Juan Luis Varona Malumbres on Thu, 31 Jan 2002 02:32:26 +0100

[Date Prev] [Date Next] [Thread Prev] [Thread Next] [Date Index] [Thread Index]

Re: A bug with factorint?


Thanks a lot for the answer.

I has been checking (with the same number) using your idea (flag=6),
and it seems that MPQS will expend about a month. So it is true,
it is not a very good MPQS implementation.

Anyway, I think it will be useful a possibility to control
the number of elliptic curves with gp code.


Juan L. Varona

We never reach MPQS method.

That is true only for large values of "never". :)  ECM in fact told
you in advance that it was going to try for 38 iterations.

You can suppress the ECM stage before MPQS through the flag argument
to factorint  (use 6),  but MPQS on a 95-bit number with the present
implementation will require on the order of many months  (sufficient
memory assumed, and a stable system and reliable power supply) - enough
to justify spending 3 weeks on ECM first, although this looks counter-
intuitive to us impatient human spectators.

Your best bet at present, if you want to do this with gp, is to
attempt the factorization simultaneously on two machines with
different flag settings, so that one will keep churning away
with ECM while the other embarks on MPQS.  You'll find that this
ECM stage will give up long before MPQS approaches the Gaussian
elimination phase.  Of course, you can run software more specialized
towards integer factorization simultaneously on a third machine...

Cheers, Gerhard