On 01/08/2011 03:55 PM, Robin Gareus wrote:
> On 01/08/2011 03:16 PM, Bernardo Barros wrote:
>> ondemand is no good because when it switches the processor actually
>> stops for a very short time. but this short drops are enough to cause
>> the terrible xruns on jack. stick to performance or conservative :-)
>
> I'm tempted to call BS.
>
> I can't find a reference or link but IIRC the "short time" a processor
> stalls when changing frequency is these days in the usec range.
>
> Anyway, discard all theory. It works empirically:
> There are no problems here with JACK at 32*2 fpp @48kHz with the
> ondemand govenour on both Intel Core Duo, Intel Core 2 Duo and Intel i5
> CPUs (yes, that is going from nought to 16 jconvolvers causing CPU and
> JACK load to jump).
anecdotal evidence: on my system(s), using ondemand is absolutely deadly
when using the ffado backend.
with the alsa backend, the system is mostly ok when running ondemand,
but then again i run fairly high latencies because jconvolver gets
really expensive when you shrink the period size...
_______________________________________________
Linux-audio-user mailing list
Linux-audio-user@email-addr-hidden
http://lists.linuxaudio.org/listinfo/linux-audio-user
Received on Mon Jan 10 12:15:03 2011
This archive was generated by hypermail 2.1.8 : Mon Jan 10 2011 - 12:15:03 EET