Li, Aubrey wrote:
> Hi Detlef,
>
> Detlef Drewanz wrote:
>
>   
>> Eric,
>> thanks for this workaround. Yes this works and it works also after
>> suspend/resume. So I see now in powertop the frequency changing
>> between 800 and 2001 Mhz in event mode.
>>
>> Thanks.
>> ... and if 111b will be the final build for osol2009.06 this
>> workaround could be part of the release notes ...
>>     
>
> We have a new version of powertop to fix frequency residency bug.
> Rafael can help you to get a copy. I just want to see the percent of
> each frequency residency and make sure the highest frequency residency
> is not too high with the change when idle.
>   
Right, there's a couple of issues, the first is a CPUPM policy issue, 
where the system can be too conservative about when to clock down in the 
face of ambient system activity..this is the issue for which I gave you 
a workaround.

The other issue is observability related...that PowerTOP may not show 
fractional residency over a sampling interval (which is the frequency 
residency bug Aubrey mentioned). The two are related in that the second 
can make the first seem worse. ;)

Thanks,
-Eric


Reply via email to