The overall goal of the TSTT Center is to enable the scientific community to more easily use modern high-order, adaptive, parallel mesh and discretization tools. To achieve this goal, we are following three distinct but related paths. The first is to work directly with a number of lead application teams (for the most part SciDAC-funded) to use such technologies in their application domains. The second is to create new technology that eases the use of such tools, not only for our designated application partners, but across a broad range of application areas that require mesh and discretization tools for scientific simulation. The main technology thrust is not to create new tools (although some of this will occur), but to create new capabilities that will allow the use of these tools interoperably. This very profound step can be compared to the shift from hand craftmanship to manufactured products with interchangable components which revolutionized the world economy one to two centuries ago. The third component of our efforts is to embed this work in a larger framework of related activities, each seeking a similar, and profound, change in the practice of computational science. To ensure the relevance of our work to the SciDAC program goals, we originally selected six application areas, and in each, one or more application projects and teams with which to work directly. One application collaboration which targeted the development of an adaptive mesh refinement capability for the oceanographic code POP was postponed and may be dropped due to unanticipated technical obstacles in the specific goal selected. One new application involving jet breakup for spray combustion was added. The initial job of establishing good working relations, agreement on a plan of action, and obtaining initial results was accomplished in all cases. In general, our work with the applications has been more difficult than anticipated, in spite of the experience of the TSTT team members in similar application-motivated collaborations. For this reason, the routes to the goals have been modified in some cases, but good progress has been obtained for all of the targeted application teams. For example, in the case of the electromagnetic code for accelerator design, the original goal of developing more stable meshes has been enlarged to include the underlying difficulty which motivated this goal: to cure or ameliorate instabilities of the time stepping algorithm. With the fusion M3D code, we decided to work initially with a related, but smaller and more easily modified code from the same application team, for initial testing and proof of principle, as the full M3D code proved difficult to work with. In several applications (astrophysics, climate), our initial technology development goals were met, and while we await their use or evaluation, further collaborative goals will be pursued. The spray breakup problem achieved initial success and awaits adaptive TSTT technology to allow refined grid simulations for its next steps. We plan to continue the intensive effort to insert our existing advanced mesh and discretization technology into existing application codes for the coming year. Our main progress towards the development of new technology has been the definition of the low level interface to a variety of mesh generation and adaptive mesh management tools. This interface provides a common calling convention that will allow an application to call any compliant mesh tool in an interchangeable fashion. Most of the TSTT advanced meshing tools have been or will be made compliant to this interface. We have also pursued one-on-one interoperability goals with the development of interoperability between the FronTier front-tracking library and the Overture mesh library. This goal, advanced from year two to year one because of its need in one of our applications, has made good progress, and will be completed in the coming year. Finally, we mention the integration of this effort (interoperability and applications) with a larger computational science effort. The importance of this broader integration goal can be understood by recalling our larger goal of influencing the practice of computational science in a general sense. We have engaged the computational science community with an invitation to comment on our interoperability plan.
