Re: new IRQ scalability changes in 2.3.48

From: Andrea Arcangeli (andrea@suse.de)
Date: Thu Mar 09 2000 - 08:25:19 EST


On Mon, 13 Mar 2000, Ingo Molnar wrote:

>of course there is some overhead here, but it's not at all as slow as you
>imagine. It's these two additional (nonlocked) instructions on x86:
>
> decl %0
> jz 1f
>.section offline.preempt
> call do_reschedule

So you if you do:

sys_xxx {
        spin_lock()
        spin_unlock()
}

then the spin_unlock for sure jumps always far away out of line and you'll
see that change in the bench numbers.

Andrea



This archive was generated by hypermail 2b29 : Wed Mar 15 2000 - 21:00:33 EST