Octant overlapping and angle-energy group expanded KBA parallelism of three-dimensional discrete ordinate method

Introduction To enhance the parallel efficiency of the conventional KBA spatial decomposition method, this study incorporates octant overlapping and angle-energy group expansion into the source iteration scheme for solving the discrete ordinate equations. Methods The performance of various KBA-based methods is theoretically analyzed in terms of parallel efficiency and speedup ratio. Furthermore, the influence of different data structures on CPU cache hit ratio and calculation time is tested through three-dimensional benchmark problems. Results and Discussion The results demonstrate that the source iteration method incorporating octant overlapping and angle-group expansion exhibits superior parallel characteristics. This enhancement can elevate the parallel efficiency of KBA for prevalent problems to over 97%. The impact of data structures on the cache hit ratio and calculation time is substantial. Organizing data according to its access order is shown to significantly improve the cache hit rate. The findings of the three-dimensional benchmark problem confirm that the proposed octant overlapping and angle-energy group expanding parallel method achieves a higher speedup ratio. This method aligns closely with the theoretical analysis results and can be employed for fine-grained, large-scale parallel computing and analysis.