The purpose of the work is to acquaint the reader about algorithms for finding the maximum flow in networks. It describes three types of algorithms, their implementations and analysis. First algorithm is Ford-Fulkerson which was one of the first algorithms for finding maximum flow in networks. Second algorithm is sequential push-relabel algorithm which is one of the fastest sequential algorithms for finding maximum flow in networks at the time. Third algorithm is parallel push-relabel algorithm which is parallel speed up of sequential push-relabel algorithm. In addition to algorithms, Chapter 2 describes network theory to help understand how algorithms work.
Chapter 5 covers all the results. In the beginning, there is a time comparison of all three algorithms. Then the results focus only on the parallel push-relabel algorithm, since we were interested in speeding up the sequential push-relabel by parallel processing and the impact of the number of processors on running the parallel push-relabel algorithm.
Speeding up the sequential algorithm has proven to be successful, despite slower performance on networks with fewer nodes. On networks with many nodes, the sequential push-relabel algorithm was speeded up to 4 times.
The parallel push-relabel algorithm could be further optimized to better speed up imprementation, but this was not the goal of this thesis.
|