It would probably be a good idea to offer something in the way of an explanation as to why a custom list of code offers significant benchmark improvement.

Did the guy who first posted that say anything about it?