Computer algebra in scientific computation squarely faces the dilemma of natural mathematical expression versus efficiency. While higher-order programming constructs and parametric polymorphism proatural and expressive language for mathematical abstractions, they can come at a considerable cost. We investigate how deeply nested type constructions may be optimized to achieve performance similar to that of hand-tuned code written in lower-level languages.
Laurentiu Dragan, Stephen M. Watt