ABSTRACT
This chapter provides a basic introduction to the present status of parallel computing, and to some of the key issues in it. The three forms of parallelism which are common in many problems are: event parallelism; geometric parallelism; and algorithmic parallelism. Parallel systems have some inbuilt advantages in cost effectiveness; they typically do not demand large numbers of special purpose chips; are amenable to standard very large scale integrated design methods; and the replication of components provides the opportunity of economies of scale even in a single machine. The present state of system and compiler technology is such that in all supercomputer or novel architecture systems, although to very different degrees, considerable effort must be made by the programmer if a performance approaching the machine’s peak is to be attained. The string dynamics is intrinsically parallel, and has been implemented by a straightforward ‘domain decomposition’ on the computing Surface.
