[Rosegarden-devel] couple of mmapped observations and audio stuff

[Richard B. <bo...@bo...>]

Been fiddling about with audio this week with some success and
I/ew need to make some changes to the way mmapping works for
audio segments.  For a start we need to be able to insert partial
audio events for audio segments that start before current playback
position and extend through it - so we play partial audio files.

Secondly we have a looping issue currently - it appears when we flyback
to the first loop marker we take several copies of the events around
the loop markers with us.  In the audio case I was seeing 10+ audio
events (hence 10 samples) playing instead of one - I hear the same
with normal events.


Thirdly we need to have advance warning of audio events at the sequencer
so we can queue them up in the disk thread and have them ready to go
when the sequencer needs to start spooling them out.  I've tried some
naieve methods of getting early warning (taking RealTimes off the current
event time) but of course we're still looking at a "slice" of events
at the getSlice() call.  I'm thinking we need a new method to handle 
audio events for both this case and the first case (partial audio files).

Fourthly (and this isn't to do with mmapped code as such) we need a 
better method of syncing the MIDI and audio together.  Currently it's
all a bit handwavey - use RealTimes and potential latency figures derived
from JACK plus the one we inject into ALSA to cue them up together.
However with JACK's frame clock we have a potentially hi resolution 
medium (not to mention the other sync methods) to tighten these up.
I think at the moment we're also seeing some natural latency in audio
playback because our audio events give no time to cue up and start
spooling the audio data to the disk thread.  

R