Computation and Visualization

As part of this project we will implement a number of approaches for obtaining efficient parallel solution of the linear systems arising from radiation transport problems. This work will necessarily include comparative studies of the effectiveness of a number of parallel Krylov subspace algorithms. Furthermore, efficient solution of these large-scale problems will require the investigation of parallel preconditioners. We will consider a number of promising approaches to preconditioning. Finally, the transport problems themselves provide an ideal setting for the investigation of ways of restructuring the iterative methods to achieve optimal scalability at the limit of available TeraScale architectures.

Several factors indicate that the collaboration's scientific goals in astrophysics and nucleosynthesis will be aided significantly by integrating visualization with a repertoire of simulation codes. A couple of recent but modest collaborations between team members at NCSA (Swesty - Baker) and ORNL (Mezzacappa - Toedte) have contributed to understanding dynamics of neutron star collisions and neutrino-driven convection in supernovae. These initial efforts prototyped new ways to custom render and display simulation output, and to share the visualization among multiple participants. We believe strongly that we have only begun to reap the benefits of incorporating visualization into the team's research process. Given the complexity of the science, simulation techniques, and corresponding data, visualization is an essential element of this project.