Supercomputers for Superproblems: An Architectural Introduction

The author examines various supercomputer architectures that might be suitable for problems involving mega, giga, and possibly tera numbers of floating-point arithmetic operations. He describes the two main supercomputer architectures that now exist for the efficient handling of matrices: (1) array processors, also known as single-instruction, multiple-data (SIMD) processors and (2) vector processors, also known as pipeline processors, and shows how they may be incorporated into both special-purpose and general-purpose machines. 21 references.