SGE parallel environment (PE) is used to specify how multiple CPU cores are allocated across one or more nodes. The syntax to specify PE is:
-pe pe_name number_of_cpu
where pe_name is one of the PE names (details below) and number_of_cpu is the total number of CPU cores you intend to use in the computation. There is no need to specify “-pe” if the job is sequential (i.e. using only 1 CPU core).
- Use “shared” for pe_name. All of the CPU cores will be on the same node.
- Example: to request 8 CPU cores: -pe shared 8
- The maximum number of CPU cores you can use is 16, or the highest per-node core count of the compute nodes that you have access to. A job may not start if the maximum is exceeded.
- Use “dc*” for pe_name. The CPU cores may be distributed on one or across multiple compute nodes.
- Example: to request 32 CPU cores (across multiple nodes): -pe dc* 32
- If you use “-pe dc* 32” on the command line, add the escape character “\” before the “*”, e.g. “-pe dc\* 32”