Subject: Re: [linux-audio-dev] LAAGA restrictions
From: Paul Davis (pbd_AT_Op.Net)
Date: Fri May 18 2001 - 18:08:42 EEST
>> An event would be anything that changed the state of the system.
>
>So, change of sample rate or similar?
that would be one, but thats much more catastrophic and wouldn't be
handled in the same way (AES actually stops and re-starts the audio
device when that happens).
right now, i can't think of any events that a typical engine would be
sensitive to. things like MIDI MMC would be destined for individual
plugins, not the engine. but it seems to me that such events might be
devised or noticed at some point, and thus 2^n block sizes cannot be
guaranteed.
>> >Why is it critical to prevent an app from responding to an event before
>> >its time?
>>
>> Sorry, I wasn't very clear. You can't allow the plugins to do their
>> thing before the *engine* has processed the event. If you did, they
>> would be processing data using the old state of the system, which
>> should have been altered before they got busy.
>
>Doesn't that mean it can't happen part way through a block?
Precisely. But if the event is supposed to be handled with sample
accuracy, then you (almost certainly) have to split the processing of
a single "period" (to use ALSA terminology) in two blocks.
Anyway, for now, its moot since there are asynchronous engine events.
--p
This archive was generated by hypermail 2b28 : Fri May 18 2001 - 19:32:36 EEST