SearcharxivSearch

arXiv subjects

Matt Knepley

Publications and source records attributed to Matt Knepley.

2 recordsLinked to original sources

The Landau Collision Integral in the Particle Basis in the PETSc Library

The Landau collision integral is often considered the gold standard in the context of kinetic plasma simulation due to its conservative properties, despite challenges involved in its discretization. The primary challenge when implementing an efficient computation of this operator is conserving physical properties of the continuum equation when the system is discretized. Recent work has achieved continuum discretizations using the method of Finite Elements which maintain conservation of mass, momentum, and energy, but which lacks monotonic entropy production. More recently, a particle discretization has been introduced which conserves mass, momentum, and energy, but maintains the benefit of monotonic entropy production necessary for the metriplecticity of the system. We present here an implementation of the particle basis Landau collision integral in the Portable Extensible Toolkit for Scientific Computing in 2 and 3V for the construction of a full geometry solver with a novel approach to computation of the entropy functional gradients. Verification of the operator is achieved with thermal equilibration and isotropization tests. All examples are available, open source, in the PETSc repository for reproduction.

physics.plasm-ph

A low memory, highly concurrent multigrid algorithm

We examine what is an efficient and scalable nonlinear solver, with low work and memory complexity, for many classes of discretized partial differential equations (PDEs) - matrix-free Full multigrid (FMG) with a Full Approximation Storage (FAS) - in the context of current trends in computer architectures. Brandt proposed an extremely low memory FMG-FAS algorithm over 25 years ago that has several attractive properties for reducing costs on modern - memory centric -- machines and has not been developed to our knowledge. This method, segmental refinement (SR), has very low memory requirements because the finest grids need not be held in memory at any one time but can be "swept" through, computing coarse grid correction and any quantities of interest, allowing for orders of magnitude reduction in memory usage. This algorithm has two useful ideas for effectively exploiting future architectures: improved data locality and reuse via "vertical" processing of the multigrid algorithms and the method of $\tau$-corrections, which allows for not storing the entire fine grids at any one time. This report develops this algorithm for a model problem and a parallel generalization of the original sweeping technique. We show that FMG-FAS-SR can work as originally predicted, solving systems accurately enough to maintain the convergence rate of the discretization with one FMG iteration, and that the parallel algorithm provides a natural approach to fully exploiting the available parallelism of FMG.

math.NA