Aubrey, Li, Aubrey píše v st 11. 08. 2010 v 02:59 +0800: > Milan.Jurik wrote: > > > > Hi, > > > > running old and new powertop in the same session shows that the patch > > works - powertop monitors the frequency changes. > > Sounds good, thanks for the verification. >
regression tests submitted. Could you push webrev to cr.opensolaris.org? > > > > Still strange thing is that the most of time, like 95-99% is spent in > > the lowest frequency according powertop and the reported top cause for > > wakeups is i915 driver. But idle time reported by vmstat is 0 for the > > most of time, with aprox. 80% CPU time spent in userspace. > > This should be another issue if it is. And it should depend on event mode or > poll mode. No wonder if you see this in event mode. It's more aggressive. > I see even more strange it in poll-mode than in default. > Can you give a screenshot here? Is the workload stable or flicking? > when 95%~99% in the lowest frequency, and vmstat shows idle is 0, what's > the number in C0? > workload seems stable according vmstat (idle 0 for long time), it is compilation of many C++ source codes. C0 is 100%. Some screenshosts should be in your inbox now (too big for mailing list). > Thanks, > -Aubrey Best regards, Milan _______________________________________________ pm-discuss mailing list [email protected] http://mail.opensolaris.org/mailman/listinfo/pm-discuss
