Theory of Parallel Programming. I. Survey of Practical Aspects

Abstract

The search for ways to enhance the performance of computer equipment eventually led, in the 1950s, to the notion of parallel data processing on multiprocessor parallel computer systems. These systems are made up of a number of computing units known as processors. Each processor executes, concurrently with the others, some fragment of the computation; the result produced by one processor can be made accessible to the others for further processing. The structure and organization of a multiprocessor computer system can vary: The processors may be of the same type or have different properties; each of the processors may have its own memory and they may exchange data over permanent or dynamically assigned channels, or the processors may all access one central memory through which data are exchanged; supervisory control of the concurrent operation of and exchange of information between processors may be centralized or allocated among the processors; finally, the system may have a hierarchic structure, in which each processor may be organized as a multiprocessor system.