The Simulation of Synchronous Reactive Systems In Ptolemy II

The Synchronous Reactive (SR) domain in Ptolemy II models systems as components with both infinite processing resources and infinite communication bandwidth. Computation within components and data transfer between components are considered to happen instantaneously. Therefore, a given system is simulated as if all of its components execute and produce outputs at the same time. This raises interesting issues, particularly in systems with feedback. To guarantee a deterministic model of computation, outputs are permitted to have undefined value, and certain constraints are imposed on component behavior. Specialized components, such as finite state machines, may execute even when some inputs are undefined, producing only those outputs that can be determined from the defined inputs. This report outlines the semantics of the SR model of computation and describes its implementation as a domain in Ptolemy II. ii Acknowledgements First, I would like to thank Professor Edward A. Lee for introducing me to the Synchronous Reactive model of computation. His continuous support and guidance made this research possible. I am extremely grateful for the valuable advice he offered me regarding this report, and also over the course of my involvement with the Ptolemy project. I would also like to thank Professor Thomas A. Henzinger for graciously consenting to act as a second reader of this report. I am greatly indebted to Stephen Edwards for his work on Synchronous Reactive systems in Ptolemy Classic. I also appreciated his willingness to discuss his component execution scheduling algorithms. This paper stands on his shoulders. I would like to thank all of my colleagues in the Ptolemy group for providing an excellent research environment and for helping me understand Ptolemy II. In particular, Jie Liu explained some of the implementation issues of fixed-point semantics. Xiaojun Liu taught me about finite state machines in Ptolemy II, allowing me to integrate them with my work. Stephen Neuendorffer clarified many fine points of the Ptolemy II kernel and sometimes helped me identify my mistakes. Yuhong Xiong answered all of my questions about the Ptolemy II type system. Christopher Hylands and Jörn Janneck reviewed a great deal of my work and provided invaluable programming advice. a draft of this report and offered suggestions.