Open samuelecancellieri opened 4 years ago
Are you using TopDown or TopDown2? I would recommend using TopDown2. TopDown using a unique optimization that allows for "over-queuing" elements while improving overall performance.
Also, I would suggest moving this discussion to: https://github.com/rapidsai/cuhornet We have made a lot of modifications to Hornet in the last year that have not been applied to this framework.
Thanks. I will download the new repository and try the alternative topdown BFS.
Hi,
I'm trying to implement an algorithm based on the top down BFS integrated in hornet, but I obtained different results every time I execute the BFS.
For example, I fix the root for the BFS, I start the algorithm and in the end I check the distance array. Every time the distances of the same vertices are different in the results array, how is that possible?
I thought about some compatibility problem, but I tested with cuda 9 and cuda 10 and I have always this problem.
Thanks for the help.