Identifying Communication Models in Process Networks derived from Weakly Dynamic Programs

Dmitry Nadezhkin and Todor Stefanov
LIACS, Leiden University


Process Networks (PNs) is an appealing computation abstraction helping to specify an application in parallel form and realize it on parallel platforms. The key questions to be answered are how a PN can be derived and how its components can be realized efficiently on a given parallel system. In this paper we present a novel approach of communication model identification in PNs derived from nested loops programs with data-dependent control statements, which we call Weakly Dynamic Programs (WDP). Identifying communication models at compile-time allows us to select the most efficient realization of communication components in a PN. We show how our approach is seamlessly integrated into our existing procedure of automated PN derivation from WDP programs. This paper can be considered as an important complementary work that makes our automated derivation of WDPs to PNs complete.