This paper uses cooperatively scheduled threads (i.e. no preemption or concurrent execution of threads), which means you don't have to deal with synchronization.