From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1751035AbWBXPUO (ORCPT ); Fri, 24 Feb 2006 10:20:14 -0500 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1751030AbWBXPUO (ORCPT ); Fri, 24 Feb 2006 10:20:14 -0500 Received: from kanga.kvack.org ([66.96.29.28]:63392 "EHLO kanga.kvack.org") by vger.kernel.org with ESMTP id S932186AbWBXPUN (ORCPT ); Fri, 24 Feb 2006 10:20:13 -0500 Date: Fri, 24 Feb 2006 10:15:10 -0500 From: Benjamin LaHaise To: Alan Stern Cc: Andrew Morton , sekharan@us.ibm.com, Kernel development list Subject: Re: [PATCH] Avoid calling down_read and down_write during startup Message-ID: <20060224151510.GC7101@kvack.org> References: <20060224144028.GB7101@kvack.org> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: User-Agent: Mutt/1.4.1i Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org On Fri, Feb 24, 2006 at 10:04:23AM -0500, Alan Stern wrote: > What do you think of the two suggestions in my previous message? Even if the read version of the lock only touches a cacheline local to the cpu, you'd still have to use the lock prefix to allow for correctness when a writer comes along. It is not cacheline bouncing that worries me, it is serialising instructions and memory barriers as those hurt immensely when the data is in the cache. I've been looking at a lot of profiles on P4s of late, and every single locked instruction is painful as it means all of the memory ordering rules come into play. Neither suggestion addresses that overhead that has been introduced. -ben -- "Ladies and gentlemen, I'm sorry to interrupt, but the police are here and they've asked us to stop the party." Don't Email: .