hapdec.c 13.9 KB
Newer Older
1 2 3
/*
 * Vidvox Hap decoder
 * Copyright (C) 2015 Vittorio Giovara <vittorio.giovara@gmail.com>
4
 * Copyright (C) 2015 Tom Butterworth <bangnoise@gmail.com>
5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39
 *
 * This file is part of Libav.
 *
 * Libav is free software; you can redistribute it and/or
 * 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.
 *
 * Libav is distributed in the hope that it will be useful,
 * 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
 * License along with Libav; if not, write to the Free Software
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
 */

/**
 * @file
 * Hap decoder
 *
 * Fourcc: Hap1, Hap5, HapY
 *
 * https://github.com/Vidvox/hap/blob/master/documentation/HapVideoDRAFT.md
 */

#include <stdint.h>

#include "libavutil/imgutils.h"

#include "avcodec.h"
#include "bytestream.h"
#include "hap.h"
#include "internal.h"
40
#include "memory.h"
41 42 43 44 45 46 47
#include "snappy.h"
#include "texturedsp.h"
#include "thread.h"

/* The first three bytes are the size of the section past the header, or zero
 * if the length is stored in the next long word. The fourth byte in the first
 * long word indicates the type of the current section. */
48 49
static int parse_section_header(GetByteContext *gbc, int *section_size,
                                enum HapSectionType *section_type)
50 51 52 53
{
    if (bytestream2_get_bytes_left(gbc) < 4)
        return AVERROR_INVALIDDATA;

54 55
    *section_size = bytestream2_get_le24(gbc);
    *section_type = bytestream2_get_byte(gbc);
56

57
    if (*section_size == 0) {
58 59
        if (bytestream2_get_bytes_left(gbc) < 4)
            return AVERROR_INVALIDDATA;
60 61

        *section_size = bytestream2_get_le32(gbc);
62 63
    }

64
    if (*section_size > bytestream2_get_bytes_left(gbc))
65
        return AVERROR_INVALIDDATA;
66 67 68
    else
        return 0;
}
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149
static int hap_parse_decode_instructions(HapContext *ctx, int size)
{
    GetByteContext *gbc = &ctx->gbc;
    int section_size;
    enum HapSectionType section_type;
    int is_first_table = 1, had_offsets = 0, had_compressors = 0, had_sizes = 0;
    int i, ret;

    while (size > 0) {
        int stream_remaining = bytestream2_get_bytes_left(gbc);
        ret = parse_section_header(gbc, &section_size, &section_type);
        if (ret != 0)
            return ret;

        size -= stream_remaining - bytestream2_get_bytes_left(gbc);

        switch (section_type) {
            case HAP_ST_COMPRESSOR_TABLE:
                ret = ff_hap_set_chunk_count(ctx, section_size, is_first_table);
                if (ret != 0)
                    return ret;
                for (i = 0; i < section_size; i++) {
                    ctx->chunks[i].compressor = bytestream2_get_byte(gbc) << 4;
                }
                had_compressors = 1;
                is_first_table = 0;
                break;
            case HAP_ST_SIZE_TABLE:
                ret = ff_hap_set_chunk_count(ctx, section_size / 4, is_first_table);
                if (ret != 0)
                    return ret;
                for (i = 0; i < section_size / 4; i++) {
                    ctx->chunks[i].compressed_size = bytestream2_get_le32(gbc);
                }
                had_sizes = 1;
                is_first_table = 0;
                break;
            case HAP_ST_OFFSET_TABLE:
                ret = ff_hap_set_chunk_count(ctx, section_size / 4, is_first_table);
                if (ret != 0)
                    return ret;
                for (i = 0; i < section_size / 4; i++) {
                    ctx->chunks[i].compressed_offset = bytestream2_get_le32(gbc);
                }
                had_offsets = 1;
                is_first_table = 0;
                break;
            default:
                break;
        }
        size -= section_size;
    }

    if (!had_sizes || !had_compressors)
        return AVERROR_INVALIDDATA;

    /* The offsets table is optional. If not present than calculate offsets by
     * summing the sizes of preceding chunks. */
    if (!had_offsets) {
        size_t running_size = 0;
        for (i = 0; i < ctx->chunk_count; i++) {
            ctx->chunks[i].compressed_offset = running_size;
            running_size += ctx->chunks[i].compressed_size;
        }
    }

    return 0;
}

static int hap_can_use_tex_in_place(HapContext *ctx)
{
    int i;
    size_t running_offset = 0;
    for (i = 0; i < ctx->chunk_count; i++) {
        if (ctx->chunks[i].compressed_offset != running_offset
            || ctx->chunks[i].compressor != HAP_COMP_NONE)
            return 0;
        running_offset += ctx->chunks[i].compressed_size;
    }
    return 1;
150 151
}

152
static int hap_parse_frame_header(AVCodecContext *avctx)
153 154 155
{
    HapContext *ctx = avctx->priv_data;
    GetByteContext *gbc = &ctx->gbc;
156 157
    int section_size;
    enum HapSectionType section_type;
158
    const char *compressorstr;
159 160 161 162 163
    int i, ret;

    ret = parse_section_header(gbc, &section_size, &section_type);
    if (ret != 0)
        return ret;
164

165 166 167
    if ((avctx->codec_tag == MKTAG('H','a','p','1') && (section_type & 0x0F) != HAP_FMT_RGBDXT1) ||
        (avctx->codec_tag == MKTAG('H','a','p','5') && (section_type & 0x0F) != HAP_FMT_RGBADXT5) ||
        (avctx->codec_tag == MKTAG('H','a','p','Y') && (section_type & 0x0F) != HAP_FMT_YCOCGDXT5)) {
168
        av_log(avctx, AV_LOG_ERROR,
169
               "Invalid texture format %#04x.\n", section_type & 0x0F);
170 171 172
        return AVERROR_INVALIDDATA;
    }

173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231
    switch (section_type & 0xF0) {
        case HAP_COMP_NONE:
        case HAP_COMP_SNAPPY:
            ret = ff_hap_set_chunk_count(ctx, 1, 1);
            if (ret == 0) {
                ctx->chunks[0].compressor = section_type & 0xF0;
                ctx->chunks[0].compressed_offset = 0;
                ctx->chunks[0].compressed_size = section_size;
            }
            if (ctx->chunks[0].compressor == HAP_COMP_NONE) {
                compressorstr = "none";
            } else {
                compressorstr = "snappy";
            }
            break;
        case HAP_COMP_COMPLEX:
            ret = parse_section_header(gbc, &section_size, &section_type);
            if (ret == 0 && section_type != HAP_ST_DECODE_INSTRUCTIONS)
                ret = AVERROR_INVALIDDATA;
            if (ret == 0)
                ret = hap_parse_decode_instructions(ctx, section_size);
            compressorstr = "complex";
            break;
        default:
            ret = AVERROR_INVALIDDATA;
            break;
    }

    if (ret != 0)
        return ret;

    /* Check the frame is valid and read the uncompressed chunk sizes */
    ctx->tex_size = 0;
    for (i = 0; i < ctx->chunk_count; i++) {
        HapChunk *chunk = &ctx->chunks[i];

        /* Check the compressed buffer is valid */
        if (chunk->compressed_offset + chunk->compressed_size > bytestream2_get_bytes_left(gbc))
            return AVERROR_INVALIDDATA;

        /* Chunks are unpacked sequentially, ctx->tex_size is the uncompressed
         * size thus far */
        chunk->uncompressed_offset = ctx->tex_size;

        /* Fill out uncompressed size */
        if (chunk->compressor == HAP_COMP_SNAPPY) {
            GetByteContext gbc_tmp;
            int64_t uncompressed_size;
            bytestream2_init(&gbc_tmp, gbc->buffer + chunk->compressed_offset,
                             chunk->compressed_size);
            uncompressed_size = ff_snappy_peek_uncompressed_length(&gbc_tmp);
            if (uncompressed_size < 0) {
                return uncompressed_size;
            }
            chunk->uncompressed_size = uncompressed_size;
        } else if (chunk->compressor == HAP_COMP_NONE) {
            chunk->uncompressed_size = chunk->compressed_size;
        } else {
            return AVERROR_INVALIDDATA;
232
        }
233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255
        ctx->tex_size += chunk->uncompressed_size;
    }

    av_log(avctx, AV_LOG_DEBUG, "%s compressor\n", compressorstr);

    return ret;
}

static int decompress_chunks_thread(AVCodecContext *avctx, void *arg,
                                    int chunk_nb, int thread_nb)
{
    HapContext *ctx = avctx->priv_data;

    HapChunk *chunk = &ctx->chunks[chunk_nb];
    GetByteContext gbc;
    uint8_t *dst = ctx->tex_buf + chunk->uncompressed_offset;

    bytestream2_init(&gbc, ctx->gbc.buffer + chunk->compressed_offset, chunk->compressed_size);

    if (chunk->compressor == HAP_COMP_SNAPPY) {
        int ret;
        int64_t uncompressed_size = ctx->tex_size;

256
        /* Uncompress the frame */
257
        ret = ff_snappy_uncompress(&gbc, dst, &uncompressed_size);
258 259 260 261
        if (ret < 0) {
             av_log(avctx, AV_LOG_ERROR, "Snappy uncompress error\n");
             return ret;
        }
262 263
    } else if (chunk->compressor == HAP_COMP_NONE) {
        bytestream2_get_buffer(&gbc, dst, chunk->compressed_size);
264 265 266 267 268 269
    }

    return 0;
}

static int decompress_texture_thread(AVCodecContext *avctx, void *arg,
270
                                     int slice, int thread_nb)
271 272 273
{
    HapContext *ctx = avctx->priv_data;
    AVFrame *frame = arg;
274 275
    const uint8_t *d = ctx->tex_data;
    int w_block = avctx->coded_width / TEXTURE_BLOCK_W;
276
    int h_block = avctx->coded_height / TEXTURE_BLOCK_H;
277 278
    int x, y;
    int start_slice, end_slice;
279 280
    int base_blocks_per_slice = h_block / ctx->slice_count;
    int remainder_blocks = h_block % ctx->slice_count;
281

282 283 284 285 286 287
    /* When the frame height (in blocks) doesn't divide evenly between the
     * number of slices, spread the remaining blocks evenly between the first
     * operations */
    start_slice = slice * base_blocks_per_slice;
    /* Add any extra blocks (one per slice) that have been added before this slice */
    start_slice += FFMIN(slice, remainder_blocks);
288

289 290 291 292
    end_slice = start_slice + base_blocks_per_slice;
    /* Add an extra block if there are still remainder blocks to be accounted for */
    if (slice < remainder_blocks)
        end_slice++;
293 294 295 296 297 298 299 300 301

    for (y = start_slice; y < end_slice; y++) {
        uint8_t *p = frame->data[0] + y * frame->linesize[0] * TEXTURE_BLOCK_H;
        int off  = y * w_block;
        for (x = 0; x < w_block; x++) {
            ctx->tex_fun(p + x * 16, frame->linesize[0],
                         d + (off + x) * ctx->tex_rat);
        }
    }
302 303 304 305 306 307 308 309 310

    return 0;
}

static int hap_decode(AVCodecContext *avctx, void *data,
                      int *got_frame, AVPacket *avpkt)
{
    HapContext *ctx = avctx->priv_data;
    ThreadFrame tframe;
311
    int ret, i;
312 313 314 315

    bytestream2_init(&ctx->gbc, avpkt->data, avpkt->size);

    /* Check for section header */
316
    ret = hap_parse_frame_header(avctx);
317 318 319 320 321 322 323 324 325 326
    if (ret < 0)
        return ret;

    /* Get the output frame ready to receive data */
    tframe.f = data;
    ret = ff_thread_get_buffer(avctx, &tframe, 0);
    if (ret < 0)
        return ret;
    ff_thread_finish_setup(avctx);

327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347
    /* Unpack the DXT texture */
    if (hap_can_use_tex_in_place(ctx)) {
        /* Only DXTC texture compression in a contiguous block */
        ctx->tex_data = ctx->gbc.buffer;
    } else {
        /* Perform the second-stage decompression */
        ret = av_reallocp(&ctx->tex_buf, ctx->tex_size);
        if (ret < 0)
            return ret;

        avctx->execute2(avctx, decompress_chunks_thread, NULL,
                        ctx->chunk_results, ctx->chunk_count);

        for (i = 0; i < ctx->chunk_count; i++) {
            if (ctx->chunk_results[i] < 0)
                return ctx->chunk_results[i];
        }

        ctx->tex_data = ctx->tex_buf;
    }

348
    /* Use the decompress function on the texture, one block per thread */
349
    avctx->execute2(avctx, decompress_texture_thread, tframe.f, NULL, ctx->slice_count);
350 351 352 353 354 355 356 357 358 359 360 361

    /* Frame is ready to be output */
    tframe.f->pict_type = AV_PICTURE_TYPE_I;
    tframe.f->key_frame = 1;
    *got_frame = 1;

    return avpkt->size;
}

static av_cold int hap_init(AVCodecContext *avctx)
{
    HapContext *ctx = avctx->priv_data;
362
    const char *texture_name;
363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379
    int ret = av_image_check_size(avctx->width, avctx->height, 0, avctx);

    if (ret < 0) {
        av_log(avctx, AV_LOG_ERROR, "Invalid video size %dx%d.\n",
               avctx->width, avctx->height);
        return ret;
    }

    /* Since codec is based on 4x4 blocks, size is aligned to 4 */
    avctx->coded_width  = FFALIGN(avctx->width,  TEXTURE_BLOCK_W);
    avctx->coded_height = FFALIGN(avctx->height, TEXTURE_BLOCK_H);

    /* Technically only one mode has alpha, but 32 bits are easier to handle */
    avctx->pix_fmt = AV_PIX_FMT_RGBA;

    ff_texturedsp_init(&ctx->dxtc);

380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401
    switch (avctx->codec_tag) {
    case MKTAG('H','a','p','1'):
        texture_name = "DXT1";
        ctx->tex_rat = 8;
        ctx->tex_fun = ctx->dxtc.dxt1_block;
        break;
    case MKTAG('H','a','p','5'):
        texture_name = "DXT5";
        ctx->tex_rat = 16;
        ctx->tex_fun = ctx->dxtc.dxt5_block;
        break;
    case MKTAG('H','a','p','Y'):
        texture_name = "DXT5-YCoCg-scaled";
        ctx->tex_rat = 16;
        ctx->tex_fun = ctx->dxtc.dxt5ys_block;
        break;
    default:
        return AVERROR_DECODER_NOT_FOUND;
    }

    av_log(avctx, AV_LOG_DEBUG, "%s texture\n", texture_name);

402 403 404
    ctx->slice_count = av_clip(avctx->thread_count, 1,
                               avctx->coded_height / TEXTURE_BLOCK_H);

405 406 407 408 409 410 411
    return 0;
}

static av_cold int hap_close(AVCodecContext *avctx)
{
    HapContext *ctx = avctx->priv_data;

412
    ff_hap_free_context(ctx);
413 414 415 416 417 418 419 420 421 422 423 424 425

    return 0;
}

AVCodec ff_hap_decoder = {
    .name           = "hap",
    .long_name      = NULL_IF_CONFIG_SMALL("Vidvox Hap decoder"),
    .type           = AVMEDIA_TYPE_VIDEO,
    .id             = AV_CODEC_ID_HAP,
    .init           = hap_init,
    .decode         = hap_decode,
    .close          = hap_close,
    .priv_data_size = sizeof(HapContext),
426 427
    .capabilities   = AV_CODEC_CAP_FRAME_THREADS | AV_CODEC_CAP_SLICE_THREADS |
                      AV_CODEC_CAP_DR1,
428 429 430
    .caps_internal  = FF_CODEC_CAP_INIT_THREADSAFE |
                      FF_CODEC_CAP_INIT_CLEANUP,
};