x86: af_volume: add SSE2-optimized s16 volume scaling
[ffmpeg.git] / libavfilter / af_resample.c
1 /*
2  *
3  * This file is part of Libav.
4  *
5  * Libav is free software; you can redistribute it and/or
6  * modify it under the terms of the GNU Lesser General Public
7  * License as published by the Free Software Foundation; either
8  * version 2.1 of the License, or (at your option) any later version.
9  *
10  * Libav is distributed in the hope that it will be useful,
11  * but WITHOUT ANY WARRANTY; without even the implied warranty of
12  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
13  * Lesser General Public License for more details.
14  *
15  * You should have received a copy of the GNU Lesser General Public
16  * License along with Libav; if not, write to the Free Software
17  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
18  */
19
20 /**
21  * @file
22  * sample format and channel layout conversion audio filter
23  */
24
25 #include "libavutil/avassert.h"
26 #include "libavutil/avstring.h"
27 #include "libavutil/common.h"
28 #include "libavutil/mathematics.h"
29 #include "libavutil/opt.h"
30
31 #include "libavresample/avresample.h"
32
33 #include "audio.h"
34 #include "avfilter.h"
35 #include "formats.h"
36 #include "internal.h"
37
38 typedef struct ResampleContext {
39     AVAudioResampleContext *avr;
40
41     int64_t next_pts;
42
43     /* set by filter_frame() to signal an output frame to request_frame() */
44     int got_output;
45 } ResampleContext;
46
47 static av_cold void uninit(AVFilterContext *ctx)
48 {
49     ResampleContext *s = ctx->priv;
50
51     if (s->avr) {
52         avresample_close(s->avr);
53         avresample_free(&s->avr);
54     }
55 }
56
57 static int query_formats(AVFilterContext *ctx)
58 {
59     AVFilterLink *inlink  = ctx->inputs[0];
60     AVFilterLink *outlink = ctx->outputs[0];
61
62     AVFilterFormats        *in_formats      = ff_all_formats(AVMEDIA_TYPE_AUDIO);
63     AVFilterFormats        *out_formats     = ff_all_formats(AVMEDIA_TYPE_AUDIO);
64     AVFilterFormats        *in_samplerates  = ff_all_samplerates();
65     AVFilterFormats        *out_samplerates = ff_all_samplerates();
66     AVFilterChannelLayouts *in_layouts      = ff_all_channel_layouts();
67     AVFilterChannelLayouts *out_layouts     = ff_all_channel_layouts();
68
69     ff_formats_ref(in_formats,  &inlink->out_formats);
70     ff_formats_ref(out_formats, &outlink->in_formats);
71
72     ff_formats_ref(in_samplerates,  &inlink->out_samplerates);
73     ff_formats_ref(out_samplerates, &outlink->in_samplerates);
74
75     ff_channel_layouts_ref(in_layouts,  &inlink->out_channel_layouts);
76     ff_channel_layouts_ref(out_layouts, &outlink->in_channel_layouts);
77
78     return 0;
79 }
80
81 static int config_output(AVFilterLink *outlink)
82 {
83     AVFilterContext *ctx = outlink->src;
84     AVFilterLink *inlink = ctx->inputs[0];
85     ResampleContext   *s = ctx->priv;
86     char buf1[64], buf2[64];
87     int ret;
88
89     if (s->avr) {
90         avresample_close(s->avr);
91         avresample_free(&s->avr);
92     }
93
94     if (inlink->channel_layout == outlink->channel_layout &&
95         inlink->sample_rate    == outlink->sample_rate    &&
96         (inlink->format        == outlink->format ||
97         (av_get_channel_layout_nb_channels(inlink->channel_layout)  == 1 &&
98          av_get_channel_layout_nb_channels(outlink->channel_layout) == 1 &&
99          av_get_planar_sample_fmt(inlink->format) ==
100          av_get_planar_sample_fmt(outlink->format))))
101         return 0;
102
103     if (!(s->avr = avresample_alloc_context()))
104         return AVERROR(ENOMEM);
105
106     av_opt_set_int(s->avr,  "in_channel_layout", inlink ->channel_layout, 0);
107     av_opt_set_int(s->avr, "out_channel_layout", outlink->channel_layout, 0);
108     av_opt_set_int(s->avr,  "in_sample_fmt",     inlink ->format,         0);
109     av_opt_set_int(s->avr, "out_sample_fmt",     outlink->format,         0);
110     av_opt_set_int(s->avr,  "in_sample_rate",    inlink ->sample_rate,    0);
111     av_opt_set_int(s->avr, "out_sample_rate",    outlink->sample_rate,    0);
112
113     if ((ret = avresample_open(s->avr)) < 0)
114         return ret;
115
116     outlink->time_base = (AVRational){ 1, outlink->sample_rate };
117     s->next_pts        = AV_NOPTS_VALUE;
118
119     av_get_channel_layout_string(buf1, sizeof(buf1),
120                                  -1, inlink ->channel_layout);
121     av_get_channel_layout_string(buf2, sizeof(buf2),
122                                  -1, outlink->channel_layout);
123     av_log(ctx, AV_LOG_VERBOSE,
124            "fmt:%s srate:%d cl:%s -> fmt:%s srate:%d cl:%s\n",
125            av_get_sample_fmt_name(inlink ->format), inlink ->sample_rate, buf1,
126            av_get_sample_fmt_name(outlink->format), outlink->sample_rate, buf2);
127
128     return 0;
129 }
130
131 static int request_frame(AVFilterLink *outlink)
132 {
133     AVFilterContext *ctx = outlink->src;
134     ResampleContext   *s = ctx->priv;
135     int ret = 0;
136
137     s->got_output = 0;
138     while (ret >= 0 && !s->got_output)
139         ret = ff_request_frame(ctx->inputs[0]);
140
141     /* flush the lavr delay buffer */
142     if (ret == AVERROR_EOF && s->avr) {
143         AVFilterBufferRef *buf;
144         int nb_samples = av_rescale_rnd(avresample_get_delay(s->avr),
145                                         outlink->sample_rate,
146                                         ctx->inputs[0]->sample_rate,
147                                         AV_ROUND_UP);
148
149         if (!nb_samples)
150             return ret;
151
152         buf = ff_get_audio_buffer(outlink, AV_PERM_WRITE, nb_samples);
153         if (!buf)
154             return AVERROR(ENOMEM);
155
156         ret = avresample_convert(s->avr, buf->extended_data,
157                                  buf->linesize[0], nb_samples,
158                                  NULL, 0, 0);
159         if (ret <= 0) {
160             avfilter_unref_buffer(buf);
161             return (ret == 0) ? AVERROR_EOF : ret;
162         }
163
164         buf->pts = s->next_pts;
165         return ff_filter_frame(outlink, buf);
166     }
167     return ret;
168 }
169
170 static int filter_frame(AVFilterLink *inlink, AVFilterBufferRef *buf)
171 {
172     AVFilterContext  *ctx = inlink->dst;
173     ResampleContext    *s = ctx->priv;
174     AVFilterLink *outlink = ctx->outputs[0];
175     int ret;
176
177     if (s->avr) {
178         AVFilterBufferRef *buf_out;
179         int delay, nb_samples;
180
181         /* maximum possible samples lavr can output */
182         delay      = avresample_get_delay(s->avr);
183         nb_samples = av_rescale_rnd(buf->audio->nb_samples + delay,
184                                     outlink->sample_rate, inlink->sample_rate,
185                                     AV_ROUND_UP);
186
187         buf_out = ff_get_audio_buffer(outlink, AV_PERM_WRITE, nb_samples);
188         if (!buf_out) {
189             ret = AVERROR(ENOMEM);
190             goto fail;
191         }
192
193         ret     = avresample_convert(s->avr, buf_out->extended_data,
194                                      buf_out->linesize[0], nb_samples,
195                                      buf->extended_data, buf->linesize[0],
196                                      buf->audio->nb_samples);
197         if (ret <= 0) {
198             avfilter_unref_buffer(buf_out);
199             if (ret < 0)
200                 goto fail;
201         }
202
203         av_assert0(!avresample_available(s->avr));
204
205         if (s->next_pts == AV_NOPTS_VALUE) {
206             if (buf->pts == AV_NOPTS_VALUE) {
207                 av_log(ctx, AV_LOG_WARNING, "First timestamp is missing, "
208                        "assuming 0.\n");
209                 s->next_pts = 0;
210             } else
211                 s->next_pts = av_rescale_q(buf->pts, inlink->time_base,
212                                            outlink->time_base);
213         }
214
215         if (ret > 0) {
216             buf_out->audio->nb_samples = ret;
217             if (buf->pts != AV_NOPTS_VALUE) {
218                 buf_out->pts = av_rescale_q(buf->pts, inlink->time_base,
219                                             outlink->time_base) -
220                                av_rescale(delay, outlink->sample_rate,
221                                           inlink->sample_rate);
222             } else
223                 buf_out->pts = s->next_pts;
224
225             s->next_pts = buf_out->pts + buf_out->audio->nb_samples;
226
227             ret = ff_filter_frame(outlink, buf_out);
228             s->got_output = 1;
229         }
230
231 fail:
232         avfilter_unref_buffer(buf);
233     } else {
234         buf->format = outlink->format;
235         ret = ff_filter_frame(outlink, buf);
236         s->got_output = 1;
237     }
238
239     return ret;
240 }
241
242 static const AVFilterPad avfilter_af_resample_inputs[] = {
243     {
244         .name           = "default",
245         .type           = AVMEDIA_TYPE_AUDIO,
246         .filter_frame   = filter_frame,
247         .min_perms      = AV_PERM_READ
248     },
249     { NULL }
250 };
251
252 static const AVFilterPad avfilter_af_resample_outputs[] = {
253     {
254         .name          = "default",
255         .type          = AVMEDIA_TYPE_AUDIO,
256         .config_props  = config_output,
257         .request_frame = request_frame
258     },
259     { NULL }
260 };
261
262 AVFilter avfilter_af_resample = {
263     .name          = "resample",
264     .description   = NULL_IF_CONFIG_SMALL("Audio resampling and conversion."),
265     .priv_size     = sizeof(ResampleContext),
266
267     .uninit         = uninit,
268     .query_formats  = query_formats,
269
270     .inputs    = avfilter_af_resample_inputs,
271     .outputs   = avfilter_af_resample_outputs,
272 };