A parallel auxiliary grid algebraic multigrid method for graphic processing units

Lu Wang, Xiaozhe Hu, Jonathan Cohen, Jinchao Xu

Research output: Contribution to journalArticle

9 Citations (Scopus)

Abstract

In this paper, we develop a new parallel auxiliary grid algebraic multigrid (AMG) method to leverage the power of graphic processing units (GPUs). In the construction of the hierarchical coarse grid, we use a simple and fixed coarsening procedure based on a region quadtree generated from an auxiliary grid. This allows us to explicitly control the sparsity patterns and operator complexities of the AMG solver. This feature provides (nearly) optimal load balancing and predictable communication patterns on shape regular grids, which makes our new algorithm suitable for parallel computing, especially on GPUs. We also design a parallel smoother based on the special coloring of the quadtree to accelerate the convergence rate and improve the parallel performance of this solver. Based on the CUDA toolkit [NVIDIA CUDA Programming Guide, NVIDIA Corp., 2010], we implemented our new parallel auxiliary grid AMG method on GPUs and the numerical results of this implementation demonstrate the efficiency of our new method for (nearly) isotropic problems. The results achieve an average speedup of over 4 on quasi-uniform grids and 2 on shape regular grids when compared to the AMG implementation in CUSP [M. Garland and N. Bell, CUSP: Generic Parallel Algorithms for Sparse Matrix and Graph Computations, http://cusplibrary.github. com/ (2010)].

Original languageEnglish (US)
Pages (from-to)C263-C283
JournalSIAM Journal on Scientific Computing
Volume35
Issue number3
DOIs
StatePublished - Oct 8 2013

Fingerprint

Algebraic multigrid Method
Graphics Processing Unit
Grid
Algebraic multigrid
Coarsening
Coloring
Parallel processing systems
Quadtree
Parallel algorithms
Resource allocation
Cusp
Communication
Sparse Graphs
Sparse matrix
Parallel Computing
Graphics processing unit
Load Balancing
Sparsity
Leverage
Parallel Algorithms

All Science Journal Classification (ASJC) codes

  • Computational Mathematics
  • Applied Mathematics

Cite this

Wang, Lu ; Hu, Xiaozhe ; Cohen, Jonathan ; Xu, Jinchao. / A parallel auxiliary grid algebraic multigrid method for graphic processing units. In: SIAM Journal on Scientific Computing. 2013 ; Vol. 35, No. 3. pp. C263-C283.
@article{e95a2eac2bb14b2eb4905a0f977c4665,
title = "A parallel auxiliary grid algebraic multigrid method for graphic processing units",
abstract = "In this paper, we develop a new parallel auxiliary grid algebraic multigrid (AMG) method to leverage the power of graphic processing units (GPUs). In the construction of the hierarchical coarse grid, we use a simple and fixed coarsening procedure based on a region quadtree generated from an auxiliary grid. This allows us to explicitly control the sparsity patterns and operator complexities of the AMG solver. This feature provides (nearly) optimal load balancing and predictable communication patterns on shape regular grids, which makes our new algorithm suitable for parallel computing, especially on GPUs. We also design a parallel smoother based on the special coloring of the quadtree to accelerate the convergence rate and improve the parallel performance of this solver. Based on the CUDA toolkit [NVIDIA CUDA Programming Guide, NVIDIA Corp., 2010], we implemented our new parallel auxiliary grid AMG method on GPUs and the numerical results of this implementation demonstrate the efficiency of our new method for (nearly) isotropic problems. The results achieve an average speedup of over 4 on quasi-uniform grids and 2 on shape regular grids when compared to the AMG implementation in CUSP [M. Garland and N. Bell, CUSP: Generic Parallel Algorithms for Sparse Matrix and Graph Computations, http://cusplibrary.github. com/ (2010)].",
author = "Lu Wang and Xiaozhe Hu and Jonathan Cohen and Jinchao Xu",
year = "2013",
month = "10",
day = "8",
doi = "10.1137/120894452",
language = "English (US)",
volume = "35",
pages = "C263--C283",
journal = "SIAM Journal of Scientific Computing",
issn = "1064-8275",
publisher = "Society for Industrial and Applied Mathematics Publications",
number = "3",

}

A parallel auxiliary grid algebraic multigrid method for graphic processing units. / Wang, Lu; Hu, Xiaozhe; Cohen, Jonathan; Xu, Jinchao.

In: SIAM Journal on Scientific Computing, Vol. 35, No. 3, 08.10.2013, p. C263-C283.

Research output: Contribution to journalArticle

TY - JOUR

T1 - A parallel auxiliary grid algebraic multigrid method for graphic processing units

AU - Wang, Lu

AU - Hu, Xiaozhe

AU - Cohen, Jonathan

AU - Xu, Jinchao

PY - 2013/10/8

Y1 - 2013/10/8

N2 - In this paper, we develop a new parallel auxiliary grid algebraic multigrid (AMG) method to leverage the power of graphic processing units (GPUs). In the construction of the hierarchical coarse grid, we use a simple and fixed coarsening procedure based on a region quadtree generated from an auxiliary grid. This allows us to explicitly control the sparsity patterns and operator complexities of the AMG solver. This feature provides (nearly) optimal load balancing and predictable communication patterns on shape regular grids, which makes our new algorithm suitable for parallel computing, especially on GPUs. We also design a parallel smoother based on the special coloring of the quadtree to accelerate the convergence rate and improve the parallel performance of this solver. Based on the CUDA toolkit [NVIDIA CUDA Programming Guide, NVIDIA Corp., 2010], we implemented our new parallel auxiliary grid AMG method on GPUs and the numerical results of this implementation demonstrate the efficiency of our new method for (nearly) isotropic problems. The results achieve an average speedup of over 4 on quasi-uniform grids and 2 on shape regular grids when compared to the AMG implementation in CUSP [M. Garland and N. Bell, CUSP: Generic Parallel Algorithms for Sparse Matrix and Graph Computations, http://cusplibrary.github. com/ (2010)].

AB - In this paper, we develop a new parallel auxiliary grid algebraic multigrid (AMG) method to leverage the power of graphic processing units (GPUs). In the construction of the hierarchical coarse grid, we use a simple and fixed coarsening procedure based on a region quadtree generated from an auxiliary grid. This allows us to explicitly control the sparsity patterns and operator complexities of the AMG solver. This feature provides (nearly) optimal load balancing and predictable communication patterns on shape regular grids, which makes our new algorithm suitable for parallel computing, especially on GPUs. We also design a parallel smoother based on the special coloring of the quadtree to accelerate the convergence rate and improve the parallel performance of this solver. Based on the CUDA toolkit [NVIDIA CUDA Programming Guide, NVIDIA Corp., 2010], we implemented our new parallel auxiliary grid AMG method on GPUs and the numerical results of this implementation demonstrate the efficiency of our new method for (nearly) isotropic problems. The results achieve an average speedup of over 4 on quasi-uniform grids and 2 on shape regular grids when compared to the AMG implementation in CUSP [M. Garland and N. Bell, CUSP: Generic Parallel Algorithms for Sparse Matrix and Graph Computations, http://cusplibrary.github. com/ (2010)].

UR - http://www.scopus.com/inward/record.url?scp=84883696475&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84883696475&partnerID=8YFLogxK

U2 - 10.1137/120894452

DO - 10.1137/120894452

M3 - Article

AN - SCOPUS:84883696475

VL - 35

SP - C263-C283

JO - SIAM Journal of Scientific Computing

JF - SIAM Journal of Scientific Computing

SN - 1064-8275

IS - 3

ER -