> > > It's based on the idea of a priority heap.
> [snip]
>
> > [...] And how do you handle VM and CPU (and who knows what type of
> > other future) affinity?
>
> I haven't thought about that yet [...]
thats one of the tough things IMO ... without those constraints we could
easily get rid of the linear behavior in the scheduler. We are trying to
avoid situations where there are lots of tasks on the runqueue _and_ there
are lots of reschedules happening, this way linearity doesnt matter.
> [...], but I think we _will_
> have to think up something to make the current scheduler
> more responsive and more predictable under load [...]
do you still see responsiveness problems under pre4-2.2.8? If yes then i'd
be very interested in reproducing it.
> [...] and to
> reduce the CPU time used by niced tasks.
this can be achieved by increasing (or changing) the scale of priorities.
(this can be done seemlessly) But i doubt this matters in most cases.
(except when there are _lots_ of niced 100%-CPU using processes running)
-- mingo
-
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.tux.org/lkml/