Actually this is a miss reading. The whole point of the first post was 
to point out that the speed up from going from Perl (or any scripting 
language) to C. I used the same Perl code to pre compute values to 
create the C header file because I already had code at this point that 
did the work and the impact of doing the pre compute was pretty much 
insignificant to the overall run time. I then used to same technique to 
get the Java implementation. I have never claimed that the C version was 
the best possible implementations only that it delivered a massive 
performance boost. I think that I have proven that.

There are optimisations that could be applied to the C version but they 
are only going to get percentage improvements and are getting much 
harder to reliably measure.

As to Ocaml versions, well they were submitted and proved to be faster 
than my C version. C is quite a fragile language to develop in and so 
any high level language that can deliver that performance is very 
interesting. To me at least.

And all this Java stuff, well that comes from the post by Charles O Nutter:

 > - Write it in C is as valid as write it in Java (as someone else 
mentioned).
 > Java is at least as fast as C for most algorithms.

It was the "at least as fast as" bit that I disputed and the whole Java 
sub thread comes from that.

The only requirement is that I can get an implementation of the language 
in question on my Macintosh.