From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1751725AbaC2NOj (ORCPT ); Sat, 29 Mar 2014 09:14:39 -0400 Received: from mail-qc0-f172.google.com ([209.85.216.172]:35138 "EHLO mail-qc0-f172.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751454AbaC2NOi (ORCPT ); Sat, 29 Mar 2014 09:14:38 -0400 Date: Sat, 29 Mar 2014 09:14:34 -0400 From: Tejun Heo To: Jianyu Zhan Cc: linux-mm@kvack.org, akpm@linux-foundation.org, cl@linux-foundation.org, linux-kernel@vger.kernel.org Subject: Re: [PATCH 2/2] mm/percpu.c: don't bother to re-walk the pcpu_slot list if nobody free space since we last drop pcpu_lock Message-ID: <20140329131434.GB5553@htj.dyndns.org> References: <1396011357-21560-1-git-send-email-nasa4836@gmail.com> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <1396011357-21560-1-git-send-email-nasa4836@gmail.com> User-Agent: Mutt/1.5.21 (2010-09-15) Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Fri, Mar 28, 2014 at 08:55:57PM +0800, Jianyu Zhan wrote: > Quoted tj: > >Hmmm... I'm not sure whether the added complexity is worthwhile. It's > >a fairly cold path. Can you show how helpful this optimization is? > > The patch is quite less intrusive in the normal path > and if we fall on the cold path, it means after satifying this allocation > the chunk may be moved to lower slot, and the follow-up allocation > of same or larger size(though rare) is likely to fail to cold path again. So > this patch could be based on to do some heuristic later. The above really doesn't show how helpful it is. This adds complexity to optimize what seemingly is a quite cold path, which often is a pretty bad idea as they tend to trade off readability and long term maintainability for almost non-existing actual gain. If you think this is a worthwhile optimization, please give justifications - use scenarios, performance numbers and so on. Thanks. -- tejun