The Influence of the Los Alamos and Livermore National Laboratories on the Development of Supercomputing

Abstract

This paper begins by reviewing the development of supercomputer architectures. It identifies a "mainstream" characterized by the gradual transformation of an originally sequential architecture in the direction of various forms of parallelism, and a set of "radical" alternatives all involving parallelism of a greater degree and different kind. The paper then describes the nature and enormous magnitude of the computational tasks of nuclear weapons design, particularly the two different problems of mesh computation and Monte Carlo modelling. The importance of the Los Alamos Laboratory in the early development of digital computing is summarized, and the role subsequently played by Livermore and Los Alamos as customers for and sponsors of supercomputing is then described. It is concluded that the Laboratories have played a major part in shaping what we understand by "supercomputer" - the set of fastest floating-point arithmetic processors permitted by the state-of-the-art at a given point in time. Paradoxically, though, their more specific architectural influences are small by comparison with their overall influence. The paper finishes with two speculations, the first concerning Seymour Cray, virtuoso of the robust supercomputer architecture, the second concerning the finding that the task of nuclear weapons design inclines the Laboratories to the support of mainstream rather than radically parallel architectures.