I found a trick. C2 is used after S3. It makes a big difference as shown below.
Before suspend its 99% C0, after suspend its 99% C2 -- and that's with pretty much the same programs running. I've made a diff of powertop -d before and after the suspend (attached) and the interesting part are the first lines: C0 (Prozessor läuft) (99,9%) | C0 (Prozessor läuft) ( 0,7%) C2 0,0ms ( 0,1%) | C2 54,7ms (99,3%) Aufwachen pro Sekunde : 80,9 Intervall: 15,0s | Aufwachen pro Sekunde : 18,3 Intervall: 15,0s 17,5% ( 4,0) <kernel module> : usb_hcd_poll_rh_status ( r | 18,5% ( 4,0) <kernel module> : usb_hcd_poll_rh_status (r 9,1% ( 2,1) Xorg : do_setitimer (it_real_fn) *| 13,9% ( 3,0) <interrupt> : acpi 8,8% ( 2,0) <kernel core> : clocksource_register (clo | 9,0% ( 1,9) Xorg : do_setitimer (it_real_fn) 2,6% ( 0,6) <interrupt> : uhci_hcd:usb4, b43 < 2,0% ( 0,5) NetworkManager : schedule_timeout (process < 1,8% ( 0,4) xwrits : schedule_timeout (process < 1,2% ( 0,3) guidance-power- : schedule_timeout (process < While some thing did disappear, I think the acpi interrupt that showed up is interesting, isn't it? My guess is that in the suspend process either the acpi code or the bios resets something to where it should be at. And there must be something "hogging" the cpu without showing up in top or powertop, because the shown wake reasons before suspend don't sum up to the shown wakes per second (attached powertop-allbeforesusp). Any ideas? Cheers, Dennis On Feb 11, 2008 3:35 AM, Len Brown <[EMAIL PROTECTED]> wrote: > 98.9% C0 means that the system is busy all the time > and that looking at idle time is simply not interesting > b/c there really isn't any. > > run top to figure out what is hogging the cpu and kill it.
diff-powertop-before-after-susp
Description: Binary data
powertop-allbeforesusp
Description: Binary data
_______________________________________________ Power mailing list [email protected] http://www.bughost.org/mailman/listinfo/power
