Well, I think we will need the java code to try and identify any obvious differences. My fist guess would be that the java version may have used tail recursion extensively. That can be a huge benefit because it nearly eliminates the overhead of a subroutine call all together. Even if he didn't there are some really cool hotspot optimizations that JDKjava 1.6 can do. For instance, the equivalent of the grid_check method would likely be compiled down to machine code and run natively during runtime, where as the perl version would always be interpreted. And again, depending on how the java version was written, many of his method calls could have been inlined.
Update: I remember hearing quite often that perl sub calls were rather slow (aside from no tail recursion support), but I cannot remember the details as to why they may be slower (it is possible that this is no longer an issue either). Hopefully someone better informed will either dispel or clarify this.
Ted Young
($$<<$$=>$$<=>$$<=$$>>$$) always returns 1. :-)