New science from high-performance computing: an introduction.

Abstract

Computational methods have for many years had a major impact on a wide range of scientific disciplines including physical, biological and environmental sciences and engineering. Indeed, the ability to model and simulate complex systems is an indispensable component of contemporary scientific research. Modern computation employs a hierarchy of systems which may be viewed as a pyramid, at the base of which are desktop/workstation facilities, while the apex is provided by 'high-end' or highperformance computing (HPC). Of course, today's HPC is tomorrow's mid-range computing, and the HPC of the 1980s had the same capabilities as contemporary desktop facilities. But at any one time, HPC defines many of the major challenges and opportunities in the field. HPC technology is currently advancing particularly rapidly, being driven in part by major initiatives, most notably the ASCI (Advanced Simulation and Computing Initiative) in the US. 'Teraflop' systems (in which total processor power achieves 1012 floating-point operations per second) are now a reality. With these developments comes an exciting range of opportunities and challenges for computational science. Let us consider first some of the characteristics of the current field of scientific applications of HPC. Three areas deserve particular emphasis. The first is the range of science encompassed by the field; second is the increasing tendency for modelling methods to investigate complex, real systems in a predictive manner. The third is the developing ability of computational methods, especially with HPC, to model whole systems, rather than one component of a complex multi-component system. The range of HPC-enabled science is amply illustrated by the articles in the present issue, and includes the following (non-exhaustive) list of topical themes: protein structures, materials design, atmospheric science, deep earth (mantle and core), drug design, fluid dynamics, ocean and atmosphere, high-energy physics and cosmology. And in addition to these topics in physical, biological and engineering sciences, there is a growing recognition of the potential importance of HPC in economic and social sciences, as illustrated by the article of Doornik in this issue (Doornik et al. 2002). The issues relating to complexity, realism and predictive power will again be apparent in every article in the issue, but can be illustrated by a number of recently reported studies.