[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20070912083203.GA5005@skynet.ie>
Date: Wed, 12 Sep 2007 09:32:04 +0100
From: mel@...net.ie (Mel Gorman)
To: KAMEZAWA Hiroyuki <kamezawa.hiroyu@...fujitsu.com>
Cc: Lee.Schermerhorn@...com, akpm@...ux-foundation.org, ak@...e.de,
clameter@....com, linux-kernel@...r.kernel.org, linux-mm@...ck.org
Subject: Re: [PATCH 4/6] Embed zone_id information within the zonelist->zones pointer
On (12/09/07 16:51), KAMEZAWA Hiroyuki didst pronounce:
> On Tue, 11 Sep 2007 22:31:27 +0100 (IST)
> Mel Gorman <mel@....ul.ie> wrote:
>
> > Using two zonelists per node requires very frequent use of zone_idx(). This
> > is costly as it involves a lookup of another structure and a substraction
> > operation. As struct zone is always word aligned and normally cache-line
> > aligned, the pointer values have a number of 0's at the least significant
> > bits of the address.
> >
> > This patch embeds the zone_id of a zone in the zonelist->zones pointers.
> > The real zone pointer is retrieved using the zonelist_zone() helper function.
> > The ID of the zone is found using zonelist_zone_idx(). To avoid accidental
> > references, the zones field is renamed to _zones and the type changed to
> > unsigned long.
> >
> At first, I welcome this patch. thanks!.
>
>
> A comment after reading all is, how about defining zonelist as following
> instead of encoding in pointer ?
> ==
> struct zone_pointer {
> struct zone *zone;
> int node_id;
> int zone_idx;
> };
>
> struct zonelist {
> struct zone_pointer _zones[MAX_ZONES_PER_ZONELIST + 1];
> };
>
> #define zonelist_zone(zl) (zl)->zone
> #define zonelist_zone_idx(zl) (zl)->zone_idx
> #ifdef CONFIG_NUMA
> #define zonelist_zone_nid(zl) (zl)->node_id
> #else
> #define zonelist_zone_nid(zl, i) (0)
> ==
>
That is an interesting idea. I'll try it out and see what happens.
Thanks
> If we really want to avoid unnecessary access to "zone" while walking zonelist,
> above may do something good. Cons is this makes sizeof zonlist bigger.
>
While this is true, it might not affect the size of struct zone because
of padding. One way to find out
--
Mel Gorman
Part-time Phd Student Linux Technology Center
University of Limerick IBM Dublin Software Lab
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
Powered by blists - more mailing lists