Supercomputers for Superproblems: An Architectural Introduction

• Published in: Computer (Long Beach, Calif.) Calif.), 1984-03, Vol.17 (3), p.62-74
• Main Author: Norrie, C
• Format: Article
• Language: English
• Subjects: Aircraft; Computer architecture; Data engineering; Differential equations; Energy resolution; Large-scale systems; Navier-Stokes equations; Nuclear physics; Physics computing; Supercomputers
• ISSN: 0018-9162; 1558-0814
• DOI: 10.1109/MC.1984.1659083

An application involving large-scale computing operations is related to the solution of continuous field problems. One example for such a problem is the simulation of the aerodynamics around a jet aircraft on the basis of the Navier-Stokes or the Euler equations. Current supercomputers are still not adequate for the required calculations. However, there are a number of supercomputer architectures which would be suitable for such applications. The concept of a supercomputer is discussed, taking into account array processors and vector, or pipeline, processors. In the development of future general-purpose supercomputers, the new frontier may be a multiple-instruction, multiple-data, or MIMD, processor. Special-purpose supercomputers are designed to solve a particular problem or class of problems. These computers can be divided into two categories. In the first of these categories the architecture reflects the physical structure of the problem to be solved. In the second category, the architecture reflects the general solution method for that class of problems.