Towards a Modular and Variability-Aware Aerodynamic Simulator.

Computational Fluid Dynamics (CFD) consists of numerically solving the fluid dynamics equations and has become a major tool in designing and evaluating any physical structures, like airplane, rotors, or even nuclear plants, where the flow of a fluid can be a critical efficiency or security aspect of these structures. Our first contribution is a brief review of the core characteristics a CFD solver should have (based on two common functionalities they usually provide) and the state of the art of CFD tools. Indeed, research on this field principally focuses on specific numerical or computation methods, software architecture is rarely discussed. Moreover, to the best of our knowledge, all CFD tools have major structural flaws that limit their capacities to integrate new methods and take advantage of new hardware. Our second contribution is a new approach that aims to solve these flaws. We exploit formal methods (namely, order-sorted algebra and Delta-Oriented Programming) to build a flexible CFD framework in which new methods can be added as modules. By exploiting dataflow automatic generation, our approach adds no runtime overhead. We implemented our approach and tested it on a simple example.