[OT] Re: rt kernel

[ailo]

On 04/06/2011 04:21 PM, Ralf Mardorf wrote:
> On Wed, 2011-04-06 at 15:36 +0200, ailo wrote:
>> Well, if you get a chance, please try them. 2.6.38, that is.
>>
>> Also, I'm hoping 2.6.39 will be even better, since it seems we will be
>> able to adjust irq priority.
> 
> No, I won't use those kernels, just because of issues that might not be
> caused by those kernels, but that unfavourable could be the putative
> cause for an issue.
> 

I'm not sure what you mean by this.
I have noticed that Ubuntu's -generic kernels performance is partly
dependent on something else, since it can vary between updates (during
the development period, this is of course not a problem).


> Bug reports for DAWs often are a problem, because of outdated libs and
> apps. The community of audio coders seems to be very small, so they
> maintain their latest software versions only. Distributions like
> openSUSE, that often update the complete version, don't provide all
> current libs and especially not audio apps. I'm still on 11.2, but the
> folks and magazines are at 11.4 at the moment. Audio coders sometimes
> use svn libs that aren't official released. For Ubuntu I noticed a trend
> to use Natty, I just installed Maverick, only because I've got serious
> issues with Ubuntu Studio Lucid, while Maverick is ok.
> 

I was very happy with Karmic. And specifically using Puredyne with the
-rt kernel included in the main repo. If all else fails, that is what I
would use for music production. The -rt kernel performs well enough for
my needs. No PA, which is what I prefer for strictly production based
system.
I've been using it only for live audio processing, and so far I have
done so little sequencing on Linux in general, so I cannot say whether
there are any problems with sync.
For my home system, I'd rather use Ubuntu or Ubuntu Studio, since I do
more than just audio processing at home.

On Ubuntu and Ubuntu Studio, I feel that what has impressed me the most
is -lowlatency, because it will give me stable performance at low
latencies, no matter which desktop I use. With 2.6.39 maybe we can get
the same nice results from firewire devices where there are irq sharing
issues.


> The choice of the used kernel, distro and apps also depends to the
> trends of the Linux community, not only to stability and use.
> 

I haven't used the different releases enough to see how they would
differ in other respects. There are probably issues with all of them.
Mostly, I find that Lucid and beyond is pretty ok. I guess it depends on
many things, like what hardware you use. Different firewire stack and
support.


> When KDE 3 became the best desktop environment that ever was programmed
> for any OS and it just needed some optimisation, they switch to KDE 4
> and gave us another unfinished desktop environment. I needed to switch
> to GNOME, with all it's disadvantages, but also with the advantage not
> to change the complete workflow.
> 
> It's hard to keep an audio Linux stable + up-to-date ... anyway ... if
> you run in issues for hard audio work and you do a bug report, you
> better use a kernel-rt (+ current libs and apps).
> 
>> Perhaps the midi problems are partly because of Alsa? Don't know much,
>> but from what I hear, alsa midi is not the most reliable.
> 
> ALSA MIDI (not audio)? Yes, perhaps. Most of the times there are no
> timing/jitter issues for ALSA MIDI and ALSA MIDI clients. JACK MIDI only
> is for internal usage and the bridges 'jackd [snip] -Xseq' or
> 'a2jmidi_bridge' usually for my needs do a good job too. At least the -X
> switch is known as less good. I can't confirm this. JACK transport again
> and again was out of sync, dunno the current state, I don't use JACK
> transport. It wasn't the bad of JACK all the time, I remember that there
> once was a bug for Hydrogen. The doubtful funny thing is, that most
> Linux musicians didn't notice this bug, while others, e.g. me, aren't
> able to make music, if sync isn't very good.
> 
> Regarding to sync, e.g. play a sequence in cycle mode and use a
> 'virtual' (internal) synth + an external MIDI standalone device in
> unison. After a while the two synth will drift completely. Jitter isn't
> the only issue.

If a -generic or a -lowlatency cannot compare to -rt, I think it would
be great to know exactly how, from a practical point of view. For us
musicians, we only care about what we can produce with it, as you
demonstrate.
I suppose this could be one of those examples, though I wonder how
directly midi sync is related to the kernel.
It is however worth gold to know what setup you can trust, in what way,
and with what hardware, even if you don't know why :) (which I seldom do).

> 
> 
> 
> 
> 
> 


-- 
ailo