High performance computing for direct turbulence simulation in computational fluid dynamics

The remarkable advances obtained in prediction and design in computational fluid dynamics (CFD) which requires scalable computing power in terms of speed and memory can currently be sustained with parallel supercomputers. This paper presents cooperative work between ONERA (France) and NAL (Japan) to investigate the potential of distributed environments for high performance computing in CFD. For this purpose a Navier-Stokes flow code was integrated into a distributed processor environment with two parallelization strategies: one based on domain partitioning with message passing; and another one based on loop-parallelization with compiler directives. We present in this paper the results obtained on different computers such as CRAY, PARAGON and NAL NWT supercomputers.