Hi Jasen! :-)

 JB> I figured squares in D would be easier to find than primes to put in 
 JB> for A and B. (not using primes could be slower... using mostly-primes 
 JB> could work I guess)
Yeah, I was thinking one could do a decode run for some combination of A
and B that looks likely and then look at the result. However, there may be
no way to discern a successful decode from gibberish.

 JB> I've not looked closely at the binary implementation of SQRT.
 JB> but IIRC the best implementation is O(n*log(n)) (for n bits)
I'd be using the gmp (GNU math precision) library. However, its
documentation doesn't say how effecient their sqrt() is.

Ciao
Pascal

--- Msged/LNX 6.1.1