Parallel processing and expert systems

The author gives an example of how to write a simple parallel program and explains the significance of the communications overhead. This leads to the idea of course grain and fine grain implementation. The author then explains the point-to-point communications system, called link in the Inmos Transputer, for sending data between programs and the four links per processor limit of the Transputer. Another method of sending data is by sharing memory between processor. The author´s department intends to build a computer based on the T800 Transputer but with a distributed memory architecture. The implementation (which will use a Unix like operating system from Perihelion Software) should be efficient for expert systems and large databases