1

I am trying to find all Hamiltonian Cycles on all Bi-Connected Chordal graphs. I have been able to generate all the graphs that I need up to 14 vertices. But now I have to search all of those graphs for the Hamiltonian Graphs.

Since the problem is NP-Complete I know there isn't any other way other than to brute force it (unless there is a better algorithm out there that I am not aware of). But I know the computations may take weeks or even longer to search all of those graphs before we get anything to research on.

So the question is, would it be faster to try and compute it on my GPU? I have a 1080ti FTW in my labs computer. Would it be worth it to try and look into trying to compute it on my GPU? If so, where should I begin looking?

Sailanarmo
  • 1,139
  • 15
  • 41
  • What do _you_ think? Do you already implemented it on a CPU? And were you able to parallelize it on the CPU? (Note: I don't know enough about graph theory to really help you. But showing your algorithm maybe makes it possible to answer even for those who are not into graph theory.) Btw even [Wikipedia](https://en.m.wikipedia.org/wiki/Hamiltonian_path_problem#Algorithms) mentions some non brute force algorithms. – BlameTheBits May 29 '17 at 06:09
  • Do you have any idea how many cycles there might be? Combinatoric problems get very large very quickly. I get very cautious when somebody wants *all* solutions of a given combinatoric problem. – TimChippingtonDerrick May 29 '17 at 08:29
  • 1
    On a complete graph, where every node it connected to each other, with 14 nodes then there would be 13! Hamilton cycles that's 6,227,020,800. You probably want the cuda library. There is a specific (nvgraph)[https://developer.nvidia.com/nvgraph] library for graph theory problems. – Salix alba May 29 '17 at 12:42
  • Shadow, I have not personally implemented it on a CPU. However my adviser has and stopped when the computations got too large. Mind you, he is not a computer scientist, but a mathematician. Thank you for the link. I'll look into it. TimChippingtonDerrick, it is what Salix has answered. It is very large, but we need them for this purpose of our research. Salixalba, Thank you. I will have a look into this library. – Sailanarmo May 29 '17 at 17:32
  • 1
    This is a good question, and it's not easily answerable. Please look at https://scicomp.stackexchange.com/questions/943/what-kinds-of-problems-lend-themselves-well-to-gpu-computing – HEKTO May 29 '17 at 18:14

0 Answers0