Assessing the impact of largescale computing on the size and complexity of firstprinciples electromagnetic models
Abstract
There is a growing need to determine the electromagnetic performance of increasingly complex systems at ever higher frequencies. The ideal approach would be some appropriate combination of measurement, analysis, and computation so that system design and assessment can be achieved to a needed degree of accuracy at some acceptable cost. Both measurement and computation benefit from the continuing growth in computer power that, since the early 1950s, has increased by a factor of more than a million in speed and storage. For example, a CRAY2 has an effective throughput (not the clock rate) of about 10(exp 11) floating point operations (FLOPs) per hour compared with the approx. 10(exp 5) provided by the UNIVAC1. The purpose of this discussion is to illustrate the computational complexity of modeling large (in wavelengths) electromagnetic problems. In particular the point is made that simply relying on faster computers for increasing the size and complexity of problems that can be modeled is less effective than might be anticipated from this raw increase in computer throughput. It is suggested that rather than depending on faster computers alone, various analytical and numerical alternatives need development for reducing the overall FLOP count required to acquire the information desired. One approach is to decrease the operation count of the basic model computation itself, by reducing the order of the frequency dependence of the various numerical operations or their multiplying coefficients. Another is to decrease the number of model evaluations that are needed, an example being the number of frequency samples required to define a wideband response, by using an auxiliary model of the expected behavior.
 Publication:

Presented at the International Conference on Directions in Electromagnetic Wave Modeling
 Pub Date:
 1990
 Bibcode:
 1990dewm.conf...22M
 Keywords:

 Computer Systems Design;
 Computer Systems Performance;
 Convergence;
 Electromagnetic Fields;
 Mathematical Models;
 Supercomputers;
 Computerized Simulation;
 Differential Equations;
 Floating Point Arithmetic;
 Integral Equations;
 Parallel Processing (Computers);
 Specifications;
 Communications and Radar