A user-controlled synchronization method

Abstract

In any system that allows the sharing of facilities between independently-running processes, it is occasionally necessary to 'synchronize' references to shared facilities. A method is presented here that enables any set of processes to achieve such synchronization without any aid from the executive system or special hardware. Only normal load and store operations on two arrays of shared data are required. No overhead is required for unsynchronized operations. And dead-locks are avoided, since it is never necessary to 'block' a process.This method can be viewed as a generalization of Dijkstra's example [D3], though the connection may not be obvious at first glance.