1 /*
2 * Copyright (c) 2013 Paul B Mahol
3 *
4 * This file is part of FFmpeg.
5 *
6 * FFmpeg is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Lesser General Public
8 * License as published by the Free Software Foundation; either
9 * version 2.1 of the License, or (at your option) any later version.
10 *
11 * FFmpeg is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * Lesser General Public License for more details.
15 *
16 * You should have received a copy of the GNU Lesser General Public
17 * License along with FFmpeg; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19 */
20
21 #include "libavutil/avstring.h"
22 #include "libavutil/eval.h"
23 #include "libavutil/opt.h"
24 #include "libavutil/samplefmt.h"
25 #include "avfilter.h"
26 #include "audio.h"
27 #include "filters.h"
28 #include "internal.h"
29
30 typedef struct ChanDelay {
31 int64_t delay;
32 size_t delay_index;
33 size_t index;
34 unsigned int samples_size;
35 uint8_t *samples;
36 } ChanDelay;
37
38 typedef struct AudioDelayContext {
39 const AVClass *class;
40 int all;
41 char *delays;
42 ChanDelay *chandelay;
43 int nb_delays;
44 int block_align;
45 int64_t padding;
46 int64_t max_delay;
47 int64_t next_pts;
48 int eof;
49
50 void (*delay_channel)(ChanDelay *d, int nb_samples,
51 const uint8_t *src, uint8_t *dst);
52 int (*resize_channel_samples)(ChanDelay *d, int64_t new_delay);
53 } AudioDelayContext;
54
55 #define OFFSET(x) offsetof(AudioDelayContext, x)
56 #define A AV_OPT_FLAG_AUDIO_PARAM|AV_OPT_FLAG_FILTERING_PARAM
57
58 static const AVOption adelay_options[] = {
59 { "delays", "set list of delays for each channel", OFFSET(delays), AV_OPT_TYPE_STRING, {.str=NULL}, 0, 0, A | AV_OPT_FLAG_RUNTIME_PARAM },
60 { "all", "use last available delay for remained channels", OFFSET(all), AV_OPT_TYPE_BOOL, {.i64=0}, 0, 1, A },
61 { NULL }
62 };
63
64 AVFILTER_DEFINE_CLASS(adelay);
65
66 #define DELAY(name, type, fill) \
67 static void delay_channel_## name ##p(ChanDelay *d, int nb_samples, \
68 const uint8_t *ssrc, uint8_t *ddst) \
69 { \
70 const type *src = (type *)ssrc; \
71 type *dst = (type *)ddst; \
72 type *samples = (type *)d->samples; \
73 \
74 while (nb_samples) { \
75 if (d->delay_index < d->delay) { \
76 const int len = FFMIN(nb_samples, d->delay - d->delay_index); \
77 \
78 memcpy(&samples[d->delay_index], src, len * sizeof(type)); \
79 memset(dst, fill, len * sizeof(type)); \
80 d->delay_index += len; \
81 src += len; \
82 dst += len; \
83 nb_samples -= len; \
84 } else { \
85 *dst = samples[d->index]; \
86 samples[d->index] = *src; \
87 nb_samples--; \
88 d->index++; \
89 src++, dst++; \
90 d->index = d->index >= d->delay ? 0 : d->index; \
91 } \
92 } \
93 }
94
95 DELAY(u8, uint8_t, 0x80)
96 DELAY(s16, int16_t, 0)
97 DELAY(s32, int32_t, 0)
98 DELAY(flt, float, 0)
99 DELAY(dbl, double, 0)
100
101 #define CHANGE_DELAY(name, type, fill) \
102 static int resize_samples_## name ##p(ChanDelay *d, int64_t new_delay) \
103 { \
104 type *samples; \
105 \
106 if (new_delay == d->delay) { \
107 return 0; \
108 } \
109 \
110 if (new_delay == 0) { \
111 av_freep(&d->samples); \
112 d->samples_size = 0; \
113 d->delay = 0; \
114 d->index = 0; \
115 d->delay_index = 0; \
116 return 0; \
117 } \
118 \
119 samples = (type *) av_fast_realloc(d->samples, &d->samples_size, new_delay * sizeof(type)); \
120 if (!samples) { \
121 return AVERROR(ENOMEM); \
122 } \
123 \
124 if (new_delay < d->delay) { \
125 if (d->index > new_delay) { \
126 d->index -= new_delay; \
127 memmove(samples, &samples[new_delay], d->index * sizeof(type)); \
128 d->delay_index = new_delay; \
129 } else if (d->delay_index > d->index) { \
130 memmove(&samples[d->index], &samples[d->index+(d->delay-new_delay)], \
131 (new_delay - d->index) * sizeof(type)); \
132 d->delay_index -= d->delay - new_delay; \
133 } \
134 } else { \
135 size_t block_size; \
136 if (d->delay_index >= d->delay) { \
137 block_size = (d->delay - d->index) * sizeof(type); \
138 memmove(&samples[d->index+(new_delay - d->delay)], &samples[d->index], block_size); \
139 d->delay_index = new_delay; \
140 } else { \
141 d->delay_index += new_delay - d->delay; \
142 } \
143 block_size = (new_delay - d->delay) * sizeof(type); \
144 memset(&samples[d->index], fill, block_size); \
145 } \
146 d->delay = new_delay; \
147 d->samples = (void *) samples; \
148 return 0; \
149 }
150
151 CHANGE_DELAY(u8, uint8_t, 0x80)
152 CHANGE_DELAY(s16, int16_t, 0)
153 CHANGE_DELAY(s32, int32_t, 0)
154 CHANGE_DELAY(flt, float, 0)
155 CHANGE_DELAY(dbl, double, 0)
156
parse_delays(char * p,char ** saveptr,int64_t * result,AVFilterContext * ctx,int sample_rate)157 static int parse_delays(char *p, char **saveptr, int64_t *result, AVFilterContext *ctx, int sample_rate) {
158 float delay, div;
159 int ret;
160 char *arg;
161 char type = 0;
162
163 if (!(arg = av_strtok(p, "|", saveptr)))
164 return 1;
165
166 ret = av_sscanf(arg, "%"SCNd64"%c", result, &type);
167 if (ret != 2 || type != 'S') {
168 div = type == 's' ? 1.0 : 1000.0;
169 if (av_sscanf(arg, "%f", &delay) != 1) {
170 av_log(ctx, AV_LOG_ERROR, "Invalid syntax for delay.\n");
171 return AVERROR(EINVAL);
172 }
173 *result = delay * sample_rate / div;
174 }
175
176 if (*result < 0) {
177 av_log(ctx, AV_LOG_ERROR, "Delay must be non negative number.\n");
178 return AVERROR(EINVAL);
179 }
180 return 0;
181 }
182
config_input(AVFilterLink * inlink)183 static int config_input(AVFilterLink *inlink)
184 {
185 AVFilterContext *ctx = inlink->dst;
186 AudioDelayContext *s = ctx->priv;
187 char *p, *saveptr = NULL;
188 int i;
189
190 s->chandelay = av_calloc(inlink->ch_layout.nb_channels, sizeof(*s->chandelay));
191 if (!s->chandelay)
192 return AVERROR(ENOMEM);
193 s->nb_delays = inlink->ch_layout.nb_channels;
194 s->block_align = av_get_bytes_per_sample(inlink->format);
195
196 p = s->delays;
197 for (i = 0; i < s->nb_delays; i++) {
198 ChanDelay *d = &s->chandelay[i];
199 int ret;
200
201 ret = parse_delays(p, &saveptr, &d->delay, ctx, inlink->sample_rate);
202 if (ret == 1)
203 break;
204 else if (ret < 0)
205 return ret;
206 p = NULL;
207 }
208
209 if (s->all && i) {
210 for (int j = i; j < s->nb_delays; j++)
211 s->chandelay[j].delay = s->chandelay[i-1].delay;
212 }
213
214 s->padding = s->chandelay[0].delay;
215 for (i = 1; i < s->nb_delays; i++) {
216 ChanDelay *d = &s->chandelay[i];
217
218 s->padding = FFMIN(s->padding, d->delay);
219 }
220
221 if (s->padding) {
222 for (i = 0; i < s->nb_delays; i++) {
223 ChanDelay *d = &s->chandelay[i];
224
225 d->delay -= s->padding;
226 }
227 }
228
229 for (i = 0; i < s->nb_delays; i++) {
230 ChanDelay *d = &s->chandelay[i];
231
232 if (!d->delay)
233 continue;
234
235 if (d->delay > SIZE_MAX) {
236 av_log(ctx, AV_LOG_ERROR, "Requested delay is too big.\n");
237 return AVERROR(EINVAL);
238 }
239
240 d->samples = av_malloc_array(d->delay, s->block_align);
241 if (!d->samples)
242 return AVERROR(ENOMEM);
243 d->samples_size = d->delay * s->block_align;
244
245 s->max_delay = FFMAX(s->max_delay, d->delay);
246 }
247
248 switch (inlink->format) {
249 case AV_SAMPLE_FMT_U8P : s->delay_channel = delay_channel_u8p ;
250 s->resize_channel_samples = resize_samples_u8p; break;
251 case AV_SAMPLE_FMT_S16P: s->delay_channel = delay_channel_s16p;
252 s->resize_channel_samples = resize_samples_s16p; break;
253 case AV_SAMPLE_FMT_S32P: s->delay_channel = delay_channel_s32p;
254 s->resize_channel_samples = resize_samples_s32p; break;
255 case AV_SAMPLE_FMT_FLTP: s->delay_channel = delay_channel_fltp;
256 s->resize_channel_samples = resize_samples_fltp; break;
257 case AV_SAMPLE_FMT_DBLP: s->delay_channel = delay_channel_dblp;
258 s->resize_channel_samples = resize_samples_dblp; break;
259 }
260
261 return 0;
262 }
263
process_command(AVFilterContext * ctx,const char * cmd,const char * args,char * res,int res_len,int flags)264 static int process_command(AVFilterContext *ctx, const char *cmd, const char *args,
265 char *res, int res_len, int flags)
266 {
267 int ret = AVERROR(ENOSYS);
268 AVFilterLink *inlink = ctx->inputs[0];
269 AudioDelayContext *s = ctx->priv;
270
271 if (!strcmp(cmd, "delays")) {
272 int64_t delay;
273 char *p, *saveptr = NULL;
274 int64_t all_delay = -1;
275 int64_t max_delay = 0;
276 char *args_cpy = av_strdup(args);
277 if (args_cpy == NULL) {
278 return AVERROR(ENOMEM);
279 }
280
281 ret = 0;
282 p = args_cpy;
283
284 if (!strncmp(args, "all:", 4)) {
285 p = &args_cpy[4];
286 ret = parse_delays(p, &saveptr, &all_delay, ctx, inlink->sample_rate);
287 if (ret == 1)
288 ret = AVERROR(EINVAL);
289 else if (ret == 0)
290 delay = all_delay;
291 }
292
293 if (!ret) {
294 for (int i = 0; i < s->nb_delays; i++) {
295 ChanDelay *d = &s->chandelay[i];
296
297 if (all_delay < 0) {
298 ret = parse_delays(p, &saveptr, &delay, ctx, inlink->sample_rate);
299 if (ret != 0) {
300 ret = 0;
301 break;
302 }
303 p = NULL;
304 }
305
306 ret = s->resize_channel_samples(d, delay);
307 if (ret)
308 break;
309 max_delay = FFMAX(max_delay, d->delay);
310 }
311 s->max_delay = FFMAX(s->max_delay, max_delay);
312 }
313 av_freep(&args_cpy);
314 }
315 return ret;
316 }
317
filter_frame(AVFilterLink * inlink,AVFrame * frame)318 static int filter_frame(AVFilterLink *inlink, AVFrame *frame)
319 {
320 AVFilterContext *ctx = inlink->dst;
321 AVFilterLink *outlink = ctx->outputs[0];
322 AudioDelayContext *s = ctx->priv;
323 AVFrame *out_frame;
324 int i;
325
326 if (ctx->is_disabled || !s->delays)
327 return ff_filter_frame(outlink, frame);
328
329 out_frame = ff_get_audio_buffer(outlink, frame->nb_samples);
330 if (!out_frame) {
331 av_frame_free(&frame);
332 return AVERROR(ENOMEM);
333 }
334 av_frame_copy_props(out_frame, frame);
335
336 for (i = 0; i < s->nb_delays; i++) {
337 ChanDelay *d = &s->chandelay[i];
338 const uint8_t *src = frame->extended_data[i];
339 uint8_t *dst = out_frame->extended_data[i];
340
341 if (!d->delay)
342 memcpy(dst, src, frame->nb_samples * s->block_align);
343 else
344 s->delay_channel(d, frame->nb_samples, src, dst);
345 }
346
347 out_frame->pts = s->next_pts;
348 s->next_pts += av_rescale_q(frame->nb_samples, (AVRational){1, outlink->sample_rate}, outlink->time_base);
349 av_frame_free(&frame);
350 return ff_filter_frame(outlink, out_frame);
351 }
352
activate(AVFilterContext * ctx)353 static int activate(AVFilterContext *ctx)
354 {
355 AVFilterLink *inlink = ctx->inputs[0];
356 AVFilterLink *outlink = ctx->outputs[0];
357 AudioDelayContext *s = ctx->priv;
358 AVFrame *frame = NULL;
359 int ret, status;
360 int64_t pts;
361
362 FF_FILTER_FORWARD_STATUS_BACK(outlink, inlink);
363
364 if (s->padding) {
365 int nb_samples = FFMIN(s->padding, 2048);
366
367 frame = ff_get_audio_buffer(outlink, nb_samples);
368 if (!frame)
369 return AVERROR(ENOMEM);
370 s->padding -= nb_samples;
371
372 av_samples_set_silence(frame->extended_data, 0,
373 frame->nb_samples,
374 outlink->ch_layout.nb_channels,
375 frame->format);
376
377 frame->pts = s->next_pts;
378 if (s->next_pts != AV_NOPTS_VALUE)
379 s->next_pts += av_rescale_q(nb_samples, (AVRational){1, outlink->sample_rate}, outlink->time_base);
380
381 return ff_filter_frame(outlink, frame);
382 }
383
384 ret = ff_inlink_consume_frame(inlink, &frame);
385 if (ret < 0)
386 return ret;
387
388 if (ret > 0)
389 return filter_frame(inlink, frame);
390
391 if (ff_inlink_acknowledge_status(inlink, &status, &pts)) {
392 if (status == AVERROR_EOF)
393 s->eof = 1;
394 }
395
396 if (s->eof && s->max_delay) {
397 int nb_samples = FFMIN(s->max_delay, 2048);
398
399 frame = ff_get_audio_buffer(outlink, nb_samples);
400 if (!frame)
401 return AVERROR(ENOMEM);
402 s->max_delay -= nb_samples;
403
404 av_samples_set_silence(frame->extended_data, 0,
405 frame->nb_samples,
406 outlink->ch_layout.nb_channels,
407 frame->format);
408
409 frame->pts = s->next_pts;
410 return filter_frame(inlink, frame);
411 }
412
413 if (s->eof && s->max_delay == 0) {
414 ff_outlink_set_status(outlink, AVERROR_EOF, s->next_pts);
415 return 0;
416 }
417
418 if (!s->eof)
419 FF_FILTER_FORWARD_WANTED(outlink, inlink);
420
421 return FFERROR_NOT_READY;
422 }
423
uninit(AVFilterContext * ctx)424 static av_cold void uninit(AVFilterContext *ctx)
425 {
426 AudioDelayContext *s = ctx->priv;
427
428 if (s->chandelay) {
429 for (int i = 0; i < s->nb_delays; i++)
430 av_freep(&s->chandelay[i].samples);
431 }
432 av_freep(&s->chandelay);
433 }
434
435 static const AVFilterPad adelay_inputs[] = {
436 {
437 .name = "default",
438 .type = AVMEDIA_TYPE_AUDIO,
439 .config_props = config_input,
440 },
441 };
442
443 static const AVFilterPad adelay_outputs[] = {
444 {
445 .name = "default",
446 .type = AVMEDIA_TYPE_AUDIO,
447 },
448 };
449
450 const AVFilter ff_af_adelay = {
451 .name = "adelay",
452 .description = NULL_IF_CONFIG_SMALL("Delay one or more audio channels."),
453 .priv_size = sizeof(AudioDelayContext),
454 .priv_class = &adelay_class,
455 .activate = activate,
456 .uninit = uninit,
457 FILTER_INPUTS(adelay_inputs),
458 FILTER_OUTPUTS(adelay_outputs),
459 FILTER_SAMPLEFMTS(AV_SAMPLE_FMT_U8P, AV_SAMPLE_FMT_S16P, AV_SAMPLE_FMT_S32P,
460 AV_SAMPLE_FMT_FLTP, AV_SAMPLE_FMT_DBLP),
461 .flags = AVFILTER_FLAG_SUPPORT_TIMELINE_INTERNAL,
462 .process_command = process_command,
463 };
464