Toward Massively Parallel Adaptive Simulations on Unstructured MeshesPresenter: Mark Shephard, RPIScientific Computation Research Center, Rensselaer Polytechnic Institute and Lori A. Diachin, Center for Applied Scientific Computing, Lawrence Livermore National Laboratory |
It is well known that adaptive methods are central to reliable simulations. The ability to perform adaptive calculations on massively parallel computers is complicated by the need to dynamically load balance as the computation grids are adapted. Several groups are investigating parallel adaptive methods that employ some form of structured mesh technique (AMR). An alternative approach is to employ unstructured meshes and mesh adaptation procedures. Although these methods employ more complex representations and algorithms they have the advantages of being able to take advantage of fully automatic mesh generation technologies and employ a wide variety of mesh adaptation procedures including highly effective anisotropic meshes with alignment on general curved geometries.
This presentation will discuss several aspects of the Interoperable Technologies for Advanced Petascale Simulations (ITAPS) CET efforts to develop interoperable tools to support parallel adaptive unstructured mesh simulations. The presentation will overview a prototype structure to support dynamically evolving distributed unstructured meshes and discuss efforts to develop parallel adaptive simulations for SciDAC applications taking advantage of the ITAPS methods.