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/avstring.h"
27 #include "libavutil/bprint.h"
28 #include "libavutil/channel_layout.h"
29 #include "libavutil/opt.h"
32 #include "bufferqueue.h"
40 int route[SWR_CH_MAX]; /**< channels routing, see copy_samples */
43 struct FFBufQueue queue;
44 int nb_ch; /**< number of channels for the input */
50 #define OFFSET(x) offsetof(AMergeContext, x)
51 #define FLAGS AV_OPT_FLAG_AUDIO_PARAM|AV_OPT_FLAG_FILTERING_PARAM
53 static const AVOption amerge_options[] = {
54 { "inputs", "specify the number of inputs", OFFSET(nb_inputs),
55 AV_OPT_TYPE_INT, { .i64 = 2 }, 1, SWR_CH_MAX, FLAGS },
59 AVFILTER_DEFINE_CLASS(amerge);
61 static av_cold void uninit(AVFilterContext *ctx)
63 AMergeContext *s = ctx->priv;
66 for (i = 0; i < s->nb_inputs; i++) {
68 ff_bufqueue_discard_all(&s->in[i].queue);
70 av_freep(&ctx->input_pads[i].name);
75 static int query_formats(AVFilterContext *ctx)
77 AMergeContext *s = ctx->priv;
78 int64_t inlayout[SWR_CH_MAX], outlayout = 0;
79 AVFilterFormats *formats;
80 AVFilterChannelLayouts *layouts;
81 int i, ret, overlap = 0, nb_ch = 0;
83 for (i = 0; i < s->nb_inputs; i++) {
84 if (!ctx->inputs[i]->in_channel_layouts ||
85 !ctx->inputs[i]->in_channel_layouts->nb_channel_layouts) {
86 av_log(ctx, AV_LOG_WARNING,
87 "No channel layout for input %d\n", i + 1);
88 return AVERROR(EAGAIN);
90 inlayout[i] = ctx->inputs[i]->in_channel_layouts->channel_layouts[0];
91 if (ctx->inputs[i]->in_channel_layouts->nb_channel_layouts > 1) {
93 av_get_channel_layout_string(buf, sizeof(buf), 0, inlayout[i]);
94 av_log(ctx, AV_LOG_INFO, "Using \"%s\" for input %d\n", buf, i + 1);
96 s->in[i].nb_ch = av_get_channel_layout_nb_channels(inlayout[i]);
97 if (outlayout & inlayout[i])
99 outlayout |= inlayout[i];
100 nb_ch += s->in[i].nb_ch;
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);
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++)
112 outlayout = av_get_default_channel_layout(nb_ch);
113 if (!outlayout && nb_ch)
114 outlayout = 0xFFFFFFFFFFFFFFFFULL >> (64 - nb_ch);
116 int *route[SWR_CH_MAX];
117 int c, out_ch_number = 0;
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++;
127 formats = ff_make_format_list(ff_packed_sample_fmts_array);
128 if ((ret = ff_set_common_formats(ctx, formats)) < 0)
130 for (i = 0; i < s->nb_inputs; i++) {
132 if ((ret = ff_add_channel_layout(&layouts, inlayout[i])) < 0)
134 if ((ret = ff_channel_layouts_ref(layouts, &ctx->inputs[i]->out_channel_layouts)) < 0)
138 if ((ret = ff_add_channel_layout(&layouts, outlayout)) < 0)
140 if ((ret = ff_channel_layouts_ref(layouts, &ctx->outputs[0]->in_channel_layouts)) < 0)
143 return ff_set_common_samplerates(ctx, ff_all_samplerates());
146 static int config_output(AVFilterLink *outlink)
148 AVFilterContext *ctx = outlink->src;
149 AMergeContext *s = ctx->priv;
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);
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;
166 av_bprint_init(&bp, 0, 1);
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);
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);
178 static int request_frame(AVFilterLink *outlink)
180 AVFilterContext *ctx = outlink->src;
181 AMergeContext *s = ctx->priv;
184 for (i = 0; i < s->nb_inputs; i++)
185 if (!s->in[i].nb_samples)
186 if ((ret = ff_request_frame(ctx->inputs[i])) < 0)
192 * Copy samples from several input streams to one output stream.
193 * @param nb_inputs number of inputs
194 * @param in inputs; used only for the nb_ch field;
195 * @param route routing values;
196 * input channel i goes to output channel route[i];
197 * i < in[0].nb_ch are the channels from the first output;
198 * i >= in[0].nb_ch are the channels from the second output
199 * @param ins pointer to the samples of each inputs, in packed format;
200 * will be left at the end of the copied samples
201 * @param outs pointer to the samples of the output, in packet format;
202 * must point to a buffer big enough;
203 * will be left at the end of the copied samples
204 * @param ns number of samples to copy
205 * @param bps bytes per sample
207 static inline void copy_samples(int nb_inputs, struct amerge_input in[],
208 int *route, uint8_t *ins[],
209 uint8_t **outs, int ns, int bps)
214 for (i = 0; i < nb_inputs; i++)
215 nb_ch += in[i].nb_ch;
218 for (i = 0; i < nb_inputs; i++) {
219 for (c = 0; c < in[i].nb_ch; c++) {
220 memcpy((*outs) + bps * *(route_cur++), ins[i], bps);
224 *outs += nb_ch * bps;
228 static int filter_frame(AVFilterLink *inlink, AVFrame *insamples)
230 AVFilterContext *ctx = inlink->dst;
231 AMergeContext *s = ctx->priv;
232 AVFilterLink *const outlink = ctx->outputs[0];
234 int nb_samples, ns, i;
235 AVFrame *outbuf, *inbuf[SWR_CH_MAX];
236 uint8_t *ins[SWR_CH_MAX], *outs;
238 for (input_number = 0; input_number < s->nb_inputs; input_number++)
239 if (inlink == ctx->inputs[input_number])
241 av_assert1(input_number < s->nb_inputs);
242 if (ff_bufqueue_is_full(&s->in[input_number].queue)) {
243 av_frame_free(&insamples);
244 return AVERROR(ENOMEM);
246 ff_bufqueue_add(ctx, &s->in[input_number].queue, av_frame_clone(insamples));
247 s->in[input_number].nb_samples += insamples->nb_samples;
248 av_frame_free(&insamples);
249 nb_samples = s->in[0].nb_samples;
250 for (i = 1; i < s->nb_inputs; i++)
251 nb_samples = FFMIN(nb_samples, s->in[i].nb_samples);
255 outbuf = ff_get_audio_buffer(ctx->outputs[0], nb_samples);
257 return AVERROR(ENOMEM);
258 outs = outbuf->data[0];
259 for (i = 0; i < s->nb_inputs; i++) {
260 inbuf[i] = ff_bufqueue_peek(&s->in[i].queue, 0);
261 ins[i] = inbuf[i]->data[0] +
262 s->in[i].pos * s->in[i].nb_ch * s->bps;
264 av_frame_copy_props(outbuf, inbuf[0]);
265 outbuf->pts = inbuf[0]->pts == AV_NOPTS_VALUE ? AV_NOPTS_VALUE :
267 av_rescale_q(s->in[0].pos,
268 av_make_q(1, ctx->inputs[0]->sample_rate),
269 ctx->outputs[0]->time_base);
271 outbuf->nb_samples = nb_samples;
272 outbuf->channel_layout = outlink->channel_layout;
273 av_frame_set_channels(outbuf, outlink->channels);
277 for (i = 0; i < s->nb_inputs; i++)
278 ns = FFMIN(ns, inbuf[i]->nb_samples - s->in[i].pos);
279 /* Unroll the most common sample formats: speed +~350% for the loop,
280 +~13% overall (including two common decoders) */
283 copy_samples(s->nb_inputs, s->in, s->route, ins, &outs, ns, 1);
286 copy_samples(s->nb_inputs, s->in, s->route, ins, &outs, ns, 2);
289 copy_samples(s->nb_inputs, s->in, s->route, ins, &outs, ns, 4);
292 copy_samples(s->nb_inputs, s->in, s->route, ins, &outs, ns, s->bps);
297 for (i = 0; i < s->nb_inputs; i++) {
298 s->in[i].nb_samples -= ns;
300 if (s->in[i].pos == inbuf[i]->nb_samples) {
302 av_frame_free(&inbuf[i]);
303 ff_bufqueue_get(&s->in[i].queue);
304 inbuf[i] = ff_bufqueue_peek(&s->in[i].queue, 0);
305 ins[i] = inbuf[i] ? inbuf[i]->data[0] : NULL;
309 return ff_filter_frame(ctx->outputs[0], outbuf);
312 static av_cold int init(AVFilterContext *ctx)
314 AMergeContext *s = ctx->priv;
317 s->in = av_calloc(s->nb_inputs, sizeof(*s->in));
319 return AVERROR(ENOMEM);
320 for (i = 0; i < s->nb_inputs; i++) {
321 char *name = av_asprintf("in%d", i);
324 .type = AVMEDIA_TYPE_AUDIO,
325 .filter_frame = filter_frame,
328 return AVERROR(ENOMEM);
329 ff_insert_inpad(ctx, i, &pad);
334 static const AVFilterPad amerge_outputs[] = {
337 .type = AVMEDIA_TYPE_AUDIO,
338 .config_props = config_output,
339 .request_frame = request_frame,
344 AVFilter ff_af_amerge = {
346 .description = NULL_IF_CONFIG_SMALL("Merge two or more audio streams into "
347 "a single multi-channel stream."),
348 .priv_size = sizeof(AMergeContext),
351 .query_formats = query_formats,
353 .outputs = amerge_outputs,
354 .priv_class = &amerge_class,
355 .flags = AVFILTER_FLAG_DYNAMIC_INPUTS,