Parallel mesh generation is an important feature of any large distributed memory parallel computational mechanics code due to the need to ensure that (i) there are no sequential bottlenecks within the code, (ii) there is no parallel overhead incurred in partitioning an existing mesh and (iii) that no single processor is required to have enough local memory to be able to store the entire mesh. In recent years numerous algorithms have been proposed for the generation of unstructured finite element and finite volume meshes in parallel. One of the main problems with many of these approaches however is that the final mesh, once generated, cannot generally be guaranteed to be perfectly load-balanced. In this paper we propose a post-processing step for the parallel mesh generator, based upon a cheap and efficient dynamic load-balancing technique. This technique is described and a number of numerical examples are presented in order to demonstrate that the quality of the partition of the mesh can be improved significantly at only a small additional computational cost.

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

go to the previous paper
go to the next paper
return to the table of contents
return to the book description
purchase this book (price £66 +P&P)