On either of dextrose or entropy,
you should be able to run the
g09 and ghelp commands to run
Gaussian 09, without
making any changes to your environment (there are
g09 and ghelp scripts in
/usr/local/bin which include the environment
variable settings required by Gaussian). On dextrose,
you can also run g03, to run Gaussian 03,
the previous release.
Remember, though, that on dextrose, you need to use
the bsub command to send your Gaussian jobs to
dextrose's compute nodes, rather than
starting them directly on the cluster's head node. See
this primer about batch jobs on dextrose.
You can run other Gaussian commands by prefixing
their name with g09util (or
g03util, on dextrose). For example, to
run the formchk utility to get a
machine-independent version of a binary checkpoint
file:
Running g09util without arguments provides
a partial list of available utilities. Gaussian
provides limited documentation for the use of these
utilities. You can get instructions for some by
providing their name as an argument to ghelp, for
example:
Some utilities are described in the Gaussian 09 Online Manual Others seem to be undocumented.
For information about input to Gaussian 09 itself, see its online manual.
You can use the %NProcShared command to
indicate that Gaussian should use multiple
shared-memory processors for the portions of the
calculation which can be parallelized. On
dextrose's compute nodes the
maximum number of processors available is twelve, i.e.
(You cannot use the %NProcLinda command
to run on processors from multiple nodes on dextrose because our
version of Gaussian does not include the Linda extension
for distributed memory parallelization.)
On entropy, the theoretical maximum number
of processors is seventy-two, but in practice you need to
take into account the
needs of other users when deciding how many
processors to use on entropy.