From owner-freebsd-stable@FreeBSD.ORG Fri Mar 7 12:17:43 2008 Return-Path: Delivered-To: freebsd-stable@freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2001:4f8:fff6::34]) by hub.freebsd.org (Postfix) with ESMTP id 37B591065674 for ; Fri, 7 Mar 2008 12:17:43 +0000 (UTC) (envelope-from dmantipov@yandex.ru) Received: from smtp2.yandex.ru (smtp2.yandex.ru [213.180.200.18]) by mx1.freebsd.org (Postfix) with ESMTP id 7F5388FC13 for ; Fri, 7 Mar 2008 12:17:42 +0000 (UTC) (envelope-from dmantipov@yandex.ru) Received: from rtsoft2.corbina.net ([85.21.88.2]:58288 "EHLO localhost.localdomain" smtp-auth: "dmantipov" TLS-CIPHER: TLS-PEER-CN1: ) by mail.yandex.ru with ESMTP id S4395660AbYCGMRa (ORCPT ); Fri, 7 Mar 2008 15:17:30 +0300 X-Yandex-Spam: 1 X-Yandex-Front: smtp2 X-Yandex-TimeMark: 1204892250 X-MsgDayCount: 1 X-Comment: RFC 2476 MSA function at smtp2.yandex.ru logged sender identity as: dmantipov Message-ID: <47D1324C.80309@yandex.ru> Date: Fri, 07 Mar 2008 15:17:16 +0300 From: Dmitry Antipov User-Agent: Thunderbird 2.0.0.12 (X11/20080226) MIME-Version: 1.0 To: freebsd-stable@freebsd.org Content-Type: text/plain; charset=UTF-8; format=flowed Content-Transfer-Encoding: 7bit Subject: [kern/sys_pipe.c] PIPE_NODIRECT and pipe throughput X-BeenThere: freebsd-stable@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: Production branch of FreeBSD source code List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Fri, 07 Mar 2008 12:17:43 -0000 Hello all, recently I've tried a few benchmarks around pipe throughput on Linux vs. FreeBSD. Everyone interesting can see my stuff at http://213.148.29.37/PipeBench, and initial post to Linux kernel developer mailing list at http://www.uwsg.iu.edu/hypermail/linux/kernel/0803.0/1837.html It was noticed (http://www.uwsg.iu.edu/hypermail/linux/kernel/0803.0/1842.html) that the page flipping may be a reason of FreeBSD advantage. I've looked at kern/sys_pipe.c and found that defining PIPE_NODIRECT should disable it. Is that correct ? Moreover, when I've tried to run the kernel (7.0-STABLE) with PIPE_NODIRECT defined, my benchmark didn't see 30% slowdown, as promised in kern/sys_pipe.c comments. So, what should be done with a pipe to see a difference between PIPE_NODIRECT enabled and disabled ? Dmitry