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 <string.h>
28
29 #ifdef IDE_COMPILE
30 #include "libavutil/internal.h"
31 #endif
32
33 #include "avcodec.h"
34 #include "audioconvert.h"
35 #include "libavutil/opt.h"
36 #include "libavutil/mem.h"
37 #include "libavutil/samplefmt.h"
38
39 #if FF_API_AVCODEC_RESAMPLE
40
41 #define MAX_CHANNELS 8
42
43 struct AVResampleContext;
44
context_to_name(void * ptr)45 static const char *context_to_name(void *ptr)
46 {
47 return "audioresample";
48 }
49
50 static const AVOption options[] = {{NULL}};
51 static const AVClass audioresample_context_class = {
52 "ReSampleContext", context_to_name, options, LIBAVUTIL_VERSION_INT
53 };
54
55 struct ReSampleContext {
56 struct AVResampleContext *resample_context;
57 short *temp[MAX_CHANNELS];
58 int temp_len;
59 float ratio;
60 /* channel convert */
61 int input_channels, output_channels, filter_channels;
62 AVAudioConvert *convert_ctx[2];
63 enum AVSampleFormat sample_fmt[2]; ///< input and output sample format
64 unsigned sample_size[2]; ///< size of one sample in sample_fmt
65 short *buffer[2]; ///< buffers used for conversion to S16
66 unsigned buffer_size[2]; ///< sizes of allocated buffers
67 };
68
69 /* n1: number of samples */
stereo_to_mono(short * output,short * input,int n1)70 static void stereo_to_mono(short *output, short *input, int n1)
71 {
72 short *p, *q;
73 int n = n1;
74
75 p = input;
76 q = output;
77 while (n >= 4) {
78 q[0] = (p[0] + p[1]) >> 1;
79 q[1] = (p[2] + p[3]) >> 1;
80 q[2] = (p[4] + p[5]) >> 1;
81 q[3] = (p[6] + p[7]) >> 1;
82 q += 4;
83 p += 8;
84 n -= 4;
85 }
86 while (n > 0) {
87 q[0] = (p[0] + p[1]) >> 1;
88 q++;
89 p += 2;
90 n--;
91 }
92 }
93
94 /* n1: number of samples */
mono_to_stereo(short * output,short * input,int n1)95 static void mono_to_stereo(short *output, short *input, int n1)
96 {
97 short *p, *q;
98 int n = n1;
99 int v;
100
101 p = input;
102 q = output;
103 while (n >= 4) {
104 v = p[0]; q[0] = v; q[1] = v;
105 v = p[1]; q[2] = v; q[3] = v;
106 v = p[2]; q[4] = v; q[5] = v;
107 v = p[3]; q[6] = v; q[7] = v;
108 q += 8;
109 p += 4;
110 n -= 4;
111 }
112 while (n > 0) {
113 v = p[0]; q[0] = v; q[1] = v;
114 q += 2;
115 p += 1;
116 n--;
117 }
118 }
119
120 /*
121 5.1 to stereo input: [fl, fr, c, lfe, rl, rr]
122 - Left = front_left + rear_gain * rear_left + center_gain * center
123 - Right = front_right + rear_gain * rear_right + center_gain * center
124 Where rear_gain is usually around 0.5-1.0 and
125 center_gain is almost always 0.7 (-3 dB)
126 */
surround_to_stereo(short ** output,short * input,int channels,int samples)127 static void surround_to_stereo(short **output, short *input, int channels, int samples)
128 {
129 int i;
130 short l, r;
131
132 for (i = 0; i < samples; i++) {
133 int fl,fr,c,rl,rr;
134 fl = input[0];
135 fr = input[1];
136 c = input[2];
137 // lfe = input[3];
138 rl = input[4];
139 rr = input[5];
140
141 l = av_clip_int16(fl + (0.5 * rl) + (0.7 * c));
142 r = av_clip_int16(fr + (0.5 * rr) + (0.7 * c));
143
144 /* output l & r. */
145 *output[0]++ = l;
146 *output[1]++ = r;
147
148 /* increment input. */
149 input += channels;
150 }
151 }
152
deinterleave(short ** output,short * input,int channels,int samples)153 static void deinterleave(short **output, short *input, int channels, int samples)
154 {
155 int i, j;
156
157 for (i = 0; i < samples; i++) {
158 for (j = 0; j < channels; j++) {
159 *output[j]++ = *input++;
160 }
161 }
162 }
163
interleave(short * output,short ** input,int channels,int samples)164 static void interleave(short *output, short **input, int channels, int samples)
165 {
166 int i, j;
167
168 for (i = 0; i < samples; i++) {
169 for (j = 0; j < channels; j++) {
170 *output++ = *input[j]++;
171 }
172 }
173 }
174
ac3_5p1_mux(short * output,short * input1,short * input2,int n)175 static void ac3_5p1_mux(short *output, short *input1, short *input2, int n)
176 {
177 int i;
178 short l, r;
179
180 for (i = 0; i < n; i++) {
181 l = *input1++;
182 r = *input2++;
183 *output++ = l; /* left */
184 *output++ = (l / 2) + (r / 2); /* center */
185 *output++ = r; /* right */
186 *output++ = 0; /* left surround */
187 *output++ = 0; /* right surroud */
188 *output++ = 0; /* low freq */
189 }
190 }
191
192 #define SUPPORT_RESAMPLE(ch1, ch2, ch3, ch4, ch5, ch6, ch7, ch8) \
193 ch8<<7 | ch7<<6 | ch6<<5 | ch5<<4 | ch4<<3 | ch3<<2 | ch2<<1 | ch1<<0
194
195 static const uint8_t supported_resampling[MAX_CHANNELS] = {
196 // output ch: 1 2 3 4 5 6 7 8
197 SUPPORT_RESAMPLE(1, 1, 0, 0, 0, 0, 0, 0), // 1 input channel
198 SUPPORT_RESAMPLE(1, 1, 0, 0, 0, 1, 0, 0), // 2 input channels
199 SUPPORT_RESAMPLE(0, 0, 1, 0, 0, 0, 0, 0), // 3 input channels
200 SUPPORT_RESAMPLE(0, 0, 0, 1, 0, 0, 0, 0), // 4 input channels
201 SUPPORT_RESAMPLE(0, 0, 0, 0, 1, 0, 0, 0), // 5 input channels
202 SUPPORT_RESAMPLE(0, 1, 0, 0, 0, 1, 0, 0), // 6 input channels
203 SUPPORT_RESAMPLE(0, 0, 0, 0, 0, 0, 1, 0), // 7 input channels
204 SUPPORT_RESAMPLE(0, 0, 0, 0, 0, 0, 0, 1), // 8 input channels
205 };
206
av_audio_resample_init(int output_channels,int input_channels,int output_rate,int input_rate,enum AVSampleFormat sample_fmt_out,enum AVSampleFormat sample_fmt_in,int filter_length,int log2_phase_count,int linear,double cutoff)207 ReSampleContext *av_audio_resample_init(int output_channels, int input_channels,
208 int output_rate, int input_rate,
209 enum AVSampleFormat sample_fmt_out,
210 enum AVSampleFormat sample_fmt_in,
211 int filter_length, int log2_phase_count,
212 int linear, double cutoff)
213 {
214 ReSampleContext *s;
215
216 if (input_channels > MAX_CHANNELS) {
217 av_log(NULL, AV_LOG_ERROR,
218 "Resampling with input channels greater than %d is unsupported.\n",
219 MAX_CHANNELS);
220 return NULL;
221 }
222 if (!(supported_resampling[input_channels-1] & (1<<(output_channels-1)))) {
223 int i;
224 av_log(NULL, AV_LOG_ERROR, "Unsupported audio resampling. Allowed "
225 "output channels for %d input channel%s", input_channels,
226 input_channels > 1 ? "s:" : ":");
227 for (i = 0; i < MAX_CHANNELS; i++)
228 if (supported_resampling[input_channels-1] & (1<<i))
229 av_log(NULL, AV_LOG_ERROR, " %d", i + 1);
230 av_log(NULL, AV_LOG_ERROR, "\n");
231 return NULL;
232 }
233
234 s = av_mallocz(sizeof(ReSampleContext));
235 if (!s) {
236 av_log(NULL, AV_LOG_ERROR, "Can't allocate memory for resample context.\n");
237 return NULL;
238 }
239
240 s->ratio = (float)output_rate / (float)input_rate;
241
242 s->input_channels = input_channels;
243 s->output_channels = output_channels;
244
245 s->filter_channels = s->input_channels;
246 if (s->output_channels < s->filter_channels)
247 s->filter_channels = s->output_channels;
248
249 s->sample_fmt[0] = sample_fmt_in;
250 s->sample_fmt[1] = sample_fmt_out;
251 s->sample_size[0] = av_get_bytes_per_sample(s->sample_fmt[0]);
252 s->sample_size[1] = av_get_bytes_per_sample(s->sample_fmt[1]);
253
254 if (s->sample_fmt[0] != AV_SAMPLE_FMT_S16) {
255 if (!(s->convert_ctx[0] = av_audio_convert_alloc(AV_SAMPLE_FMT_S16, 1,
256 s->sample_fmt[0], 1, NULL, 0))) {
257 av_log(s, AV_LOG_ERROR,
258 "Cannot convert %s sample format to s16 sample format\n",
259 av_get_sample_fmt_name(s->sample_fmt[0]));
260 av_free(s);
261 return NULL;
262 }
263 }
264
265 if (s->sample_fmt[1] != AV_SAMPLE_FMT_S16) {
266 if (!(s->convert_ctx[1] = av_audio_convert_alloc(s->sample_fmt[1], 1,
267 AV_SAMPLE_FMT_S16, 1, NULL, 0))) {
268 av_log(s, AV_LOG_ERROR,
269 "Cannot convert s16 sample format to %s sample format\n",
270 av_get_sample_fmt_name(s->sample_fmt[1]));
271 av_audio_convert_free(s->convert_ctx[0]);
272 av_free(s);
273 return NULL;
274 }
275 }
276
277 s->resample_context = av_resample_init(output_rate, input_rate,
278 filter_length, log2_phase_count,
279 linear, cutoff);
280
281 *(const AVClass**)s->resample_context = &audioresample_context_class;
282
283 return s;
284 }
285
286 /* resample audio. 'nb_samples' is the number of input samples */
287 /* XXX: optimize it ! */
audio_resample(ReSampleContext * s,short * output,short * input,int nb_samples)288 int audio_resample(ReSampleContext *s, short *output, short *input, int nb_samples)
289 {
290 int i, nb_samples1;
291 short *bufin[MAX_CHANNELS];
292 short *bufout[MAX_CHANNELS];
293 short *buftmp2[MAX_CHANNELS], *buftmp3[MAX_CHANNELS];
294 short *output_bak = NULL;
295 int lenout;
296
297 if (s->input_channels == s->output_channels && s->ratio == 1.0 && 0) {
298 /* nothing to do */
299 memcpy(output, input, nb_samples * s->input_channels * sizeof(short));
300 return nb_samples;
301 }
302
303 if (s->sample_fmt[0] != AV_SAMPLE_FMT_S16) {
304 int istride[1] = { s->sample_size[0] };
305 int ostride[1] = { 2 };
306 const void *ibuf[1] = { input };
307 void *obuf[1];
308 unsigned input_size = nb_samples * s->input_channels * 2;
309
310 if (!s->buffer_size[0] || s->buffer_size[0] < input_size) {
311 av_free(s->buffer[0]);
312 s->buffer_size[0] = input_size;
313 s->buffer[0] = av_malloc(s->buffer_size[0]);
314 if (!s->buffer[0]) {
315 av_log(s->resample_context, AV_LOG_ERROR, "Could not allocate buffer\n");
316 return 0;
317 }
318 }
319
320 obuf[0] = s->buffer[0];
321
322 if (av_audio_convert(s->convert_ctx[0], obuf, ostride,
323 ibuf, istride, nb_samples * s->input_channels) < 0) {
324 av_log(s->resample_context, AV_LOG_ERROR,
325 "Audio sample format conversion failed\n");
326 return 0;
327 }
328
329 input = s->buffer[0];
330 }
331
332 lenout= 2*s->output_channels*nb_samples * s->ratio + 16;
333
334 if (s->sample_fmt[1] != AV_SAMPLE_FMT_S16) {
335 int out_size = lenout * av_get_bytes_per_sample(s->sample_fmt[1]) *
336 s->output_channels;
337 output_bak = output;
338
339 if (!s->buffer_size[1] || s->buffer_size[1] < out_size) {
340 av_free(s->buffer[1]);
341 s->buffer_size[1] = out_size;
342 s->buffer[1] = av_malloc(s->buffer_size[1]);
343 if (!s->buffer[1]) {
344 av_log(s->resample_context, AV_LOG_ERROR, "Could not allocate buffer\n");
345 return 0;
346 }
347 }
348
349 output = s->buffer[1];
350 }
351
352 /* XXX: move those malloc to resample init code */
353 for (i = 0; i < s->filter_channels; i++) {
354 bufin[i] = av_malloc_array((nb_samples + s->temp_len), sizeof(short));
355 bufout[i] = av_malloc_array(lenout, sizeof(short));
356
357 if (!bufin[i] || !bufout[i]) {
358 av_log(s->resample_context, AV_LOG_ERROR, "Could not allocate buffer\n");
359 nb_samples1 = 0;
360 goto fail;
361 }
362
363 memcpy(bufin[i], s->temp[i], s->temp_len * sizeof(short));
364 buftmp2[i] = bufin[i] + s->temp_len;
365 }
366
367 if (s->input_channels == 2 && s->output_channels == 1) {
368 buftmp3[0] = output;
369 stereo_to_mono(buftmp2[0], input, nb_samples);
370 } else if (s->output_channels >= 2 && s->input_channels == 1) {
371 buftmp3[0] = bufout[0];
372 memcpy(buftmp2[0], input, nb_samples * sizeof(short));
373 } else if (s->input_channels == 6 && s->output_channels ==2) {
374 buftmp3[0] = bufout[0];
375 buftmp3[1] = bufout[1];
376 surround_to_stereo(buftmp2, input, s->input_channels, nb_samples);
377 } else if (s->output_channels >= s->input_channels && s->input_channels >= 2) {
378 for (i = 0; i < s->input_channels; i++) {
379 buftmp3[i] = bufout[i];
380 }
381 deinterleave(buftmp2, input, s->input_channels, nb_samples);
382 } else {
383 buftmp3[0] = output;
384 memcpy(buftmp2[0], input, nb_samples * sizeof(short));
385 }
386
387 nb_samples += s->temp_len;
388
389 /* resample each channel */
390 nb_samples1 = 0; /* avoid warning */
391 for (i = 0; i < s->filter_channels; i++) {
392 int consumed;
393 int is_last = i + 1 == s->filter_channels;
394
395 nb_samples1 = av_resample(s->resample_context, buftmp3[i], bufin[i],
396 &consumed, nb_samples, lenout, is_last);
397 s->temp_len = nb_samples - consumed;
398 s->temp[i] = av_realloc_array(s->temp[i], s->temp_len, sizeof(short));
399 memcpy(s->temp[i], bufin[i] + consumed, s->temp_len * sizeof(short));
400 }
401
402 if (s->output_channels == 2 && s->input_channels == 1) {
403 mono_to_stereo(output, buftmp3[0], nb_samples1);
404 } else if (s->output_channels == 6 && s->input_channels == 2) {
405 ac3_5p1_mux(output, buftmp3[0], buftmp3[1], nb_samples1);
406 } else if ((s->output_channels == s->input_channels && s->input_channels >= 2) ||
407 (s->output_channels == 2 && s->input_channels == 6)) {
408 interleave(output, buftmp3, s->output_channels, nb_samples1);
409 }
410
411 if (s->sample_fmt[1] != AV_SAMPLE_FMT_S16) {
412 int istride[1] = { 2 };
413 int ostride[1] = { s->sample_size[1] };
414 const void *ibuf[1] = { output };
415 void *obuf[1] = { output_bak };
416
417 if (av_audio_convert(s->convert_ctx[1], obuf, ostride,
418 ibuf, istride, nb_samples1 * s->output_channels) < 0) {
419 av_log(s->resample_context, AV_LOG_ERROR,
420 "Audio sample format conversion failed\n");
421 return 0;
422 }
423 }
424
425 fail:
426 for (i = 0; i < s->filter_channels; i++) {
427 av_free(bufin[i]);
428 av_free(bufout[i]);
429 }
430
431 return nb_samples1;
432 }
433
audio_resample_close(ReSampleContext * s)434 void audio_resample_close(ReSampleContext *s)
435 {
436 int i;
437 av_resample_close(s->resample_context);
438 for (i = 0; i < s->filter_channels; i++)
439 av_freep(&s->temp[i]);
440 av_freep(&s->buffer[0]);
441 av_freep(&s->buffer[1]);
442 av_audio_convert_free(s->convert_ctx[0]);
443 av_audio_convert_free(s->convert_ctx[1]);
444 av_free(s);
445 }
446
447 #endif
448