Re: 2.4.10pre7aa1

Dipankar Sarma (dipankar@in.ibm.com)
Mon, 17 Sep 2001 14:43:15 +0530


In article <Pine.LNX.4.33L.0109161433530.9536-100000@imladris.rielhome.conectiva> you wrote:
> On Sun, 16 Sep 2001, Andrea Arcangeli wrote:

>> However the issue with keventd and the fact we can get away with a
>> single per-cpu counter increase in the scheduler fast path made us to
>> think it's cleaner to just spend such cycle for each schedule rather
>> than having yet another 8k per cpu wasted and longer taskslists (a
>> local cpu increase is cheaper than a conditional jump).

> So why don't we put the test+branch inside keventd ?

> wakeup_krcud(void)
> {
> krcud_wanted = 1;
> wakeup(&keventd);
> }

> cheers,

> Rik
> --

keventd is not suitable for RCU at all. It can get starved out by RT
threads and that can result in either memory pressure or performance
degradation depending on how RCU is being used.

I have a patch that uses a per-cpu quiescent state counter. Cost of
this on schedule() path is one per-cpu counter increment. I will
mail out the patch as soon as I can complete testing Andrea's review
comments on a bigger SMP box.

Most impartantly :-) it doesn't use kernel threads.

Thanks
Dipankar

-- 
Dipankar Sarma  <dipankar@in.ibm.com> Project: http://lse.sourceforge.net
Linux Technology Center, IBM Software Lab, Bangalore, India.
-
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/