Cluster:Running Gaussian in parallel

From Earlham CS Department
Revision as of 14:04, 5 July 2005 by Skylar (talk | contribs)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to navigation Jump to search

Understand the distinction between nprocs and nprocl. nprocs is for a shared-memory (single-image) machine, whereas nprocl is for a distributed-memory (cluster) system. You want to use nprocl on all our computing systems. For example, if you want to run on eight processes (four nodes), add a line like this 

%nprocl=8

to your input file. You should then specify qsub -l nodes=4:ppn=2 script on the command line to submit your PBS script with a requirement of four dual-processor machines.

Once you have the number of CPUs specified in the input file, make sure that your nodelist in GAUSS_LFLAGS is set properly. See Gaussian environment for details.

If you have a bunch of input files you want to modify, use the add_nproc.pl script in /cluster/generic/bin. First cd to the directory with your input files, and then do: 

perl /cluster/generic/bin/add_nproc.pl n

where n is the number proecssors you want to run on.

After that, fire it up with $g03root/bsd/g03l.

Retrieved from "https://wiki.cs.earlham.edu/index.php?title=Cluster:Running_Gaussian_in_parallel&oldid=5369"