From mboxrd@z Thu Jan 1 00:00:00 1970 From: Andi Kleen Subject: Re: [PATCH] NET : change layout of ehash table Date: Fri, 9 Feb 2007 09:57:24 +0100 Message-ID: <200702090957.25110.ak@suse.de> References: <200702071159.34935.dada1@cosmosbay.com> <20070209.004019.75431263.davem@davemloft.net> Mime-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Cc: dada1@cosmosbay.com, linux@horizon.com, akepner@sgi.com, netdev@vger.kernel.org To: David Miller Return-path: Received: from mx2.suse.de ([195.135.220.15]:58513 "EHLO mx2.suse.de" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1946230AbXBII5g (ORCPT ); Fri, 9 Feb 2007 03:57:36 -0500 In-Reply-To: <20070209.004019.75431263.davem@davemloft.net> Content-Disposition: inline Sender: netdev-owner@vger.kernel.org List-Id: netdev.vger.kernel.org On Friday 09 February 2007 09:40, David Miller wrote: > From: Andi Kleen > Date: 09 Feb 2007 10:18:03 +0100 > > > David Miller writes: > > > > > > I've applied this, but I _REALLY_ don't like the new multiply > > > instructions that are used now in the hash indexing paths when > > > CONFIG_SMP is set. > > > > > > I think that's a higher cost than the memory waste. > > > > You're serious? multiply on a modern CPU is _much_ cheaper than a cache miss > > e.g. a K8 can do a arbitary 64bit multiplication in 3-7 cycles. > > Any cache miss will be in the three to four digits at least. > > I'm not thinking of modern CPUs, I'm think of the little > guys :-) Even for them it's true, although the ratios are smaller (unless you go back to m68000 or VAX days) -Andi