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 #include "libavutil/audioconvert.h"
27 #include "libavutil/bprint.h"
28 #include "libavutil/opt.h"
29 #include "libswresample/swresample.h" // only for SWR_CH_MAX
32 #include "bufferqueue.h"
38 int route[SWR_CH_MAX]; /**< channels routing, see copy_samples */
41 struct FFBufQueue queue;
42 int nb_ch; /**< number of channels for the input */
48 #define OFFSET(x) offsetof(AMergeContext, x)
49 #define FLAGS AV_OPT_FLAG_AUDIO_PARAM|AV_OPT_FLAG_FILTERING_PARAM
51 static const AVOption amerge_options[] = {
52 { "inputs", "specify the number of inputs", OFFSET(nb_inputs),
53 AV_OPT_TYPE_INT, { .dbl = 2 }, 2, SWR_CH_MAX, FLAGS },
57 AVFILTER_DEFINE_CLASS(amerge);
59 static av_cold void uninit(AVFilterContext *ctx)
61 AMergeContext *am = ctx->priv;
64 for (i = 0; i < am->nb_inputs; i++)
65 ff_bufqueue_discard_all(&am->in[i].queue);
69 static int query_formats(AVFilterContext *ctx)
71 AMergeContext *am = ctx->priv;
72 int64_t inlayout[SWR_CH_MAX], outlayout = 0;
73 AVFilterFormats *formats;
74 AVFilterChannelLayouts *layouts;
75 int i, overlap = 0, nb_ch = 0;
77 for (i = 0; i < am->nb_inputs; i++) {
78 if (!ctx->inputs[i]->in_channel_layouts ||
79 !ctx->inputs[i]->in_channel_layouts->nb_channel_layouts) {
80 av_log(ctx, AV_LOG_ERROR,
81 "No channel layout for input %d\n", i + 1);
82 return AVERROR(EINVAL);
84 inlayout[i] = ctx->inputs[i]->in_channel_layouts->channel_layouts[0];
85 if (ctx->inputs[i]->in_channel_layouts->nb_channel_layouts > 1) {
87 av_get_channel_layout_string(buf, sizeof(buf), 0, inlayout[i]);
88 av_log(ctx, AV_LOG_INFO, "Using \"%s\" for input %d\n", buf, i + 1);
90 am->in[i].nb_ch = av_get_channel_layout_nb_channels(inlayout[i]);
91 if (outlayout & inlayout[i])
93 outlayout |= inlayout[i];
94 nb_ch += am->in[i].nb_ch;
96 if (nb_ch > SWR_CH_MAX) {
97 av_log(ctx, AV_LOG_ERROR, "Too many channels (max %d)\n", SWR_CH_MAX);
98 return AVERROR(EINVAL);
101 av_log(ctx, AV_LOG_WARNING,
102 "Inputs overlap: output layout will be meaningless\n");
103 for (i = 0; i < nb_ch; i++)
105 outlayout = av_get_default_channel_layout(nb_ch);
107 outlayout = ((int64_t)1 << nb_ch) - 1;
109 int *route[SWR_CH_MAX];
110 int c, out_ch_number = 0;
112 route[0] = am->route;
113 for (i = 1; i < am->nb_inputs; i++)
114 route[i] = route[i - 1] + am->in[i - 1].nb_ch;
115 for (c = 0; c < 64; c++)
116 for (i = 0; i < am->nb_inputs; i++)
117 if ((inlayout[i] >> c) & 1)
118 *(route[i]++) = out_ch_number++;
120 formats = ff_make_format_list(ff_packed_sample_fmts_array);
121 ff_set_common_formats(ctx, formats);
122 for (i = 0; i < am->nb_inputs; i++) {
124 ff_add_channel_layout(&layouts, inlayout[i]);
125 ff_channel_layouts_ref(layouts, &ctx->inputs[i]->out_channel_layouts);
128 ff_add_channel_layout(&layouts, outlayout);
129 ff_channel_layouts_ref(layouts, &ctx->outputs[0]->in_channel_layouts);
130 ff_set_common_samplerates(ctx, ff_all_samplerates());
134 static int config_output(AVFilterLink *outlink)
136 AVFilterContext *ctx = outlink->src;
137 AMergeContext *am = ctx->priv;
141 for (i = 1; i < am->nb_inputs; i++) {
142 if (ctx->inputs[i]->sample_rate != ctx->inputs[0]->sample_rate) {
143 av_log(ctx, AV_LOG_ERROR,
144 "Inputs must have the same sample rate "
145 "%d for in%d vs %d\n",
146 ctx->inputs[i]->sample_rate, i, ctx->inputs[0]->sample_rate);
147 return AVERROR(EINVAL);
150 am->bps = av_get_bytes_per_sample(ctx->outputs[0]->format);
151 outlink->sample_rate = ctx->inputs[0]->sample_rate;
152 outlink->time_base = ctx->inputs[0]->time_base;
154 av_bprint_init(&bp, 0, 1);
155 for (i = 0; i < am->nb_inputs; i++) {
156 av_bprintf(&bp, "%sin%d:", i ? " + " : "", i);
157 av_bprint_channel_layout(&bp, -1, ctx->inputs[i]->channel_layout);
159 av_bprintf(&bp, " -> out:");
160 av_bprint_channel_layout(&bp, -1, ctx->outputs[0]->channel_layout);
161 av_log(ctx, AV_LOG_VERBOSE, "%s\n", bp.str);
166 static int request_frame(AVFilterLink *outlink)
168 AVFilterContext *ctx = outlink->src;
169 AMergeContext *am = ctx->priv;
172 for (i = 0; i < am->nb_inputs; i++)
173 if (!am->in[i].nb_samples)
174 if ((ret = ff_request_frame(ctx->inputs[i])) < 0)
180 * Copy samples from several input streams to one output stream.
181 * @param nb_inputs number of inputs
182 * @param in inputs; used only for the nb_ch field;
183 * @param route routing values;
184 * input channel i goes to output channel route[i];
185 * i < in[0].nb_ch are the channels from the first output;
186 * i >= in[0].nb_ch are the channels from the second output
187 * @param ins pointer to the samples of each inputs, in packed format;
188 * will be left at the end of the copied samples
189 * @param outs pointer to the samples of the output, in packet format;
190 * must point to a buffer big enough;
191 * will be left at the end of the copied samples
192 * @param ns number of samples to copy
193 * @param bps bytes per sample
195 static inline void copy_samples(int nb_inputs, struct amerge_input in[],
196 int *route, uint8_t *ins[],
197 uint8_t **outs, int ns, int bps)
202 for (i = 0; i < nb_inputs; i++)
203 nb_ch += in[i].nb_ch;
206 for (i = 0; i < nb_inputs; i++) {
207 for (c = 0; c < in[i].nb_ch; c++) {
208 memcpy((*outs) + bps * *(route_cur++), ins[i], bps);
212 *outs += nb_ch * bps;
216 static int filter_samples(AVFilterLink *inlink, AVFilterBufferRef *insamples)
218 AVFilterContext *ctx = inlink->dst;
219 AMergeContext *am = ctx->priv;
220 AVFilterLink *const outlink = ctx->outputs[0];
222 int nb_samples, ns, i;
223 AVFilterBufferRef *outbuf, *inbuf[SWR_CH_MAX];
224 uint8_t *ins[SWR_CH_MAX], *outs;
226 for (input_number = 0; input_number < am->nb_inputs; input_number++)
227 if (inlink == ctx->inputs[input_number])
229 av_assert1(input_number < am->nb_inputs);
230 ff_bufqueue_add(ctx, &am->in[input_number].queue, insamples);
231 am->in[input_number].nb_samples += insamples->audio->nb_samples;
232 nb_samples = am->in[0].nb_samples;
233 for (i = 1; i < am->nb_inputs; i++)
234 nb_samples = FFMIN(nb_samples, am->in[i].nb_samples);
238 outbuf = ff_get_audio_buffer(ctx->outputs[0], AV_PERM_WRITE, nb_samples);
239 outs = outbuf->data[0];
240 for (i = 0; i < am->nb_inputs; i++) {
241 inbuf[i] = ff_bufqueue_peek(&am->in[i].queue, 0);
242 ins[i] = inbuf[i]->data[0] +
243 am->in[i].pos * am->in[i].nb_ch * am->bps;
245 avfilter_copy_buffer_ref_props(outbuf, inbuf[0]);
246 outbuf->pts = inbuf[0]->pts == AV_NOPTS_VALUE ? AV_NOPTS_VALUE :
248 av_rescale_q(am->in[0].pos,
249 (AVRational){ 1, ctx->inputs[0]->sample_rate },
250 ctx->outputs[0]->time_base);
252 outbuf->audio->nb_samples = nb_samples;
253 outbuf->audio->channel_layout = outlink->channel_layout;
257 for (i = 0; i < am->nb_inputs; i++)
258 ns = FFMIN(ns, inbuf[i]->audio->nb_samples - am->in[i].pos);
259 /* Unroll the most common sample formats: speed +~350% for the loop,
260 +~13% overall (including two common decoders) */
263 copy_samples(am->nb_inputs, am->in, am->route, ins, &outs, ns, 1);
266 copy_samples(am->nb_inputs, am->in, am->route, ins, &outs, ns, 2);
269 copy_samples(am->nb_inputs, am->in, am->route, ins, &outs, ns, 4);
272 copy_samples(am->nb_inputs, am->in, am->route, ins, &outs, ns, am->bps);
277 for (i = 0; i < am->nb_inputs; i++) {
278 am->in[i].nb_samples -= ns;
280 if (am->in[i].pos == inbuf[i]->audio->nb_samples) {
282 avfilter_unref_buffer(inbuf[i]);
283 ff_bufqueue_get(&am->in[i].queue);
284 inbuf[i] = ff_bufqueue_peek(&am->in[i].queue, 0);
285 ins[i] = inbuf[i] ? inbuf[i]->data[0] : NULL;
289 return ff_filter_samples(ctx->outputs[0], outbuf);
292 static av_cold int init(AVFilterContext *ctx, const char *args)
294 AMergeContext *am = ctx->priv;
298 am->class = &amerge_class;
299 av_opt_set_defaults(am);
300 ret = av_set_options_string(am, args, "=", ":");
302 av_log(ctx, AV_LOG_ERROR, "Error parsing options: '%s'\n", args);
305 am->in = av_calloc(am->nb_inputs, sizeof(*am->in));
307 return AVERROR(ENOMEM);
308 for (i = 0; i < am->nb_inputs; i++) {
311 .type = AVMEDIA_TYPE_AUDIO,
312 .filter_samples = filter_samples,
313 .min_perms = AV_PERM_READ | AV_PERM_PRESERVE,
315 snprintf(name, sizeof(name), "in%d", i);
316 ff_insert_inpad(ctx, i, &pad);
321 AVFilter avfilter_af_amerge = {
323 .description = NULL_IF_CONFIG_SMALL("Merge two audio streams into "
324 "a single multi-channel stream."),
325 .priv_size = sizeof(AMergeContext),
328 .query_formats = query_formats,
330 .inputs = (const AVFilterPad[]) { { .name = NULL } },
331 .outputs = (const AVFilterPad[]) {
333 .type = AVMEDIA_TYPE_AUDIO,
334 .config_props = config_output,
335 .request_frame = request_frame, },
338 .priv_class = &amerge_class,