Solution to PDEs using radial basis function finite-differences (RBF-FD) on multiple GPUs

Authors:
Evan F. Bollig;Natasha Flyer;Gordon Erlebacher
Affiliations:
Department of Scientific Computing, Florida State University, 400 Dirac Science Library, Tallahassee, FL 32306, USA;Institute for Mathematics Applied to Geosciences, National Center for Atmospheric Research, 1850 Table Mesa Dr., Boulder, CO 80305, USA;Department of Scientific Computing, Florida State University, 400 Dirac Science Library, Tallahassee, FL 32306, USA
Venue:
Journal of Computational Physics
Year:
2012

Citing 23
Cited 1

Spectral transform solutions to the shallow water test set

Journal of Computational Physics
A Fast and High Quality Multilevel Scheme for Partitioning Irregular Graphs

SIAM Journal on Scientific Computing
Smooth surface reconstruction from noisy range data

Proceedings of the 1st international conference on Computer graphics and interactive techniques in Australasia and South East Asia
Climate Modeling with Spherical Geodesic Grids

Computing in Science and Engineering
Radial Basis Functions

Radial Basis Functions
Radial basis function interpolation: numerical and analytical developments

Radial basis function interpolation: numerical and analytical developments
Numerical methods for high dimensional Hamilton-Jacobi equations using radial basis functions

Journal of Computational Physics
Hardware-Assisted Feature Analysis and Visualization of Procedurally Encoded Multifield Volumetric Data

IEEE Computer Graphics and Applications
Scattered node compact finite difference-type formulas generated from radial basis functions

Journal of Computational Physics
On choosing a radial basis function and a shape parameter when solving a convective PDE on a sphere

Journal of Computational Physics
A Stable Algorithm for Flat Radial Basis Functions on a Sphere

SIAM Journal on Scientific Computing
The use of PDE centres in the local RBF Hermitian method for 3D convective-diffusion problems

Journal of Computational Physics
Implementing sparse matrix-vector multiplication on throughput-oriented processors

Proceedings of the Conference on High Performance Computing Networking, Storage and Analysis
Stable computation of multiquadric interpolants for all values of the shape parameter

Computers & Mathematics with Applications
Fast Construction of k-Nearest Neighbor Graphs for Point Clouds

IEEE Transactions on Visualization and Computer Graphics
Modeling of tsunami waves and atmospheric swirling flows with graphics processing unit (GPU) and radial basis functions (RBF)

Concurrency and Computation: Practice & Experience - Proceedings of the 6th ACES Symposium, May 11–16, 2008, Cairns, Australia
Interactive SPH simulation and rendering on the GPU

Proceedings of the 2010 ACM SIGGRAPH/Eurographics Symposium on Computer Animation
Stabilization of RBF-generated finite difference methods for convective PDEs

Journal of Computational Physics
Keeneland: Bringing Heterogeneous GPU Computing to the Computational Science Community

Computing in Science and Engineering
Stable Computations with Gaussian Radial Basis Functions

SIAM Journal on Scientific Computing
Fast GPU-based locality sensitive hashing for k-nearest neighbor computation

Proceedings of the 19th ACM SIGSPATIAL International Conference on Advances in Geographic Information Systems
A guide to RBF-generated finite differences for nonlinear transport: Shallow water simulations on a sphere

Journal of Computational Physics
Radial Basis Function Networks GPU-Based Implementation

IEEE Transactions on Neural Networks

A guide to RBF-generated finite differences for nonlinear transport: Shallow water simulations on a sphere

Journal of Computational Physics

Quantified Score

Hi-index	31.45

Visualization

Abstract

This paper presents parallelization strategies for the radial basis function-finite difference (RBF-FD) method. As a generalized finite differencing scheme, the RBF-FD method functions without the need for underlying meshes to structure nodes. It offers high-order accuracy approximation and scales as O(N) per time step, with N being with the total number of nodes. To our knowledge, this is the first implementation of the RBF-FD method to leverage GPU accelerators for the solution of PDEs. Additionally, this implementation is the first to span both multiple CPUs and multiple GPUs. OpenCL kernels target the GPUs and inter-processor communication and synchronization is managed by the Message Passing Interface (MPI). We verify our implementation of the RBF-FD method with two hyperbolic PDEs on the sphere, and demonstrate up to 9x speedup on a commodity GPU with unoptimized kernel implementations. On a high performance cluster, the method achieves up to 7x speedup for the maximum problem size of 27,556 nodes.