Parallel Numerical Algorithms (Prentice-Hall International Series in Computer Science)

With the increasing use of more powerful multiprocessor computer systems comes the need to develop parallel implementations of numerical methods which were originally developed for use with uniprocessors. Focusing on shared and local memory MIMD parallel computer systems, this volume is designed to help applications programmers make the not-so-straightforward transition from a serial to a parallel environment. It covers both synchronization and message-passing paradigms; discusses code development using parallel extensions to Fortran; emphasizes linear algebra, but covers other areas - ordinary differential equations, optimization - in detail.