Previous abstract Next abstract
The past decade has seen a general maturation of the basic numerical algorithms for modeling the time-dependent, compressible flows which play such an important role in astrophysics. Examples of such simulations can be found in the poster presentations for this topical session. The most pressing challenges now seem to involve other aspects of these computations, such as the manner in which the computation is decomposed so that it may proceed in parallel on hundreds or on thousands of independent processors working from independent memories. Once these challenges are met, others immediately present themselves. Massively parallel computers produce massive data sets. These must be stored, analyzed, and visualized. The new parallel machines present more mundane challenges which are of no lesser importance. These machines all seem to require different Fortran dialects and/or extensions, so that the generation of portable program code requires extensive code preprocessing. Our efforts at Minnesota to meet these new challenges of code parallelization, data handling and visualization, and machine performance will be described. The results of our labors are high resolution simulations of turbulent astrophysical flows carried out on grids of up to a billion computational zones. Selected animations of these simulation results will be shown to illustrate the computational techniques involved.
Wednesday program listing