[Cado-nfs-discuss] multithreaded merge

Guy - pleasechoose at hotmail.com
Wed Jan 3 20:06:01 CET 2018


Nevertheless, multithreaded merge would be nice:)


I've tried running opal-test/optimize.sh but its really slow, its running las blackbox on just one process.  the NUM_THREADS controls the number of threads of 'las' but I would like to keep it not more than 4 threads.

I've tried to change it to nomad.mpi and set the mpi parameters, but that did not spawn any more las processes (only multiple nomad.mpi processes)

Any idea how to speed up this optimization to utilize all of the processors?


Thanks!


________________________________
From: paul zimmermann <Paul.Zimmermann at inria.fr>
Sent: Tuesday, December 12, 2017 12:46 PM
To: Guy -
Cc: emmanuel.thome at inria.fr; cado-nfs-discuss at lists.gforge.inria.fr
Subject: Re: [Cado-nfs-discuss] multithreaded merge

> I am examining what it would take to run cado+msieve for many (hundreds and more)
> large (150-200 digits) numbers. So for me any improvements (of hours per number)
> will be multiplied in the long term and can have big impact...

if you really want to save time, try to optimize the sieving parameters using
scripts/opal/optimize.sh. They might not be optimal for 150-200 digits.

Paul
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.gforge.inria.fr/pipermail/cado-nfs-discuss/attachments/20180103/da731ff5/attachment.html>


More information about the Cado-nfs-discuss mailing list