NERSCPowering Scientific Discovery Since 1974

Using Hybrid MPI/OpenMP

Using OpenMP

Franklin has 4 cores sharing the memory on each node. OpenMP is supported within the node. To use OpenMP, a specific compiler option to turn on OpenMP is needed to compile the code with each compiler.  A torque batch script need to specify keywords "-l mppnppn=1" (use single-core for one MPI task per node) and "-l mppdepth=4" (use 4 threads per MPI task). Also need to set OpenMP environment variable OMP_NUM_THREADS to 4, and use "-N 1 -d 4" option for the "aprun" command.

More details on how to compile and run with different programming environment and supported thread level etc. can be found at Using Hybrid MPI/OpenMP.