[taeric]

It would be interesting to see timing results of this implementation. Similarly, comparisons to other implementations/algorithms. I will have to fire up erlang when I get home this evening.

I am basically a broken record on this topic, but I do like sending more people to see the Dancing Links algorithms solving this (and other related problems, such as sudoku.)[1] The techniques in DLX cut the search space for the 8-queens down to just 1,048 board states. (For 13, the number of boards is up to 1,046,318)

[1] http://taeric.github.io/DancingLinks.html

[tacticiankerala]

8-queens : 0.19s user 0.09s system 127% cpu 0.222 total

12-queens : 0.21s user 0.11s system 129% cpu 0.244 total

16-queens : 0.29s user 0.10s system 119% cpu 0.325 total

20-queens : 2.79s user 0.10s system 102% cpu 2.826 total

24-queens : 7.50s user 0.10s system 100% cpu 7.533 total

Thanks a lot for sharing the Dancing links algorithm. It looks really interesting. I wanted to try sudoku with backtracking,but now I would like to give dancing link a try!

[taeric]

Are those times for finding a solution, or for all solutions? If all, that is truly impressive!

Dancing links is a backtracking algorithm, just a nice way to optimize the backtracking.

[tacticiankerala]

It is for finding the first solution. Sorry that I didn't mention it.

[taeric]

No worries. I could have guessed, if I'd paid attention to how large of an N you looked at. :)

Also, I should explicitly say thanks for sharing! It is fun reading solutions to this puzzle. Also fun to read Erlang.