search

NTUDDSNLab / WBPR

3 stars 1 forks source link

Quick question of some lines of code in GloabalRelabel function #5

Open GiovanniRaseraF opened 1 week ago

GiovanniRaseraF commented 1 week ago

Hi, Giovanni again :), I'm performing some tests on your code and i have difficulties understand why this code in the global relable is needed. I'm testing the global relable without this line of code and it seams to me that the function produces the same results. Please let me know what Im missing :).

https://github.com/NTUDDSNLab/WBPR/blob/6cd8663e5b3d0e4ac70418725a32a3736a344215/maxflow-cuda/src/global_relabel.cu#L9-L33

PunchShadow commented 1 week ago

Dear Giovanni,

The global relabeling function does not affect the maximum flow result; instead, it can accelerate the execution time of the next push-relabel kernel process, which is proven in the original global relabeling paper written by Derig et al.

Our implementation references He's paper (2010 IPDPS).

Best wishes, PunchShadow