From: Andrew Morton Subject: Re: [RFC/PATCH] ext3: remove inode constructor Date: Fri, 4 May 2007 13:02:41 -0700 Message-ID: <20070504130241.c436eb43.akpm@linux-foundation.org> References: Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Cc: linux-ext4@vger.kernel.org, linux-kernel@vger.kernel.org, sct@redhat.com, adilger@clusterfs.com, clameter@sgi.com To: Pekka J Enberg Return-path: Received: from smtp1.linux-foundation.org ([65.172.181.25]:52563 "EHLO smtp1.linux-foundation.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1031503AbXEDUD2 (ORCPT ); Fri, 4 May 2007 16:03:28 -0400 In-Reply-To: Sender: linux-ext4-owner@vger.kernel.org List-Id: linux-ext4.vger.kernel.org On Fri, 4 May 2007 13:14:35 +0300 (EEST) Pekka J Enberg wrote: > As explained by Christoph Lameter, ext3_alloc_inode() touches the same > cache line as init_once() so we gain nothing from using slab > constructors. The SLUB allocator will be more effective without it > (free pointer can be placed inside the free'd object), so move inode > initialization to ext3_alloc_inode completely. I got 100% rejects against this because Christoph has already had his paws all over the slab constructor code everywhere. Was going to fix it up but then decided that we ought to make changes like this to ext4 as well. Ideally beforehand, but simultaneously is OK as long as it's simple enough. btw, for a benchmark I'd suggest just a silly create-10000-files tight loop rather than something more complex like postmark.