[OT] Re: rt kernel

[Ralf]

On Sun, 2011-04-10 at 06:37 +0200, ailo wrote:
> [snip]
> What I need is a system that works, flawlessly.
> Flawlessly means, I don't suffer crashes in the middle of a performance,
> and I don't hear irregularities caused by the system.
> 
> Finding that out may be a different experience for people, because of a
> range of variables. One of them being what hardware is used. Another
> being what the system is used for.
> 
> I have yet not worked with -lowlatency for long periods of time, like I
> have with previous -rt kernels. So far, it does seem as -lowlatency will
> serve my purposes well.
> Only one -rt kernel has done that for me so far, but I have not been in
> this game for very long.

If I remember correctly, we had just one bad kernel-rt in recent years.
I guess this was at a time when there was a big step forward for
real-time audio. Instability on my computer is caused by JACK or JACK
clients. For example, Qtractor (any version I did use) crashes very
often, if you set up FluidSynth DSSI (any version I did use) during a
production. Once it's set up, Qtractor + FluidSynth DSSI are stable for
performance.

I'm using a script to kill'n'launch all audio apps during a production
session, if an app should crash.

Two days ago I add 2GB RAM, now I've got 4GB main memory. The 4GB are
accessible by my old 64-bit Linux installs, but Maverick is a 32-bit
install on my 64-bit machine and here I only get ...

edubuntu at edubuntu:/boot$ cat config-2.6.33.7.2-rt30 | grep HIGHMEM
# CONFIG_NOHIGHMEM is not set
CONFIG_HIGHMEM4G=y
# CONFIG_HIGHMEM64G is not set
CONFIG_HIGHMEM=y
edubuntu at edubuntu:/boot$ hwinfo --memory
01: None 00.0: 10102 Main Memory                                
  [Created at memory.61]
  Unique ID: rdCR.CxwsZFjVASF
  Hardware Class: memory
  Model: "Main Memory"
  Memory Range: 0x00000000-0xcb5abfff (rw)
  Memory Size: 3 GB + 256 MB
  Config Status: cfg=new, avail=yes, need=no, active=unknown

... I wonder if I need to enable 64G HIGHMEM and if this would be ok
with the RT-patch. Special wishes usually are no big deal for default
kernels, but those could be a pain for the RT-patch. This is the first
time I do have a special wish, usually I'm fine with RT-PREEMPT-kernels
for everything I do with a computer. Ok, there's one additional
exception for a short while, I need the generic kernel on Ubuntu to get
NVIDIA 3D support.

Btw. on JACK devel mailing list we were ask to test JACK from svn and to
run a JACK MIDI latency test. If the measurements on my machine for ...

edubuntu at edubuntu:/boot$ jackd -V
jackdmp 1.9.8

... and my PCI card should be ok, my computer is able to do hard
real-time for MIDI. I got an averaged MIDI jitter of 0.00 ms (0.02
frames) and a peak MIDI jitter of 0.06 ms (3 frames).
This would mean, that jitter was caused by JACK MIDI or ALSA MIDI.
Regarding to it's protocol or hardware, jitter still is caused by USB
MIDI devices. Until now I only did the measurements but no audio test.

Is somebody is interested in doing the test too, I'll write how to build
JACK dummy packages and how to install and uninstall JACK from SVN.
Building JACK dummy packages and compiling JACK from svn just takes some
minutes and is easy to do.

Too funny, Rui, coder of the 'Q'-audio-apps, qjackctl, qsynth etc. said
that latest PREEMPT kernel release candidate version 'feels' like RT
PREEMPT for his desktop.

Cheers,

Ralf