ROSS implementation of TSP solver.
Note: I've stopped working on this project for numerous reasons. The prospect of solving TSP in general is a challenging one - doing it efficiently in any manner is a long shot. This project was to evaluate the idea of discrete event simulation being used to solve TSP and general integer programming problems. It also was aimed as being a stepping stone for a TSP solver on neuromorphic hardware. Both were lofty goals and unfortunately most graduate students don't have unlimited funding for lofty goal projects. For a full writeup of this project and preliminary results, please see the report pdf in this repo.
This model requires the installation of ROSS. After cloning the ROSS repository, follow the installation instructions in the ROSS repository to install.
Then symlink the TSP-ROSS repository into the models folder of your ROSS source. Be sure to configure your ROSS installation to 'Build Models' in the CCMAKE configurator.
You can make the model by using the command:
`make tsps'
Once you've built the program you can run it using:
`mpirun -n 4 ./tsps --cities=10 --synch=3'
This runs the program to solve a random 10 city network with 4 PEs.
You may need to increase the amount of extra events allowed by using --extramem=1000000 as well as increasing the size of the AVL tree via --avl-size=26