[proaudio] Re: [proaudio] Re: [proaudio] Re: [proaudio] Has anyone experimented with different rlimits values in /etc/security/limits.conf?

[Dominique Michel <dominique.michel@xxxxxxxxxxxx>]

Le Sun, 11 Jun 2006 21:35:20 -0500,
"Colin Shapiro" <cmshapiro@xxxxxxxxx> a écrit :

> Well, I've now downgraded to pam-0.78-r5 and switched to realtime-lsm, and
> I've successfully played my digital piano with LinuxSampler for two straight
> hours without any xruns.
> 
> So, I think it's safe to say (at least for myself and Dominique), that
> realtime-lsm just works better than PAM rlimits.
> 
> This leads me to believe that more can be done to optimize rlimits for pro
> audio work.  I've scoured the internet for hours, and barely found any good
> information on rlimits.  The documentation is scarce, and people who are
> using it are just guessing.  Of course, this is new, so we'll probably just
> have to wait as it enters the Linux audio world.
> 
> If anyone has anything to add, I'd be glad to hear it.
> 
> - Colin

I believe at it is (at least) 2 things to take in account:

1) We have to manage both the hardware and software priorities. The lsm module use a simplest way to do that as PAM that is very complex. And I am not sure at Pam-rlimits manage the hardware priorities as well as the rt-lsm do.

2) The lsm module seam to have have less CPU overload as PAM, so I don't think at it will be possible to archive the same good result with PAM-rlimits as with the rt.lsm. (I am not sure about that because I don't know the internal of those 2 programs, but the result is at the rt-limits is better for realtime operations. The rt-lsm was developed for that from the beginning, not pam that is a login-access management system.)

But I am glad to ear at your problem seam to be fixed.

Best,
Dominique