Alex Granovsky
gran@classic.chem.msu.su
to run mpich1/statically linked version on a single node it is
sufficient to put the following into your procgrp file:
local 3
To run across several nodes you need a bit more complex procgrp file,
e.g. the following one:
local 3 node2 4 /home/mohamed/firefly810/bin/firefly810
will run 4 processes on local node and 4 processes on node "node2",
i.e. 8 processes in total. You can find a bit more info on this here:
http://classic.chem.msu.su/gran/gamess/linux_mpich.html
Anyway, I'd suggest you to run Firefly using more recent MPI implementations like mpich2 or intel MPI. The performance should
be better, and the procedure to run Firefly in parallel is simpler.
Kind regards,
Alex Granovsky
On Wed Jun 24 '15 10:29pm, makhyoun wrote
-----------------------------------------
>Dear all;
>I succeeded to run firefly 8.1 on one node/4cpu as follows: (statically linked with mpich)
>
>
>
>mpirun -p4pg procgrp /home/mohamed/firefly810/bin/firefly810 -ex /home/mohamed/firefly810/bin -t/home/mohamed/firefly810/EXE/tmp
>
>
>
>
>The structure of the procgrp file for one node is:
>linux-jeem 0 /home/mohamed/firefly810/bin/firefly810 mohamed
>linux-jeem 1 /home/mohamed/firefly810/bin/firefly810 mohamed
>linux-jeem 2 /home/mohamed/firefly810/bin/firefly810 mohamed
>Now I don't know how to run it on more than one node.
>My system is an ordinary cluster.
>
>
>
>
>