From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1752541AbbAGK6I (ORCPT ); Wed, 7 Jan 2015 05:58:08 -0500 Received: from mail-wi0-f175.google.com ([209.85.212.175]:33375 "EHLO mail-wi0-f175.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752080AbbAGK5w (ORCPT ); Wed, 7 Jan 2015 05:57:52 -0500 Date: Wed, 7 Jan 2015 11:57:49 +0100 From: Michal Hocko To: Vlastimil Babka Cc: Andrew Morton , linux-mm@kvack.org, linux-kernel@vger.kernel.org, Mel Gorman , Zhang Yanfei , Minchan Kim , David Rientjes , Rik van Riel , "Aneesh Kumar K.V" , "Kirill A. Shutemov" , Johannes Weiner , Joonsoo Kim Subject: Re: [PATCH V4 4/4] mm: microoptimize zonelist operations Message-ID: <20150107105749.GC16553@dhcp22.suse.cz> References: <1420478263-25207-1-git-send-email-vbabka@suse.cz> <1420478263-25207-5-git-send-email-vbabka@suse.cz> <20150106150920.GE20860@dhcp22.suse.cz> <54ACF93B.3060801@suse.cz> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <54ACF93B.3060801@suse.cz> User-Agent: Mutt/1.5.23 (2014-03-12) Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org 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. -- Michal Hocko SUSE Labs