As noted in the Introduction, there are three major aspects to synthesizing our task graph representation for a parallel program: (1) synthesizing the static, symbolic task graph; (2) condensing the task graph; and (3) optionally instantiating a dynamic task graph representing an execution on a particular program input. Each of these steps relies on information about the message-passing program gathered by the compiler, although for many programs the third step can be performed purely by inspecting the static task graph, as explained in Section 3.3. These steps are described in detail in the following three subsections.