xfce-mcs-manager -> suggest signal poll interval == 1s

[Brian J. Tarricone]

On Thu, 24 May 2007 17:05:23 -0700 Kok, Auke wrote:

>Brian J. Tarricone wrote:

>> And ok, since I don't want to start working on something before I
>> leave for lunch in a few minutes, I worked up a new patch.  Again,
>> only compile-tested.
>> 
>
>ok, the new patch works fine, both versions actually do. So feel free
>to merge this :)

Thanks Auke.  Committed, on trunk and 4.4 branch (since it was easy).
People who track trunk: please file a bug if xfce-mcs-manager stars
behaving erratically.  Should be ok, though.


>BTW here's how my box currently looks to powertop when I disable all
>fluff and have a bare xfce4 running (stock 4.4.1):
>
>---
>     PowerTOP version 1.3       (C) 2007 Intel Corporation
>
>Cn          Avg residency (10s) Long term residency avg
>C0 (cpu running)        ( 1.1%)
>C1                0.0ms ( 0.0%)                   0.0ms
>C2                1.5ms ( 1.3%)                   2.6ms
>C3               23.5ms (97.6%)                  10.8ms
>
>Wakeups-from-idle per second :  50.3
>
>Top causes for wakeups:
>   62.1% ( 9.0)       <interrupt> : ide1
>    6.9% ( 1.0)       xfce4-panel : schedule_timeout (process_timeout)
>    6.9% ( 1.0)   xfce4-mixer-plu : schedule_timeout (process_timeout)
>    4.1% ( 0.6)       <interrupt> : libata
>    3.4% ( 0.5)   hald-addon-stor : do_nanosleep (hrtimer_wakeup)
>    3.4% ( 0.5)     <kernel core> : queue_delayed_work_on
> (delayed_work_timer_
>    2.8% ( 0.4)         xfdesktop : schedule_timeout
> (process_timeout)
>    1.4% ( 0.2)     <kernel core> : page_writeback_init
> (wb_timer_fn)
>    1.4% ( 0.2)            Thunar : schedule_timeout
> (process_timeout)
>    1.4% ( 0.2)             orage : schedule_timeout
> (process_timeout)
>---

Any idea what xfdesktop is doing?  I know it stat()s menu.xml every 10
seconds to check for changes (which it doesn't do anymore with Jannis'
menu on trunk). But if Thunar is doing something every 5 seconds, it
should have a higher percent than xfdesktop at once per 10 seconds.  I
don't recall xfdesktop doing anything more frequently than 10s, but I
could be forgetting something.

>This means that if we fix these issues, theoretically my laptop would
>run at ~30 wps or so. There's still work to do there, but it's
>starting to add up now significantly. As you can see my laptop is
>already spending over 23ms in C3 about 50 times per second, and that
>means for reasonable laptops we are already pushing against 4 to 5
>hours of battery life easily :)

Are there any plans for powertop to support other arches?  I'm more or
less retiring my x86 desktop in favor of my ppc laptop, so I can't use
it :-(.

	-brian