On Mon, Mar 30, 2015 at 09:46:35PM -0300, Arnaldo Carvalho de Melo wrote: > Em Tue, Mar 31, 2015 at 09:27:30AM +0900, Namhyung Kim escreveu: > > On Mon, Mar 30, 2015 at 12:13:03PM -0300, Arnaldo Carvalho de Melo wrote: > > > Em Mon, Mar 30, 2015 at 11:58:05AM -0300, Arnaldo Carvalho de Melo > > > escreveu: > > > > Em Mon, Mar 30, 2015 at 09:48:52PM +0900, Namhyung Kim escreveu: > > > > > But this makes every sample processing grabs and releases the lock so > > > > > might cause high overhead. It can be a problem if such processing is > > > > > done parallelly like my multi-thread work. :-/ > > > > > > > > Still untested, using rw lock, next step is auditing the > > > > machine__findnew_thread users that really should be using > > > > machine__find_thread, i.e. grabbing just the reader lock, and measuring > > > > the overhead of using a pthread rw lock instead of pthread_mutex_t as > > > > Jiri is doing. > > > > > > Don't bother trying it, doesn't even compile ;-\ > > > > OK. :) > > > > But I think rw lock still has not-so-low overhead as it involves > > atomic operations and cache misses. > > But we will have to serialize access to the data structure at some > point...
Yes, as long as we keep ref-counting. I'm guessing if we only focus on the perf top case, there might be a way to cleanup dead threads without ref-counting (i.e. w/o affecting fastpath on the perf report). Thanks, Namhyung -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
