could anyone review this code "search routine" ?

[mar]

Interesting, just ran a quick benchmark on 100 random numbers and indeed shellsort was nearly 3x faster than insertion sort,
maybe even more because I couldn't subtract the time to generate PR numbers (my generator is pretty fast though).

I agree that jump tables should be part of the standard
as it's trivial to do something like that in assembly and it would be damn useful.