From owner-freebsd-current@FreeBSD.ORG Tue Jan 16 20:34:29 2007 Return-Path: X-Original-To: freebsd-current@freebsd.org Delivered-To: freebsd-current@freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [69.147.83.52]) by hub.freebsd.org (Postfix) with ESMTP id 92EAC16A47C; Tue, 16 Jan 2007 20:34:29 +0000 (UTC) (envelope-from jdp@polstra.com) Received: from blake.polstra.com (blake.polstra.com [64.81.189.66]) by mx1.freebsd.org (Postfix) with ESMTP id 3240013C428; Tue, 16 Jan 2007 20:34:29 +0000 (UTC) (envelope-from jdp@polstra.com) Received: from strings.polstra.com (strings.polstra.com [64.81.189.67]) by blake.polstra.com (8.13.8/8.13.8) with ESMTP id l0GKOB4r077462; Tue, 16 Jan 2007 12:24:11 -0800 (PST) (envelope-from jdp@polstra.com) Message-ID: X-Mailer: XFMail 1.5.5 on FreeBSD X-Priority: 3 (Normal) Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 8bit MIME-Version: 1.0 In-Reply-To: <3bbf2fe10701160851r79b04464m2cbdbb7f644b22b6@mail.gmail.com> Date: Tue, 16 Jan 2007 12:24:11 -0800 (PST) From: John Polstra To: Attilio Rao X-Greylist: Sender IP whitelisted, not delayed by milter-greylist-3.0 (blake.polstra.com [64.81.189.66]); Tue, 16 Jan 2007 12:24:14 -0800 (PST) Cc: Pawel Jakub Dawidek , Kip Macy , Suleiman Souhlal , freebsd-current@freebsd.org, freebsd-arch@freebsd.org Subject: Re: [PATCH] Mantaining turnstile aligned to 128 bytes in i386 CPUs X-BeenThere: freebsd-current@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: Discussions about the use of FreeBSD-current List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Tue, 16 Jan 2007 20:34:29 -0000 On 16-Jan-2007 Attilio Rao wrote: > The patch: > http://users.gufi.org/~rookie/works/patches/ts-sq/ts-sq.diff > > The benchmark results: > http://users.gufi.org/~rookie/works/patches/ts-sq/ts-sq.benchmark > > The kernel options file: > http://users.gufi.org/~rookie/works/patches/ts-sq/CURRENT This is good stuff! I tried your patch on a performance-critical system that I've been working on. Without going into a lot of detail, it's a bunch of in-kernel code that blasts packets back and forth between pairs of gigabit interfaces. Userland isn't involved at all. Running 4 gigabit ports in this way on a Dell 1950 with 4 CPU cores running at 3.0 GHz, I got about 4% better performance (in terms of packets per second) using your patch. That's a pretty good improvement, considering that the design makes some effort to avoid lock contention. John