connecting to input/output files

If your job reads from input files or writes to output files then you need to specify them with a full path. This is because the SGE job script is launched from your home directory on a compute node. An alternative approach is to put a change of directory at the top of your job script.

For example suppose you want to execute a code from the directory /users/nick/TESTING, assuming that this directory contains input files for the code xdlu specified without a path (i.e the code assumes to find the input files in the current directory) then a suitable PBS job script would look like:

#! /bin/sh
cd /users/nick/TESTING
./benchmark

You can use the same method for both SCore MPI and non-SCore jobs.