...
binaries needs to be compiled using mpi compilers.
We have intel mpi and sgi mpi.
intel mpi
=========
####For intel mpi, use the following syntax####
#module load intel/2019up5/mpi
#mpiexec -n $PROCS $PROGRAM NAME < $INPUT FILE >& $OUTPUT FILE
module load intel-fc-11/12.0.2.137
module load intel-mpi/4.0.1.007
...
| No Format |
|---|
module load mpi/openMPI/1.4.3-gnu
###For openmpi, use the following syntax####
#module load mpi/openmpi/4.0.2
#mpiexec $PROGRAM NAME < $INPUT FILE >& $OUTPUT FILE |
Sample run
openMPI has a different mechanism to pass the local environmental variables from master node to slave nodes. You have three options:
...
As this is not the most intuitive command, the following table is provided as guidance as to how this command works:
select | ncpus | mpiprocs | description |
|---|---|---|---|
4 | 8 | 8 | 32 Processor job, using 4 nodes and 8 processors per node |
4 | 4 | 4 | 16 Processor job, using 4 nodes and 4 processors per node |
16 | 1 | 1 | 16 Processor job, using 16 nodes running 1 mpi process per processor and utilising 1 processor per node |
16 | 8 | 8 | 128 Processor job, using 16 nodes and 8 processors per node (each running an mpi process) |
Ref
1. http://www.hpcu.uq.edu.au/hpc/content/view/225/34/
2. http://www.oerc.ox.ac.uk/computing-resources/osc/support/documentation-help/job-schedulers/pbs/pbs-job-submission-scripts
3. http://www.cardiff.ac.uk/arcca/services/equipment/User-Guide/user-guide.html
...