[Cado-nfs-discuss] Using GPU in Sieving Step

Emmanuel Thomé emmanuel.thome at gmail.com
Thu Aug 16 10:06:43 CEST 2012


Hi,

Cofactorisation in nfs software has special-purpose code adapted to the
composite sizes encountered.

Also, cofactorisation does early-abort every so often.

Gmp-ecm is neither designed for, nor adapted to the cofactoring situation.
Different problems, different software.

To OP: no SLA implied with list membership. Thanks.

E.
Le 15 août 2012 15:33, "Shi Bai" <shih.bai at gmail.com> a écrit :

> Hi Hamid, Zach,
>
> On Sat, Aug 11, 2012 at 11:06 PM, Zachary Harris
> <zacharyharris at hotmail.com> wrote:
> > Hamid,
> >
> >   Someone with greater expertise can correct me if I'm wrong, but I
> believe
> > that, as a complete factorization algorithm itself, ECM is independent of
> > NFS and therefore your answer is "No", it is in no way useful for the
> > sieving part of NFS (i.e. ECM is not a sieve).
> >   As a rough heuristic, if I have a large number to factor, I'll hit it
> with
> > ECM first to quickly pull out any small (~25 digit) factors, and then if
> the
> > remaining factor is 60 digits or more I'll let NFS handle that.
>
> It's more likely that Hamid mean the ECM for the cofactorization step?
>
> > On 08/11/2012 01:40 AM, hamid reza arkian wrote:
> >
> > Hi,
> > I have sent some questions 4 days ago, but I have not received any answer
> > yet.
> > Can anyone help me, please?
>
> I know that some members of the CADO-NFS team are on holidays these
> days so they may not see this.
>
> > Thanks,
> > Hamid
> >
> >
> >
> > On Tue, Aug 7, 2012 at 11:36 AM, hamid reza arkian <
> hamid.arkian at gmail.com>
> > wrote:
> >>
> >> Hi,
> >>
> >> I tried GMP-ECM project (the branch that is currently under active
> >> development for GPU devices) on a GPU cluster and got good results.
> >> I want to replace ecm part of sieving step in CADO-NFS with GMP-ECM(that
> >> support GPU), but I have the following questions:
> >>
> >> - Basically, it's a good idea or not?
> I personally think it's a good idea (though I admit I haven't wrote
> codes for the sieving part.) But there could be some issues, such as
> the stage 2 memory requirement. PS: I've CC-ed this email to Alex
> Kruppa, who can answer this much better.
>
> Kind regards,
> Shi
>
> >> - Is the GMP-ECM faster than ecm part of CADO-NFS? and if so, why the
> >> developers don't do this replacement,yet?
>
> _______________________________________________
> Cado-nfs-discuss mailing list
> Cado-nfs-discuss at lists.gforge.inria.fr
> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/cado-nfs-discuss
>
-------------- section suivante --------------
Une pi?ce jointe HTML a ?t? nettoy?e...
URL: <http://lists.gforge.inria.fr/pipermail/cado-nfs-discuss/attachments/20120816/5e2fb8ec/attachment.html>


More information about the Cado-nfs-discuss mailing list