From mboxrd@z Thu Jan 1 00:00:00 1970 From: Patrick McHardy Subject: Re: [RFC SKBUFF]: Keep track of writable header len of headerless clones Date: Mon, 25 Jun 2007 10:53:52 +0200 Message-ID: <467F82A0.3070801@trash.net> References: <467E6950.4020707@trash.net> <20070624.193751.46198776.davem@davemloft.net> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Cc: netdev@vger.kernel.org, netfilter-devel@lists.netfilter.org To: David Miller Return-path: Received: from stinky.trash.net ([213.144.137.162]:62632 "EHLO stinky.trash.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751094AbXFYIya (ORCPT ); Mon, 25 Jun 2007 04:54:30 -0400 In-Reply-To: <20070624.193751.46198776.davem@davemloft.net> Sender: netdev-owner@vger.kernel.org List-Id: netdev.vger.kernel.org David Miller wrote: > From: Patrick McHardy > Date: Sun, 24 Jun 2007 14:53:36 +0200 > > >> - sendmsg eth0, no NAT: sys 0m2.508s >> - sendmsg eth0, NAT: sys 0m2.539s >> - sendmsg eth0, NAT + patch: sys 0m2.445s (no change) >> > > This is probably because we're touching all the data anyways > and if the resident set size is small enough (as a gigabit > or 100MB ethernet TCP stream window would be) then it all fits > mostly in the L2 caches of the cpu. Yeah, thats what I suspected too. On loopback it seems to exceed the cache size, which is a bit strange since its 1MB. Or maybe there is some loopback optimization interfering.