|
|
Computational Science, Engineering & Technology Series
ISSN 1759-3158
CSETS: 31
DEVELOPMENTS IN PARALLEL, DISTRIBUTED, GRID AND CLOUD COMPUTING FOR ENGINEERING
Edited by: B.H.V. Topping and P. Iványi
Chapter 8 A Parallel Incompressible Navier-Stokes Solver: Implementation Issues G. Houzeaux, H. Owen, B. Eguzkitza, C. Samaniego, R. de la Cruz, H. Calmet, M. Vázquez and M. Ávila CASE - Physical and Numerical Modelling, Barcelona Supercomputing Centre, Barcelona, Spain Full Bibliographic Reference for this chapter
G. Houzeaux, H. Owen, B. Eguzkitza, C. Samaniego, R. de la Cruz, H. Calmet, M. Vázquez, M. Ávila, "A Parallel Incompressible Navier-Stokes Solver: Implementation Issues", in B.H.V. Topping and P. Iványi, (Editor), "Developments in Parallel, Distributed, Grid and Cloud Computing for Engineering", Saxe-Coburg Publications, Stirlingshire, UK, Chapter 8, pp 171-201, 2013. doi:10.4203/csets.31.8
Keywords: incompressible flow, Navier-Stokes, parallelization, iterative solver, MPI, Chimera, Lagrangian particles, mesh multiplication.
Abstract
We present some implementation aspects of a parallel incompressible Navier-Stokes solver, one of the physical modules of an in-house code, Alya - High Performance Computational Mechanics, developed at Barcelona Supercomputing Center, Spain. We will not only treat the solver itself but also three of its surrounding components that can be useful in specific situations: a mesh multiplication algorithm, a Chimera method and a Lagrangian particle tracking. These three components are usually used as pre or post-process tools. However, as a resault of the ever growing weight of pre and post-process tasks, the tendency is to include these tools in the CFD solvers. This allows us to take advantage of the distributed parallel structure of these codes so that the operations of the components are carried out in runtime.
The chapter is organized into seven main sections and covers: the flow equations and the associated numericalmethod; the algebraic strategy that decouples the velocity and pressure; the parallelization strategy for distributed memory supercomputers; the algebraic solver to solve the pressure equation and the three aforementioned components: mesh multiplication, Chimera method and Lagrangian particles. We illustrate these concepts with examples in each section.
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) |
