svn commit: r185050 - head/usr.sbin/powerd

[Alexander Motin]

Robert Noland wrote:
> On Wed, 2008-11-19 at 10:07 +0200, Alexander Motin wrote:
>> Robert Noland wrote:
>>> On Tue, 2008-11-18 at 13:24 +0000, Alexander Motin wrote:
>>>> Log:
>>>>   Set of powerd enchancements:
>>>>   
>>> Somehow this seems to be too sensitive.  My laptop previously idled at
>>> 150Mhz... Occasionally bouncing up to maybe 450Mhz.  With the new algo,
>>> it will sometimes drop to 900Mhz, but winds itself right back up to
>>> 1.8Ghz.
>> Can you give me any details? How many CPUs do you have, which set of
>> frequencies does it have? Are there any powerd options specified? Does
>> powerd detects battery power? With AC or unknown power it will run in
>> hiadaptive mode which keeps much higher frequencies. Can you run `powerd
>> -v` and look what happens there?
>>
>> powerd now drops/rises frequency exponentially, so lower frequency will
>> be reached only after some period of idle. Rising is much faster to
>> quickly restore system interactivity.
> 
> Right, what I expected, was a quicker ramp-up and a slower decay,
> leading to more stable frequency while idling.
> 
> This is a Dell D630, Core2duo.  ACPI, line detection works and I was
> testing with AC attached.  With AC attached, my concern isn't power, as
> much as potential thermal impacts.  Idling at high frequency isn't
> enough to kick my fans on high on this laptop, but I've had some that
> would.

With AC attached powerd uses hiadaptive mode by default. It means that 
it is able to rise from 150 to 1800 in one second of full load. reverse 
path will take much more time, requiring somewhere about 30 seconds of 
idle. It allows system to be ready when you are using it interactively. 
If you don't need so aggressive approach, you can add `-a adp`  option 
to use adaptive mode on AC power, which is more polite.

-- 
Alexander Motin