The goal of the TSTT Center is to develop and deploy a set of computational tools that will significantly and positively impact the ability of SciDAC applications researchers to employ advanced terascale mesh-based simulation technology in their research. Because of the sophistication of most existing advanced tools in this area, their availability has until now been limited to only the most ambitious application developers. We are addressing this deficiency by offering significant capability for generating different kinds of meshes for complex geometry and employing these meshes in scientific simulations through easy-to-use application-appropriate interfaces. Since many important Office of Science applications are characterized by strongly non-uniform solution features, these tools will include capabilities for solution-adaptive mesh and solution improvement capabilities, including mesh refinement and front tracking. Interaction with a broad cross-section of SciDAC applications scientists early in the design phase of our software is an essential step in ensuring that the interfaces we develop address the needs of a broad class of Office of Science applications.
Our application work will form the initial phase in the development of a common interface for meshing and discretization technologies. This phase will be “fire tested” through integration of TSTT tools (primarily existing tools to achieve early deliverables) with application codes. Acceptance of the common interface is central aspect to the plug and play usability of adaptive grid technologies that we are proposing. As our common interface evolves, we will assist the application codes in a migration to its use, while considering deeper uses for adaptivity and other advanced technologies within the same application areas. We will also broaden our involvement with applications, including ones not covered in the initial phases of our work, to ensure the widest attainable acceptance of these standards.
We have invested a significant portion of our resources in meeting with scientists from each of the SciDAC application areas, analyzing their needs for advanced meshing and discretization technologies, and working with them to demonstrate the promise of such techniques in their application domains. We highlight these interactions here.
- For accelerator design, we have focused on helping to increase the stability of the Tau3P simulation code used for computational electromagnetics simulations at SLAC. We have adopted a two-pronged approach. First, we are seeking to understand the dependence of the simulation code stability on properties of the meshes used for accelerator design simulations and are using TSTT mesh generation technology to shorten the time required to generate high-quality, all-hexahedral meshes. Second, we are working to understand and improve the stability of the simulation code by analyzing the DSI discretization strategy used in Tau3P. Early successes of this work include the construction of a stable, first-order accurate method for triangular meshes and the ability to stabilize the DSI scheme by adding high-order artificial dissipation. Because the addition of artificial diffusion can affect long term accuracy of these simulations, more work is needed to complete this aspect of the project.
- For two of the astrophysics SciDAC centers, we have demonstrated the potential impact of high-order discretization methods for both the hydrodynamics and neutrino transport aspects of their simulations. In both cases, adaptive Discontinuous Galerkin discretizations were implemented for the appropriate test problems and comparison to the currently used techniques shows that a significant benefit in terms of accuracy and time to solution can be obtained by using these techniques.
- In our interactions with climate scientists, we have focused our efforts in two main areas. First, we have worked on mesh generation strategies which create high-quality grids whose vertices are focused over regions of interest, and, in this case in particular, over regions of high altitude. Second, joint work between scientists at NCAR and at ANL has resulted in a new preconditioner for spectral element simulations based on low-order finite element discretizations that has accelerated the integration rate in the shallow water equations test problem.
- TSTT members at BNL and SUNY SB have been working with basic energy scientists at Argonne to create a simulation code that models spray formation in diesel jet break up using an enhanced version of the FronTier front-tracking code. Early simulation results have helped determine the sensitivity of the model to various input parameters and is a primary motivator for our work to create interoperable meshing technologies.
- In our work with fusion scientists at PPPL, we have targeted the use of high-order adaptive discretization technologies in MHD fusion simulations. To date, we have examined the M3D code to determine if it can be installed at SCOREC, implemented a specific MHD capability within the SCOREC Trellis framework, and tested this capability with an appropriate test problem. This work provides the foundation for the algorithms that we plan to directly insert into the M3D code.
- We are also working to cultivate a relationship between TSTT and computational biologists by incorporating TSTT technology into two DOE programs in this area. The focus of our efforts are in image reconstruction and feature extraction for complex biological systems, and we have had initial successes in microbial cell and human lung modeling along with geometry extraction and mesh generation for rat olfactory systems.