Alex van Denzel <bitbucket@adenzel.demon.nl> writes:

> He was astounded at the speed the program executed, until he
> unassembled the output and only got a RET...

A friend and myself once tried to hand-optimize a FORTRAN program
under VAX/VMS, on the knowledge that if some particular input
parameters to a calculation involving a deeply nested loop were zero,
the calculation could be bypassed.  Strangely, we didn't get any
improvement at all.  Studying the generated assembly code, we found,
to our amazement, that the FORTRAN compiler included our tests in the
output even if we didn't have them in the source code...  :-)

-tih
-- 
Popularity is the hallmark of mediocrity.  --Niles Crane, "Frasier"