Home | History | Annotate | Download | only in tests
      1 /*
      2  * Copyright (C) 2014 The Android Open Source Project
      3  *
      4  * Licensed under the Apache License, Version 2.0 (the "License");
      5  * you may not use this file except in compliance with the License.
      6  * You may obtain a copy of the License at
      7  *
      8  *      http://www.apache.org/licenses/LICENSE-2.0
      9  *
     10  * Unless required by applicable law or agreed to in writing, software
     11  * distributed under the License is distributed on an "AS IS" BASIS,
     12  * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
     13  * See the License for the specific language governing permissions and
     14  * limitations under the License.
     15  */
     16 
     17 #include <stdio.h>
     18 #include <inttypes.h>
     19 #include <math.h>
     20 #include <vector>
     21 #include <audio_utils/primitives.h>
     22 #include <audio_utils/sndfile.h>
     23 #include <media/AudioBufferProvider.h>
     24 #include "AudioMixer.h"
     25 #include "test_utils.h"
     26 
     27 /* Testing is typically through creation of an output WAV file from several
     28  * source inputs, to be later analyzed by an audio program such as Audacity.
     29  *
     30  * Sine or chirp functions are typically more useful as input to the mixer
     31  * as they show up as straight lines on a spectrogram if successfully mixed.
     32  *
     33  * A sample shell script is provided: mixer_to_wave_tests.sh
     34  */
     35 
     36 using namespace android;
     37 
     38 static void usage(const char* name) {
     39     fprintf(stderr, "Usage: %s [-f] [-m] [-c channels]"
     40                     " [-s sample-rate] [-o <output-file>] [-a <aux-buffer-file>] [-P csv]"
     41                     " (<input-file> | <command>)+\n", name);
     42     fprintf(stderr, "    -f    enable floating point input track by default\n");
     43     fprintf(stderr, "    -m    enable floating point mixer output\n");
     44     fprintf(stderr, "    -c    number of mixer output channels\n");
     45     fprintf(stderr, "    -s    mixer sample-rate\n");
     46     fprintf(stderr, "    -o    <output-file> WAV file, pcm16 (or float if -m specified)\n");
     47     fprintf(stderr, "    -a    <aux-buffer-file>\n");
     48     fprintf(stderr, "    -P    # frames provided per call to resample() in CSV format\n");
     49     fprintf(stderr, "    <input-file> is a WAV file\n");
     50     fprintf(stderr, "    <command> can be 'sine:[(i|f),]<channels>,<frequency>,<samplerate>'\n");
     51     fprintf(stderr, "                     'chirp:[(i|f),]<channels>,<samplerate>'\n");
     52 }
     53 
     54 static int writeFile(const char *filename, const void *buffer,
     55         uint32_t sampleRate, uint32_t channels, size_t frames, bool isBufferFloat) {
     56     if (filename == NULL) {
     57         return 0; // ok to pass in NULL filename
     58     }
     59     // write output to file.
     60     SF_INFO info;
     61     info.frames = 0;
     62     info.samplerate = sampleRate;
     63     info.channels = channels;
     64     info.format = SF_FORMAT_WAV | (isBufferFloat ? SF_FORMAT_FLOAT : SF_FORMAT_PCM_16);
     65     printf("saving file:%s  channels:%u  samplerate:%u  frames:%zu\n",
     66             filename, info.channels, info.samplerate, frames);
     67     SNDFILE *sf = sf_open(filename, SFM_WRITE, &info);
     68     if (sf == NULL) {
     69         perror(filename);
     70         return EXIT_FAILURE;
     71     }
     72     if (isBufferFloat) {
     73         (void) sf_writef_float(sf, (float*)buffer, frames);
     74     } else {
     75         (void) sf_writef_short(sf, (short*)buffer, frames);
     76     }
     77     sf_close(sf);
     78     return EXIT_SUCCESS;
     79 }
     80 
     81 const char *parseFormat(const char *s, bool *useFloat) {
     82     if (!strncmp(s, "f,", 2)) {
     83         *useFloat = true;
     84         return s + 2;
     85     }
     86     if (!strncmp(s, "i,", 2)) {
     87         *useFloat = false;
     88         return s + 2;
     89     }
     90     return s;
     91 }
     92 
     93 int main(int argc, char* argv[]) {
     94     const char* const progname = argv[0];
     95     bool useInputFloat = false;
     96     bool useMixerFloat = false;
     97     bool useRamp = true;
     98     uint32_t outputSampleRate = 48000;
     99     uint32_t outputChannels = 2; // stereo for now
    100     std::vector<int> Pvalues;
    101     const char* outputFilename = NULL;
    102     const char* auxFilename = NULL;
    103     std::vector<int32_t> names;
    104     std::vector<SignalProvider> providers;
    105     std::vector<audio_format_t> formats;
    106 
    107     for (int ch; (ch = getopt(argc, argv, "fmc:s:o:a:P:")) != -1;) {
    108         switch (ch) {
    109         case 'f':
    110             useInputFloat = true;
    111             break;
    112         case 'm':
    113             useMixerFloat = true;
    114             break;
    115         case 'c':
    116             outputChannels = atoi(optarg);
    117             break;
    118         case 's':
    119             outputSampleRate = atoi(optarg);
    120             break;
    121         case 'o':
    122             outputFilename = optarg;
    123             break;
    124         case 'a':
    125             auxFilename = optarg;
    126             break;
    127         case 'P':
    128             if (parseCSV(optarg, Pvalues) < 0) {
    129                 fprintf(stderr, "incorrect syntax for -P option\n");
    130                 return EXIT_FAILURE;
    131             }
    132             break;
    133         case '?':
    134         default:
    135             usage(progname);
    136             return EXIT_FAILURE;
    137         }
    138     }
    139     argc -= optind;
    140     argv += optind;
    141 
    142     if (argc == 0) {
    143         usage(progname);
    144         return EXIT_FAILURE;
    145     }
    146     if ((unsigned)argc > AudioMixer::MAX_NUM_TRACKS) {
    147         fprintf(stderr, "too many tracks: %d > %u", argc, AudioMixer::MAX_NUM_TRACKS);
    148         return EXIT_FAILURE;
    149     }
    150 
    151     size_t outputFrames = 0;
    152 
    153     // create providers for each track
    154     names.resize(argc);
    155     providers.resize(argc);
    156     formats.resize(argc);
    157     for (int i = 0; i < argc; ++i) {
    158         static const char chirp[] = "chirp:";
    159         static const char sine[] = "sine:";
    160         static const double kSeconds = 1;
    161         bool useFloat = useInputFloat;
    162 
    163         if (!strncmp(argv[i], chirp, strlen(chirp))) {
    164             std::vector<int> v;
    165             const char *s = parseFormat(argv[i] + strlen(chirp), &useFloat);
    166 
    167             parseCSV(s, v);
    168             if (v.size() == 2) {
    169                 printf("creating chirp(%d %d)\n", v[0], v[1]);
    170                 if (useFloat) {
    171                     providers[i].setChirp<float>(v[0], 0, v[1]/2, v[1], kSeconds);
    172                     formats[i] = AUDIO_FORMAT_PCM_FLOAT;
    173                 } else {
    174                     providers[i].setChirp<int16_t>(v[0], 0, v[1]/2, v[1], kSeconds);
    175                     formats[i] = AUDIO_FORMAT_PCM_16_BIT;
    176                 }
    177                 providers[i].setIncr(Pvalues);
    178             } else {
    179                 fprintf(stderr, "malformed input '%s'\n", argv[i]);
    180             }
    181         } else if (!strncmp(argv[i], sine, strlen(sine))) {
    182             std::vector<int> v;
    183             const char *s = parseFormat(argv[i] + strlen(sine), &useFloat);
    184 
    185             parseCSV(s, v);
    186             if (v.size() == 3) {
    187                 printf("creating sine(%d %d %d)\n", v[0], v[1], v[2]);
    188                 if (useFloat) {
    189                     providers[i].setSine<float>(v[0], v[1], v[2], kSeconds);
    190                     formats[i] = AUDIO_FORMAT_PCM_FLOAT;
    191                 } else {
    192                     providers[i].setSine<int16_t>(v[0], v[1], v[2], kSeconds);
    193                     formats[i] = AUDIO_FORMAT_PCM_16_BIT;
    194                 }
    195                 providers[i].setIncr(Pvalues);
    196             } else {
    197                 fprintf(stderr, "malformed input '%s'\n", argv[i]);
    198             }
    199         } else {
    200             printf("creating filename(%s)\n", argv[i]);
    201             if (useInputFloat) {
    202                 providers[i].setFile<float>(argv[i]);
    203                 formats[i] = AUDIO_FORMAT_PCM_FLOAT;
    204             } else {
    205                 providers[i].setFile<short>(argv[i]);
    206                 formats[i] = AUDIO_FORMAT_PCM_16_BIT;
    207             }
    208             providers[i].setIncr(Pvalues);
    209         }
    210         // calculate the number of output frames
    211         size_t nframes = (int64_t) providers[i].getNumFrames() * outputSampleRate
    212                 / providers[i].getSampleRate();
    213         if (i == 0 || outputFrames > nframes) { // choose minimum for outputFrames
    214             outputFrames = nframes;
    215         }
    216     }
    217 
    218     // create the output buffer.
    219     const size_t outputFrameSize = outputChannels
    220             * (useMixerFloat ? sizeof(float) : sizeof(int16_t));
    221     const size_t outputSize = outputFrames * outputFrameSize;
    222     const audio_channel_mask_t outputChannelMask =
    223             audio_channel_out_mask_from_count(outputChannels);
    224     void *outputAddr = NULL;
    225     (void) posix_memalign(&outputAddr, 32, outputSize);
    226     memset(outputAddr, 0, outputSize);
    227 
    228     // create the aux buffer, if needed.
    229     const size_t auxFrameSize = sizeof(int32_t); // Q4.27 always
    230     const size_t auxSize = outputFrames * auxFrameSize;
    231     void *auxAddr = NULL;
    232     if (auxFilename) {
    233         (void) posix_memalign(&auxAddr, 32, auxSize);
    234         memset(auxAddr, 0, auxSize);
    235     }
    236 
    237     // create the mixer.
    238     const size_t mixerFrameCount = 320; // typical numbers may range from 240 or 960
    239     AudioMixer *mixer = new AudioMixer(mixerFrameCount, outputSampleRate);
    240     audio_format_t mixerFormat = useMixerFloat
    241             ? AUDIO_FORMAT_PCM_FLOAT : AUDIO_FORMAT_PCM_16_BIT;
    242     float f = AudioMixer::UNITY_GAIN_FLOAT / providers.size(); // normalize volume by # tracks
    243     static float f0; // zero
    244 
    245     // set up the tracks.
    246     for (size_t i = 0; i < providers.size(); ++i) {
    247         //printf("track %d out of %d\n", i, providers.size());
    248         uint32_t channelMask = audio_channel_out_mask_from_count(providers[i].getNumChannels());
    249         int32_t name = mixer->getTrackName(channelMask,
    250                 formats[i], AUDIO_SESSION_OUTPUT_MIX);
    251         ALOG_ASSERT(name >= 0);
    252         names[i] = name;
    253         mixer->setBufferProvider(name, &providers[i]);
    254         mixer->setParameter(name, AudioMixer::TRACK, AudioMixer::MAIN_BUFFER,
    255                 (void *)outputAddr);
    256         mixer->setParameter(
    257                 name,
    258                 AudioMixer::TRACK,
    259                 AudioMixer::MIXER_FORMAT,
    260                 (void *)(uintptr_t)mixerFormat);
    261         mixer->setParameter(
    262                 name,
    263                 AudioMixer::TRACK,
    264                 AudioMixer::FORMAT,
    265                 (void *)(uintptr_t)formats[i]);
    266         mixer->setParameter(
    267                 name,
    268                 AudioMixer::TRACK,
    269                 AudioMixer::MIXER_CHANNEL_MASK,
    270                 (void *)(uintptr_t)outputChannelMask);
    271         mixer->setParameter(
    272                 name,
    273                 AudioMixer::TRACK,
    274                 AudioMixer::CHANNEL_MASK,
    275                 (void *)(uintptr_t)channelMask);
    276         mixer->setParameter(
    277                 name,
    278                 AudioMixer::RESAMPLE,
    279                 AudioMixer::SAMPLE_RATE,
    280                 (void *)(uintptr_t)providers[i].getSampleRate());
    281         if (useRamp) {
    282             mixer->setParameter(name, AudioMixer::VOLUME, AudioMixer::VOLUME0, &f0);
    283             mixer->setParameter(name, AudioMixer::VOLUME, AudioMixer::VOLUME1, &f0);
    284             mixer->setParameter(name, AudioMixer::RAMP_VOLUME, AudioMixer::VOLUME0, &f);
    285             mixer->setParameter(name, AudioMixer::RAMP_VOLUME, AudioMixer::VOLUME1, &f);
    286         } else {
    287             mixer->setParameter(name, AudioMixer::VOLUME, AudioMixer::VOLUME0, &f);
    288             mixer->setParameter(name, AudioMixer::VOLUME, AudioMixer::VOLUME1, &f);
    289         }
    290         if (auxFilename) {
    291             mixer->setParameter(name, AudioMixer::TRACK, AudioMixer::AUX_BUFFER,
    292                     (void *) auxAddr);
    293             mixer->setParameter(name, AudioMixer::VOLUME, AudioMixer::AUXLEVEL, &f0);
    294             mixer->setParameter(name, AudioMixer::RAMP_VOLUME, AudioMixer::AUXLEVEL, &f);
    295         }
    296         mixer->enable(name);
    297     }
    298 
    299     // pump the mixer to process data.
    300     size_t i;
    301     for (i = 0; i < outputFrames - mixerFrameCount; i += mixerFrameCount) {
    302         for (size_t j = 0; j < names.size(); ++j) {
    303             mixer->setParameter(names[j], AudioMixer::TRACK, AudioMixer::MAIN_BUFFER,
    304                     (char *) outputAddr + i * outputFrameSize);
    305             if (auxFilename) {
    306                 mixer->setParameter(names[j], AudioMixer::TRACK, AudioMixer::AUX_BUFFER,
    307                         (char *) auxAddr + i * auxFrameSize);
    308             }
    309         }
    310         mixer->process(AudioBufferProvider::kInvalidPTS);
    311     }
    312     outputFrames = i; // reset output frames to the data actually produced.
    313 
    314     // write to files
    315     writeFile(outputFilename, outputAddr,
    316             outputSampleRate, outputChannels, outputFrames, useMixerFloat);
    317     if (auxFilename) {
    318         // Aux buffer is always in q4_27 format for now.
    319         // memcpy_to_i16_from_q4_27(), but with stereo frame count (not sample count)
    320         ditherAndClamp((int32_t*)auxAddr, (int32_t*)auxAddr, outputFrames >> 1);
    321         writeFile(auxFilename, auxAddr, outputSampleRate, 1, outputFrames, false);
    322     }
    323 
    324     delete mixer;
    325     free(outputAddr);
    326     free(auxAddr);
    327     return EXIT_SUCCESS;
    328 }
    329