How to ... Run a Program

You may either run a program interactively from a shell window or by submitting a script to the batch system. For instructions on running on a given system, see the web page for that system.

Interactive Processing

Interactive jobs should mainly be used for program development and testing. Each machine has relatively strict limits on interactive memory usage and CPU time.

To run a serial program or command interactively, type the name of the program or command at the machine prompt (%). If the location of the program is in your command PATH, you can just type the filename, otherwise type the complete pathname. Since the current working directory (.) is not in the search path by default, the correct way to execute the program named my_program in the current directory is

 % ./my_program

Batch Processing

Most production runs should be executed as jobs under control of the batch system. Batch (non-interactive) jobs are run by submitting a text file (script) that contains a list of commands. The script must also specify various parameters such amount of CPU needs, memory requirements, etc.

There are different batch systems at NERSC. You should read the "Running Jobs" page on the system you intend to use. See NERSC Systems.