Any Documentation available on internals?

cuGraph, and now the new ReOpt. It says it has parallel Lin/Kernighan and parallel ant colony. Does NVidia document any of this, perhaps even make the code readable? I understand the value and necessity of closed-source software, but can they at least document WHAT and HOW things are parallelized? Just saying “parallel algorithms” doesn’t tell me much, you can still “parallelize” in a smart way and in a dumb way. There are several ways to implement centrality of ant colony or Dijkstra or page rank, can we know more please?