Object-oriented programming for linear algebra

Abstract

Many numerical analysts and Lisp/Smalltalk programmers share the assumption that languages like Fortran are more appropriate for traditional, quantitative scientific programming than object-oriented languages. To show how straightforward application of object-oriented design to standard algorithms in numerical linear algebra improves clarity and expressiveness, without sacrificing speed or accuracy, I describe parts of Cactus, a system for numerical linear algebra and constrained optimization, implemented in Common Lisp and CLOS. © 1989, ACM. All rights reserved.