GPU-portable Hybrid-VPIC code for multiphysics simulations with kinetic ions
POSTER
Abstract
Hybrid-VPIC [1] is an open-source kinetic-ion/fluid-electron version of the Vector Particle-in-Cell (VPIC) code. Here, we present new Hybrid-VPIC modeling capabilities (https://github.com/lanl/vpic-kokkos/tree/hybridVPIC-K) built on VPIC 2.0 [2], which uses the Kokkos library to achieve high performance and portability to GPU-based machines. The basic collisionless hybrid plasma simulations run 10 to 20 times faster on GPU nodes with 4 GPUs compared to CPU nodes with ~100 cores. In addition, new multiphysics capabilities are being added to the code: Coulomb, charge exchange, and other collision models have been implemented between particles or particles and a fluid species; an interface is under development to allow two-way coupling of the plasma dynamics to radiation transport handled by the Jayenne implicit Monte Carlo code; and an air chemistry model is being developed to treat processes in the upper atmosphere. Example calculations will be shown from space, ionospheric, high energy density laboratory, and magnetic mirror fusion applications.
[1] Le, et al. Physics of plasmas 30.6 (2023).
[2] Bird et al. IEEE Transactions on Parallel and Distributed Systems 33.4 (2021).
[1] Le, et al. Physics of plasmas 30.6 (2023).
[2] Bird et al. IEEE Transactions on Parallel and Distributed Systems 33.4 (2021).
*LA-UR-25-27460. Work supported by LANL LDRD, DOE, and Realta Fusion and used LANL Institutional Computing resources and resources of the NERSC, a DOE Office of Science User Facility supported under Contract No. DE-AC02-05CH11231 using NERSC award FES-ERCAP0026655.
Presenters
-
Ari Le
- Los Alamos National Laboratory (LANL)
- Los Alamos National Laboratory