Parallel Adaptive Techniques for Transient Partial Differential Equations.
Abstract
As scientific and engineering problems increase in complexity, there is a corresponding need to automate a greater portion of the solution process. Adaptive methods are capable of providing many of the decisions - that arise when solving problems involving partial differential equations. When automatic mesh generation, refinement and coarsening is combined with method-order variation, it is possible to construct methods having spectral convergence rates. Accuracy and solution reliability are verified by asymptotically correct estimates of discretization errors. With parallel computation becoming more widespread and necessary, it is important to consider parallel adaptive strategies. Much more difficult to parallelize than traditional algorithms, adaptive strategies necessarily require dynamic processor scheduling because of load changes introduced by periodic solution enrichment. (AN)
Document Details
- Document Type
- Technical Report
- Publication Date
- Nov 10, 1994
- Accession Number
- ADA290460
Entities
People
- J. E. Flaherty
- Mark S. Shephard
Organizations
- Rensselaer Polytechnic Institute