On Thu, 1 Jun 2000, Horst von Brand wrote:
> Ville Herva <vherva@niksula.hut.fi> said:
>
> [...]
>
> > I remember the one Be demo I saw 3-4 years ago, where the os developer
> > showed us a dual powerpc BeBox and a threaded fractal program he had made.
> > He showed the control panel where you could turn cpus on and off. He then
> > quickly switched cpus on and off to show how the mandelbrot renderer
> > reacted.
>
> Good for demos, even for laughs (as you tell). Can't think of any real use
> for such a feature, and unless there is _pressing_ need for it, the cost
> (probably a slower, more complex scheduler) just isn't worth it. Even if
> the cost is only a dozen extra lines in the kernel and no impact at all I'd
> be against it.
Provided the scheduler code respects the CPU affinity of processes, could
you not just schedule a realtime idle loop, with hard affinity for the CPU
in question? Better still, you could just "cli, hlt" on that CPU. Of
course, you'll need to disarm the NMI watchdog, and have some way to wake
up again. Skip the cli and just mask off all the IRQs except timer? Or can
one CPU kick-start another like this?
> Featuritis leads to bloated, unstable systems. Linus' genius is in large
> part that he has kept Linux away from there. Let's keep it that way.
Agreed; in this particular case, though, I'm not sure this would be a case
of featuritis.
James.
-
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/
This archive was generated by hypermail 2b29 : Wed Jun 07 2000 - 21:00:14 EST