The field of high performance computing is developing at an extremely rapid pace. Massively parallel computers offering orders of magnitude increase in performance are under development by all the major computer vendors. Many sites now have production facilities that include massively parallel hardware. Molecular modeling methodologies (both quantum and classical) are also advancing at a brisk pace. The transition of molecular modeling software to a massively parallel computing environment offers many exciting opportunities, such as the accurate treatment of larger more complex molecular systems in routine fashion, and a viable, cost effective route to study physical, biological, and chemical ``grand challenge'' problems that are impractical on traditional vector supercomputers. This will have a broad effect on all areas of basic chemical science at academic research institutions and chemical, petroleum and pharmaceutical industries in the U.S., as well as chemical waste and environmental remediation processes. But, this transition also poses significant challenges: architectural issues (SIMD, MIMD, local memory, global memory, etc.) remain poorly understood and software development tools (compilers, debuggers, performance monitors, etc.) are not well developed. In addition, researchers that understand and wish to pursue the benefits offered by massively parallel computing are often hindered by the lack of expertise, hardware, and/or information at their site.
A Conference and Workshop organized to focus on these issues was held at the National Institute of Health, Bethesda (February, 1993). The specific aim of the Conference was to provide a snapshot of current state-of-the-art and background material on the following:
This report is the culmination of the organized workshop and the efforts of the participants; it discusses issues (1) and (3) at some length, and provides an update of activities that have taken place since the workshop . The main conclusion of the report is that a drastic acceleration in the present rate of progress is required for the chemistry community to be positioned to exploit fully the emerging class of Teraflop computers, even allowing for the significant work to date by the community in developing software for parallel architectures.