Li, Aubrey wrote: > Rafael.Vanoni wrote: > >> I've updated the webrev @ >> http://cr.opensolaris.org/~rafaelv/ptop-freq/ >> >> I'll file a separate CR to change the DTrace script in cpufreq.c, we >> have other issues I think are more important at this time. >> >> This latest webrev is against PowerTOP's gate, not ON. I've added a >> check to make sure we're deducting the accounted time for the correct >> p-state. Let me know what you think. >> > > I don't think the patch is doing the right thing to deal with accounted time. > > Example: > > Round1: 5 seconds > 1) 2s in P1 > 2) 3s in P0 > > Round2: 5 seconds > 1) 3s in P0 > 2) 2s in P1 > > Here, in the second round, accounted time is apparantly for P0, but > cpu_pow->current_pstate is P1. So the suggested patch is still wrong.
What accounted time? Your scenario is self contained. In both cases, time_accounted would be zero since the aggregation results account for the entire interval. Rafael
