The process network domain, created by Thomas M. Parks and documented in his Ph.D. thesis [Par95], implements Kahn process networks, a generalization of dataflow where processes replace actors. It has some of the flavor of the recently removed CP domain, in that it implements concurrent processes, but unlike the CP domain, it is determinate and has no model of time. The PN domain is implemented using POSIX threads. In principle, PN systems can run in parallel on multiprocessor workstations with appropriate OS support for threads.
The SDF, BDF and DDF domains are subdomains of PN, which means that these stars can be used directly in PN systems. When stars from these domains are used in a PN system, each dataflow actor becomes a dataflow process [Lee95]. For greater efficiency, dataflow domains can be mixed with PN using the Ptolemy hierarchy. A galaxy within a PN system can be SDF, BDF, or DDF, using a scheduler appropriate for that domain. The galaxy as a whole becomes a single process in the PN system.
Copyright © 1990-1997, University of California. All rights