>Please read the post you replied to. I don't believe that threaded UP
>is better than non-threaded UP, I believe that if you are going to have
>a source base that is threaded at all, you should run the same code
>everywhere, SMP and UP. That way you have one code base to debug and
The only reasoning for doing the preemptive kernel only on UP is that in
SMP we should add explicit preemption-locks also around each very SMP
friendly per-cpu data structure (on UP this is a non-issue since the CPU
can't change from under us even if we get rescheduled ;).
On UP instead ""only"" spin_lock/spin_unlock gets bloated by preemption
lock/unlock and by explicit checks for current->need_resched after that
last unlock (and that is just enough bloating and inlined too...).
I believe we could add the preemptive kernel as a kernel option
at first and add the explicit locks around the per-cpu data structures too,
then we could choose to build UP/SMP and preemptive/non-preemptive kernel
with both option orthogonals.
Andrea
-
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/