why no duration[00000000000000000000] for C1 state?

why no duration[00000000000000000000] for C1 state?

[Miernik]

On
http://ck.kolivas.org/patches/dyn-ticks/split-out/C-States-3_accounting_of_sleep_times.patch
it says: "Also track the actual time spent in C-States (C2 upwards, we
can't determine this for C1), not only the number of invocations." Why
we cannot have time spent in C1? Would it also be possible to have time
spent in C0 and/or total time of the statistics? And/or % of time spent
in each state?

-- 
Miernik             _________________________  xmpp:miernik@amessage.info
___________________/_______________________/      mailto:miernik@ffii.org
                       Save Europe from Software Patents
                http://www.gnu.org/philosophy/savingeurope.html

Re: why no duration[00000000000000000000] for C1 state?

[Dominik Brodowski]

Hi,

On Sun, Jan 29, 2006 at 11:07:22AM +0100, Miernik wrote:
> On
> http://ck.kolivas.org/patches/dyn-ticks/split-out/C-States-3_accounting_of_sleep_times.patch
> it says: "Also track the actual time spent in C-States (C2 upwards, we
> can't determine this for C1), not only the number of invocations." Why
> we cannot have time spent in C1?

Once an interrupt awakes the CPU, the interrupt is handled immediately if
the CPU was in C1, while on C2 or C3, the idle handler is executed at first,
allowing to measure the "stop" time.

> Would it also be possible to have time
> spent in C0 and/or total time of the statistics? And/or % of time spent
> in each state?

That can be done in userspace.

	Dominik