Open Access Open Access  Restricted Access Subscription Access

High Performance Computational Engineering: Putting the ‘E’ Back in ‘CSE’

D. J. MAVRIPLIS

Abstract


Over the last decade, advocacy for high performance computing has increasingly been taken up by the science community with the argument that computational methods are becoming a third pillar of scientific discovery alongside theory and experiment. On the other hand, the capabilities of computational engineering problems have become increasingly stagnant as efforts have tended to concentrate on reducing simulation cost rather than increasing simulation capability. Computational engineering problems differ in many aspects from discovery-based computational science problems, and specific techniques for addressing these differences must be developed in order for engineering problems to benefit from the upcoming paradigm shift to massively parallel multicore architectures. Arguing that computational fluid dynamics in general, and computational aerodynamics in particular, represent some of the most advanced computational engineering applications, we identify specific barriers impeding the deployment of these problems to current and future HPC architectures and suggest several strategies for overcoming these impediments. The long term objective is to enable the development of significantly more capable production level computational fluid dynamic engineering tools by harnessing the power of the upcoming generation of massively parallel multicore architectures.

Full Text:

PDF