Tools for Assessing Multiprocessing

Abstract

In the past few years, the computer industry has seen a dramatic increase in the number and variety of hardware architectures with parallel processing capabilities. These computer systems vary dramatically in terms of the size and number of processors, the type of the processor (scalar or vector), the memory structure, the processor/memory interconnect strategy, and the synchronization hardware. Parallel processors tend to fall into several different categories. Many systems fall into the Parallel Von Neumann category, where the system is a collection of processors acting on a single global shared memory. Larger numbers of processors can be connected, if the system is composed of a cluster of processors acting on a shared cluster memory and having connections to other Clustered processors. A further extension of this approach isolates each processor/memory pair in a completely Distributed system much like the Floating Point System’s T-Series. Other categories, which will not be discussed here, are Data Flow, Systolic Array, and Special Purpose systems designed for specific applications.