proresdec_lgpl.c 25.1 KB
Newer Older
1 2 3 4 5
/*
 * Apple ProRes compatible decoder
 *
 * Copyright (c) 2010-2011 Maxim Poliakovski
 *
6
 * This file is part of FFmpeg.
7
 *
8
 * FFmpeg is free software; you can redistribute it and/or
9 10 11 12
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
 * version 2.1 of the License, or (at your option) any later version.
 *
13
 * FFmpeg is distributed in the hope that it will be useful,
14 15 16 17 18
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * Lesser General Public License for more details.
 *
 * You should have received a copy of the GNU Lesser General Public
19
 * License along with FFmpeg; if not, write to the Free Software
20 21 22 23 24 25 26 27 28 29 30
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
 */

/**
 * @file
 * This is a decoder for Apple ProRes 422 SD/HQ/LT/Proxy and ProRes 4444.
 * It is used for storing and editing high definition video data in Apple's Final Cut Pro.
 *
 * @see http://wiki.multimedia.cx/index.php?title=Apple_ProRes
 */

31
#define LONG_BITSTREAM_READER // some ProRes vlc codes require up to 28 bits to be read at once
32 33 34 35 36

#include <stdint.h>

#include "libavutil/intmath.h"
#include "avcodec.h"
37
#include "idctdsp.h"
38
#include "internal.h"
39
#include "proresdata.h"
40
#include "proresdsp.h"
41 42
#include "get_bits.h"

43
typedef struct ProresThreadData {
44 45 46 47
    const uint8_t *index;            ///< pointers to the data of this slice
    int slice_num;
    int x_pos, y_pos;
    int slice_width;
48
    int prev_slice_sf;               ///< scalefactor of the previous decoded slice
Diego Biurrun's avatar
Diego Biurrun committed
49
    DECLARE_ALIGNED(16, int16_t, blocks)[8 * 4 * 64];
50 51
    DECLARE_ALIGNED(16, int16_t, qmat_luma_scaled)[64];
    DECLARE_ALIGNED(16, int16_t, qmat_chroma_scaled)[64];
52
} ProresThreadData;
53

54
typedef struct ProresContext {
55
    ProresDSPContext dsp;
56
    AVFrame    *frame;
57 58 59 60 61 62 63 64 65
    ScanTable  scantable;
    int        scantable_type;           ///< -1 = uninitialized, 0 = progressive, 1/2 = interlaced

    int        frame_type;               ///< 0 = progressive, 1 = top-field first, 2 = bottom-field first
    int        pic_format;               ///< 2 = 422, 3 = 444
    uint8_t    qmat_luma[64];            ///< dequantization matrix for luma
    uint8_t    qmat_chroma[64];          ///< dequantization matrix for chroma
    int        qmat_changed;             ///< 1 - global quantization matrices changed
    int        total_slices;            ///< total number of slices in a picture
66 67
    ProresThreadData *slice_data;
    int        pic_num;
68 69 70 71 72 73 74 75 76
    int        chroma_factor;
    int        mb_chroma_factor;
    int        num_chroma_blocks;       ///< number of chrominance blocks in a macroblock
    int        num_x_slices;
    int        num_y_slices;
    int        slice_width_factor;
    int        slice_height_factor;
    int        num_x_mbs;
    int        num_y_mbs;
77
    int        alpha_info;
78 79 80 81 82 83 84 85
} ProresContext;


static av_cold int decode_init(AVCodecContext *avctx)
{
    ProresContext *ctx = avctx->priv_data;

    ctx->total_slices     = 0;
86
    ctx->slice_data       = NULL;
87

88
    avctx->bits_per_raw_sample = PRORES_BITS_PER_SAMPLE;
89
    ff_proresdsp_init(&ctx->dsp, avctx);
90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106

    ctx->scantable_type = -1;   // set scantable type to uninitialized
    memset(ctx->qmat_luma, 4, 64);
    memset(ctx->qmat_chroma, 4, 64);

    return 0;
}


static int decode_frame_header(ProresContext *ctx, const uint8_t *buf,
                               const int data_size, AVCodecContext *avctx)
{
    int hdr_size, version, width, height, flags;
    const uint8_t *ptr;

    hdr_size = AV_RB16(buf);
    if (hdr_size > data_size) {
107
        av_log(avctx, AV_LOG_ERROR, "frame data too small\n");
108
        return AVERROR_INVALIDDATA;
109 110 111 112 113 114
    }

    version = AV_RB16(buf + 2);
    if (version >= 2) {
        av_log(avctx, AV_LOG_ERROR,
               "unsupported header version: %d\n", version);
115
        return AVERROR_INVALIDDATA;
116 117 118 119 120 121
    }

    width  = AV_RB16(buf + 8);
    height = AV_RB16(buf + 10);
    if (width != avctx->width || height != avctx->height) {
        av_log(avctx, AV_LOG_ERROR,
122
               "picture dimension changed: old: %d x %d, new: %d x %d\n",
123
               avctx->width, avctx->height, width, height);
124
        return AVERROR_INVALIDDATA;
125 126 127 128 129
    }

    ctx->frame_type = (buf[12] >> 2) & 3;
    if (ctx->frame_type > 2) {
        av_log(avctx, AV_LOG_ERROR,
130
               "unsupported frame type: %d\n", ctx->frame_type);
131
        return AVERROR_INVALIDDATA;
132 133 134 135 136
    }

    ctx->chroma_factor     = (buf[12] >> 6) & 3;
    ctx->mb_chroma_factor  = ctx->chroma_factor + 2;
    ctx->num_chroma_blocks = (1 << ctx->chroma_factor) >> 1;
137 138 139 140 141 142
    ctx->alpha_info        = buf[17] & 0xf;

    if (ctx->alpha_info > 2) {
        av_log(avctx, AV_LOG_ERROR, "Invalid alpha mode %d\n", ctx->alpha_info);
        return AVERROR_INVALIDDATA;
    }
143
    if (avctx->skip_alpha) ctx->alpha_info = 0;
144

145 146
    switch (ctx->chroma_factor) {
    case 2:
147 148
        avctx->pix_fmt = ctx->alpha_info ? AV_PIX_FMT_YUVA422P10
                                         : AV_PIX_FMT_YUV422P10;
149 150
        break;
    case 3:
151 152
        avctx->pix_fmt = ctx->alpha_info ? AV_PIX_FMT_YUVA444P10
                                         : AV_PIX_FMT_YUV444P10;
153 154 155
        break;
    default:
        av_log(avctx, AV_LOG_ERROR,
156
               "unsupported picture format: %d\n", ctx->pic_format);
157
        return AVERROR_INVALIDDATA;
158 159 160 161 162
    }

    if (ctx->scantable_type != ctx->frame_type) {
        if (!ctx->frame_type)
            ff_init_scantable(ctx->dsp.idct_permutation, &ctx->scantable,
163
                              ff_prores_progressive_scan);
164 165
        else
            ff_init_scantable(ctx->dsp.idct_permutation, &ctx->scantable,
166
                              ff_prores_interlaced_scan);
167 168 169 170
        ctx->scantable_type = ctx->frame_type;
    }

    if (ctx->frame_type) {      /* if interlaced */
171 172
        ctx->frame->interlaced_frame = 1;
        ctx->frame->top_field_first  = ctx->frame_type & 1;
173
    } else {
174
        ctx->frame->interlaced_frame = 0;
175 176
    }

177 178 179 180
    avctx->color_primaries = buf[14];
    avctx->color_trc       = buf[15];
    avctx->colorspace      = buf[16];

181 182 183 184 185
    ctx->qmat_changed = 0;
    ptr   = buf + 20;
    flags = buf[19];
    if (flags & 2) {
        if (ptr - buf > hdr_size - 64) {
186
            av_log(avctx, AV_LOG_ERROR, "header data too small\n");
187
            return AVERROR_INVALIDDATA;
188 189 190 191 192 193 194 195 196 197 198 199 200
        }
        if (memcmp(ctx->qmat_luma, ptr, 64)) {
            memcpy(ctx->qmat_luma, ptr, 64);
            ctx->qmat_changed = 1;
        }
        ptr += 64;
    } else {
        memset(ctx->qmat_luma, 4, 64);
        ctx->qmat_changed = 1;
    }

    if (flags & 1) {
        if (ptr - buf > hdr_size - 64) {
201
            av_log(avctx, AV_LOG_ERROR, "header data too small\n");
202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226
            return -1;
        }
        if (memcmp(ctx->qmat_chroma, ptr, 64)) {
            memcpy(ctx->qmat_chroma, ptr, 64);
            ctx->qmat_changed = 1;
        }
    } else {
        memset(ctx->qmat_chroma, 4, 64);
        ctx->qmat_changed = 1;
    }

    return hdr_size;
}


static int decode_picture_header(ProresContext *ctx, const uint8_t *buf,
                                 const int data_size, AVCodecContext *avctx)
{
    int   i, hdr_size, pic_data_size, num_slices;
    int   slice_width_factor, slice_height_factor;
    int   remainder, num_x_slices;
    const uint8_t *data_ptr, *index_ptr;

    hdr_size = data_size > 0 ? buf[0] >> 3 : 0;
    if (hdr_size < 8 || hdr_size > data_size) {
227
        av_log(avctx, AV_LOG_ERROR, "picture header too small\n");
228
        return AVERROR_INVALIDDATA;
229 230 231 232
    }

    pic_data_size = AV_RB32(buf + 1);
    if (pic_data_size > data_size) {
233
        av_log(avctx, AV_LOG_ERROR, "picture data too small\n");
234
        return AVERROR_INVALIDDATA;
235 236 237 238 239 240
    }

    slice_width_factor  = buf[7] >> 4;
    slice_height_factor = buf[7] & 0xF;
    if (slice_width_factor > 3 || slice_height_factor) {
        av_log(avctx, AV_LOG_ERROR,
241
               "unsupported slice dimension: %d x %d\n",
242
               1 << slice_width_factor, 1 << slice_height_factor);
243
        return AVERROR_INVALIDDATA;
244 245 246 247 248 249
    }

    ctx->slice_width_factor  = slice_width_factor;
    ctx->slice_height_factor = slice_height_factor;

    ctx->num_x_mbs = (avctx->width + 15) >> 4;
250
    ctx->num_y_mbs = (avctx->height +
251 252
                      (1 << (4 + ctx->frame->interlaced_frame)) - 1) >>
                     (4 + ctx->frame->interlaced_frame);
253 254 255 256 257 258 259

    remainder    = ctx->num_x_mbs & ((1 << slice_width_factor) - 1);
    num_x_slices = (ctx->num_x_mbs >> slice_width_factor) + (remainder & 1) +
                   ((remainder >> 1) & 1) + ((remainder >> 2) & 1);

    num_slices = num_x_slices * ctx->num_y_mbs;
    if (num_slices != AV_RB16(buf + 5)) {
260
        av_log(avctx, AV_LOG_ERROR, "invalid number of slices\n");
261
        return AVERROR_INVALIDDATA;
262 263 264
    }

    if (ctx->total_slices != num_slices) {
265
        av_freep(&ctx->slice_data);
266
        ctx->slice_data = av_malloc_array(num_slices + 1, sizeof(ctx->slice_data[0]));
267
        if (!ctx->slice_data)
268 269 270 271 272
            return AVERROR(ENOMEM);
        ctx->total_slices = num_slices;
    }

    if (hdr_size + num_slices * 2 > data_size) {
273
        av_log(avctx, AV_LOG_ERROR, "slice table too small\n");
274
        return AVERROR_INVALIDDATA;
275 276 277 278 279 280 281
    }

    /* parse slice table allowing quick access to the slice data */
    index_ptr = buf + hdr_size;
    data_ptr = index_ptr + num_slices * 2;

    for (i = 0; i < num_slices; i++) {
282
        ctx->slice_data[i].index = data_ptr;
283
        ctx->slice_data[i].prev_slice_sf = 0;
284 285
        data_ptr += AV_RB16(index_ptr + i * 2);
    }
286
    ctx->slice_data[i].index = data_ptr;
287
    ctx->slice_data[i].prev_slice_sf = 0;
288 289

    if (data_ptr > buf + data_size) {
290
        av_log(avctx, AV_LOG_ERROR, "out of slice data\n");
291 292 293 294 295 296 297 298 299 300
        return -1;
    }

    return pic_data_size;
}


/**
 * Read an unsigned rice/exp golomb codeword.
 */
301
static inline int decode_vlc_codeword(GetBitContext *gb, unsigned codebook)
302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324
{
    unsigned int rice_order, exp_order, switch_bits;
    unsigned int buf, code;
    int log, prefix_len, len;

    OPEN_READER(re, gb);
    UPDATE_CACHE(re, gb);
    buf = GET_CACHE(re, gb);

    /* number of prefix bits to switch between Rice and expGolomb */
    switch_bits = (codebook & 3) + 1;
    rice_order  = codebook >> 5;        /* rice code order */
    exp_order   = (codebook >> 2) & 7;  /* exp golomb code order */

    log = 31 - av_log2(buf); /* count prefix bits (zeroes) */

    if (log < switch_bits) { /* ok, we got a rice code */
        if (!rice_order) {
            /* shortcut for faster decoding of rice codes without remainder */
            code = log;
            LAST_SKIP_BITS(re, gb, log + 1);
        } else {
            prefix_len = log + 1;
325
            code = (log << rice_order) + NEG_USR32(buf << prefix_len, rice_order);
326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344
            LAST_SKIP_BITS(re, gb, prefix_len + rice_order);
        }
    } else { /* otherwise we got a exp golomb code */
        len  = (log << 1) - switch_bits + exp_order + 1;
        code = NEG_USR32(buf, len) - (1 << exp_order) + (switch_bits << rice_order);
        LAST_SKIP_BITS(re, gb, len);
    }

    CLOSE_READER(re, gb);

    return code;
}

#define LSB2SIGN(x) (-((x) & 1))
#define TOSIGNED(x) (((x) >> 1) ^ LSB2SIGN(x))

/**
 * Decode DC coefficients for all blocks in a slice.
 */
Diego Biurrun's avatar
Diego Biurrun committed
345
static inline void decode_dc_coeffs(GetBitContext *gb, int16_t *out,
346 347
                                    int nblocks)
{
Diego Biurrun's avatar
Diego Biurrun committed
348
    int16_t prev_dc;
349 350 351 352 353 354 355 356 357 358 359
    int     i, sign;
    int16_t delta;
    unsigned int code;

    code   = decode_vlc_codeword(gb, FIRST_DC_CB);
    out[0] = prev_dc = TOSIGNED(code);

    out   += 64; /* move to the DC coeff of the next block */
    delta  = 3;

    for (i = 1; i < nblocks; i++, out += 64) {
360
        code = decode_vlc_codeword(gb, ff_prores_dc_codebook[FFMIN(FFABS(delta), 3)]);
361 362 363 364 365 366 367 368

        sign     = -(((delta >> 15) & 1) ^ (code & 1));
        delta    = (((code + 1) >> 1) ^ sign) - sign;
        prev_dc += delta;
        out[0]   = prev_dc;
    }
}

369
#define MAX_PADDING 16
370 371 372 373

/**
 * Decode AC coefficients for all blocks in a slice.
 */
374 375 376 377
static inline int decode_ac_coeffs(GetBitContext *gb, int16_t *out,
                                   int blocks_per_slice,
                                   int plane_size_factor,
                                   const uint8_t *scan)
378 379 380 381 382 383 384 385 386 387 388 389
{
    int pos, block_mask, run, level, sign, run_cb_index, lev_cb_index;
    int max_coeffs, bits_left;

    /* set initial prediction values */
    run   = 4;
    level = 2;

    max_coeffs = blocks_per_slice << 6;
    block_mask = blocks_per_slice - 1;

    for (pos = blocks_per_slice - 1; pos < max_coeffs;) {
390 391
        run_cb_index = ff_prores_run_to_cb_index[FFMIN(run, 15)];
        lev_cb_index = ff_prores_lev_to_cb_index[FFMIN(level, 9)];
392 393

        bits_left = get_bits_left(gb);
394
        if (bits_left <= 0 || (bits_left <= MAX_PADDING && !show_bits(gb, bits_left)))
395
            return 0;
396

397
        run = decode_vlc_codeword(gb, ff_prores_ac_codebook[run_cb_index]);
398 399
        if (run < 0)
            return AVERROR_INVALIDDATA;
400 401

        bits_left = get_bits_left(gb);
402
        if (bits_left <= 0 || (bits_left <= MAX_PADDING && !show_bits(gb, bits_left)))
403
            return AVERROR_INVALIDDATA;
404

405
        level = decode_vlc_codeword(gb, ff_prores_ac_codebook[lev_cb_index]) + 1;
406 407
        if (level < 0)
            return AVERROR_INVALIDDATA;
408 409 410 411 412 413 414 415 416

        pos += run + 1;
        if (pos >= max_coeffs)
            break;

        sign = get_sbits(gb, 1);
        out[((pos & block_mask) << 6) + scan[pos >> plane_size_factor]] =
            (level ^ sign) - sign;
    }
417 418

    return 0;
419 420 421 422 423 424
}


/**
 * Decode a slice plane (luma or chroma).
 */
425 426 427 428 429 430
static int decode_slice_plane(ProresContext *ctx, ProresThreadData *td,
                              const uint8_t *buf,
                              int data_size, uint16_t *out_ptr,
                              int linesize, int mbs_per_slice,
                              int blocks_per_mb, int plane_size_factor,
                              const int16_t *qmat, int is_chroma)
431 432
{
    GetBitContext gb;
Diego Biurrun's avatar
Diego Biurrun committed
433
    int16_t *block_ptr;
434
    int mb_num, blocks_per_slice, ret;
435 436 437

    blocks_per_slice = mbs_per_slice * blocks_per_mb;

438
    memset(td->blocks, 0, 8 * 4 * 64 * sizeof(*td->blocks));
439 440 441

    init_get_bits(&gb, buf, data_size << 3);

442
    decode_dc_coeffs(&gb, td->blocks, blocks_per_slice);
443

444 445 446 447
    ret = decode_ac_coeffs(&gb, td->blocks, blocks_per_slice,
                           plane_size_factor, ctx->scantable.permutated);
    if (ret < 0)
        return ret;
448 449

    /* inverse quantization, inverse transform and output */
450
    block_ptr = td->blocks;
Maxim Poliakovski's avatar
Maxim Poliakovski committed
451

452 453 454
    if (!is_chroma) {
        for (mb_num = 0; mb_num < mbs_per_slice; mb_num++, out_ptr += blocks_per_mb * 4) {
            ctx->dsp.idct_put(out_ptr,                    linesize, block_ptr, qmat);
455
            block_ptr += 64;
456 457 458 459 460
            if (blocks_per_mb > 2) {
                ctx->dsp.idct_put(out_ptr + 8,            linesize, block_ptr, qmat);
                block_ptr += 64;
            }
            ctx->dsp.idct_put(out_ptr + linesize * 4,     linesize, block_ptr, qmat);
461
            block_ptr += 64;
462 463 464 465
            if (blocks_per_mb > 2) {
                ctx->dsp.idct_put(out_ptr + linesize * 4 + 8, linesize, block_ptr, qmat);
                block_ptr += 64;
            }
466
        }
467 468 469 470 471
    } else {
        for (mb_num = 0; mb_num < mbs_per_slice; mb_num++, out_ptr += blocks_per_mb * 4) {
            ctx->dsp.idct_put(out_ptr,                    linesize, block_ptr, qmat);
            block_ptr += 64;
            ctx->dsp.idct_put(out_ptr + linesize * 4,     linesize, block_ptr, qmat);
472
            block_ptr += 64;
473 474 475 476 477 478
            if (blocks_per_mb > 2) {
                ctx->dsp.idct_put(out_ptr + 8,            linesize, block_ptr, qmat);
                block_ptr += 64;
                ctx->dsp.idct_put(out_ptr + linesize * 4 + 8, linesize, block_ptr, qmat);
                block_ptr += 64;
            }
479
        }
480
    }
481
    return 0;
482 483 484
}


485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509
static void unpack_alpha(GetBitContext *gb, uint16_t *dst, int num_coeffs,
                         const int num_bits)
{
    const int mask = (1 << num_bits) - 1;
    int i, idx, val, alpha_val;

    idx       = 0;
    alpha_val = mask;
    do {
        do {
            if (get_bits1(gb))
                val = get_bits(gb, num_bits);
            else {
                int sign;
                val  = get_bits(gb, num_bits == 16 ? 7 : 4);
                sign = val & 1;
                val  = (val + 2) >> 1;
                if (sign)
                    val = -val;
            }
            alpha_val = (alpha_val + val) & mask;
            if (num_bits == 16)
                dst[idx++] = alpha_val >> 6;
            else
                dst[idx++] = (alpha_val << 2) | (alpha_val >> 6);
510
            if (idx >= num_coeffs) {
511
                break;
512
            }
513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557
        } while (get_bits1(gb));
        val = get_bits(gb, 4);
        if (!val)
            val = get_bits(gb, 11);
        if (idx + val > num_coeffs)
            val = num_coeffs - idx;
        if (num_bits == 16)
            for (i = 0; i < val; i++)
                dst[idx++] = alpha_val >> 6;
        else
            for (i = 0; i < val; i++)
                dst[idx++] = (alpha_val << 2) | (alpha_val >> 6);
    } while (idx < num_coeffs);
}

/**
 * Decode alpha slice plane.
 */
static void decode_alpha_plane(ProresContext *ctx, ProresThreadData *td,
                               const uint8_t *buf, int data_size,
                               uint16_t *out_ptr, int linesize,
                               int mbs_per_slice)
{
    GetBitContext gb;
    int i;
    uint16_t *block_ptr;

    memset(td->blocks, 0, 8 * 4 * 64 * sizeof(*td->blocks));

    init_get_bits(&gb, buf, data_size << 3);

    if (ctx->alpha_info == 2)
        unpack_alpha(&gb, td->blocks, mbs_per_slice * 4 * 64, 16);
    else
        unpack_alpha(&gb, td->blocks, mbs_per_slice * 4 * 64, 8);

    block_ptr = td->blocks;

    for (i = 0; i < 16; i++) {
        memcpy(out_ptr, block_ptr, 16 * mbs_per_slice * sizeof(*out_ptr));
        out_ptr   += linesize >> 1;
        block_ptr += 16 * mbs_per_slice;
    }
}

558
static int decode_slice(AVCodecContext *avctx, void *tdata)
559
{
560
    ProresThreadData *td = tdata;
561 562 563 564 565 566
    ProresContext *ctx = avctx->priv_data;
    int mb_x_pos  = td->x_pos;
    int mb_y_pos  = td->y_pos;
    int pic_num   = ctx->pic_num;
    int slice_num = td->slice_num;
    int mbs_per_slice = td->slice_width;
567
    const uint8_t *buf;
568
    uint8_t *y_data, *u_data, *v_data, *a_data;
569
    AVFrame *pic = ctx->frame;
570
    int i, sf, slice_width_factor;
571 572 573
    int slice_data_size, hdr_size;
    int y_data_size, u_data_size, v_data_size, a_data_size;
    int y_linesize, u_linesize, v_linesize, a_linesize;
574
    int coff[4];
575
    int ret;
576

577 578
    buf             = ctx->slice_data[slice_num].index;
    slice_data_size = ctx->slice_data[slice_num + 1].index - buf;
579 580 581 582 583 584

    slice_width_factor = av_log2(mbs_per_slice);

    y_data     = pic->data[0];
    u_data     = pic->data[1];
    v_data     = pic->data[2];
585
    a_data     = pic->data[3];
586 587 588
    y_linesize = pic->linesize[0];
    u_linesize = pic->linesize[1];
    v_linesize = pic->linesize[2];
589
    a_linesize = pic->linesize[3];
590 591 592 593 594 595

    if (pic->interlaced_frame) {
        if (!(pic_num ^ pic->top_field_first)) {
            y_data += y_linesize;
            u_data += u_linesize;
            v_data += v_linesize;
596 597
            if (a_data)
                a_data += a_linesize;
598 599 600 601
        }
        y_linesize <<= 1;
        u_linesize <<= 1;
        v_linesize <<= 1;
602
        a_linesize <<= 1;
603
    }
604 605 606 607 608
    y_data += (mb_y_pos << 4) * y_linesize + (mb_x_pos << 5);
    u_data += (mb_y_pos << 4) * u_linesize + (mb_x_pos << ctx->mb_chroma_factor);
    v_data += (mb_y_pos << 4) * v_linesize + (mb_x_pos << ctx->mb_chroma_factor);
    if (a_data)
        a_data += (mb_y_pos << 4) * a_linesize + (mb_x_pos << 5);
609 610

    if (slice_data_size < 6) {
611
        av_log(avctx, AV_LOG_ERROR, "slice data too small\n");
612
        return AVERROR_INVALIDDATA;
613 614 615 616
    }

    /* parse slice header */
    hdr_size    = buf[0] >> 3;
617
    coff[0]     = hdr_size;
618
    y_data_size = AV_RB16(buf + 2);
619
    coff[1]     = coff[0] + y_data_size;
620
    u_data_size = AV_RB16(buf + 4);
621 622 623
    coff[2]     = coff[1] + u_data_size;
    v_data_size = hdr_size > 7 ? AV_RB16(buf + 6) : slice_data_size - coff[2];
    coff[3]     = coff[2] + v_data_size;
624
    a_data_size = ctx->alpha_info ? slice_data_size - coff[3] : 0;
625 626 627 628

    /* if V or alpha component size is negative that means that previous
       component sizes are too large */
    if (v_data_size < 0 || a_data_size < 0 || hdr_size < 6) {
629
        av_log(avctx, AV_LOG_ERROR, "invalid data size\n");
630
        return AVERROR_INVALIDDATA;
631 632 633 634 635 636
    }

    sf = av_clip(buf[1], 1, 224);
    sf = sf > 128 ? (sf - 96) << 2 : sf;

    /* scale quantization matrixes according with slice's scale factor */
637
    /* TODO: this can be SIMD-optimized a lot */
638 639
    if (ctx->qmat_changed || sf != td->prev_slice_sf) {
        td->prev_slice_sf = sf;
640
        for (i = 0; i < 64; i++) {
641 642
            td->qmat_luma_scaled[ctx->dsp.idct_permutation[i]]   = ctx->qmat_luma[i]   * sf;
            td->qmat_chroma_scaled[ctx->dsp.idct_permutation[i]] = ctx->qmat_chroma[i] * sf;
643 644 645 646
        }
    }

    /* decode luma plane */
647 648 649 650 651 652 653
    ret = decode_slice_plane(ctx, td, buf + coff[0], y_data_size,
                             (uint16_t*) y_data, y_linesize,
                             mbs_per_slice, 4, slice_width_factor + 2,
                             td->qmat_luma_scaled, 0);

    if (ret < 0)
        return ret;
654 655

    /* decode U chroma plane */
656 657 658 659 660 661 662
    ret = decode_slice_plane(ctx, td, buf + coff[1], u_data_size,
                             (uint16_t*) u_data, u_linesize,
                             mbs_per_slice, ctx->num_chroma_blocks,
                             slice_width_factor + ctx->chroma_factor - 1,
                             td->qmat_chroma_scaled, 1);
    if (ret < 0)
        return ret;
663 664

    /* decode V chroma plane */
665 666 667 668 669 670 671
    ret = decode_slice_plane(ctx, td, buf + coff[2], v_data_size,
                             (uint16_t*) v_data, v_linesize,
                             mbs_per_slice, ctx->num_chroma_blocks,
                             slice_width_factor + ctx->chroma_factor - 1,
                             td->qmat_chroma_scaled, 1);
    if (ret < 0)
        return ret;
672

673 674
    /* decode alpha plane if available */
    if (a_data && a_data_size)
675 676
        decode_alpha_plane(ctx, td, buf + coff[3], a_data_size,
                           (uint16_t*) a_data, a_linesize,
677 678
                           mbs_per_slice);

679 680 681 682 683 684 685 686 687 688 689
    return 0;
}


static int decode_picture(ProresContext *ctx, int pic_num,
                          AVCodecContext *avctx)
{
    int slice_num, slice_width, x_pos, y_pos;

    slice_num = 0;

690
    ctx->pic_num = pic_num;
691 692 693 694 695 696 697 698
    for (y_pos = 0; y_pos < ctx->num_y_mbs; y_pos++) {
        slice_width = 1 << ctx->slice_width_factor;

        for (x_pos = 0; x_pos < ctx->num_x_mbs && slice_width;
             x_pos += slice_width) {
            while (ctx->num_x_mbs - x_pos < slice_width)
                slice_width >>= 1;

699 700 701 702
            ctx->slice_data[slice_num].slice_num   = slice_num;
            ctx->slice_data[slice_num].x_pos       = x_pos;
            ctx->slice_data[slice_num].y_pos       = y_pos;
            ctx->slice_data[slice_num].slice_width = slice_width;
703 704 705 706 707

            slice_num++;
        }
    }

708
    return avctx->execute(avctx, decode_slice,
709 710
                          ctx->slice_data, NULL, slice_num,
                          sizeof(ctx->slice_data[0]));
711 712 713 714 715
}


#define MOVE_DATA_PTR(nbytes) buf += (nbytes); buf_size -= (nbytes)

716
static int decode_frame(AVCodecContext *avctx, void *data, int *got_frame,
717 718 719 720 721 722 723
                        AVPacket *avpkt)
{
    ProresContext *ctx = avctx->priv_data;
    const uint8_t *buf = avpkt->data;
    int buf_size       = avpkt->size;
    int frame_hdr_size, pic_num, pic_data_size;

724 725 726 727
    ctx->frame            = data;
    ctx->frame->pict_type = AV_PICTURE_TYPE_I;
    ctx->frame->key_frame = 1;

728 729 730 731
    /* check frame atom container */
    if (buf_size < 28 || buf_size < AV_RB32(buf) ||
        AV_RB32(buf + 4) != FRAME_ID) {
        av_log(avctx, AV_LOG_ERROR, "invalid frame\n");
732
        return AVERROR_INVALIDDATA;
733 734 735 736 737 738
    }

    MOVE_DATA_PTR(8);

    frame_hdr_size = decode_frame_header(ctx, buf, buf_size, avctx);
    if (frame_hdr_size < 0)
739
        return AVERROR_INVALIDDATA;
740 741 742

    MOVE_DATA_PTR(frame_hdr_size);

743
    if (ff_get_buffer(avctx, ctx->frame, 0) < 0)
744 745
        return -1;

746
    for (pic_num = 0; ctx->frame->interlaced_frame - pic_num + 1; pic_num++) {
747 748
        pic_data_size = decode_picture_header(ctx, buf, buf_size, avctx);
        if (pic_data_size < 0)
749
            return AVERROR_INVALIDDATA;
750 751 752 753 754 755 756

        if (decode_picture(ctx, pic_num, avctx))
            return -1;

        MOVE_DATA_PTR(pic_data_size);
    }

757 758
    ctx->frame = NULL;
    *got_frame = 1;
759 760 761 762 763 764 765 766 767

    return avpkt->size;
}


static av_cold int decode_close(AVCodecContext *avctx)
{
    ProresContext *ctx = avctx->priv_data;

768
    av_freep(&ctx->slice_data);
769 770 771 772 773 774

    return 0;
}


AVCodec ff_prores_lgpl_decoder = {
775
    .name           = "prores_lgpl",
776
    .long_name      = NULL_IF_CONFIG_SMALL("Apple ProRes (iCodec Pro)"),
777
    .type           = AVMEDIA_TYPE_VIDEO,
778
    .id             = AV_CODEC_ID_PRORES,
779 780 781 782
    .priv_data_size = sizeof(ProresContext),
    .init           = decode_init,
    .close          = decode_close,
    .decode         = decode_frame,
783
    .capabilities   = AV_CODEC_CAP_DR1 | AV_CODEC_CAP_SLICE_THREADS,
784
};