ForestClaw : Parallel, adaptive mesh refinement for large scale simulations
Donna Calhoun, Scott Aiton, Carsten Burstedde
Date :
SNuBIC! Research Seminar, SNuBIC Research Group (Aachen, Bochum, Cologne, Düsseldorf,Munich,Stuttgart), Virtual (Invited)
Abstract
Large scale simulations place extreme computatational demands on computing platforms. Parallel computing capabilities are essential. On the software side, there is continued effort to develop faster and more accurate solvers, as well as improve software engineering practices that can take advantage of the latest compilers and languages. Our software platform, ForestClaw (D. Calhoun and C. Burstedde) provides parallel, adaptive capabilities using the state-of-the-art meshing library p4est (C. Burstedde) for finite volume solvers on a nested hierarchicy of logically Cartesian grids. This talk will focus on the adaptive mesh capabilities of ForestClaw, applications that use ForestClaw to accelerate geophysical simulations, recent developments in developing higher order methods using the Active Flux method (E. Chudzik, C. Helzel), and if time permits our GPU implementation of ForestClaw solvers. We will demonstrate applications from tsunami modeling, overland flooding, volcanic eruptions and ash transport, wildfire smoke transport and acoustic wave propagation in the atmosphere.
Slides : snubic-2023.pdf