2 * Sierra VMD Audio & Video Decoders
3 * Copyright (C) 2004 the ffmpeg project
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
24 * Sierra VMD audio & video decoders
25 * by Vladimir "VAG" Gneushev (vagsoft at mail.ru)
26 * for more information on the Sierra VMD format, visit:
27 * http://www.pcisys.net/~melanson/codecs/
29 * The video decoder outputs PAL8 colorspace data. The decoder expects
30 * a 0x330-byte VMD file header to be transmitted via extradata during
31 * codec initialization. Each encoded frame that is sent to this decoder
32 * is expected to be prepended with the appropriate 16-byte frame
33 * information record from the VMD file.
35 * The audio decoder, like the video decoder, expects each encoded data
36 * chunk to be prepended with the appropriate 16-byte frame information
37 * record from the VMD file. It does not require the 0x330-byte VMD file
38 * header, but it does need the audio setup parameters passed in through
39 * normal libavcodec API means.
46 #include "libavutil/avassert.h"
47 #include "libavutil/channel_layout.h"
48 #include "libavutil/common.h"
49 #include "libavutil/intreadwrite.h"
52 #include "bytestream.h"
54 #define VMD_HEADER_SIZE 0x330
55 #define PALETTE_COUNT 256
61 typedef struct VmdVideoContext {
63 AVCodecContext *avctx;
66 const unsigned char *buf;
69 unsigned char palette[PALETTE_COUNT * 4];
70 unsigned char *unpack_buffer;
71 int unpack_buffer_size;
76 #define QUEUE_SIZE 0x1000
77 #define QUEUE_MASK 0x0FFF
79 static void lz_unpack(const unsigned char *src, int src_len,
80 unsigned char *dest, int dest_len)
84 unsigned char queue[QUEUE_SIZE];
86 unsigned int dataleft;
87 unsigned int chainofs;
88 unsigned int chainlen;
94 bytestream2_init(&gb, src, src_len);
97 dataleft = bytestream2_get_le32(&gb);
98 memset(queue, 0x20, QUEUE_SIZE);
99 if (bytestream2_get_bytes_left(&gb) < 4)
101 if (bytestream2_peek_le32(&gb) == 0x56781234) {
102 bytestream2_skipu(&gb, 4);
107 speclen = 100; /* no speclen */
110 while (dataleft > 0 && bytestream2_get_bytes_left(&gb) > 0) {
111 tag = bytestream2_get_byteu(&gb);
112 if ((tag == 0xFF) && (dataleft > 8)) {
113 if (d_end - d < 8 || bytestream2_get_bytes_left(&gb) < 8)
115 for (i = 0; i < 8; i++) {
116 queue[qpos++] = *d++ = bytestream2_get_byteu(&gb);
121 for (i = 0; i < 8; i++) {
125 if (d_end - d < 1 || bytestream2_get_bytes_left(&gb) < 1)
127 queue[qpos++] = *d++ = bytestream2_get_byteu(&gb);
131 chainofs = bytestream2_get_byte(&gb);
132 chainofs |= ((bytestream2_peek_byte(&gb) & 0xF0) << 4);
133 chainlen = (bytestream2_get_byte(&gb) & 0x0F) + 3;
134 if (chainlen == speclen) {
135 chainlen = bytestream2_get_byte(&gb) + 0xF + 3;
137 if (d_end - d < chainlen)
139 for (j = 0; j < chainlen; j++) {
140 *d = queue[chainofs++ & QUEUE_MASK];
141 queue[qpos++] = *d++;
144 dataleft -= chainlen;
151 static int rle_unpack(const unsigned char *src, unsigned char *dest,
152 int src_count, int src_size, int dest_len)
156 unsigned char *dest_end = dest + dest_len;
160 bytestream2_init(&gb, src, src_size);
163 if (bytestream2_get_bytes_left(&gb) < 1)
165 *pd++ = bytestream2_get_byteu(&gb);
170 if (bytestream2_get_bytes_left(&gb) < 1)
172 l = bytestream2_get_byteu(&gb);
175 if (dest_end - pd < l || bytestream2_get_bytes_left(&gb) < l)
176 return bytestream2_tell(&gb);
177 bytestream2_get_bufferu(&gb, pd, l);
180 if (dest_end - pd < 2*l || bytestream2_get_bytes_left(&gb) < 2)
181 return bytestream2_tell(&gb);
182 run_val = bytestream2_get_ne16(&gb);
183 for (i = 0; i < l; i++) {
184 AV_WN16(pd, run_val);
190 } while (used < src_count);
192 return bytestream2_tell(&gb);
195 static int vmd_decode(VmdVideoContext *s, AVFrame *frame)
198 unsigned int *palette32;
199 unsigned char r, g, b;
204 unsigned char *dp; /* pointer to current frame */
205 unsigned char *pp; /* pointer to previous frame */
209 int frame_x, frame_y;
210 int frame_width, frame_height;
212 frame_x = AV_RL16(&s->buf[6]);
213 frame_y = AV_RL16(&s->buf[8]);
214 frame_width = AV_RL16(&s->buf[10]) - frame_x + 1;
215 frame_height = AV_RL16(&s->buf[12]) - frame_y + 1;
217 if ((frame_width == s->avctx->width && frame_height == s->avctx->height) &&
218 (frame_x || frame_y)) {
226 if (frame_x < 0 || frame_width < 0 ||
227 frame_x >= s->avctx->width ||
228 frame_width > s->avctx->width ||
229 frame_x + frame_width > s->avctx->width) {
230 av_log(s->avctx, AV_LOG_ERROR,
231 "Invalid horizontal range %d-%d\n",
232 frame_x, frame_width);
233 return AVERROR_INVALIDDATA;
235 if (frame_y < 0 || frame_height < 0 ||
236 frame_y >= s->avctx->height ||
237 frame_height > s->avctx->height ||
238 frame_y + frame_height > s->avctx->height) {
239 av_log(s->avctx, AV_LOG_ERROR,
240 "Invalid vertical range %d-%d\n",
241 frame_x, frame_width);
242 return AVERROR_INVALIDDATA;
245 /* if only a certain region will be updated, copy the entire previous
246 * frame before the decode */
247 if (s->prev_frame.data[0] &&
248 (frame_x || frame_y || (frame_width != s->avctx->width) ||
249 (frame_height != s->avctx->height))) {
251 memcpy(frame->data[0], s->prev_frame.data[0],
252 s->avctx->height * frame->linesize[0]);
255 /* check if there is a new palette */
256 bytestream2_init(&gb, s->buf + 16, s->size - 16);
257 if (s->buf[15] & 0x02) {
258 bytestream2_skip(&gb, 2);
259 palette32 = (unsigned int *)s->palette;
260 if (bytestream2_get_bytes_left(&gb) >= PALETTE_COUNT * 3) {
261 for (i = 0; i < PALETTE_COUNT; i++) {
262 r = bytestream2_get_byteu(&gb) * 4;
263 g = bytestream2_get_byteu(&gb) * 4;
264 b = bytestream2_get_byteu(&gb) * 4;
265 palette32[i] = 0xFFU << 24 | (r << 16) | (g << 8) | (b);
266 palette32[i] |= palette32[i] >> 6 & 0x30303;
269 av_log(s->avctx, AV_LOG_ERROR, "Incomplete palette\n");
270 return AVERROR_INVALIDDATA;
277 /* originally UnpackFrame in VAG's code */
278 if (bytestream2_get_bytes_left(&gb) < 1)
279 return AVERROR_INVALIDDATA;
280 meth = bytestream2_get_byteu(&gb);
282 if (!s->unpack_buffer_size) {
283 av_log(s->avctx, AV_LOG_ERROR,
284 "Trying to unpack LZ-compressed frame with no LZ buffer\n");
285 return AVERROR_INVALIDDATA;
287 lz_unpack(gb.buffer, bytestream2_get_bytes_left(&gb),
288 s->unpack_buffer, s->unpack_buffer_size);
290 bytestream2_init(&gb, s->unpack_buffer, s->unpack_buffer_size);
293 dp = &frame->data[0][frame_y * frame->linesize[0] + frame_x];
294 pp = &s->prev_frame.data[0][frame_y * s->prev_frame.linesize[0] + frame_x];
297 for (i = 0; i < frame_height; i++) {
300 len = bytestream2_get_byte(&gb);
302 len = (len & 0x7F) + 1;
303 if (ofs + len > frame_width ||
304 bytestream2_get_bytes_left(&gb) < len)
305 return AVERROR_INVALIDDATA;
306 bytestream2_get_bufferu(&gb, &dp[ofs], len);
309 /* interframe pixel copy */
310 if (ofs + len + 1 > frame_width || !s->prev_frame.data[0])
311 return AVERROR_INVALIDDATA;
312 memcpy(&dp[ofs], &pp[ofs], len + 1);
315 } while (ofs < frame_width);
316 if (ofs > frame_width) {
317 av_log(s->avctx, AV_LOG_ERROR,
318 "offset > width (%d > %d)\n",
320 return AVERROR_INVALIDDATA;
322 dp += frame->linesize[0];
323 pp += s->prev_frame.linesize[0];
328 for (i = 0; i < frame_height; i++) {
329 bytestream2_get_buffer(&gb, dp, frame_width);
330 dp += frame->linesize[0];
331 pp += s->prev_frame.linesize[0];
336 for (i = 0; i < frame_height; i++) {
339 len = bytestream2_get_byte(&gb);
341 len = (len & 0x7F) + 1;
342 if (bytestream2_peek_byte(&gb) == 0xFF) {
344 bytestream2_get_byte(&gb);
345 len = rle_unpack(gb.buffer, &dp[ofs],
346 len, bytestream2_get_bytes_left(&gb),
349 bytestream2_skip(&gb, len);
351 bytestream2_get_buffer(&gb, &dp[ofs], len);
355 /* interframe pixel copy */
356 if (ofs + len + 1 > frame_width || !s->prev_frame.data[0])
357 return AVERROR_INVALIDDATA;
358 memcpy(&dp[ofs], &pp[ofs], len + 1);
361 } while (ofs < frame_width);
362 if (ofs > frame_width) {
363 av_log(s->avctx, AV_LOG_ERROR,
364 "offset > width (%d > %d)\n",
366 return AVERROR_INVALIDDATA;
368 dp += frame->linesize[0];
369 pp += s->prev_frame.linesize[0];
376 static av_cold int vmdvideo_decode_init(AVCodecContext *avctx)
378 VmdVideoContext *s = avctx->priv_data;
380 unsigned int *palette32;
381 int palette_index = 0;
382 unsigned char r, g, b;
383 unsigned char *vmd_header;
384 unsigned char *raw_palette;
387 avctx->pix_fmt = AV_PIX_FMT_PAL8;
389 /* make sure the VMD header made it */
390 if (s->avctx->extradata_size != VMD_HEADER_SIZE) {
391 av_log(s->avctx, AV_LOG_ERROR, "expected extradata size of %d\n",
393 return AVERROR_INVALIDDATA;
395 vmd_header = (unsigned char *)avctx->extradata;
397 s->unpack_buffer_size = AV_RL32(&vmd_header[800]);
398 if (s->unpack_buffer_size) {
399 s->unpack_buffer = av_malloc(s->unpack_buffer_size);
400 if (!s->unpack_buffer)
401 return AVERROR(ENOMEM);
404 /* load up the initial palette */
405 raw_palette = &vmd_header[28];
406 palette32 = (unsigned int *)s->palette;
407 for (i = 0; i < PALETTE_COUNT; i++) {
408 r = raw_palette[palette_index++] * 4;
409 g = raw_palette[palette_index++] * 4;
410 b = raw_palette[palette_index++] * 4;
411 palette32[i] = (r << 16) | (g << 8) | (b);
414 avcodec_get_frame_defaults(&s->prev_frame);
419 static int vmdvideo_decode_frame(AVCodecContext *avctx,
420 void *data, int *got_frame,
423 const uint8_t *buf = avpkt->data;
424 int buf_size = avpkt->size;
425 VmdVideoContext *s = avctx->priv_data;
426 AVFrame *frame = data;
433 return AVERROR_INVALIDDATA;
435 if ((ret = ff_get_buffer(avctx, frame, AV_GET_BUFFER_FLAG_REF)) < 0)
438 if ((ret = vmd_decode(s, frame)) < 0)
441 /* make the palette available on the way out */
442 memcpy(frame->data[1], s->palette, PALETTE_COUNT * 4);
445 av_frame_unref(&s->prev_frame);
446 if ((ret = av_frame_ref(&s->prev_frame, frame)) < 0)
451 /* report that the buffer was completely consumed */
455 static av_cold int vmdvideo_decode_end(AVCodecContext *avctx)
457 VmdVideoContext *s = avctx->priv_data;
459 av_frame_unref(&s->prev_frame);
460 av_free(s->unpack_buffer);
470 #define BLOCK_TYPE_AUDIO 1
471 #define BLOCK_TYPE_INITIAL 2
472 #define BLOCK_TYPE_SILENCE 3
474 typedef struct VmdAudioContext {
479 static const uint16_t vmdaudio_table[128] = {
480 0x000, 0x008, 0x010, 0x020, 0x030, 0x040, 0x050, 0x060, 0x070, 0x080,
481 0x090, 0x0A0, 0x0B0, 0x0C0, 0x0D0, 0x0E0, 0x0F0, 0x100, 0x110, 0x120,
482 0x130, 0x140, 0x150, 0x160, 0x170, 0x180, 0x190, 0x1A0, 0x1B0, 0x1C0,
483 0x1D0, 0x1E0, 0x1F0, 0x200, 0x208, 0x210, 0x218, 0x220, 0x228, 0x230,
484 0x238, 0x240, 0x248, 0x250, 0x258, 0x260, 0x268, 0x270, 0x278, 0x280,
485 0x288, 0x290, 0x298, 0x2A0, 0x2A8, 0x2B0, 0x2B8, 0x2C0, 0x2C8, 0x2D0,
486 0x2D8, 0x2E0, 0x2E8, 0x2F0, 0x2F8, 0x300, 0x308, 0x310, 0x318, 0x320,
487 0x328, 0x330, 0x338, 0x340, 0x348, 0x350, 0x358, 0x360, 0x368, 0x370,
488 0x378, 0x380, 0x388, 0x390, 0x398, 0x3A0, 0x3A8, 0x3B0, 0x3B8, 0x3C0,
489 0x3C8, 0x3D0, 0x3D8, 0x3E0, 0x3E8, 0x3F0, 0x3F8, 0x400, 0x440, 0x480,
490 0x4C0, 0x500, 0x540, 0x580, 0x5C0, 0x600, 0x640, 0x680, 0x6C0, 0x700,
491 0x740, 0x780, 0x7C0, 0x800, 0x900, 0xA00, 0xB00, 0xC00, 0xD00, 0xE00,
492 0xF00, 0x1000, 0x1400, 0x1800, 0x1C00, 0x2000, 0x3000, 0x4000
495 static av_cold int vmdaudio_decode_init(AVCodecContext *avctx)
497 VmdAudioContext *s = avctx->priv_data;
499 if (avctx->channels < 1 || avctx->channels > 2) {
500 av_log(avctx, AV_LOG_ERROR, "invalid number of channels\n");
501 return AVERROR(EINVAL);
503 if (avctx->block_align < 1 || avctx->block_align % avctx->channels) {
504 av_log(avctx, AV_LOG_ERROR, "invalid block align\n");
505 return AVERROR(EINVAL);
508 avctx->channel_layout = avctx->channels == 1 ? AV_CH_LAYOUT_MONO :
511 if (avctx->bits_per_coded_sample == 16)
512 avctx->sample_fmt = AV_SAMPLE_FMT_S16;
514 avctx->sample_fmt = AV_SAMPLE_FMT_U8;
515 s->out_bps = av_get_bytes_per_sample(avctx->sample_fmt);
517 s->chunk_size = avctx->block_align + avctx->channels * (s->out_bps == 2);
519 av_log(avctx, AV_LOG_DEBUG, "%d channels, %d bits/sample, "
520 "block align = %d, sample rate = %d\n",
521 avctx->channels, avctx->bits_per_coded_sample, avctx->block_align,
527 static void decode_audio_s16(int16_t *out, const uint8_t *buf, int buf_size,
531 const uint8_t *buf_end = buf + buf_size;
533 int st = channels - 1;
535 /* decode initial raw sample */
536 for (ch = 0; ch < channels; ch++) {
537 predictor[ch] = (int16_t)AV_RL16(buf);
539 *out++ = predictor[ch];
542 /* decode DPCM samples */
544 while (buf < buf_end) {
547 predictor[ch] -= vmdaudio_table[b & 0x7F];
549 predictor[ch] += vmdaudio_table[b];
550 predictor[ch] = av_clip_int16(predictor[ch]);
551 *out++ = predictor[ch];
556 static int vmdaudio_decode_frame(AVCodecContext *avctx, void *data,
557 int *got_frame_ptr, AVPacket *avpkt)
559 AVFrame *frame = data;
560 const uint8_t *buf = avpkt->data;
561 const uint8_t *buf_end;
562 int buf_size = avpkt->size;
563 VmdAudioContext *s = avctx->priv_data;
564 int block_type, silent_chunks, audio_chunks;
566 uint8_t *output_samples_u8;
567 int16_t *output_samples_s16;
570 av_log(avctx, AV_LOG_WARNING, "skipping small junk packet\n");
576 if (block_type < BLOCK_TYPE_AUDIO || block_type > BLOCK_TYPE_SILENCE) {
577 av_log(avctx, AV_LOG_ERROR, "unknown block type: %d\n", block_type);
578 return AVERROR(EINVAL);
583 /* get number of silent chunks */
585 if (block_type == BLOCK_TYPE_INITIAL) {
588 av_log(avctx, AV_LOG_ERROR, "packet is too small\n");
589 return AVERROR(EINVAL);
591 flags = AV_RB32(buf);
592 silent_chunks = av_popcount(flags);
595 } else if (block_type == BLOCK_TYPE_SILENCE) {
597 buf_size = 0; // should already be zero but set it just to be sure
600 /* ensure output buffer is large enough */
601 audio_chunks = buf_size / s->chunk_size;
603 /* drop incomplete chunks */
604 buf_size = audio_chunks * s->chunk_size;
606 /* get output buffer */
607 frame->nb_samples = ((silent_chunks + audio_chunks) * avctx->block_align) /
609 if ((ret = ff_get_buffer(avctx, frame, 0)) < 0)
611 output_samples_u8 = frame->data[0];
612 output_samples_s16 = (int16_t *)frame->data[0];
614 /* decode silent chunks */
615 if (silent_chunks > 0) {
616 int silent_size = avctx->block_align * silent_chunks;
617 av_assert0(avctx->block_align * silent_chunks <= frame->nb_samples * avctx->channels);
619 if (s->out_bps == 2) {
620 memset(output_samples_s16, 0x00, silent_size * 2);
621 output_samples_s16 += silent_size;
623 memset(output_samples_u8, 0x80, silent_size);
624 output_samples_u8 += silent_size;
628 /* decode audio chunks */
629 if (audio_chunks > 0) {
630 buf_end = buf + buf_size;
631 av_assert0((buf_size & (avctx->channels > 1)) == 0);
632 while (buf_end - buf >= s->chunk_size) {
633 if (s->out_bps == 2) {
634 decode_audio_s16(output_samples_s16, buf, s->chunk_size,
636 output_samples_s16 += avctx->block_align;
638 memcpy(output_samples_u8, buf, s->chunk_size);
639 output_samples_u8 += avctx->block_align;
641 buf += s->chunk_size;
652 * Public Data Structures
655 AVCodec ff_vmdvideo_decoder = {
657 .type = AVMEDIA_TYPE_VIDEO,
658 .id = AV_CODEC_ID_VMDVIDEO,
659 .priv_data_size = sizeof(VmdVideoContext),
660 .init = vmdvideo_decode_init,
661 .close = vmdvideo_decode_end,
662 .decode = vmdvideo_decode_frame,
663 .capabilities = CODEC_CAP_DR1,
664 .long_name = NULL_IF_CONFIG_SMALL("Sierra VMD video"),
667 AVCodec ff_vmdaudio_decoder = {
669 .type = AVMEDIA_TYPE_AUDIO,
670 .id = AV_CODEC_ID_VMDAUDIO,
671 .priv_data_size = sizeof(VmdAudioContext),
672 .init = vmdaudio_decode_init,
673 .decode = vmdaudio_decode_frame,
674 .capabilities = CODEC_CAP_DR1,
675 .long_name = NULL_IF_CONFIG_SMALL("Sierra VMD audio"),