Pypar is an efficient but easy-to-use module that allows programs written in Python to run in parallel on multiple processors and communicate using message passing. Pypar provides bindings to a subset of the message passing interface standard MPI.
Note that it offers only a subset of MPI commands and lacks groups, communicators and non-blocking sends and receives. MPI4Pyoffers a more fully-featured MPI implementation although pypar is easier to use.
Note also that due to the dynamic typing of Python care must be taken with receive commands to ensure the correct type is seen by this command as the syntax is
variable=pypar.receive(source=x,tag=y) and the type is inferred. Thus if a particular size of floating point object is required (float or double) it must be made clear by assigning variable to an example, or in creating arrays before calling the receive.
module load site-local-development
module load python/pypar/2.1.4_94/py27/02
Example Job Scripts
module load site-local
This version uses an out-of-date MPI linkage, and should not be used. The associated module will be withdrawn in future.