Home | History | Annotate | Download | only in media
      1 /*
      2 **
      3 ** Copyright 2007, The Android Open Source Project
      4 **
      5 ** Licensed under the Apache License, Version 2.0 (the "License");
      6 ** you may not use this file except in compliance with the License.
      7 ** You may obtain a copy of the License at
      8 **
      9 **     http://www.apache.org/licenses/LICENSE-2.0
     10 **
     11 ** Unless required by applicable law or agreed to in writing, software
     12 ** distributed under the License is distributed on an "AS IS" BASIS,
     13 ** WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
     14 ** See the License for the specific language governing permissions and
     15 ** limitations under the License.
     16 */
     17 
     18 #ifndef ANDROID_AUDIO_MIXER_H
     19 #define ANDROID_AUDIO_MIXER_H
     20 
     21 #include <pthread.h>
     22 #include <sstream>
     23 #include <stdint.h>
     24 #include <sys/types.h>
     25 #include <unordered_map>
     26 
     27 #include <media/AudioBufferProvider.h>
     28 #include <media/AudioResampler.h>
     29 #include <media/AudioResamplerPublic.h>
     30 #include <media/BufferProviders.h>
     31 #include <media/nblog/NBLog.h>
     32 #include <system/audio.h>
     33 #include <utils/Compat.h>
     34 #include <utils/threads.h>
     35 
     36 // FIXME This is actually unity gain, which might not be max in future, expressed in U.12
     37 #define MAX_GAIN_INT AudioMixer::UNITY_GAIN_INT
     38 
     39 // This must match frameworks/av/services/audioflinger/Configuration.h
     40 #define FLOAT_AUX
     41 
     42 namespace android {
     43 
     44 // ----------------------------------------------------------------------------
     45 
     46 class AudioMixer
     47 {
     48 public:
     49     // Do not change these unless underlying code changes.
     50     // This mixer has a hard-coded upper limit of 8 channels for output.
     51     static constexpr uint32_t MAX_NUM_CHANNELS = FCC_8;
     52     static constexpr uint32_t MAX_NUM_VOLUMES = FCC_2; // stereo volume only
     53     // maximum number of channels supported for the content
     54     static const uint32_t MAX_NUM_CHANNELS_TO_DOWNMIX = AUDIO_CHANNEL_COUNT_MAX;
     55 
     56     static const uint16_t UNITY_GAIN_INT = 0x1000;
     57     static const CONSTEXPR float UNITY_GAIN_FLOAT = 1.0f;
     58 
     59     enum { // names
     60         // setParameter targets
     61         TRACK           = 0x3000,
     62         RESAMPLE        = 0x3001,
     63         RAMP_VOLUME     = 0x3002, // ramp to new volume
     64         VOLUME          = 0x3003, // don't ramp
     65         TIMESTRETCH     = 0x3004,
     66 
     67         // set Parameter names
     68         // for target TRACK
     69         CHANNEL_MASK    = 0x4000,
     70         FORMAT          = 0x4001,
     71         MAIN_BUFFER     = 0x4002,
     72         AUX_BUFFER      = 0x4003,
     73         DOWNMIX_TYPE    = 0X4004,
     74         MIXER_FORMAT    = 0x4005, // AUDIO_FORMAT_PCM_(FLOAT|16_BIT)
     75         MIXER_CHANNEL_MASK = 0x4006, // Channel mask for mixer output
     76         // for target RESAMPLE
     77         SAMPLE_RATE     = 0x4100, // Configure sample rate conversion on this track name;
     78                                   // parameter 'value' is the new sample rate in Hz.
     79                                   // Only creates a sample rate converter the first time that
     80                                   // the track sample rate is different from the mix sample rate.
     81                                   // If the new sample rate is the same as the mix sample rate,
     82                                   // and a sample rate converter already exists,
     83                                   // then the sample rate converter remains present but is a no-op.
     84         RESET           = 0x4101, // Reset sample rate converter without changing sample rate.
     85                                   // This clears out the resampler's input buffer.
     86         REMOVE          = 0x4102, // Remove the sample rate converter on this track name;
     87                                   // the track is restored to the mix sample rate.
     88         // for target RAMP_VOLUME and VOLUME (8 channels max)
     89         // FIXME use float for these 3 to improve the dynamic range
     90         VOLUME0         = 0x4200,
     91         VOLUME1         = 0x4201,
     92         AUXLEVEL        = 0x4210,
     93         // for target TIMESTRETCH
     94         PLAYBACK_RATE   = 0x4300, // Configure timestretch on this track name;
     95                                   // parameter 'value' is a pointer to the new playback rate.
     96     };
     97 
     98     AudioMixer(size_t frameCount, uint32_t sampleRate)
     99         : mSampleRate(sampleRate)
    100         , mFrameCount(frameCount) {
    101         pthread_once(&sOnceControl, &sInitRoutine);
    102     }
    103 
    104     // Create a new track in the mixer.
    105     //
    106     // \param name        a unique user-provided integer associated with the track.
    107     //                    If name already exists, the function will abort.
    108     // \param channelMask output channel mask.
    109     // \param format      PCM format
    110     // \param sessionId   Session id for the track. Tracks with the same
    111     //                    session id will be submixed together.
    112     //
    113     // \return OK        on success.
    114     //         BAD_VALUE if the format does not satisfy isValidFormat()
    115     //                   or the channelMask does not satisfy isValidChannelMask().
    116     status_t    create(
    117             int name, audio_channel_mask_t channelMask, audio_format_t format, int sessionId);
    118 
    119     bool        exists(int name) const {
    120         return mTracks.count(name) > 0;
    121     }
    122 
    123     // Free an allocated track by name.
    124     void        destroy(int name);
    125 
    126     // Enable or disable an allocated track by name
    127     void        enable(int name);
    128     void        disable(int name);
    129 
    130     void        setParameter(int name, int target, int param, void *value);
    131 
    132     void        setBufferProvider(int name, AudioBufferProvider* bufferProvider);
    133 
    134     void        process() {
    135         (this->*mHook)();
    136     }
    137 
    138     size_t      getUnreleasedFrames(int name) const;
    139 
    140     std::string trackNames() const {
    141         std::stringstream ss;
    142         for (const auto &pair : mTracks) {
    143             ss << pair.first << " ";
    144         }
    145         return ss.str();
    146     }
    147 
    148     void        setNBLogWriter(NBLog::Writer *logWriter) {
    149         mNBLogWriter = logWriter;
    150     }
    151 
    152     static inline bool isValidFormat(audio_format_t format) {
    153         switch (format) {
    154         case AUDIO_FORMAT_PCM_8_BIT:
    155         case AUDIO_FORMAT_PCM_16_BIT:
    156         case AUDIO_FORMAT_PCM_24_BIT_PACKED:
    157         case AUDIO_FORMAT_PCM_32_BIT:
    158         case AUDIO_FORMAT_PCM_FLOAT:
    159             return true;
    160         default:
    161             return false;
    162         }
    163     }
    164 
    165     static inline bool isValidChannelMask(audio_channel_mask_t channelMask) {
    166         return audio_channel_mask_is_valid(channelMask); // the RemixBufferProvider is flexible.
    167     }
    168 
    169 private:
    170 
    171     /* For multi-format functions (calls template functions
    172      * in AudioMixerOps.h).  The template parameters are as follows:
    173      *
    174      *   MIXTYPE     (see AudioMixerOps.h MIXTYPE_* enumeration)
    175      *   USEFLOATVOL (set to true if float volume is used)
    176      *   ADJUSTVOL   (set to true if volume ramp parameters needs adjustment afterwards)
    177      *   TO: int32_t (Q4.27) or float
    178      *   TI: int32_t (Q4.27) or int16_t (Q0.15) or float
    179      *   TA: int32_t (Q4.27)
    180      */
    181 
    182     enum {
    183         // FIXME this representation permits up to 8 channels
    184         NEEDS_CHANNEL_COUNT__MASK   = 0x00000007,
    185     };
    186 
    187     enum {
    188         NEEDS_CHANNEL_1             = 0x00000000,   // mono
    189         NEEDS_CHANNEL_2             = 0x00000001,   // stereo
    190 
    191         // sample format is not explicitly specified, and is assumed to be AUDIO_FORMAT_PCM_16_BIT
    192 
    193         NEEDS_MUTE                  = 0x00000100,
    194         NEEDS_RESAMPLE              = 0x00001000,
    195         NEEDS_AUX                   = 0x00010000,
    196     };
    197 
    198     // hook types
    199     enum {
    200         PROCESSTYPE_NORESAMPLEONETRACK, // others set elsewhere
    201     };
    202 
    203     enum {
    204         TRACKTYPE_NOP,
    205         TRACKTYPE_RESAMPLE,
    206         TRACKTYPE_NORESAMPLE,
    207         TRACKTYPE_NORESAMPLEMONO,
    208     };
    209 
    210     // process hook functionality
    211     using process_hook_t = void(AudioMixer::*)();
    212 
    213     struct Track;
    214     using hook_t = void(Track::*)(int32_t* output, size_t numOutFrames, int32_t* temp, int32_t* aux);
    215 
    216     struct Track {
    217         Track()
    218             : bufferProvider(nullptr)
    219         {
    220             // TODO: move additional initialization here.
    221         }
    222 
    223         ~Track()
    224         {
    225             // bufferProvider, mInputBufferProvider need not be deleted.
    226             mResampler.reset(nullptr);
    227             // Ensure the order of destruction of buffer providers as they
    228             // release the upstream provider in the destructor.
    229             mTimestretchBufferProvider.reset(nullptr);
    230             mPostDownmixReformatBufferProvider.reset(nullptr);
    231             mDownmixerBufferProvider.reset(nullptr);
    232             mReformatBufferProvider.reset(nullptr);
    233         }
    234 
    235         bool        needsRamp() { return (volumeInc[0] | volumeInc[1] | auxInc) != 0; }
    236         bool        setResampler(uint32_t trackSampleRate, uint32_t devSampleRate);
    237         bool        doesResample() const { return mResampler.get() != nullptr; }
    238         void        resetResampler() { if (mResampler.get() != nullptr) mResampler->reset(); }
    239         void        adjustVolumeRamp(bool aux, bool useFloat = false);
    240         size_t      getUnreleasedFrames() const { return mResampler.get() != nullptr ?
    241                                                     mResampler->getUnreleasedFrames() : 0; };
    242 
    243         status_t    prepareForDownmix();
    244         void        unprepareForDownmix();
    245         status_t    prepareForReformat();
    246         void        unprepareForReformat();
    247         bool        setPlaybackRate(const AudioPlaybackRate &playbackRate);
    248         void        reconfigureBufferProviders();
    249 
    250         static hook_t getTrackHook(int trackType, uint32_t channelCount,
    251                 audio_format_t mixerInFormat, audio_format_t mixerOutFormat);
    252 
    253         void track__nop(int32_t* out, size_t numFrames, int32_t* temp, int32_t* aux);
    254 
    255         template <int MIXTYPE, bool USEFLOATVOL, bool ADJUSTVOL,
    256             typename TO, typename TI, typename TA>
    257         void volumeMix(TO *out, size_t outFrames, const TI *in, TA *aux, bool ramp);
    258 
    259         uint32_t    needs;
    260 
    261         // TODO: Eventually remove legacy integer volume settings
    262         union {
    263         int16_t     volume[MAX_NUM_VOLUMES]; // U4.12 fixed point (top bit should be zero)
    264         int32_t     volumeRL;
    265         };
    266 
    267         int32_t     prevVolume[MAX_NUM_VOLUMES];
    268         int32_t     volumeInc[MAX_NUM_VOLUMES];
    269         int32_t     auxInc;
    270         int32_t     prevAuxLevel;
    271         int16_t     auxLevel;       // 0 <= auxLevel <= MAX_GAIN_INT, but signed for mul performance
    272 
    273         uint16_t    frameCount;
    274 
    275         uint8_t     channelCount;   // 1 or 2, redundant with (needs & NEEDS_CHANNEL_COUNT__MASK)
    276         uint8_t     unused_padding; // formerly format, was always 16
    277         uint16_t    enabled;        // actually bool
    278         audio_channel_mask_t channelMask;
    279 
    280         // actual buffer provider used by the track hooks, see DownmixerBufferProvider below
    281         //  for how the Track buffer provider is wrapped by another one when dowmixing is required
    282         AudioBufferProvider*                bufferProvider;
    283 
    284         mutable AudioBufferProvider::Buffer buffer; // 8 bytes
    285 
    286         hook_t      hook;
    287         const void  *mIn;             // current location in buffer
    288 
    289         std::unique_ptr<AudioResampler> mResampler;
    290         uint32_t            sampleRate;
    291         int32_t*           mainBuffer;
    292         int32_t*           auxBuffer;
    293 
    294         /* Buffer providers are constructed to translate the track input data as needed.
    295          *
    296          * TODO: perhaps make a single PlaybackConverterProvider class to move
    297          * all pre-mixer track buffer conversions outside the AudioMixer class.
    298          *
    299          * 1) mInputBufferProvider: The AudioTrack buffer provider.
    300          * 2) mReformatBufferProvider: If not NULL, performs the audio reformat to
    301          *    match either mMixerInFormat or mDownmixRequiresFormat, if the downmixer
    302          *    requires reformat. For example, it may convert floating point input to
    303          *    PCM_16_bit if that's required by the downmixer.
    304          * 3) mDownmixerBufferProvider: If not NULL, performs the channel remixing to match
    305          *    the number of channels required by the mixer sink.
    306          * 4) mPostDownmixReformatBufferProvider: If not NULL, performs reformatting from
    307          *    the downmixer requirements to the mixer engine input requirements.
    308          * 5) mTimestretchBufferProvider: Adds timestretching for playback rate
    309          */
    310         AudioBufferProvider*     mInputBufferProvider;    // externally provided buffer provider.
    311         std::unique_ptr<PassthruBufferProvider> mReformatBufferProvider;
    312         std::unique_ptr<PassthruBufferProvider> mDownmixerBufferProvider;
    313         std::unique_ptr<PassthruBufferProvider> mPostDownmixReformatBufferProvider;
    314         std::unique_ptr<PassthruBufferProvider> mTimestretchBufferProvider;
    315 
    316         int32_t     sessionId;
    317 
    318         audio_format_t mMixerFormat;     // output mix format: AUDIO_FORMAT_PCM_(FLOAT|16_BIT)
    319         audio_format_t mFormat;          // input track format
    320         audio_format_t mMixerInFormat;   // mix internal format AUDIO_FORMAT_PCM_(FLOAT|16_BIT)
    321                                          // each track must be converted to this format.
    322         audio_format_t mDownmixRequiresFormat;  // required downmixer format
    323                                                 // AUDIO_FORMAT_PCM_16_BIT if 16 bit necessary
    324                                                 // AUDIO_FORMAT_INVALID if no required format
    325 
    326         float          mVolume[MAX_NUM_VOLUMES];     // floating point set volume
    327         float          mPrevVolume[MAX_NUM_VOLUMES]; // floating point previous volume
    328         float          mVolumeInc[MAX_NUM_VOLUMES];  // floating point volume increment
    329 
    330         float          mAuxLevel;                     // floating point set aux level
    331         float          mPrevAuxLevel;                 // floating point prev aux level
    332         float          mAuxInc;                       // floating point aux increment
    333 
    334         audio_channel_mask_t mMixerChannelMask;
    335         uint32_t             mMixerChannelCount;
    336 
    337         AudioPlaybackRate    mPlaybackRate;
    338 
    339     private:
    340         // hooks
    341         void track__genericResample(int32_t* out, size_t numFrames, int32_t* temp, int32_t* aux);
    342         void track__16BitsStereo(int32_t* out, size_t numFrames, int32_t* temp, int32_t* aux);
    343         void track__16BitsMono(int32_t* out, size_t numFrames, int32_t* temp, int32_t* aux);
    344 
    345         void volumeRampStereo(int32_t* out, size_t frameCount, int32_t* temp, int32_t* aux);
    346         void volumeStereo(int32_t* out, size_t frameCount, int32_t* temp, int32_t* aux);
    347 
    348         // multi-format track hooks
    349         template <int MIXTYPE, typename TO, typename TI, typename TA>
    350         void track__Resample(TO* out, size_t frameCount, TO* temp __unused, TA* aux);
    351         template <int MIXTYPE, typename TO, typename TI, typename TA>
    352         void track__NoResample(TO* out, size_t frameCount, TO* temp __unused, TA* aux);
    353     };
    354 
    355     // TODO: remove BLOCKSIZE unit of processing - it isn't needed anymore.
    356     static constexpr int BLOCKSIZE = 16;
    357 
    358     bool setChannelMasks(int name,
    359             audio_channel_mask_t trackChannelMask, audio_channel_mask_t mixerChannelMask);
    360 
    361     // Called when track info changes and a new process hook should be determined.
    362     void invalidate() {
    363         mHook = &AudioMixer::process__validate;
    364     }
    365 
    366     void process__validate();
    367     void process__nop();
    368     void process__genericNoResampling();
    369     void process__genericResampling();
    370     void process__oneTrack16BitsStereoNoResampling();
    371 
    372     template <int MIXTYPE, typename TO, typename TI, typename TA>
    373     void process__noResampleOneTrack();
    374 
    375     static process_hook_t getProcessHook(int processType, uint32_t channelCount,
    376             audio_format_t mixerInFormat, audio_format_t mixerOutFormat);
    377 
    378     static void convertMixerFormat(void *out, audio_format_t mixerOutFormat,
    379             void *in, audio_format_t mixerInFormat, size_t sampleCount);
    380 
    381     static void sInitRoutine();
    382 
    383     // initialization constants
    384     const uint32_t mSampleRate;
    385     const size_t mFrameCount;
    386 
    387     NBLog::Writer *mNBLogWriter = nullptr;   // associated NBLog::Writer
    388 
    389     process_hook_t mHook = &AudioMixer::process__nop;   // one of process__*, never nullptr
    390 
    391     // the size of the type (int32_t) should be the largest of all types supported
    392     // by the mixer.
    393     std::unique_ptr<int32_t[]> mOutputTemp;
    394     std::unique_ptr<int32_t[]> mResampleTemp;
    395 
    396     // track names grouped by main buffer, in no particular order of main buffer.
    397     // however names for a particular main buffer are in order (by construction).
    398     std::unordered_map<void * /* mainBuffer */, std::vector<int /* name */>> mGroups;
    399 
    400     // track names that are enabled, in increasing order (by construction).
    401     std::vector<int /* name */> mEnabled;
    402 
    403     // track smart pointers, by name, in increasing order of name.
    404     std::map<int /* name */, std::shared_ptr<Track>> mTracks;
    405 
    406     static pthread_once_t sOnceControl; // initialized in constructor by first new
    407 };
    408 
    409 // ----------------------------------------------------------------------------
    410 } // namespace android
    411 
    412 #endif // ANDROID_AUDIO_MIXER_H
    413