Minisymposium

This talk will discuss two efficiency-enhancements of the discontinuous Galerkin solution of the Vlasov equation: communication-avoiding boundary exchange as well as an arbitrary Lagrangian-Eulerian implementation of the gyration about the magnetic field lines. The communication of boundary data in a distributed solution constitute a major bottleneck in the solution of high-dimensional problems due to the increased surface-to-volume ratio. We will discuss a reduced communication rate combined with extrapolated boundary data. On the other hand, we combine the Lagrangian treatment of the vxB term with a DG treatment of the rest of the equation to enable increased time steps.

The simulation of turbulence in the boundary region of a tokamak is crucial for understanding and optimizing the performance of fusion reactors. However, these simulations pose significant computational challenges due to the complex plasma dynamics at the boundary. In this talk, we explore how low-rank linear algebra techniques can enhance the efficiency of boundary simulations, specifically by accelerating the computation of neutral particle dynamics. When modeled deterministically, this problem leads to the solution of integral equations, which typically result in dense linear systems upon discretization. We employ hierarchical matrix approximations to significantly reduce the computational cost of assembling and solving linear systems, leading to substantial savings in both time and memory. Our method is implemented and tested within the GBS framework, achieving over 90% reduction in computation time and memory and enabling simulations with unprecedented spatial resolution for neutral particles. The scalability of this approach represents a critical step towards simulating turbulence in larger fusion reactors, facilitating high-resolution studies of plasma-neutral interactions and enhancing the safety and operational efficiency of fusion devices.

The GEMPICX code is a geometric particle in cell solver for plasma physics models. It can handle both electrostatic and fully magnetic simulations. The code is able to leverage modern (including GPU-based) high performance computing systems by utilizing the AMREX framework for block structured adaptive mesh refinement. In this talk, we are going to present recent developments to add relativistic effects to the GEMPICX code. We start with a Hamiltonian splitting and further decompose the particle part into the gyromotion and the free streaming part, which allows us to treat both parts exactly.

Struphy (STRUcture-Preserving HYbrid codes) is an open-source Python package for solving plasma physics PDEs based on finite element exterior calculus (FEEC), particle-in-cell (PIC) and purely Lagrangian methods (smoothed-particle hydrodynamics, SPH). Its purpose is to provide a unified development platform for "prototyping to production in a single framework", being suitable for teaching and research. Struphy is object-oriented and features an instructive API that allows for seamless integration with other programs and easy learning of Struphy objects in Jupyter notebooks. These same objects can then be used in HPC applications within the suite of available "models". In my talk I will focus on the scalability of some of the Struphy models and present the recent addition of GPU acceleration of compute kernels via OpenMP.