27.04.2011 22:41, Josef Weidendorfer пишет: > On Thursday 21 April 2011, Вадим Воеводин wrote: >>> Long answer is, the question is kind-of meaningless. Cache misses are >>> a function of the overall memory behaviour of your program. So the >>> misses on b[] also depend on how the program accesses a[], x[], ip[], >>> etc, and you can't really measure each in isolation. >>> >> Of course I agree with you, it depends on other accesses, but it is >> interesting for my research to get caches misses for particular arrays >> as well as for the whole program. >> And in simple programs like this influence of different arrays on each >> other can be approximately estimated. > Not sure I understand this statement. I am curious. How do you suggest to > estimate the influence of different arrays on each other? Even if you > only have 2 arrays, there still could be accesses to the stack, and the > code around the inner loop also matters. With 3 arrays as in this example, > separate influences of one array on the other is very difficult to define > in a meaningful way, and probably would be mostly useless to come up with > code improvements. I suppose you also would need to take into account how > array accesses are covered by the different cache sets. > OK, I would say it like that: in my research I'm trying to understand, can I (and how precise) estimate data locality using several characteristics. By now I'm working with independent arrays, just because it's simlper. Of course in future array interference need to be considered. So I was interested in number of cache misses for an array in order to understand, whether these characteristics can give rather close estimation. > Josef > > >>> J >>> >>> >> >> >
-- С уважением, Вадим Воеводин. ------------------------------------------------------------------------------ WhatsUp Gold - Download Free Network Management Software The most intuitive, comprehensive, and cost-effective network management toolset available today. Delivers lowest initial acquisition cost and overall TCO of any competing solution. http://p.sf.net/sfu/whatsupgold-sd _______________________________________________ Valgrind-users mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/valgrind-users
