well - i may not use MUL - i will probably use the extended addressing modes :P
for now, i am just writing a test to 8,589,934,591 so i can compare it with Eratosthenes odd-bit
with a little luck, i can develop a theorem to eliminate mod-7's
then, i can use an "Atkin-like" method to create a new "Sheldon-Atkin-Bernstein" sieve