WASAPI VoiceEngine implementation

lib/audiodev/AudioVoiceEngine.cpp

@@ -28,12 +28,6 @@ void BaseAudioVoiceEngine::_pumpAndMixVoices(size_t frames, float* dataOut)
             vox->pumpAndMix(m_mixInfo, frames, dataOut);
 }
 
-BaseAudioVoiceEngine::BaseAudioVoiceEngine
-    (const std::function<AudioVoiceEngineMixInfo()>& getEngineMixInfo)
-{
-    m_mixInfo = getEngineMixInfo();
-}
-
 std::unique_ptr<IAudioVoice>
 BaseAudioVoiceEngine::allocateNewMonoVoice(double sampleRate,
                                        IAudioVoiceCallback* cb,

lib/audiodev/AudioVoiceEngine.hpp

@@ -31,8 +31,6 @@ protected:
     void _pumpAndMixVoices(size_t frames, float* dataOut);
 
 public:
-    BaseAudioVoiceEngine(const std::function<AudioVoiceEngineMixInfo()>& getEngineMixInfo);
-
     std::unique_ptr<IAudioVoice> allocateNewMonoVoice(double sampleRate,
                                                       IAudioVoiceCallback* cb,
                                                       bool dynamicPitch=false);

lib/audiodev/WASAPI.cpp

@@ -1,5 +1,5 @@
 #include "../win/Win32Common.hpp"
-#include "boo/audiodev/IAudioVoiceAllocator.hpp"
+#include "AudioVoiceEngine.hpp"
 #include "logvisor/logvisor.hpp"
 
 #include <Mmdeviceapi.h>
@@ -14,177 +14,16 @@ namespace boo
 {
 static logvisor::Module Log("boo::WASAPI");
 
-struct WASAPIAudioVoice : IAudioVoice
+struct WASAPIAudioVoiceEngine : BaseAudioVoiceEngine
-{
-    struct WASAPIAudioVoiceAllocator& m_parent;
-    std::list<WASAPIAudioVoice*>::iterator m_parentIt;
-
-    ChannelMap m_map;
-    IAudioVoiceCallback* m_cb;
-    ComPtr<IAudioClient> m_audClient;
-    ComPtr<IAudioRenderClient> m_renderClient;
-    UINT32  m_bufferFrames = 1024;
-    size_t m_frameSize;
-
-    const ChannelMap& channelMap() const {return m_map;}
-
-    WASAPIAudioVoice(WASAPIAudioVoiceAllocator& parent, IMMDevice* dev, AudioChannelSet set,
-                     unsigned sampleRate, IAudioVoiceCallback* cb)
-    : m_parent(parent), m_cb(cb)
-    {
-        unsigned chCount = ChannelCount(set);
-
-        WAVEFORMATEX desc = {};
-        desc.wFormatTag = WAVE_FORMAT_PCM;
-        desc.nChannels = chCount;
-        desc.nSamplesPerSec = sampleRate;
-        desc.wBitsPerSample = 16;
-        desc.nBlockAlign = desc.nChannels * desc.wBitsPerSample / 8;
-        desc.nAvgBytesPerSec = desc.nSamplesPerSec * desc.nBlockAlign;
-
-        if (FAILED(dev->Activate(IID_IAudioClient, CLSCTX_ALL,
-                                 nullptr, &m_audClient)))
-        {
-            Log.report(logvisor::Fatal, "unable to create audio client");
-            return;
-        }
-
-        WAVEFORMATEX* works;
-        m_audClient->IsFormatSupported(AUDCLNT_SHAREMODE_SHARED, &desc, &works);
-
-        HRESULT hr = m_audClient->Initialize(AUDCLNT_SHAREMODE_SHARED, 0,
-                                             1000000, 0, &desc, nullptr);
-
-        if (FAILED(hr))
-        {
-            Log.report(logvisor::Fatal, "unable to initialize audio client");
-            return;
-        }
-
-        if (FAILED(m_audClient->GetBufferSize(&m_bufferFrames)))
-        {
-            Log.report(logvisor::Fatal, "unable to obtain audio buffer size");
-            return;
-        }
-
-        if (FAILED(m_audClient->GetService(IID_IAudioRenderClient, &m_renderClient)))
-        {
-            Log.report(logvisor::Fatal, "unable to create audio render client");
-            return;
-        }
-
-        switch (chCount)
-        {
-        case 2:
-            m_map.m_channelCount = 2;
-            m_map.m_channels[0] = AudioChannel::FrontLeft;
-            m_map.m_channels[1] = AudioChannel::FrontRight;
-            break;
-        case 4:
-            m_map.m_channelCount = 4;
-            m_map.m_channels[0] = AudioChannel::FrontLeft;
-            m_map.m_channels[1] = AudioChannel::FrontRight;
-            m_map.m_channels[2] = AudioChannel::RearLeft;
-            m_map.m_channels[3] = AudioChannel::RearRight;
-            break;
-        case 5:
-            m_map.m_channelCount = 5;
-            m_map.m_channels[0] = AudioChannel::FrontLeft;
-            m_map.m_channels[1] = AudioChannel::FrontRight;
-            m_map.m_channels[2] = AudioChannel::FrontCenter;
-            m_map.m_channels[3] = AudioChannel::RearLeft;
-            m_map.m_channels[4] = AudioChannel::RearRight;
-            break;
-        case 6:
-            m_map.m_channelCount = 6;
-            m_map.m_channels[0] = AudioChannel::FrontLeft;
-            m_map.m_channels[1] = AudioChannel::FrontRight;
-            m_map.m_channels[2] = AudioChannel::FrontCenter;
-            m_map.m_channels[3] = AudioChannel::LFE;
-            m_map.m_channels[4] = AudioChannel::RearLeft;
-            m_map.m_channels[5] = AudioChannel::RearRight;
-            break;
-        case 8:
-            m_map.m_channelCount = 8;
-            m_map.m_channels[0] = AudioChannel::FrontLeft;
-            m_map.m_channels[1] = AudioChannel::FrontRight;
-            m_map.m_channels[2] = AudioChannel::FrontCenter;
-            m_map.m_channels[3] = AudioChannel::LFE;
-            m_map.m_channels[4] = AudioChannel::RearLeft;
-            m_map.m_channels[5] = AudioChannel::RearRight;
-            m_map.m_channels[6] = AudioChannel::SideLeft;
-            m_map.m_channels[7] = AudioChannel::SideRight;
-            break;
-        default:
-            Log.report(logvisor::Error, "unknown channel layout %u; using stereo", chCount);
-            m_map.m_channelCount = 2;
-            m_map.m_channels[0] = AudioChannel::FrontLeft;
-            m_map.m_channels[1] = AudioChannel::FrontRight;
-            break;
-        }
-
-        while (m_map.m_channelCount < chCount)
-            m_map.m_channels[m_map.m_channelCount++] = AudioChannel::Unknown;
-
-        m_frameSize = chCount * 2;
-
-        for (unsigned i=0 ; i<3 ; ++i)
-            m_cb->needsNextBuffer(*this, m_bufferFrames);
-    }
-
-    void bufferSampleData(const int16_t* data, size_t frames)
-    {
-        BYTE* dataOut;
-        if (FAILED(m_renderClient->GetBuffer(frames, &dataOut)))
-        {
-            Log.report(logvisor::Fatal, L"unable to obtain audio buffer");
-            return;
-        }
-
-        memcpy(dataOut, data, frames * m_frameSize);
-
-        if (FAILED(m_renderClient->ReleaseBuffer(frames, 0)))
-        {
-            Log.report(logvisor::Fatal, L"unable to release audio buffer");
-            return;
-        }
-    }
-
-    void start()
-    {
-        m_audClient->Start();
-    }
-
-    void stop()
-    {
-        m_audClient->Stop();
-    }
-
-    void pump()
-    {
-        UINT32 padding;
-        if (FAILED(m_audClient->GetCurrentPadding(&padding)))
-        {
-            Log.report(logvisor::Fatal, L"unable to obtain audio buffer padding");
-            return;
-        }
-        INT32 available = m_bufferFrames - padding;
-        m_cb->needsNextBuffer(*this, available);
-    }
-
-    ~WASAPIAudioVoice();
-};
-
-struct WASAPIAudioVoiceAllocator : IAudioVoiceAllocator
 {
     ComPtr<IMMDevice> m_device;
-    AudioChannelSet m_maxSet = AudioChannelSet::Unknown;
+    ComPtr<IAudioClient> m_audClient;
-    std::list<WASAPIAudioVoice*> m_allocatedVoices;
+    ComPtr<IAudioRenderClient> m_renderClient;
 
-    WASAPIAudioVoiceAllocator()
+    WASAPIAudioVoiceEngine()
     {
+        /* Enumerate default audio device */
         ComPtr<IMMDeviceEnumerator> pEnumerator;
-
         if (FAILED(CoCreateInstance(CLSID_MMDeviceEnumerator, nullptr,
                                     CLSCTX_ALL, IID_IMMDeviceEnumerator,
                                     &pEnumerator)))
@@ -196,79 +35,229 @@ struct WASAPIAudioVoiceAllocator : IAudioVoiceAllocator
         if (FAILED(pEnumerator->GetDefaultAudioEndpoint(eRender, eConsole, &m_device)))
         {
             Log.report(logvisor::Fatal, L"unable to obtain default audio device");
+            m_device.Reset();
             return;
         }
 
-        ComPtr<IAudioClient> pAudioClient;
+        if (FAILED(m_device->Activate(IID_IAudioClient, CLSCTX_ALL, nullptr, &m_audClient)))
-        if (FAILED(m_device->Activate(IID_IAudioClient, CLSCTX_ALL, nullptr, &pAudioClient)))
         {
             Log.report(logvisor::Fatal, L"unable to create audio client from device");
+            m_device.Reset();
             return;
         }
 
         WAVEFORMATEXTENSIBLE* pwfx;
-        if (FAILED(pAudioClient->GetMixFormat((WAVEFORMATEX**)&pwfx)))
+        if (FAILED(m_audClient->GetMixFormat((WAVEFORMATEX**)&pwfx)))
         {
             Log.report(logvisor::Fatal, L"unable to obtain audio mix format from device");
+            m_device.Reset();
             return;
         }
 
+        /* Get channel information */
         if ((pwfx->dwChannelMask & (SPEAKER_FRONT_LEFT|SPEAKER_FRONT_RIGHT)) == (SPEAKER_FRONT_LEFT|SPEAKER_FRONT_RIGHT))
         {
-            m_maxSet = AudioChannelSet::Stereo;
+            m_mixInfo.m_channels = AudioChannelSet::Stereo;
             if ((pwfx->dwChannelMask & (SPEAKER_BACK_LEFT|SPEAKER_BACK_RIGHT)) == (SPEAKER_BACK_LEFT|SPEAKER_BACK_RIGHT))
             {
-                m_maxSet = AudioChannelSet::Quad;
+                m_mixInfo.m_channels = AudioChannelSet::Quad;
                 if ((pwfx->dwChannelMask & (SPEAKER_FRONT_CENTER|SPEAKER_LOW_FREQUENCY)) == (SPEAKER_FRONT_CENTER|SPEAKER_LOW_FREQUENCY))
                 {
-                    m_maxSet = AudioChannelSet::Surround51;
+                    m_mixInfo.m_channels = AudioChannelSet::Surround51;
                     if ((pwfx->dwChannelMask & (SPEAKER_SIDE_LEFT|SPEAKER_SIDE_RIGHT)) == (SPEAKER_SIDE_LEFT|SPEAKER_SIDE_RIGHT))
                     {
-                        m_maxSet = AudioChannelSet::Surround71;
+                        m_mixInfo.m_channels = AudioChannelSet::Surround71;
                     }
                 }
             }
         }
 
+        ChannelMap& chMapOut = m_mixInfo.m_channelMap;
+        switch (pwfx->Format.nChannels)
+        {
+        case 2:
+            chMapOut.m_channelCount = 2;
+            chMapOut.m_channels[0] = AudioChannel::FrontLeft;
+            chMapOut.m_channels[1] = AudioChannel::FrontRight;
+            break;
+        case 4:
+            chMapOut.m_channelCount = 4;
+            chMapOut.m_channels[0] = AudioChannel::FrontLeft;
+            chMapOut.m_channels[1] = AudioChannel::FrontRight;
+            chMapOut.m_channels[2] = AudioChannel::RearLeft;
+            chMapOut.m_channels[3] = AudioChannel::RearRight;
+            break;
+        case 5:
+            chMapOut.m_channelCount = 5;
+            chMapOut.m_channels[0] = AudioChannel::FrontLeft;
+            chMapOut.m_channels[1] = AudioChannel::FrontRight;
+            chMapOut.m_channels[2] = AudioChannel::FrontCenter;
+            chMapOut.m_channels[3] = AudioChannel::RearLeft;
+            chMapOut.m_channels[4] = AudioChannel::RearRight;
+            break;
+        case 6:
+            chMapOut.m_channelCount = 6;
+            chMapOut.m_channels[0] = AudioChannel::FrontLeft;
+            chMapOut.m_channels[1] = AudioChannel::FrontRight;
+            chMapOut.m_channels[2] = AudioChannel::FrontCenter;
+            chMapOut.m_channels[3] = AudioChannel::LFE;
+            chMapOut.m_channels[4] = AudioChannel::RearLeft;
+            chMapOut.m_channels[5] = AudioChannel::RearRight;
+            break;
+        case 8:
+            chMapOut.m_channelCount = 8;
+            chMapOut.m_channels[0] = AudioChannel::FrontLeft;
+            chMapOut.m_channels[1] = AudioChannel::FrontRight;
+            chMapOut.m_channels[2] = AudioChannel::FrontCenter;
+            chMapOut.m_channels[3] = AudioChannel::LFE;
+            chMapOut.m_channels[4] = AudioChannel::RearLeft;
+            chMapOut.m_channels[5] = AudioChannel::RearRight;
+            chMapOut.m_channels[6] = AudioChannel::SideLeft;
+            chMapOut.m_channels[7] = AudioChannel::SideRight;
+            break;
+        default:
+            Log.report(logvisor::Warning, "unknown channel layout %u; using stereo", pwfx->Format.nChannels);
+            chMapOut.m_channelCount = 2;
+            chMapOut.m_channels[0] = AudioChannel::FrontLeft;
+            chMapOut.m_channels[1] = AudioChannel::FrontRight;
+            break;
+        }
+
+        /* Initialize audio client */
+        if (FAILED(m_audClient->Initialize(
+                   AUDCLNT_SHAREMODE_SHARED,
+                   0,
+                   1000000,
+                   0,
+                   (WAVEFORMATEX*)pwfx,
+                   nullptr)))
+        {
+            Log.report(logvisor::Fatal, L"unable to initialize audio client");
+            m_device.Reset();
+            CoTaskMemFree(pwfx);
+            return;
+        }
+        m_mixInfo.m_sampleRate = pwfx->Format.nSamplesPerSec;
+
+        if (pwfx->Format.wFormatTag == WAVE_FORMAT_PCM ||
+            (pwfx->Format.wFormatTag == WAVE_FORMAT_EXTENSIBLE && pwfx->SubFormat == KSDATAFORMAT_SUBTYPE_PCM))
+        {
+            if (pwfx->Format.wBitsPerSample == 16)
+            {
+                m_mixInfo.m_sampleFormat = SOXR_INT16_I;
+                m_mixInfo.m_bitsPerSample = 16;
+            }
+            else if (pwfx->Format.wBitsPerSample == 32)
+            {
+                m_mixInfo.m_sampleFormat = SOXR_INT32_I;
+                m_mixInfo.m_bitsPerSample = 32;
+            }
+            else
+            {
+                Log.report(logvisor::Fatal, L"unsupported bits-per-sample %d", pwfx->Format.wBitsPerSample);
+                m_device.Reset();
+                return;
+            }
+        }
+        else if (pwfx->Format.wFormatTag == WAVE_FORMAT_IEEE_FLOAT ||
+                 (pwfx->Format.wFormatTag == WAVE_FORMAT_EXTENSIBLE && pwfx->SubFormat == KSDATAFORMAT_SUBTYPE_IEEE_FLOAT))
+        {
+            if (pwfx->Format.wBitsPerSample == 32)
+            {
+                m_mixInfo.m_sampleFormat = SOXR_FLOAT32_I;
+                m_mixInfo.m_bitsPerSample = 32;
+            }
+            else
+            {
+                Log.report(logvisor::Fatal, L"unsupported floating-point bits-per-sample %d", pwfx->Format.wBitsPerSample);
+                m_device.Reset();
+                return;
+            }
+        }
+
         CoTaskMemFree(pwfx);
+
+        UINT32 bufferFrameCount;
+        if (FAILED(m_audClient->GetBufferSize(&bufferFrameCount)))
+        {
+            Log.report(logvisor::Fatal, L"unable to get audio buffer frame count");
+            m_device.Reset();
+            return;
+        }
+        m_mixInfo.m_periodFrames = bufferFrameCount;
+
+        if (FAILED(m_audClient->GetService(IID_IAudioRenderClient, &m_renderClient)))
+        {
+            Log.report(logvisor::Fatal, L"unable to create audio render client");
+            m_device.Reset();
+            return;
+        }
     }
 
-    ~WASAPIAudioVoiceAllocator()
+    bool m_started = false;
+
+    void pumpAndMixVoices()
     {
+        UINT32 numFramesPadding;
+        if (FAILED(m_audClient->GetCurrentPadding(&numFramesPadding)))
+        {
+            Log.report(logvisor::Fatal, L"unable to get available buffer frames");
+            return;
         }
 
-    AudioChannelSet getAvailableSet()
+        size_t frames = m_mixInfo.m_periodFrames - numFramesPadding;
+        if (frames <= 0)
+            return;
+
+        BYTE* bufOut;
+        if (FAILED(m_renderClient->GetBuffer(frames, &bufOut)))
         {
-        return m_maxSet;
+            Log.report(logvisor::Fatal, L"unable to map audio buffer");
+            return;
         }
 
-    std::unique_ptr<IAudioVoice> allocateNewVoice(AudioChannelSet layoutOut,
+        DWORD flags = 0;
-                                                  unsigned sampleRate,
+        switch (m_mixInfo.m_sampleFormat)
-                                                  IAudioVoiceCallback* cb)
         {
-        WASAPIAudioVoice* newVoice = new WASAPIAudioVoice(*this, m_device.Get(), layoutOut, sampleRate, cb);
+        case SOXR_INT16_I:
-        newVoice->m_parentIt = m_allocatedVoices.insert(m_allocatedVoices.end(), newVoice);
+            _pumpAndMixVoices(frames, reinterpret_cast<int16_t*>(bufOut));
-        std::unique_ptr<IAudioVoice> ret(newVoice);
+            break;
-        if (!newVoice->m_audClient)
+        case SOXR_INT32_I:
+            _pumpAndMixVoices(frames, reinterpret_cast<int32_t*>(bufOut));
+            break;
+        case SOXR_FLOAT32_I:
+            _pumpAndMixVoices(frames, reinterpret_cast<float*>(bufOut));
+            break;
+        default:
+            flags = AUDCLNT_BUFFERFLAGS_SILENT;
+            break;
+        }
+
+        if (FAILED(m_renderClient->ReleaseBuffer(frames, flags)))
+        {
+            Log.report(logvisor::Fatal, L"unable to unmap audio buffer");
+            return;
+        }
+
+        if (!m_started)
+        {
+            if (FAILED(m_audClient->Start()))
+            {
+                Log.report(logvisor::Fatal, L"unable to start audio client");
+                m_device.Reset();
+                return;
+            }
+            m_started = true;
+        }
+    }
+};
+
+std::unique_ptr<IAudioVoiceEngine> NewAudioVoiceEngine()
+{
+    std::unique_ptr<IAudioVoiceEngine> ret = std::make_unique<WASAPIAudioVoiceEngine>();
+    if (!static_cast<WASAPIAudioVoiceEngine&>(*ret).m_device)
         return {};
     return ret;
 }
 
-    void pumpVoices()
-    {
-        for (WASAPIAudioVoice* vox : m_allocatedVoices)
-            vox->pump();
-    }
-};
-
-WASAPIAudioVoice::~WASAPIAudioVoice()
-{
-    m_parent.m_allocatedVoices.erase(m_parentIt);
-}
-
-std::unique_ptr<IAudioVoiceAllocator> NewAudioVoiceAllocator()
-{
-    return std::make_unique<WASAPIAudioVoiceAllocator>();
-}
-
 }