PREFACE

In 1983, Los Alamos National Laboratory cosponsored the first Frontiers of Supercomputing conference and, in August 1990, cosponsored Frontiers of Supercomputing II: A National Reassessment, along with the National Security Agency, the Defense Advanced Research Projects Agency, the Department of Energy, the National Aeronautics and Space Administration, the National Science Foundation, and the Supercomputing Research Center.

Continued leadership in supercomputing is vital to U.S. technological progress, to domestic economic growth, to international industrial competitiveness, and to a strong defense posture. In the seven years that passed since the first conference, the U.S. was able to maintain this lead, although that lead has significantly eroded in several key areas. To help maintain and extend a leadership position, the 1990 conference aimed to facilitate a national reassessment of U.S. supercomputing and of the economic, technical, educational, and governmental barriers to continued progress. The conference addressed events and progress since 1983, problems in the U.S. supercomputing industry today, R&D priorities for high-performance computing in the U.S., and policy at the national level.

The challenges in 1983 were to develop computer hardware and software based on parallel processing, to build a massively parallel computer, and to write new schemes and algorithms for such machines. In the 1990s, the dream of computers with parallel processors is being realized. Some computers, such as Thinking Machines Corporation's Connection Machine, have more than 65,000 parallel processors and thus are massively parallel.

Participants and speakers at the 1990 conference included senior managers and policy makers, chief executive officers and presidents of companies, computer vendors, industrial users, U.S. senators, high-level federal officials, national laboratory directors, and renowned academicians.

The discussions published here incorporate much of the widely ranging, often spontaneous, and invariably lively exchanges that took place among this diverse group of conferees.

Specifically, Frontiers of Supercomputing II features presentations on the prospects for and limits of hardware technology, systems architecture, and software; new mathematical models and algorithms for parallel processing; the structure of the U.S. supercomputing industry for competition in today's international industrial climate; the status of U.S. supercomputer use; and highlights from the international scene. The proceedings conclude with a session focused on government initiatives necessary to preserve and extend the U.S. lead in high-performance computing.

Conferees faced a new challenge—a dichotomy in the computing world. The supercomputers of today are huge, centrally located, expensive mainframes that "crunch numbers." These computers are very good at solving intensive calculations, such as those associated with nuclear weapons design, global climate, and materials science. Some computer scientists consider these mainframes to be dinosaurs, and they look to the powerful new microcomputers, scientific workstations, and minicomputers as the "supercomputers" of the future. Today's desktop computers can be as powerful as early versions of the Cray supercomputers and are much cheaper than mainframes.

Conference participants expressed their views that the mainframes and the powerful new microcomputers have complementary roles. The challenge is to develop an environment in which the ease and usefulness of desktop computers are tied to the enormous capacity and performance of mainframes. Developments must include new user interfaces, high-speed networking, graphics, and visualization. Future users may sit at their desktop computers and, without knowing it, have their work parceled out to mainframes, or they may access databases around the world.

Los Alamos National Laboratory and the National Security Agency wish to thank all of the conference cosponsors and participants. The 1990 conference was a tremendous success. When the next Frontiers of Supercomputing conference convenes, the vision of a seamless, comprehensive computing environment may then be a reality. The challenge now is to focus the energies of government, industry, national laboratories, and universities to accomplish this task.