Rémi Forax wrote: > I've perhaps found why there is no difference in your test. > I've re-read the code of your ruby Classoader and someting is weird, > the last instruction of the static block should be
Some actual numbers now, showing difference. The test instantiates 100k new Ruby methods and immediately causes them to JIT. This effectively creates 100k classes, normally with each loaded in its own classloader. Heap memory did not appear to be a whole lot different, though Runtime.totalMemory reported 26MB with ACL versus 31MB without. That's not going to be very accurate though...jconsole reported both processes using about 210MB of heap. Non-heap memory was smaller for the ACL case, 28MB versus 31.3MB. Permgen was down to 25.2MB from 28.5MB. So there's a definite improvement on non-heap and permgen memory use. Excellent! - Charlie _______________________________________________ mlvm-dev mailing list [email protected] http://mail.openjdk.java.net/mailman/listinfo/mlvm-dev
