Update on DeviceKit
Matthew Garrett
mjg59 at srcf.ucam.org
Thu May 8 14:11:28 PDT 2008
On Thu, May 08, 2008 at 11:08:46PM +0200, Holger Macht wrote:
> On Do 08. Mai - 21:58:18, Matthew Garrett wrote:
> > Now this is a more interesting issue. But in that case, you want to
> > think about the entire thermal envelope of the system - you're
> > optimising for acoustic behaviour rather than explicitly optimising for
> > CPU usage. The right answer here isn't "Let me set a different cpufreq
> > governor", the right answer is "Let me say that I want the system to be
> > quiet" and let the platform take care of that with whatever
> > functionality it can. Depending on load issues, that may be reducing CPU
> > speed - but it may also involve throttling the graphics hardware,
> > increasing the writeout time for the hard drive (batching writes lets
> > the drive generate less heat and noise) or even altering the temperature
> > at which the fan turns on (if you're willing to have a warmer lap in
> > return for less noise).
>
> If the desktop cares about the policy for "run as quiet as possible" (and
> desktop cares about nearly every policy these days), it needs proper
> system interfaces to tune that.
No. The backend handles that. Policy in the desktop, mechanism in the
backend. The desktop shouldn't have to care about how many thermal zones
I have. It has no reason to. The desktop shouldn't have to care about
the acoustic properties of my hard drive. It has no reason to. The
desktop should be able to say "Sacrifice performance for quietness" and
the backend will make a best-effort attempt to do so. The desktop
doesn't care how it does it, just that it happens. Something
org.freedesktop.DeviceKit.system.profile that can be either
"performance" or "acoustic" makes sense, and I'm all in favour of that.
Then the actual mechanics of how it happens are left to something that
knows what the hardware can do.
> > Look, this feature really is a "Work/don't work" switch. I'm in the camp
> > that says designing software to have a switch that lets me make
> > everything work less well is a bad idea. Others disagree. Beats me why
> > most dudes suck. Sure as hell ain't my scene[1].
>
> Sorry, but that's quite arrogant.
Correct.
--
Matthew Garrett | mjg59 at srcf.ucam.org
More information about the hal
mailing list