This paper describes parallelization techniques for a multigrid solver for finite difference analysis of three-dimensional Poisson equations. We first apply our block red-black ordering for parallelization of a Gauss-Seidel (GS) smoother, whose sequentiality is often problematic in parallelization of multigrid methods. Furthermore, we introduce a new multiplicative Schwarz smoother, in which multiple GS iterations are performed in each of red-black ordered blocks. Numerical tests are conducted on a cluster of multi-processor nodes comprising four quad-core AMD Opteron processors to examine the effectiveness of these parallel smoothers. The multi-process test using 216 processes in flat-MPI model shows that the block red-black GS smoother and its multiplicative Schwarz variant achieve 1.3 and 1.8 times better performance than the conventional red-black GS smoother, respectively.
[1]
Ulrich Rüde,et al.
Cache-Aware Multigrid Methods for Solving Poisson's Equation in Two Dimensions
,
2000,
Computing.
[2]
Peter Thoman.
Multigrid Methods on GPUs
,
2008
.
[3]
河合 直聡,et al.
Parallel Multigrid Poisson Solver Based on Block Red-black Ordering
,
2012
.
[4]
Takeshi Iwashita,et al.
Block Red-Black Ordering: A New Ordering Strategy for Parallelization of ICCG Method
,
2004,
International Journal of Parallel Programming.
[5]
V. E. Henson,et al.
BoomerAMG: a parallel algebraic multigrid solver and preconditioner
,
2002
.