1 /*
2 * Copyright (c) 2011 Nicolas George <nicolas.george@normalesup.org>
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
14 * GNU 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 /**
22 * @file
23 * Audio merging filter
24 */
25
26 #include "libavutil/avstring.h"
27 #include "libavutil/bprint.h"
28 #include "libavutil/channel_layout.h"
29 #include "libavutil/opt.h"
30 #include "avfilter.h"
31 #include "filters.h"
32 #include "audio.h"
33 #include "internal.h"
34
35 #define SWR_CH_MAX 64
36
37 typedef struct AMergeContext {
38 const AVClass *class;
39 int nb_inputs;
40 int route[SWR_CH_MAX]; /**< channels routing, see copy_samples */
41 int bps;
42 struct amerge_input {
43 int nb_ch; /**< number of channels for the input */
44 } *in;
45 } AMergeContext;
46
47 #define OFFSET(x) offsetof(AMergeContext, x)
48 #define FLAGS AV_OPT_FLAG_AUDIO_PARAM|AV_OPT_FLAG_FILTERING_PARAM
49
50 static const AVOption amerge_options[] = {
51 { "inputs", "specify the number of inputs", OFFSET(nb_inputs),
52 AV_OPT_TYPE_INT, { .i64 = 2 }, 1, SWR_CH_MAX, FLAGS },
53 { NULL }
54 };
55
56 AVFILTER_DEFINE_CLASS(amerge);
57
uninit(AVFilterContext * ctx)58 static av_cold void uninit(AVFilterContext *ctx)
59 {
60 AMergeContext *s = ctx->priv;
61 int i;
62
63 for (i = 0; i < s->nb_inputs; i++) {
64 if (ctx->input_pads)
65 av_freep(&ctx->input_pads[i].name);
66 }
67 av_freep(&s->in);
68 }
69
query_formats(AVFilterContext * ctx)70 static int query_formats(AVFilterContext *ctx)
71 {
72 AMergeContext *s = ctx->priv;
73 int64_t inlayout[SWR_CH_MAX], outlayout = 0;
74 AVFilterFormats *formats;
75 AVFilterChannelLayouts *layouts;
76 int i, ret, overlap = 0, nb_ch = 0;
77
78 for (i = 0; i < s->nb_inputs; i++) {
79 if (!ctx->inputs[i]->in_channel_layouts ||
80 !ctx->inputs[i]->in_channel_layouts->nb_channel_layouts) {
81 av_log(ctx, AV_LOG_WARNING,
82 "No channel layout for input %d\n", i + 1);
83 return AVERROR(EAGAIN);
84 }
85 inlayout[i] = ctx->inputs[i]->in_channel_layouts->channel_layouts[0];
86 if (ctx->inputs[i]->in_channel_layouts->nb_channel_layouts > 1) {
87 char buf[256];
88 av_get_channel_layout_string(buf, sizeof(buf), 0, inlayout[i]);
89 av_log(ctx, AV_LOG_INFO, "Using \"%s\" for input %d\n", buf, i + 1);
90 }
91 s->in[i].nb_ch = FF_LAYOUT2COUNT(inlayout[i]);
92 if (s->in[i].nb_ch) {
93 overlap++;
94 } else {
95 s->in[i].nb_ch = av_get_channel_layout_nb_channels(inlayout[i]);
96 if (outlayout & inlayout[i])
97 overlap++;
98 outlayout |= inlayout[i];
99 }
100 nb_ch += s->in[i].nb_ch;
101 }
102 if (nb_ch > SWR_CH_MAX) {
103 av_log(ctx, AV_LOG_ERROR, "Too many channels (max %d)\n", SWR_CH_MAX);
104 return AVERROR(EINVAL);
105 }
106 if (overlap) {
107 av_log(ctx, AV_LOG_WARNING,
108 "Input channel layouts overlap: "
109 "output layout will be determined by the number of distinct input channels\n");
110 for (i = 0; i < nb_ch; i++)
111 s->route[i] = i;
112 outlayout = av_get_default_channel_layout(nb_ch);
113 if (!outlayout && nb_ch)
114 outlayout = 0xFFFFFFFFFFFFFFFFULL >> (64 - nb_ch);
115 } else {
116 int *route[SWR_CH_MAX];
117 int c, out_ch_number = 0;
118
119 route[0] = s->route;
120 for (i = 1; i < s->nb_inputs; i++)
121 route[i] = route[i - 1] + s->in[i - 1].nb_ch;
122 for (c = 0; c < 64; c++)
123 for (i = 0; i < s->nb_inputs; i++)
124 if ((inlayout[i] >> c) & 1)
125 *(route[i]++) = out_ch_number++;
126 }
127 formats = ff_make_format_list(ff_packed_sample_fmts_array);
128 if ((ret = ff_set_common_formats(ctx, formats)) < 0)
129 return ret;
130 for (i = 0; i < s->nb_inputs; i++) {
131 layouts = NULL;
132 if ((ret = ff_add_channel_layout(&layouts, inlayout[i])) < 0)
133 return ret;
134 if ((ret = ff_channel_layouts_ref(layouts, &ctx->inputs[i]->out_channel_layouts)) < 0)
135 return ret;
136 }
137 layouts = NULL;
138 if ((ret = ff_add_channel_layout(&layouts, outlayout)) < 0)
139 return ret;
140 if ((ret = ff_channel_layouts_ref(layouts, &ctx->outputs[0]->in_channel_layouts)) < 0)
141 return ret;
142
143 return ff_set_common_samplerates(ctx, ff_all_samplerates());
144 }
145
config_output(AVFilterLink * outlink)146 static int config_output(AVFilterLink *outlink)
147 {
148 AVFilterContext *ctx = outlink->src;
149 AMergeContext *s = ctx->priv;
150 AVBPrint bp;
151 int i;
152
153 for (i = 1; i < s->nb_inputs; i++) {
154 if (ctx->inputs[i]->sample_rate != ctx->inputs[0]->sample_rate) {
155 av_log(ctx, AV_LOG_ERROR,
156 "Inputs must have the same sample rate "
157 "%d for in%d vs %d\n",
158 ctx->inputs[i]->sample_rate, i, ctx->inputs[0]->sample_rate);
159 return AVERROR(EINVAL);
160 }
161 }
162 s->bps = av_get_bytes_per_sample(ctx->outputs[0]->format);
163 outlink->sample_rate = ctx->inputs[0]->sample_rate;
164 outlink->time_base = ctx->inputs[0]->time_base;
165
166 av_bprint_init(&bp, 0, AV_BPRINT_SIZE_AUTOMATIC);
167 for (i = 0; i < s->nb_inputs; i++) {
168 av_bprintf(&bp, "%sin%d:", i ? " + " : "", i);
169 av_bprint_channel_layout(&bp, -1, ctx->inputs[i]->channel_layout);
170 }
171 av_bprintf(&bp, " -> out:");
172 av_bprint_channel_layout(&bp, -1, ctx->outputs[0]->channel_layout);
173 av_log(ctx, AV_LOG_VERBOSE, "%s\n", bp.str);
174
175 return 0;
176 }
177
178 /**
179 * Copy samples from several input streams to one output stream.
180 * @param nb_inputs number of inputs
181 * @param in inputs; used only for the nb_ch field;
182 * @param route routing values;
183 * input channel i goes to output channel route[i];
184 * i < in[0].nb_ch are the channels from the first output;
185 * i >= in[0].nb_ch are the channels from the second output
186 * @param ins pointer to the samples of each inputs, in packed format;
187 * will be left at the end of the copied samples
188 * @param outs pointer to the samples of the output, in packet format;
189 * must point to a buffer big enough;
190 * will be left at the end of the copied samples
191 * @param ns number of samples to copy
192 * @param bps bytes per sample
193 */
copy_samples(int nb_inputs,struct amerge_input in[],int * route,uint8_t * ins[],uint8_t ** outs,int ns,int bps)194 static inline void copy_samples(int nb_inputs, struct amerge_input in[],
195 int *route, uint8_t *ins[],
196 uint8_t **outs, int ns, int bps)
197 {
198 int *route_cur;
199 int i, c, nb_ch = 0;
200
201 for (i = 0; i < nb_inputs; i++)
202 nb_ch += in[i].nb_ch;
203 while (ns--) {
204 route_cur = route;
205 for (i = 0; i < nb_inputs; i++) {
206 for (c = 0; c < in[i].nb_ch; c++) {
207 memcpy((*outs) + bps * *(route_cur++), ins[i], bps);
208 ins[i] += bps;
209 }
210 }
211 *outs += nb_ch * bps;
212 }
213 }
214
free_frames(int nb_inputs,AVFrame ** input_frames)215 static void free_frames(int nb_inputs, AVFrame **input_frames)
216 {
217 int i;
218 for (i = 0; i < nb_inputs; i++)
219 av_frame_free(&input_frames[i]);
220 }
221
try_push_frame(AVFilterContext * ctx,int nb_samples)222 static int try_push_frame(AVFilterContext *ctx, int nb_samples)
223 {
224 AMergeContext *s = ctx->priv;
225 AVFilterLink *outlink = ctx->outputs[0];
226 int i, ret;
227 AVFrame *outbuf, *inbuf[SWR_CH_MAX] = { NULL };
228 uint8_t *outs, *ins[SWR_CH_MAX];
229
230 for (i = 0; i < ctx->nb_inputs; i++) {
231 ret = ff_inlink_consume_samples(ctx->inputs[i], nb_samples, nb_samples, &inbuf[i]);
232 if (ret < 0) {
233 free_frames(i, inbuf);
234 return ret;
235 }
236 ins[i] = inbuf[i]->data[0];
237 }
238
239 outbuf = ff_get_audio_buffer(ctx->outputs[0], nb_samples);
240 if (!outbuf) {
241 free_frames(s->nb_inputs, inbuf);
242 return AVERROR(ENOMEM);
243 }
244
245 outs = outbuf->data[0];
246 outbuf->pts = inbuf[0]->pts;
247
248 outbuf->nb_samples = nb_samples;
249 outbuf->channel_layout = outlink->channel_layout;
250 outbuf->channels = outlink->channels;
251
252 while (nb_samples) {
253 /* Unroll the most common sample formats: speed +~350% for the loop,
254 +~13% overall (including two common decoders) */
255 switch (s->bps) {
256 case 1:
257 copy_samples(s->nb_inputs, s->in, s->route, ins, &outs, nb_samples, 1);
258 break;
259 case 2:
260 copy_samples(s->nb_inputs, s->in, s->route, ins, &outs, nb_samples, 2);
261 break;
262 case 4:
263 copy_samples(s->nb_inputs, s->in, s->route, ins, &outs, nb_samples, 4);
264 break;
265 default:
266 copy_samples(s->nb_inputs, s->in, s->route, ins, &outs, nb_samples, s->bps);
267 break;
268 }
269
270 nb_samples = 0;
271 }
272
273 free_frames(s->nb_inputs, inbuf);
274 return ff_filter_frame(ctx->outputs[0], outbuf);
275 }
276
activate(AVFilterContext * ctx)277 static int activate(AVFilterContext *ctx)
278 {
279 int i, status;
280 int ret, nb_samples;
281 int64_t pts;
282
283 FF_FILTER_FORWARD_STATUS_BACK_ALL(ctx->outputs[0], ctx);
284
285 nb_samples = ff_inlink_queued_samples(ctx->inputs[0]);
286 for (i = 1; i < ctx->nb_inputs && nb_samples > 0; i++) {
287 nb_samples = FFMIN(ff_inlink_queued_samples(ctx->inputs[i]), nb_samples);
288 }
289
290 if (nb_samples) {
291 ret = try_push_frame(ctx, nb_samples);
292 if (ret < 0)
293 return ret;
294 }
295
296 for (i = 0; i < ctx->nb_inputs; i++) {
297 if (ff_inlink_queued_samples(ctx->inputs[i]))
298 continue;
299
300 if (ff_inlink_acknowledge_status(ctx->inputs[i], &status, &pts)) {
301 ff_outlink_set_status(ctx->outputs[0], status, pts);
302 return 0;
303 } else if (ff_outlink_frame_wanted(ctx->outputs[0])) {
304 ff_inlink_request_frame(ctx->inputs[i]);
305 return 0;
306 }
307 }
308
309 return 0;
310 }
311
init(AVFilterContext * ctx)312 static av_cold int init(AVFilterContext *ctx)
313 {
314 AMergeContext *s = ctx->priv;
315 int i, ret;
316
317 s->in = av_calloc(s->nb_inputs, sizeof(*s->in));
318 if (!s->in)
319 return AVERROR(ENOMEM);
320 for (i = 0; i < s->nb_inputs; i++) {
321 char *name = av_asprintf("in%d", i);
322 AVFilterPad pad = {
323 .name = name,
324 .type = AVMEDIA_TYPE_AUDIO,
325 };
326 if (!name)
327 return AVERROR(ENOMEM);
328 if ((ret = ff_insert_inpad(ctx, i, &pad)) < 0) {
329 av_freep(&pad.name);
330 return ret;
331 }
332 }
333 return 0;
334 }
335
336 static const AVFilterPad amerge_outputs[] = {
337 {
338 .name = "default",
339 .type = AVMEDIA_TYPE_AUDIO,
340 .config_props = config_output,
341 },
342 { NULL }
343 };
344
345 AVFilter ff_af_amerge = {
346 .name = "amerge",
347 .description = NULL_IF_CONFIG_SMALL("Merge two or more audio streams into "
348 "a single multi-channel stream."),
349 .priv_size = sizeof(AMergeContext),
350 .init = init,
351 .uninit = uninit,
352 .query_formats = query_formats,
353 .activate = activate,
354 .inputs = NULL,
355 .outputs = amerge_outputs,
356 .priv_class = &amerge_class,
357 .flags = AVFILTER_FLAG_DYNAMIC_INPUTS,
358 };
359