Parallel Algorithms in Computational Sciences

Aims to show physicists, mathematicians, chemists and biologists, who work with, or plan to work with parallel computers, how to make optimum use of the available parallelism when designing numerical simulations.

---

The authors present techniques and algorithms in computational science that have been developed for parallel computers. The numerical methods that have evolved in recent years to investigate problems in physics, mathematics, chemistry and biology by means of computer simulations are now being extended to make use of the opportunities opened up by the development of parallel computers. To utilize this potentially vast increase in computer power, it is necessary to devise algorithms that are appropriate to the particular problem being investigated. This volume combines a detailed consideration of the computer architecture (hardware and software feature as well as inter-processor communication schemes) with an understanding of the physical system being studied. These two aspects together make it possible to develop techniques that make optimum use of the available parallelism. Attention is focused on the Monte Carlo and molecular dynamics methods, and the techniques and algorithms are illustrated with examples from the statistical mechanics of spin systems, polymers and other many-particle systems.