Re: [LAD] [Fwd: Re: Summercode 2008: LASH, pt. 3]

From: Stéphane Letz <letz@email-addr-hidden>
Date: Tue Feb 05 2008 - 09:24:31 EET

Le 5 févr. 08 à 00:10, Krzysztof Foltman a écrit :

> Paul Davis wrote:
>> there are many parts of the JACK API that are not RT-bound. in fact,
>> basically *all* of it except for a clients process() callback,
>> which a
>> client is not required to register.
>>
> I guess the "synchronous" response to port appearance/connection
> events
> (like application connecting ports before next event is generated by
> JACK) pretty much requires RT in those applications. Is that right?
>
> Or are those callbacks called in non-audio thread context and don't
> block audio? (I suppose they block audio, because I see several
> watchdog
> timeouts which seem to be related to patchbay applications doing
> screen
> updates etc.)
>
> Krzysztof
>

jackdmp used a 2 threads model on client side: one RT thread for
audio (call the "process" callback) and one non RT thread for
"notification" callback (like graph reorder, port register, connect/
disconnect...)

Stéphane
_______________________________________________
Linux-audio-dev mailing list
Linux-audio-dev@email-addr-hidden
http://lists.linuxaudio.org/mailman/listinfo/linux-audio-dev
Received on Tue Feb 5 12:15:01 2008

This archive was generated by hypermail 2.1.8 : Tue Feb 05 2008 - 12:15:01 EET