I put a post on our blog a little while back discussing this.

   http://ocaml.janestreet.com/?q=node/71

There are a number of tricks you can do, including loop unrolling, and using
a counter to keep track of the number of stack frames, to get code that
behaves well on small-to-medium lists, uses a bounded number of stack
frames, and is faster than the standard List.map even for small lists.

y