/*
- * Sample rate convertion for both audio and video
- * Copyright (c) 2000 Fabrice Bellard.
+ * samplerate conversion for both audio and video
+ * Copyright (c) 2000 Fabrice Bellard
*
- * This library is free software; you can redistribute it and/or
+ * This file is part of Libav.
+ *
+ * Libav is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
- * version 2 of the License, or (at your option) any later version.
+ * version 2.1 of the License, or (at your option) any later version.
*
- * This library is distributed in the hope that it will be useful,
+ * Libav 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
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
- * License along with this library; if not, write to the Free Software
- * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
+ * License along with Libav; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
/**
- * @file resample.c
- * Sample rate convertion for both audio and video.
+ * @file
+ * samplerate conversion for both audio and video
*/
-#include "avcodec.h"
-#include "os_support.h"
-
-typedef struct {
- /* fractional resampling */
- uint32_t incr; /* fractional increment */
- uint32_t frac;
- int last_sample;
- /* integer down sample */
- int iratio; /* integer divison ratio */
- int icount, isum;
- int inv;
-} ReSampleChannelContext;
+#include <string.h>
-struct ReSampleContext {
- ReSampleChannelContext channel_ctx[2];
- float ratio;
- /* channel convert */
- int input_channels, output_channels, filter_channels;
-};
+#include "avcodec.h"
+#include "audioconvert.h"
+#include "libavutil/opt.h"
+#include "libavutil/mem.h"
+#include "libavutil/samplefmt.h"
+#if FF_API_AVCODEC_RESAMPLE
-#define FRAC_BITS 16
-#define FRAC (1 << FRAC_BITS)
+#define MAX_CHANNELS 8
-static void init_mono_resample(ReSampleChannelContext *s, float ratio)
-{
- ratio = 1.0 / ratio;
- s->iratio = (int)floorf(ratio);
- if (s->iratio == 0)
- s->iratio = 1;
- s->incr = (int)((ratio / s->iratio) * FRAC);
- s->frac = FRAC;
- s->last_sample = 0;
- s->icount = s->iratio;
- s->isum = 0;
- s->inv = (FRAC / s->iratio);
-}
+struct AVResampleContext;
-/* fractional audio resampling */
-static int fractional_resample(ReSampleChannelContext *s, short *output, short *input, int nb_samples)
+static const char *context_to_name(void *ptr)
{
- unsigned int frac, incr;
- int l0, l1;
- short *q, *p, *pend;
-
- l0 = s->last_sample;
- incr = s->incr;
- frac = s->frac;
-
- p = input;
- pend = input + nb_samples;
- q = output;
-
- l1 = *p++;
- for(;;) {
- /* interpolate */
- *q++ = (l0 * (FRAC - frac) + l1 * frac) >> FRAC_BITS;
- frac = frac + s->incr;
- while (frac >= FRAC) {
- frac -= FRAC;
- if (p >= pend)
- goto the_end;
- l0 = l1;
- l1 = *p++;
- }
- }
- the_end:
- s->last_sample = l1;
- s->frac = frac;
- return q - output;
+ return "audioresample";
}
-static int integer_downsample(ReSampleChannelContext *s, short *output, short *input, int nb_samples)
-{
- short *q, *p, *pend;
- int c, sum;
-
- p = input;
- pend = input + nb_samples;
- q = output;
-
- c = s->icount;
- sum = s->isum;
+static const AVOption options[] = {{NULL}};
+static const AVClass audioresample_context_class = {
+ "ReSampleContext", context_to_name, options, LIBAVUTIL_VERSION_INT
+};
- for(;;) {
- sum += *p++;
- if (--c == 0) {
- *q++ = (sum * s->inv) >> FRAC_BITS;
- c = s->iratio;
- sum = 0;
- }
- if (p >= pend)
- break;
- }
- s->isum = sum;
- s->icount = c;
- return q - output;
-}
+struct ReSampleContext {
+ struct AVResampleContext *resample_context;
+ short *temp[MAX_CHANNELS];
+ int temp_len;
+ float ratio;
+ /* channel convert */
+ int input_channels, output_channels, filter_channels;
+ AVAudioConvert *convert_ctx[2];
+ enum AVSampleFormat sample_fmt[2]; ///< input and output sample format
+ 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
+};
/* n1: number of samples */
static void stereo_to_mono(short *output, short *input, int n1)
}
}
-/* XXX: should use more abstract 'N' channels system */
-static void stereo_split(short *output1, short *output2, short *input, int n)
+static void deinterleave(short **output, short *input, int channels, int samples)
{
- int i;
+ int i, j;
- for(i=0;i<n;i++) {
- *output1++ = *input++;
- *output2++ = *input++;
+ for (i = 0; i < samples; i++) {
+ for (j = 0; j < channels; j++) {
+ *output[j]++ = *input++;
+ }
}
}
-static void stereo_mux(short *output, short *input1, short *input2, int n)
+static void interleave(short *output, short **input, int channels, int samples)
{
- int i;
+ int i, j;
- for(i=0;i<n;i++) {
- *output++ = *input1++;
- *output++ = *input2++;
+ for (i = 0; i < samples; i++) {
+ for (j = 0; j < channels; j++) {
+ *output++ = *input[j]++;
+ }
}
}
-static int mono_resample(ReSampleChannelContext *s, short *output, short *input, int nb_samples)
+static void ac3_5p1_mux(short *output, short *input1, short *input2, int n)
{
- short *buf1;
- short *buftmp;
-
- buf1= (short*)av_malloc( nb_samples * sizeof(short) );
-
- /* first downsample by an integer factor with averaging filter */
- if (s->iratio > 1) {
- buftmp = buf1;
- nb_samples = integer_downsample(s, buftmp, input, nb_samples);
- } else {
- buftmp = input;
- }
-
- /* then do a fractional resampling with linear interpolation */
- if (s->incr != FRAC) {
- nb_samples = fractional_resample(s, output, buftmp, nb_samples);
- } else {
- memcpy(output, buftmp, nb_samples * sizeof(short));
+ int i;
+ 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 */
}
- av_free(buf1);
- return nb_samples;
}
-ReSampleContext *audio_resample_init(int output_channels, int input_channels,
- int output_rate, int input_rate)
+ReSampleContext *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)
{
ReSampleContext *s;
- int i;
-
- if (output_channels > 2 || input_channels > 2)
+
+ if (input_channels > MAX_CHANNELS) {
+ av_log(NULL, AV_LOG_ERROR,
+ "Resampling with input channels greater than %d is unsupported.\n",
+ MAX_CHANNELS);
+ return NULL;
+ }
+ if (output_channels != input_channels &&
+ (input_channels > 2 ||
+ output_channels > 2 &&
+ !(output_channels == 6 && input_channels == 2))) {
+ av_log(NULL, AV_LOG_ERROR,
+ "Resampling output channel count must be 1 or 2 for mono input; 1, 2 or 6 for stereo input; or N for N channel input.\n");
return NULL;
+ }
s = av_mallocz(sizeof(ReSampleContext));
- if (!s)
+ if (!s) {
+ av_log(NULL, AV_LOG_ERROR, "Can't allocate memory for resample context.\n");
return NULL;
+ }
s->ratio = (float)output_rate / (float)input_rate;
-
+
s->input_channels = input_channels;
s->output_channels = output_channels;
-
+
s->filter_channels = s->input_channels;
if (s->output_channels < s->filter_channels)
s->filter_channels = s->output_channels;
- for(i=0;i<s->filter_channels;i++) {
- init_mono_resample(&s->channel_ctx[i], s->ratio);
+ s->sample_fmt[0] = sample_fmt_in;
+ s->sample_fmt[1] = sample_fmt_out;
+ 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]);
+
+ if (s->sample_fmt[0] != AV_SAMPLE_FMT_S16) {
+ if (!(s->convert_ctx[0] = av_audio_convert_alloc(AV_SAMPLE_FMT_S16, 1,
+ s->sample_fmt[0], 1, NULL, 0))) {
+ av_log(s, AV_LOG_ERROR,
+ "Cannot convert %s sample format to s16 sample format\n",
+ av_get_sample_fmt_name(s->sample_fmt[0]));
+ av_free(s);
+ return NULL;
+ }
}
+
+ if (s->sample_fmt[1] != AV_SAMPLE_FMT_S16) {
+ if (!(s->convert_ctx[1] = av_audio_convert_alloc(s->sample_fmt[1], 1,
+ AV_SAMPLE_FMT_S16, 1, NULL, 0))) {
+ av_log(s, AV_LOG_ERROR,
+ "Cannot convert s16 sample format to %s sample format\n",
+ av_get_sample_fmt_name(s->sample_fmt[1]));
+ av_audio_convert_free(s->convert_ctx[0]);
+ av_free(s);
+ return NULL;
+ }
+ }
+
+ s->resample_context = av_resample_init(output_rate, input_rate,
+ filter_length, log2_phase_count,
+ linear, cutoff);
+
+ *(const AVClass**)s->resample_context = &audioresample_context_class;
+
return s;
}
/* resample audio. 'nb_samples' is the number of input samples */
/* XXX: optimize it ! */
-/* XXX: do it with polyphase filters, since the quality here is
- HORRIBLE. Return the number of samples available in output */
int audio_resample(ReSampleContext *s, short *output, short *input, int nb_samples)
{
int i, nb_samples1;
- short *bufin[2];
- short *bufout[2];
- short *buftmp2[2], *buftmp3[2];
+ short *bufin[MAX_CHANNELS];
+ short *bufout[MAX_CHANNELS];
+ short *buftmp2[MAX_CHANNELS], *buftmp3[MAX_CHANNELS];
+ short *output_bak = NULL;
int lenout;
- if (s->input_channels == s->output_channels && s->ratio == 1.0) {
+ if (s->input_channels == s->output_channels && s->ratio == 1.0 && 0) {
/* nothing to do */
memcpy(output, input, nb_samples * s->input_channels * sizeof(short));
return nb_samples;
}
+ if (s->sample_fmt[0] != AV_SAMPLE_FMT_S16) {
+ int istride[1] = { s->sample_size[0] };
+ int ostride[1] = { 2 };
+ const void *ibuf[1] = { input };
+ void *obuf[1];
+ unsigned input_size = nb_samples * s->input_channels * 2;
+
+ 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]) {
+ av_log(s->resample_context, AV_LOG_ERROR, "Could not allocate buffer\n");
+ return 0;
+ }
+ }
+
+ obuf[0] = s->buffer[0];
+
+ if (av_audio_convert(s->convert_ctx[0], obuf, ostride,
+ ibuf, istride, nb_samples * s->input_channels) < 0) {
+ av_log(s->resample_context, AV_LOG_ERROR,
+ "Audio sample format conversion failed\n");
+ return 0;
+ }
+
+ input = s->buffer[0];
+ }
+
+ lenout = 4 * nb_samples * s->ratio + 16;
+
+ if (s->sample_fmt[1] != AV_SAMPLE_FMT_S16) {
+ int out_size = lenout * av_get_bytes_per_sample(s->sample_fmt[1]) *
+ s->output_channels;
+ output_bak = output;
+
+ if (!s->buffer_size[1] || s->buffer_size[1] < out_size) {
+ av_free(s->buffer[1]);
+ s->buffer_size[1] = out_size;
+ s->buffer[1] = av_malloc(s->buffer_size[1]);
+ if (!s->buffer[1]) {
+ av_log(s->resample_context, AV_LOG_ERROR, "Could not allocate buffer\n");
+ return 0;
+ }
+ }
+
+ output = s->buffer[1];
+ }
+
/* XXX: move those malloc to resample init code */
- bufin[0]= (short*) av_malloc( nb_samples * sizeof(short) );
- bufin[1]= (short*) av_malloc( nb_samples * sizeof(short) );
-
- /* make some zoom to avoid round pb */
- lenout= (int)(nb_samples * s->ratio) + 16;
- bufout[0]= (short*) av_malloc( lenout * sizeof(short) );
- bufout[1]= (short*) av_malloc( lenout * sizeof(short) );
-
- if (s->input_channels == 2 &&
- s->output_channels == 1) {
- buftmp2[0] = bufin[0];
+ for (i = 0; i < s->filter_channels; i++) {
+ bufin[i] = av_malloc((nb_samples + s->temp_len) * sizeof(short));
+ memcpy(bufin[i], s->temp[i], s->temp_len * sizeof(short));
+ buftmp2[i] = bufin[i] + s->temp_len;
+ bufout[i] = av_malloc(lenout * sizeof(short));
+ }
+
+ if (s->input_channels == 2 && s->output_channels == 1) {
buftmp3[0] = output;
stereo_to_mono(buftmp2[0], input, nb_samples);
- } else if (s->output_channels == 2 && s->input_channels == 1) {
- buftmp2[0] = input;
+ } else if (s->output_channels >= 2 && s->input_channels == 1) {
buftmp3[0] = bufout[0];
- } else if (s->output_channels == 2) {
- buftmp2[0] = bufin[0];
- buftmp2[1] = bufin[1];
- buftmp3[0] = bufout[0];
- buftmp3[1] = bufout[1];
- stereo_split(buftmp2[0], buftmp2[1], input, nb_samples);
+ memcpy(buftmp2[0], input, nb_samples * sizeof(short));
+ } 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);
} else {
- buftmp2[0] = input;
buftmp3[0] = output;
+ memcpy(buftmp2[0], input, nb_samples * sizeof(short));
}
+ nb_samples += s->temp_len;
+
/* resample each channel */
nb_samples1 = 0; /* avoid warning */
- for(i=0;i<s->filter_channels;i++) {
- nb_samples1 = mono_resample(&s->channel_ctx[i], buftmp3[i], buftmp2[i], nb_samples);
+ for (i = 0; i < s->filter_channels; i++) {
+ int consumed;
+ int is_last = i + 1 == s->filter_channels;
+
+ 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));
}
if (s->output_channels == 2 && s->input_channels == 1) {
mono_to_stereo(output, buftmp3[0], nb_samples1);
- } else if (s->output_channels == 2) {
- stereo_mux(output, buftmp3[0], buftmp3[1], nb_samples1);
+ } else if (s->output_channels == 6 && s->input_channels == 2) {
+ ac3_5p1_mux(output, buftmp3[0], buftmp3[1], nb_samples1);
+ } else if (s->output_channels == s->input_channels && s->input_channels >= 2) {
+ interleave(output, buftmp3, s->output_channels, nb_samples1);
}
- av_free(bufin[0]);
- av_free(bufin[1]);
+ if (s->sample_fmt[1] != AV_SAMPLE_FMT_S16) {
+ 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,
+ ibuf, istride, nb_samples1 * s->output_channels) < 0) {
+ av_log(s->resample_context, AV_LOG_ERROR,
+ "Audio sample format conversion failed\n");
+ return 0;
+ }
+ }
+
+ for (i = 0; i < s->filter_channels; i++) {
+ av_free(bufin[i]);
+ av_free(bufout[i]);
+ }
- av_free(bufout[0]);
- av_free(bufout[1]);
return nb_samples1;
}
void audio_resample_close(ReSampleContext *s)
{
+ int i;
+ av_resample_close(s->resample_context);
+ for (i = 0; i < s->filter_channels; i++)
+ av_freep(&s->temp[i]);
+ 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]);
av_free(s);
}
+
+#endif
projects / ffmpeg / blobdiff