[PlanetCCRMA] System slowing down with Ardour FC4/edge-kernel (softirq-tasklet)

[chromdildo]

update:

"rtload" seems to be missing for FC4, my latest version is
rtload-0.0.2-1.rhfc3.ccrma

$$ /etc/rc.d/init.d/rtload status
=> "realtime module does not exist on this kernel [warning]"


Where can I get a valid rtload for the latest kernel?
(I think/hope his _is_ the case with my troubles.)


Thanks again for your time

Best regards
./chrom





----------------

BTW. during startup (runlevel 5) it tells
"realtime module does not exist on this kernel [warning]"

What's wrong with my FC4/CCRMA?


......


Thanks for your reply,

Tried your approach with the pci-latency... but I think my problem has
nothing to do with that.

I try to explain:
    - running CCRMA-kernel-edge + jack & Ardour (Realtime), rosegarden...:
       there's a process called "softirq-tasklet" consuming all the
available cpu-time (according to top),
       the whole system gets sloppy - but only when in "playing"-mode of
ardour, rosegarden, muse...
    - 2nd try was the stock kernel of FC4 (oh... updated via yum, I
think) - same thing,
       but the disturbing process wasn't "softirq-tasklet", it was "X"
consuming all cpu when an app was playing.
      ...also tried cinelerra with same results (video output hangs
while playing)...

Audio is clear and fine, good latency, but I can't really use anything
together with ardour....
whats the matter, am I doing something wrong?

System is: AMD 2800+, 1GB RAM, IDE HD, Delta 1010LT, geforce FX5200
(Dualhead), FedoraCore4 (upgraded from FC3)
using the ccrma-core-edge kernel


Thanks again for any replys

Best regards
./chrom


Tracey Hytry wrote:

>On Thu, 14 Jul 2005 10:46:40 +0200
>chromdildo <chromdildo@t-online.de> wrote:
>
>  
>
>>Whats the matter? On FC3 everything was ok. btw. installed a new nVidia
>>FX5200 (Dualhead/Xinerama).
>>Got something todo with it?
>>    
>>
>
>I'm not sure if this has anything to do with it;  all I can say is I'll pass on some more of my experiences.  This happened all with FC3:
>
>I'm not very trusting of the nvidia drivers, especially on top of these kernels.  I feel like we're taking them to places they were never designed for.
>
>I was using an FX5200 with the NVIDIA-Linux-x86-1.0-7167 drivers before.  When I tried to upgrade the drivers I had trouble trying to get Xwindows to work in 1920x1200 mode on a DFP;  so I had to stick to the 1.0-7167 version.  It seemed that the newer version of the drivers couldn't comprehend what I told the earlier one to do.
>
>We also had a couple of what apeared to be lockups on another machine here with the same edge kernel, same nvidia driver, but different card.  It turned out it was the driver doing loops and other things, consuming most of the cpu.  Somewhere I think I read that these drivers may not be built for this kind of schedualing.
>
>As an aside, whenever X starts, the nvidia driver resets the pci latency timer to the maximum value.  This overides any setup I tell the machine to do during boot.  Until I can figure out where in the X startup proccess to change it back;  I run this from a term after X starts:
>
>#######################################################
>
>#!/bin/sh
>#
># used to reset latency timers to 64
># after the nvidia driver rudely sets itself to 248
># when xwindows is started
>
>setpci -v -d *:* latency_timer=40
>
>#######################################################
>
>We found that setting everything from the normal 32 to 64 worked pretty well for us.
>
>Tracey
>
>_______________________________________________
>PlanetCCRMA mailing list
>PlanetCCRMA@ccrma.stanford.edu
>http://ccrma-mail.stanford.edu/mailman/listinfo/planetccrma
>
>  
>