A GPU-based multi-level algorithm for boundary value problems

DOI10.1016/J.CAM.2019.112481zbMATH Open1433.78028arXiv1703.07206OpenAlexW2977881823WikidataQ127216545 ScholiaQ127216545MaRDI QIDQ2291986FDOQ2291986

Authors: J. Becerra-Sagredo, Carlos Málaga, Francisco Mandujano 

Publication date: 31 January 2020

Published in: Journal of Computational and Applied Mathematics (Search for Journal in Brave)

Abstract: A novel and scalable geometric multi-level algorithm is presented for the numerical solution of elliptic partial differential equations, specially designed to run with high occupancy of streaming processors inside Graphics Processing Units(GPUs). The algorithm consists of iterative, superposed operations on a single grid, and it is composed of two simple full-grid routines: a restriction and a coarsened interpolation-relaxation. The restriction is used to collect sources using recursive coarsened averages, and the interpolation-relaxation simultaneously applies coarsened finite-difference operators and interpolations. The routines are scheduled in a saw-like refining cycle. Convergence to machine precision is achieved repeating the full cycle using accumulated residuals and successively collecting the solution. Its total number of operations scale linearly with the number of nodes. It provides an attractive fast solver for Boundary Value Problems (BVPs), specially for simulations running entirely in the GPU. Applications shown in this work include the deformation of two-dimensional grids, the computation of three-dimensional streamlines for a singular trifoil-knot vortex and the calculation of three-dimensional electric potentials in heterogeneous dielectric media.

Full work available at URL: https://arxiv.org/abs/1703.07206

Recommendations

• Numerical study of geometric multigrid methods on CPU-GPU heterogeneous computers
• Multi-GPU acceleration of algebraic multigrid preconditioners
• GPU-accelerated boundary element method for Helmholtz' equation in three dimensions
• Architecting the finite element method pipeline for the GPU
• Parallel computations on GPU in 3D using the vortex particle method

zbMATH Keywords

CUDAGPUparallel multilevel

Mathematics Subject Classification ID

Multigrid methods; domain decomposition for boundary value problems involving PDEs (65N55) Numerical interpolation (65D05) Laplace operator, Helmholtz equation (reduced wave equation), Poisson equation (35J05) Electro- and magnetostatics (78A30) Finite difference methods applied to problems in optics and electromagnetic theory (78M20) PDEs in connection with optics and electromagnetic theory (35Q60) Finite difference methods for boundary value problems involving PDEs (65N06) Numerical algorithms for specific classes of architectures (65Y10)

Cites Work

• AmgX: a library for GPU accelerated algebraic multigrid and preconditioned iterative methods
• Title not available (Why is that?)
• Title not available (Why is that?)
• A Multigrid Tutorial, Second Edition
• Black box multigrid
• Large calculation of the flow over a hypersonic vehicle using a GPU
• Finite-difference ghost-point multigrid methods on Cartesian grids for elliptic problems in arbitrary domains
• The Immersed Interface/Multigrid Methods for Interface Problems
• Spectral multigrid methods for elliptic equations
• Numerical analysis and implementational aspects of a new multilevel grid deformation method
• A quadtree-adaptive multigrid solver for the Serre-Green-Naghdi equations
• Image registration by a regularized gradient flow. A streaming implementation in DX9 graphics hardware
• Multigrid methods for space fractional partial differential equations
• General relaxation schemes in multigrid algorithms for higher-order singularity methods
• Moments preserving and high-resolution semi-Lagrangian advection scheme
• General theory of domain decomposition: Beyond Schwarz methods

Cited In (3)

• The collocation basis of compact finite differences for moment-preserving interpolations: review, extension and applications
• A GPU accelerated level set reinitialization for an adaptive discontinuous Galerkin method
• Application of graphics processing units for self-consistent modelling of shallow water dynamics and sediment transport

Uses Software

• AmgX
• libmultigrid
• GitHub

This page was built for publication: A GPU-based multi-level algorithm for boundary value problems

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2291986)