Bug 1262753: P6. Perform downmixing in DecodeAudioDataSink. r=kinetik

Rather than performing the audio processing a few frames at a time, we perform the operation when popping the audio data block.
The only downmixing task left to AudioStream is to handle the force mono preference.

Resampling is currently disabled.

MozReview-Commit-ID: GNQA1h0xcp7

--HG--
extra : rebase_source : c58aa4514a0dc4de06be6dc1da6a598742caa262

dom/media/AudioStream.cpp

@@ -331,7 +331,7 @@ AudioStream::Init(uint32_t aNumChannels, uint32_t aRate,
     ("%s  channels: %d, rate: %d for %p", __FUNCTION__, aNumChannels, aRate, this));
   mInRate = mOutRate = aRate;
   mChannels = aNumChannels;
-  mOutChannels = (aNumChannels > 2) ? 2 : aNumChannels;
+  mOutChannels = mIsMonoAudioEnabled ? 1 : aNumChannels;
 
   mDumpFile = OpenDumpFile(this);
 
@@ -353,9 +353,11 @@ AudioStream::Init(uint32_t aNumChannels, uint32_t aRate,
   params.format = ToCubebFormat<AUDIO_OUTPUT_FORMAT>::value;
   mAudioClock.Init();
 
-  AudioConfig inConfig(mChannels, mInRate);
-  AudioConfig outConfig(mOutChannels, mOutRate);
-  mAudioConverter = MakeUnique<AudioConverter>(inConfig, outConfig);
+  if (mIsMonoAudioEnabled) {
+    AudioConfig inConfig(mChannels, mInRate);
+    AudioConfig outConfig(mOutChannels, mOutRate);
+    mAudioConverter = MakeUnique<AudioConverter>(inConfig, outConfig);
+  }
   return OpenCubeb(params);
 }
 
@@ -561,15 +563,10 @@ AudioStream::Downmix(Chunk* aChunk)
     return false;
   }
 
-  if (aChunk->Channels() > 2) {
-    MOZ_ASSERT(mAudioConverter);
+  if (mAudioConverter) {
     mAudioConverter->Process(aChunk->GetWritable(), aChunk->Frames());
   }
 
-  if (aChunk->Channels() >= 2 && mIsMonoAudioEnabled) {
-    DownmixStereoToMono(aChunk->GetWritable(), aChunk->Frames());
-  }
-
   return true;
 }
 

dom/media/mediasink/DecodedAudioDataSink.cpp

@@ -8,6 +8,7 @@
 #include "MediaQueue.h"
 #include "DecodedAudioDataSink.h"
 #include "VideoUtils.h"
+#include "AudioConverter.h"
 
 #include "mozilla/CheckedInt.h"
 #include "mozilla/DebugOnly.h"
@@ -39,6 +40,9 @@ DecodedAudioDataSink::DecodedAudioDataSink(MediaQueue<MediaData>& aAudioQueue,
   , mChannel(aChannel)
   , mPlaying(true)
   , mPlaybackComplete(false)
+  , mConverter(MakeUnique<AudioConverter>(
+                AudioConfig(mInfo.mChannels, mInfo.mRate),
+                AudioConfig(mInfo.mChannels > 2 ? 2 : mInfo.mChannels, mInfo.mRate)))
 {
 }
 
@@ -136,7 +140,9 @@ nsresult
 DecodedAudioDataSink::InitializeAudioStream(const PlaybackParams& aParams)
 {
   mAudioStream = new AudioStream(*this);
-  nsresult rv = mAudioStream->Init(mInfo.mChannels, mInfo.mRate, mChannel);
+  nsresult rv = mAudioStream->Init(mConverter->OutputConfig().Channels(),
+                                   mConverter->OutputConfig().Rate(),
+                                   mChannel);
   if (NS_FAILED(rv)) {
     mAudioStream->Shutdown();
     mAudioStream = nullptr;
@@ -156,7 +162,8 @@ DecodedAudioDataSink::InitializeAudioStream(const PlaybackParams& aParams)
 int64_t
 DecodedAudioDataSink::GetEndTime() const
 {
-  CheckedInt64 playedUsecs = FramesToUsecs(mWritten, mInfo.mRate) + mStartTime;
+  CheckedInt64 playedUsecs =
+    FramesToUsecs(mWritten, mConverter->OutputConfig().Rate()) + mStartTime;
   if (!playedUsecs.isValid()) {
     NS_WARNING("Int overflow calculating audio end time");
     return -1;
@@ -231,9 +238,11 @@ DecodedAudioDataSink::PopFrames(uint32_t aFrames)
     // audio hardware, so we can play across the gap.
     // Calculate the timestamp of the next chunk of audio in numbers of
     // samples.
-    CheckedInt64 sampleTime = UsecsToFrames(AudioQueue().PeekFront()->mTime, mInfo.mRate);
+    CheckedInt64 sampleTime = UsecsToFrames(AudioQueue().PeekFront()->mTime,
+                                            mConverter->OutputConfig().Rate());
     // Calculate the number of frames that have been pushed onto the audio hardware.
-    CheckedInt64 playedFrames = UsecsToFrames(mStartTime, mInfo.mRate) +
+    CheckedInt64 playedFrames = UsecsToFrames(mStartTime,
+                                              mConverter->OutputConfig().Rate()) +
                                 static_cast<int64_t>(mWritten);
     CheckedInt64 missingFrames = sampleTime - playedFrames;
 
@@ -243,6 +252,9 @@ DecodedAudioDataSink::PopFrames(uint32_t aFrames)
       return MakeUnique<Chunk>();
     }
 
+    const uint32_t rate = mConverter->OutputConfig().Rate();
+    const uint32_t channels = mConverter->OutputConfig().Channels();
+
     if (missingFrames.value() > AUDIO_FUZZ_FRAMES) {
       // The next audio chunk begins some time after the end of the last chunk
       // we pushed to the audio hardware. We must push silence into the audio
@@ -251,10 +263,26 @@ DecodedAudioDataSink::PopFrames(uint32_t aFrames)
       missingFrames = std::min<int64_t>(UINT32_MAX, missingFrames.value());
       auto framesToPop = std::min<uint32_t>(missingFrames.value(), aFrames);
       mWritten += framesToPop;
-      return MakeUnique<SilentChunk>(framesToPop, mInfo.mChannels, mInfo.mRate);
+      return MakeUnique<SilentChunk>(framesToPop, channels, rate);
+    }
+
+    RefPtr<AudioData> data =
+      dont_AddRef(AudioQueue().PopFront().take()->As<AudioData>());
+    if (mConverter->InputConfig() != mConverter->OutputConfig()) {
+      AlignedAudioBuffer convertedData =
+        mConverter->Process(AudioSampleBuffer(Move(data->mAudioData))).Forget();
+      mCurrentData =
+        new AudioData(data->mOffset,
+                      data->mTime,
+                      data->mDuration,
+                      convertedData.Length() / channels,
+                      Move(convertedData),
+                      channels,
+                      rate);
+    } else {
+      mCurrentData = Move(data);
     }
 
-    mCurrentData = dont_AddRef(AudioQueue().PopFront().take()->As<AudioData>());
     mCursor = MakeUnique<AudioBufferCursor>(mCurrentData->mAudioData.get(),
                                             mCurrentData->mChannels,
                                             mCurrentData->mFrames);

dom/media/mediasink/DecodedAudioDataSink.h

@@ -21,6 +21,8 @@
 
 namespace mozilla {
 
+class AudioConverter;
+
 namespace media {
 
 class DecodedAudioDataSink : public AudioSink,
@@ -97,13 +99,15 @@ private:
    */
   // The AudioData at which AudioStream::DataSource is reading.
   RefPtr<AudioData> mCurrentData;
-  // Keep track of the read positoin of mCurrentData.
+  // Keep track of the read position of mCurrentData.
   UniquePtr<AudioBufferCursor> mCursor;
   // True if there is any error in processing audio data like overflow.
   bool mErrored = false;
 
   // Set on the callback thread of cubeb once the stream has drained.
   Atomic<bool> mPlaybackComplete;
+
+  UniquePtr<AudioConverter> mConverter;
 };
 
 } // namespace media