Enabling technologies for robust high-performance simulations in computational fluid dynamics.

Abstract

In computational science and engineering, the role of computer science includes the mechanical generation of programs for the fast computation of accurate derivatives and the efficient utilization of parallel computer architectures. Automatic differentiation and parallel computing are two technologies enabling robust high-performance simulations in various scientific and engineering disciplines. This article gives a survey of selected results of an interdisciplinary research project where automatic differentiation, parallel computing, and their interplay are investigated in the context of two computational fluid dynamics packages developed at RWTH Aachen University.