Restores larger output buffer size when output stream requiring smaller size is closed

media/audio/mac/audio_manager_mac.h

Index: media/audio/mac/audio_manager_mac.h
diff --git a/media/audio/mac/audio_manager_mac.h b/media/audio/mac/audio_manager_mac.h
index f5159e956cc44a249af38a698cf4e3399969f6b5..a7a74a3a4a2f7254b88fecf32cb83e947d14d6e1 100644
--- a/media/audio/mac/audio_manager_mac.h
+++ b/media/audio/mac/audio_manager_mac.h
@@ -10,6 +10,7 @@
 #include <stddef.h>
 
 #include <list>
+#include <map>
 #include <memory>
 #include <string>
 
@@ -127,6 +128,22 @@ class MEDIA_EXPORT AudioManagerMac : public AudioManagerBase {
   // sample rate has changed, otherwise does nothing.
   void HandleDeviceChanges();
 
+  // Returns true if any active input stream is using the specified |device_id|.
+  bool AudioDeviceIsUsedForInput(AudioDeviceID device_id);
+
+  // This method is called when an output stream has been released and it takes
+  // the given |device_id| and scans all active output streams that are
+  // using this id. The goal is to find a new (larger) I/O buffer size which
+  // can be applied to all active output streams since doing so will save
+  // system resources.
+  // Note that, it is only called if no input stream is also using the device.
+  // Example: two active output streams where #1 wants 1024 as buffer size but
+  // is using 256 since stream #2 wants it. Now, if stream #2 is closed down,
+  // the native I/O buffer size will be increased to 1024 instead of 256.
+  // TODO(henrika): possibly extend the scheme to also take input streams into
+  // account.
+  bool IncreaseIOBufferSizeIfPossible(AudioDeviceID device_id);

[Henrik Grunell, 2016/04/28 12:42:50]

Would it be possible to merge this function with MaybeChangeBufferSize()? Just
one general function that maybe changes the size to the optimal one.

[henrika (OOO until Aug 14), 2016/04/28 12:46:23]

I would like to avoid that at this stage since it would make it even more
complicated. This new method is only designed for output. Maybe if input side is
included as well.

At this stage I can't see any benefit by doing such a merge.

+
   std::unique_ptr<AudioDeviceListenerMac> output_device_listener_;
 
   // Track the output sample-rate and the default output device
@@ -149,6 +166,10 @@ class MEDIA_EXPORT AudioManagerMac : public AudioManagerBase {
   std::list<AUAudioInputStream*> low_latency_input_streams_;
   std::list<AUHALStream*> output_streams_;
 
+  // Maps device IDs and their corresponding actual (I/O) buffer sizes for
+  // all output streams using the specific device.
+  std::map<AudioDeviceID, size_t> output_io_buffer_size_map_;
+
   DISALLOW_COPY_AND_ASSIGN(AudioManagerMac);
 };