summaryrefslogtreecommitdiffstats
path: root/audio/filter/af_lavrresample.c
blob: 56257edb8497f5d42c748281de66f2b969a48bbc (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
/*
 * Copyright (c) 2004 Michael Niedermayer <michaelni@gmx.at>
 * Copyright (c) 2013 Stefano Pigozzi <stefano.pigozzi@gmail.com>
 *
 * Based on Michael Niedermayer's lavcresample.
 *
 * This file is part of mpv.
 *
 * mpv is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation; either version 2 of the License, or
 * (at your option) any later version.
 *
 * mpv is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License along
 * with mpv.  If not, see <http://www.gnu.org/licenses/>.
 */

#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <inttypes.h>
#include <math.h>
#include <assert.h>

#include <libavutil/opt.h>
#include <libavutil/common.h>
#include <libavutil/samplefmt.h>
#include <libavutil/channel_layout.h>
#include <libavutil/mathematics.h>

#include "common/common.h"
#include "config.h"

#if HAVE_LIBAVRESAMPLE
#include <libavresample/avresample.h>
#elif HAVE_LIBSWRESAMPLE
#include <libswresample/swresample.h>
#define AVAudioResampleContext SwrContext
#define avresample_alloc_context swr_alloc
#define avresample_open swr_init
#define avresample_close(x) do { } while(0)
#define avresample_free swr_free
#define avresample_available(x) 0
#define avresample_convert(ctx, out, out_planesize, out_samples, in, in_planesize, in_samples) \
    swr_convert(ctx, out, out_samples, (const uint8_t**)(in), in_samples)
#define avresample_set_channel_mapping swr_set_channel_mapping
#define avresample_set_compensation swr_set_compensation
#else
#error "config.h broken or no resampler found"
#endif

#include "common/av_common.h"
#include "common/msg.h"
#include "options/m_option.h"
#include "audio/filter/af.h"
#include "audio/fmt-conversion.h"
#include "osdep/endian.h"

struct af_resample_opts {
    int filter_size;
    int phase_shift;
    int linear;
    double cutoff;
    int normalize;
};

struct af_resample {
    int allow_detach;
    char **avopts;
    double playback_speed;
    struct AVAudioResampleContext *avrctx;
    struct mp_audio avrctx_fmt; // output format of avrctx
    struct mp_audio pool_fmt; // format used to allocate frames for avrctx output
    struct mp_audio pre_out_fmt; // format before final conversion (S24)
    struct AVAudioResampleContext *avrctx_out; // for output channel reordering
    struct af_resample_opts opts;  // opts requested by the user
    // At least libswresample keeps a pointer around for this:
    int reorder_in[MP_NUM_CHANNELS];
    int reorder_out[MP_NUM_CHANNELS];
    struct mp_audio_pool *reorder_buffer;

    int in_rate_af; // filter input sample rate
    int in_rate;    // actual rate (used by lavr), adjusted for playback speed
    int in_format;
    struct mp_chmap in_channels;
    int out_rate;
    int out_format;
    struct mp_chmap out_channels;
};

#if HAVE_LIBAVRESAMPLE
static double get_delay(struct af_resample *s)
{
    return avresample_get_delay(s->avrctx) / (double)s->in_rate +
           avresample_available(s->avrctx) / (double)s->out_rate;
}
static void drop_all_output(struct af_resample *s)
{
    while (avresample_read(s->avrctx, NULL, 1000) > 0) {}
}
static int get_out_samples(struct af_resample *s, int in_samples)
{
    return avresample_get_out_samples(s->avrctx, in_samples);
}
#else
static double get_delay(struct af_resample *s)
{
    int64_t base = s->in_rate * (int64_t)s->out_rate;
    return swr_get_delay(s->avrctx, base) / (double)base;
}
static void drop_all_output(struct af_resample *s)
{
    while (swr_drop_output(s->avrctx, 1000) > 0) {}
}
static int get_out_samples(struct af_resample *s, int in_samples)
{
#if LIBSWRESAMPLE_VERSION_MAJOR > 1 || LIBSWRESAMPLE_VERSION_MINOR >= 2
    return swr_get_out_samples(s->avrctx, in_samples);
#else
    return av_rescale_rnd(in_samples, s->out_rate, s->in_rate, AV_ROUND_UP)
           + swr_get_delay(s->avrctx, s->out_rate);
#endif
}
#endif

static void close_lavrr(struct af_instance *af)
{
    struct af_resample *s = af->priv;

    if (s->avrctx)
        avresample_close(s->avrctx);
    avresample_free(&s->avrctx);
    if (s->avrctx_out)
        avresample_close(s->avrctx_out);
    avresample_free(&s->avrctx_out);
}

static int resample_frame(struct AVAudioResampleContext *r,
                          struct mp_audio *out, struct mp_audio *in)
{
    return avresample_convert(r,
        out ? (uint8_t **)out->planes : NULL,
        out ? mp_audio_get_allocated_size(out) : 0,
        out ? out->samples : 0,
        in ? (uint8_t **)in->planes : NULL,
        in ? mp_audio_get_allocated_size(in) : 0,
        in ? in->samples : 0);
}

static double af_resample_default_cutoff(int filter_size)
{
    return FFMAX(1.0 - 6.5 / (filter_size + 8), 0.80);
}

static int rate_from_speed(int rate, double speed)
{
    return lrint(rate * speed);
}

// Return the format libavresample should convert to, given the final output
// format mp_format. In some cases (S24) we perform an extra conversion step,
// and signal here what exactly libavresample should output. It will be the
// input to the final conversion to mp_format.
static int check_output_conversion(int mp_format)
{
    if (mp_format == AF_FORMAT_S24)
        return AV_SAMPLE_FMT_S32;
    return af_to_avformat(mp_format);
}

bool af_lavrresample_test_conversion(int src_format, int dst_format)
{
    return af_to_avformat(src_format) != AV_SAMPLE_FMT_NONE &&
           check_output_conversion(dst_format) != AV_SAMPLE_FMT_NONE;
}

static struct mp_chmap fudge_pairs[][2] = {
    {MP_CHMAP2(BL,  BR),  MP_CHMAP2(SL,  SR)},
    {MP_CHMAP2(SL,  SR),  MP_CHMAP2(BL,  BR)},
    {MP_CHMAP2(SDL, SDR), MP_CHMAP2(SL,  SR)},
    {MP_CHMAP2(SL,  SR),  MP_CHMAP2(SDL, SDR)},
};

// Modify out_layout and return the new value. The intention is reducing the
// loss libswresample's rematrixing will cause by exchanging similar, but
// strictly speaking incompatible channel pairs. For example, 7.1 should be
// changed to 7.1(wide) without dropping the SL/SR channels. (We still leave
// it to libswresample to create the remix matrix.)
static uint64_t fudge_layout_conversion(struct af_instance *af,
                                        uint64_t in, uint64_t out)
{
    for (int n = 0; n < MP_ARRAY_SIZE(fudge_pairs); n++) {
        uint64_t a = mp_chmap_to_lavc(&fudge_pairs[n][0]);
        uint64_t b = mp_chmap_to_lavc(&fudge_pairs[n][1]);
        if ((in & a) == a && (in & b) == 0 &&
            (out & a) == 0 && (out & b) == b)
        {
            out = (out & ~b) | a;

            MP_VERBOSE(af, "Fudge: %s -> %s\n",
                       mp_chmap_to_str(&fudge_pairs[n][0]),
                       mp_chmap_to_str(&fudge_pairs[n][1]));
        }
    }
    return out;
}

// mp_chmap_get_reorder() performs:
//  to->speaker[n] = from->speaker[src[n]]
// but libavresample does:
//  to->speaker[dst[n]] = from->speaker[n]
static void transpose_order(int *map, int num)
{
    int nmap[MP_NUM_CHANNELS] = {0};
    for (int n = 0; n < num; n++) {
        for (int i = 0; i < num; i++) {
            if (map[n] == i)
                nmap[i] = n;
        }
    }
    memcpy(map, nmap, sizeof(nmap));
}

static int configure_lavrr(struct af_instance *af, struct mp_audio *in,
                           struct mp_audio *out, bool verbose)
{
    struct af_resample *s = af->priv;

    close_lavrr(af);

    s->avrctx = avresample_alloc_context();
    s->avrctx_out = avresample_alloc_context();
    if (!s->avrctx || !s->avrctx_out)
        goto error;

    enum AVSampleFormat in_samplefmt = af_to_avformat(in->format);
    enum AVSampleFormat out_samplefmt = check_output_conversion(out->format);
    enum AVSampleFormat out_samplefmtp = av_get_planar_sample_fmt(out_samplefmt);

    if (in_samplefmt == AV_SAMPLE_FMT_NONE ||
        out_samplefmt == AV_SAMPLE_FMT_NONE ||
        out_samplefmtp == AV_SAMPLE_FMT_NONE)
        goto error;

    s->out_rate    = out->rate;
    s->in_rate_af  = in->rate;
    s->in_rate     = rate_from_speed(in->rate, s->playback_speed);
    s->out_format  = out->format;
    s->in_format   = in->format;
    s->out_channels= out->channels;
    s->in_channels = in->channels;

    av_opt_set_int(s->avrctx, "filter_size",        s->opts.filter_size, 0);
    av_opt_set_int(s->avrctx, "phase_shift",        s->opts.phase_shift, 0);
    av_opt_set_int(s->avrctx, "linear_interp",      s->opts.linear, 0);

    av_opt_set_double(s->avrctx, "cutoff",          s->opts.cutoff, 0);

    struct mp_chmap map_in = in->channels;
    struct mp_chmap map_out = out->channels;

    // Try not to do any remixing if at least one is "unknown".
    if (mp_chmap_is_unknown(&map_in) || mp_chmap_is_unknown(&map_out)) {
        mp_chmap_set_unknown(&map_in, map_in.num);
        mp_chmap_set_unknown(&map_out, map_out.num);
    }

    // unchecked: don't take any channel reordering into account
    uint64_t in_ch_layout = mp_chmap_to_lavc_unchecked(&map_in);
    uint64_t out_ch_layout = mp_chmap_to_lavc_unchecked(&map_out);

    struct mp_chmap in_lavc, out_lavc;
    mp_chmap_from_lavc(&in_lavc, in_ch_layout);
    mp_chmap_from_lavc(&out_lavc, out_ch_layout);

    if (verbose && !mp_chmap_equals(&in_lavc, &out_lavc)) {
        MP_VERBOSE(af, "Remix: %s -> %s\n", mp_chmap_to_str(&in_lavc),
                                            mp_chmap_to_str(&out_lavc));
    }

    if (in_lavc.num != map_in.num) {
        // For handling NA channels, we would have to add a planarization step.
        MP_FATAL(af, "Unsupported channel remapping.\n");
        goto error;
    }

    mp_chmap_get_reorder(s->reorder_in, &map_in, &in_lavc);
    transpose_order(s->reorder_in, map_in.num);

    if (mp_chmap_equals(&out_lavc, &map_out)) {
        // No intermediate step required - output new format directly.
        out_samplefmtp = out_samplefmt;
    } else {
        // Verify that we really just reorder and/or insert NA channels.
        struct mp_chmap withna = out_lavc;
        mp_chmap_fill_na(&withna, map_out.num);
        if (withna.num != map_out.num)
            goto error;
    }
    mp_chmap_get_reorder(s->reorder_out, &out_lavc, &map_out);

    s->avrctx_fmt = *out;
    mp_audio_set_channels(&s->avrctx_fmt, &out_lavc);
    mp_audio_set_format(&s->avrctx_fmt, af_from_avformat(out_samplefmtp));

    s->pre_out_fmt = *out;
    mp_audio_set_format(&s->pre_out_fmt, af_from_avformat(out_samplefmt));

    // If there are NA channels, the final output will have more channels than
    // the avrctx output. Also, avrctx will output planar (out_samplefmtp was
    // not overwritten). Allocate the output frame with more channels, so the
    // NA channels can be trivially added.
    s->pool_fmt = s->avrctx_fmt;
    if (map_out.num > out_lavc.num)
        mp_audio_set_channels(&s->pool_fmt, &map_out);

    out_ch_layout = fudge_layout_conversion(af, in_ch_layout, out_ch_layout);

    int normalize = s->opts.normalize;
    if (normalize < 0)
        normalize = af->opts->audio_normalize;
#if HAVE_LIBSWRESAMPLE
    av_opt_set_double(s->avrctx, "rematrix_maxval", normalize ? 1 : 1000, 0);
#else
    av_opt_set_int(s->avrctx, "normalize_mix_level", !!normalize, 0);
#endif

    if (mp_set_avopts(af->log, s->avrctx, s->avopts) < 0)
        goto error;

    // Real conversion; output is input to avrctx_out.
    av_opt_set_int(s->avrctx, "in_channel_layout",  in_ch_layout, 0);
    av_opt_set_int(s->avrctx, "out_channel_layout", out_ch_layout, 0);
    av_opt_set_int(s->avrctx, "in_sample_rate",     s->in_rate, 0);
    av_opt_set_int(s->avrctx, "out_sample_rate",    s->out_rate, 0);
    av_opt_set_int(s->avrctx, "in_sample_fmt",      in_samplefmt, 0);
    av_opt_set_int(s->avrctx, "out_sample_fmt",     out_samplefmtp, 0);

    // Just needs the correct number of channels for deplanarization.
    struct mp_chmap fake_chmap;
    mp_chmap_set_unknown(&fake_chmap, map_out.num);
    uint64_t fake_out_ch_layout = mp_chmap_to_lavc_unchecked(&fake_chmap);
    if (!fake_out_ch_layout)
        goto error;
    av_opt_set_int(s->avrctx_out, "in_channel_layout",  fake_out_ch_layout, 0);
    av_opt_set_int(s->avrctx_out, "out_channel_layout", fake_out_ch_layout, 0);

    av_opt_set_int(s->avrctx_out, "in_sample_fmt",      out_samplefmtp, 0);
    av_opt_set_int(s->avrctx_out, "out_sample_fmt",     out_samplefmt, 0);
    av_opt_set_int(s->avrctx_out, "in_sample_rate",     s->out_rate, 0);
    av_opt_set_int(s->avrctx_out, "out_sample_rate",    s->out_rate, 0);

    // API has weird requirements, quoting avresample.h:
    //  * This function can only be called when the allocated context is not open.
    //  * Also, the input channel layout must have already been set.
    avresample_set_channel_mapping(s->avrctx, s->reorder_in);

    if (avresample_open(s->avrctx) < 0 || avresample_open(s->avrctx_out) < 0) {
        MP_ERR(af, "Cannot open Libavresample Context. \n");
        goto error;
    }
    return AF_OK;

error:
    close_lavrr(af);
    return AF_ERROR;
}


static int control(struct af_instance *af, int cmd, void *arg)
{
    struct af_resample *s = af->priv;

    switch (cmd) {
    case AF_CONTROL_REINIT: {
        struct mp_audio *in = arg;
        struct mp_audio *out = af->data;
        struct mp_audio orig_in = *in;

        if (((out->rate    == in->rate) || (out->rate == 0)) &&
            (out->format   == in->format) &&
            (mp_chmap_equals(&out->channels, &in->channels) || out->nch == 0) &&
            s->allow_detach && s->playback_speed == 1.0)
            return AF_DETACH;

        if (out->rate == 0)
            out->rate = in->rate;

        if (mp_chmap_is_empty(&out->channels))
            mp_audio_set_channels(out, &in->channels);

        if (af_to_avformat(in->format) == AV_SAMPLE_FMT_NONE)
            mp_audio_set_format(in, AF_FORMAT_FLOAT);
        if (check_output_conversion(out->format) == AV_SAMPLE_FMT_NONE)
            mp_audio_set_format(out, in->format);

        int r = ((in->format == orig_in.format) &&
                mp_chmap_equals(&in->channels, &orig_in.channels))
                ? AF_OK : AF_FALSE;

        if (r == AF_OK)
            r = configure_lavrr(af, in, out, true);
        return r;
    }
    case AF_CONTROL_SET_FORMAT: {
        int format = *(int *)arg;
        if (format && check_output_conversion(format) == AV_SAMPLE_FMT_NONE)
            return AF_FALSE;

        mp_audio_set_format(af->data, format);
        return AF_OK;
    }
    case AF_CONTROL_SET_CHANNELS: {
        mp_audio_set_channels(af->data, (struct mp_chmap *)arg);
        return AF_OK;
    }
    case AF_CONTROL_SET_RESAMPLE_RATE:
        af->data->rate = *(int *)arg;
        return AF_OK;
    case AF_CONTROL_SET_PLAYBACK_SPEED_RESAMPLE: {
        s->playback_speed = *(double *)arg;
        return AF_OK;
    }
    case AF_CONTROL_RESET:
        if (s->avrctx)
            drop_all_output(s);
        return AF_OK;
    }
    return AF_UNKNOWN;
}

static void uninit(struct af_instance *af)
{
    close_lavrr(af);
}

// The LSB is always ignored.
#if BYTE_ORDER == BIG_ENDIAN
#define SHIFT24(x) ((3-(x))*8)
#else
#define SHIFT24(x) (((x)+1)*8)
#endif

static void extra_output_conversion(struct af_instance *af, struct mp_audio *mpa)
{
    if (mpa->format == AF_FORMAT_S32 && af->data->format == AF_FORMAT_S24) {
        size_t len = mp_audio_psize(mpa) / mpa->bps;
        for (int s = 0; s < len; s++) {
            uint32_t val = *((uint32_t *)mpa->planes[0] + s);
            uint8_t *ptr = (uint8_t *)mpa->planes[0] + s * 3;
            ptr[0] = val >> SHIFT24(0);
            ptr[1] = val >> SHIFT24(1);
            ptr[2] = val >> SHIFT24(2);
        }
        mp_audio_set_format(mpa, AF_FORMAT_S24);
    }

    for (int p = 0; p < mpa->num_planes; p++) {
        void *ptr = mpa->planes[p];
        int total = mpa->samples * mpa->spf;
        if (af_fmt_from_planar(mpa->format) == AF_FORMAT_FLOAT) {
            for (int s = 0; s < total; s++)
                ((float *)ptr)[s] = av_clipf(((float *)ptr)[s], -1.0f, 1.0f);
        } else if (af_fmt_from_planar(mpa->format) == AF_FORMAT_DOUBLE) {
            for (int s = 0; s < total; s++)
                ((double *)ptr)[s] = MPCLAMP(((double *)ptr)[s], -1.0, 1.0);
        }
    }
}

// This relies on the tricky way mpa was allocated.
static void reorder_planes(struct mp_audio *mpa, int *reorder,
                           struct mp_chmap *newmap)
{
    struct mp_audio prev = *mpa;
    mp_audio_set_channels(mpa, newmap);

    // The trailing planes were never written by avrctx, they're the NA channels.
    int next_na = prev.num_planes;

    for (int n = 0; n < mpa->num_planes; n++) {
        int src = reorder[n];
        assert(src >= -1 && src < prev.num_planes);
        if (src >= 0) {
            mpa->planes[n] = prev.planes[src];
        } else {
            assert(next_na < mpa->num_planes);
            mpa->planes[n] = prev.planes[next_na++];
            af_fill_silence(mpa->planes[n], mpa->sstride * mpa->samples,
                            mpa->format);
        }
    }
}

static int filter_resample(struct af_instance *af, struct mp_audio *in)
{
    struct af_resample *s = af->priv;

    int samples = get_out_samples(s, in ? in->samples : 0);

    struct mp_audio out_format = s->pool_fmt;
    struct mp_audio *out = mp_audio_pool_get(af->out_pool, &out_format, samples);
    if (!out)
        goto error;
    if (in)
        mp_audio_copy_attributes(out, in);

    if (!s->avrctx)
        goto error;

    if (out->samples) {
        out->samples = resample_frame(s->avrctx, out, in);
        if (out->samples < 0)
            goto error;
    }

    struct mp_audio real_out = *out;
    mp_audio_copy_config(out, &s->avrctx_fmt);

    if (out->samples && !mp_audio_config_equals(out, &s->pre_out_fmt)) {
        assert(af_fmt_is_planar(out->format) && out->format == real_out.format);
        reorder_planes(out, s->reorder_out, &s->pool_fmt.channels);
        if (!mp_audio_config_equals(out, &s->pre_out_fmt)) {
            struct mp_audio *new = mp_audio_pool_get(s->reorder_buffer,
                                                     &s->pre_out_fmt,
                                                     out->samples);
            if (!new)
                goto error;
            mp_audio_copy_attributes(new, out);
            int out_samples = resample_frame(s->avrctx_out, new, out);
            talloc_free(out);
            out = new;
            if (out_samples != new->samples)
                goto error;
        }
    }

    extra_output_conversion(af, out);

    talloc_free(in);
    if (out->samples) {
        af_add_output_frame(af, out);
    } else {
        talloc_free(out);
    }

    af->delay = get_delay(s);

    return 0;
error:
    talloc_free(in);
    talloc_free(out);
    return -1;
}

static int filter(struct af_instance *af, struct mp_audio *in)
{
    struct af_resample *s = af->priv;

    int new_rate = rate_from_speed(s->in_rate_af, s->playback_speed);
    bool need_reinit = fabs(new_rate / (double)s->in_rate - 1) > 0.01;

    if (s->avrctx) {
        AVRational r = av_d2q(s->playback_speed * s->in_rate_af / s->in_rate,
                              INT_MAX / 2);
        // Essentially, swr/avresample_set_compensation() does 2 things:
        // - adjust output sample rate by sample_delta/compensation_distance
        // - reset the adjustment after compensation_distance output samples
        // Increase the compensation_distance to avoid undesired reset
        // semantics - we want to keep the ratio for the whole frame we're
        // feeding it, until the next filter() call.
        int mult = INT_MAX / 2 / MPMAX(MPMAX(abs(r.num), abs(r.den)), 1);
        r = (AVRational){ r.num * mult, r.den * mult };
        if (avresample_set_compensation(s->avrctx, r.den - r.num, r.den) < 0)
            need_reinit = true;
    }

    if (need_reinit && new_rate != s->in_rate) {
        // Before reconfiguring, drain the audio that is still buffered
        // in the resampler.
        filter_resample(af, NULL);
        // Reinitialize resampler.
        configure_lavrr(af, &af->fmt_in, &af->fmt_out, false);
    }

    return filter_resample(af, in);
}

static int af_open(struct af_instance *af)
{
    struct af_resample *s = af->priv;

    af->control = control;
    af->uninit  = uninit;
    af->filter_frame = filter;

    if (s->opts.cutoff <= 0.0)
        s->opts.cutoff = af_resample_default_cutoff(s->opts.filter_size);

    s->reorder_buffer = mp_audio_pool_create(s);

    return AF_OK;
}

#define OPT_BASE_STRUCT struct af_resample

const struct af_info af_info_lavrresample = {
    .info = "Sample frequency conversion using libavresample",
    .name = "lavrresample",
    .open = af_open,
    .priv_size = sizeof(struct af_resample),
    .priv_defaults = &(const struct af_resample) {
        .opts = {
            .filter_size = 16,
            .cutoff      = 0.0,
            .phase_shift = 10,
            .normalize   = -1,
        },
        .playback_speed = 1.0,
        .allow_detach = 1,
    },
    .options = (const struct m_option[]) {
        OPT_INTRANGE("filter-size", opts.filter_size, 0, 0, 32),
        OPT_INTRANGE("phase-shift", opts.phase_shift, 0, 0, 30),
        OPT_FLAG("linear", opts.linear, 0),
        OPT_DOUBLE("cutoff", opts.cutoff, M_OPT_RANGE, .min = 0, .max = 1),
        OPT_FLAG("detach", allow_detach, 0),
        OPT_CHOICE("normalize", opts.normalize, 0,
                   ({"no", 0}, {"yes", 1}, {"auto", -1})),
        OPT_KEYVALUELIST("o", avopts, 0),
        {0}
    },
};