RE: bug in sched.c:activate_task()

[Chen, Kenneth W]

Ingo Molnar wrote on Monday, October 04, 2004 11:55 PM
> On Tue, 5 Oct 2004, Con Kolivas wrote:
>
> > 	unsigned long long delta = now - next->timestamp;
> >
> > 	if (next->activated == 1)
> > 		delta = delta * (ON_RUNQUEUE_WEIGHT * 128 / 100) / 128;
> >
> > is in schedule() before we update the timestamp, no?
>
> indeed ... so the patch is just random incorrect damage that happened to
> distrub the scheduler fixing some balancing problem. Kenneth, what
> precisely is the balancing problem you are seeing?

We are seeing truck load of move_task() which was originated from newly
idle load balance.  The problem is load balancing going nuts moving tons
of tasks around and what we are seeing is cache misses for the application
shots up to the roof and suffering from cache threshing.

- Ken


-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/