MPI and PETSc for Python

LISANDRO D. DALCÍN; PABLO A. KLER; MARIO A. STORTI; RODRIGO R. PAZ

Computational Mechanics Research Trends

Nova Science Publishers

Lugar: New York; Año: 2010; p. 453 - 493

This work reports our attempts to facilitate the access to high-performance parallel computing resources within a Python program-ming environment. The outcome of this effort are two open source andpublic domain packages, MPI for Python (known in short as mpi4py)and PETSc for Python (known in short as petsc4py).MPI for Python [1, 2, 3], is an open-source software project thatprovides bindings of the Message Passing Interface (MPI) standardfor the Python programming language and targets the development ofparallel application codes in Python. Its facilities allow parallel Pythonprograms to easily exploit multiple processors. MPI for Python employsany back-end MPI implementation, thus being immediately available onany parallel environment providing access to any MPI library.PETSc for Python [4] is an open-source software project that pro-vides access to the Portable, Extensible Toolkit for Scientific Compu-tation (PETSc) libraries within the Python programming language.Its facilities allow sequential and parallel Python applications to ex-ploit state of the art algorithms and data structures readily available inPETSc.