Re: [PATCH 3/6] sched: high-res preemption tick

From: Peter Zijlstra
Date: Thu Nov 01 2007 - 06:12:39 EST


On Wed, 2007-10-31 at 22:53 +0100, Andi Kleen wrote:
> Peter Zijlstra <a.p.zijlstra@xxxxxxxxx> writes:
>
> > Use HR-timers (when available) to deliver an accurate preemption tick.
> >
> > The regular scheduler tick that runs at 1/HZ can be too coarse when nice
> > level are used. The fairness system will still keep the cpu utilisation 'fair'
> > by then delaying the task that got an excessive amount of CPU time but try to
> > minimize this by delivering preemption points spot-on.
>
> This might be costly when hrtimers happen to use an more expensive
> to reprogram time source. Even an APIC timer access is fairly slow.
> And you'll potentially add the to lots of context switces.
>
> Not sure that is a good idea for performance in general.

Right, now I remember.

The idea was to run the rest of the kernel at HZ=50 or so, nothing but
scheduling needs it anymore, and with this patch the scheduler doesn't
need it anymore either. Should be good for power. This new hrtick thing
only does a lot of ticks when there are a lot of runnable tasks, it
starts at 2 with a tick per latency/2.

Attachment: signature.asc
Description: This is a digitally signed message part