Summary: Programming environments for scientific and engineering applications should combine all aspects of reliable computations on different computer architectures. Software implementation requires powerful programming tools providing, for example, a wide variety of data structures with operations in mathematical notation. More reliable calculations are possible by self-validating numerical algorithms which base on easy access to rounding control and sometimes on extensive use of accurate dotproduct operations. Combining floating-point algorithms and computer algebra systems can even improve the reliability of computations.
The minisymposium is intended to give insight into recent developments in programming environments for reliable computations and to report experiences with these.
Day: July 4, 1995 Time: 09.30
C.P. Ullrich, Universität Basel, Switzerland:
Scientific programming language concepts
S.M. Rump, Technische Universität Hamburg-Harburg, Germany;
R. Lindemann, Technische Universität Hamburg-Harburg, Germany:
C extension for scientific computing
B.M. Verdonk, University of Antwerpen, Belgium;
A. Cuyt, B. Naudts, J. Verelest, University of Antwerp, Belgium:
On the integration of software tools for scientific computation
B. Philippe, INRIA-IRISA, France;
J. Erhel, INRIA/IRISA, Rennes, France;
N. Mallejac, CEA, Villeneuve Saint Georges, France:
AQUARELS: a problem-solving environment for numerical quality --
review of the project and first use on realistic problems --
H. Burkhart, Universität Basel, Switzerland:
Software engineering techniques for parallel
computing in science and engineering
J.W. Wolff von Gudenberg, Universität Würzburg, Germany:
Structure of a C++ library for parallel accurate linear algebra