Parthenon
Benjamin R. Ryan, Jonah M Miller, Philipp Grete, Luke F. Roberts, Ben S. Prather, Nirmal Prajapati, Patrick Dean Mullen, Alex Long, Forrest W. Glines, Marissa Sorkin · 2024
Parthenon* is an open-source** performance portability library for high performance multiphysics applications written in C++.Provides data abstractions that support block-adaptive mesh refinement and domain decomposition for physics algorithms that use cell-, face-, edge-centered fields, explicit and implicit solvers, and particle methods.Dependencies: • Kokkos for performance portability (CUDA backend) • MPI for inter-/intra-node parallelism • HDF5 for I/O (openPMD support in development) • Catch2 for testing Here we are focusing on improving parallel scaling on GPU HPC machines Downstream applications