Re: [LAU] realtime kernel

Hi...
> Hi everyone!
> Short question only: If I compile a realtime kernel, with Ingo's patches, do
> I need to set the timer to 1000Hz.
It still worths to use 1000 Hz even if you use realtime kernel. real
time kernel, AFAIK, makes interrupts management (especially timer) more
fluid (read: fast). However, that also depends on the timer granularity
itself. That's why using 1000Hz is important: finer granularity.

If you care about power consumption, enable no_hz so clock ticks are
skipped whenever no timers need to be handled at certain time. This is
to compensate the fact that higher HZ means more time spent to handle
interrupts.

In my experiment (for still unpublished article), making HZ higher makes
UI (User Interface) somewhat more responsible to mouse movement/click
and keyboard press. This is probably due to the way signal is handled in
Linux: they are piggybacking kernel-to-user space code path

And don't forget to use full preemption. So the full recipe is:
1000HZ + no_hz + full preemption

AFAIK, most MIDI applications now use RTC, so they don't rely on HZ
granularity. But still, using HZ=1000 bring advantage for another aspect.

> Or is it best to do so. My system is about
> 2 and a half years old athlon 2500+ CPU 1.8gHz.
>
Single CPU single core, then? Make sure not to compile your kernel for
SMP, that will make your kernel a bit smaller and somewhat faster.

Hope this helps.

regards,

Mulyadi

_______________________________________________
Linux-audio-user mailing list
Linux-audio-user@email-addr-hidden
http://lists.linuxaudio.org/mailman/listinfo/linux-audio-user
Received on Tue Nov 20 12:15:01 2007