bertini.parallel

MPI parallelism support for Bertini2.

Requires an MPI installation (OpenMPI or MPICH) and mpi4py for distributed computing. When used without MPI, all functions return serial defaults (rank=0, size=1, is_manager=True).

bertini.parallel.rank() → int :

Return the rank of this process. Returns 0 in serial (non-MPI) builds.

bertini.parallel.size() → int :

Return the total number of processes. Returns 1 in serial (non-MPI) builds.

bertini.parallel.is_manager() → bool :

Return True if this process is the manager (rank 0). Always True in serial builds.

bertini.parallel.is_worker() → bool :

Return True if this process is a worker (rank > 0). Always False in serial builds.

bertini.parallel.initialize() → None :

Initialize MPI (or no-op if already initialized). Safe to call with mpi4py pre-init.

bertini.parallel.finalize() → None :

Finalize MPI (or no-op if already finalized).