Re: [PATCH 2/4] sched,numa: weigh nearby nodes for task placement on complex NUMA topologies

[Rik van Riel]

On 05/09/2014 05:53 AM, Peter Zijlstra wrote:
> On Thu, May 08, 2014 at 01:23:29PM -0400, riel@xxxxxxxxxx wrote:
> > +		/*
> > +		 * For nodes with distances in-between LOCAL_DISTANCE
> > +		 * and max_distance, we count the faults on those nodes
> > +		 * in proportion to their distance, using this formula:
> > +		 *
> > +		 * max_distance - node_distance
> > +		 * -----------------------------
> > +		 * max_distance - LOCAL_DISTANCE
> > +		 */
> > +		if (task)
> > +			faults = task_faults(p, node);
> > +		else
> > +			faults = group_faults(p, node);
> > +
> > +		score += 1000 * faults *
> > +				(max_distance - distance) /
> > +				(max_distance - LOCAL_DISTANCE);
>
> OK that makes sense, except I would suggest you use a power-of-two scale
> factor :-)

I guess we could build a NUMA distance table that
counts the number of hops, and use that.

That is likely to result in better/easier values
for grouping than the (somewhat arbitrary) distances
in the SLIT table, anyway...
--
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/