From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: with ECARTIS (v1.0.0; list xfs); Tue, 25 Jul 2006 03:51:44 -0700 (PDT) Received: from pentafluge.infradead.org (pentafluge.infradead.org [213.146.154.40]) by oss.sgi.com (8.12.10/8.12.10/SuSE Linux 0.7) with ESMTP id k6PAp7DX001440 for ; Tue, 25 Jul 2006 03:51:10 -0700 Date: Tue, 25 Jul 2006 10:38:42 +0100 From: Christoph Hellwig Subject: Re: review: rework bulkstat readahead Message-ID: <20060725093842.GA29615@infradead.org> References: <20060725134651.D2116482@wobbly.melbourne.sgi.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20060725134651.D2116482@wobbly.melbourne.sgi.com> Sender: xfs-bounce@oss.sgi.com Errors-To: xfs-bounce@oss.sgi.com List-Id: xfs To: Nathan Scott Cc: vapo@melbourne.sgi.com, xfs@oss.sgi.com On Tue, Jul 25, 2006 at 01:46:51PM +1000, Nathan Scott wrote: > Hi all, > > This patch rearranges the logic inside xfs_bulkstat to issue > readahead in a way that performs alot better. The most critical > piece is moving the readahead out of the final read-and-format > loop into the earlier build-up-irbuf-array loop. This means: > > - buffer readaheads are all issued at once, up front, before we > start to issue the blocking reads waiting on each inode cluster; > > - as a by-product, the readahead which is done for the internal > nodes of the AGI btree (from within xfs_inobt_increment) is now > issued in the correct order relative to all the leaf readahead. > > I've also removed the incorrect endian swapping on the irbuf array > elements, that was always unnecessary as they're only ever held in > memory here. looks good and sorry for not noticing the unessecary swapping when doing the sparse annotations.