Discussion related to cado-nfs

[Paul Leyland <paul.leyland@gmail.com>]

I would like to suggest that the documentation on how to run SNFS with
CADO-NFS be expanded.  Unless I've missed something very important the
present documentation can be summarized as "create a polynomial file by
hand instead of using the GNFS polynomial finder".

Now that's fine as far as it goes, but it doesn't go far enough in my
opinion.  In particular, no guidance is given on how to select the other
parameters such as prime bounds and sieving area.  Old hands at NFS,
such as myself and many (all?) of the developers can work it out but
only laboriously and through rules of thumb such as the 3/2 heuristic.

Could a set of files be created in params{,_py}/ with names such as
params.s150, params.s155, params.s160, .... params.s250, params.s255,
params.s260 ... and so on up to a reasonable upper limit?  Reasonable
here might be params.s310 to allow for kilobit factorizations.  I'd
certainly find it useful up to C260 or so.

The existing gnfs params and those shipped with msieve may be a good
starting point.

I am aware that not all SNFS polynomials for a particular size sieve
equally well.  A program to calculate the parameters would be ideal, but
a simple set of static tables would much better than nothing.


Paul (the other one).

P.S. A GNFS c143 started with alpha1 was completed successfully with the
beta release earlier this morning.  8-)