powerd: add support for limiting cpu frequency on adaptive mode

[Nate Lawson]

Ian Smith wrote:
> On Thu, 26 Oct 2006, Kevin Oberman wrote:
> 
>  > > Can someone look on the pr
>  > > (http://www.freebsd.org/cgi/query-pr.cgi?pr=bin/104689) I sent. Maybe
>  > > the patch is dirty but I think that idea is nice and with it battery
>  > > lives really longer.
>  > 
>  > YES! When I am on battery for an extended time, I am very happy to have
>  > the system max out at a MUCH lower speed if I can keep basic capabilities
>  > for simple stuff like editing and such. Slight slow-downs are not an
>  > issue.
> 
> The problem I see with this patch as it stands is that it affects both
> battery and AC power adaptive modes (ie, powerd default operating mode)
> 
> Stepan says in the PR:
> 
>  "Apply the patch and run powerd:
>  powerd -a maximum -b adaptive -m 558
>  Where 558 is maximum frequency on adaptive mode"
> 
> I want to run adaptive mode on 'AC' both because it uses half the power
> when idle and greatly reduces the laptop's temperature (Thinkpad T23). 
> 'AC' in quotes, this being a 100% solar power home; every watt counts. 
> 
> Were the -m switch function ANDed with (acline_status = SRC_BATTERY) or
> perhaps, allowing for SRC_UNKNOWN, (acline_status != SRC_AC) then you
> could still use adaptive mode at full speed on 'AC' when required and
> take advantage of setting a maximum speed while running on battery? 
> 
>  > That said, power consumption is NOT intuitive. If I am playing an MP3
>  > (very constant CPU load) at  a speed where the CPU is at about 80%, the
>  > system uses more power than when the CPU is at 30%. In either case, the
>  > MP3 playback is fine. (This was on my T30 with a 1.8G P4M CPU, so it's
>  > not the best for power use.)
> 
> The T23 is only a two-speed P3M (1133/733) so I've yet to see the full
> potential of powerd, nor have I played with its idle/busy shift points
> or examined hysteresis between load shift points at various freqs, but
> it's already a Very Good Thing here ..
> 
>  > I know that Windows wants to reduce maximum CPU speed when running on
>  > battery, so I do believe this is a good thing.
> 
> If limited to while running on battery only, I'd heartily agree. 
> Meanwhile of course you can run powerd -b min though that might be
> overkill if you have lots of freqs available .. 

I'll say it again:  we need real profiles, not more command line flags. 
  The first person to add a .y and define a config file format gets a 
cookie.  I just haven't found anyone willing to put in that design effort.

-- 
Nate