Natty and RT Kernel (was "Maverick and RT")

[Ralf Mardorf]

On Thu, 2010-09-30 at 08:19 +0200, Alessio Igor Bogani wrote:
> Hi Ronan,
> 
> 2010/9/30 Ronan Jouchet <ronan at jouchet.fr>:
> [...]
> > latencies I reached with no xruns during a 10min rakarrack session on my
> > test laptop (Dell Vostro V13 with a TI firewire card):
> > -generic: 16ms
> > -lowlatency: 4ms (@2ms: tons of xruns)
> > -realtime: 4ms (@2ms: many xruns)
> >
> > To put it simply: -lowlatency all the way! It delivers impressive
> > results for maintenance requirements way lower than -realtime (or -rt
> > even more), meaning less work for maintainers and new kernel candy for
> > users.
> [...]
> > A few diehard performance fans may appreciate a PPA with -rt, but if
> > there must be one sustainable and supported priority, it is -lowlatency.
> 
> At least at the end someone have noticed it! :-)
> 
> Ciao,
> Alessio

That's bad reasoning. Just because an app isn't ok, when using a
kernel-rt, low latency without rt isn't the better solution.
Independently, did you ensure that the kernel-rt runs with CPU frequency
scaling set up to performance? And did you test what will happen, if you
don't use rakarrack, but a heavy audio and MIDI set up? Did you compare
JACK1 and JACK2? Etc.?
Especially for external MIDI devices the so called Linux rt is far away
from hard rt.
Resume, even when using a kernel-rt, Linux is far away from hard rt, we
do need support of the kernel-rt for multimedia work and all the apps,
that do cause issues, when using a kernel-rt, need rework.

- Ralf