Short English Information On U. Kulisch#

Ulrich Kulisch was director of the Institute of Applied Mathematics at the University of Karlsruhe for over 30 years.

Prof. Kulisch is well known in the area of floating-point arithmetic. But before his dedication to this subject he had performed other pioneer work. One example is the introduction of computer science teaching in 1968, which later led to the foundation of the Faculty of Informatics at the University of Karlsruhe, the first one in Germany.

Later he was probably the first one who gave students in his programming classes access to a kind of personal computer, while batch processing of punch cards and punch tapes were standard. These were Z80 based systems running CP/M, and the Pascal compiler resided on a single floppy disk.

In his publications Prof. Kulisch laid the foundation for a theoretical approach to computer arithmetic by defining the mathematical concept of a semimorphism. One crucial point was the application of this theory not only to operations on real numbers, but also on vectors and matrices as well as on complex numbers. This led to requirements on the implementation of floating-point arithmetic on computers. Several software and hardware solutions were developed.

The usefulness of this theory is proven by the fact that algorithms exist which allow the computation of narrow bounds for the solution of mathematical problems in spite of the rounding errors inherent in floating-point arithmetic. Implementing these algorithms is facilitated by programming languages which allow easy access to the features of the floating-point arithmetic. Intervals for scalars, vectors, and matrices are intrinsic data types in these languages and arithmetic operations are defined for them. Thus Prof. Kulisch has always been interested in programming languages.

Extensions to Pascal, Fortran, and C have been defined and implemented. Over time a rich library of programs has been built which deliver guaranteed bounds for solutions of a wide spectrum of mathematical problems. Prof. Kulisch has always been successful in finding partners for his projects. He is easily able to convince an audience of the importance of doing mathematics on a computer in the right way.

In his presentations there are always examples where standard methods for solving a mathematical problem on a computer deliver huge errors without any warning. Several collaborations with industry resulted from his activities. A long running cooperation was the development of ACRITH with IBM Boblingen. This was a Fortran based package for S/370 systems providing the base floating-point arithmetic along with a library of problem-solving functions. There was hardware assist available on several S/370 systems.