Instructors: Christopher Beattie, Department of Mathematics and Calvin Ribbens, Department of Computer Science.
More information on UH3004: High Performance and Scientific Computing
This Honors Colloquium in "High Performance Scientific Computing" is being offered for the first time in Spring, 1995. Co-taught by Professors Chris Beattie (Math) and Cal Ribbens (Computer Science), the course is based on materials developed at the University of Colorado. The focus is on computational science, vector and parallel computers, and scientific visualization.
The past decade has seen the rise of computational science as a major component of science and engineering. While computing has been an important tool in many disciplines for almost fifty years, the role of computation has grown especially significant with the recent widespread availability of large-scale, high-performance computers (i.e., "supercomputers," "massively parallel computers," etc.). For example, aircraft designers now use computational models to validate, and in some cases replace, wind tunnel experiments to study air flow around proposed designs. In this and many other disciplines computational methods have joined experimental and theoretical approaches as a third major paradigm for doing science and engineering.
The main goal of this course is to give students an understanding of B high-performance computing systems, of the algorithms designed to run on them, and of the important issues which must be dealt with in modern large-scale scientific computation. Furthermore, we are studying these systems, algorithms, and issues without divorcing them from the motivating problem contexts. High-performance scientific computing is an exciting and rapidly-changing field. We hope to give a sense of where computation fits into the work of science, and of where science fits into the work of computing.
Given the time constraints of one semester we cannot hope to cover a great many topics in detail. Instead our approach is to choose two high-performance machines and two (nearly) real-world problems, and study them in some depth. The machines we at are the Cray Y-MP and the Intel Paragon. The problems are taken from molecular dynamics and advection. We also examine the role of high-end workstations in modern scientific computing. Finally, since virtually all nontrivial scientific computations produce and/or consume large amounts of data, we introduce some concepts and tools used in the increasingly important field of scientific visualization.