To run the code on a network of workstations using the Parallel Virtual Machine (PVM) software, the executable pccm2 must be located in the users directory pvm3/bin/$ARCH, where $ARCH specifies the particular machine, eg. RS6K. Since passing environment variables through PVM is a bit tricky, the user should either specify the location of the binary files directly in the namelist file or provide links to the appropriate files in the users home directory. Output will be written to the users home directory.
To start a PVM session, it is assumed that PVM has been installed on a cluster of workstations (or the MPP) and that appropriate links and modifications to path statements have been made. The user starts the PVM console by typing pvm in some window. This responds with a prompt and allows the user to add machines, i.e. other workstations, defining the virtual machine. In another window, the user executes the pccm2 code which will spawn the pccm2 on the other processors of the virtual machine. Logically, the program is still defined on a mesh of processors though, in fact, the virtual machine may consist of a heterogeneous collection of processors linked with a local area network.
Further information on PVM may be found by browsing
http://www.epm.ornl.gov/pvm/ .
PCCM2.1 may also be used under MPI on networks of IBM or other vendor's workstations. Consult
http://www.mcs.anl.gov/mpi/
for information on obtaining and installing MPICH, the publicly available version of MPI, and for running parallel programs under this version.