A PhD position is proposed at Onera, Châtillon (suburb of Paris) in the CFD
department DSNA.

The full title is "Analysis of modal and nodal discontinuous Galerkin methods
for high performance computing" (Ref. DSNA2013-01).

The use of Computational Fluid Dynamics (CFD) in Aeronautics is continuously
increasing due to decreasing computational costs and to more reliable numerical
results and physical models. However, obtaining accurate results with a
high-level description of physical phenomena in a short execution time requires
significant improvements in the CFD methods. Developing new solvers based on
high-order schemes and designing efficient algorithms for modern architectures
dedicated to high performance computing (HPC) turns out to be necessary to
tackle these new challenges. 

A CFD solver based on high-order discontinuous Galerkin (DG) schemes is under
development in our team. In recent years, these methods have become very popular
for the solution of nonlinear convection dominated flow problems. Indeed, these
methods offer high-order spatial accuracy, compactness and flexibility, and are
particularly well-suited to unstructured meshes, parallel computing, etc.
However, these may lead to long execution time and high memory requirements.

This work aims at investigating the properties of the DG methods in order to
improve their efficiency for complex flow simulations. This will be done by
considering different polynomial expansions (modal or nodal) and different time
integration strategies. An in-depth analysis of the properties of these methods
will be done in terms of performance (floating point operations, memory
accesses, cache miss rate, execution time, parallel efficiency at large scale).
Several optimization techniques will be implemented in relation with the HPC
architectural components.

The candidate should be motivated, dynamic, independent and willing to work in a
team. He (she)  should have finished or nearly finished a Master degree in
mechanical engineering, applied mathematics or computer sciences. Coding skills
are also necessary and knowledge in high-performance parallel computing will be
appreciated.

More information, contact and application at
http://www.onera.fr/fr/content/propositions-de-theses (-> MFE -> DSNA -> 
DSNA2013-01).