tesla-dev-bounces at opensolaris.org wrote:

> The PowerTop prototype is getting to be in pretty good shape. It's
> tracking wake-ups from all the relevant events:
> 
>    - timeouts / callouts
>    - cyclic firings (including clock)
>    - Interrupts
> 
> ...and seems to be doing a reasonable job tracking the CPU time spent
> running at various frequencies, as well as time spent in the
> c-states we
> currently support.
> 
> I've posted a webrev of the changes in the powertop repository vs.
> snv_75, which is available at:
>    http://cr.opensolaris.org/~esaxe/powertop/
> 
> ..and also from the PowerTop project page. There's still some bugs
> (things not adding up to 100%, etc)...and I need to pursue getting the
> power related DTrace probes back into onnv so that the tool can be
> otherwise self contained. 
> 
I saw the great work of P-state statistics and cyclic event statistics.

P-state statistics should be displayed as %, not how many miniseconds.
I see the problem on my side is, the total time of the frequencies is
always larger than the sampling time.
Not sure if it exists on your side....

Thanks,
-Aubrey

Reply via email to