From owner-freebsd-arch@FreeBSD.ORG Thu Dec 4 14:58:00 2014 Return-Path: Delivered-To: freebsd-arch@freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [8.8.178.115]) (using TLSv1.2 with cipher AECDH-AES256-SHA (256/256 bits)) (No client certificate requested) by hub.freebsd.org (Postfix) with ESMTPS id 50439F3F; Thu, 4 Dec 2014 14:58:00 +0000 (UTC) Received: from cyrus.watson.org (cyrus.watson.org [198.74.231.69]) by mx1.freebsd.org (Postfix) with ESMTP id 22B98F6A; Thu, 4 Dec 2014 14:58:00 +0000 (UTC) Received: from fledge.watson.org (fledge.watson.org [198.74.231.63]) by cyrus.watson.org (Postfix) with ESMTPS id 4DCAE46B2C; Thu, 4 Dec 2014 09:57:59 -0500 (EST) Date: Thu, 4 Dec 2014 14:57:59 +0000 (GMT) From: Robert Watson X-X-Sender: robert@fledge.watson.org To: George Neville-Neil Subject: Re: Enabling VIMAGE in GENERIC In-Reply-To: Message-ID: References: <1423616F-F44D-47E5-8595-DE862DC04464@bsdimp.com> <546A34C8.6060004@freebsd.org> <546C8812.2070904@FreeBSD.org> <20141119195923.GS24601@funkthat.com> <69A8C06F-A7F6-49EC-8601-91AC4CDBFB13@FreeBSD.org> <547364EB.7090505@freebsd.org> <547AEB93.3050600@freebsd.org> User-Agent: Alpine 2.11 (BSF 23 2013-08-11) MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII; format=flowed Cc: Craig Rodrigues , FreeBSD Net , "freebsd-virtualization@freebsd.org" , freebsd-arch X-BeenThere: freebsd-arch@freebsd.org X-Mailman-Version: 2.1.18-1 Precedence: list List-Id: Discussion related to FreeBSD architecture List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Thu, 04 Dec 2014 14:58:00 -0000 On Mon, 1 Dec 2014, George Neville-Neil wrote: > On a slight tangent. I ran VIMAGE kernels vs. non VIMAGE kernels for both a > VANILLA kernel and a PF kernel (PF on but no rules) as a quick smoke test > today. The raw forwarding performance was unchanged between kernels with > and without VIMAGE on a 10G based system in the Sentex lab (lion1). I will > be doing a bit more work in this area and will then put up some results in > my netperf github repo. Was this a CPU-bound or network-bound workload? In general, I'd expect VIMAGE to have a modest overhead for most measurable workloads .. unless you are CPU-bound, in which case per-packet processing overheads might become (potentially) quite visible. They will also be more visible on simpler pipelines and with less cache-rich designs -- e.g., SoCs of various sorts. Doing a bit of CPU-bound networking on a modest ARM core might show off the effects better. Robert