Commit 3e00abab authored by Alex Converse's avatar Alex Converse Committed by Alex Converse

Allow resampling with no channel count change for up to 8 channels.

parent 918a5409
...@@ -29,6 +29,8 @@ ...@@ -29,6 +29,8 @@
#include "libavutil/opt.h" #include "libavutil/opt.h"
#include "libavutil/samplefmt.h" #include "libavutil/samplefmt.h"
#define MAX_CHANNELS 8
struct AVResampleContext; struct AVResampleContext;
static const char *context_to_name(void *ptr) static const char *context_to_name(void *ptr)
...@@ -41,7 +43,7 @@ static const AVClass audioresample_context_class = { "ReSampleContext", context_ ...@@ -41,7 +43,7 @@ static const AVClass audioresample_context_class = { "ReSampleContext", context_
struct ReSampleContext { struct ReSampleContext {
struct AVResampleContext *resample_context; struct AVResampleContext *resample_context;
short *temp[2]; short *temp[MAX_CHANNELS];
int temp_len; int temp_len;
float ratio; float ratio;
/* channel convert */ /* channel convert */
...@@ -104,24 +106,25 @@ static void mono_to_stereo(short *output, short *input, int n1) ...@@ -104,24 +106,25 @@ static void mono_to_stereo(short *output, short *input, int n1)
} }
} }
/* XXX: should use more abstract 'N' channels system */ static void deinterleave(short **output, short *input, int channels, int samples)
static void stereo_split(short *output1, short *output2, short *input, int n)
{ {
int i; int i, j;
for(i=0;i<n;i++) { for (i = 0; i < samples; i++) {
*output1++ = *input++; for (j = 0; j < channels; j++) {
*output2++ = *input++; *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++) { for (i = 0; i < samples; i++) {
*output++ = *input1++; for (j = 0; j < channels; j++) {
*output++ = *input2++; *output++ = *input[j]++;
}
} }
} }
...@@ -151,14 +154,18 @@ ReSampleContext *av_audio_resample_init(int output_channels, int input_channels, ...@@ -151,14 +154,18 @@ ReSampleContext *av_audio_resample_init(int output_channels, int input_channels,
{ {
ReSampleContext *s; ReSampleContext *s;
if ( input_channels > 2) if (input_channels > MAX_CHANNELS)
{ {
av_log(NULL, AV_LOG_ERROR, "Resampling with input channels greater than 2 unsupported.\n"); av_log(NULL, AV_LOG_ERROR,
"Resampling with input channels greater than %d is unsupported.\n",
MAX_CHANNELS);
return NULL; return NULL;
} }
if (output_channels > 2 && !(output_channels == 6 && input_channels == 2)) { if ( output_channels > 2 &&
!(output_channels == 6 && input_channels == 2) &&
output_channels != input_channels) {
av_log(NULL, AV_LOG_ERROR, av_log(NULL, AV_LOG_ERROR,
"Resampling output channel count must be 1 or 2 for mono input and 1, 2 or 6 for stereo input.\n"); "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; return NULL;
} }
...@@ -206,14 +213,6 @@ ReSampleContext *av_audio_resample_init(int output_channels, int input_channels, ...@@ -206,14 +213,6 @@ ReSampleContext *av_audio_resample_init(int output_channels, int input_channels,
} }
} }
/*
* AC-3 output is the only case where filter_channels could be greater than 2.
* input channels can't be greater than 2, so resample the 2 channels and then
* expand to 6 channels after the resampling.
*/
if(s->filter_channels>2)
s->filter_channels = 2;
#define TAPS 16 #define TAPS 16
s->resample_context= av_resample_init(output_rate, input_rate, s->resample_context= av_resample_init(output_rate, input_rate,
filter_length, log2_phase_count, linear, cutoff); filter_length, log2_phase_count, linear, cutoff);
...@@ -228,9 +227,9 @@ ReSampleContext *av_audio_resample_init(int output_channels, int input_channels, ...@@ -228,9 +227,9 @@ ReSampleContext *av_audio_resample_init(int output_channels, int input_channels,
int audio_resample(ReSampleContext *s, short *output, short *input, int nb_samples) int audio_resample(ReSampleContext *s, short *output, short *input, int nb_samples)
{ {
int i, nb_samples1; int i, nb_samples1;
short *bufin[2]; short *bufin[MAX_CHANNELS];
short *bufout[2]; short *bufout[MAX_CHANNELS];
short *buftmp2[2], *buftmp3[2]; short *buftmp2[MAX_CHANNELS], *buftmp3[MAX_CHANNELS];
short *output_bak = NULL; short *output_bak = NULL;
int lenout; int lenout;
...@@ -291,12 +290,9 @@ int audio_resample(ReSampleContext *s, short *output, short *input, int nb_sampl ...@@ -291,12 +290,9 @@ int audio_resample(ReSampleContext *s, short *output, short *input, int nb_sampl
bufin[i]= av_malloc( (nb_samples + s->temp_len) * sizeof(short) ); bufin[i]= av_malloc( (nb_samples + s->temp_len) * sizeof(short) );
memcpy(bufin[i], s->temp[i], s->temp_len * sizeof(short)); memcpy(bufin[i], s->temp[i], s->temp_len * sizeof(short));
buftmp2[i] = bufin[i] + s->temp_len; buftmp2[i] = bufin[i] + s->temp_len;
bufout[i] = av_malloc(lenout * sizeof(short));
} }
/* make some zoom to avoid round pb */
bufout[0]= av_malloc( lenout * sizeof(short) );
bufout[1]= av_malloc( lenout * sizeof(short) );
if (s->input_channels == 2 && if (s->input_channels == 2 &&
s->output_channels == 1) { s->output_channels == 1) {
buftmp3[0] = output; buftmp3[0] = output;
...@@ -304,10 +300,11 @@ int audio_resample(ReSampleContext *s, short *output, short *input, int nb_sampl ...@@ -304,10 +300,11 @@ int audio_resample(ReSampleContext *s, short *output, short *input, int nb_sampl
} else if (s->output_channels >= 2 && s->input_channels == 1) { } else if (s->output_channels >= 2 && s->input_channels == 1) {
buftmp3[0] = bufout[0]; buftmp3[0] = bufout[0];
memcpy(buftmp2[0], input, nb_samples*sizeof(short)); memcpy(buftmp2[0], input, nb_samples*sizeof(short));
} else if (s->output_channels >= 2) { } else if (s->output_channels >= s->input_channels && s->input_channels >= 2) {
buftmp3[0] = bufout[0]; for (i = 0; i < s->input_channels; i++) {
buftmp3[1] = bufout[1]; buftmp3[i] = bufout[i];
stereo_split(buftmp2[0], buftmp2[1], input, nb_samples); }
deinterleave(buftmp2, input, s->input_channels, nb_samples);
} else { } else {
buftmp3[0] = output; buftmp3[0] = output;
memcpy(buftmp2[0], input, nb_samples*sizeof(short)); memcpy(buftmp2[0], input, nb_samples*sizeof(short));
...@@ -329,10 +326,10 @@ int audio_resample(ReSampleContext *s, short *output, short *input, int nb_sampl ...@@ -329,10 +326,10 @@ int audio_resample(ReSampleContext *s, short *output, short *input, int nb_sampl
if (s->output_channels == 2 && s->input_channels == 1) { if (s->output_channels == 2 && s->input_channels == 1) {
mono_to_stereo(output, buftmp3[0], nb_samples1); mono_to_stereo(output, buftmp3[0], nb_samples1);
} else if (s->output_channels == 2) { } else if (s->output_channels == 6 && s->input_channels == 2) {
stereo_mux(output, buftmp3[0], buftmp3[1], nb_samples1);
} else if (s->output_channels == 6) {
ac3_5p1_mux(output, buftmp3[0], buftmp3[1], nb_samples1); 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);
} }
if (s->sample_fmt[1] != AV_SAMPLE_FMT_S16) { if (s->sample_fmt[1] != AV_SAMPLE_FMT_S16) {
...@@ -348,19 +345,20 @@ int audio_resample(ReSampleContext *s, short *output, short *input, int nb_sampl ...@@ -348,19 +345,20 @@ int audio_resample(ReSampleContext *s, short *output, short *input, int nb_sampl
} }
} }
for(i=0; i<s->filter_channels; i++) for (i = 0; i < s->filter_channels; i++) {
av_free(bufin[i]); av_free(bufin[i]);
av_free(bufout[i]);
}
av_free(bufout[0]);
av_free(bufout[1]);
return nb_samples1; return nb_samples1;
} }
void audio_resample_close(ReSampleContext *s) void audio_resample_close(ReSampleContext *s)
{ {
int i;
av_resample_close(s->resample_context); av_resample_close(s->resample_context);
av_freep(&s->temp[0]); for (i = 0; i < s->filter_channels; i++)
av_freep(&s->temp[1]); av_freep(&s->temp[i]);
av_freep(&s->buffer[0]); av_freep(&s->buffer[0]);
av_freep(&s->buffer[1]); av_freep(&s->buffer[1]);
av_audio_convert_free(s->convert_ctx[0]); av_audio_convert_free(s->convert_ctx[0]);
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment