ON THE COMPARISON OF NUMERICAL METHODS

Abstract

Serious comparisons of numerical methods are important for scientists who develop new codes as well as for those who use programs. Historical considerations show some errors which were made in the past and should be avoided in the future. Every numerical code is based not only on numerical but also on analytical considerations. Both of them have to be taken into account. As a result, benchmarks for complicated topics (numerical calculations of electromagnetic fields) should give more information than just numbers like 'speed', 'memory requirement', etc. [Vol. 3, No. 2, pp. 79-82 (1988)]