We present Matrix Distributed Processing 1.0, a collection of classes and functions written in C++ for implementing a certain type of numerical algorithms on parallel computers and cluster of PCs. Any problem defined in terms of field variables (on a discretized compact space/grid) and a system of coupled partial differential equations (in the field variables) can be implemented, rapidly and optimally, using MDP 1.0. MDP 1.0 can simulate multiple grids, each of them of arbitrary dimension, size and topology. Multiple fields can be defined on each grid including scalar fields, vector fields, tensors and spinors. MDP 1.0 provides the functions for distributing the field variables and the computation over the different processes and deals automatically with communication issues. As an example of application a full Lattice QCD simulation program has been developed at Fermilab using MDP 1.0. Some other possible applicative fields include classical electrodynamics, thermodynamics and astrophysics. Documentation is available at http://xxx.lanl.gov/abs/hep-lat/0004007

Back to Program

Back to ACAT Web page