1/*
2 * samplerate conversion for both audio and video
3 * Copyright (c) 2000 Fabrice Bellard
4 *
5 * This file is part of FFmpeg.
6 *
7 * FFmpeg is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
10 * version 2.1 of the License, or (at your option) any later version.
11 *
12 * FFmpeg is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
15 * Lesser General Public License for more details.
16 *
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with FFmpeg; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20 */
21
22/**
23 * @file
24 * samplerate conversion for both audio and video
25 */
26
27#include "avcodec.h"
28#include "audioconvert.h"
29#include "opt.h"
30
31struct AVResampleContext;
32
33static const char *context_to_name(void *ptr)
34{
35    return "audioresample";
36}
37
38static const AVOption options[] = {{NULL}};
39static const AVClass audioresample_context_class = { "ReSampleContext", context_to_name, options, LIBAVUTIL_VERSION_INT };
40
41struct ReSampleContext {
42    struct AVResampleContext *resample_context;
43    short *temp[2];
44    int temp_len;
45    float ratio;
46    /* channel convert */
47    int input_channels, output_channels, filter_channels;
48    AVAudioConvert *convert_ctx[2];
49    enum SampleFormat sample_fmt[2]; ///< input and output sample format
50    unsigned sample_size[2];         ///< size of one sample in sample_fmt
51    short *buffer[2];                ///< buffers used for conversion to S16
52    unsigned buffer_size[2];         ///< sizes of allocated buffers
53};
54
55/* n1: number of samples */
56static void stereo_to_mono(short *output, short *input, int n1)
57{
58    short *p, *q;
59    int n = n1;
60
61    p = input;
62    q = output;
63    while (n >= 4) {
64        q[0] = (p[0] + p[1]) >> 1;
65        q[1] = (p[2] + p[3]) >> 1;
66        q[2] = (p[4] + p[5]) >> 1;
67        q[3] = (p[6] + p[7]) >> 1;
68        q += 4;
69        p += 8;
70        n -= 4;
71    }
72    while (n > 0) {
73        q[0] = (p[0] + p[1]) >> 1;
74        q++;
75        p += 2;
76        n--;
77    }
78}
79
80/* n1: number of samples */
81static void mono_to_stereo(short *output, short *input, int n1)
82{
83    short *p, *q;
84    int n = n1;
85    int v;
86
87    p = input;
88    q = output;
89    while (n >= 4) {
90        v = p[0]; q[0] = v; q[1] = v;
91        v = p[1]; q[2] = v; q[3] = v;
92        v = p[2]; q[4] = v; q[5] = v;
93        v = p[3]; q[6] = v; q[7] = v;
94        q += 8;
95        p += 4;
96        n -= 4;
97    }
98    while (n > 0) {
99        v = p[0]; q[0] = v; q[1] = v;
100        q += 2;
101        p += 1;
102        n--;
103    }
104}
105
106/* XXX: should use more abstract 'N' channels system */
107static void stereo_split(short *output1, short *output2, short *input, int n)
108{
109    int i;
110
111    for(i=0;i<n;i++) {
112        *output1++ = *input++;
113        *output2++ = *input++;
114    }
115}
116
117static void stereo_mux(short *output, short *input1, short *input2, int n)
118{
119    int i;
120
121    for(i=0;i<n;i++) {
122        *output++ = *input1++;
123        *output++ = *input2++;
124    }
125}
126
127static void ac3_5p1_mux(short *output, short *input1, short *input2, int n)
128{
129    int i;
130    short l,r;
131
132    for(i=0;i<n;i++) {
133      l=*input1++;
134      r=*input2++;
135      *output++ = l;           /* left */
136      *output++ = (l/2)+(r/2); /* center */
137      *output++ = r;           /* right */
138      *output++ = 0;           /* left surround */
139      *output++ = 0;           /* right surroud */
140      *output++ = 0;           /* low freq */
141    }
142}
143
144ReSampleContext *av_audio_resample_init(int output_channels, int input_channels,
145                                        int output_rate, int input_rate,
146                                        enum SampleFormat sample_fmt_out,
147                                        enum SampleFormat sample_fmt_in,
148                                        int filter_length, int log2_phase_count,
149                                        int linear, double cutoff)
150{
151    ReSampleContext *s;
152
153    if ( input_channels > 2)
154      {
155        av_log(NULL, AV_LOG_ERROR, "Resampling with input channels greater than 2 unsupported.\n");
156        return NULL;
157      }
158
159    s = av_mallocz(sizeof(ReSampleContext));
160    if (!s)
161      {
162        av_log(NULL, AV_LOG_ERROR, "Can't allocate memory for resample context.\n");
163        return NULL;
164      }
165
166    s->ratio = (float)output_rate / (float)input_rate;
167
168    s->input_channels = input_channels;
169    s->output_channels = output_channels;
170
171    s->filter_channels = s->input_channels;
172    if (s->output_channels < s->filter_channels)
173        s->filter_channels = s->output_channels;
174
175    s->sample_fmt [0] = sample_fmt_in;
176    s->sample_fmt [1] = sample_fmt_out;
177    s->sample_size[0] = av_get_bits_per_sample_format(s->sample_fmt[0])>>3;
178    s->sample_size[1] = av_get_bits_per_sample_format(s->sample_fmt[1])>>3;
179
180    if (s->sample_fmt[0] != SAMPLE_FMT_S16) {
181        if (!(s->convert_ctx[0] = av_audio_convert_alloc(SAMPLE_FMT_S16, 1,
182                                                         s->sample_fmt[0], 1, NULL, 0))) {
183            av_log(s, AV_LOG_ERROR,
184                   "Cannot convert %s sample format to s16 sample format\n",
185                   avcodec_get_sample_fmt_name(s->sample_fmt[0]));
186            av_free(s);
187            return NULL;
188        }
189    }
190
191    if (s->sample_fmt[1] != SAMPLE_FMT_S16) {
192        if (!(s->convert_ctx[1] = av_audio_convert_alloc(s->sample_fmt[1], 1,
193                                                         SAMPLE_FMT_S16, 1, NULL, 0))) {
194            av_log(s, AV_LOG_ERROR,
195                   "Cannot convert s16 sample format to %s sample format\n",
196                   avcodec_get_sample_fmt_name(s->sample_fmt[1]));
197            av_audio_convert_free(s->convert_ctx[0]);
198            av_free(s);
199            return NULL;
200        }
201    }
202
203/*
204 * AC-3 output is the only case where filter_channels could be greater than 2.
205 * input channels can't be greater than 2, so resample the 2 channels and then
206 * expand to 6 channels after the resampling.
207 */
208    if(s->filter_channels>2)
209      s->filter_channels = 2;
210
211#define TAPS 16
212    s->resample_context= av_resample_init(output_rate, input_rate,
213                         filter_length, log2_phase_count, linear, cutoff);
214
215    *(const AVClass**)s->resample_context = &audioresample_context_class;
216
217    return s;
218}
219
220#if LIBAVCODEC_VERSION_MAJOR < 53
221ReSampleContext *audio_resample_init(int output_channels, int input_channels,
222                                     int output_rate, int input_rate)
223{
224    return av_audio_resample_init(output_channels, input_channels,
225                                  output_rate, input_rate,
226                                  SAMPLE_FMT_S16, SAMPLE_FMT_S16,
227                                  TAPS, 10, 0, 0.8);
228}
229#endif
230
231/* resample audio. 'nb_samples' is the number of input samples */
232/* XXX: optimize it ! */
233int audio_resample(ReSampleContext *s, short *output, short *input, int nb_samples)
234{
235    int i, nb_samples1;
236    short *bufin[2];
237    short *bufout[2];
238    short *buftmp2[2], *buftmp3[2];
239    short *output_bak = NULL;
240    int lenout;
241
242    if (s->input_channels == s->output_channels && s->ratio == 1.0 && 0) {
243        /* nothing to do */
244        memcpy(output, input, nb_samples * s->input_channels * sizeof(short));
245        return nb_samples;
246    }
247
248    if (s->sample_fmt[0] != SAMPLE_FMT_S16) {
249        int istride[1] = { s->sample_size[0] };
250        int ostride[1] = { 2 };
251        const void *ibuf[1] = { input };
252        void       *obuf[1];
253        unsigned input_size = nb_samples*s->input_channels*2;
254
255        if (!s->buffer_size[0] || s->buffer_size[0] < input_size) {
256            av_free(s->buffer[0]);
257            s->buffer_size[0] = input_size;
258            s->buffer[0] = av_malloc(s->buffer_size[0]);
259            if (!s->buffer[0]) {
260                av_log(s->resample_context, AV_LOG_ERROR, "Could not allocate buffer\n");
261                return 0;
262            }
263        }
264
265        obuf[0] = s->buffer[0];
266
267        if (av_audio_convert(s->convert_ctx[0], obuf, ostride,
268                             ibuf, istride, nb_samples*s->input_channels) < 0) {
269            av_log(s->resample_context, AV_LOG_ERROR, "Audio sample format conversion failed\n");
270            return 0;
271        }
272
273        input  = s->buffer[0];
274    }
275
276    lenout= 4*nb_samples * s->ratio + 16;
277
278    if (s->sample_fmt[1] != SAMPLE_FMT_S16) {
279        output_bak = output;
280
281        if (!s->buffer_size[1] || s->buffer_size[1] < lenout) {
282            av_free(s->buffer[1]);
283            s->buffer_size[1] = lenout;
284            s->buffer[1] = av_malloc(s->buffer_size[1]);
285            if (!s->buffer[1]) {
286                av_log(s->resample_context, AV_LOG_ERROR, "Could not allocate buffer\n");
287                return 0;
288            }
289        }
290
291        output = s->buffer[1];
292    }
293
294    /* XXX: move those malloc to resample init code */
295    for(i=0; i<s->filter_channels; i++){
296        bufin[i]= av_malloc( (nb_samples + s->temp_len) * sizeof(short) );
297        memcpy(bufin[i], s->temp[i], s->temp_len * sizeof(short));
298        buftmp2[i] = bufin[i] + s->temp_len;
299    }
300
301    /* make some zoom to avoid round pb */
302    bufout[0]= av_malloc( lenout * sizeof(short) );
303    bufout[1]= av_malloc( lenout * sizeof(short) );
304
305    if (s->input_channels == 2 &&
306        s->output_channels == 1) {
307        buftmp3[0] = output;
308        stereo_to_mono(buftmp2[0], input, nb_samples);
309    } else if (s->output_channels >= 2 && s->input_channels == 1) {
310        buftmp3[0] = bufout[0];
311        memcpy(buftmp2[0], input, nb_samples*sizeof(short));
312    } else if (s->output_channels >= 2) {
313        buftmp3[0] = bufout[0];
314        buftmp3[1] = bufout[1];
315        stereo_split(buftmp2[0], buftmp2[1], input, nb_samples);
316    } else {
317        buftmp3[0] = output;
318        memcpy(buftmp2[0], input, nb_samples*sizeof(short));
319    }
320
321    nb_samples += s->temp_len;
322
323    /* resample each channel */
324    nb_samples1 = 0; /* avoid warning */
325    for(i=0;i<s->filter_channels;i++) {
326        int consumed;
327        int is_last= i+1 == s->filter_channels;
328
329        nb_samples1 = av_resample(s->resample_context, buftmp3[i], bufin[i], &consumed, nb_samples, lenout, is_last);
330        s->temp_len= nb_samples - consumed;
331        s->temp[i]= av_realloc(s->temp[i], s->temp_len*sizeof(short));
332        memcpy(s->temp[i], bufin[i] + consumed, s->temp_len*sizeof(short));
333    }
334
335    if (s->output_channels == 2 && s->input_channels == 1) {
336        mono_to_stereo(output, buftmp3[0], nb_samples1);
337    } else if (s->output_channels == 2) {
338        stereo_mux(output, buftmp3[0], buftmp3[1], nb_samples1);
339    } else if (s->output_channels == 6) {
340        ac3_5p1_mux(output, buftmp3[0], buftmp3[1], nb_samples1);
341    }
342
343    if (s->sample_fmt[1] != SAMPLE_FMT_S16) {
344        int istride[1] = { 2 };
345        int ostride[1] = { s->sample_size[1] };
346        const void *ibuf[1] = { output };
347        void       *obuf[1] = { output_bak };
348
349        if (av_audio_convert(s->convert_ctx[1], obuf, ostride,
350                             ibuf, istride, nb_samples1*s->output_channels) < 0) {
351            av_log(s->resample_context, AV_LOG_ERROR, "Audio sample format convertion failed\n");
352            return 0;
353        }
354    }
355
356    for(i=0; i<s->filter_channels; i++)
357        av_free(bufin[i]);
358
359    av_free(bufout[0]);
360    av_free(bufout[1]);
361    return nb_samples1;
362}
363
364void audio_resample_close(ReSampleContext *s)
365{
366    av_resample_close(s->resample_context);
367    av_freep(&s->temp[0]);
368    av_freep(&s->temp[1]);
369    av_freep(&s->buffer[0]);
370    av_freep(&s->buffer[1]);
371    av_audio_convert_free(s->convert_ctx[0]);
372    av_audio_convert_free(s->convert_ctx[1]);
373    av_free(s);
374}
375