Comment by ChickeNES

The former is discussed here: https://devblogs.nvidia.com/thinking-parallel-part-ii-tree-t...