Re: [patch v2] mm, tmp: only collapse hugepages to nodes with affinity for zone_reclaim_mode

[David Rientjes]

On Wed, 16 Jul 2014, Vlastimil Babka wrote:

> I wonder if you could do this here?
> 
> if (khugepaged_node_load[nid])
> 	return false;
> 
> If the condition is true, it means you already checked the 'nid' node against
> all other nodes present in the pmd in a previous khugepaged_scan_pmd iteration.
> And if it passed then, it would also pass now. If meanwhile a new node was found
> and recorded, it was also checked against everything present at that point,
> including 'nid'. So it should be safe?
> 
> The worst case (perfect interleaving page per page, so that "node != last_node"
> is true in each iteration) complexity then reduces from O(HPAGE_PMD_NR *
> MAX_NUMNODES) to O(HPAGE_PMD_NR + MAX_NUMNODES) iterations.
> 

Excellent suggestion, thanks Vlastimil!
--
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/