From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id ; Tue, 26 Jun 2001 18:36:04 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id ; Tue, 26 Jun 2001 18:35:43 -0400 Received: from enhanced.ppp.eticomm.net ([206.228.183.5]:21754 "EHLO intech19.enhanced.com") by vger.kernel.org with ESMTP id ; Tue, 26 Jun 2001 18:35:33 -0400 To: "Albert D. Cahalan" cc: linux-kernel@vger.kernel.org Subject: Re: User space zero copy HOWTO? In-Reply-To: <200106260243.f5Q2hfJ177651@saturn.cs.uml.edu> From: Camm Maguire Date: 26 Jun 2001 18:35:28 -0400 In-Reply-To: "Albert D. Cahalan"'s message of "Mon, 25 Jun 2001 22:43:41 -0400 (EDT)" Message-ID: <54d77qg98v.fsf@intech19.enhanced.com> X-Mailer: Gnus v5.7/Emacs 20.7 Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org Greetings, and thanks for this. If I recall, there is some facility for this in the tulip cards too, no? Can one get any benefit with 100Mbps, or is the copy too fast anyway? Is the source code for Tux available somewhere? This is probably the best bet. It seems as though the source for X15 is not available. Any pointers appreciated. I'm investigating implementing something like this for the lam MPI libs. Take care, "Albert D. Cahalan" writes: > > Greetings! Is there any faq/sample code somewhere showing how to get > > zero copy tcp/ip with kernel 2.4, and what special hardware if any is > > required? Any information most appreciated. Kindly cc me directly. > > The hardware must do scatter-gather and IP checksuming. > > The Alteon-based gigabit cards do this well. They are fully > programmable, and even have a developer's kit that can be used > to implement non-IP message passing protocols. > > > > -- Camm Maguire camm@enhanced.com ========================================================================== "The earth is but one country, and mankind its citizens." -- Baha'u'llah