Tekkotsu Homepage
Demos
Overview
Downloads
Dev. Resources
Reference
Credits

SoundManager.h

Go to the documentation of this file.
00001 //-*-c++-*-
00002 #ifndef INCLUDED_SoundManager_h_
00003 #define INCLUDED_SoundManager_h_
00004 
00005 #include "IPC/RCRegion.h"
00006 #include "Shared/ODataFormats.h"
00007 
00008 #include <string>
00009 #include <vector>
00010 #include "IPC/ListMemBuf.h"
00011 #include "IPC/MutexLock.h"
00012 #include "SoundManagerMsg.h"
00013 #include "IPC/ProcessID.h"
00014 #include "Shared/attributes.h"
00015 
00016 #ifdef PLATFORM_APERIOS
00017 class OSubject;
00018 class ONotifyEvent;
00019 #else
00020 #  include "IPC/MessageQueue.h"
00021 #endif
00022 
00023 //! Provides sound effects and caching services, as well as mixing buffers for the SoundPlay process
00024 /*! Provides easy methods for playing back sounds, either from files
00025  *  on the memory stick, or from dynamically generated buffers.  You
00026  *  can chain playback commands so that when one sound finishes,
00027  *  another picks up automatically.  This might be handy if, say,
00028  *  someone wants to write an MP3 player ;) The sounds would be too
00029  *  large to load into memory all at once, but you could load a block
00030  *  at a time and chain them so it seamlessly moves from one to the
00031  *  other.
00032  *  
00033  *  You can also preload sounds (loadFile()) before playing them (play() / playFile()) so
00034  *  there's no delay between requesting a sound and having it start playing while it is loaded from disk/memory stick.
00035  *  Just be sure to release the file (releaseFile()) again
00036  *  when you're done with it ;)
00037  *
00038  *  All functions will attempt to lock the SoundManager.  Remember,
00039  *  this is running in a shared memory region, accessible by the
00040  *  SoundPlay process and both the Main and Motion processes (so
00041  *  MotionCommands can play sounds!)
00042  *
00043  *  One could be tempted to draw parallels to the MotionManager, and
00044  *  envision a system with SoundCommands that are handed over and can
00045  *  dynamically compute sound buffers as needed.  If you have the time
00046  *  and inclination, the job's all yours... (Midi players, speech
00047  *  synthesizer, ...?)
00048  *
00049  *  @todo Volume control, variable playback speed, support more wav
00050  *  file formats (latter two are the same thing if you think about it - need
00051  *  to be able to resample on the fly)
00052  *  
00053  *  @todo Add functions to hand out regions to be filled out to avoid
00054  *  copying into the buffer.
00055  *
00056  *  @see <a href="http://www.cs.cmu.edu/~dst/Tekkotsu/Tutorial/sound.shtml">David Touretzky's "Playing Sounds" Chapter</a>
00057  */
00058 class SoundManager {
00059   friend void speechDoneCleanup(long playid);
00060 public:
00061   //! destructor
00062   virtual ~SoundManager();
00063 
00064   //!constructor, should only be called by the receiving process (SoundPlay)
00065   SoundManager();
00066 #ifdef PLATFORM_APERIOS
00067   //!Each process needs to call this before it can send sounds to the SoundPlay process
00068   void InitAccess(OSubject* subj);
00069 #else //PLATFORM_LOCAL
00070   //!Each process (except SoundPlay) needs to call this before it can send sounds to the SoundPlay process
00071   void InitAccess(MessageQueueBase& sndbufq);
00072 #endif
00073   
00074   //!This is used for referring to sound data so you can start playing it or release it
00075   typedef SoundManagerMsg::Snd_ID Snd_ID;
00076   static const Snd_ID invalid_Snd_ID=(Snd_ID)-1; //!< for reporting errors
00077   static const Snd_ID MAX_SND=50; //!< the number of sounds that can be loaded at any given time
00078   
00079   //!This is for referring to instances of the play command so you can stop, pause, or monitor progress (later versions will send events upon completion)
00080   typedef unsigned short Play_ID;
00081   static const Play_ID invalid_Play_ID=(Play_ID)-1; //!< for reporting errors
00082   static const Play_ID MAX_PLAY=256; //!< the number of sounds that can be enqueued at the same time (see MixMode_t)
00083 
00084   static const unsigned int MAX_NAME_LEN=128;   //!<maximum length of a path
00085 
00086   //!Used to set the mode for mixing multiple sound channels
00087   /*!Feel free to add a higher quality mixer if you're an audiophile - I'm pretty new to sound processing*/
00088   enum MixMode_t {
00089     // there's some prototype code for a bit-shifting 'Faster' quality level, but it hasn't been finished... 'Fast' is the default for now.
00090     Fast,    //!< uses real division to maintain volume level, without increasing intermediary precision, which causes low-order bit error in exchange for less CPU usage
00091     Quality  //!< uses real division to maintain volume level, using an intermediary higher precision buffer for mixing
00092   };
00093 
00094   //! indicates how to handle channel overflow (trying to play more sounds than maximum number of mixing channels). See #queue_mode
00095   enum QueueMode_t {
00096     Enqueue,        //!< newer sounds are played when a channel opens up (when old sound finishes)
00097     Pause,          //!< newer sounds pause oldest sound, which continues when a channel opens
00098     Stop,           //!< newer sounds stop oldest sound
00099     Override,       //!< older sounds have play heads advanced, but don't get mixed until a channel opens
00100   };
00101 
00102   //!loads a wav file (if it matches Config::sound_config settings - can't do resampling yet)
00103   /*!Since the SoundManager does the loading, if the same file is being played more than once, only once copy is stored in memory 
00104    * @param name can be either a full path, or a partial path relative to Config::sound_config::root
00105    * @return ID number for future references (can also use name)
00106    * The sound data will be cached until releaseFile() or release() is called a matching number of times*/
00107   Snd_ID loadFile(std::string const &name);
00108 
00109   //!loads raw samples from a buffer (assumes matches Config::sound_config settings)
00110   /*!The sound data will be cached until release() is called a matching number of times.\n
00111    * This function is useful for dynamic sound sources.  A copy will be made. */
00112   Snd_ID loadBuffer(const char buf[], unsigned int len);
00113   
00114   //!Marks the sound buffer to be released after the last play command completes (or right now if not being played)
00115   void releaseFile(std::string const &name);
00116 
00117   //!Marks the sound buffer to be released after the last play command completes (or right now if not being played)
00118   void release(Snd_ID id);
00119   
00120   //!play a wav file (if it matches Config::sound_config settings - can't do resampling yet)
00121   /*!Will do a call to loadFile() first, and then automatically release the sound again when complete.
00122    * @param name can be either a full path, or a partial path relative to Config::sound_config::root
00123    * @return ID number for future references
00124    * The sound data will not be cached after done playing unless a call to loadFile is made*/
00125   Play_ID playFile(std::string const &name);
00126 
00127   //!loads raw samples from a buffer (assumes buffer matches Config::sound_config settings)
00128   /*!The sound data will be released after done playing*/
00129   Play_ID playBuffer(const char buf[], unsigned int len);
00130   
00131   //!plays a previously loaded buffer or file
00132   Play_ID play(Snd_ID id);
00133   
00134   //!allows automatic queuing of sounds - good for dynamic sound sources!
00135   /*!if you chain more than once to the same base, the new buffers are appended
00136    * to the end of the chain - the new buffer doesn't replace the current chain
00137    * @return @a base - just for convenience of multiple calls*/
00138   Play_ID chainFile(Play_ID base, std::string const &next);
00139 
00140   //!allows automatic queuing of sounds - good for dynamic sound sources!
00141   /*!if you chain more than once to the same base, the new buffers are appended
00142    * to the end of the chain - the new buffer doesn't replace the current chain
00143    * @return @a base - just for convenience of multiple calls*/
00144   Play_ID chainBuffer(Play_ID base, const char buf[], unsigned int len);
00145 
00146   //!allows automatic queuing of sounds - good for dynamic sound sources!
00147   /*!if you chain more than once to the same base, the new buffers are appended
00148    * to the end of the chain - the new buffer doesn't replace the current chain
00149    * @return @a base - just for convenience of multiple calls*/
00150   Play_ID chain(Play_ID base, Snd_ID next);
00151   
00152   //! Speaks its argument using the Mary text-to-speech system to generate a WAV file
00153   /*! On Aperios or unsupported platforms, simply displays the text on stdout */
00154   SoundManager::Play_ID speak(const std::string& text, bool showText=true, const std::string& voice="female");
00155 
00156   //!Lets you stop playback of all sounds
00157   void stopPlay();
00158 
00159   //!Lets you stop playback of a sound
00160   void stopPlay(Play_ID id);
00161   
00162   //!Lets you pause playback
00163   void pausePlay(Play_ID id);
00164   
00165   //!Lets you resume playback
00166   void resumePlay(Play_ID id);
00167   
00168   //!Lets you control the maximum number of channels (currently playing sounds), method for mixing (applies when max_chan>1), and queuing method (for when overflow channels)
00169   void setMode(unsigned int max_channels, MixMode_t mixer_mode, QueueMode_t queuing_mode);
00170 
00171   //!Gives the time until the sound finishes, in milliseconds.  Subtract 32 to get guarranteed valid time for this ID.
00172   /*!You should be passing the beginning of a chain to get proper results...\n
00173    * May be slightly conservative (will report too small a time) because this
00174    * does not account for delay until SoundPlay picks up the message that a
00175    * sound has been added.\n
00176    * However, it is slighly optimistic (will report too large a time) because
00177    * it processes a buffer all at one go, so it could mark the sound as finished
00178    * (and cause the ID to go invalid) up to RobotInfo::SoundBufferTime (32 ms)
00179    * before the sound finishes.  So subtract SoundBufferTime if you want to be
00180    * absolutely sure the ID will still valid. */
00181   unsigned int getRemainTime(Play_ID id) const;
00182   
00183 #ifdef PLATFORM_APERIOS
00184   //!Copies the sound data to the OPENR buffer, ready to be passed to the system, only called by SoundPlay
00185   /*!@return the number of active sounds */
00186   unsigned int CopyTo(OSoundVectorData* data);
00187 
00188   //!updates internal data structures on the SoundPlay side - you shouldn't be calling this
00189   void ReceivedMsg(const ONotifyEvent& event);
00190 #endif
00191 
00192   //!Copies the sound data to the specified memory buffer, ready to be passed to the system
00193   /*!@return the number of active sounds */
00194   unsigned int CopyTo(void * dest, size_t destSize);
00195   
00196   //!updates internal data structures on the SoundPlay side - you shouldn't be calling this
00197   void ProcessMsg(RCRegion * rcr);
00198   
00199   //! returns number of sounds currently playing
00200   unsigned int getNumPlaying() { return chanlist.size(); }
00201   
00202   //! return the next region serial number -- doesn't actually increment it though, repeated calls will return the same value until the value is actually used
00203   virtual unsigned int getNextKey() { return sn+1; }
00204   
00205 protected:
00206   //!Mixes the channel into the buffer
00207   void mixChannel(Play_ID channelId, void* buf, size_t size);
00208   
00209   //!Mixes the channel into the buffer additively
00210   /*!If mode is Quality, then the size of the buffer should be double the normal
00211   * size. */
00212   void mixChannelAdditively(Play_ID channelId, int bitsPerSample, MixMode_t mode, short scalingFactor, void* buf, size_t size);
00213   
00214   //!The intermediate mixer buffer used for Quality mode mixing
00215   int* mixerBuffer;
00216   
00217   //!Size (in bytes) of the intermediate mixer buffer
00218   size_t mixerBufferSize;
00219 
00220   //!Sets up a shared region to hold a sound - rounds to nearest page size
00221   RCRegion* initRegion(unsigned int size);
00222 
00223   //!Looks to see if @a name matches any of the sounds in sndlist (converts to absolute path if not already)
00224   Snd_ID lookupPath(std::string const &name) const;
00225 
00226   //!selects which of the channels are actually to be mixed together, depending on queue_mode
00227   void selectChannels(std::vector<Play_ID>& mix);
00228 
00229   //!update the offsets of sounds which weren't mixed (when needed depending on queue_mode)
00230   void updateChannels(const std::vector<Play_ID>& mixs,size_t used);
00231 
00232   //!called when a buffer end is reached, may reset buffer to next in chain, or just stopPlay()
00233   bool endPlay(Play_ID id);
00234   
00235   //!Holds data about the loaded sounds
00236   struct SoundData {
00237     SoundData();                             //!<constructor
00238     RCRegion * rcr;                          //!<shared region - don't need to share among processes, just collect in SoundPlay
00239     byte* data;                              //!<point to data in region (for convenience, only valid in SoundPlay)
00240     unsigned int len;                        //!<size of the sound
00241     unsigned int ref;                        //!<reference counter
00242     unsigned int sn;                         //!<serial number, allows us to verify that a given message buffer does indeed match this sound, and wasn't delayed in processing
00243 #ifdef __APPLE__
00244     struct MacSpeechState* macSpeech; //!< if non-NULL, stores text string and OS context for speech request, #data and #rcr will be NULL (no PCM samples)
00245 #endif
00246     char name[SoundManager::MAX_NAME_LEN];   //!<stores the path to the file, empty if from a buffer
00247   private:
00248     SoundData(const SoundData&);             //!< don't call
00249     SoundData operator=(const SoundData&);   //!< don't call
00250   };
00251   //!For convenience
00252   typedef ListMemBuf<SoundData,MAX_SND,Snd_ID> sndlist_t;
00253   //!Holds a list of all currently loaded sounds
00254   sndlist_t sndlist;
00255   
00256   //!Holds data about sounds currently being played
00257   struct PlayState {
00258     PlayState();            //!<constructor
00259     Snd_ID snd_id;          //!<index of sound
00260     unsigned int offset;    //!<position in the sound
00261     unsigned int cumulative;//!<total time of playing (over queued sounds)
00262     Play_ID next_id;        //!<lets you queue for continuous sound, or loop
00263   };
00264   //!For convenience
00265   typedef ListMemBuf<PlayState,MAX_PLAY,Play_ID> playlist_t;
00266   //!Holds a list of all sounds currently enqueued
00267   playlist_t playlist;
00268   //!For convenience
00269   typedef ListMemBuf<Play_ID,MAX_PLAY,Play_ID> chanlist_t;
00270   //!Holds a list of all currently playing sounds, ordered newest (front) to oldest(back)
00271   chanlist_t chanlist;
00272   
00273   //!Current mixing mode, set by setMode();
00274   MixMode_t mix_mode;
00275 
00276   //!Current queuing mode, set by setMode();
00277   QueueMode_t queue_mode;
00278 
00279   //!Current maximum number of sounds to mix together
00280   unsigned int max_chan;
00281 
00282 #ifndef __APPLE__
00283   //!Prevents multiple processes from accessing at the same time
00284   mutable MutexLock<ProcessID::NumProcesses> lock;
00285 #else
00286   //!Prevents multiple processes from accessing at the same time
00287   mutable struct Lock : public MutexLock<ProcessID::NumProcesses> {
00288     Lock() : MutexLock<ProcessID::NumProcesses>(), initiatedSpeech(), completedSpeech() {}
00289     std::vector<struct MacSpeechState*> initiatedSpeech; //!< buffered speech contexts which should be started
00290     std::vector<struct MacSpeechState*> completedSpeech; //!< buffered speech contexts which need to be freed
00291     virtual void releaseResource(Data& d);
00292     //! this buffers launching OS resources until after lock is released so we don't deadlock with OS mutex in CopyTo callback
00293     void initiated(struct MacSpeechState* mss);
00294     //! this buffers freeing OS resources until after lock is released so we don't deadlock with OS mutex in CopyTo callback
00295     void completed(struct MacSpeechState* mss);
00296   } lock;
00297 #endif
00298   
00299   //!A serial number, incremented for each sound which is created
00300   /*! This is used to verify that a sound message from a process
00301    *  refers to a current sound.  If you imaging a pathological
00302    *  process, which rapidly creates and releases sounds, it would 
00303    *  run through the sndlist ids, possibly before the sound process
00304    *  can process the incoming buffers.  So this is used to ensure
00305    *  that a given message refers to the current sound, and not one
00306    *  that was already released and then reassigned. */
00307   unsigned int sn;
00308 
00309   //!the size of a SoundManagerMsg, which is prefixed before each region sent/received by SoundManager (rounded up to nearest even word boundary)
00310   static const unsigned int MSG_SIZE=((sizeof(SoundManagerMsg)-1)/8+1)*8;
00311 
00312 #ifdef PLATFORM_APERIOS
00313   //!Storage of each process's subject object, used to internally transmit sound buffers to SoundPlay
00314   OSubject * subjs[ProcessID::NumProcesses];
00315 #else //PLATFORM_LOCAL
00316   //!Storage of each process's attachment of the message queue, used to internally transmit sound buffers to SoundPlay
00317   MessageQueueBase * subjs[ProcessID::NumProcesses];
00318 #endif
00319 
00320 private:
00321   SoundManager(const SoundManager&);           //!< don't call
00322   SoundManager operator=(const SoundManager&); //!< don't call
00323 };
00324 
00325 //! lets you play a sound from anywhere in your code - just a one liner!
00326 extern SoundManager * sndman;
00327 
00328 /*! @file
00329  * @brief Describes SoundManager, which provides sound effects and caching services, as well as mixing buffers for the SoundPlay process
00330  * @author ejt (Creator)
00331  */
00332 
00333 #endif

Tekkotsu v5.1CVS
Generated Mon May 9 04:58:51 2016 by Doxygen 1.6.3