libavcodec/resample.c
de6d9b64
 /*
f1ea5c2a
  * samplerate conversion for both audio and video
406792e7
  * Copyright (c) 2000 Fabrice Bellard
de6d9b64
  *
b78e7197
  * This file is part of FFmpeg.
  *
  * FFmpeg is free software; you can redistribute it and/or
ff4ec49e
  * modify it under the terms of the GNU Lesser General Public
  * License as published by the Free Software Foundation; either
b78e7197
  * version 2.1 of the License, or (at your option) any later version.
de6d9b64
  *
b78e7197
  * FFmpeg is distributed in the hope that it will be useful,
de6d9b64
  * but WITHOUT ANY WARRANTY; without even the implied warranty of
ff4ec49e
  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  * Lesser General Public License for more details.
de6d9b64
  *
ff4ec49e
  * You should have received a copy of the GNU Lesser General Public
b78e7197
  * License along with FFmpeg; if not, write to the Free Software
5509bffa
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
de6d9b64
  */
983e3246
 
 /**
ba87f080
  * @file
f1ea5c2a
  * samplerate conversion for both audio and video
983e3246
  */
 
1d9c2dc8
 #include <string.h>
 
de6d9b64
 #include "avcodec.h"
d1e3c6fd
 #include "audioconvert.h"
6ed04040
 #include "libavutil/opt.h"
1d9c2dc8
 #include "libavutil/mem.h"
7ffe76e5
 #include "libavutil/samplefmt.h"
69db4e10
 
30223b3b
 #if FF_API_AVCODEC_RESAMPLE
 
3e00abab
 #define MAX_CHANNELS 8
 
aaaf1635
 struct AVResampleContext;
de6d9b64
 
d1e3c6fd
 static const char *context_to_name(void *ptr)
 {
     return "audioresample";
 }
 
 static const AVOption options[] = {{NULL}};
ffc437c0
 static const AVClass audioresample_context_class = {
     "ReSampleContext", context_to_name, options, LIBAVUTIL_VERSION_INT
 };
d1e3c6fd
 
de6d9b64
 struct ReSampleContext {
aaaf1635
     struct AVResampleContext *resample_context;
3e00abab
     short *temp[MAX_CHANNELS];
aaaf1635
     int temp_len;
de6d9b64
     float ratio;
     /* channel convert */
     int input_channels, output_channels, filter_channels;
d1e3c6fd
     AVAudioConvert *convert_ctx[2];
5d6e4c16
     enum AVSampleFormat sample_fmt[2]; ///< input and output sample format
ffc437c0
     unsigned sample_size[2];           ///< size of one sample in sample_fmt
     short *buffer[2];                  ///< buffers used for conversion to S16
     unsigned buffer_size[2];           ///< sizes of allocated buffers
de6d9b64
 };
 
 /* n1: number of samples */
 static void stereo_to_mono(short *output, short *input, int n1)
 {
     short *p, *q;
     int n = n1;
 
     p = input;
     q = output;
     while (n >= 4) {
         q[0] = (p[0] + p[1]) >> 1;
         q[1] = (p[2] + p[3]) >> 1;
         q[2] = (p[4] + p[5]) >> 1;
         q[3] = (p[6] + p[7]) >> 1;
         q += 4;
         p += 8;
         n -= 4;
     }
     while (n > 0) {
         q[0] = (p[0] + p[1]) >> 1;
         q++;
         p += 2;
         n--;
     }
 }
 
 /* n1: number of samples */
 static void mono_to_stereo(short *output, short *input, int n1)
 {
     short *p, *q;
     int n = n1;
     int v;
 
     p = input;
     q = output;
     while (n >= 4) {
         v = p[0]; q[0] = v; q[1] = v;
         v = p[1]; q[2] = v; q[3] = v;
         v = p[2]; q[4] = v; q[5] = v;
         v = p[3]; q[6] = v; q[7] = v;
         q += 8;
         p += 4;
         n -= 4;
     }
     while (n > 0) {
         v = p[0]; q[0] = v; q[1] = v;
         q += 2;
         p += 1;
         n--;
     }
 }
 
02260ccc
 /*
 5.1 to stereo input: [fl, fr, c, lfe, rl, rr]
 - Left = front_left + rear_gain * rear_left + center_gain * center
 - Right = front_right + rear_gain * rear_right + center_gain * center
 Where rear_gain is usually around 0.5-1.0 and
       center_gain is almost always 0.7 (-3 dB)
 */
 static void surround_to_stereo(short **output, short *input, int channels, int samples)
 {
     int i;
     short l, r;
 
     for (i = 0; i < samples; i++) {
b85e9d31
         int fl,fr,c,rl,rr;
02260ccc
         fl = input[0];
         fr = input[1];
         c = input[2];
b85e9d31
         // lfe = input[3];
02260ccc
         rl = input[4];
         rr = input[5];
 
         l = av_clip_int16(fl + (0.5 * rl) + (0.7 * c));
         r = av_clip_int16(fr + (0.5 * rr) + (0.7 * c));
 
         /* output l & r. */
         *output[0]++ = l;
         *output[1]++ = r;
 
         /* increment input. */
         input += channels;
     }
 }
 
3e00abab
 static void deinterleave(short **output, short *input, int channels, int samples)
de6d9b64
 {
3e00abab
     int i, j;
de6d9b64
 
3e00abab
     for (i = 0; i < samples; i++) {
         for (j = 0; j < channels; j++) {
             *output[j]++ = *input++;
         }
de6d9b64
     }
 }
 
3e00abab
 static void interleave(short *output, short **input, int channels, int samples)
de6d9b64
 {
3e00abab
     int i, j;
de6d9b64
 
3e00abab
     for (i = 0; i < samples; i++) {
         for (j = 0; j < channels; j++) {
             *output++ = *input[j]++;
         }
de6d9b64
     }
 }
 
743739d2
 static void ac3_5p1_mux(short *output, short *input1, short *input2, int n)
 {
     int i;
ffc437c0
     short l, r;
 
     for (i = 0; i < n; i++) {
         l = *input1++;
         r = *input2++;
         *output++ = l;                  /* left */
         *output++ = (l / 2) + (r / 2);  /* center */
         *output++ = r;                  /* right */
         *output++ = 0;                  /* left surround */
         *output++ = 0;                  /* right surroud */
         *output++ = 0;                  /* low freq */
743739d2
     }
 }
 
e8876902
 #define SUPPORT_RESAMPLE(ch1, ch2, ch3, ch4, ch5, ch6, ch7, ch8) \
     ch8<<7 | ch7<<6 | ch6<<5 | ch5<<4 | ch4<<3 | ch3<<2 | ch2<<1 | ch1<<0
 
 static const uint8_t supported_resampling[MAX_CHANNELS] = {
acc3c380
     // output ch:    1  2  3  4  5  6  7  8
e8876902
     SUPPORT_RESAMPLE(1, 1, 0, 0, 0, 0, 0, 0), // 1 input channel
     SUPPORT_RESAMPLE(1, 1, 0, 0, 0, 1, 0, 0), // 2 input channels
     SUPPORT_RESAMPLE(0, 0, 1, 0, 0, 0, 0, 0), // 3 input channels
     SUPPORT_RESAMPLE(0, 0, 0, 1, 0, 0, 0, 0), // 4 input channels
     SUPPORT_RESAMPLE(0, 0, 0, 0, 1, 0, 0, 0), // 5 input channels
     SUPPORT_RESAMPLE(0, 1, 0, 0, 0, 1, 0, 0), // 6 input channels
     SUPPORT_RESAMPLE(0, 0, 0, 0, 0, 0, 1, 0), // 7 input channels
     SUPPORT_RESAMPLE(0, 0, 0, 0, 0, 0, 0, 1), // 8 input channels
 };
 
d1e3c6fd
 ReSampleContext *av_audio_resample_init(int output_channels, int input_channels,
                                         int output_rate, int input_rate,
5d6e4c16
                                         enum AVSampleFormat sample_fmt_out,
                                         enum AVSampleFormat sample_fmt_in,
d1e3c6fd
                                         int filter_length, int log2_phase_count,
                                         int linear, double cutoff)
de6d9b64
 {
     ReSampleContext *s;
115329f1
 
ffc437c0
     if (input_channels > MAX_CHANNELS) {
3e00abab
         av_log(NULL, AV_LOG_ERROR,
                "Resampling with input channels greater than %d is unsupported.\n",
                MAX_CHANNELS);
bb270c08
         return NULL;
ffc437c0
     }
e8876902
     if (!(supported_resampling[input_channels-1] & (1<<(output_channels-1)))) {
         int i;
         av_log(NULL, AV_LOG_ERROR, "Unsupported audio resampling. Allowed "
                "output channels for %d input channel%s", input_channels,
                input_channels > 1 ? "s:" : ":");
         for (i = 0; i < MAX_CHANNELS; i++)
             if (supported_resampling[input_channels-1] & (1<<i))
                 av_log(NULL, AV_LOG_ERROR, " %d", i + 1);
         av_log(NULL, AV_LOG_ERROR, "\n");
bb270c08
         return NULL;
918a5409
     }
de6d9b64
 
     s = av_mallocz(sizeof(ReSampleContext));
ffc437c0
     if (!s) {
30dc5541
         av_log(NULL, AV_LOG_ERROR, "Can't allocate memory for resample context.\n");
bb270c08
         return NULL;
ffc437c0
     }
de6d9b64
 
     s->ratio = (float)output_rate / (float)input_rate;
115329f1
 
de6d9b64
     s->input_channels = input_channels;
     s->output_channels = output_channels;
115329f1
 
de6d9b64
     s->filter_channels = s->input_channels;
     if (s->output_channels < s->filter_channels)
         s->filter_channels = s->output_channels;
 
ffc437c0
     s->sample_fmt[0]  = sample_fmt_in;
     s->sample_fmt[1]  = sample_fmt_out;
e6c52cee
     s->sample_size[0] = av_get_bytes_per_sample(s->sample_fmt[0]);
     s->sample_size[1] = av_get_bytes_per_sample(s->sample_fmt[1]);
d1e3c6fd
 
5d6e4c16
     if (s->sample_fmt[0] != AV_SAMPLE_FMT_S16) {
         if (!(s->convert_ctx[0] = av_audio_convert_alloc(AV_SAMPLE_FMT_S16, 1,
d1e3c6fd
                                                          s->sample_fmt[0], 1, NULL, 0))) {
             av_log(s, AV_LOG_ERROR,
                    "Cannot convert %s sample format to s16 sample format\n",
ba7d6e79
                    av_get_sample_fmt_name(s->sample_fmt[0]));
d1e3c6fd
             av_free(s);
             return NULL;
         }
     }
 
5d6e4c16
     if (s->sample_fmt[1] != AV_SAMPLE_FMT_S16) {
d1e3c6fd
         if (!(s->convert_ctx[1] = av_audio_convert_alloc(s->sample_fmt[1], 1,
5d6e4c16
                                                          AV_SAMPLE_FMT_S16, 1, NULL, 0))) {
d1e3c6fd
             av_log(s, AV_LOG_ERROR,
                    "Cannot convert s16 sample format to %s sample format\n",
ba7d6e79
                    av_get_sample_fmt_name(s->sample_fmt[1]));
d1e3c6fd
             av_audio_convert_free(s->convert_ctx[0]);
             av_free(s);
             return NULL;
         }
     }
 
ffc437c0
     s->resample_context = av_resample_init(output_rate, input_rate,
                                            filter_length, log2_phase_count,
                                            linear, cutoff);
d1e3c6fd
 
844d17fb
     *(const AVClass**)s->resample_context = &audioresample_context_class;
115329f1
 
de6d9b64
     return s;
 }
 
 /* resample audio. 'nb_samples' is the number of input samples */
 /* XXX: optimize it ! */
 int audio_resample(ReSampleContext *s, short *output, short *input, int nb_samples)
 {
     int i, nb_samples1;
3e00abab
     short *bufin[MAX_CHANNELS];
     short *bufout[MAX_CHANNELS];
     short *buftmp2[MAX_CHANNELS], *buftmp3[MAX_CHANNELS];
d1e3c6fd
     short *output_bak = NULL;
1a565432
     int lenout;
de6d9b64
 
b9d2085b
     if (s->input_channels == s->output_channels && s->ratio == 1.0 && 0) {
de6d9b64
         /* nothing to do */
         memcpy(output, input, nb_samples * s->input_channels * sizeof(short));
         return nb_samples;
     }
 
5d6e4c16
     if (s->sample_fmt[0] != AV_SAMPLE_FMT_S16) {
d1e3c6fd
         int istride[1] = { s->sample_size[0] };
         int ostride[1] = { 2 };
         const void *ibuf[1] = { input };
         void       *obuf[1];
ffc437c0
         unsigned input_size = nb_samples * s->input_channels * 2;
d1e3c6fd
 
         if (!s->buffer_size[0] || s->buffer_size[0] < input_size) {
             av_free(s->buffer[0]);
             s->buffer_size[0] = input_size;
             s->buffer[0] = av_malloc(s->buffer_size[0]);
             if (!s->buffer[0]) {
89bc05d1
                 av_log(s->resample_context, AV_LOG_ERROR, "Could not allocate buffer\n");
d1e3c6fd
                 return 0;
             }
         }
 
         obuf[0] = s->buffer[0];
 
         if (av_audio_convert(s->convert_ctx[0], obuf, ostride,
ffc437c0
                              ibuf, istride, nb_samples * s->input_channels) < 0) {
             av_log(s->resample_context, AV_LOG_ERROR,
                    "Audio sample format conversion failed\n");
d1e3c6fd
             return 0;
         }
 
ffc437c0
         input = s->buffer[0];
d1e3c6fd
     }
 
a61bc91d
     lenout= 2*s->output_channels*nb_samples * s->ratio + 16;
d1e3c6fd
 
5d6e4c16
     if (s->sample_fmt[1] != AV_SAMPLE_FMT_S16) {
6ca3b248
         int out_size = lenout * av_get_bytes_per_sample(s->sample_fmt[1]) *
                        s->output_channels;
d1e3c6fd
         output_bak = output;
 
6ca3b248
         if (!s->buffer_size[1] || s->buffer_size[1] < out_size) {
d1e3c6fd
             av_free(s->buffer[1]);
6ca3b248
             s->buffer_size[1] = out_size;
d1e3c6fd
             s->buffer[1] = av_malloc(s->buffer_size[1]);
             if (!s->buffer[1]) {
89bc05d1
                 av_log(s->resample_context, AV_LOG_ERROR, "Could not allocate buffer\n");
d1e3c6fd
                 return 0;
             }
         }
 
         output = s->buffer[1];
     }
 
1a565432
     /* XXX: move those malloc to resample init code */
ffc437c0
     for (i = 0; i < s->filter_channels; i++) {
         bufin[i] = av_malloc((nb_samples + s->temp_len) * sizeof(short));
aaaf1635
         memcpy(bufin[i], s->temp[i], s->temp_len * sizeof(short));
         buftmp2[i] = bufin[i] + s->temp_len;
3e00abab
         bufout[i] = av_malloc(lenout * sizeof(short));
aaaf1635
     }
115329f1
 
ffc437c0
     if (s->input_channels == 2 && s->output_channels == 1) {
de6d9b64
         buftmp3[0] = output;
         stereo_to_mono(buftmp2[0], input, nb_samples);
743739d2
     } else if (s->output_channels >= 2 && s->input_channels == 1) {
de6d9b64
         buftmp3[0] = bufout[0];
ffc437c0
         memcpy(buftmp2[0], input, nb_samples * sizeof(short));
02260ccc
     } else if (s->input_channels == 6 && s->output_channels ==2) {
         buftmp3[0] = bufout[0];
         buftmp3[1] = bufout[1];
         surround_to_stereo(buftmp2, input, s->input_channels, nb_samples);
3e00abab
     } else if (s->output_channels >= s->input_channels && s->input_channels >= 2) {
         for (i = 0; i < s->input_channels; i++) {
             buftmp3[i] = bufout[i];
         }
         deinterleave(buftmp2, input, s->input_channels, nb_samples);
de6d9b64
     } else {
         buftmp3[0] = output;
ffc437c0
         memcpy(buftmp2[0], input, nb_samples * sizeof(short));
de6d9b64
     }
 
aaaf1635
     nb_samples += s->temp_len;
 
de6d9b64
     /* resample each channel */
     nb_samples1 = 0; /* avoid warning */
ffc437c0
     for (i = 0; i < s->filter_channels; i++) {
aaaf1635
         int consumed;
ffc437c0
         int is_last = i + 1 == s->filter_channels;
aaaf1635
 
ffc437c0
         nb_samples1 = av_resample(s->resample_context, buftmp3[i], bufin[i],
                                   &consumed, nb_samples, lenout, is_last);
         s->temp_len = nb_samples - consumed;
         s->temp[i] = av_realloc(s->temp[i], s->temp_len * sizeof(short));
         memcpy(s->temp[i], bufin[i] + consumed, s->temp_len * sizeof(short));
de6d9b64
     }
 
     if (s->output_channels == 2 && s->input_channels == 1) {
         mono_to_stereo(output, buftmp3[0], nb_samples1);
3e00abab
     } else if (s->output_channels == 6 && s->input_channels == 2) {
743739d2
         ac3_5p1_mux(output, buftmp3[0], buftmp3[1], nb_samples1);
02260ccc
     } else if ((s->output_channels == s->input_channels && s->input_channels >= 2) ||
                (s->output_channels == 2 && s->input_channels == 6)) {
3e00abab
         interleave(output, buftmp3, s->output_channels, nb_samples1);
de6d9b64
     }
 
5d6e4c16
     if (s->sample_fmt[1] != AV_SAMPLE_FMT_S16) {
d1e3c6fd
         int istride[1] = { 2 };
         int ostride[1] = { s->sample_size[1] };
         const void *ibuf[1] = { output };
         void       *obuf[1] = { output_bak };
 
         if (av_audio_convert(s->convert_ctx[1], obuf, ostride,
ffc437c0
                              ibuf, istride, nb_samples1 * s->output_channels) < 0) {
             av_log(s->resample_context, AV_LOG_ERROR,
511cf612
                    "Audio sample format conversion failed\n");
d1e3c6fd
             return 0;
         }
     }
 
3e00abab
     for (i = 0; i < s->filter_channels; i++) {
dca97cbe
         av_free(bufin[i]);
3e00abab
         av_free(bufout[i]);
     }
1a565432
 
de6d9b64
     return nb_samples1;
 }
 
 void audio_resample_close(ReSampleContext *s)
 {
3e00abab
     int i;
aaaf1635
     av_resample_close(s->resample_context);
3e00abab
     for (i = 0; i < s->filter_channels; i++)
         av_freep(&s->temp[i]);
d1e3c6fd
     av_freep(&s->buffer[0]);
     av_freep(&s->buffer[1]);
     av_audio_convert_free(s->convert_ctx[0]);
     av_audio_convert_free(s->convert_ctx[1]);
6000abfa
     av_free(s);
de6d9b64
 }
30223b3b
 
 #endif