Francois Pussault writes:
I found a procedure:
[..]
That seems to be an experiment in how many stack errors and fundamental flaws can be packed in a single definition. Where did you find that? Adding some swaps and fixing the 32/64 incompatibility may fix this shiftregister based algorithm, but better throw it away altogether. By reading the coding it is obvious that when the 'register' encounters a zero it will stick there. In this case you didn't execute randomize, so it sticks right away. I think Gforth has random ( u1 -- u2 ), and 2 CHOOSE should work for a 1/0 sequence (or 2 choose invert). -marcel
