Some HPC challenges for multi-physics extended CFD computations
| Autor: | Vuillot, F., Refloch, A. |
|---|---|
| Přispěvatelé: | ONERA - The French Aerospace Lab [Châtillon], ONERA-Université Paris Saclay (COmUE), André, Cécile |
| Jazyk: | angličtina |
| Rok vydání: | 2014 |
| Předmět: | |
| Zdroj: | The 16th IEEE International Conference on High Performance Computing and Communications (HPCC 2014) The 16th IEEE International Conference on High Performance Computing and Communications (HPCC 2014), Aug 2014, PARIS, France |
| Popis: | International audience; As the numerical simulations become more and more used for prediction and analysis of complex non stationary flows and as the computer architectures steadily evolve towards massively parallel, there is a real need to adapt computational codes to make them ready for intensive use in a HPC cluster environment. In the same time there is a strong tendency in the CFD community to enlarge the scope of computations by associating and possibly by directly coupling different physics in a unique computation. Such multi-physics computations open the way to needed sizing, analysis and optimization of complex systems. Common examples are fluid/solid interactions (conjugate heat transfer, aeroelasticity, aeromechanics), aeroacoustics, two-phase flows, combustion. This reality puts severe demands especially for multi-physics codes that are at stakes to provide HPC performances while addressing several physics that are discretized on the same computational domain. The CEDRE code developed at ONERA as the reference code for energetics and propulsion is particularly concerned by these challenges. The present version CEDRE 5.1 is already daily used on clusters with thousands cores with a very good scalability. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|