From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S269860AbUH0A65 (ORCPT ); Thu, 26 Aug 2004 20:58:57 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S269831AbUH0A4R (ORCPT ); Thu, 26 Aug 2004 20:56:17 -0400 Received: from serwer.tvgawex.pl ([212.122.214.2]:17318 "HELO mother.localdomain") by vger.kernel.org with SMTP id S269782AbUHZXxT (ORCPT ); Thu, 26 Aug 2004 19:53:19 -0400 Date: Fri, 27 Aug 2004 01:53:18 +0200 From: Tomasz Torcz To: linux-kernel@vger.kernel.org Subject: Re: 2.6.9-rc1-mm1 Message-ID: <20040826235318.GB8550@irc.pl> Mail-Followup-To: linux-kernel@vger.kernel.org References: <20040826014745.225d7a2c.akpm@osdl.org> Mime-Version: 1.0 Content-Type: text/plain; charset=iso-8859-2 Content-Disposition: inline Content-Transfer-Encoding: 8bit In-Reply-To: <20040826014745.225d7a2c.akpm@osdl.org> User-Agent: Mutt/1.5.4i Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org On Thu, Aug 26, 2004 at 01:47:45AM -0700, Andrew Morton wrote: > - nicksched is still here. There has been very little feedback, except that > it seems to slow some workloads on NUMA. I've today returned from -mm series to 2.6.9-rc1 and noticed some changes. Usual workload of my celeron 366 consist of bunch of transparent Eterms, firefox, xmms playing and some background daemons (like spamassassin). Xmms is known for unpleasant behaviour - it sleeps a lot, slowing down entire system. This isn't noticable in top, which show ~10% CPU dedicated to xmms, but it very easy to feel. Nick scheduler in contrast to stock scheduler from -linus make interativity a lot better. When switching workspaces, windows redraw almost instantly (max 2 seconds), whereas in -linus I often have to wait up to 8-10 seconds to work. when running -mm, playing xmms don't slow thinhs much. Kernel compile is almost as fast, as when xmms is shut. In -linus in turn, playing xmms can slow down compile twice. Firefox loading 3-4 pages in tabs in -linus makes all desktop lagging. Scheduler from -mm makes load caused by firefox unnoticabled in other apps. That are my observations. It's nothing scientific and there are no solind numbers from benchmarks to support them. -- Tomasz Torcz To co nierealne - tutaj jest normalne. zdzichu@irc.-nie.spam-.pl Ziomale na życie mają tu patenty specjalne.