On Fri, 3 Jan 2014, Viresh Kumar wrote: > On 2 January 2014 23:08, Mikulas Patocka <mpatocka@xxxxxxxxxx> wrote: > > Flushing the cache and changing frequency takes approximatelly 500us. The > > patch increases policy->cpuinfo.transition_latency to that value. > > Its not about how fast caches get cleaned but how much time would > be wasted to get them filled again as same data could be required again > which is just flushed out. That would impact performance more than > flushing caches. I didn't see any performance degradation when I tried changing the frequency manually with or without the cache flush patch - the overhead of running cpufreq (8ms) is far worse than the frequency transition itself. Mikulas -- 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
