The aim with SIQS is to make switching polynomial very fast such that many more polynomials can be used than MPQS. However, I believe that my implementation still spends 10% of its time switching A. The more factors you have in A the less you will need to switch B, however, the factors get smaller the more you have and are excluded from sieving so they can't get too small.
Are you using large primes?
At 50 digits I use about 11 As. This increases rapidly from here as numbers get bigger. Unless you are spending all your time generating them I wouldn't worry about having a few As. Each A had 6 factors allowing for 64 Bs per A.
Last fiddled with by henryzz on 20190516 at 08:27
