[Cado-nfs-discuss] Fwd: bug in polyselect2l.c parameters

Zimmermann Paul Paul.Zimmermann at loria.fr
Mon Oct 1 15:23:50 CEST 2012


       Dear Jayson,

thank you for your report. It indeed makes perfectly sense.

> For the fix, it's easy to bound each q when -lq is normal. I'm still
> thinking how to deal with large -lq and any suggestions/ideas would be
> more than welcome.

Shi, I suggest you discard all q's such that p1*p2*q > m0, i.e. q > m0/(p1*p2).

If too expensive to check for each p1, p2, you can simply discard all q's
such that (2P)^2*q > m0, i.e., q > m0/(4P^2).

However this has the disadvantage that for some (bad) parameter choices, we
might discard almost all q's, or even all q's.

Possible workarounds:

(1) reduce the upper bound for the special-q's (251 so far) so that
    q <= m0/(4P^2) in all cases. In this case no check is needed.

(2) abort if q > m0/(4P^2) for the largest possible q

Paul



More information about the Cado-nfs-discuss mailing list