Keywords: high-performance-computing, computational fluid dynamics, lattice Boltzmann, general purpose graphics processing units.

In recent years it has become obvious that modern computer hardware essentially will rely on multi- and many-core architectures to increase computational performance. Yet, only very few computational frameworks are ready to meet this challenge from an algorithmic and software engineering point of view. This chapter gives an overview about the recent work undertaken by our group in the context of high-performancecomputing (HPC) simulations in the area of computational fluid dynamics (CFD) in civil engineering. We will address different aspects of the simulation pipeline ranging from the efficiency of explicit numerical kernels capable of exploiting modern central processing units (CPU) as well as general purpose graphics processing units (GPGPU) architectures as well as domain-decomposition aspects for local time stepping schemes on Eulerian grids and the importance of hardware aware data structures. Our basic modeling approach for CFD is based on various lattice-Boltzmann models (LBM) tailored for turbulent flows. After a short introduction into the basic methods we will discuss several computational examples of complex flow simulations with up to five billion degrees of freedom. These examples demonstrate the feasibility of LBM for complex flow problems and the efficiency of GPGPU based simulations which allow simulations with more than a billion degrees of freedom (DOF) on desktop systems. Thus kinetic techniques for transport simulation may play a more prominent role in computational engineering due to their intrinsic suitability for present and future multi- and many-core architectures.

purchase the full-text of this chapter (price £20)

go to the previous chapter
go to the next chapter
return to the table of contents
return to the book description
purchase this book (price £85 +P&P)