[Cado-nfs-discuss] Filter Parallelization
Paul.Zimmermann at inria.fr
Fri Jul 24 09:14:06 CEST 2015
> From: "Wilson, Andrew K." <Andrew.Wilson at noblis.org>
> Date: Thu, 23 Jul 2015 18:09:32 +0000
> My name is Andrew Wilson. I am a data scientist in the Big Data Lab at Noblis<http://www.noblis.org/>. We are currently implementing CADO 2.1.1 in a client project and needed to parallelize the filtering steps (among others). I noticed on your website that it was an "ongoing work" and that it currently "is not parallel." Is this near finished or in progress? If not, could you give me a few pointers on how to parallelize this code?
> I noticed the switch for force-posix-threads in all of the source files and tried using this but didn't see any performance improvements. I assume, then, that this is the in progress work?
> Thanks for your time. I look forward to your response.
> Andrew K. Wilson
> Junior Data Scientist
the filtering step now can use several threads (default is 4) in the "purge"
command. However with n threads you won't get a speedup of a factor n.
This has little priority for us, since the most time consuming steps
are the sieving and linear algebra step.
Thus if you want to work on this, feel free!
More information about the Cado-nfs-discuss