From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1751155AbWF3Gb7 (ORCPT ); Fri, 30 Jun 2006 02:31:59 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1751161AbWF3Gb7 (ORCPT ); Fri, 30 Jun 2006 02:31:59 -0400 Received: from smtp.osdl.org ([65.172.181.4]:4557 "EHLO smtp.osdl.org") by vger.kernel.org with ESMTP id S1751155AbWF3Gb6 (ORCPT ); Fri, 30 Jun 2006 02:31:58 -0400 Date: Thu, 29 Jun 2006 23:31:51 -0700 From: Andrew Morton To: Christoph Lameter Cc: linux-kernel@vger.kernel.org, Trond Myklebust Subject: Re: ZVC: Increase threshold for larger processor configurationss Message-Id: <20060629233151.31a12d81.akpm@osdl.org> In-Reply-To: References: <20060628154911.6e035153.akpm@osdl.org> X-Mailer: Sylpheed version 2.2.4 (GTK+ 2.8.17; i686-pc-linux-gnu) Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org On Thu, 29 Jun 2006 23:15:55 -0700 (PDT) Christoph Lameter wrote: > On Thu, 29 Jun 2006, Christoph Lameter wrote: > > > > Did you consider my earlier suggestion about these counters? That, over the > > > short-term, they tend to count in only one direction? So we can do > > Uhh... We are overcompensating right? Pretty funky idea that is new to me > > and that would require some thought. > > > > This would basically increase the stepping by 50% if we are only going in > > one direction. > > A patch that does this: > > > ZVC: overcompensate while incrementing ZVC counters > > Overcompensate by a balance factor when incrementing or decrementing > ZVC counters anticipating continual increase in the same direction. Looks sensible. Please check that none of this is racy wrt memory hotplug (process_zones->vm_stat_setup). > Note that I have not been able to see any effect off this approach on > an 8p system where I tested this. > I probably will have a chance to test it on larger systems (160p) tomorrow. OK. But let's not rush - it's only fine-tuning. I'm thinking we should get what we have in -mm4 into -rc1 - I think it's stable enough, and we don't want to be carrying all those changes splatered across the VM for the next two months. Let's aim to get the well-measured fine-tuning in place for -rc2, OK? Are you aware of any to-do items remaining in the -mm4 patches? The NFS changes need a review from Trond - hopefully he'll be able to find 5-10 minutes to do that sometime?