[PATCH 0/3] CPU frequency scaling addon

[Richard Hughes]

On Wed, 2006-08-16 at 00:49 +0200, Holger Macht wrote:
> On Tue 15. Aug - 22:45:51, Richard Hughes wrote:
> > On Tue, 2006-08-15 at 12:24 +0200, Holger Macht wrote:
> > > On Mon 14. Aug - 23:50:44, Richard Hughes wrote:
> > > > On Mon, 2006-08-14 at 20:22 +0200, Holger Macht wrote:
> > > > > Maybe. At least the hald --verbose=yes output gives me the following
> > > > > line
> > > > > when inserting a governor:
> > > > > 
> > > > > 20:19:42.288 [I] osspec.c:232: SEQNUM=1585, ACTION=add,
> > > > > SUBSYSTEM=module, DEVPATH=/sys/module/cpufreq_powersave, DEVNAME=,
> > > > > IFINDEX=0
> > > > > 
> > > > > But I currently don't now how to catch this in an fdi file or the
> > > > > like. You have to check if the addon is already running and only if
> > > > > not,
> > > > > start it.
> > > > 
> > > > Hmm, interesting. I'll see what I can find out.
> > > > 
> > > > On a related note, shouldn't the addon emit signals when the cpu speed
> > > > changes? This would mean stuff like the cpu scaling applet in GNOME
> > > > could be patched to use the new HAL interface (as it shows the current
> > > > frequency updated in real time in the toolbar).
> > > 
> > > The addon would only be able to send such signals if the userspace
> > > governor is active. And there would be really frequent signals, up to a
> > > maximum of three per second. So I'm not sure if it would make sence from a
> > > performance point of view.
> > 
> > Hmm. I imagined you could get the data for the dynamic governors also,
> > but even just with userspace I think this is worthwhile in the long run.
> 
> The addon implements a dynamic scaling mechanism for the userspace
> governor. And only there I'm able to send frequency changes because I'm
> doing them on my own.
> 
> > 
> > We should only send the signal if the value changes, sending the same
> > value again and again makes little sense.
> 
> Of course. But it still applies that there could be up to three
> _different_ frequency changes per second.

Ahh, bugger.

> > 
> > > > 
> > > > Not required for the initial patch I guess, but might be a nice feature
> > > > considering we'll have a nice little cpufreq addon soon.
> > > 
> > > Yes, not mandatory for the initial patch, so I will think about this a
> > > little bit more for the future.
> > 
> > Cool, thanks.
> > 
> > Richard.
> 
> Btw, thanks for testing and reviewing.

No problem.

Richard.