3 * Copyright (c) 2005 Jeff Muizelaar
5 * This file is part of FFmpeg.
7 * FFmpeg is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
10 * version 2.1 of the License, or (at your option) any later version.
12 * FFmpeg is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * Lesser General Public License for more details.
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with FFmpeg; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
25 * @author Jeff Muizelaar
31 #include "bytestream.h"
36 #define MAX_CHANNELS 8
37 #define MAX_BLOCKSIZE 65535
39 #define OUT_BUFFER_SIZE 16384
43 #define WAVE_FORMAT_PCM 0x0001
45 #define DEFAULT_BLOCK_SIZE 256
51 #define BITSHIFTSIZE 2
64 #define V2LPCQOFFSET (1 << LPCQUANT)
72 #define FN_BLOCKSIZE 5
78 /** indicates if the FN_* command is audio or non-audio */
79 static const uint8_t is_audio_command[10] = { 1, 1, 1, 1, 0, 0, 0, 1, 1, 0 };
81 #define VERBATIM_CKSIZE_SIZE 5
82 #define VERBATIM_BYTE_SIZE 8
83 #define CANONICAL_HEADER_SIZE 44
85 typedef struct ShortenContext {
86 AVCodecContext *avctx;
89 int min_framesize, max_framesize;
92 int32_t *decoded[MAX_CHANNELS];
93 int32_t *decoded_base[MAX_CHANNELS];
94 int32_t *offset[MAX_CHANNELS];
99 unsigned int allocated_bitstream_size;
101 uint8_t header[OUT_BUFFER_SIZE];
112 int got_quit_command;
115 static av_cold int shorten_decode_init(AVCodecContext *avctx)
117 ShortenContext *s = avctx->priv_data;
123 static int allocate_buffers(ShortenContext *s)
127 for (chan = 0; chan < s->channels; chan++) {
128 if (FFMAX(1, s->nmean) >= UINT_MAX / sizeof(int32_t)) {
129 av_log(s->avctx, AV_LOG_ERROR, "nmean too large\n");
130 return AVERROR_INVALIDDATA;
132 if (s->blocksize + s->nwrap >= UINT_MAX / sizeof(int32_t) ||
133 s->blocksize + s->nwrap <= (unsigned)s->nwrap) {
134 av_log(s->avctx, AV_LOG_ERROR,
135 "s->blocksize + s->nwrap too large\n");
136 return AVERROR_INVALIDDATA;
139 if ((err = av_reallocp(&s->offset[chan],
141 FFMAX(1, s->nmean))) < 0)
144 if ((err = av_reallocp(&s->decoded_base[chan], (s->blocksize + s->nwrap) *
145 sizeof(s->decoded_base[0][0]))) < 0)
147 for (i = 0; i < s->nwrap; i++)
148 s->decoded_base[chan][i] = 0;
149 s->decoded[chan] = s->decoded_base[chan] + s->nwrap;
152 if ((err = av_reallocp(&s->coeffs, s->nwrap * sizeof(*s->coeffs))) < 0)
158 static inline unsigned int get_uint(ShortenContext *s, int k)
161 k = get_ur_golomb_shorten(&s->gb, ULONGSIZE);
162 return get_ur_golomb_shorten(&s->gb, k);
165 static void fix_bitshift(ShortenContext *s, int32_t *buffer)
169 if (s->bitshift != 0)
170 for (i = 0; i < s->blocksize; i++)
171 buffer[i] <<= s->bitshift;
174 static int init_offset(ShortenContext *s)
178 int nblock = FFMAX(1, s->nmean);
179 /* initialise offset */
180 switch (s->internal_ftype) {
182 s->avctx->sample_fmt = AV_SAMPLE_FMT_U8P;
187 s->avctx->sample_fmt = AV_SAMPLE_FMT_S16P;
190 av_log(s->avctx, AV_LOG_ERROR, "unknown audio type\n");
191 return AVERROR_PATCHWELCOME;
194 for (chan = 0; chan < s->channels; chan++)
195 for (i = 0; i < nblock; i++)
196 s->offset[chan][i] = mean;
200 static int decode_wave_header(AVCodecContext *avctx, const uint8_t *header,
207 bytestream2_init(&gb, header, header_size);
209 if (bytestream2_get_le32(&gb) != MKTAG('R', 'I', 'F', 'F')) {
210 av_log(avctx, AV_LOG_ERROR, "missing RIFF tag\n");
211 return AVERROR_INVALIDDATA;
214 bytestream2_skip(&gb, 4); /* chunk size */
216 if (bytestream2_get_le32(&gb) != MKTAG('W', 'A', 'V', 'E')) {
217 av_log(avctx, AV_LOG_ERROR, "missing WAVE tag\n");
218 return AVERROR_INVALIDDATA;
221 while (bytestream2_get_le32(&gb) != MKTAG('f', 'm', 't', ' ')) {
222 len = bytestream2_get_le32(&gb);
223 bytestream2_skip(&gb, len);
224 if (len < 0 || bytestream2_get_bytes_left(&gb) < 16) {
225 av_log(avctx, AV_LOG_ERROR, "no fmt chunk found\n");
226 return AVERROR_INVALIDDATA;
229 len = bytestream2_get_le32(&gb);
232 av_log(avctx, AV_LOG_ERROR, "fmt chunk was too short\n");
233 return AVERROR_INVALIDDATA;
236 wave_format = bytestream2_get_le16(&gb);
238 switch (wave_format) {
239 case WAVE_FORMAT_PCM:
242 av_log(avctx, AV_LOG_ERROR, "unsupported wave format\n");
243 return AVERROR(ENOSYS);
246 bytestream2_skip(&gb, 2); // skip channels (already got from shorten header)
247 avctx->sample_rate = bytestream2_get_le32(&gb);
248 bytestream2_skip(&gb, 4); // skip bit rate (represents original uncompressed bit rate)
249 bytestream2_skip(&gb, 2); // skip block align (not needed)
250 bps = bytestream2_get_le16(&gb);
251 avctx->bits_per_coded_sample = bps;
253 if (bps != 16 && bps != 8) {
254 av_log(avctx, AV_LOG_ERROR, "unsupported number of bits per sample: %d\n", bps);
255 return AVERROR(ENOSYS);
260 av_log(avctx, AV_LOG_INFO, "%d header bytes unparsed\n", len);
265 static const int fixed_coeffs[][3] = {
272 static int decode_subframe_lpc(ShortenContext *s, int command, int channel,
273 int residual_size, int32_t coffset)
275 int pred_order, sum, qshift, init_sum, i, j;
278 if (command == FN_QLPC) {
279 /* read/validate prediction order */
280 pred_order = get_ur_golomb_shorten(&s->gb, LPCQSIZE);
281 if (pred_order > s->nwrap) {
282 av_log(s->avctx, AV_LOG_ERROR, "invalid pred_order %d\n",
284 return AVERROR(EINVAL);
286 /* read LPC coefficients */
287 for (i = 0; i < pred_order; i++)
288 s->coeffs[i] = get_sr_golomb_shorten(&s->gb, LPCQUANT);
293 /* fixed LPC coeffs */
294 pred_order = command;
295 if (pred_order >= FF_ARRAY_ELEMS(fixed_coeffs)) {
296 av_log(s->avctx, AV_LOG_ERROR, "invalid pred_order %d\n",
298 return AVERROR_INVALIDDATA;
300 coeffs = fixed_coeffs[pred_order];
304 /* subtract offset from previous samples to use in prediction */
305 if (command == FN_QLPC && coffset)
306 for (i = -pred_order; i < 0; i++)
307 s->decoded[channel][i] -= coffset;
309 /* decode residual and do LPC prediction */
310 init_sum = pred_order ? (command == FN_QLPC ? s->lpcqoffset : 0) : coffset;
311 for (i = 0; i < s->blocksize; i++) {
313 for (j = 0; j < pred_order; j++)
314 sum += coeffs[j] * s->decoded[channel][i - j - 1];
315 s->decoded[channel][i] = get_sr_golomb_shorten(&s->gb, residual_size) +
319 /* add offset to current samples */
320 if (command == FN_QLPC && coffset)
321 for (i = 0; i < s->blocksize; i++)
322 s->decoded[channel][i] += coffset;
327 static int read_header(ShortenContext *s)
331 /* shorten signature */
332 if (get_bits_long(&s->gb, 32) != AV_RB32("ajkg")) {
333 av_log(s->avctx, AV_LOG_ERROR, "missing shorten magic 'ajkg'\n");
334 return AVERROR_INVALIDDATA;
338 s->blocksize = DEFAULT_BLOCK_SIZE;
340 s->version = get_bits(&s->gb, 8);
341 s->internal_ftype = get_uint(s, TYPESIZE);
343 s->channels = get_uint(s, CHANSIZE);
345 av_log(s->avctx, AV_LOG_ERROR, "No channels reported\n");
346 return AVERROR_INVALIDDATA;
348 if (s->channels > MAX_CHANNELS) {
349 av_log(s->avctx, AV_LOG_ERROR, "too many channels: %d\n", s->channels);
351 return AVERROR_INVALIDDATA;
353 s->avctx->channels = s->channels;
355 /* get blocksize if version > 0 */
356 if (s->version > 0) {
360 blocksize = get_uint(s, av_log2(DEFAULT_BLOCK_SIZE));
361 if (!blocksize || blocksize > MAX_BLOCKSIZE) {
362 av_log(s->avctx, AV_LOG_ERROR,
363 "invalid or unsupported block size: %d\n",
365 return AVERROR(EINVAL);
367 s->blocksize = blocksize;
369 maxnlpc = get_uint(s, LPCQSIZE);
370 s->nmean = get_uint(s, 0);
372 skip_bytes = get_uint(s, NSKIPSIZE);
373 for (i = 0; i < skip_bytes; i++)
374 skip_bits(&s->gb, 8);
376 s->nwrap = FFMAX(NWRAP, maxnlpc);
378 if ((ret = allocate_buffers(s)) < 0)
381 if ((ret = init_offset(s)) < 0)
385 s->lpcqoffset = V2LPCQOFFSET;
387 if (get_ur_golomb_shorten(&s->gb, FNSIZE) != FN_VERBATIM) {
388 av_log(s->avctx, AV_LOG_ERROR,
389 "missing verbatim section at beginning of stream\n");
390 return AVERROR_INVALIDDATA;
393 s->header_size = get_ur_golomb_shorten(&s->gb, VERBATIM_CKSIZE_SIZE);
394 if (s->header_size >= OUT_BUFFER_SIZE ||
395 s->header_size < CANONICAL_HEADER_SIZE) {
396 av_log(s->avctx, AV_LOG_ERROR, "header is wrong size: %d\n",
398 return AVERROR_INVALIDDATA;
401 for (i = 0; i < s->header_size; i++)
402 s->header[i] = (char)get_ur_golomb_shorten(&s->gb, VERBATIM_BYTE_SIZE);
404 if ((ret = decode_wave_header(s->avctx, s->header, s->header_size)) < 0)
415 static int shorten_decode_frame(AVCodecContext *avctx, void *data,
416 int *got_frame_ptr, AVPacket *avpkt)
418 AVFrame *frame = data;
419 const uint8_t *buf = avpkt->data;
420 int buf_size = avpkt->size;
421 ShortenContext *s = avctx->priv_data;
422 int i, input_buf_size = 0;
425 /* allocate internal bitstream buffer */
426 if (s->max_framesize == 0) {
428 s->max_framesize = 8192; // should hopefully be enough for the first header
429 tmp_ptr = av_fast_realloc(s->bitstream, &s->allocated_bitstream_size,
430 s->max_framesize + FF_INPUT_BUFFER_PADDING_SIZE);
432 av_log(avctx, AV_LOG_ERROR, "error allocating bitstream buffer\n");
433 return AVERROR(ENOMEM);
435 memset(tmp_ptr, 0, s->allocated_bitstream_size);
436 s->bitstream = tmp_ptr;
439 /* append current packet data to bitstream buffer */
440 if (1 && s->max_framesize) { //FIXME truncated
441 buf_size = FFMIN(buf_size, s->max_framesize - s->bitstream_size);
442 input_buf_size = buf_size;
444 if (s->bitstream_index + s->bitstream_size + buf_size + FF_INPUT_BUFFER_PADDING_SIZE >
445 s->allocated_bitstream_size) {
446 memmove(s->bitstream, &s->bitstream[s->bitstream_index],
448 s->bitstream_index = 0;
451 memcpy(&s->bitstream[s->bitstream_index + s->bitstream_size], buf,
453 buf = &s->bitstream[s->bitstream_index];
454 buf_size += s->bitstream_size;
455 s->bitstream_size = buf_size;
457 /* do not decode until buffer has at least max_framesize bytes or
458 * the end of the file has been reached */
459 if (buf_size < s->max_framesize && avpkt->data) {
461 return input_buf_size;
464 /* init and position bitstream reader */
465 init_get_bits(&s->gb, buf, buf_size * 8);
466 skip_bits(&s->gb, s->bitindex);
468 /* process header or next subblock */
469 if (!s->got_header) {
470 if ((ret = read_header(s)) < 0)
476 /* if quit command was read previously, don't decode anything */
477 if (s->got_quit_command) {
483 while (s->cur_chan < s->channels) {
487 if (get_bits_left(&s->gb) < 3 + FNSIZE) {
492 cmd = get_ur_golomb_shorten(&s->gb, FNSIZE);
494 if (cmd > FN_VERBATIM) {
495 av_log(avctx, AV_LOG_ERROR, "unknown shorten function %d\n", cmd);
500 if (!is_audio_command[cmd]) {
501 /* process non-audio command */
504 len = get_ur_golomb_shorten(&s->gb, VERBATIM_CKSIZE_SIZE);
506 get_ur_golomb_shorten(&s->gb, VERBATIM_BYTE_SIZE);
509 unsigned bitshift = get_ur_golomb_shorten(&s->gb, BITSHIFTSIZE);
511 av_log(avctx, AV_LOG_ERROR, "bitshift %d is invalid\n",
513 return AVERROR_PATCHWELCOME;
515 s->bitshift = bitshift;
519 unsigned blocksize = get_uint(s, av_log2(s->blocksize));
520 if (blocksize > s->blocksize) {
521 av_log(avctx, AV_LOG_ERROR,
522 "Increasing block size is not supported\n");
523 return AVERROR_PATCHWELCOME;
525 if (!blocksize || blocksize > MAX_BLOCKSIZE) {
526 av_log(avctx, AV_LOG_ERROR, "invalid or unsupported "
527 "block size: %d\n", blocksize);
528 return AVERROR(EINVAL);
530 s->blocksize = blocksize;
534 s->got_quit_command = 1;
537 if (cmd == FN_BLOCKSIZE || cmd == FN_QUIT) {
542 /* process audio command */
543 int residual_size = 0;
544 int channel = s->cur_chan;
547 /* get Rice code for residual decoding */
548 if (cmd != FN_ZERO) {
549 residual_size = get_ur_golomb_shorten(&s->gb, ENERGYSIZE);
550 /* This is a hack as version 0 differed in the definition
551 * of get_sr_golomb_shorten(). */
556 /* calculate sample offset using means from previous blocks */
558 coffset = s->offset[channel][0];
560 int32_t sum = (s->version < 2) ? 0 : s->nmean / 2;
561 for (i = 0; i < s->nmean; i++)
562 sum += s->offset[channel][i];
563 coffset = sum / s->nmean;
565 coffset = s->bitshift == 0 ? coffset : coffset >> s->bitshift - 1 >> 1;
568 /* decode samples for this channel */
569 if (cmd == FN_ZERO) {
570 for (i = 0; i < s->blocksize; i++)
571 s->decoded[channel][i] = 0;
573 if ((ret = decode_subframe_lpc(s, cmd, channel,
574 residual_size, coffset)) < 0)
578 /* update means with info from the current block */
580 int32_t sum = (s->version < 2) ? 0 : s->blocksize / 2;
581 for (i = 0; i < s->blocksize; i++)
582 sum += s->decoded[channel][i];
584 for (i = 1; i < s->nmean; i++)
585 s->offset[channel][i - 1] = s->offset[channel][i];
588 s->offset[channel][s->nmean - 1] = sum / s->blocksize;
590 s->offset[channel][s->nmean - 1] = (sum / s->blocksize) << s->bitshift;
593 /* copy wrap samples for use with next block */
594 for (i = -s->nwrap; i < 0; i++)
595 s->decoded[channel][i] = s->decoded[channel][i + s->blocksize];
597 /* shift samples to add in unused zero bits which were removed
599 fix_bitshift(s, s->decoded[channel]);
601 /* if this is the last channel in the block, output the samples */
603 if (s->cur_chan == s->channels) {
605 int16_t *samples_s16;
608 /* get output buffer */
609 frame->nb_samples = s->blocksize;
610 if ((ret = ff_get_buffer(avctx, frame, 0)) < 0)
613 for (chan = 0; chan < s->channels; chan++) {
614 samples_u8 = ((uint8_t **)frame->extended_data)[chan];
615 samples_s16 = ((int16_t **)frame->extended_data)[chan];
616 for (i = 0; i < s->blocksize; i++) {
617 switch (s->internal_ftype) {
619 *samples_u8++ = av_clip_uint8(s->decoded[chan][i]);
623 *samples_s16++ = av_clip_int16(s->decoded[chan][i]);
633 if (s->cur_chan < s->channels)
637 s->bitindex = get_bits_count(&s->gb) - 8 * (get_bits_count(&s->gb) / 8);
638 i = get_bits_count(&s->gb) / 8;
640 av_log(s->avctx, AV_LOG_ERROR, "overread: %d\n", i - buf_size);
641 s->bitstream_size = 0;
642 s->bitstream_index = 0;
643 return AVERROR_INVALIDDATA;
645 if (s->bitstream_size) {
646 s->bitstream_index += i;
647 s->bitstream_size -= i;
648 return input_buf_size;
653 static av_cold int shorten_decode_close(AVCodecContext *avctx)
655 ShortenContext *s = avctx->priv_data;
658 for (i = 0; i < s->channels; i++) {
659 s->decoded[i] = NULL;
660 av_freep(&s->decoded_base[i]);
661 av_freep(&s->offset[i]);
663 av_freep(&s->bitstream);
664 av_freep(&s->coeffs);
669 AVCodec ff_shorten_decoder = {
671 .long_name = NULL_IF_CONFIG_SMALL("Shorten"),
672 .type = AVMEDIA_TYPE_AUDIO,
673 .id = AV_CODEC_ID_SHORTEN,
674 .priv_data_size = sizeof(ShortenContext),
675 .init = shorten_decode_init,
676 .close = shorten_decode_close,
677 .decode = shorten_decode_frame,
678 .capabilities = CODEC_CAP_DELAY | CODEC_CAP_DR1,
679 .sample_fmts = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_S16P,
681 AV_SAMPLE_FMT_NONE },