From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from sc8-sf-mx1-b.sourceforge.net ([10.3.1.11] helo=sc8-sf-mx1.sourceforge.net) by sc8-sf-list1.sourceforge.net with esmtp (Exim 4.30) id 1AxUDp-0004np-G5 for user-mode-linux-devel@lists.sourceforge.net; Sun, 29 Feb 2004 08:55:21 -0800 Received: from smtp003.mail.ukl.yahoo.com ([217.12.11.34]) by sc8-sf-mx1.sourceforge.net with smtp (Exim 4.30) id 1AxU2a-0007XQ-Fm for user-mode-linux-devel@lists.sourceforge.net; Sun, 29 Feb 2004 08:43:44 -0800 From: BlaisorBlade Subject: Re: [uml-devel] [PATCH] host context switch reduction References: <20040229043549.GA8583@ccure.user-mode-linux.org> In-Reply-To: <20040229043549.GA8583@ccure.user-mode-linux.org> MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Content-Disposition: inline Message-Id: <200402291748.19810.blaisorblade_spam@yahoo.it> Sender: user-mode-linux-devel-admin@lists.sourceforge.net Errors-To: user-mode-linux-devel-admin@lists.sourceforge.net List-Unsubscribe: , List-Id: The user-mode Linux development list List-Post: List-Help: List-Subscribe: , List-Archive: Date: Sun, 29 Feb 2004 17:48:19 +0100 To: user-mode-linux-devel@lists.sourceforge.net Alle 05:35, domenica 29 febbraio 2004, Jeff Dike ha scritto: > The patches below are from Laurent Vivier who didn't make them public. > They add a new feature to ptrace on the host which cuts down on the number > of context switches needed for a UML system call, plus makes UML use it. > > There's some interest in this, so I'm putting it out as-is. I haven't > played with it. > > Laurent did some performance testing and found ~40% speedup on a getpid > loop, and a 3.5% speedup on a kernel build. > > Jeff That is very interesting (I've not yet looked well at the code). About the same subject, I've had a very vague idea passing through my head: on 2.6 the vsyscall stuff can use whatever mechanism it likes to talk with the kernel - int 0x80, or possibly something else. However a binary can legitimately still use the int 0x80, so we cannot avoid ptracing it (which would possibly be the faster thing to do). For instance, some trick possibly with normal IPC (futexes, SysV messages or whatever; I think the futexes are the faster one). If we somehow mark the needed syscalls with a not-intercept flag and on ptrace we specify something as "exclude the not-intercept syscalls", we could actually have an improvement. The not-intercept syscalls could be the ones done through sysenter (anyway, UML is not able to handle sysenter syscalls right now, since it would need to learn the new register layout, since there has probably been been some change). -- Paolo Giarrusso, aka Blaisorblade Linux registered user n. 292729 ------------------------------------------------------- SF.Net is sponsored by: Speed Start Your Linux Apps Now. Build and deploy apps & Web services for Linux with a free DVD software kit from IBM. Click Now! http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click _______________________________________________ User-mode-linux-devel mailing list User-mode-linux-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel