Parallel algorithms in computational science (Q1188833)

The book results from a one-semester course given at the University of Wuppertal and from research by the authors. Techniques and methods connected with computer simulations are presented particularly in the light of the possibilities opened by parallel computers. Special interest is devoted to problems in physics. The simulation methods relate especially to the Monte Carlo method (including importance of sampling) and molecular dynamics simulation with emphasis on lattice and polymer systems. Corresponding algorithms and the underlying hardware are developed. Accuracy considerations and finite-size problems, implementing the Monte Carlo algorithm for the Ising model, ideas behind the parallelization methods and their application to simulation problems are dealt with. Geometrically inspired parallelizations predominate. The geometric reduction or partitioning of systems is a key to perform simulation on parallel computers. Parallel programming languages (especially the Occam language) are discussed. In the Appendix parallel programs for the Ising model, random number generation and for molecular dynamics are listed.