Juergen Keil wrote: > 2009/4/13 Rafael Vanoni <Rafael.Vanoni at sun.com>: > >> After applying your patch, the p-state numbers would display negative values >> when the sampling interval was shorter than the interval between firings of >> the cpu-change-speed probe. I added a check in pt_cpufreq_stat_account() to >> fix that. I'm including the diff below. > > Something like this? > > P-states (frequencies) > 996 Mhz 131,2% > 1328 Mhz 0,0% > 1660 Mhz 0,0% > 1992 Mhz -31,2% > > I noticed this earlier today, and just wanted to start looking what's wrong; > but it seems you already fixed this...
Yeah, I'm still looking at this other issue. I'll try to get it integrated today. Rafael >> I'm still seeing the original issue when observing CPUs other than CPU 0. >> The --cpu option has been on the project gate for a while, but was only >> putback last Friday.
