Parallelization using ordered sweep, angle, and space domain decompositions in 2D and 3D SN methods

Abstract

The object of this study is to speedup $S_N$ transport calculation using parallel computation. To achieve that goal, several parallelization methods were used which are based on the ordered sweep, angle, and space domain decompositions methods. Parallelization using the ordered sweep draws many attentions at the moment because its capacity to provide massive parallelization. However, the ordered sweep itself still has some disadvantages that may be improved. Angle and space domain decompositions were used to improve the ordered sweep’s performance. The angle domain decomposition decomposes the angle domain within one octant into smaller angle subdomains and sweeps each subdomain simultaneously. The angle domain decomposition improves the ordered sweep’s performance especially in problems that require large number of discrete angles. The space domain decomposition decomposes the space domain into smaller space subdomains and sweeps each subdomain simultaneously. The space decomposition improves the ordered sweep’s performance especially in problems that require large number of mesh. Both methods also add the possibility to use more processors to speedup the calculation.