Oh, yeah. The call overhead specifically isn’t all that onerous though is it? For your example you’re also talking about making a memory copy, and unless you have hardware multiply you’re doing looping addition.

Most Forths I’ve dealt with also offer inline assembly as part of a word definition, so I suppose you could do it that way if really desired. I can see what you mean though about the penalty being completely acceptable, because it shouldn’t be super large.