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/bprint.h"
27 #include "libavutil/opt.h"
28 #include "libswresample/swresample.h" // only for SWR_CH_MAX
31 #include "bufferqueue.h"
37 int route[SWR_CH_MAX]; /**< channels routing, see copy_samples */
40 struct FFBufQueue queue;
41 int nb_ch; /**< number of channels for the input */
47 #define OFFSET(x) offsetof(AMergeContext, x)
49 static const AVOption amerge_options[] = {
50 { "inputs", "specify the number of inputs", OFFSET(nb_inputs),
51 AV_OPT_TYPE_INT, { .dbl = 2 }, 2, SWR_CH_MAX },
55 static const AVClass amerge_class = {
56 .class_name = "amerge",
57 .item_name = av_default_item_name,
58 .option = amerge_options,
59 .version = LIBAVUTIL_VERSION_INT,
60 .category = AV_CLASS_CATEGORY_FILTER,
63 static av_cold void uninit(AVFilterContext *ctx)
65 AMergeContext *am = ctx->priv;
68 for (i = 0; i < am->nb_inputs; i++)
69 ff_bufqueue_discard_all(&am->in[i].queue);
73 static int query_formats(AVFilterContext *ctx)
75 AMergeContext *am = ctx->priv;
76 int64_t inlayout[SWR_CH_MAX], outlayout = 0;
77 AVFilterFormats *formats;
78 AVFilterChannelLayouts *layouts;
79 int i, overlap = 0, nb_ch = 0;
81 for (i = 0; i < am->nb_inputs; i++) {
82 if (!ctx->inputs[i]->in_channel_layouts ||
83 !ctx->inputs[i]->in_channel_layouts->nb_channel_layouts) {
84 av_log(ctx, AV_LOG_ERROR,
85 "No channel layout for input %d\n", i + 1);
86 return AVERROR(EINVAL);
88 inlayout[i] = ctx->inputs[i]->in_channel_layouts->channel_layouts[0];
89 if (ctx->inputs[i]->in_channel_layouts->nb_channel_layouts > 1) {
91 av_get_channel_layout_string(buf, sizeof(buf), 0, inlayout[i]);
92 av_log(ctx, AV_LOG_INFO, "Using \"%s\" for input %d\n", buf, i + 1);
94 am->in[i].nb_ch = av_get_channel_layout_nb_channels(inlayout[i]);
95 if (outlayout & inlayout[i])
97 outlayout |= inlayout[i];
98 nb_ch += am->in[i].nb_ch;
100 if (nb_ch > SWR_CH_MAX) {
101 av_log(ctx, AV_LOG_ERROR, "Too many channels (max %d)\n", SWR_CH_MAX);
102 return AVERROR(EINVAL);
105 av_log(ctx, AV_LOG_WARNING,
106 "Inputs overlap: output layout will be meaningless\n");
107 for (i = 0; i < nb_ch; i++)
109 outlayout = av_get_default_channel_layout(nb_ch);
111 outlayout = ((int64_t)1 << nb_ch) - 1;
113 int *route[SWR_CH_MAX];
114 int c, out_ch_number = 0;
116 route[0] = am->route;
117 for (i = 1; i < am->nb_inputs; i++)
118 route[i] = route[i - 1] + am->in[i - 1].nb_ch;
119 for (c = 0; c < 64; c++)
120 for (i = 0; i < am->nb_inputs; i++)
121 if ((inlayout[i] >> c) & 1)
122 *(route[i]++) = out_ch_number++;
124 formats = ff_make_format_list(ff_packed_sample_fmts_array);
125 ff_set_common_formats(ctx, formats);
126 for (i = 0; i < am->nb_inputs; i++) {
128 ff_add_channel_layout(&layouts, inlayout[i]);
129 ff_channel_layouts_ref(layouts, &ctx->inputs[i]->out_channel_layouts);
132 ff_add_channel_layout(&layouts, outlayout);
133 ff_channel_layouts_ref(layouts, &ctx->outputs[0]->in_channel_layouts);
134 ff_set_common_samplerates(ctx, ff_all_samplerates());
138 static int config_output(AVFilterLink *outlink)
140 AVFilterContext *ctx = outlink->src;
141 AMergeContext *am = ctx->priv;
145 for (i = 1; i < am->nb_inputs; i++) {
146 if (ctx->inputs[i]->sample_rate != ctx->inputs[0]->sample_rate) {
147 av_log(ctx, AV_LOG_ERROR,
148 "Inputs must have the same sample rate "
149 "(%"PRIi64" for in%d vs %"PRIi64")\n",
150 ctx->inputs[i]->sample_rate, i, ctx->inputs[0]->sample_rate);
151 return AVERROR(EINVAL);
154 am->bps = av_get_bytes_per_sample(ctx->outputs[0]->format);
155 outlink->sample_rate = ctx->inputs[0]->sample_rate;
156 outlink->time_base = ctx->inputs[0]->time_base;
158 av_bprint_init(&bp, 0, 1);
159 for (i = 0; i < am->nb_inputs; i++) {
160 av_bprintf(&bp, "%sin%d:", i ? " + " : "", i);
161 av_bprint_channel_layout(&bp, -1, ctx->inputs[i]->channel_layout);
163 av_bprintf(&bp, " -> out:");
164 av_bprint_channel_layout(&bp, -1, ctx->outputs[0]->channel_layout);
165 av_log(ctx, AV_LOG_INFO, "%s\n", bp.str);
170 static int request_frame(AVFilterLink *outlink)
172 AVFilterContext *ctx = outlink->src;
173 AMergeContext *am = ctx->priv;
176 for (i = 0; i < am->nb_inputs; i++)
177 if (!am->in[i].nb_samples)
178 if ((ret = avfilter_request_frame(ctx->inputs[i])) < 0)
184 * Copy samples from several input streams to one output stream.
185 * @param nb_inputs number of inputs
186 * @param in inputs; used only for the nb_ch field;
187 * @param route routing values;
188 * input channel i goes to output channel route[i];
189 * i < in[0].nb_ch are the channels from the first output;
190 * i >= in[0].nb_ch are the channels from the second output
191 * @param ins pointer to the samples of each inputs, in packed format;
192 * will be left at the end of the copied samples
193 * @param outs pointer to the samples of the output, in packet format;
194 * must point to a buffer big enough;
195 * will be left at the end of the copied samples
196 * @param ns number of samples to copy
197 * @param bps bytes per sample
199 static inline void copy_samples(int nb_inputs, struct amerge_input in[],
200 int *route, uint8_t *ins[],
201 uint8_t **outs, int ns, int bps)
206 for (i = 0; i < nb_inputs; i++)
207 nb_ch += in[i].nb_ch;
210 for (i = 0; i < nb_inputs; i++) {
211 for (c = 0; c < in[i].nb_ch; c++) {
212 memcpy((*outs) + bps * *(route_cur++), ins[i], bps);
216 *outs += nb_ch * bps;
220 static void filter_samples(AVFilterLink *inlink, AVFilterBufferRef *insamples)
222 AVFilterContext *ctx = inlink->dst;
223 AMergeContext *am = ctx->priv;
224 AVFilterLink *const outlink = ctx->outputs[0];
226 int nb_samples, ns, i;
227 AVFilterBufferRef *outbuf, *inbuf[SWR_CH_MAX];
228 uint8_t *ins[SWR_CH_MAX], *outs;
230 for (input_number = 0; input_number < am->nb_inputs; input_number++)
231 if (inlink == ctx->inputs[input_number])
233 av_assert1(input_number < am->nb_inputs);
234 ff_bufqueue_add(ctx, &am->in[input_number].queue, insamples);
235 am->in[input_number].nb_samples += insamples->audio->nb_samples;
236 nb_samples = am->in[0].nb_samples;
237 for (i = 1; i < am->nb_inputs; i++)
238 nb_samples = FFMIN(nb_samples, am->in[i].nb_samples);
242 outbuf = ff_get_audio_buffer(ctx->outputs[0], AV_PERM_WRITE, nb_samples);
243 outs = outbuf->data[0];
244 for (i = 0; i < am->nb_inputs; i++) {
245 inbuf[i] = ff_bufqueue_peek(&am->in[i].queue, 0);
246 ins[i] = inbuf[i]->data[0] +
247 am->in[i].pos * am->in[i].nb_ch * am->bps;
249 outbuf->pts = inbuf[0]->pts == AV_NOPTS_VALUE ? AV_NOPTS_VALUE :
251 av_rescale_q(am->in[0].pos,
252 (AVRational){ 1, ctx->inputs[0]->sample_rate },
253 ctx->outputs[0]->time_base);
255 avfilter_copy_buffer_ref_props(outbuf, inbuf[0]);
256 outbuf->audio->nb_samples = nb_samples;
257 outbuf->audio->channel_layout = outlink->channel_layout;
261 for (i = 0; i < am->nb_inputs; i++)
262 ns = FFMIN(ns, inbuf[i]->audio->nb_samples - am->in[i].pos);
263 /* Unroll the most common sample formats: speed +~350% for the loop,
264 +~13% overall (including two common decoders) */
267 copy_samples(am->nb_inputs, am->in, am->route, ins, &outs, ns, 1);
270 copy_samples(am->nb_inputs, am->in, am->route, ins, &outs, ns, 2);
273 copy_samples(am->nb_inputs, am->in, am->route, ins, &outs, ns, 4);
276 copy_samples(am->nb_inputs, am->in, am->route, ins, &outs, ns, am->bps);
281 for (i = 0; i < am->nb_inputs; i++) {
282 am->in[i].nb_samples -= ns;
284 if (am->in[i].pos == inbuf[i]->audio->nb_samples) {
286 avfilter_unref_buffer(inbuf[i]);
287 ff_bufqueue_get(&am->in[i].queue);
288 inbuf[i] = ff_bufqueue_peek(&am->in[i].queue, 0);
289 ins[i] = inbuf[i] ? inbuf[i]->data[0] : NULL;
293 ff_filter_samples(ctx->outputs[0], outbuf);
296 static av_cold int init(AVFilterContext *ctx, const char *args, void *opaque)
298 AMergeContext *am = ctx->priv;
302 am->class = &amerge_class;
303 av_opt_set_defaults(am);
304 ret = av_set_options_string(am, args, "=", ":");
306 av_log(ctx, AV_LOG_ERROR, "Error parsing options: '%s'\n", args);
309 am->in = av_calloc(am->nb_inputs, sizeof(*am->in));
311 return AVERROR(ENOMEM);
312 for (i = 0; i < am->nb_inputs; i++) {
315 .type = AVMEDIA_TYPE_AUDIO,
316 .filter_samples = filter_samples,
317 .min_perms = AV_PERM_READ | AV_PERM_PRESERVE,
319 snprintf(name, sizeof(name), "in%d", i);
320 ff_insert_inpad(ctx, i, &pad);
325 AVFilter avfilter_af_amerge = {
327 .description = NULL_IF_CONFIG_SMALL("Merge two audio streams into "
328 "a single multi-channel stream."),
329 .priv_size = sizeof(AMergeContext),
332 .query_formats = query_formats,
334 .inputs = (const AVFilterPad[]) { { .name = NULL } },
335 .outputs = (const AVFilterPad[]) {
337 .type = AVMEDIA_TYPE_AUDIO,
338 .config_props = config_output,
339 .request_frame = request_frame, },