Rodrigo R. Paz

MPI for Python

MPI for Python provides bindings of the Message Passing Interface (MPI) standard for the Python programming language and allows any Python program to exploit multiple processors. This package is constructed on top of the MPI-1 specification and defines an object-oriented interface which closely follows MPI-2 C++bindings. It supports point-to-point (sends, receives) and collective (broadcasts, scatters, gathers) communications of general Python objects. Efficiency has been tested in a Beowulf class cluster and satisfying results were obtained. MPI for Python is open source and available for download on the web (http://www.cimec.org.ar/python).

MPI for Python: Performance improvements and MPI-2 extensions

MPI for Python provides bindings of the message passing interface (MPI) standard for the Python programming language and allows any Python program to exploit multiple processors. In its first release, MPI for Python was constructed on top of the MPI-1 specification defining an object-oriented interface that closely followed the MPI-2 C++ bindings, and provided support for communications of general Python objects. In the latest release, this package is improved to enable direct blocking/non-blocking communication of numeric arrays, and to support almost all MPI-2 features. Improvements in communication performance have been tested in a Beowulf class cluster. Results showed a negligible overhead in comparison to compiled C code. MPI for Python is open source and available for download on the web (http://mpi4py.scipy.org/).

A preconditioner for the Schur complement matrix

Abstract A preconditioner for iterative solution of the interface problem in Schur Complement Domain Decomposition Methods is presented. This preconditioner is based on solving a global problem in a narrow strip around the interface. It requires much less memory and computing time than classical Neumann–Neumann preconditioner and its variants, and handles correctly the flux splitting among subdomains that share the interface. The aim of this work is to present a theoretical basis (regarding the behavior of Schur complement matrix spectra) and some simple numerical experiments conducted in a sequential environment as a motivation for adopting the proposed preconditioner. Efficiency, scalability, and implementation details on a production parallel finite element code [Sonzogni V, Yommi A, Nigro N, Storti M. A parallel finite element program on a Beowulf cluster. Adv Eng Software 2002;33(7–10):427–43; Storti M, Nigro N, Paz R, Dalcin L. PETSc-FEM: a general purpose, parallel, multi-physics FEM program, 1999–2006] can be found in works [Paz R, Storti M. An interface strip preconditioner for domain decomposition methods: application to hydrology. Int J Numer Methods Eng 2005;62(13):1873–94; Paz R, Nigro N, Storti M. On the efficiency and quality of numerical solutions in cfd problems using the interface strip preconditioner for domain decomposition methods. Int J Numer Methods Fluids, in press].

GPGPU implementation of the BFECC algorithm for pure advection equations

In the present work an implementation of the Back and Forth Error Compensation and Correction (BFECC) algorithm specially suited for running on General-Purpose Graphics Processing Units (GPGPUs) through Nvidia’s Compute Unified Device Architecture (CUDA) is analyzed in order to solve transient pure advection equations. The objective is to compare it to a previous explicit version used in a Navier-Stokes solver fully written in CUDA. It turns out that BFECC could be implemented with unconditional stable stability using Semi-Lagrangian time integration allowing larger time steps than Eulerian ones.