General discussion of EcaSound multitrack audio editor

John Rigg | 11 Oct 2006 19:49

Re: stabilty & reliability

On Wed, Oct 11, 2006 at 12:00:41PM +0200, Jure Pe??ar wrote:
> My current setup:
> 
> linux 2.6.16-rt16 smp
> 2x delta 1010LT
> alsa 1.0.11
> jackd version 0.101.0 with mmap_complex-patch-0.100.7
> ecasound-2.4.4
> 
> So, question for all that are doing similiar multichannel recordings ... what combination of kernel - rt
patch - alsa - jack - ecasound works 100% ok for you? How much is smp an issue with rt? Should I try non-smp kernel?

Are you using x86_64? There are lots of issues here. I'm using
3 x Delta 1010 with a dual Opteron. I've never been able to get -rt
kernels to work with this setup. ALSA pcm_multi causes problems for -rt
and locks up my system as soon as I start jackd. I use vanilla SMP kernel
with CONFIG_PREEMPT=y. This works well.

Another problem is pcm_multi doesn't work with jackd in duplex mode
after alsa-lib-1.0.8. There's a patch here (tested with alsa-lib 1.0.9
through 1.0.12):

http://www.sound-man.co.uk/linuxaudio/pcm_multi-patch

Another (x86_64 specific) problem is that jackd 0.101.0 and earlier
generates a spurious divide-by-zero error at seemingly random times that
kills jackd. Sometimes it runs for hours before it happens, sometimes
only a few seconds. A floating point exception message from jackd
appears on the console, and a divide trap error message from the kernel
appears in syslog when it occurs.

(Continue reading)

headers

Jure Pečar | 13 Oct 2006 02:21

Re: stabilty & reliability

On Wed, 11 Oct 2006 18:49:43 +0100
John Rigg <eca <at> sound-man.co.uk> wrote:

> I've been recording for around 5 months using the patches and workarounds
> described above, with excellent results and reliability.

Sorry, I've somehow managed to miss your first mail ...

No, I don't have a 64bit machine. It's a plain "old" dual xeon rack server. Also, I'm doing recording only, so
no duplex here. I'm surely going to dig about that alsa pcm_multi interaction with realtime kernels to see
if I come up with anything similiar to what I'm seeing ...

You say that realtime kernel patch is not strictly necessary for multichannel recordings? That would
certainly simplify some things ...

Also, your site was a great help for putting things together. Keep up the good work!

--

-- 

Jure Pečar
http://jure.pecar.org/

-------------------------------------------------------------------------
Using Tomcat but need to do more? Need to support web services, security?
Get stuff done quickly with pre-integrated technology to make your job easier
Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642
_______________________________________________
Ecasound-list mailing list
Ecasound-list <at> lists.sourceforge.net

(Continue reading)

headers

John Rigg | 13 Oct 2006 10:18

Re: stabilty & reliability

On Fri, Oct 13, 2006 at 02:21:06AM +0200, Jure Pe??ar wrote:
> You say that realtime kernel patch is not strictly necessary for multichannel recordings? That would
certainly simplify some things ...

If you are just recording and not doing anything that needs very low
latency (like MIDI), then a non-rt SMP kernel should work well.
In fact the -rt patch makes use of some of the kernel's SMP facilities
and applies them to single CPUs, so if you have real SMP to start with
you're most of the way there.

John

-------------------------------------------------------------------------
Using Tomcat but need to do more? Need to support web services, security?
Get stuff done quickly with pre-integrated technology to make your job easier
Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642
_______________________________________________
Ecasound-list mailing list
Ecasound-list <at> lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/ecasound-list