Re: [PATCH V4 4/4] mm: microoptimize zonelist operations

[Mel Gorman]

On Wed, Jan 07, 2015 at 11:57:49AM +0100, Michal Hocko wrote:
> On Wed 07-01-15 10:15:39, Vlastimil Babka wrote:
> > On 01/06/2015 04:09 PM, Michal Hocko wrote:
> > > On Mon 05-01-15 18:17:43, Vlastimil Babka wrote:
> > >> The function next_zones_zonelist() returns zoneref pointer, as well as zone
> > >> pointer via extra parameter. Since the latter can be trivially obtained by
> > >> dereferencing the former, the overhead of the extra parameter is unjustified.
> > >> 
> > >> This patch thus removes the zone parameter from next_zones_zonelist(). Both
> > >> callers happen to be in the same header file, so it's simple to add the
> > >> zoneref dereference inline. We save some bytes of code size.
> > > 
> > > Dunno. It makes first_zones_zonelist and next_zones_zonelist look
> > > different which might be a bit confusing. It's not a big deal but
> > > I am not sure it is worth it.
> > 
> > Yeah I thought that nobody uses them directly anyway thanks to
> > for_each_zone_zonelist* so it's not a big deal.
> 
> OK, I have checked why we need the whole struct zoneref when it
> only caches zone_idx. dd1a239f6f2d (mm: have zonelist contains
> structs with both a zone pointer and zone_idx) claims this will
> reduce cache contention by reducing pointer chasing because we
> do not have to dereference pgdat so often in hot paths. Fair
> enough but I do not see any numbers in the changelog nor in the
> original discussion (https://lkml.org/lkml/2007/11/20/547 resp.
> https://lkml.org/lkml/2007/9/28/170). Maybe Mel remembers what was the
> benchmark which has shown the difference so that we can check whether
> this is still relevant and caching the index is still worth it.
> 

IIRC, the difference was a few percent on instruction profiles and cache
profiles when driven from a systemtap microbenchmark but I no longer have
the data and besides it would have been based on an ancient machine by
todays standards. When zeroing of pages is taken into account it's going
to be marginal so a userspace test would probably show nothing. Still,
I see little motivation to replace a single deference with multiple
dereferences and pointer arithmetic when zonelist_zone_idx() is called.

-- 
Mel Gorman
SUSE Labs
--
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/