Joined: 18 Aug 2004
|Posted: Sat Oct 15, 2005 4:51 pm Post subject:
|Hi Kernel9x -
I guess it all depends on what your expectations are for using the 2 PCs for parallel processing.
The usual way of dong this is to use a message passing library and MPI is generally the one used.
In order to use MPI, your code needs to be written with various MPI functions to send and receive data. Assuming that your code is already an MPI based code, you then need to have the MPI libraries installed on your systems. If you purchased the PGI CDK (cluster development toolkit) it comes with a version of MPI and the install instructions should guide you through the installation.
If you don't have the CDK, you can still download and install MPI yourself. Most people use MPICH for MPI - which is available from Argone National Lab http://www-unix.mcs.anl.gov/mpi/mpich
The Tips and Techniques section on this web site has some information to help you build MPI: http://www.pgroup.com/resources/mpich/mpich126_pgi60.htm
The MPICH web page has instructions on the several different versions and how to use them to launch parallel applications.
If your code isn't written using MPI but you would like to learn about MPI - there are numerous books on the subject, and a number of tutorials that you can find on the web. It takes a bit of work to get your hands around it when you first start, but after you get used to it, you can write very complex applications.
The methods of parallel computing that the compiler itself supports - automatic loop parallelization and OpenMP style parallelization, both work only with a single address space (that is multiple processors on a single system).
Hope that helps you get started.