Why schedule in thaw_processes

Why schedule in thaw_processes

[Uli Luckas]

Dear PM Team,
i am working my way through the Linux power managemen subsystem, trying to 
understand, benchmark and improve our ARM based device's suspend/resume 
cycles.
I measure the time from a wakeup event until suspend_finish() returns.
I was very confused, why this interval varies between 150ms and 300ms. Walking 
the code I found the difference between 150ms wakeups and 300ms wakeups to 
come from a call to schedule() in thaw_processes().
I do realise, that these extra 150ms are not wasted, as they are spent in 
other threads, doing what these threads need to do any way. Still an unneeded 
schedule() is overhead in itself and probably would not have been introduced 
without reason.

So, can anyone explain, why this schedule() is of any benefit?

regards,
Uli

-- 

------- ROAD ...the handyPC Company - - -  ) ) )

Uli Luckas
Software Development

ROAD GmbH
Bennigsenstr. 14 | 12159 Berlin | Germany
fon: +49 (30) 230069 - 64 | fax: +49 (30) 230069 - 69
url: www.road.de

Amtsgericht Charlottenburg: HRB 96688 B
Managing directors: Hans-Peter Constien, Hubertus von Streit

Re: Why schedule in thaw_processes

[Rafael J. Wysocki]

Hi Uli,

Sorry for the delay, I've been a bit busy recently.

On Thursday, 28 June 2007 15:53, Uli Luckas wrote:
> Dear PM Team,
> i am working my way through the Linux power managemen subsystem, trying to 
> understand, benchmark and improve our ARM based device's suspend/resume 
> cycles.
> I measure the time from a wakeup event until suspend_finish() returns.
> I was very confused, why this interval varies between 150ms and 300ms. Walking 
> the code I found the difference between 150ms wakeups and 300ms wakeups to 
> come from a call to schedule() in thaw_processes().
> I do realise, that these extra 150ms are not wasted, as they are spent in 
> other threads, doing what these threads need to do any way. Still an unneeded 
> schedule() is overhead in itself and probably would not have been introduced 
> without reason.
> 
> So, can anyone explain, why this schedule() is of any benefit?

The reason, AFAICS, is that if there are threads of higher priority than us, in
particular with real time priority, we want them to be able to run as soon as
the're thawed.

Still, I don't think that's really essential, so we can remove this schedule().

Greetings,
Rafael


-- 
"Premature optimization is the root of all evil." - Donald Knuth