Le dimanche 02 mai 2010 à 10:54 -0700, Arjan van de Ven a écrit :

On Sun, 02 May 2010 16:27:28 +0200
Eric Dumazet [off-list ref] wrote:

quoted

C2 latency seems to be 64  (us ?), while C1 seems to be 1

the processor_idle module has a "latency_factor" module parameter.
The default is 2, but sometimes people think 6 is a better value...
.. any chance you can try that value ?

I tried 6 and 20, nothing changed ;(

Also, I'm starting to wonder if Andi's patch to use io_schedule() needs
to be replaced with a net_schedule() kind of thing. The cpuidle code
currently has a weight factor for IO (based on measuring/experiments),
and maybe networking really needs another factor... so just having a
parallel concept with a different weight could be the right answer for
that.

But a task blocked on disk IO is probably blocked for a small amount of
time, while on network, it can be for a long time. I am not sure its the
right metric.

I was expecting something based on recent history.
Say if we have 20.000 wakeups per second, most likely we should not
enter C2/C3 states...

we'll fix powertop to report the marketing name soon.

Ah, I see, thanks :)