Ingo Molnar 12341234123412341324123412wrote:
>
> > i guess this refers to something like binding processes to CPUS like
>
> yes it exits, it's called the 'pset patch'.
looked at that a while ago, but i won't go into that now (the details
doesn't matter much as adding the functionality (or a subset thereof)
would be relatively simple and cheap anyway)
> The pset
> patch is handy for eg. SMP realtime applications where some CPU is
> dedicated to a RT process. But i think it's safe to say that in any other
> case explicit scheduling-control is a loss.
even in this case i don't see this being a clear win ;) [the RT thread
blocking could make a difference, but if the wakeup is handled properly
it shouldn't be a problem (assuming anything that blocks also can take
the cache refill hits etc seems reasonable).] With a cpu dedicated to
more than one RT thread pset could make more difference, but it's not
that obvious either.
What could be useful would be 'isolation', ie being able to prevent
the dedicated cpu (set) from processing interrupts. Would dynamic irq
routing be possible on ia32?
-
Linux SMP list: FIRST see FAQ at http://www.irisa.fr/prive/dmentre/smp-howto/
To Unsubscribe: send "unsubscribe linux-smp" to [EMAIL PROTECTED]
