[SNIP me saying that low priority stuff should have extremely
long slices to lower memory pressure etc.]
> IMO you are mixing two things: priorities and lengths of
> timeslices. These are separate: you want low-priority short-timeslice
> for screensavers so that multiple screensavers move smoothly; you want
> low-priority long-timeslice for simulations. You might even want
> high-priority long-timeslice for important simulations :-). Do not mix
> that up.
- large simulations should have long slices, since they're large
- for small processes, we don't really care about the lenght of
the timeslice (they don't give much memory pressure)
- the border between small and large is dynamically adjusted
according to memory pressure (not a 'hard' border, but a full
spectrum of slice lengths)
- a screen saver will voluntarily give up CPU (even an OpenGL
xlockmore thingy :) so the rules don't really apply.
- a very important simulation won't be run under SCHED_IDLE :-)
Taking these things into account, we might be able to produce
quite a workable system...
Alternatively, we adapt/rewrite the multilevel QNX scheduling
code to do this for us (and to avoid the process starvation
that happened in the original QNX-style scheduler)
Rik.
+-------------------------------------------------------------------+
| Linux memory management tour guide. H.H.vanRiel@phys.uu.nl |
| Scouting Vries cubscout leader. http://www.phys.uu.nl/~riel/ |
+-------------------------------------------------------------------+
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.altern.org/andrebalsa/doc/lkml-faq.html