extend resampling API, add S16 internal conversion

Originally committed as revision 17163 to svn://svn.ffmpeg.org/ffmpeg/trunk

ffmpeg.c

@@ -555,12 +555,12 @@ static void do_audio_out(AVFormatContext *s,
         ost->audio_resample = 1;
 
     if (ost->audio_resample && !ost->resample) {
-        if (dec->sample_fmt != SAMPLE_FMT_S16) {
-            fprintf(stderr, "Audio resampler only works with 16 bits per sample, patch welcome.\n");
-            av_exit(1);
-        }
-        ost->resample = audio_resample_init(enc->channels,    dec->channels,
-                                            enc->sample_rate, dec->sample_rate);
+        if (dec->sample_fmt != SAMPLE_FMT_S16)
+            fprintf(stderr, "Warning, using s16 intermediate sample format for resampling\n");
+        ost->resample = av_audio_resample_init(enc->channels,    dec->channels,
+                                               enc->sample_rate, dec->sample_rate,
+                                               enc->sample_fmt,  dec->sample_fmt,
+                                               16, 10, 0, 0.8);
         if (!ost->resample) {
             fprintf(stderr, "Can not resample %d channels @ %d Hz to %d channels @ %d Hz\n",
                     dec->channels, dec->sample_rate,
@@ -570,7 +570,7 @@ static void do_audio_out(AVFormatContext *s,
     }
 
 #define MAKE_SFMT_PAIR(a,b) ((a)+SAMPLE_FMT_NB*(b))
-    if (dec->sample_fmt!=enc->sample_fmt &&
+    if (!ost->audio_resample && dec->sample_fmt!=enc->sample_fmt &&
         MAKE_SFMT_PAIR(enc->sample_fmt,dec->sample_fmt)!=ost->reformat_pair) {
         if (!audio_out2)
             audio_out2 = av_malloc(audio_out_size);
@@ -647,7 +647,7 @@ static void do_audio_out(AVFormatContext *s,
         size_out = size;
     }
 
-    if (dec->sample_fmt!=enc->sample_fmt) {
+    if (!ost->audio_resample && dec->sample_fmt!=enc->sample_fmt) {
         const void *ibuf[6]= {buftmp};
         void *obuf[6]= {audio_out2};
         int istride[6]= {isize};

libavcodec/avcodec.h

@@ -30,7 +30,7 @@
 #include "libavutil/avutil.h"
 
 #define LIBAVCODEC_VERSION_MAJOR 52
-#define LIBAVCODEC_VERSION_MINOR 14
+#define LIBAVCODEC_VERSION_MINOR 15
 #define LIBAVCODEC_VERSION_MICRO  0
 
 #define LIBAVCODEC_VERSION_INT  AV_VERSION_INT(LIBAVCODEC_VERSION_MAJOR, \
@@ -2443,8 +2443,36 @@ struct AVResampleContext;
 
 typedef struct ReSampleContext ReSampleContext;
 
-ReSampleContext *audio_resample_init(int output_channels, int input_channels,
-                                     int output_rate, int input_rate);
+#if LIBAVCODEC_VERSION_MAJOR < 53
+/**
+ * @deprecated Use av_audio_resample_init() instead.
+ */
+attribute_deprecated ReSampleContext *audio_resample_init(int output_channels, int input_channels,
+                                                          int output_rate, int input_rate);
+#endif
+/**
+ *  Initializes audio resampling context
+ *
+ * @param output_channels  number of output channels
+ * @param input_channels   number of input channels
+ * @param output_rate      output sample rate
+ * @param input_rate       input sample rate
+ * @param sample_fmt_out   requested output sample format
+ * @param sample_fmt_in    input sample format
+ * @param filter_length    length of each FIR filter in the filterbank relative to the cutoff freq
+ * @param log2_phase_count log2 of the number of entries in the polyphase filterbank
+ * @param linear           If 1 then the used FIR filter will be linearly interpolated
+                           between the 2 closest, if 0 the closest will be used
+ * @param cutoff           cutoff frequency, 1.0 corresponds to half the output sampling rate
+ * @return allocated ReSampleContext, NULL if error occured
+ */
+ReSampleContext *av_audio_resample_init(int output_channels, int input_channels,
+                                        int output_rate, int input_rate,
+                                        enum SampleFormat sample_fmt_out,
+                                        enum SampleFormat sample_fmt_in,
+                                        int filter_length, int log2_phase_count,
+                                        int linear, double cutoff);
+
 int audio_resample(ReSampleContext *s, short *output, short *input, int nb_samples);
 void audio_resample_close(ReSampleContext *s);
 

libavcodec/resample.c

@@ -25,16 +25,32 @@
  */
 
 #include "avcodec.h"
+#include "audioconvert.h"
+#include "opt.h"
 
 struct AVResampleContext;
 
+static const char *context_to_name(void *ptr)
+{
+    return "audioresample";
+}
+
+static const AVOption options[] = {{NULL}};
+static const AVClass audioresample_context_class = { "ReSampleContext", context_to_name, options };
+
 struct ReSampleContext {
+    const AVClass *av_class;
     struct AVResampleContext *resample_context;
     short *temp[2];
     int temp_len;
     float ratio;
     /* channel convert */
     int input_channels, output_channels, filter_channels;
+    AVAudioConvert *convert_ctx[2];
+    enum SampleFormat 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 */
@@ -126,8 +142,12 @@ static void ac3_5p1_mux(short *output, short *input1, short *input2, int n)
     }
 }
 
-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 SampleFormat sample_fmt_out,
+                                        enum SampleFormat sample_fmt_in,
+                                        int filter_length, int log2_phase_count,
+                                        int linear, double cutoff)
 {
     ReSampleContext *s;
 
@@ -153,6 +173,34 @@ ReSampleContext *audio_resample_init(int output_channels, int input_channels,
     if (s->output_channels < s->filter_channels)
         s->filter_channels = s->output_channels;
 
+    s->sample_fmt [0] = sample_fmt_in;
+    s->sample_fmt [1] = sample_fmt_out;
+    s->sample_size[0] = av_get_bits_per_sample_format(s->sample_fmt[0])>>3;
+    s->sample_size[1] = av_get_bits_per_sample_format(s->sample_fmt[1])>>3;
+
+    if (s->sample_fmt[0] != SAMPLE_FMT_S16) {
+        if (!(s->convert_ctx[0] = av_audio_convert_alloc(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",
+                   avcodec_get_sample_fmt_name(s->sample_fmt[0]));
+            av_free(s);
+            return NULL;
+        }
+    }
+
+    if (s->sample_fmt[1] != SAMPLE_FMT_S16) {
+        if (!(s->convert_ctx[1] = av_audio_convert_alloc(s->sample_fmt[1], 1,
+                                                         SAMPLE_FMT_S16, 1, NULL, 0))) {
+            av_log(s, AV_LOG_ERROR,
+                   "Cannot convert s16 sample format to %s sample format\n",
+                   avcodec_get_sample_fmt_name(s->sample_fmt[1]));
+            av_audio_convert_free(s->convert_ctx[0]);
+            av_free(s);
+            return NULL;
+        }
+    }
+
 /*
  * 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
@@ -162,11 +210,25 @@ ReSampleContext *audio_resample_init(int output_channels, int input_channels,
       s->filter_channels = 2;
 
 #define TAPS 16
-    s->resample_context= av_resample_init(output_rate, input_rate, TAPS, 10, 0, 0.8);
+    s->resample_context= av_resample_init(output_rate, input_rate,
+                         filter_length, log2_phase_count, linear, cutoff);
+
+    s->av_class= &audioresample_context_class;
 
     return s;
 }
 
+#if LIBAVCODEC_VERSION_MAJOR < 53
+ReSampleContext *audio_resample_init(int output_channels, int input_channels,
+                                     int output_rate, int input_rate)
+{
+    return av_audio_resample_init(output_channels, input_channels,
+                                  output_rate, input_rate,
+                                  SAMPLE_FMT_S16, SAMPLE_FMT_S16,
+                                  TAPS, 10, 0, 0.8);
+}
+#endif
+
 /* 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)
@@ -175,6 +237,7 @@ int audio_resample(ReSampleContext *s, short *output, short *input, int nb_sampl
     short *bufin[2];
     short *bufout[2];
     short *buftmp2[2], *buftmp3[2];
+    short *output_bak = NULL;
     int lenout;
 
     if (s->input_channels == s->output_channels && s->ratio == 1.0 && 0) {
@@ -183,6 +246,52 @@ int audio_resample(ReSampleContext *s, short *output, short *input, int nb_sampl
         return nb_samples;
     }
 
+    if (s->sample_fmt[0] != 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*s->sample_size[0];
+
+        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, 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, 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] != SAMPLE_FMT_S16) {
+        output_bak = output;
+
+        if (!s->buffer_size[1] || s->buffer_size[1] < lenout) {
+            av_free(s->buffer[1]);
+            s->buffer_size[1] = lenout;
+            s->buffer[1] = av_malloc(s->buffer_size[1]);
+            if (!s->buffer[1]) {
+                av_log(s, AV_LOG_ERROR, "Could not allocate buffer\n");
+                return 0;
+            }
+        }
+
+        output = s->buffer[1];
+    }
+
     /* XXX: move those malloc to resample init code */
     for(i=0; i<s->filter_channels; i++){
         bufin[i]= av_malloc( (nb_samples + s->temp_len) * sizeof(short) );
@@ -191,7 +300,6 @@ int audio_resample(ReSampleContext *s, short *output, short *input, int nb_sampl
     }
 
     /* make some zoom to avoid round pb */
-    lenout= 4*nb_samples * s->ratio + 16;
     bufout[0]= av_malloc( lenout * sizeof(short) );
     bufout[1]= av_malloc( lenout * sizeof(short) );
 
@@ -233,6 +341,19 @@ int audio_resample(ReSampleContext *s, short *output, short *input, int nb_sampl
         ac3_5p1_mux(output, buftmp3[0], buftmp3[1], nb_samples1);
     }
 
+    if (s->sample_fmt[1] != 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, AV_LOG_ERROR, "Audio sample format convertion failed\n");
+            return 0;
+        }
+    }
+
     for(i=0; i<s->filter_channels; i++)
         av_free(bufin[i]);
 
@@ -246,5 +367,9 @@ void audio_resample_close(ReSampleContext *s)
     av_resample_close(s->resample_context);
     av_freep(&s->temp[0]);
     av_freep(&s->temp[1]);
+    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);
 }