I wrote a simple program in Java in order to test the speed of java in Windows and Linux. All that my program does is to run some loops some times. I used windows xp sp2 and fedora core 3 64bit with 64bit java. None of the systems had any other load while executing the program.

In windows the program finished in 1:36.73 and in Linux in 2:12.96. I didn't expect to see so big difference (36 seconds) especially because Linux was a 64bit system.

Does anyone know if execution of java is slower in Linux or this was just a coincidence for my program?

Here is the source code for anyone who has a dual boot system and would like to test it. Please let me know what the results were on your system. I have an amd 64bit 3400+ with512MB of RAM.

http://egnatia.ee.auth.gr/~adns/benchmark.java

Are you running a 64 bit version of Windows?
Are you running a version of Java that takes advantage of your 64bit processor?
Have you tried running this on any other distro besides Fedora? Its not exactly known for its speed.

Those might explain the difference.

I run 32bit Windows with 32bit jsdk and 64bit fc3 with 64bit jsdk. I wanted to test the speed in those two since i have to do some 'big runs' of java programs and therefore the fastest the best. I don't have any other distro installed to make a check. That's why i asked if someone else has a dual boot system with any distro to test it.

It might just be the different compilers then. They don't all work alike.

I didn't mention that i have downloaded from sun's site jsdk 1.5.0 for windows and 1.5.0_04 amd64 for Linux. I supposed that the one for Linux should be faster. I don't know if it has only to do with the difference in Linux-Windows system or if there is a problem with the 64bit version also.

Your benchmark source code is quite buggy, all the loops are reusing the same i variable so in fact only the innermost for loop is fully executed, all other are useless.
Even corrected, this program doesn't represent a real life Java application so any numbers you got will not really represent the JVMs performance.

Oups you are right... The new code is here http://egnatia.ee.auth.gr/~adns/benchmark.java but i haven't tested it yet.

I know that this program doesn't represent a real life Java application. I just wanted a way to measure how fast the compiled code runs so i assumed that this was a simple way.