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 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)
50 static const AVOption amerge_options[] = {
51 { "inputs", "specify the number of inputs", OFFSET(nb_inputs),
52 AV_OPT_TYPE_INT, { .dbl = 2 }, 2, SWR_CH_MAX },
56 AVFILTER_DEFINE_CLASS(amerge);
58 static av_cold void uninit(AVFilterContext *ctx)
60 AMergeContext *am = ctx->priv;
63 for (i = 0; i < am->nb_inputs; i++)
64 ff_bufqueue_discard_all(&am->in[i].queue);
68 static int query_formats(AVFilterContext *ctx)
70 AMergeContext *am = ctx->priv;
71 int64_t inlayout[SWR_CH_MAX], outlayout = 0;
72 AVFilterFormats *formats;
73 AVFilterChannelLayouts *layouts;
74 int i, overlap = 0, nb_ch = 0;
76 for (i = 0; i < am->nb_inputs; i++) {
77 if (!ctx->inputs[i]->in_channel_layouts ||
78 !ctx->inputs[i]->in_channel_layouts->nb_channel_layouts) {
79 av_log(ctx, AV_LOG_ERROR,
80 "No channel layout for input %d\n", i + 1);
81 return AVERROR(EINVAL);
83 inlayout[i] = ctx->inputs[i]->in_channel_layouts->channel_layouts[0];
84 if (ctx->inputs[i]->in_channel_layouts->nb_channel_layouts > 1) {
86 av_get_channel_layout_string(buf, sizeof(buf), 0, inlayout[i]);
87 av_log(ctx, AV_LOG_INFO, "Using \"%s\" for input %d\n", buf, i + 1);
89 am->in[i].nb_ch = av_get_channel_layout_nb_channels(inlayout[i]);
90 if (outlayout & inlayout[i])
92 outlayout |= inlayout[i];
93 nb_ch += am->in[i].nb_ch;
95 if (nb_ch > SWR_CH_MAX) {
96 av_log(ctx, AV_LOG_ERROR, "Too many channels (max %d)\n", SWR_CH_MAX);
97 return AVERROR(EINVAL);
100 av_log(ctx, AV_LOG_WARNING,
101 "Inputs overlap: output layout will be meaningless\n");
102 for (i = 0; i < nb_ch; i++)
104 outlayout = av_get_default_channel_layout(nb_ch);
106 outlayout = ((int64_t)1 << nb_ch) - 1;
108 int *route[SWR_CH_MAX];
109 int c, out_ch_number = 0;
111 route[0] = am->route;
112 for (i = 1; i < am->nb_inputs; i++)
113 route[i] = route[i - 1] + am->in[i - 1].nb_ch;
114 for (c = 0; c < 64; c++)
115 for (i = 0; i < am->nb_inputs; i++)
116 if ((inlayout[i] >> c) & 1)
117 *(route[i]++) = out_ch_number++;
119 formats = ff_make_format_list(ff_packed_sample_fmts_array);
120 ff_set_common_formats(ctx, formats);
121 for (i = 0; i < am->nb_inputs; i++) {
123 ff_add_channel_layout(&layouts, inlayout[i]);
124 ff_channel_layouts_ref(layouts, &ctx->inputs[i]->out_channel_layouts);
127 ff_add_channel_layout(&layouts, outlayout);
128 ff_channel_layouts_ref(layouts, &ctx->outputs[0]->in_channel_layouts);
129 ff_set_common_samplerates(ctx, ff_all_samplerates());
133 static int config_output(AVFilterLink *outlink)
135 AVFilterContext *ctx = outlink->src;
136 AMergeContext *am = ctx->priv;
140 for (i = 1; i < am->nb_inputs; i++) {
141 if (ctx->inputs[i]->sample_rate != ctx->inputs[0]->sample_rate) {
142 av_log(ctx, AV_LOG_ERROR,
143 "Inputs must have the same sample rate "
144 "%d for in%d vs %d\n",
145 ctx->inputs[i]->sample_rate, i, ctx->inputs[0]->sample_rate);
146 return AVERROR(EINVAL);
149 am->bps = av_get_bytes_per_sample(ctx->outputs[0]->format);
150 outlink->sample_rate = ctx->inputs[0]->sample_rate;
151 outlink->time_base = ctx->inputs[0]->time_base;
153 av_bprint_init(&bp, 0, 1);
154 for (i = 0; i < am->nb_inputs; i++) {
155 av_bprintf(&bp, "%sin%d:", i ? " + " : "", i);
156 av_bprint_channel_layout(&bp, -1, ctx->inputs[i]->channel_layout);
158 av_bprintf(&bp, " -> out:");
159 av_bprint_channel_layout(&bp, -1, ctx->outputs[0]->channel_layout);
160 av_log(ctx, AV_LOG_VERBOSE, "%s\n", bp.str);
165 static int request_frame(AVFilterLink *outlink)
167 AVFilterContext *ctx = outlink->src;
168 AMergeContext *am = ctx->priv;
171 for (i = 0; i < am->nb_inputs; i++)
172 if (!am->in[i].nb_samples)
173 if ((ret = ff_request_frame(ctx->inputs[i])) < 0)
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
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)
201 for (i = 0; i < nb_inputs; i++)
202 nb_ch += in[i].nb_ch;
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);
211 *outs += nb_ch * bps;
215 static int filter_samples(AVFilterLink *inlink, AVFilterBufferRef *insamples)
217 AVFilterContext *ctx = inlink->dst;
218 AMergeContext *am = ctx->priv;
219 AVFilterLink *const outlink = ctx->outputs[0];
221 int nb_samples, ns, i;
222 AVFilterBufferRef *outbuf, *inbuf[SWR_CH_MAX];
223 uint8_t *ins[SWR_CH_MAX], *outs;
225 for (input_number = 0; input_number < am->nb_inputs; input_number++)
226 if (inlink == ctx->inputs[input_number])
228 av_assert1(input_number < am->nb_inputs);
229 ff_bufqueue_add(ctx, &am->in[input_number].queue, insamples);
230 am->in[input_number].nb_samples += insamples->audio->nb_samples;
231 nb_samples = am->in[0].nb_samples;
232 for (i = 1; i < am->nb_inputs; i++)
233 nb_samples = FFMIN(nb_samples, am->in[i].nb_samples);
237 outbuf = ff_get_audio_buffer(ctx->outputs[0], AV_PERM_WRITE, nb_samples);
238 outs = outbuf->data[0];
239 for (i = 0; i < am->nb_inputs; i++) {
240 inbuf[i] = ff_bufqueue_peek(&am->in[i].queue, 0);
241 ins[i] = inbuf[i]->data[0] +
242 am->in[i].pos * am->in[i].nb_ch * am->bps;
244 outbuf->pts = inbuf[0]->pts == AV_NOPTS_VALUE ? AV_NOPTS_VALUE :
246 av_rescale_q(am->in[0].pos,
247 (AVRational){ 1, ctx->inputs[0]->sample_rate },
248 ctx->outputs[0]->time_base);
250 avfilter_copy_buffer_ref_props(outbuf, inbuf[0]);
251 outbuf->audio->nb_samples = nb_samples;
252 outbuf->audio->channel_layout = outlink->channel_layout;
256 for (i = 0; i < am->nb_inputs; i++)
257 ns = FFMIN(ns, inbuf[i]->audio->nb_samples - am->in[i].pos);
258 /* Unroll the most common sample formats: speed +~350% for the loop,
259 +~13% overall (including two common decoders) */
262 copy_samples(am->nb_inputs, am->in, am->route, ins, &outs, ns, 1);
265 copy_samples(am->nb_inputs, am->in, am->route, ins, &outs, ns, 2);
268 copy_samples(am->nb_inputs, am->in, am->route, ins, &outs, ns, 4);
271 copy_samples(am->nb_inputs, am->in, am->route, ins, &outs, ns, am->bps);
276 for (i = 0; i < am->nb_inputs; i++) {
277 am->in[i].nb_samples -= ns;
279 if (am->in[i].pos == inbuf[i]->audio->nb_samples) {
281 avfilter_unref_buffer(inbuf[i]);
282 ff_bufqueue_get(&am->in[i].queue);
283 inbuf[i] = ff_bufqueue_peek(&am->in[i].queue, 0);
284 ins[i] = inbuf[i] ? inbuf[i]->data[0] : NULL;
288 return ff_filter_samples(ctx->outputs[0], outbuf);
291 static av_cold int init(AVFilterContext *ctx, const char *args)
293 AMergeContext *am = ctx->priv;
297 am->class = &amerge_class;
298 av_opt_set_defaults(am);
299 ret = av_set_options_string(am, args, "=", ":");
301 av_log(ctx, AV_LOG_ERROR, "Error parsing options: '%s'\n", args);
304 am->in = av_calloc(am->nb_inputs, sizeof(*am->in));
306 return AVERROR(ENOMEM);
307 for (i = 0; i < am->nb_inputs; i++) {
310 .type = AVMEDIA_TYPE_AUDIO,
311 .filter_samples = filter_samples,
312 .min_perms = AV_PERM_READ | AV_PERM_PRESERVE,
314 snprintf(name, sizeof(name), "in%d", i);
315 ff_insert_inpad(ctx, i, &pad);
320 AVFilter avfilter_af_amerge = {
322 .description = NULL_IF_CONFIG_SMALL("Merge two audio streams into "
323 "a single multi-channel stream."),
324 .priv_size = sizeof(AMergeContext),
327 .query_formats = query_formats,
329 .inputs = (const AVFilterPad[]) { { .name = NULL } },
330 .outputs = (const AVFilterPad[]) {
332 .type = AVMEDIA_TYPE_AUDIO,
333 .config_props = config_output,
334 .request_frame = request_frame, },