Discussion related to cado-nfs

[Владимир Мартьянов <vilgeforce@gmail.com>]

2016-07-13 10:53 GMT+03:00 paul zimmermann <Paul.Zimmermann@inria.fr>:

       Hi,

> From: Владимир Мартьянов
>  <vilgeforce@gmail.com>
> Date: Tue, 12 Jul 2016 17:17:31 +0300
>
> Hello again!
>
> I faced with some troubles in client-server mode. I have a snapshot of the
> work in /tmp/cado.sfjjfw3j/c270.parameters_snapshot.0 I tried to add a line
> server.port=1313 to this line to specify a new port, it doesn't work.

please can you give more details? The following works with the git version:

1) $ ./cado-nfs.py 8312294295031399876237581072267611210232959213949587843322514862035389516592913 server.port=1312

Oh, I just misspelled: server.port=131 instead of 1313, and the Linux denied to open the port. Sorry...

 

the number of threads per work unit cannot be changed, since the master cannot
know which client will process which work unit. However you can start as many
cado-nfs-client.py's you want on each node (see README):

   The programs (for the polynomial selection and sieving steps) can run
   multithreaded, but it is better to have them run with a capped number of
   threads (say, 2), and run several clients per node.

Yes, I used the solution with multiple clients on a PCs, but it's not convenient in case of 32-core PCs. Person, who starts the slave, definitely know how many threads he wants to give to the client.