A partitioning strategy for explicit difference methods
Iterative numerical methods for linear systems (65F10) Finite difference methods for boundary value problems involving PDEs (65N06) Numerical solution of discretized equations for boundary value problems involving PDEs (65N22) Mesh generation, refinement, and adaptive methods for boundary value problems involving PDEs (65N50) Parallel numerical computation (65Y05)
Two different partition shapes, slices and rectangles, are analyzed for the case of explicit difference methods on regular grids. Two different work factors are assumed, \({\mathcal W}_ I\) at interior points and \({\mathcal W}_ B\) at boundary points. It is shown that a good strategy is to nevertheless do the partitioning as if the work was equal at all points. This strategy is nearly optimal, in the sense that it will in most cases give only a small reduction of the efficiency. An algorithm which simplifies the choice of partition shape for problems with different work factors is suggested. The algorithm has been implemented in a computer program. It is remarked that the algorithm should not be applied if \({\mathcal W}_ B\) is much larger than \({\mathcal W}_ I\).
- Straightforward partitioning of composite grids for explicit difference methods
- Mesh partitioning algorithms for the parallel solution of partial differential equations
- scientific article; zbMATH DE number 1215254
- Optimal equi-partition of rectangular domains for parallel computation
- Partitioning PDE computations: Methods and performance evaluation
This page was built for publication: A partitioning strategy for explicit difference methods
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q921877)