Thread: SoundManagerAlsa.cpp and multi-threaded applications

I am trying to understand a few things about this SoundManagerAlsa script and multi-threading.

Here is how the SoundManagerAlsa script is being called:

Code:

{
  SoundManager::SoundManager soundManager(portNumber);
  soundManager.openSoundDevice();
  soundManager.startServer();

  cmdQueue.get(); // blocks until a command is available in the queue

  soundManager.shutDown();
  }

So from the top here's openSoundDevice():

Code:

void SoundManager::openSoundDevice()
  {
    initiateSoundDevice(); // Close then reopen the device.
  }

now initiateSoundDevice():

Code:

void SoundManager::initiateSoundDevice()
  {
    using namespace TimeStuff;

    if (!m_toneDev || m_audioSetup.threadIsRunning) // If the thread is already running then shut it down and reopen the audio device.
    {
      printf("%10.3f Shut down thread.\n", fToHMS(now()));
      shutDown();

      assert(!m_audioSetup.threadIsRunning);
      if (m_audioSetup.threadIsRunning)
      {
        perror("Unable to shut down audio update thread");
      }
      else
      {
        printf("%10.3f Open sound device.\n", fToHMS(now()));

        for (int i=0 ; audioDeviceName[i] && !m_toneDev ; i++)
        {
          int result = snd_pcm_open(&m_toneDev, audioDeviceName[i], SND_PCM_STREAM_PLAYBACK, 0);
          if (result < 0)
          {
            std::cerr << "Unable to open sound device(" << audioDeviceName[i] << "), tone canceled. Errmsg: " << snd_strerror(result) << std::endl;
            m_toneDev = NULL;
          }
          else
          {
            printf("%10.3f Sound device (%s) open. Handle: %p\n", fToHMS(now()), audioDeviceName[i], m_toneDev);
          }
        }
      }
    }

    if (m_toneDev && !m_audioSetup.threadIsRunning) // Audio device is currently open.
    {
      m_audioSetup.clear();
      /**
       * We only want 2 periods in the ring buffer.
       * This is enough to allow uninterrupted playback, and it allows us to interrupt ongoing tones reasonably fast.
       *
       * We can have two periods in the buffer, and a third waiting to get access to the buffer.
       * This means that if we want to interrupt an ongoing tone, the reaction time will be between two and three periods.
       *
       * With
       *   sampleRate = 44100 samples per second
       *      2 periods
       *   1024 samples pr period
       * we get:
       * timePerPeriod = nSamplesPerPeriod / sampleRate = 1024 / 44100 = 23.2ms
       *
       * This means that this thread will run once per 23,2ms when a tone is being played.
       * And if the tone must be interrupted the response time is in the range [2*23.2 - 3*23.2] = [46.4 - 69.6]ms.
       */
      snd_pcm_hw_params_t *hwparams;
      snd_pcm_sw_params_t *swparams;
      snd_pcm_hw_params_alloca(&hwparams); // Allocation (alloca) is deleted when function exits.
      snd_pcm_sw_params_alloca(&swparams);

      // Get all parameters:
      if (m_toneDev) handleSetupError(snd_pcm_hw_params_any(                 m_toneDev, hwparams                                ), "Unable to read hardware parameters for sound device.");
      if (m_toneDev) handleSetupError(snd_pcm_hw_params_set_rate_resample(   m_toneDev, hwparams, 1                             ), "Unable to set re-sampling for sound device.");
      if (m_toneDev) handleSetupError(snd_pcm_hw_params_set_access(          m_toneDev, hwparams, SND_PCM_ACCESS_RW_INTERLEAVED ), "Unable to set interleaved format for sound device.");
      if (m_toneDev) handleSetupError(snd_pcm_hw_params_set_format(          m_toneDev, hwparams, SND_PCM_FORMAT_S16_LE         ), "Unable to set sample format for sound device.");
      if (m_toneDev) handleSetupError(snd_pcm_hw_params_set_channels(        m_toneDev, hwparams, m_audioSetup.nChannels        ), "Unable to set number of channels for sound device.");
      if (m_toneDev) handleSetupError(snd_pcm_hw_params_set_rate_near(       m_toneDev, hwparams, &m_audioSetup.sampleRate, NULL), "Unable to set sample rate for sound device.");
      if (m_toneDev) handleSetupError(snd_pcm_hw_params_set_buffer_size_near(m_toneDev, hwparams, &m_audioSetup.bufferSize      ), "Unable to set buffer size for sound device.");
      if (m_toneDev) handleSetupError(snd_pcm_hw_params_set_period_size_near(m_toneDev, hwparams, &m_audioSetup.periodSize, NULL), "Unable to set period size for sound device.");
      if (m_toneDev) handleSetupError(snd_pcm_hw_params(                     m_toneDev, hwparams                                ), "Unable to write hw parameters to sound device.");

      if (m_toneDev)
      {
        m_audioSetup.msPerSample = 1000.0 / static_cast<double>(m_audioSetup.sampleRate);
        printf("%10.3f Sound device Hardware setup:\n", fToHMS(now()));
        printf("          Millisec per sample: %fms\n", m_audioSetup.msPerSample);
        printf("          Millisec per period: %fms\n", m_audioSetup.msPerSample * static_cast<double>(m_audioSetup.periodSize));
        printf("             Samples per period: %u\n", static_cast<unsigned>(m_audioSetup.periodSize));
        printf("             Samples per buffer: %u\n", static_cast<unsigned>(m_audioSetup.bufferSize));

        m_audioSetup.audioBuffer = new int16_t[m_audioSetup.periodSize * m_audioSetup.nChannels];
        assert(m_audioSetup.audioBuffer);

        if (!m_audioSetup.audioBuffer)
        {
          std::cerr << "Unable to allocate audio buffer. Tone cancelled." << std::endl;
          snd_pcm_close(m_toneDev);
          m_toneDev = NULL;
        }
      }

      if (m_toneDev)
      {
        m_audioSetup.threadIsRunning = true;
        m_toneThread = threadfunc::ThreadCreate( 6, threadUpdateTone, this );
      }
    }
  }

The portion that confuses me is:

Code:

if (!m_toneDev || m_audioSetup.threadIsRunning) // If the thread is already running then shut it down and reopen the audio device.
    {
      printf("%10.3f Shut down thread.\n", fToHMS(now()));
      shutDown();

      assert(!m_audioSetup.threadIsRunning);
      if (m_audioSetup.threadIsRunning)
      {
        perror("Unable to shut down audio update thread");
      }
      else
      {
        printf("%10.3f Open sound device.\n", fToHMS(now()));

        for (int i=0 ; audioDeviceName[i] && !m_toneDev ; i++)
        {
          int result = snd_pcm_open(&m_toneDev, audioDeviceName[i], SND_PCM_STREAM_PLAYBACK, 0);
          if (result < 0)
          {
            std::cerr << "Unable to open sound device(" << audioDeviceName[i] << "), tone canceled. Errmsg: " << snd_strerror(result) << std::endl;
            m_toneDev = NULL;
          }
          else
          {
            printf("%10.3f Sound device (%s) open. Handle: %p\n", fToHMS(now()), audioDeviceName[i], m_toneDev);
          }
        }
      }
    }

What is going on here? Why would you have a multi-threaded audio device? You only have one speaker, you can't play two tones at once, so whats the point? I mean it even stops the currently running thread from running whenever initiateSoundDevice() is called doesn't it? Why would this be a threaded application?

I think the hint lies in the comment: "This means that this thread will run once per 23,2ms when a tone is being played." So, it looks like the thread exists so that it can be perdiocally put to sleep then run.

Originally Posted by laserlight

I think the hint lies in the comment: "This means that this thread will run once per 23,2ms when a tone is being played." So, it looks like the thread exists so that it can be perdiocally put to sleep then run.

I always miss the most obvious things... thank you. Okay so the thread is there just for debugging purposes you are thinking?

I am thinking I can stop the multithreading just by saying something like:

Code:

void SoundManager::reopenSoundDevice()
  {
    if (m_audioSetup.threadIsRunning){}
    else{
       initiateSoundDevice();
    }
  }

Does that sound right? I am new to threading... maybe I should just say: m_toneMutex.wait()