This volume contains a selection of papers which were presented during a Dagstuhl Seminar in the Internationales Begegnungs-und Forschungszentrum f ur Informatik at Dagstuhl Castle. The title of the Seminar was ``Parallel Processing in the Engineering Sciences ยฑ Methods and Applications''. Parallel computers are meanwhile a generally accepted tool in all parts of applied scientiยฎc research. Unfortunately the speed up of algorithms is in practice often not as huge as theoretically awaited.
The intention of the seminar was to bring together scientists from the ยฎeld of Numerical Analysis, Computer Science, Engineering and Natural Science, respectively, in order to discuss the state of the art and future developments of parallel processing in the applied sciences. The meeting provided a forum of exchange between these dierent research ยฎelds.
In 24 talks various parallel algorithms for dierent computer architectures and parallel software for mathematical modeling of real life problems were presented.
Due to the variety of attacked problems and to the many dierent existing parallel computers no uniform methodology has evolved during the last years. Therefore dierent methods have to be used for concrete problems. This will also be demonstrated by nearly all the papers of this volume.
For example, in contrast to standard ยฎnite element methods, the recursive substructuring technique assembles the element matrices recursively in several levels. For automated parallelization the method is using functional programming. Adaptivity requires dynamic load balancing.
Adaptivity of multigrid methods for partial dierential equations can be attached by the concept of hash-table storage techniques.
The ecient eigenvalue and singular value computation on shared memory machines is described by two dierent techniques depending on the information that is required.
Nonlinear problems from structural ยฎnite element analysis usually are tackled with Newton-like methods. For a parallel implementation domain decomposition methods are used.
Hyper-systolic algorithms represent a new class of parallel computing structures. Their application to N-body computations and distributed matrix multiplication is discussed.