Executing concurrent specifications on sequential hardware is important for both simulation of systems that are eventually implemented on concurrent hardware and for those most conveniently described as a set of concurrent processes. As with most forms of simulation, this is easy to do correctly but difficult to do efficiently. Solutions such as preemptive operating systems and discrete-event simulators present significant overhead.
In this talk, I present our technique for compiling the concurrent language Esterel into very efficient C code. Our technique minimizes runtime overhead by making most scheduling decisions at compile time and using a very simple linked-list-based event queue at runtime.
While these techniques work particularly well for Esterel with its high-level concurrent semantics, the same technique could also be applied to efficiently execute other concurrent specifications.