28 Jan
2006
28 Jan
'06
12:49 a.m.
I love Ubuntu for other people, although I'm still a Debian user by
force of habit. One of the problems I can see with using released Ubuntu
for audio is that you can get up to six months behind, and what you have
for those six months is whatever was in debian unstable at the time,
which you'd better hope wasn't broken. This is the case for anything
non-mainstream that you have to get from universe and that certainly
includes audio work.
If you're willing to build from source of course this is less of a
concern. If you're deb savvy you can but a debian testing/unstable
deb-src line in there and that way you can usually get testing/unstable
packages without messing things up.
The good news of course is that you're never more than 6 months behind
using released software. :-) Most of us Debian users are tracking
testing or unstable and although we get teased relentlessly about ice
ages and ice skating in hell, being six months behind just feels
awkward.
And in both cases, much of the quality debian packages available for
debian and ubuntu comes from the hard work of the demudi team. I've been
doing audio on debian since before demudi rolled around, and I've helped
out on a few packages, and let me tell you the difference they've made
is tremendous.
On Fri, 27 Jan 2006 at 15:20 -0800, Jon Hoskins wrote:
Hey all, thought I would share some of my recent
discoveries:
As demudi 1.2.1 has been lacking a kernel source for awhile and demudi
1.30rc1 refuses to boot on my new hardware I've had to seek alternatives.
After reading several threads on using Ubuntu as a music production OS I
decided to give Ubuntu 5.10 'breezy' a shot. Installing
"realtime-lsm"
with module-assistant was easy as pie and I got relatively good latency
with the stock kernel, but real improvements came from two things:
I decided to give Ubuntu 6.04 "Dapper Flight 3" a run and noticed that
#cat /boot/config-2.6.15-4-386 | grep PREEMPT gives CONFIG_PREEMPT=y and
has been the same in all the "dapper" kernels I've tried, something that
was not enabled in the stable "breezy" kernels. Does this mean the ubuntu
guys are looking at releasing a stable kernel with PREEMPT enabled?
Anyway the second thing I found was using 'rtlimits' instead of
"realtime-lsm". Making use of a small app I found at:
http://www.physics.adelaide.edu.au/~jwoithe/
makes assigning realtime access on an app-per-app basis and even
user/group basis a breeze. This combination has lowered my latency *under
load* to 5.8ms in KDE (Kubuntu is what I installed) and I can squeeze
2.9ms out of it in a very conservative fluxbox or even xfce4 enviroment
with no xruns and less than 10% cpu/dps usage (as reported by qjackctl).
Most of the good sound apps are available in the repositories as well with
the exceptions of Seq24 (conflicting libs) that I easily built from
scratch, Freecycle, which I'm still trying to get to compile correctly,
and Willem's wonderful DSSI packages.
It looks like Ubuntu might have a stable release on the horizon suitable
for serious audio work! You are more than welcome to check my blog as
well for an up-to-date journal of my ubuntu "studio" experiments and
installed software list: http://oktyabr.blogspot.com
Best,
Jon Hoskins
In a world without walls who needs windows or gates? --author unknown
--
Hans Fugal ; http://hans.fugal.net
There's nothing remarkable about it. All one has to do is hit the
right keys at the right time and the instrument plays itself.
-- Johann Sebastian Bach