3 Feb
2014
3 Feb
'14
10:13 p.m.
On 02/03/2014 09:38 PM, F Tux wrote:
Hi all!
In the past few months I've been fiddling with the kernel config to
get the best latency out of my laptop (core2 t6400, 2GHz 1MB L2, 4GB
of RAM).
The best settings I've got so far are 128*3 with a sample rate of
48000Hz. All of this with the rt-patch and following the guide in the
linux audio wiki.
My question is regarding the method described here
http://proaudio.tuxfamily.org/wiki/index.php?title=DAW_Digital_Audio_Workst…
I've seen this described in many a forum, and people say they got
better results with it than with the rt-patch.
Do you have any experiences with the cgroups solution? Is it stable
enough for live performance?
I will be testing it some time in the future, but I thought it would
be a good idea to fire up the discussion here and make it for a common
profit.
Thanks a lot to all of you. It's great to know there's a place for the
really independent musician to learn and share for free with his
fellow hackers around the globe.
Bye
Hello,
The information in the proaudio wiki seems wrong to me. cgroups do not
yield the same result as using the RT patchset as the RT patchset
targets a completely different goal. Also cgroups add an unnecessary
layer of complexity. I've never had to use them (from what I've
understood cgroups are disabled by default on Ubuntu and maybe also
Debian) and that's also one of the reasons there's nothing about them in
the linuxaudio.org wiki (of which I'm the main author). What I think
that still works best is to use a RT kernel or low-latency (PREEMPT__LL)
kernel, use rtirq and raise JACK's priority to a reasonable number. And
not the default prio of 10 that figures in the proaudio wiki, another
reason that makes me doubt if the author understands how setting rtprio
works. But I'm open to a discussion about this as I don't have any
hands-on experience with cgroups so I could be completely mistaken :)
Best,
Jeremy