On Wed, Sep 29, 2010 at 03:41:59PM +0100, Mel Gorman wrote: > On Wed, Sep 29, 2010 at 09:34:09AM -0500, Christoph Lameter wrote: > > On Wed, 29 Sep 2010, Mel Gorman wrote: > > > > > > Updating the threshold also is expensive. > > > > > > Even if it's moved to a read-mostly part of the zone such as after > > > lowmem_reserve? > > > > The threshold is stored in the hot part of the per cpu page structure. > > > > And the consequences of moving it? In terms of moving, it would probably > work out better to move percpu_drift_mark after the lowmem_reserve and > put the threshold after it so they're at least similarly hot across > CPUs. > I should be clearer here. Initially, I'm thinking the consequences of moving it are not terrible bad so I'm wondering if you see some problem I have not thought of. If the threshold value is sharing the cache line with watermark or lowmem_reserve, then it should still have the same hotness in the path we really care about (zone_watermark_ok for example) without necessarily needing to be part of the per-cpu structure. The real badness would be if an additional cache line was required due to the move but I don't think this is the case (but I didn't double check with pahole or the back of an envelope either). The line will be dirtied and cause a bounce when kswapd wakes or goes to sleep but this should not be a severe problem. -- Mel Gorman Part-time Phd Student Linux Technology Center University of Limerick IBM Dublin Software Lab -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@xxxxxxxxxx For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>