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

[Alessio Igor Bogani]

Hi Ralf,

2010/9/30 Ralf Mardorf <ralf.mardorf at alice-dsl.net>:

[...]
> First of all, "on-demand" for the CPU frequency scaling vs "performance"
> cat config-[...] | grep CONFIG_CPU_FREQ_DEFAULT_GOV_PERFORMANCE
> should be set up to 'y'.
[...]
> I experienced that Rui's rtirq script doesn't have much impact, but it
> anyway should be included.
[...]
> On some machines JACK1 seem to work better and on other machines JACK2.
[...]
> A big problem is MIDI, when controlling stand alone devices, then jitter
> very often is an issue.
[...]
> If we run uname -a and the kernel is just a
>
> PREEMPT
> but a
> PREEMPT RT
>
> 'things' are more worse. IMO we only do need 'real' real-time kernels.
[...]
> We don't need latest Desktop candy supported by generic kernels, but
> well tuned disros + kernel-rt, to get a good audio workstation.

Life is a trade-off: someone want a very hard real-time system without
adopt other hardware architectures than pc (MCU, FPGA/ASIC, and so
on), others want use normal and cheap pc, others want use also
accelerated drivers also, others want a real-system which take care of
they laptop's batteries too. And at the end we have limited
resources(*) for give an answer on these needs (we are all
volunteers).

Obviously We can't do all these things so we have to choose: It is sad
but it is the life.

It is on leader, developers team and comunity choose that trade-off.

Ciao,
Alessio

(*) In ideal world we certainly choose to work on -rt instead than on
-lowlatency.