[Cado-nfs-discuss] cadofct.pm parameter issue

Zimmermann Paul Paul.Zimmermann at loria.fr
Thu Jul 21 19:50:26 CEST 2011


       Jim,

> Working with version 882, the cadofct.pl in trunk does not properly parse
> kjM, kjl and kjkeep while the one in branches/1.0 does. Parameter files
> (param.c155) in both trunk and branches seem to contain these values and
> therefore throw an =93Unknown argument=94 error when using the cadofct.pm i=
> n
> trunk. Is it advisable to comment these parameters out? Or use the
> branches/1.0 code?

I have now updated all the parameters files in the trunk so that they use the
new parameters (kjP instead of kjM). Note however that for large numbers they
might not be optimal or worse the polynomial selection might fail.

If you factor a large number with CADO-NFS, please run the scripts/cpu_time.sh
at the end in the wdir directory, and send us the output, together with the
*.poly file. This will help us improving the parameters.

As a rule of thumb the polynomial selection should take 3% to 5% of the total
time, and the maxnorm value should be such that each polynomial selection run
keeps about 10 polynomials (see "xxx below maxnorm" in *.poly).

Hope this helps,
Paul Zimmermann




More information about the Cado-nfs-discuss mailing list