[Cado-nfs-discuss] Will hyperthreading result in faster or slower sieve times?

ROBERT DUCHATEAU rjd1212 at cox.net
Fri Jan 19 12:13:15 CET 2018


Thanks for the quick reply, Paul,

OK - then, that's good news concerning the hyperthreading increasing the wall clock time & that there is no network bottleneck.

On each of the seven clients that are performing the sieving, all have 16GB of DDR4 3000 memory - while the server has 64GB of the same memory running 16 threads. ( 8 cores )

Best Regards,

RJD

> On January 19, 2018 at 5:44 AM paul zimmermann <Paul.Zimmermann at inria.fr> wrote:
>
>
> Robert,
>
> > Currently, I am factoring RSA C180, utilizing a 10GB network using eight machines
> > for a total of 72 threads.
> >
> > Would eliminating the hyperthreading portion speed up the lattice sieving, in
> > addition to possibly tamping down any possible network bottleneck, or not?
> >
> > Best Regards,
> >
> > RJD
>
> my experience is that using hyperthreading helps to reduce the wall-clock time,
> assuming the las jobs do not exceed the available memory of each node.
>
> In the sieving phase there is no network bottleneck.
>
> Best regards,
> Paul Zimmermann
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.gforge.inria.fr/pipermail/cado-nfs-discuss/attachments/20180119/f5079174/attachment.html>


More information about the Cado-nfs-discuss mailing list