>Agreed, but a much more plausible possible optimization is the >introduction of memo-functions. This is not a completely trivial Unfortunately, we are talking about C with side effects and aliasses and all those other nasty-nasties that give optimisers fits.