I'm confused about how to properly patch a kernel. I tried to patch a
kernel recently with Mingo's patch, and it failed. In researching the
problem, I came up with this:
http://kernelnewbies.org/FAQ/HowToApplyAPatch
>From what I can understand, this document is telling me that I would
need to download kernel 2.6 (no incremental versions), and apply the
most recent Mingo patch (as of today, 2.6.16-rt20) from there. Prior
to reading this, I always thought that you downloaded the most recent
full kernel (as of today, 2.6.16.15), the most recent patch (as of
today, 2.6.16-rt20), and go from there. Which way is the proper way?
--
Josh Lawrence
http://www.hardbop200.com
Better late than never. I configured an Asterisk conference room at the URI lau(a)ash.97montrose.org
Let me know if you can connect to it. I can through a SIP phone that is already registered to the server but I'm not sure about one that is not. This is a recreational server, be nice. I have lots of bandwidth but not too many admin hours.
-lee
Lee Revell wrote:
> On Tue, 2006-04-25 at 08:48 -0700, Tim Howard wrote:
> > Thanks a bunch, Fernando! I just realized last night that my system
> > (FC5 OOTB + Jack and Ardour) has about 42ms of latency, which I don't
> > think qualifies as low latency...
> >
> > I do really like FC5, though. We'll stay tuned for a kernel release!
> >
>
> Are you sure you're running realtime? How do you test latency? What
> happens if you try to go lower than 42ms?
>
> Lee
>
Lee, I'm a real newbie at this... <smacks forehead> But I'm learning slowly.
I didn't realize that the "42ms" of latency displayed by qjackctl in
the setup section was NOT my system's "absolute lowest latency,
take-it-or-leave-it."
So when I loaded CCRMA's RT kernel for FC5, I was surprised to see
that latency number stay the same... Then I started messing around
with the frames per second setting, and got the number down to 5.333
msec (128 frames/sec), with a noticeably improved response time for
softsynths, for example.
I have tested this setting for xruns, and I haven't gotten any yet. I
was playing back with 8 tracks, about 12 LADSPA effects, and recording
1 stereo track, and it was fine. (That's about the most I ever do.)
What I'm wondering now is whether the stock kernel would have also
been able to work acceptably, since it could also run in RT... Hmm,
I'll have to go back and check that out.
Is there a better way to check latency, BTW?
Thanks for clueing me in, Lee.
-TimH
Hi,
I'll do a concert at May 18 2006 as part of the Scene NRW: Baltics
festival. It will be an evening with literature and music, where
besides me the Latvian writer Pauls Bankovskis will read and Selfish
from Lativa will do a DJ set.
All this will happen at the Elektra bar in Cologne, Germany. I will
perform an extended version of the stuff I did at the Linux Audio
Conference 2006, which means lots of physical modelling with [msd2D]
and [msd3D] for visuals and sound in Pure Data.
Some more info, pics and URLs at http://footils.org/cms/show/53
I still haven't found a good name for this performance. Code name
always was: "Faltig" but today another name occured to me with may be
even cuter:
"Einstürzende Bauklötze".
Or maybe not.
Ciao
--
Frank Barknecht _ ______footils.org_ __goto10.org__
Hi,
Yet Another Stupid Question: I have a Creative Labs Soundblaster Audigy,
with no soundfonts loaded onto it. I have placed the CL CD in and found
no soundfonts with it. I do have the PC51f.sf2 soundfont, which I assume
you load onto the soundcard with asfxload, but the last time I tried
this I couldn't make my any noise.
Could anyone point out what I should do/have done wrong/which friendly
manual I should read?
Jonty
Hello, all!
Me and my girlfriend are planning a trip to Europe for a month between
May and June.
It will be our first trip abroad, so please help if you can with a good advice.
Since it's not so easy to get out of here (russia), and I can't visit
all that great & numerous openlab's etc., one of the main goals is
"community building":
I'd like to meet (at least some of ) you, LADs/LAUs, in person.
Current plan is (yes, it's huge):
Moscow
St.Petersurg
--
Helsinki
Stockholm
Berlin
Koeln
Brussels
London
Paris
Bourdeaux
Madrid
Barcelona,Figeras
Genova
Milan?
Rome
Venice
Vienna
Prague
--
Moscow
There are two main questions
(lonelyplanet.com helps a bit, but residents can say much more):
1) what is the cheapest way to travel inside europe?
Bus? Airplane? If it's by air, how can we buy cheapest tickets? Which
company? How much days before the flight? etc. Or may be hitchhiking?
2) how hard is to find cheap place to sleep?
AND
Where can we meet YOU?
Sorry, we don't know exact dates as we are researching on european
visa (trip begins from Helsinki, since it seems the easiest way to get
into europe).
Sincerely,
Dmitry.
On Wed, 2006-05-10 at 14:37 +0200, Thomas Gleixner wrote:
>
> It should be not that hard to seperate them also for non highres mode,
> but I never thought about it though.
OK. It's not a problem to require high res timers for this, however
many users disable them as the code was not stable until recently.
I think the separation would be useful if high res timers and realtime
preemption will be independently submitted upstream.
Lee
On Wed, 2006-05-10 at 14:13 +0200, Thomas Gleixner wrote:
> On Wed, 2006-05-10 at 13:32 +0200, Wolfgang Hoffmann wrote:
> > Thomas, does this dynamic priority support work when running softirq-timer/0
> > at SCHED_OTHER? I tried, and it seems to work here, I just want to make sure
> > it's valid to do so.
>
> The softirq-timer/0 is not affected by this. Only the timers which are
> in the hrtimer subsystem are handled that way.
>
> Thats currently: nano_sleep, itimer, posix_timers
Sorry, I'm more confused than I was at the beginning of the thread.
How can the OP's problem (setting softirq-timer to SCHED_FIFO priority
99 is required to get sleep() to work, but causes large latencies when
rt_secret_rebuild() runs) be addressed? Is the only solution to enable
high res timers?
Lee
On Wed, 2006-05-10 at 08:25 +0200, Thomas Gleixner wrote:
> On Wed, 2006-05-10 at 02:15 -0400, Steven Rostedt wrote:
> > I've CC'd Thomas Gleixner. I beleive that the hrtimers code now uses
> > priorities. That is, when a timer is to go off, the hrtimer softirq
> > inherits the priority of the task that created the timer. Thomas, correct
> > me if I'm wrong. But I also believe this code is a work in progress, and
> > the more feedback of what people want/need the better we can make the code
> > do just that.
> >
> > Is the high res timers turned on?
>
> http://www.linutronix.de/index.php?module=News&id=cntnt01&cntnt01action=det…
>
> This is only true, when high resolution timers are enabled.
Is there any solution for the !CONFIG_HIGH_RES_TIMERS case? I for one
can't use it because my hardware doesn't support it (no local APIC).
Lee
On Wed, 2006-05-10 at 12:56 +0200, Thomas Gleixner wrote:
> On Wed, 2006-05-10 at 06:49 -0400, Lee Revell wrote:
> > >
> > > This is only true, when high resolution timers are enabled.
> >
> > Is there any solution for the !CONFIG_HIGH_RES_TIMERS case? I for one
> > can't use it because my hardware doesn't support it (no local APIC).
>
> What hardware platform ?
>
Via C3 mini ITX board.
> HRT works with PIT only too.
I was not aware of that.
> Otherwise you really need to chrt the
> hrtimers softirq prio statically.
>
The problem was that setting the thread's priority to 99 caused long
latencies every 10 minutes when rt_secret_rebuild() ran, as if the timer
callbacks that do real work and the ones that just deliver a signal are
all running from the softirq timer thread.
Lee
> tglx
>
>
>