Hi Daniel,
Interesting point here. Do you mean you edited the /etc/security/limits.conf even when running the non rt kernel and that it made things better than with the non-edited /etc/security/limits.conf ?
I already had realtime permissions enabled while using the rt kernel.
Yup, I basically was messing around trying to figure out why my suspend/sleep was broken in Gutsy when it had worked with the Live cd ... then I realized mabey its the rt kernel I had installed and reinstalled the generic. Of course, that fixed these problems and I tried pd + jack and noticed that it also ran those really well ... better then the rt kernel which prompted me to wonder if anyone else had noticed this behaviour? I thought it was some bug with pd or jack, in that pd was giving me noticeable xruns using the *realtime* kernel when in using such a kernel in dapper, edgy, and fiesty was the only way to *avoid* them. I remember reading somewhere that the Ubuntu kernel team has been integrating some of the realtime kernel patches and enabling pre-emption to keep the kernel fast for the whole "multimedia desktop experience" etc ... and it seems that this works with realtime audio in pd and jack, at least for me.
I also noticed that the realtime kernel has become more focused on *real* realtime requirements like robotics, etc on the ubuntu page, dont recall the link off hand.
On a side note, merry christmas boys and boys (are there any girls on the list?). My parents just bestowed a solid state harddrive upon my performance computer, so now I can bounce around without worrying about physical disc crashes! Happy nerdy xmas all.
That's definitely one thing I must try then...
To get even more off-topic (even though I'm not really convenced it is) I noticed as well that rt kernels are really aggressive. For instance when I just play mp3s through jack with the hdsp driver it eats all performance for the rest of desktop apps which then get really really sluggish (on a PIV 3000MHz...).
Jé
Daniel Wilcox a écrit :
To all those Pd Ubuntu-ers out there,
I have been using the realtime (lowlatency) kernel for a while, back since Feisty, and compiled my own in Dapper.
I have noticed that, at least with Gusty, the generic kernel works better then the rt kernel, giving me much *fewer*? xruns and artifacting with Pd ... while using the rt kernel, Pd seems to "sput" xruns while using Jack. It's important to note that I did enable the realtime privledges by writing to the security config as noted on the Ubuntu studio page: sudo su -c 'echo @audio - rtprio 99 >> /etc/security/limits.conf' sudo su -c 'echo @audio - memlock 250000 >>
/etc/security/limits.conf'
sudo su -c 'echo @audio - nice -10 >> /etc/security/limits.conf'
Could it be that the preemption enabled in the rt kernel is a bit *too* aggressive? Realtime audio in PD with Jack seems to run much better with the generic kernel! Even at a 8ms latency ... in fact I'm testing the same on my slow performance computer (PIII 500MHz) with a fresh install of a Minimal command line Gutsy and it works *soooo* much better then the realtime kernel. (Plus suspend and hibernation work, while they don't using the realtime kernel.)
-- Dan Wilcox danomatika www.robotcowboy.com http://www.robotcowboy.com