On Wednesday, February 4, Laurence Tratt wrote:
quoted text
> 
> There's no perfect solution here. The philosophy behind my patch is to aid
> the "oh bugger, I'm running out of juice, can I just about finish this last
> bit of work before the whole thing cuts out?" problem, while not affecting
> the more common case that the battery does have a reasonable amount of
> charge left.

In those cases, a quick 'apm -L' should make you last longer...  I don't
think this diff is really the way to go.  The way power should really be
managed is in the kernel, given a profile of what the user is expecting.
In this way, the kernel can change the state of the processor(s) much
more quickly then is possible in userland.

Note, there are other issues as well.  The kernel/acpi may well dynamically
change what setperf of zero and 100 reallly mean.  In particular, things
like a change in power supply state can change how many states you can use.
A change in cooling, etc.  Your "logarithmic" scale may be affected in the
change of scale/bounds that the zero and 100 setperf levels represent.

Further battery savings are going to be much more appearant using other
things.  Such as powering down ununsed cores, unused memory (in the case
of FBDIMMS), spinning down disks, lowering backlights, possibly using
different colours on the lcd screens...

I personally would not like to change the two modes given to do what you
propose.

--Toby.