How changing high performance computing technology changes the way in which we do computational chemistry

Developments in Computational Quantum Chemistry have been driven by the developments in hardware and software technology. Quantum chemical applications use 30-40% of the computational resources available in National and Regional supercomputer centers, therefore, it is important to discuss in the SUPERCOMPUTING ’92 forum how new technological developments in parallel and distributed computing, will affect the computational methodologies used in Quantum Chemistry. This panel will focus on “ab initio” types of quantum chemical calculations, “Traditional” quantum chemical applications have been built around external files containing the so called “two electron integrals”, These integrals are computed once and then stored in auxiliary storage devices. Thereafter they are used multiple times during the calculations and well developed methods exist to make efficient use of hardware and software using this approach. 111 recent years CPU technology has advanced much faster than storage technology, therefore new methods (“direct” methods) in which the two electron integrals are calculated when needed, “on the fly”, have been introduced. Of course this is a trade off of 1/0 for integrals may calculated many 1063-9535/92 $3.00 Q 1992 IEEE CPU; these s am ~ be needed and times. While these techniques are not yet as well developed as their predecessors, big advances have been made in recent years. The use of “traditional” vs “direct” techniques is very much dictated by the trade off governed by the cost, availability, and performance of different computer technologies, mainly CPU and auxiliary storage. This panel will discuss this trade off in view of recent computational advances such as parallel/distributed computing (hardware and software technology) and new storage technologies (for example new disk array technologies).