A lot of improvements to OS X output plugin #14
Conversation
which ensures mpd do bit perfect playback on OS X
do not hog system device.
…initialization problem
Wang-Yue
changed the title
| @@ -101,6 +103,8 @@ osx_output_configure(OSXOutput *oo, const ConfigBlock &block) | |||
| } | |||
|
|
|||
| oo->channel_map = block.GetBlockValue("channel_map"); | |||
| oo->hog_device = block.GetBlockValue("hog_device", false); | |||
| oo->sync_sample_rate = block.GetBlockValue("sync_sample_rate", false); | |||
There was a problem hiding this comment.
These must be documented in doc/user.xml
| @@ -665,9 +616,9 @@ osx_output_disable(AudioOutput *ao) | |||
|
|
|||
| AudioComponentInstanceDispose(oo->au); | |||
|
|
|||
| if (oo->hog_device) { | |||
| if (oo->hog_device) { | |||
There was a problem hiding this comment.
a tab. I previously used spaces for indent.
| osx_output_hog_device(oo->dev_id, false); | ||
| } | ||
| } |
|
Thanks @Wang-Yue ! I am too late to review this but I am going to ask a few questions so I can learn. |
|
|
||
| return noErr; | ||
| int count = in_number_frames * od->asbd.mBytesPerFrame; | ||
| buffer_list->mBuffers[0].mDataByteSize = |
There was a problem hiding this comment.
How do you know there is only one output buffer? Is this documented by Apple anywhere?
To be clear, I am not asking because I think this is wrong, but because I wonder how one is supposed to find out about these things.
There was a problem hiding this comment.
I know about the comments in /System/Library/Frameworks/CoreAudio.framework/Headers/*.h but it is hard to learn from those comments if you don't know what to look for.
There was a problem hiding this comment.
This is undocumented (The entire CoreAudio is poorly document) but I was told by an Apple engineer.
| int count = in_number_frames * od->asbd.mBytesPerFrame; | ||
| buffer_list->mBuffers[0].mDataByteSize = | ||
| od->ring_buffer->pop((uint8_t *)buffer_list->mBuffers[0].mData, | ||
| count); |
There was a problem hiding this comment.
Maybe I missed it but when reading the spsc_queue docs I nowhere saw any promises about how many elements will be popped. How do we know we get count bytes, or at least a multiple of mBytesPerFrame?
There was a problem hiding this comment.
You don't get this promise. But if it's not multiple mBytesPerFrame it won't cause any trouble. During the next call CoreAudio will get the remaining bytes.
|
|
||
| // We set the frame size to a power of two integer that | ||
| // is larger than buffer_frame_size. | ||
| while (*frame_size < buffer_frame_size + 1) { |
There was a problem hiding this comment.
Would it make sense to call this variable buffer_size instead of frame_size? When I read frame_size I think of the size of a single frame. The buffer (spsc_queue) holds multiple frames.
There was a problem hiding this comment.
Feel free to rename it.
Add sample rate synchronization, which ensures mpd play bit perfectly on OS X. User can enable this by setting sync_sample_rate to "true".
Add device hogging capability. This forbids other process interfere with selected user devices. User may enable this by setting "hog_device" to "true".
Kill the mutex, locks, extra buffer in original code. Now you'll see greater than 3X performance improvement.
Set the right buffer length for the device.
Fix an audio unit initialization problem.
Code borrowed from the CMUS project. I am the author of cmus coreaudio output plugin. So I have full rights to redistribute the code. The coding style may be slightly different, but we can always polish/refactor it in another PR.
@jacobvosmaer Please take a look.
@MaxKellermann please merge.