Impact of computing platforms on the performance of the asymmetric traveling salesman problem

This article discusses a serial and a parallel version of a heuristic algorithm for a solution to the traveling salesman problem, implementation of these algorithms, and their performance in terms of finding the most favorable solution, and an average solution. The algorithms implemented in C sequential programming language, and in Occam concurrent programming language were run on different computing platforms to study their impact on the comparative value of the solutions each platform provided. The programs were tested on reference data from TSPLIB data files, a collection of sample instances of traveling salesman problem. Experimental results, using asymmetric cost matrix, indicate good performance compared to the optimal solutions provided in the TSPLIB reference library.