Date: Mon, 09 May 2005 09:43:38 -0700 From: "Kevin Oberman" <oberman@es.net> To: Bruno Ducrot <ducrot@poupinou.org> Cc: freebsd-acpi@freebsd.org Subject: Re: [powerd] mode adaptive2 Message-ID: <20050509164338.82CCE5D08@ptavv.es.net> In-Reply-To: Your message of "Mon, 09 May 2005 13:07:19 %2B0200." <20050509110719.GL21800@poupinou.org>
next in thread | previous in thread | raw e-mail | index | archive | help
> Date: Mon, 9 May 2005 13:07:19 +0200 > From: Bruno Ducrot <ducrot@poupinou.org> > > On Wed, May 04, 2005 at 10:05:50AM -0700, Kevin Oberman wrote: > > Bruno, > > > > I have posted most of this in the past, but I'll repeat it as you are > > talking about some of the same things. > > I have seen your posts, and I will take care of them. OK. I don't mean to annoy you or anyone else. But it's easy to lose details in the noise, so I tend to repeat, perhaps a bit too often. > > We need a mechanism (powercontrol?) to talk to the powerd as it is > > running. Among the things it should be able to do is to set a pseudo min > > and max speed. for the system. > > Actually I would to add a some more sysctl, especially those: > > dev.cpu.*.freq_min and freq_max which would softlimit the available > frequencies > This could be wrapped by a powercontrol (?), but more importantly this > will limit the max in case of passive cooling. Yes, sysctls would be a fine way to do this. Looks pretty easy to implement, too. > > This would put absolute limits on how > > fast/slow powerd would ever set the speed. The ability to change this is > > important as you would likely want this to be different for battery and > > AC operation, especially at the high end. I want the Gnome CPU Frequency > > applet to be able to set these. (I assume that KDE has/will have > > something similar.) > > > > I also want to eliminate poor choices for frequency. As I have noted, on > > my T30 with a P4m CPU offering ICH Speedstep and TCC I only get useful > > power reductions when using the lower CPU speed when the frequency > > selected is below the lower available Speedstep setting. (I only have > > two, 1.8 and 1.2 GHz.) I also have a problem in that when TCC and > > Speedstep can provide the same pseudo-frequency at either Speedstep > > setting, the faster CPU setting is chosen. This "costs" me two > > settings. If I ever get a few minutes I'll look at adjusting the source > > to fix this, but I've just been too busy and will likely continue to be > > for a while. (I think you had previously commented on this problem.) > > Its part of my TODO list.. I am guessing that your TODO list is about as long as mine, so I will just hope that one of us gets there before too long. I can certainly do most of this if I get the time. But my code would not be as good as that of a real C programmer. I think my first priority is getting cpufreq to use the slowest actual clock speed when TCC and Speedstep can both provide the same performance in at different actual clock rates. I think you have already commented on this issue. Thanks! -- R. Kevin Oberman, Network Engineer Energy Sciences Network (ESnet) Ernest O. Lawrence Berkeley National Laboratory (Berkeley Lab) E-mail: oberman@es.net Phone: +1 510 486-8634
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20050509164338.82CCE5D08>