> This implementation additionally can distribute pools of parallel processes over native OS level threads, thus taking advantage of multicore architectures. Locking overhead should be small, as data is normally not shared among processes executing on different threads.

From: http://www.call-with-current-continuation.org/fleng/fleng.ht...

It seems the author has taken this into account, it's not a 1:1 mapping of processes to OS threads.