A low memory, highly concurrent multigrid algorithm

TL;DR

This paper presents a low-memory, highly concurrent multigrid algorithm based on Brandt's segmental refinement, optimized for modern architectures, enabling efficient PDE solutions with reduced memory and improved parallelism.

Contribution

It develops a parallel, low-memory multigrid method using segmental refinement and $ au$-corrections, extending Brandt's original algorithm for modern high-performance computing.

Findings

Maintains convergence rate with one FMG iteration.

Reduces memory usage significantly through segmental refinement.

Provides a parallel implementation exploiting modern architectures.

Abstract

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.