Parallelization of travelling salesman problem (TSP) using PyCSP
Used as an example in the presentation: "Communicating Sequential Processes (CSP) - An alternative to the actor model"
The implementation models TSP as a complete, undirected weighted graph, and implements the branch-and-bound optimization. The parallelization uses a master-slave pattern, and requires Python 2.6 or 2.7.
python run_sequential.py <num_cities>
, where
num_cities
is number of cites/node in TSP
python run_parallel.py <num_cities><task depth><num workers>
, where
num_cities
is number of cities/nodes in TSPtask depth
is the length of sub-routes the master will use as tasks in the parallelization. A larger task depth results in more tasks.num workers
is number of workers that will be spawned
- iter1: Does not use branch-and-bound optimization in parallelization. Less complexity.
- cluster: Minor modification for using the cluster module in PyCSP.