That's a different speedup technique, though - Transit Node Routing works quite differently from Contraction Hierarchies.

Edit: it seems like the paper you linked uses a technique for parallelizing CH construction that I mentioned in my other comment, https://news.ycombinator.com/item?id=12642961