[LAU] : jack CPU % is going crazy !

[Pieter Palmers]

Mysth-R wrote:
> 
> 
> 2008/4/7, Pieter Palmers <pieterp at joow.be <mailto:pieterp at joow.be>>:
> 
>     Mysth-R wrote:
> 
> 
> 
>         2008/4/7, Pieter Palmers <pieterp at joow.be
>         <mailto:pieterp at joow.be> <mailto:pieterp at joow.be
>         <mailto:pieterp at joow.be>>>:
> 
> 
>            Mysth-R wrote:
> 
>                I forgot to precise that my sound card is a Presonus
>         Firebox and
>                run with Freebob.
>                I got the idea to test with my U-control Uca202
>         (Behringer) Usb
>                sound-card, and there is no problems. The cpu percent is
>         quite
>                stable around 2,1%.
>                It is a bit strange because I got the rtirq script
>         congiured for
>                my firewire card. So it should not work better than my
>         firewire ...
> 
>                I am a bit lost !!  :O[
> 
> 
>            Is the 'real' cpu time (e.g. from top) also that high? It
>         could be
>            that there is a bug in the time reporting of the freebob backend,
>            resulting in a bad calculation of the cpu time.
> 
>            Greets,
> 
>            Pieter
> 
> 
> 
>         Hi,
>         It is very strange. I can't see a real correlation between cpu
>         load displayed by htop and the DSP load displayed by qjackctl.
>         In htop the cpu is quite unstable too, but it doesn'tt change at
>         the same time as the dsp load in qjackctl. Perhaps it is due to
>         a latency. but I can't say if is linked.
>         I have made some more test :
> 
>         1- runing qjackctl with freebob backend : both dsp load
>         (qjackctl) and cpu load (htop) are unstable but not really linked.
>         2- runing qjackctl with alsa backend : dsp load is quite stable
>         but cpu load is unstable.
>         3- running jackdmp in a console with Freebob then alsa : cpu
>         load is unstable.
>         4- when jackdmp or qjackctl are killed : cpu load is normal
>         around 2% and stable.
> 
>         So :
>         Is it normal that the cpu load displayed in htop is unstable ?
>         If it is, then as you said pieter, could it be a bug with
>         freebob ??? I never saw this before, on the same laptop but with
>         32bits OS.
> 
>     FreeBoB is guaranteed 110% bug-free.
> 
> 
> 
> héhé :D
> 
>         Is there a problem in my kernel config ?
>         Is there a something wrong with jack or freebob with the 64bits
>         arch ?
> 
>     It could very well be a 64bit issue somewhere. I haven't tested
>     freebob on 64bit yet.
> 
> 
> 
>         Well, if someone as some idea for me to make more tests ... I
>         would be happy ;)
> 
> 
>     It might be interesting if you could figure out exactly what the
>     differences are between your setups. Mainly software versions etc...
> 
> 
> well, on the same laptop I got a 32bits openSuse system and a gentoo 
> 64bits in dual boot.
> I have tested on OpenSuse, with both Kde3.5 and Fvwm, just using 
> qjackctl 0.3.2 with jackd 0.109 and it works perfectly.
> Now on my gentoo 64 bits I tryed on both Kde4.0 and Fvwm-crystal, just 
> using qjackctl 0.3.2 and jackdmp0.70. I also tryed with qjackctl 0.2.23 
> (in case it was due to QT4 library) and jackdmp-svn and jackd 0.109 : in 
> all the case I got the cpu load problem.
> 
> Do you think this is a firewire problem ? perhaps I forgot an important 
> option when I compiled my kernel (2.6.24-rt1)
> I think I will try with another kernel. Will see if there is new kernel 
> available. otherwise I will try with an older (2.6.22 ...)
> 
> Thank you for your answer.
> 
> ps : pieter if you haven't tested freebob in 64 bits, perhaps you could 
> give me some guidelines, to test it and make a report for you ?

I don't really have a clue about what's going on, so it's pretty 
difficult to do so.

Can you try running jack/freebob with a very high priority (e.g. -P99) 
to see if this changes things?

Greets,

Pieter