get rid of the last svn mentions
[libav.git] / libavformat / dvenc.c
CommitLineData
d4fc91b4
PI
1/*
2 * General DV muxer/demuxer
3 * Copyright (c) 2003 Roman Shaposhnik
4 *
5 * Many thanks to Dan Dennedy <dan@dennedy.org> for providing wealth
6 * of DV technical info.
7 *
8 * Raw DV format
406792e7 9 * Copyright (c) 2002 Fabrice Bellard
d4fc91b4
PI
10 *
11 * 50 Mbps (DVCPRO50) support
12 * Copyright (c) 2006 Daniel Maas <dmaas@maasdigital.com>
13 *
b78e7197
DB
14 * This file is part of FFmpeg.
15 *
16 * FFmpeg is free software; you can redistribute it and/or
d4fc91b4
PI
17 * modify it under the terms of the GNU Lesser General Public
18 * License as published by the Free Software Foundation; either
b78e7197 19 * version 2.1 of the License, or (at your option) any later version.
d4fc91b4 20 *
b78e7197 21 * FFmpeg is distributed in the hope that it will be useful,
d4fc91b4
PI
22 * but WITHOUT ANY WARRANTY; without even the implied warranty of
23 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
24 * Lesser General Public License for more details.
25 *
26 * You should have received a copy of the GNU Lesser General Public
b78e7197 27 * License along with FFmpeg; if not, write to the Free Software
d4fc91b4
PI
28 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
29 */
30#include <time.h>
e950141a 31#include <stdarg.h>
e1ae1642 32
d4fc91b4 33#include "avformat.h"
e1ae1642 34#include "internal.h"
245976da 35#include "libavcodec/dvdata.h"
d4fc91b4 36#include "dv.h"
245976da 37#include "libavutil/fifo.h"
d4fc91b4
PI
38
39struct DVMuxContext {
75f2c209
DB
40 const DVprofile* sys; /* current DV profile, e.g.: 525/60, 625/50 */
41 int n_ast; /* number of stereo audio streams (up to 2) */
42 AVStream *ast[2]; /* stereo audio streams */
41dd680d 43 AVFifoBuffer *audio_data[2]; /* FIFO for storing excessive amounts of PCM */
75f2c209
DB
44 int frames; /* current frame number */
45 time_t start_time; /* recording start time */
f94036f1
DB
46 int has_audio; /* frame under contruction has audio */
47 int has_video; /* frame under contruction has video */
48 uint8_t frame_buf[DV_MAX_FRAME_SIZE]; /* frame under contruction */
d4fc91b4
PI
49};
50
51static const int dv_aaux_packs_dist[12][9] = {
52 { 0xff, 0xff, 0xff, 0x50, 0x51, 0x52, 0x53, 0xff, 0xff },
53 { 0x50, 0x51, 0x52, 0x53, 0xff, 0xff, 0xff, 0xff, 0xff },
54 { 0xff, 0xff, 0xff, 0x50, 0x51, 0x52, 0x53, 0xff, 0xff },
55 { 0x50, 0x51, 0x52, 0x53, 0xff, 0xff, 0xff, 0xff, 0xff },
56 { 0xff, 0xff, 0xff, 0x50, 0x51, 0x52, 0x53, 0xff, 0xff },
57 { 0x50, 0x51, 0x52, 0x53, 0xff, 0xff, 0xff, 0xff, 0xff },
58 { 0xff, 0xff, 0xff, 0x50, 0x51, 0x52, 0x53, 0xff, 0xff },
59 { 0x50, 0x51, 0x52, 0x53, 0xff, 0xff, 0xff, 0xff, 0xff },
60 { 0xff, 0xff, 0xff, 0x50, 0x51, 0x52, 0x53, 0xff, 0xff },
61 { 0x50, 0x51, 0x52, 0x53, 0xff, 0xff, 0xff, 0xff, 0xff },
62 { 0xff, 0xff, 0xff, 0x50, 0x51, 0x52, 0x53, 0xff, 0xff },
63 { 0x50, 0x51, 0x52, 0x53, 0xff, 0xff, 0xff, 0xff, 0xff },
64};
65
66static int dv_audio_frame_size(const DVprofile* sys, int frame)
67{
f94036f1 68 return sys->audio_samples_dist[frame % (sizeof(sys->audio_samples_dist) /
d4fc91b4
PI
69 sizeof(sys->audio_samples_dist[0]))];
70}
71
e950141a 72static int dv_write_pack(enum dv_pack_type pack_id, DVMuxContext *c, uint8_t* buf, ...)
d4fc91b4
PI
73{
74 struct tm tc;
75 time_t ct;
76 int ltc_frame;
e950141a 77 va_list ap;
d4fc91b4 78
d4fc91b4
PI
79 buf[0] = (uint8_t)pack_id;
80 switch (pack_id) {
81 case dv_timecode:
f94036f1
DB
82 ct = (time_t)av_rescale_rnd(c->frames, c->sys->time_base.num,
83 c->sys->time_base.den, AV_ROUND_DOWN);
bc60f09b
RD
84 brktimegm(ct, &tc);
85 /*
86 * LTC drop-frame frame counter drops two frames (0 and 1) every
87 * minute, unless it is exactly divisible by 10
88 */
f94036f1 89 ltc_frame = (c->frames + 2 * ct / 60 - 2 * ct / 600) % c->sys->ltc_divisor;
75f2c209
DB
90 buf[1] = (0 << 7) | /* color frame: 0 - unsync; 1 - sync mode */
91 (1 << 6) | /* drop frame timecode: 0 - nondrop; 1 - drop */
92 ((ltc_frame / 10) << 4) | /* tens of frames */
93 (ltc_frame % 10); /* units of frames */
94 buf[2] = (1 << 7) | /* biphase mark polarity correction: 0 - even; 1 - odd */
95 ((tc.tm_sec / 10) << 4) | /* tens of seconds */
96 (tc.tm_sec % 10); /* units of seconds */
97 buf[3] = (1 << 7) | /* binary group flag BGF0 */
98 ((tc.tm_min / 10) << 4) | /* tens of minutes */
99 (tc.tm_min % 10); /* units of minutes */
100 buf[4] = (1 << 7) | /* binary group flag BGF2 */
101 (1 << 6) | /* binary group flag BGF1 */
102 ((tc.tm_hour / 10) << 4) | /* tens of hours */
103 (tc.tm_hour % 10); /* units of hours */
bc60f09b 104 break;
d4fc91b4 105 case dv_audio_source: /* AAUX source pack */
bc60f09b
RD
106 va_start(ap, buf);
107 buf[1] = (1 << 7) | /* locked mode -- SMPTE only supports locked mode */
108 (1 << 6) | /* reserved -- always 1 */
109 (dv_audio_frame_size(c->sys, c->frames) -
110 c->sys->audio_min_samples[0]);
111 /* # of samples */
112 buf[2] = (0 << 7) | /* multi-stereo */
113 (0 << 5) | /* #of audio channels per block: 0 -- 1 channel */
114 (0 << 4) | /* pair bit: 0 -- one pair of channels */
115 !!va_arg(ap, int); /* audio mode */
116 buf[3] = (1 << 7) | /* res */
117 (1 << 6) | /* multi-language flag */
118 (c->sys->dsf << 5) | /* system: 60fields/50fields */
119 (c->sys->n_difchan & 2); /* definition: 0 -- 25Mbps, 2 -- 50Mbps */
120 buf[4] = (1 << 7) | /* emphasis: 1 -- off */
121 (0 << 6) | /* emphasis time constant: 0 -- reserved */
910f02a0 122 (0 << 3) | /* frequency: 0 -- 48kHz, 1 -- 44,1kHz, 2 -- 32kHz */
bc60f09b
RD
123 0; /* quantization: 0 -- 16bit linear, 1 -- 12bit nonlinear */
124 va_end(ap);
125 break;
d4fc91b4 126 case dv_audio_control:
bc60f09b
RD
127 buf[1] = (0 << 6) | /* copy protection: 0 -- unrestricted */
128 (1 << 4) | /* input source: 1 -- digital input */
129 (3 << 2) | /* compression: 3 -- no information */
130 0; /* misc. info/SMPTE emphasis off */
131 buf[2] = (1 << 7) | /* recording start point: 1 -- no */
132 (1 << 6) | /* recording end point: 1 -- no */
133 (1 << 3) | /* recording mode: 1 -- original */
134 7;
135 buf[3] = (1 << 7) | /* direction: 1 -- forward */
136 (c->sys->pix_fmt == PIX_FMT_YUV420P ? 0x20 : /* speed */
f94036f1 137 c->sys->ltc_divisor * 4);
bc60f09b
RD
138 buf[4] = (1 << 7) | /* reserved -- always 1 */
139 0x7f; /* genre category */
140 break;
d4fc91b4
PI
141 case dv_audio_recdate:
142 case dv_video_recdate: /* VAUX recording date */
4e92dabf 143 ct = c->start_time + av_rescale_rnd(c->frames, c->sys->time_base.num,
f94036f1 144 c->sys->time_base.den, AV_ROUND_DOWN);
bc60f09b
RD
145 brktimegm(ct, &tc);
146 buf[1] = 0xff; /* ds, tm, tens of time zone, units of time zone */
147 /* 0xff is very likely to be "unknown" */
148 buf[2] = (3 << 6) | /* reserved -- always 1 */
149 ((tc.tm_mday / 10) << 4) | /* Tens of day */
150 (tc.tm_mday % 10); /* Units of day */
151 buf[3] = /* we set high 4 bits to 0, shouldn't we set them to week? */
152 ((tc.tm_mon / 10) << 4) | /* Tens of month */
153 (tc.tm_mon % 10); /* Units of month */
154 buf[4] = (((tc.tm_year % 100) / 10) << 4) | /* Tens of year */
155 (tc.tm_year % 10); /* Units of year */
156 break;
d4fc91b4
PI
157 case dv_audio_rectime: /* AAUX recording time */
158 case dv_video_rectime: /* VAUX recording time */
4e92dabf
RS
159 ct = c->start_time + av_rescale_rnd(c->frames, c->sys->time_base.num,
160 c->sys->time_base.den, AV_ROUND_DOWN);
bc60f09b
RD
161 brktimegm(ct, &tc);
162 buf[1] = (3 << 6) | /* reserved -- always 1 */
163 0x3f; /* tens of frame, units of frame: 0x3f - "unknown" ? */
164 buf[2] = (1 << 7) | /* reserved -- always 1 */
165 ((tc.tm_sec / 10) << 4) | /* Tens of seconds */
166 (tc.tm_sec % 10); /* Units of seconds */
167 buf[3] = (1 << 7) | /* reserved -- always 1 */
168 ((tc.tm_min / 10) << 4) | /* Tens of minutes */
169 (tc.tm_min % 10); /* Units of minutes */
170 buf[4] = (3 << 6) | /* reserved -- always 1 */
171 ((tc.tm_hour / 10) << 4) | /* Tens of hours */
172 (tc.tm_hour % 10); /* Units of hours */
173 break;
d4fc91b4 174 default:
bc60f09b 175 buf[1] = buf[2] = buf[3] = buf[4] = 0xff;
d4fc91b4
PI
176 }
177 return 5;
178}
179
180static void dv_inject_audio(DVMuxContext *c, int channel, uint8_t* frame_ptr)
181{
182 int i, j, d, of, size;
183 size = 4 * dv_audio_frame_size(c->sys, c->frames);
184 frame_ptr += channel * c->sys->difseg_size * 150 * 80;
185 for (i = 0; i < c->sys->difseg_size; i++) {
bc60f09b
RD
186 frame_ptr += 6 * 80; /* skip DIF segment header */
187 for (j = 0; j < 9; j++) {
188 dv_write_pack(dv_aaux_packs_dist[i][j], c, &frame_ptr[3], i >= c->sys->difseg_size/2);
189 for (d = 8; d < 80; d+=2) {
190 of = c->sys->audio_shuffle[i][j] + (d - 8)/2 * c->sys->audio_stride;
191 if (of*2 >= size)
192 continue;
d4fc91b4 193
41dd680d
MN
194 frame_ptr[d] = av_fifo_peek(c->audio_data[channel], of*2+1); // FIXME: maybe we have to admit
195 frame_ptr[d+1] = av_fifo_peek(c->audio_data[channel], of*2); // that DV is a big-endian PCM
bc60f09b
RD
196 }
197 frame_ptr += 16 * 80; /* 15 Video DIFs + 1 Audio DIF */
198 }
d4fc91b4
PI
199 }
200}
201
202static void dv_inject_metadata(DVMuxContext *c, uint8_t* frame)
203{
204 int j, k;
205 uint8_t* buf;
206
207 for (buf = frame; buf < frame + c->sys->frame_size; buf += 150 * 80) {
208 /* DV subcode: 2nd and 3d DIFs */
209 for (j = 80; j < 80 * 3; j += 80) {
210 for (k = 6; k < 6 * 8; k += 8)
211 dv_write_pack(dv_timecode, c, &buf[j+k]);
212
213 if (((long)(buf-frame)/(c->sys->frame_size/(c->sys->difseg_size*c->sys->n_difchan))%c->sys->difseg_size) > 5) { /* FIXME: is this really needed ? */
214 dv_write_pack(dv_video_recdate, c, &buf[j+14]);
215 dv_write_pack(dv_video_rectime, c, &buf[j+22]);
216 dv_write_pack(dv_video_recdate, c, &buf[j+38]);
217 dv_write_pack(dv_video_rectime, c, &buf[j+46]);
218 }
219 }
220
221 /* DV VAUX: 4th, 5th and 6th 3DIFs */
222 for (j = 80*3 + 3; j < 80*6; j += 80) {
223 dv_write_pack(dv_video_recdate, c, &buf[j+5*2]);
224 dv_write_pack(dv_video_rectime, c, &buf[j+5*3]);
225 dv_write_pack(dv_video_recdate, c, &buf[j+5*11]);
226 dv_write_pack(dv_video_rectime, c, &buf[j+5*12]);
227 }
228 }
229}
230
231/*
232 * The following 3 functions constitute our interface to the world
233 */
234
f0a86769
DEP
235static int dv_assemble_frame(DVMuxContext *c, AVStream* st,
236 uint8_t* data, int data_size, uint8_t** frame)
d4fc91b4
PI
237{
238 int i, reqasize;
239
240 *frame = &c->frame_buf[0];
241 reqasize = 4 * dv_audio_frame_size(c->sys, c->frames);
242
243 switch (st->codec->codec_type) {
72415b2a 244 case AVMEDIA_TYPE_VIDEO:
bc60f09b
RD
245 /* FIXME: we have to have more sensible approach than this one */
246 if (c->has_video)
247 av_log(st->codec, AV_LOG_ERROR, "Can't process DV frame #%d. Insufficient audio data or severe sync problem.\n", c->frames);
d4fc91b4 248
bc60f09b
RD
249 memcpy(*frame, data, c->sys->frame_size);
250 c->has_video = 1;
251 break;
72415b2a 252 case AVMEDIA_TYPE_AUDIO:
bc60f09b 253 for (i = 0; i < c->n_ast && st != c->ast[i]; i++);
d4fc91b4
PI
254
255 /* FIXME: we have to have more sensible approach than this one */
41dd680d 256 if (av_fifo_size(c->audio_data[i]) + data_size >= 100*AVCODEC_MAX_AUDIO_FRAME_SIZE)
bc60f09b 257 av_log(st->codec, AV_LOG_ERROR, "Can't process DV frame #%d. Insufficient video data or severe sync problem.\n", c->frames);
41dd680d 258 av_fifo_generic_write(c->audio_data[i], data, data_size, NULL);
d4fc91b4 259
ca74c0a1 260 /* Let us see if we've got enough audio for one DV frame. */
41dd680d 261 c->has_audio |= ((reqasize <= av_fifo_size(c->audio_data[i])) << i);
d4fc91b4 262
bc60f09b 263 break;
d4fc91b4 264 default:
bc60f09b 265 break;
d4fc91b4
PI
266 }
267
ca74c0a1 268 /* Let us see if we have enough data to construct one DV frame. */
f94036f1 269 if (c->has_video == 1 && c->has_audio + 1 == 1 << c->n_ast) {
d4fc91b4 270 dv_inject_metadata(c, *frame);
2e374f12 271 c->has_audio = 0;
f94036f1 272 for (i=0; i < c->n_ast; i++) {
bc60f09b 273 dv_inject_audio(c, i, *frame);
41dd680d
MN
274 av_fifo_drain(c->audio_data[i], reqasize);
275 c->has_audio |= ((reqasize <= av_fifo_size(c->audio_data[i])) << i);
d4fc91b4
PI
276 }
277
278 c->has_video = 0;
2e374f12 279
d4fc91b4
PI
280 c->frames++;
281
282 return c->sys->frame_size;
283 }
284
285 return 0;
286}
287
f0a86769 288static DVMuxContext* dv_init_mux(AVFormatContext* s)
d4fc91b4 289{
e4141433 290 DVMuxContext *c = s->priv_data;
d4fc91b4
PI
291 AVStream *vst = NULL;
292 int i;
293
294 /* we support at most 1 video and 2 audio streams */
295 if (s->nb_streams > 3)
296 return NULL;
297
f94036f1 298 c->n_ast = 0;
d4fc91b4
PI
299 c->ast[0] = c->ast[1] = NULL;
300
301 /* We have to sort out where audio and where video stream is */
302 for (i=0; i<s->nb_streams; i++) {
bc60f09b 303 switch (s->streams[i]->codec->codec_type) {
72415b2a 304 case AVMEDIA_TYPE_VIDEO:
bc60f09b
RD
305 if (vst) return NULL;
306 vst = s->streams[i];
307 break;
72415b2a 308 case AVMEDIA_TYPE_AUDIO:
bc60f09b
RD
309 if (c->n_ast > 1) return NULL;
310 c->ast[c->n_ast++] = s->streams[i];
311 break;
312 default:
313 goto bail_out;
314 }
d4fc91b4
PI
315 }
316
317 /* Some checks -- DV format is very picky about its incoming streams */
318 if (!vst || vst->codec->codec_id != CODEC_ID_DVVIDEO)
319 goto bail_out;
320 for (i=0; i<c->n_ast; i++) {
f94036f1 321 if (c->ast[i] && (c->ast[i]->codec->codec_id != CODEC_ID_PCM_S16LE ||
d4fc91b4 322 c->ast[i]->codec->sample_rate != 48000 ||
f94036f1 323 c->ast[i]->codec->channels != 2))
d4fc91b4
PI
324 goto bail_out;
325 }
c76911bd 326 c->sys = ff_dv_codec_profile(vst->codec);
d4fc91b4
PI
327 if (!c->sys)
328 goto bail_out;
329
f94036f1 330 if ((c->n_ast > 1) && (c->sys->n_difchan < 2)) {
d4fc91b4
PI
331 /* only 1 stereo pair is allowed in 25Mbps mode */
332 goto bail_out;
333 }
334
335 /* Ok, everything seems to be in working order */
f94036f1
DB
336 c->frames = 0;
337 c->has_audio = 0;
338 c->has_video = 0;
d4fc91b4
PI
339 c->start_time = (time_t)s->timestamp;
340
f94036f1 341 for (i=0; i < c->n_ast; i++) {
41dd680d 342 if (c->ast[i] && !(c->audio_data[i]=av_fifo_alloc(100*AVCODEC_MAX_AUDIO_FRAME_SIZE))) {
f94036f1 343 while (i > 0) {
d4fc91b4 344 i--;
41dd680d 345 av_fifo_free(c->audio_data[i]);
d4fc91b4
PI
346 }
347 goto bail_out;
348 }
349 }
350
351 return c;
352
353bail_out:
d4fc91b4
PI
354 return NULL;
355}
356
f0a86769 357static void dv_delete_mux(DVMuxContext *c)
d4fc91b4
PI
358{
359 int i;
360 for (i=0; i < c->n_ast; i++)
41dd680d 361 av_fifo_free(c->audio_data[i]);
d4fc91b4
PI
362}
363
d4fc91b4
PI
364static int dv_write_header(AVFormatContext *s)
365{
0008afc5 366 if (!dv_init_mux(s)) {
d4fc91b4
PI
367 av_log(s, AV_LOG_ERROR, "Can't initialize DV format!\n"
368 "Make sure that you supply exactly two streams:\n"
910f02a0 369 " video: 25fps or 29.97fps, audio: 2ch/48kHz/PCM\n"
d4fc91b4
PI
370 " (50Mbps allows an optional second audio stream)\n");
371 return -1;
372 }
373 return 0;
374}
375
376static int dv_write_packet(struct AVFormatContext *s, AVPacket *pkt)
377{
378 uint8_t* frame;
379 int fsize;
380
e4141433 381 fsize = dv_assemble_frame(s->priv_data, s->streams[pkt->stream_index],
d4fc91b4
PI
382 pkt->data, pkt->size, &frame);
383 if (fsize > 0) {
77eb5504 384 avio_write(s->pb, frame, fsize);
b7f2fdde 385 avio_flush(s->pb);
d4fc91b4
PI
386 }
387 return 0;
388}
389
390/*
391 * We might end up with some extra A/V data without matching counterpart.
392 * E.g. video data without enough audio to write the complete frame.
393 * Currently we simply drop the last frame. I don't know whether this
394 * is the best strategy of all
395 */
396static int dv_write_trailer(struct AVFormatContext *s)
397{
e4141433 398 dv_delete_mux(s->priv_data);
d4fc91b4
PI
399 return 0;
400}
d4fc91b4 401
c6610a21 402AVOutputFormat ff_dv_muxer = {
d4fc91b4 403 "dv",
bde15e74 404 NULL_IF_CONFIG_SMALL("DV video format"),
d4fc91b4
PI
405 NULL,
406 "dv",
407 sizeof(DVMuxContext),
408 CODEC_ID_PCM_S16LE,
409 CODEC_ID_DVVIDEO,
410 dv_write_header,
411 dv_write_packet,
412 dv_write_trailer,
413};