## Block-Structured Multigrid for the Navier-Stokes Equations: Experiences and Scalability Questions (1992)

Venue: | Proceedings of the Conference on Parallel Computational Fluid Dynamics 1992, Elsevier Science Publishers B.V |

Citations: | 3 - 1 self |

### BibTeX

@INPROCEEDINGS{Linden92block-structuredmultigrid,

author = {J. Linden and G. Lonsdale and H. Ritzdorf and A. Schüller},

title = {Block-Structured Multigrid for the Navier-Stokes Equations: Experiences and Scalability Questions},

booktitle = {Proceedings of the Conference on Parallel Computational Fluid Dynamics 1992, Elsevier Science Publishers B.V},

year = {1992},

publisher = {Elsevier Science Publishers B.V}

}

### OpenURL

### Abstract

: This paper summarizes investigations concerning the algorithmic scalability of multigrid methods for partial differential equations on MIMD distributed memory systems. It is shown that even multigrid methods which are distinguished by h-independent convergence rates are not scalable in a rigorous sense. We develop their parallel asymptotic computational complexity for different types of multigrid cycles and analyze their critical components with respect to scalability. Experimental results for two Navier-Stokes test problems presented in the last section of this paper show, however, that the theoretically predicted dependency of the combined numerical and parallel efficiencies of multigrid methods on the number of processors employed is in fact very weak. This leads to the conclusion that multigrid is also with respect to scalability an appropriate candidate for solving partial differential equations on massively parallel machines. i Contents 1 Introduction 1 2 Asymptotic complexi...

### Citations

737 | PVM: a framework for parallel distributed computing. Concurrency: Pract
- Sunderam
- 1990
(Show Context)
Citation Context ...lting in low communication capabilities. In addition, since a native PARMACS implementation for interprocess communication was not available on workstation networks, a simple PARMACS interface to PVM =-=[15]-=- was developed and used throughout the tests employing workstations. This PVM workstation implementation was not optimized in any respect. Moreover, the workstations have powerful processors which req... |

165 |
Multigrid techniques: 1984 guide with applications to fluid dynamics
- Brandt
- 1984
(Show Context)
Citation Context ...or the numerical tests. Due to space limitations we do not describe any details concerning multigrid, grid partitioning and their use with block-structured grids. The interested reader is referred to =-=[2, 5, 6, 7, 8, 9, 10, 12, 14]-=-. 2 Asymptotic complexity analysis In order to analyze the effect of parallelization by grid partitioning on the computational cost of a multigrid algorithm, we will consider a linear, 2-D equation an... |

114 |
Multigrid Methods: Fundamental Algorithms, Model Problem Analysis and Applications, Multigrid Methods
- Trottenberg
- 1981
(Show Context)
Citation Context ...or the numerical tests. Due to space limitations we do not describe any details concerning multigrid, grid partitioning and their use with block-structured grids. The interested reader is referred to =-=[2, 5, 6, 7, 8, 9, 10, 12, 14]-=-. 2 Asymptotic complexity analysis In order to analyze the effect of parallelization by grid partitioning on the computational cost of a multigrid algorithm, we will consider a linear, 2-D equation an... |

48 |
Accelerated multigrid convergence and high-Reynolds recirculating flows
- Brandt, Yavneh
- 1993
(Show Context)
Citation Context ...more coarse grid intensive cycle types would drastically increase the computational costs. Detailed analyses of such coarse grid correction problems, together with practical examples, can be found in =-=[3, 13]-=-. In general, block-structuring introduces certain limitations with respect to agglomeration or other continued coarsening approaches. In complicated geometries, the global coarsest grid may, for exam... |

30 |
U.: Multigrid methods on parallel computers – a survey of recent developments
- McBryan, Frederickson, et al.
- 1991
(Show Context)
Citation Context ...or the numerical tests. Due to space limitations we do not describe any details concerning multigrid, grid partitioning and their use with block-structured grids. The interested reader is referred to =-=[2, 5, 6, 7, 8, 9, 10, 12, 14]-=-. 2 Asymptotic complexity analysis In order to analyze the effect of parallelization by grid partitioning on the computational cost of a multigrid algorithm, we will consider a linear, 2-D equation an... |

25 |
The Argonne/GMD macros in FORTRAN for portable parallel programming and their implementation on the Intel iPSC/2
- Bomans, Roose, et al.
- 1990
(Show Context)
Citation Context ...ion networks The L i SS package uses the GMD Communications Library for all communications tasks. This library is implemented in terms of the portable PARMACS interface for interprocess communication =-=[1]-=-, which is available on many parallel systems. The parallel code could thus be executed on quite a large variety of parallel machines. Hence, in addition to the numerical experiments on the Intel iPSC... |

9 |
A multigrid method for steady incompressible Navier-Stokes equations based on flux-difference splitting
- Dick, Linden
- 1992
(Show Context)
Citation Context ...uare cavity at Rayleigh number 30000 (Fig. 3). In both cases the discretization was based on the flux difference splitting method for the equations in terms of the primitive variables as described in =-=[4]-=-. inflow outflow solid wall solid wall Figure 2: Boundary conditions for the circular bump problem. 6 solid wall T = 0:5 solid wall T = \Gamma0:5 solid wall, T n = 0 solid wall, T n = 0 ? g Figure 3: ... |

8 |
Parallel multigrid solution of the Navier-Stokes equations on general 2D-domains
- Linden, Steckel, et al.
- 1988
(Show Context)
Citation Context |

5 |
Multigrid methods: An overview with emphasis on grid generation processes
- Linden, Stuben
- 1986
(Show Context)
Citation Context |

4 |
Parallel and vector aspects of a multigrid Navier-Stokes solver, Arbeitspapiere der GMD
- Lonsdale, Schuller
- 1991
(Show Context)
Citation Context |

3 |
The GMD Communications Subroutine Library for grid-oriented problems, Arbeitspapiere der GMD
- Hempel, Ritzdorf
- 1991
(Show Context)
Citation Context |

2 |
Maintaining multigrid and parallel efficiencies for the Navier-Stokes equations
- Lonsdale, Schuller
- 1992
(Show Context)
Citation Context |

2 |
The L i SS package, Arbeitspapiere der GMD
- Lonsdale, Stuben
- 1991
(Show Context)
Citation Context ...tal results on the dependency of computing times on the number of processors on Suprenum and the Intel iPSC/860. These results were obtained with the parallel block-structured multigrid solver L i SS =-=[11]-=-. They confirm that a careful treatment of the coarse grid problems yields practical scalability at least within a wide range of processor numbers. Results on other parallel systems, including worksta... |

1 |
Experiments with parallel multigrid using the
- Hempel, Schuller
- 1988
(Show Context)
Citation Context |

1 |
Mehrgitterverfahren fur Schalenprobleme, GMD-Bericht Nr
- Schuller
- 1988
(Show Context)
Citation Context ...more coarse grid intensive cycle types would drastically increase the computational costs. Detailed analyses of such coarse grid correction problems, together with practical examples, can be found in =-=[3, 13]-=-. In general, block-structuring introduces certain limitations with respect to agglomeration or other continued coarsening approaches. In complicated geometries, the global coarsest grid may, for exam... |