Effective Heterogeneous Design and Co-simulation 25 6. Acknowledgments

" Architecture-driven synthesis techniques for mapping digital signal processing algorithms into silicon, " FSMs have a clear and unambiguous notion of a quantum of computation (one state transition), the problem reduces to determining what a quantum of computation is in the semantic model used to manage the concurrency and communication. This model could be a synchronous model, a dataflow model, or a discrete-event model, yielding three variants of the statecharts idea. But unlike the mono-lithic variants described in [41], these variants are modular. They are created conveniently when needed simply by choosing the appropriate model of computation at each level of the hierarchy. We would like to thank the Ptolemy team for building a magnificent laboratory for experimenting with the concepts discussed in this chapter. The above examples of combinations suggest that mixing models of computation is reasonably straightforward. In fact, these examples have been carefully chosen to avoid numerous treacherous pitfalls. A key problem is to determine what it means to execute a quantum of computation in a given model of computation. When any of the above models of computation are nested within another, the outer domain invokes a quantum of computation in the inner domain. For this to have meaning, the inner domain must have a clear notion of quantum of computation. For ordinary finite-state machines, which are sequential, it is clear; a quantum of computation consists of a single state transition. But what about concurrent state machines? Should a quantum of computation be a single state transition in each state machine? What if the state machines run asynchronously, and a state transition in one only gets triggered by events in another? What if they are synchronous , but operate at different rates? The answer seems to depend on the model used to manage the concurrency and communication in the FSMs. Indeed these models are typically either dataflow, synchronous (in the sense of synchronous languages), cycle-based, or discrete-event. Thus, hierarchical FSMs that support concurrency can be viewed as manifestations of precisely the sort of het-erogeneity we are discussing in this chapter. Figure 13 identifies three orthogonal semantic properties of statecharts, FSMs, hierarchy, and concurrency. Although it is not possible with statecharts (because of transitions that cross hierarchical boundaries), a simpler model would separate the semantics of concurrency from the semantics of FSMs. Thus, a model only slightly weaker than statecharts would be, in fact, a composition of two …