Parallel algorithms for graph cycle extraction using the cyclical conjunction operator

With a view to reducing the computational cost of extracting all the cycles from complex graphs, the authors have examined the viability here of parallel processing. Based on the cyclical conjunction operator, which uses an iterative process to extract every cycle from a graph, a study was performed of the factors intervening in the parallelization of this algorithm, namely the following: granularity of the parallel algorithm, requirements for synchronization points, and the spreading of the load across different processors. Tests were performed on two granularities and four different load distributions. Algorithm implementation is carried out using SGI MP and OpenMP libraries, and, in the light of the present findings, the authors propose a dynamically distributed fine-grain algorithm using that allows all the cycles in a complex graph to be found in an acceptable computational time.