Re: Stack vs stackless operation

I have done some work on optimisation on ciforth.
This work has stalled, but the infamous byte prime benchmark,
was in the ballpark of swiftforth and mpeforth.
(Disingenuous, because this was the example I used.)
See https://home.hccnet.nl/a.w.m.van.der.horst/forthlecture5.html
This is about folding, a generalisation of constant folding.
This requires that you know the properties of the Forth Words,
i.e. that you can execute + at compile time, if the inputs
are constant.
[..]
Then I got stalled. I introduced complicated rules [..]

This is more sophisticated way.
My proposal is rather humble: a modest completion
of Forth programming paradigm, from "every data goes
through the stack" to "any data can go through the
stack, still it's not strictly mandatory in every
single case".
Now I'm pondering about DO..LOOP construct; actually
probably it doesn't necessarily need to rely on return
stack. I wonder how much "lightweight" can loop become
by rewriting it using OOS words.
--