I'm working on a Sudoku solver at school and we're having a little performance contest. Right now, my algorithm is pretty fast on the first run (about 2.5ms), but even faster when I solve the same puzzle 10 000 times (about 0.5ms for each run). Those timing are, of course, depend of the puzzle being solved. I know the JVM do some optimization when a method is called multiple time, and this is what I suspect is happening.

I don't think I can further optimize the algorithm itself (though I'll keep looking), so I was wondering if I could replicate some of the optimizations done by the JVM.

Note : compiling to native code is not an option

Thanks!