The following subsections explain in detail the steps involved in running NAMD on particular platforms.
In order to run NAMD on network of workstations, a host program (called conv-host) is needed in order to launch NAMD on individual workstations. This program is bundled with the binary distribution. The names of individual workstations as well as other control information such as how many processes to create on those workstations (should they happen to be multiprocessor workstations) is specified to the host program through a file named .nodelist in the user's home directory or can be superseded by the nodelist file in the current directory. Details about the syntax of nodelist file can be found in Converse Installation and Usage Manual at http://charm.cs.uiuc.edu/.
On IBM SP3, you would have to use the scheduler available on the front end
of the SP system in order to run NAMD. Below, we give the description of
the commands we use at Argonne National Laboratory's IBM SP3 installation.
Consult the system documentation or your local system administrator about the
details at your site. We use the command spsubmit to submit NAMD runs to the SP scheduler at Argonne National Laboratory.
spsubmit -np N -maxtime mins namd2 <config-file>
N is the number of processors we request NAMD to run on, and
mins i the maximum time we expect NAMD to take during this run.
mpprun is the name of the command used to run parallel jobs on CRAY
T3E. This command can be used as:
mpprun namd2 <config-file> +pN
where N is the number of requested processors. However, this can be
used only when the number of requested processors is less than or equal
to 32. For jobs larger than this, one has to use queuing facility provided
at the site of installation. Description of the queuing facilities is out
of this document's scope, and we leave that up to you to explore.
Origin2000 version of NAMD is a shared memory version, and does not need
any particular host program to launch NAMD. Thus, in order to run
NAMD on Origin2000, we use the following command:
namd2 <config-file> +pN
where, N is the number of processes we plan to use. However, note that
depending on the load and the queuing strategy employed by the system some of
the processes may be interleaved on some processors, thus causing the
performance to suffer. Some installations of Origin2000 employ an external
queuing facility to avoid this problem. You are strongly encouraged to
consult your local documentation for queuing facility used.