I suspect you are right. I'd also like to note that this is ground so
thoroughly trodden that the grass is flat. Realtime schedulers are a well
researched topic, it's just too bad that committees don't design them as well
as engineers would.
Thinking strictly about the needs of sound processing, what's needed is a
guarantee of so much cpu time each time the timer fires, and a user limit to
prevent cpu hogging. It's worth pondering the difference between that and
rate-of-forward-progress. I suspect some simple improvements to the current
scheduler can be made to do the job, and at the same time, avoid the
priorty-based starvation issue that seems to have been practically mandated
by POSIX.
> A dynamic-window-constrained
> scheduler (that guarantees not that you'll run until you sleep, but
> that in any (settable) time period you'll get the opportunity to run
> for at least (a smaller settable period)) is closer to what's wanted.
It's possible that may be equivalent to what I said :-)
> See http://www.cs.bu.edu/fac/richwest/dwcs.html
This is an interesting link. One of the design rules has to be that O(1)
performance is never degraded, at least when there are no realtime processes.
Also, I want to be clear that I'm not suggesting this sort of thing has
anything to do with the current cycle, unless tweaking of the incumbent
sheduler fails for some reason, which it seems unlikely to do.
Regards,
Daniel
>
> --
> Dr Peter Chubb http://www.gelato.unsw.edu.au peterc AT gelato.unsw.edu.au
> You are lost in a maze of BitKeeper repositories, all slightly different.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/