On 17 September 2013 07:43, karthik vm <meetvm@xxxxxxxxx> wrote: > Yeah actually that is what my concern is. It seems that for changing > the frequency we need to acquire a lock (in cpufreq_driver_target()). > Hence even though the hardware is capable of ns level DVFS, OS > overhead may be large enough to take advantage of it. Can we measure > this overhead? There are a lot of other things that we really need to look before this one.. The way we are changing frequency currently is very delayed... There are few other threads that are going around power aware scheduler which you might be interested in.. Look for them in Archives.. Coming back to your question, these kind of overheads will always be there, isn't it? Normally that lock will be acquired fairly quickly as there might be no contention.. Yeah, you can measure it but don't know who is your best friend in term of tools.. Maybe ftrace can get you some details. -- To unsubscribe from this list: send the line "unsubscribe cpufreq" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
