Discussion related to cado-nfs

[hermann@stamm-wilbrandt.de]

On 2023-07-24 12:54, Paul Zimmermann wrote:
> Dear Hermann (and others),
>
> Emmanuel Thomé investigated more, and you can follow this interesting 
> bug at
> <https://gitlab.inria.fr/cado-nfs/cado-nfs/-/issues/30068>.
Thank you, interesting to read Emmanuels's investigation.

Regarding my previous email:
Is it safe to use the method I described on 64bit PiOS until bug fix 
will be available?

1) run "./cado-nfs.py someNumber"
2) if that does not complete, determine parameters file in /tmp/cado...
3) repeat running "./cado-nfs.py --parameters determinedFile" until 
successful factorization completion


Btw, I factored RSA-140 on my fast 6C/12T AMD 7600X CPU PC in only 
11.1h(!) with cado-nfs.py. Then I factored RSA-140 on an old 24C/48T 
server in 8.3h(!!) with cado-nfs.py. So I gave RSA-150 a try on that 
server, and cado-nfs.py factorization completed in 38.5h(!).

I will factor RSA-150 on my 7600X PC as well, but need to wait until 
current loooong running sequential job (running at 5.3GHz boost CPU 
frequency) will have completed after 75.4 days ;-)
https://github.com/Hermann-SW/9383761-digit-prime#readme

3d 20h are already completed:
https://twitter.com/HermannSW/status/1683448242934013952

PC does compute "sqrt(-1) (mod p)" for biggest known prime p which is =1 
(mod 4), the 9,383,761 digit prime being rank 9 of biggest known primes 
list (https://t5k.org/primes/lists/all.txt).


Regards,

Hermann.

> > Date: Fri, 21 Jul 2023 20:36:12 +0200
> > From: hermann@stamm-wilbrandt.de
> > Cc: cado-nfs@inria.fr
> >
> > On 2023-07-20 15:27, Paul Zimmermann wrote:
> > > At this point I cannot debug further, I will pass on our local expert.
> > >
> > Thank you.
> >
> > > However it is not fully deterministic, sometimes it works:
> > >
> > I did factor RSA-100 on the Pi400 again.
> > This time overclocked with 2.2GHz for comparable results with 
> > sequential
> > msieve, overclocked Pi400 did factor RSA-100 in 12.1h:
> > https://forums.raspberrypi.com/viewtopic.php?t=342544
> >
> > Again the first run stopped with error after 112min:
> > https://stamm-wilbrandt.de/en/forum/pi400_2200MHz,RSA-100.log
> >
> > You said "not fully deterministic", and so I did give it a 2nd chance 
> > by
> > restarting with "--parameters", and it failed quickly.
> >
> > Finally I did another run, and this time factorization completed!
> >
> > i@pi400-64:~/cado-nfs $ grep "time ./cado-" pi400_2200MHz,RSA-100.log
> > pi@pi400-64:~/cado-nfs $ time ./cado-nfs.py
> > 1522605027922533360535618378132637429718068114961380688657908494580122963258952897654000350692006139
> > pi@pi400-64:~/cado-nfs $ time ./cado-nfs.py --parameters
> > /tmp/cado.axcxb7kz/c100.parameters_snapshot.0
> > pi@pi400-64:~/cado-nfs $ time ./cado-nfs.py --parameters
> > /tmp/cado.axcxb7kz/c100.parameters_snapshot.0
> > pi@pi400-64:~/cado-nfs $
> >
> >
> > With this runtimes:
> >
> > pi@pi400-64:~/cado-nfs $ grep "^real" pi400_2200MHz,RSA-100.log
> > real    112m28.891s
> > real    0m4.184s
> > real    1m20.698s
> > pi@pi400-64:~/cado-nfs $
> >
> >
> > So in total only 1:53:54h, compared to 12.1h with msieve -- really 
> > cool
> > (Pi400 has 4 cores, but only 1 thread per core).
> >
> >
> > Regards,
> >
> > Hermann.