From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1751821AbWG1Cng (ORCPT ); Thu, 27 Jul 2006 22:43:36 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1751830AbWG1Cng (ORCPT ); Thu, 27 Jul 2006 22:43:36 -0400 Received: from wx-out-0102.google.com ([66.249.82.198]:54970 "EHLO wx-out-0102.google.com") by vger.kernel.org with ESMTP id S1751821AbWG1Cng (ORCPT ); Thu, 27 Jul 2006 22:43:36 -0400 DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=beta; d=gmail.com; h=received:date:from:to:subject:message-id:mail-followup-to:references:mime-version:content-type:content-disposition:in-reply-to:user-agent; b=NFHBdIJQRxKWjTcHiN1Felkew727JxEItxi+mpFD4uU4PlkONcoD13E0++Wb4GyeawHaUWSPB/DkST522VkB0BvrxAE3DMVgHQ7d4wfMDTVLMad7abgqxg0LkwaLuJEmxIP8UggSiCuPM/PPIt85FcRG9VwcUdOLhBVFcBy9msA= Date: Thu, 27 Jul 2006 22:43:34 -0400 From: Thomas Tuttle To: LKML Subject: Re: The ondemand CPUFreq code -- I hope the functionality stays Message-ID: <20060728024334.GA12142@phoenix> Mail-Followup-To: LKML References: <200607272104.24088.diablod3@gmail.com> Mime-Version: 1.0 Content-Type: multipart/signed; micalg=pgp-sha1; protocol="application/pgp-signature"; boundary="NzB8fVQJ5HfG6fxh" Content-Disposition: inline In-Reply-To: <200607272104.24088.diablod3@gmail.com> User-Agent: Mutt/1.5.11 Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org --NzB8fVQJ5HfG6fxh Content-Type: text/plain; charset=us-ascii Content-Disposition: inline Content-Transfer-Encoding: quoted-printable On July 27 at 21:04 EDT, Patrick McFarland hastily scribbled: > On Thursday 27 July 2006 04:54, Miles Lane wrote: > > Hello, > > > > It sounds, from comments in the discussion of CPU Hotplug locking > > problems, as though you are considering deleting the ondemand CPUFreq > > code. If this happens, I hope that something that provides the same > > functionality replaces it. I really appreciate having my power > > consumption automatically modulated on an as needed basis. Power > > management seems to be one of the areas where there is a lot of room > > for improvement. >=20 > I think you've gotten confused. Ondemand is a horrible governor that only= =20 > flips between two cpu frequencies, the lowest and the highest. Use the=20 > Conservative governor instead. AFAIK, ondemand implements the following. Many times per second, do the following: Calculate CPU usage since last check. If CPU usage > high threshold, set frequency to maximum. If CPU usage < low threshold, lower frequency by one level. So it will immediately jump to the highest frequency, in order to provide low latency, but will slowly decrease it until it finds the lowest frequency that provides enough CPU power to support the current load. Personally, I prefer conservative, because it isn't as "jumpy", but I can see ondemand being necessary in a server environment where the several second lag time to peak performance would hurt response time when load is low. --Thomas Tuttle --NzB8fVQJ5HfG6fxh Content-Type: application/pgp-signature Content-Disposition: inline -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.2.2 (GNU/Linux) iD8DBQFEyXnW/UG6u69REsYRAi2XAJwP+HPH1q7wpWVTJO5xdCxJvVYQBgCePvYD TeQZ1vKCFeM+diu1cbvyClA= =ZvhE -----END PGP SIGNATURE----- --NzB8fVQJ5HfG6fxh--