Hardware synchronization of massively parallel processes in distributed systems

In this paper a new method is proposed to synchronize massively parallel processes in distributed multiprocessor systems. The method is an extension of that used in arbitration systems like Futurebus+. It also uses three global synchronization lines and a distributed synchronizer, and can be applied without changes to the existing hardware. The method allows to carry out two alternative synchronization protocols where the end of the operation is either forced by any processor or individual a moment when all processors completed their individual parts. Application of this method, e.g., to the arbitration process allows to reduce the arbitration time in average by a factor of 2.