Chemistry, computers, and microelectronics: Present and future prospects

Computer architecture is being driven by developments in the semiconductor industry. The market for mass‐produced microprocessors in video games, personal computers, real‐time control, etc., is leading to rapid advances in the complexity of circuitry that can be placed on a single silicon chip. Present estimates suggest that within ten years it may be possible to incorporate 107 transistors on a chip (most current mainframe CPUs contain less than 106 transistors). These developments in very large‐scale integration (VLSI) will have a major impact on computer science and, subsequently, on computational chemistry. Quantum chemistry has an essentially infinite demand for processing power. Economic ways of obtaining this processing power must take advantage of VLSI. One possibility that is being actively explored is to connect hundreds (and perhaps thousands) of mass‐produced microprocessors together to form a multiprocessor. Such a multiprocessor would then execute parallel versions of the existent serial algorithms of quantum chemistry. Developing parallel algorithms to take advantage of these new architectures will require a substantial revision of the computational methods of quantum chemistry. This article will review developments in microprocessors and their likely impact on quantum chemistry. The C.mmp and Cm* multiprocessors, built in the Computer Science Department of Carnegie‐Mellon University, will be described as well as present progress in designing parallel versions of some common quantum chemistry algorithms for execution on these machines, particularly Cm*.