The Castro Adaptive Mesh Refinement Hydrodynamics Code: Applications, Algorithm Development, and Performance Portability

Autor: Zingale, Michael, Katz, Max, Sazo, Maria Barrios, Harpole, Alice, Almgren, Ann, Bell, John, Willcox, Don, Zhang, Weiqun
Rok vydání: 2019
DOI: 10.5281/zenodo.2537406
Popis: We describe the recent developments in the open source Castro adaptive mesh refinement (radiation) hydrodynamics code, with full self-gravity, general equations of state, and arbitrary nuclear reaction networks. Since its initial release, Castro has been used for models of core-collapse and Type Ia supernovae (in the Chandra, sub-Chandra, and double degenerate progenitor scenarios), X-ray bursts, and exoplanet dynamics. The core solvers have evolved significantly since the initial release. We detail a new spectral deferred correction time-integration framework, enabling a push to higher order accuracy. We show some initial tests of this new methodology. We also highlight the work done on performance portability, allowing us to reuse the same computational kernels on CPUs and GPUs. Finally, we discuss the development model for Castro, including open code repos, regular regression testing, and distribution of all files needed to rerun our science models in the core repository.
Databáze: OpenAIRE