2 * Copyright (c) 2011 Nicolas George <nicolas.george@normalesup.org>
4 * This file is part of FFmpeg.
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.
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.
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
23 * Audio merging filter
26 #define FF_INTERNAL_FIELDS 1
27 #include "framequeue.h"
29 #include "libavutil/avstring.h"
30 #include "libavutil/bprint.h"
31 #include "libavutil/channel_layout.h"
32 #include "libavutil/opt.h"
35 #include "bufferqueue.h"
40 typedef struct AMergeContext {
43 int route[SWR_CH_MAX]; /**< channels routing, see copy_samples */
46 struct FFBufQueue queue;
47 int nb_ch; /**< number of channels for the input */
53 #define OFFSET(x) offsetof(AMergeContext, x)
54 #define FLAGS AV_OPT_FLAG_AUDIO_PARAM|AV_OPT_FLAG_FILTERING_PARAM
56 static const AVOption amerge_options[] = {
57 { "inputs", "specify the number of inputs", OFFSET(nb_inputs),
58 AV_OPT_TYPE_INT, { .i64 = 2 }, 1, SWR_CH_MAX, FLAGS },
62 AVFILTER_DEFINE_CLASS(amerge);
64 static av_cold void uninit(AVFilterContext *ctx)
66 AMergeContext *s = ctx->priv;
69 for (i = 0; i < s->nb_inputs; i++) {
71 ff_bufqueue_discard_all(&s->in[i].queue);
73 av_freep(&ctx->input_pads[i].name);
78 static int query_formats(AVFilterContext *ctx)
80 AMergeContext *s = ctx->priv;
81 int64_t inlayout[SWR_CH_MAX], outlayout = 0;
82 AVFilterFormats *formats;
83 AVFilterChannelLayouts *layouts;
84 int i, ret, overlap = 0, nb_ch = 0;
86 for (i = 0; i < s->nb_inputs; i++) {
87 if (!ctx->inputs[i]->in_channel_layouts ||
88 !ctx->inputs[i]->in_channel_layouts->nb_channel_layouts) {
89 av_log(ctx, AV_LOG_WARNING,
90 "No channel layout for input %d\n", i + 1);
91 return AVERROR(EAGAIN);
93 inlayout[i] = ctx->inputs[i]->in_channel_layouts->channel_layouts[0];
94 if (ctx->inputs[i]->in_channel_layouts->nb_channel_layouts > 1) {
96 av_get_channel_layout_string(buf, sizeof(buf), 0, inlayout[i]);
97 av_log(ctx, AV_LOG_INFO, "Using \"%s\" for input %d\n", buf, i + 1);
99 s->in[i].nb_ch = FF_LAYOUT2COUNT(inlayout[i]);
100 if (s->in[i].nb_ch) {
103 s->in[i].nb_ch = av_get_channel_layout_nb_channels(inlayout[i]);
104 if (outlayout & inlayout[i])
106 outlayout |= inlayout[i];
108 nb_ch += s->in[i].nb_ch;
110 if (nb_ch > SWR_CH_MAX) {
111 av_log(ctx, AV_LOG_ERROR, "Too many channels (max %d)\n", SWR_CH_MAX);
112 return AVERROR(EINVAL);
115 av_log(ctx, AV_LOG_WARNING,
116 "Input channel layouts overlap: "
117 "output layout will be determined by the number of distinct input channels\n");
118 for (i = 0; i < nb_ch; i++)
120 outlayout = av_get_default_channel_layout(nb_ch);
121 if (!outlayout && nb_ch)
122 outlayout = 0xFFFFFFFFFFFFFFFFULL >> (64 - nb_ch);
124 int *route[SWR_CH_MAX];
125 int c, out_ch_number = 0;
128 for (i = 1; i < s->nb_inputs; i++)
129 route[i] = route[i - 1] + s->in[i - 1].nb_ch;
130 for (c = 0; c < 64; c++)
131 for (i = 0; i < s->nb_inputs; i++)
132 if ((inlayout[i] >> c) & 1)
133 *(route[i]++) = out_ch_number++;
135 formats = ff_make_format_list(ff_packed_sample_fmts_array);
136 if ((ret = ff_set_common_formats(ctx, formats)) < 0)
138 for (i = 0; i < s->nb_inputs; i++) {
140 if ((ret = ff_add_channel_layout(&layouts, inlayout[i])) < 0)
142 if ((ret = ff_channel_layouts_ref(layouts, &ctx->inputs[i]->out_channel_layouts)) < 0)
146 if ((ret = ff_add_channel_layout(&layouts, outlayout)) < 0)
148 if ((ret = ff_channel_layouts_ref(layouts, &ctx->outputs[0]->in_channel_layouts)) < 0)
151 return ff_set_common_samplerates(ctx, ff_all_samplerates());
154 static int config_output(AVFilterLink *outlink)
156 AVFilterContext *ctx = outlink->src;
157 AMergeContext *s = ctx->priv;
161 for (i = 1; i < s->nb_inputs; i++) {
162 if (ctx->inputs[i]->sample_rate != ctx->inputs[0]->sample_rate) {
163 av_log(ctx, AV_LOG_ERROR,
164 "Inputs must have the same sample rate "
165 "%d for in%d vs %d\n",
166 ctx->inputs[i]->sample_rate, i, ctx->inputs[0]->sample_rate);
167 return AVERROR(EINVAL);
170 s->bps = av_get_bytes_per_sample(ctx->outputs[0]->format);
171 outlink->sample_rate = ctx->inputs[0]->sample_rate;
172 outlink->time_base = ctx->inputs[0]->time_base;
174 av_bprint_init(&bp, 0, 1);
175 for (i = 0; i < s->nb_inputs; i++) {
176 av_bprintf(&bp, "%sin%d:", i ? " + " : "", i);
177 av_bprint_channel_layout(&bp, -1, ctx->inputs[i]->channel_layout);
179 av_bprintf(&bp, " -> out:");
180 av_bprint_channel_layout(&bp, -1, ctx->outputs[0]->channel_layout);
181 av_log(ctx, AV_LOG_VERBOSE, "%s\n", bp.str);
186 static int request_frame(AVFilterLink *outlink)
188 AVFilterContext *ctx = outlink->src;
189 AMergeContext *s = ctx->priv;
192 for (i = 0; i < s->nb_inputs; i++)
193 if (!s->in[i].nb_samples ||
194 /* detect EOF immediately */
195 (ctx->inputs[i]->status_in && !ctx->inputs[i]->status_out))
196 if ((ret = ff_request_frame(ctx->inputs[i])) < 0)
202 * Copy samples from several input streams to one output stream.
203 * @param nb_inputs number of inputs
204 * @param in inputs; used only for the nb_ch field;
205 * @param route routing values;
206 * input channel i goes to output channel route[i];
207 * i < in[0].nb_ch are the channels from the first output;
208 * i >= in[0].nb_ch are the channels from the second output
209 * @param ins pointer to the samples of each inputs, in packed format;
210 * will be left at the end of the copied samples
211 * @param outs pointer to the samples of the output, in packet format;
212 * must point to a buffer big enough;
213 * will be left at the end of the copied samples
214 * @param ns number of samples to copy
215 * @param bps bytes per sample
217 static inline void copy_samples(int nb_inputs, struct amerge_input in[],
218 int *route, uint8_t *ins[],
219 uint8_t **outs, int ns, int bps)
224 for (i = 0; i < nb_inputs; i++)
225 nb_ch += in[i].nb_ch;
228 for (i = 0; i < nb_inputs; i++) {
229 for (c = 0; c < in[i].nb_ch; c++) {
230 memcpy((*outs) + bps * *(route_cur++), ins[i], bps);
234 *outs += nb_ch * bps;
238 static int filter_frame(AVFilterLink *inlink, AVFrame *insamples)
240 AVFilterContext *ctx = inlink->dst;
241 AMergeContext *s = ctx->priv;
242 AVFilterLink *const outlink = ctx->outputs[0];
244 int nb_samples, ns, i;
245 AVFrame *outbuf, *inbuf[SWR_CH_MAX];
246 uint8_t *ins[SWR_CH_MAX], *outs;
248 for (input_number = 0; input_number < s->nb_inputs; input_number++)
249 if (inlink == ctx->inputs[input_number])
251 av_assert1(input_number < s->nb_inputs);
252 if (ff_bufqueue_is_full(&s->in[input_number].queue)) {
253 av_frame_free(&insamples);
254 return AVERROR(ENOMEM);
256 ff_bufqueue_add(ctx, &s->in[input_number].queue, av_frame_clone(insamples));
257 s->in[input_number].nb_samples += insamples->nb_samples;
258 av_frame_free(&insamples);
259 nb_samples = s->in[0].nb_samples;
260 for (i = 1; i < s->nb_inputs; i++)
261 nb_samples = FFMIN(nb_samples, s->in[i].nb_samples);
265 outbuf = ff_get_audio_buffer(ctx->outputs[0], nb_samples);
267 return AVERROR(ENOMEM);
268 outs = outbuf->data[0];
269 for (i = 0; i < s->nb_inputs; i++) {
270 inbuf[i] = ff_bufqueue_peek(&s->in[i].queue, 0);
271 ins[i] = inbuf[i]->data[0] +
272 s->in[i].pos * s->in[i].nb_ch * s->bps;
274 av_frame_copy_props(outbuf, inbuf[0]);
275 outbuf->pts = inbuf[0]->pts == AV_NOPTS_VALUE ? AV_NOPTS_VALUE :
277 av_rescale_q(s->in[0].pos,
278 av_make_q(1, ctx->inputs[0]->sample_rate),
279 ctx->outputs[0]->time_base);
281 outbuf->nb_samples = nb_samples;
282 outbuf->channel_layout = outlink->channel_layout;
283 outbuf->channels = outlink->channels;
287 for (i = 0; i < s->nb_inputs; i++)
288 ns = FFMIN(ns, inbuf[i]->nb_samples - s->in[i].pos);
289 /* Unroll the most common sample formats: speed +~350% for the loop,
290 +~13% overall (including two common decoders) */
293 copy_samples(s->nb_inputs, s->in, s->route, ins, &outs, ns, 1);
296 copy_samples(s->nb_inputs, s->in, s->route, ins, &outs, ns, 2);
299 copy_samples(s->nb_inputs, s->in, s->route, ins, &outs, ns, 4);
302 copy_samples(s->nb_inputs, s->in, s->route, ins, &outs, ns, s->bps);
307 for (i = 0; i < s->nb_inputs; i++) {
308 s->in[i].nb_samples -= ns;
310 if (s->in[i].pos == inbuf[i]->nb_samples) {
312 av_frame_free(&inbuf[i]);
313 ff_bufqueue_get(&s->in[i].queue);
314 inbuf[i] = ff_bufqueue_peek(&s->in[i].queue, 0);
315 ins[i] = inbuf[i] ? inbuf[i]->data[0] : NULL;
319 return ff_filter_frame(ctx->outputs[0], outbuf);
322 static av_cold int init(AVFilterContext *ctx)
324 AMergeContext *s = ctx->priv;
327 s->in = av_calloc(s->nb_inputs, sizeof(*s->in));
329 return AVERROR(ENOMEM);
330 for (i = 0; i < s->nb_inputs; i++) {
331 char *name = av_asprintf("in%d", i);
334 .type = AVMEDIA_TYPE_AUDIO,
335 .filter_frame = filter_frame,
338 return AVERROR(ENOMEM);
339 if ((ret = ff_insert_inpad(ctx, i, &pad)) < 0) {
347 static const AVFilterPad amerge_outputs[] = {
350 .type = AVMEDIA_TYPE_AUDIO,
351 .config_props = config_output,
352 .request_frame = request_frame,
357 AVFilter ff_af_amerge = {
359 .description = NULL_IF_CONFIG_SMALL("Merge two or more audio streams into "
360 "a single multi-channel stream."),
361 .priv_size = sizeof(AMergeContext),
364 .query_formats = query_formats,
366 .outputs = amerge_outputs,
367 .priv_class = &amerge_class,
368 .flags = AVFILTER_FLAG_DYNAMIC_INPUTS,