Research on the Optimization of Local Node Connectivity Computation Using GPGPU

Abstract

The local node connectivity of a graph indicates its stability and robustness as well as the importance of its vertices. This study presents the process and results of designing local node connectivity computation algorithms for graphs using parallel processing in GPGPU (General-Purpose computing on Graphics Processing Units). These algorithms focus on reducing latency by optimizing the computation for finding the shortest paths. Here, the algorithms “Variable Threads” and “Fixed Threads” are proposed and compared with a CPU-based baseline algorithm. Experimental results show that the GPU-based approach achieved latency reductions of up to 37% and 15% in sparse and dense graphs, respectively. These findings elucidate the effectiveness of parallel processing on GPUs for improving the efficiency of local node connectivity computations.