Type of project:
Diplomarbeit 10/28/02 - 02/28/03

Peter Häfliger, student of Computational Science and Engineering (D-MATH)

Exploration of the Suitability of O-O Techniques for the Design and Implementation of a Numeric Math Library using Eiffel



The diploma thesis investigates the `Scientific Software Paradox', a problem stated by Bertrand Meyer in his foreword to [1]: Scientific Software has benefited less from the advance of software technology in the last three decades than other applications of software engineering. A number of reasons can be found for this paradox. On the one hand, there are the age and tradition of the field. Early leadership in the development of software technology (first high-level languages like FORTRAN and Algol) has allowed the field to reach a comparatively high level at an early stage. Such a head start always bears the danger of ignoring competitors which might catch up and eventually get ahead. On the other hand, management information systems are built by software engineers, compilers are built by software engineers, scientific software is built by -- mathematicians, nuclear physicists, naval engineers, all highly educated scientists in their respective discipline, but usually only roughly familiar with the state of the art in software technology.

In the first half of the ninety-nineties, Paul Dubois created EiffelMath, a numeric math library which presents a well-designed object-oriented interface to the user but delegates the actual computations to the respected procedural NAG C library.

The diploma thesis shows how a purely object-oriented approach was successfully applied to the construction of a component for numeric integration. Of vital importance is the practical proof (through benchmark tests) that the performance price for using object technology is low. It discusses the benefits and possible limitations of object technology for scientific computing.

[1] Paul F. Dubois: Object Technology for Scientific Computing, PTR Prentice Hall, 1996.

Diploma thesis (PDF)