Iue Annual Review 2006

Biography:
René Heinzl was born in Vienna, Austria, in 1977. He studied electrical engineering at the Technische Universität Wien, where he received the degree of Diplomingenieur in 2003. He joined the Institute for Microelectronics in November 2003, where he is currently working on his doctoral degree. In April 2005 he achieved first place at the doctoral competition at the EEICT in Brno. His research interests include modern programming paradigms, high performance programming techniques, data structural aspects of scientific computing, performance analysis, process simulation, solid modeling, scientific visualization, and mesh-related tasks for TCAD.

A Generic Scientific Simulation Environment (GSSE)

In the last decades different libraries and software environments have been developed to handle various areas in the field of scientific computing. Due to the great diversity of physical phenomena, which can be described by differential equations of considerable complexity, present in different areas of scientific computing, the development of several discretization schemes has been necessary in order to best model the underlying physics and to accommodate the mathematical peculiarities of each of these equations while transferring them to the discrete world of digital computing. The recurring tasks of utilizing topological mechanisms and the development of corresponding data structures are tedious and error-prone due to code replication. The formal description of discretized equations combined with efficiency considerations, in particular in higher dimensions, has not been supported by a library up until now.
This unnecessarily complicates the actual research and software development step as well as maintenance. From all approaches developed at our institute and from other research groups we have extracted the basic concepts and have developed an orthogonal, layer-based, generic scientific simulation environment, GSSE. The most basic concept is the rigorous separation of traversals and data access, as in the C++ STL. Whereas the topological traversal attends the question how to get the requested data, the data manipulation aspect is reflected by our functional specification approach. Finally, the generic programming paradigm consolidates these two aspects of getting and manipulating different kinds of data in a dimension- and topologically neutral way. The generic programming paradigm and the C++ parametric polymorphism establish homogeneous interfaces between algorithms and data structures. Functional programming eases the specification of equations and offers easily extensible expressions while retaining the functional dependence of formulae due to higher order functions.

The following subjects are integrated into this new approach:

Solid modeling, mesh generation, and adaptation
A priori and a posteriori error estimation
Topology and functional equation specification for FD, FV, and FE
Fully generic and functional programming approach
Realtime visualization

With respect to performance analysis we can show that our highly expressive code in C++, which runs successfully on different platforms and computer architectures, does not show any abstraction penalty.

Multidimensional simulation with a two- and
three-dimensional solution for the Laplace equation.

Performance analysis of different functional description possibilities.