The program uses all the fifteen bit primes
to generate an endless stream of high quality
low cost random bits with a very long
period. The permutation group element size
that was actually applied to
ALL of the fifteen bit primes
has a cycle size greater than
ten to the seventy.
That permutation subgroup size
may in fact be closer to
ten to the seventy five.
BTW, adding all sixteen bit primes
does very little additional good,
and does add high transient cost,
although normal per bit cost
would go slightly lower.

There is one portion of the routine
that needs improvement but tis
all of three instructions in length
so I did nay bopther right now.

The tests on the prime squareds field
were really quite good and the
results should compare favorably
with EVERY other known
random number generator.

Thats not a boast,
its a very highly educated guess.

Sincerely, Daniel