It's not surprising that the performance of JDK 6 has improved significantly as they have introduced escape analysis where the JVM can convert heap allocations to stack allocation where possible.
Sanjiv On 10/13/07, Allan Ang <[EMAIL PROTECTED]> wrote: > > > true. > > but the implementation for a bubblesort is relatively simple and the c++ > and > java code match each other very very closely (i.e 2 simple for loops and a > swap). Let me know if you want the codes that i used. Naturally we only > started the timing when after the arrays were initialized. > > I was also surprised with the difference in times for different jvm > implementations. (an order of magnitude of 6!). I retested numerous times > to > make sure that I wasn't making a mistake. The compiler makes a difference > too. When I compiled with the BEA 1.5 and ran using the Sun 1.6, I got a > slower time (13 s) than if I compiled and ran using the Sun 1.6 (5s). > approx 3x as fast. > > I'll probably be making a few more of these tests and perhaps post the > results online sometime. > > Of course it's a little too presumptuous to make any conclusions at this > point, but I was wondering whether it would indeed be more beneficial to > just simply upgrade to a higher version jvm (if possible) to improve > performance rather than going through our codes with a fine tooth comb. > > -- > View this message in context: > http://www.nabble.com/Sun-java-JDK-1.6-performance-tf4611665s2369.html#a13191122 > Sent from the AppFuse - User mailing list archive at Nabble.com. > > --------------------------------------------------------------------- > To unsubscribe, e-mail: [EMAIL PROTECTED] > For additional commands, e-mail: [EMAIL PROTECTED] > >
