Paul:

> I wonder why PFGW did not implement Baillie-PSW.

I guess because then the fast PrP test would be 3 times slower.

And the -tc test is stronger than B-PSW, I believe.

If you have a fair pecentage of N^2-1 it's *very* strong.

If you have enough for BLS it's a proof.

Why should one want something intermediate between

a fast PrP and a slower-and-better-than-B-PSW test?

After all, only a fraction=O(log(sieve_depth)/log(N))

get through the fast test, so it doesn't matter that

the slow one is even better (and hence slower) than B-PSW.

[Except to Chris, who is suffering right now.]

Maybe you do not see it this way, because your log(N)

is so small. But PFGW is for finding large primes,

not small pseudoprimes.

Or have I missed something?

David