {
1807 1808 1809 1810 1811 1812
                            int i;
                            for (i = 0; i < 4; i++) {
                                int sx = mb_x * 16 + 4 + 8 * (i & 1);
                                int sy = mb_y * 16 + 4 + 8 * (i >> 1);
                                int xy = (mb_x * 2 + (i & 1) +
                                          (mb_y * 2 + (i >> 1)) * mv_stride) << (mv_sample_log2 - 1);
1813 1814
                                int mx = (motion_val[direction][xy][0] >> shift) + sx;
                                int my = (motion_val[direction][xy][1] >> shift) + sy;
1815
                                draw_arrow(ptr, sx, sy, mx, my, width,
1816
                                           height, pict->linesize[0], 100, 0, direction);
1817
                            }
1818
                        } else if (IS_16X8(mbtype_table[mb_index])) {
1819 1820 1821 1822 1823
                            int i;
                            for (i = 0; i < 2; i++) {
                                int sx = mb_x * 16 + 8;
                                int sy = mb_y * 16 + 4 + 8 * i;
                                int xy = (mb_x * 2 + (mb_y * 2 + i) * mv_stride) << (mv_sample_log2 - 1);
1824 1825
                                int mx = (motion_val[direction][xy][0] >> shift);
                                int my = (motion_val[direction][xy][1] >> shift);
1826

1827
                                if (IS_INTERLACED(mbtype_table[mb_index]))
1828 1829
                                    my *= 2;

1830
                                draw_arrow(ptr, sx, sy, mx + sx, my + sy, width,
1831
                                           height, pict->linesize[0], 100, 0, direction);
1832
                            }
1833
                        } else if (IS_8X16(mbtype_table[mb_index])) {
1834 1835 1836 1837 1838
                            int i;
                            for (i = 0; i < 2; i++) {
                                int sx = mb_x * 16 + 4 + 8 * i;
                                int sy = mb_y * 16 + 8;
                                int xy = (mb_x * 2 + i + mb_y * 2 * mv_stride) << (mv_sample_log2 - 1);
1839 1840
                                int mx = motion_val[direction][xy][0] >> shift;
                                int my = motion_val[direction][xy][1] >> shift;
1841

1842
                                if (IS_INTERLACED(mbtype_table[mb_index]))
1843 1844 1845
                                    my *= 2;

                                draw_arrow(ptr, sx, sy, mx + sx, my + sy, width,
1846
                                           height, pict->linesize[0], 100, 0, direction);
1847 1848
                            }
                        } else {
1849 1850 1851
                              int sx= mb_x * 16 + 8;
                              int sy= mb_y * 16 + 8;
                              int xy= (mb_x + mb_y * mv_stride) << mv_sample_log2;
1852 1853
                              int mx= (motion_val[direction][xy][0]>>shift) + sx;
                              int my= (motion_val[direction][xy][1]>>shift) + sy;
1854
                              draw_arrow(ptr, sx, sy, mx, my, width, height, pict->linesize[0], 100, 0, direction);
1855
                        }
1856
                    }
1857
                }
1858
#endif
1859
                if ((avctx->debug & FF_DEBUG_VIS_QP)) {
1860
                    uint64_t c = (qscale_table[mb_index] * 128 / 31) *
1861
                                 0x0101010101010101ULL;
1862
                    int y;
1863 1864 1865 1866 1867 1868 1869
                    for (y = 0; y < block_height; y++) {
                        *(uint64_t *)(pict->data[1] + 8 * mb_x +
                                      (block_height * mb_y + y) *
                                      pict->linesize[1]) = c;
                        *(uint64_t *)(pict->data[2] + 8 * mb_x +
                                      (block_height * mb_y + y) *
                                      pict->linesize[2]) = c;
1870 1871
                    }
                }
1872
                if ((avctx->debug & FF_DEBUG_VIS_MB_TYPE) &&
1873 1874
                    motion_val[0]) {
                    int mb_type = mbtype_table[mb_index];
1875 1876
                    uint64_t u,v;
                    int y;
1877
#define COLOR(theta, r) \
1878 1879
    u = (int)(128 + r * cos(theta * M_PI / 180)); \
    v = (int)(128 + r * sin(theta * M_PI / 180));
1880 1881 1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904


                    u = v = 128;
                    if (IS_PCM(mb_type)) {
                        COLOR(120, 48)
                    } else if ((IS_INTRA(mb_type) && IS_ACPRED(mb_type)) ||
                               IS_INTRA16x16(mb_type)) {
                        COLOR(30, 48)
                    } else if (IS_INTRA4x4(mb_type)) {
                        COLOR(90, 48)
                    } else if (IS_DIRECT(mb_type) && IS_SKIP(mb_type)) {
                        // COLOR(120, 48)
                    } else if (IS_DIRECT(mb_type)) {
                        COLOR(150, 48)
                    } else if (IS_GMC(mb_type) && IS_SKIP(mb_type)) {
                        COLOR(170, 48)
                    } else if (IS_GMC(mb_type)) {
                        COLOR(190, 48)
                    } else if (IS_SKIP(mb_type)) {
                        // COLOR(180, 48)
                    } else if (!USES_LIST(mb_type, 1)) {
                        COLOR(240, 48)
                    } else if (!USES_LIST(mb_type, 0)) {
                        COLOR(0, 48)
                    } else {
1905
                        av_assert2(USES_LIST(mb_type, 0) && USES_LIST(mb_type, 1));
1906 1907 1908
                        COLOR(300,48)
                    }

1909 1910 1911 1912 1913 1914 1915
                    u *= 0x0101010101010101ULL;
                    v *= 0x0101010101010101ULL;
                    for (y = 0; y < block_height; y++) {
                        *(uint64_t *)(pict->data[1] + 8 * mb_x +
                                      (block_height * mb_y + y) * pict->linesize[1]) = u;
                        *(uint64_t *)(pict->data[2] + 8 * mb_x +
                                      (block_height * mb_y + y) * pict->linesize[2]) = v;
1916 1917
                    }

1918 1919 1920 1921 1922 1923
                    // segmentation
                    if (IS_8X8(mb_type) || IS_16X8(mb_type)) {
                        *(uint64_t *)(pict->data[0] + 16 * mb_x + 0 +
                                      (16 * mb_y + 8) * pict->linesize[0]) ^= 0x8080808080808080ULL;
                        *(uint64_t *)(pict->data[0] + 16 * mb_x + 8 +
                                      (16 * mb_y + 8) * pict->linesize[0]) ^= 0x8080808080808080ULL;
1924
                    }
1925 1926 1927 1928
                    if (IS_8X8(mb_type) || IS_8X16(mb_type)) {
                        for (y = 0; y < 16; y++)
                            pict->data[0][16 * mb_x + 8 + (16 * mb_y + y) *
                                          pict->linesize[0]] ^= 0x80;
1929
                    }
1930 1931 1932 1933 1934 1935 1936 1937
                    if (IS_8X8(mb_type) && mv_sample_log2 >= 2) {
                        int dm = 1 << (mv_sample_log2 - 2);
                        for (i = 0; i < 4; i++) {
                            int sx = mb_x * 16 + 8 * (i & 1);
                            int sy = mb_y * 16 + 8 * (i >> 1);
                            int xy = (mb_x * 2 + (i & 1) +
                                     (mb_y * 2 + (i >> 1)) * mv_stride) << (mv_sample_log2 - 1);
                            // FIXME bidir
1938
                            int32_t *mv = (int32_t *) &motion_val[0][xy];
1939 1940 1941 1942 1943 1944 1945
                            if (mv[0] != mv[dm] ||
                                mv[dm * mv_stride] != mv[dm * (mv_stride + 1)])
                                for (y = 0; y < 8; y++)
                                    pict->data[0][sx + 4 + (sy + y) * pict->linesize[0]] ^= 0x80;
                            if (mv[0] != mv[dm * mv_stride] || mv[dm] != mv[dm * (mv_stride + 1)])
                                *(uint64_t *)(pict->data[0] + sx + (sy + 4) *
                                              pict->linesize[0]) ^= 0x8080808080808080ULL;
1946 1947
                        }
                    }
1948

1949
                    if (IS_INTERLACED(mb_type) &&
1950
                        avctx->codec->id == AV_CODEC_ID_H264) {
1951 1952
                        // hmm
                    }
1953
                }
1954 1955
                if (mbskip_table)
                    mbskip_table[mb_index] = 0;
1956 1957 1958
            }
        }
    }
1959 1960
}

1961
void ff_print_debug_info(MpegEncContext *s, Picture *p, AVFrame *pict)
1962
{
1963 1964
    ff_print_debug_info2(s->avctx, pict, s->mbskip_table, p->mb_type,
                         p->qscale_table, p->motion_val, &s->low_delay,
1965
                         s->mb_width, s->mb_height, s->mb_stride, s->quarter_sample);
1966 1967
}

1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979
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);
}

1980 1981 1982 1983
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,
1984
                                     int width, int height, ptrdiff_t stride,
1985 1986 1987 1988 1989
                                     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;
1990
    const int op_index = FFMIN(lowres, 3);
1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008
    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)) {
2009
        s->vdsp.emulated_edge_mc(s->sc.edge_emu_buffer, src,
2010 2011 2012 2013
                                 s->linesize, s->linesize,
                                 w + 1, (h + 1) << field_based,
                                 src_x, src_y   << field_based,
                                 h_edge_pos, v_edge_pos);
2014
        src = s->sc.edge_emu_buffer;
2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039
        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;
2040 2041
    int mx, my, src_x, src_y, uvsrc_x, uvsrc_y, sx, sy, uvsx, uvsy;
    ptrdiff_t uvlinesize, linesize;
2042
    const int lowres     = s->avctx->lowres;
2043
    const int op_index   = FFMIN(lowres-1+s->chroma_x_shift, 3);
2044 2045 2046 2047
    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;
2048 2049
    linesize   = s->current_picture.f->linesize[0] << field_based;
    uvlinesize = s->current_picture.f->linesize[1] << field_based;
2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075 2076 2077 2078 2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106 2107 2108

    // 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;

2109
    if ((unsigned) src_x > FFMAX( h_edge_pos - (!!sx) - 2 * block_s,       0) || uvsrc_y<0 ||
2110
        (unsigned) src_y > FFMAX((v_edge_pos >> field_based) - (!!sy) - h, 0)) {
2111
        s->vdsp.emulated_edge_mc(s->sc.edge_emu_buffer, ptr_y,
2112 2113
                                 linesize >> field_based, linesize >> field_based,
                                 17, 17 + field_based,
2114 2115
                                src_x, src_y << field_based, h_edge_pos,
                                v_edge_pos);
2116
        ptr_y = s->sc.edge_emu_buffer;
2117
        if (!CONFIG_GRAY || !(s->avctx->flags & AV_CODEC_FLAG_GRAY)) {
2118
            uint8_t *ubuf = s->sc.edge_emu_buffer + 18 * s->linesize;
2119 2120
            uint8_t *vbuf =ubuf + 9 * s->uvlinesize;
            s->vdsp.emulated_edge_mc(ubuf,  ptr_cb,
2121 2122
                                     uvlinesize >> field_based, uvlinesize >> field_based,
                                     9, 9 + field_based,
2123 2124
                                    uvsrc_x, uvsrc_y << field_based,
                                    h_edge_pos >> 1, v_edge_pos >> 1);
2125
            s->vdsp.emulated_edge_mc(vbuf,  ptr_cr,
2126 2127
                                     uvlinesize >> field_based,uvlinesize >> field_based,
                                     9, 9 + field_based,
2128 2129
                                    uvsrc_x, uvsrc_y << field_based,
                                    h_edge_pos >> 1, v_edge_pos >> 1);
2130 2131
            ptr_cb = ubuf;
            ptr_cr = vbuf;
2132 2133 2134
        }
    }

2135
    // FIXME use this for field pix too instead of the obnoxious hack which changes picture.f->data
2136 2137 2138 2139 2140 2141 2142 2143 2144 2145 2146 2147 2148 2149 2150 2151
    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);

2152
    if (!CONFIG_GRAY || !(s->avctx->flags & AV_CODEC_FLAG_GRAY)) {
2153
        int hc = s->chroma_y_shift ? (h+1-bottom_field)>>1 : h;
2154 2155
        uvsx = (uvsx << 2) >> lowres;
        uvsy = (uvsy << 2) >> lowres;
2156 2157 2158
        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);
2159 2160 2161 2162 2163 2164 2165 2166 2167 2168 2169 2170
        }
    }
    // 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;
2171
    const int op_index   = FFMIN(lowres, 3);
2172 2173 2174 2175
    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;
2176 2177
    int emu = 0, src_x, src_y, sx, sy;
    ptrdiff_t offset;
2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189 2190 2191 2192 2193 2194 2195 2196
    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;
2197 2198
    if ((unsigned) src_x > FFMAX(h_edge_pos - (!!sx) - block_s, 0) ||
        (unsigned) src_y > FFMAX(v_edge_pos - (!!sy) - block_s, 0)) {
2199
        s->vdsp.emulated_edge_mc(s->sc.edge_emu_buffer, ptr,
2200 2201 2202
                                 s->uvlinesize, s->uvlinesize,
                                 9, 9,
                                 src_x, src_y, h_edge_pos, v_edge_pos);
2203
        ptr = s->sc.edge_emu_buffer;
2204
        emu = 1;
2205 2206 2207 2208 2209 2210 2211
    }
    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) {
2212
        s->vdsp.emulated_edge_mc(s->sc.edge_emu_buffer, ptr,
2213 2214 2215
                                 s->uvlinesize, s->uvlinesize,
                                 9, 9,
                                 src_x, src_y, h_edge_pos, v_edge_pos);
2216
        ptr = s->sc.edge_emu_buffer;
2217 2218 2219 2220 2221 2222 2223 2224 2225 2226 2227 2228 2229 2230 2231 2232 2233 2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245 2246 2247 2248 2249 2250 2251 2252 2253 2254 2255 2256 2257 2258 2259 2260 2261 2262 2263 2264 2265 2266 2267 2268 2269 2270 2271
    }
    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];
        }

2272
        if (!CONFIG_GRAY || !(s->avctx->flags & AV_CODEC_FLAG_GRAY))
2273 2274 2275 2276 2277 2278 2279 2280 2281 2282 2283 2284 2285 2286 2287 2288 2289 2290 2291 2292
            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) {
2293
                ref_picture = s->current_picture_ptr->f->data;
2294 2295 2296 2297 2298 2299 2300 2301 2302 2303 2304 2305 2306 2307 2308 2309 2310

            }
            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 {
2311
                ref2picture = s->current_picture_ptr->f->data;
2312 2313 2314 2315 2316 2317 2318 2319 2320 2321 2322 2323 2324 2325 2326 2327 2328 2329 2330 2331 2332 2333 2334 2335 2336
            }

            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);
                }
2337
                pix_op = s->h264chroma.avg_h264_chroma_pixels_tab;
2338 2339 2340 2341 2342 2343 2344 2345 2346 2347
            }
        } 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
2348
                pix_op = s->h264chroma.avg_h264_chroma_pixels_tab;
2349 2350 2351 2352

                // opposite parity is always in the same
                // frame if this is second field
                if (!s->first_field) {
2353
                    ref_picture = s->current_picture_ptr->f->data;
2354 2355 2356 2357 2358
                }
            }
        }
        break;
    default:
2359
        av_assert2(0);
2360 2361 2362
    }
}

2363 2364 2365
/**
 * find the lowest MB row referenced in the MVs
 */
2366
static int lowest_referenced_row(MpegEncContext *s, int dir)
2367 2368 2369 2370
{
    int my_max = INT_MIN, my_min = INT_MAX, qpel_shift = !s->quarter_sample;
    int my, off, i, mvs;

2371 2372
    if (s->picture_structure != PICT_FRAME || s->mcsel)
        goto unhandled;
2373 2374 2375 2376 2377 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388

    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++) {
2389
        my = s->mv[dir][i][1];
2390 2391 2392 2393
        my_max = FFMAX(my_max, my);
        my_min = FFMIN(my_min, my);
    }

2394
    off = ((FFMAX(-my_min, my_max)<<qpel_shift) + 63) >> 6;
2395

2396
    return av_clip(s->mb_y + off, 0, s->mb_height - 1);
2397 2398 2399 2400
unhandled:
    return s->mb_height-1;
}

2401 2402
/* put block[] to dest[] */
static inline void put_dct(MpegEncContext *s,
Diego Biurrun's avatar
Diego Biurrun committed
2403
                           int16_t *block, int i, uint8_t *dest, int line_size, int qscale)
2404
{
2405
    s->dct_unquantize_intra(s, block, i, qscale);
2406
    s->idsp.idct_put(dest, line_size, block);
2407
}
2408

2409 2410
/* add block[] to dest[] */
static inline void add_dct(MpegEncContext *s,
Diego Biurrun's avatar
Diego Biurrun committed
2411
                           int16_t *block, int i, uint8_t *dest, int line_size)
2412 2413
{
    if (s->block_last_index[i] >= 0) {
2414
        s->idsp.idct_add(dest, line_size, block);
2415 2416
    }
}
2417

2418
static inline void add_dequant_dct(MpegEncContext *s,
Diego Biurrun's avatar
Diego Biurrun committed
2419
                           int16_t *block, int i, uint8_t *dest, int line_size, int qscale)
2420
{
Fabrice Bellard's avatar
Fabrice Bellard committed
2421
    if (s->block_last_index[i] >= 0) {
2422
        s->dct_unquantize_inter(s, block, i, qscale);
2423

2424
        s->idsp.idct_add(dest, line_size, block);
Fabrice Bellard's avatar
Fabrice Bellard committed
2425 2426 2427
    }
}

2428
/**
2429
 * Clean dc, ac, coded_block for the current non-intra MB.
2430 2431 2432
 */
void ff_clean_intra_table_entries(MpegEncContext *s)
{
2433
    int wrap = s->b8_stride;
2434
    int xy = s->block_index[0];
2435 2436 2437

    s->dc_val[0][xy           ] =
    s->dc_val[0][xy + 1       ] =
2438 2439 2440
    s->dc_val[0][xy     + wrap] =
    s->dc_val[0][xy + 1 + wrap] = 1024;
    /* ac pred */
2441 2442
    memset(s->ac_val[0][xy       ], 0, 32 * sizeof(int16_t));
    memset(s->ac_val[0][xy + wrap], 0, 32 * sizeof(int16_t));
2443 2444 2445 2446 2447 2448 2449
    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 */
2450 2451
    wrap = s->mb_stride;
    xy = s->mb_x + s->mb_y * wrap;
2452 2453 2454
    s->dc_val[1][xy] =
    s->dc_val[2][xy] = 1024;
    /* ac pred */
2455 2456
    memset(s->ac_val[1][xy], 0, 16 * sizeof(int16_t));
    memset(s->ac_val[2][xy], 0, 16 * sizeof(int16_t));
2457

2458
    s->mbintra_table[xy]= 0;
2459 2460
}

Fabrice Bellard's avatar
Fabrice Bellard committed
2461 2462 2463 2464 2465 2466 2467 2468 2469 2470
/* 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)
 */
2471
static av_always_inline
2472
void mpv_decode_mb_internal(MpegEncContext *s, int16_t block[12][64],
2473
                            int lowres_flag, int is_mpeg12)
Fabrice Bellard's avatar
Fabrice Bellard committed
2474
{
2475
    const int mb_xy = s->mb_y * s->mb_stride + s->mb_x;
2476

2477 2478 2479
    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
2480 2481
        return;
    }
Fabrice Bellard's avatar
Fabrice Bellard committed
2482

2483
    if(s->avctx->debug&FF_DEBUG_DCT_COEFF) {
2484
       /* print DCT coefficients */
2485
       int i,j;
2486 2487 2488
       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++){
2489 2490
               av_log(s->avctx, AV_LOG_DEBUG, "%5d",
                      block[i][s->idsp.idct_permutation[j]]);
2491 2492 2493
           }
           av_log(s->avctx, AV_LOG_DEBUG, "\n");
       }
2494 2495
    }

2496
    s->current_picture.qscale_table[mb_xy] = s->qscale;
2497

Fabrice Bellard's avatar
Fabrice Bellard committed
2498 2499
    /* update DC predictors for P macroblocks */
    if (!s->mb_intra) {
Keiji Costantini's avatar
Keiji Costantini committed
2500
        if (!is_mpeg12 && (s->h263_pred || s->h263_aic)) {
2501
            if(s->mbintra_table[mb_xy])
2502
                ff_clean_intra_table_entries(s);
Fabrice Bellard's avatar
Fabrice Bellard committed
2503
        } else {
2504 2505
            s->last_dc[0] =
            s->last_dc[1] =
Fabrice Bellard's avatar
Fabrice Bellard committed
2506 2507 2508
            s->last_dc[2] = 128 << s->intra_dc_precision;
        }
    }
Keiji Costantini's avatar
Keiji Costantini committed
2509
    else if (!is_mpeg12 && (s->h263_pred || s->h263_aic))
2510
        s->mbintra_table[mb_xy]=1;
2511

2512
    if ((s->avctx->flags & AV_CODEC_FLAG_PSNR) || s->frame_skip_threshold || s->frame_skip_factor ||
2513 2514
        !(s->encoding && (s->intra_only || s->pict_type == AV_PICTURE_TYPE_B) &&
          s->avctx->mb_decision != FF_MB_DECISION_RD)) { // FIXME precalc
2515
        uint8_t *dest_y, *dest_cb, *dest_cr;
2516
        int dct_linesize, dct_offset;
Michael Niedermayer's avatar
Michael Niedermayer committed
2517 2518
        op_pixels_func (*op_pix)[4];
        qpel_mc_func (*op_qpix)[16];
2519 2520
        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];
2521 2522
        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;
2523

Michael Niedermayer's avatar
Michael Niedermayer committed
2524 2525 2526
        /* 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){
2527
            uint8_t *mbskip_ptr = &s->mbskip_table[mb_xy];
Michael Niedermayer's avatar
Michael Niedermayer committed
2528

2529 2530
            if (s->mb_skipped) {
                s->mb_skipped= 0;
2531
                av_assert2(s->pict_type!=AV_PICTURE_TYPE_I);
2532
                *mbskip_ptr = 1;
2533
            } else if(!s->current_picture.reference) {
2534
                *mbskip_ptr = 1;
2535
            } else{
2536 2537
                *mbskip_ptr = 0; /* not skipped */
            }
Michael Niedermayer's avatar
Michael Niedermayer committed
2538
        }
2539

2540
        dct_linesize = linesize << s->interlaced_dct;
2541
        dct_offset   = s->interlaced_dct ? linesize : linesize * block_size;
2542

2543 2544 2545 2546 2547
        if(readable){
            dest_y=  s->dest[0];
            dest_cb= s->dest[1];
            dest_cr= s->dest[2];
        }else{
2548 2549 2550
            dest_y = s->sc.b_scratchpad;
            dest_cb= s->sc.b_scratchpad+16*linesize;
            dest_cr= s->sc.b_scratchpad+32*linesize;
2551
        }
2552

Fabrice Bellard's avatar
Fabrice Bellard committed
2553 2554
        if (!s->mb_intra) {
            /* motion handling */
2555
            /* decoding or more than one mb_type (MC was already done otherwise) */
2556
            if(!s->encoding){
2557

2558
                if(HAVE_THREADS && s->avctx->active_thread_type&FF_THREAD_FRAME) {
2559
                    if (s->mv_dir & MV_DIR_FORWARD) {
2560
                        ff_thread_await_progress(&s->last_picture_ptr->tf,
2561
                                                 lowest_referenced_row(s, 0),
2562
                                                 0);
2563 2564
                    }
                    if (s->mv_dir & MV_DIR_BACKWARD) {
2565
                        ff_thread_await_progress(&s->next_picture_ptr->tf,
2566
                                                 lowest_referenced_row(s, 1),
2567
                                                 0);
2568 2569 2570
                    }
                }

2571
                if(lowres_flag){
2572
                    h264_chroma_mc_func *op_pix = s->h264chroma.put_h264_chroma_pixels_tab;
2573 2574

                    if (s->mv_dir & MV_DIR_FORWARD) {
2575
                        MPV_motion_lowres(s, dest_y, dest_cb, dest_cr, 0, s->last_picture.f->data, op_pix);
2576
                        op_pix = s->h264chroma.avg_h264_chroma_pixels_tab;
2577 2578
                    }
                    if (s->mv_dir & MV_DIR_BACKWARD) {
2579
                        MPV_motion_lowres(s, dest_y, dest_cb, dest_cr, 1, s->next_picture.f->data, op_pix);
2580
                    }
2581
                }else{
2582
                    op_qpix = s->me.qpel_put;
2583
                    if ((!s->no_rounding) || s->pict_type==AV_PICTURE_TYPE_B){
2584
                        op_pix = s->hdsp.put_pixels_tab;
2585
                    }else{
2586
                        op_pix = s->hdsp.put_no_rnd_pixels_tab;
2587 2588
                    }
                    if (s->mv_dir & MV_DIR_FORWARD) {
2589
                        ff_mpv_motion(s, dest_y, dest_cb, dest_cr, 0, s->last_picture.f->data, op_pix, op_qpix);
2590
                        op_pix = s->hdsp.avg_pixels_tab;
2591 2592 2593
                        op_qpix= s->me.qpel_avg;
                    }
                    if (s->mv_dir & MV_DIR_BACKWARD) {
2594
                        ff_mpv_motion(s, dest_y, dest_cb, dest_cr, 1, s->next_picture.f->data, op_pix, op_qpix);
2595
                    }
2596
                }
Fabrice Bellard's avatar
Fabrice Bellard committed
2597 2598
            }

2599
            /* skip dequant / idct if we are really late ;) */
Michael Niedermayer's avatar
Michael Niedermayer committed
2600
            if(s->avctx->skip_idct){
2601 2602
                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
2603 2604 2605
                   || s->avctx->skip_idct >= AVDISCARD_ALL)
                    goto skip_idct;
            }
2606

Fabrice Bellard's avatar
Fabrice Bellard committed
2607
            /* add dct residue */
2608 2609
            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))){
2610
                add_dequant_dct(s, block[0], 0, dest_y                          , dct_linesize, s->qscale);
2611
                add_dequant_dct(s, block[1], 1, dest_y              + block_size, dct_linesize, s->qscale);
2612
                add_dequant_dct(s, block[2], 2, dest_y + dct_offset             , dct_linesize, s->qscale);
2613
                add_dequant_dct(s, block[3], 3, dest_y + dct_offset + block_size, dct_linesize, s->qscale);
2614

2615
                if (!CONFIG_GRAY || !(s->avctx->flags & AV_CODEC_FLAG_GRAY)) {
2616 2617 2618 2619 2620 2621 2622 2623 2624 2625 2626
                    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);
                    }
2627
                }
2628
            } else if(is_mpeg12 || (s->codec_id != AV_CODEC_ID_WMV2)){
2629
                add_dct(s, block[0], 0, dest_y                          , dct_linesize);
2630
                add_dct(s, block[1], 1, dest_y              + block_size, dct_linesize);
2631
                add_dct(s, block[2], 2, dest_y + dct_offset             , dct_linesize);
2632
                add_dct(s, block[3], 3, dest_y + dct_offset + block_size, dct_linesize);
Fabrice Bellard's avatar
Fabrice Bellard committed
2633

2634
                if (!CONFIG_GRAY || !(s->avctx->flags & AV_CODEC_FLAG_GRAY)) {
2635 2636 2637 2638 2639 2640
                    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;
2641
                        dct_offset   = s->interlaced_dct ? uvlinesize : uvlinesize*block_size;
2642 2643 2644 2645 2646 2647

                        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
2648 2649 2650 2651
                            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);
2652 2653 2654 2655
                        }
                    }
                }//fi gray
            }
2656
            else if (CONFIG_WMV2_DECODER || CONFIG_WMV2_ENCODER) {
Michael Niedermayer's avatar
Michael Niedermayer committed
2657
                ff_wmv2_add_mb(s, block, dest_y, dest_cb, dest_cr);
2658
            }
Fabrice Bellard's avatar
Fabrice Bellard committed
2659 2660
        } else {
            /* dct only in intra block */
2661
            if(s->encoding || !(s->codec_id==AV_CODEC_ID_MPEG1VIDEO || s->codec_id==AV_CODEC_ID_MPEG2VIDEO)){
2662 2663 2664 2665
                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);
2666

2667
                if (!CONFIG_GRAY || !(s->avctx->flags & AV_CODEC_FLAG_GRAY)) {
2668 2669 2670 2671 2672 2673 2674 2675 2676
                    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);
2677
                        put_dct(s, block[7], 7, dest_cr + dct_offset, dct_linesize, s->chroma_qscale);
2678 2679 2680
                    }
                }
            }else{
2681 2682 2683 2684
                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]);
2685

2686
                if (!CONFIG_GRAY || !(s->avctx->flags & AV_CODEC_FLAG_GRAY)) {
2687
                    if(s->chroma_y_shift){
2688 2689
                        s->idsp.idct_put(dest_cb, uvlinesize, block[4]);
                        s->idsp.idct_put(dest_cr, uvlinesize, block[5]);
2690
                    }else{
2691

2692
                        dct_linesize = uvlinesize << s->interlaced_dct;
2693
                        dct_offset   = s->interlaced_dct ? uvlinesize : uvlinesize*block_size;
2694

2695 2696 2697 2698
                        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]);
2699
                        if(!s->chroma_x_shift){//Chroma444
2700 2701 2702 2703
                            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]);
2704 2705 2706
                        }
                    }
                }//gray
2707
            }
2708
        }
2709 2710
skip_idct:
        if(!readable){
2711
            s->hdsp.put_pixels_tab[0][0](s->dest[0], dest_y ,   linesize,16);
2712
            if (!CONFIG_GRAY || !(s->avctx->flags & AV_CODEC_FLAG_GRAY)) {
2713 2714
                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);
2715
            }
2716
        }
2717 2718 2719
    }
}

2720 2721
void ff_mpv_decode_mb(MpegEncContext *s, int16_t block[12][64])
{
2722
#if !CONFIG_SMALL
Keiji Costantini's avatar
Keiji Costantini committed
2723
    if(s->out_format == FMT_MPEG1) {
2724 2725
        if(s->avctx->lowres) mpv_decode_mb_internal(s, block, 1, 1);
        else                 mpv_decode_mb_internal(s, block, 0, 1);
Keiji Costantini's avatar
Keiji Costantini committed
2726 2727
    } else
#endif
2728 2729
    if(s->avctx->lowres) mpv_decode_mb_internal(s, block, 1, 0);
    else                  mpv_decode_mb_internal(s, block, 0, 0);
2730 2731
}

2732 2733
void ff_mpeg_draw_horiz_band(MpegEncContext *s, int y, int h)
{
2734 2735
    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,
2736
                       s->first_field, s->low_delay);
2737 2738
}

2739
void ff_init_block_index(MpegEncContext *s){ //FIXME maybe rename
2740 2741
    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];
Michael Niedermayer's avatar
Michael Niedermayer committed
2742
    const int mb_size= 4 - s->avctx->lowres;
2743

2744 2745 2746 2747 2748 2749 2750
    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
2751

2752 2753 2754
    s->dest[0] = s->current_picture.f->data[0] + (int)((s->mb_x - 1U) <<  mb_size);
    s->dest[1] = s->current_picture.f->data[1] + (int)((s->mb_x - 1U) << (mb_size - s->chroma_x_shift));
    s->dest[2] = s->current_picture.f->data[2] + (int)((s->mb_x - 1U) << (mb_size - s->chroma_x_shift));
2755

2756
    if(!(s->pict_type==AV_PICTURE_TYPE_B && s->avctx->draw_horiz_band && s->picture_structure==PICT_FRAME))
2757
    {
2758
        if(s->picture_structure==PICT_FRAME){
2759 2760 2761
        s->dest[0] += s->mb_y *   linesize << mb_size;
        s->dest[1] += s->mb_y * uvlinesize << (mb_size - s->chroma_y_shift);
        s->dest[2] += s->mb_y * uvlinesize << (mb_size - s->chroma_y_shift);
2762 2763 2764 2765
        }else{
            s->dest[0] += (s->mb_y>>1) *   linesize << mb_size;
            s->dest[1] += (s->mb_y>>1) * uvlinesize << (mb_size - s->chroma_y_shift);
            s->dest[2] += (s->mb_y>>1) * uvlinesize << (mb_size - s->chroma_y_shift);
2766
            av_assert1((s->mb_y&1) == (s->picture_structure == PICT_BOTTOM_FIELD));
2767
        }
2768 2769
    }
}
2770

2771 2772 2773
void ff_mpeg_flush(AVCodecContext *avctx){
    int i;
    MpegEncContext *s = avctx->priv_data;
2774

2775
    if (!s || !s->picture)
2776
        return;
2777

2778
    for (i = 0; i < MAX_PICTURE_COUNT; i++)
2779
        ff_mpeg_unref_picture(s->avctx, &s->picture[i]);
2780
    s->current_picture_ptr = s->last_picture_ptr = s->next_picture_ptr = NULL;
2781

2782 2783 2784
    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);
2785

2786
    s->mb_x= s->mb_y= 0;
2787
    s->closed_gop= 0;
2788

2789 2790 2791 2792 2793 2794 2795 2796
    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
2797 2798
}

2799 2800 2801 2802 2803 2804 2805 2806 2807 2808 2809 2810 2811 2812 2813 2814
/**
 * 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 ];
}
2815

2816
void ff_mpv_report_decode_progress(MpegEncContext *s)
2817
{
2818
    if (s->pict_type != AV_PICTURE_TYPE_B && !s->partitioned_frame && !s->er.error_occurred)
2819
        ff_thread_report_progress(&s->current_picture_ptr->tf, s->mb_y, 0);
2820
}