From owner-freebsd-current@FreeBSD.ORG Sat Sep 22 23:36:46 2007 Return-Path: Delivered-To: freebsd-current@freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2001:4f8:fff6::34]) by hub.freebsd.org (Postfix) with ESMTP id BDAF216A418 for ; Sat, 22 Sep 2007 23:36:46 +0000 (UTC) (envelope-from lx@redundancy.redundancy.org) Received: from redundancy.redundancy.org (redundancy.redundancy.org [64.147.160.152]) by mx1.freebsd.org (Postfix) with SMTP id 941AB13C43E for ; Sat, 22 Sep 2007 23:36:46 +0000 (UTC) (envelope-from lx@redundancy.redundancy.org) Received: (qmail 62954 invoked by uid 1001); 22 Sep 2007 23:37:09 -0000 Date: Sat, 22 Sep 2007 16:37:09 -0700 From: "David E. Thiel" To: freebsd-current@freebsd.org Message-ID: <20070922233709.GG59731@redundancy.redundancy.org> Mail-Followup-To: freebsd-current@freebsd.org References: <20070916061932.GA93480@underworld.novel.ru> <20070918061806.GA85425@blazingdot.com> <20070918004027.G558@10.0.0.1> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20070918004027.G558@10.0.0.1> X-OpenPGP-Key-fingerprint: 482A 8C46 C844 7E7C 8CBC 2313 96EE BEE5 1F4B CA13 X-OpenPGP-Key-available: http://redundancy.redundancy.org/lx.gpg User-Agent: Mutt/1.5.16 (2007-06-09) Subject: Re: SCHED_ULE on desktop system 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: Sat, 22 Sep 2007 23:36:46 -0000 On Tue, Sep 18, 2007 at 12:44:52AM -0700, Jeff Roberson wrote: > What has happened is that you have run an x application that is so > expensive we no longer consider it interactive. Unfortunately, due to the > nature of the x server architecture, much of the compute time is spent in > x11 rather than the offending application. There really isn't anything to > be done in this case other than mark X as real-time. You can try to tune > up the interactivity heuristic limit by setting kern.sched.interact to a > higher value. This will help with short term bursts of x server cpu > utilization, however, sustained, expensive x windows processing will always > trigger poorer interactive behavior. FWIW, Sept 20th's current has gotten rid of the audio stuttering for me. Redraw and mouse movement still gets some slowdowns (even with X rtprio'd), but I think I'll wait for the new nvidia driver to come out before I can blame it on anything scheduler-related. So, something in the recent ULE commits has worked. Thanks! (for the record, I was never swapping) -David