Corresponding author: Alessio Bonfietti, DISI, University of Bologna. Tel.: +390512093271; Fax: +390512093953; E-mail: [email protected].
Abstract: As the number of processors integrated on a single chip increases with the fast pace dictated by Moore’s Law, multi-core systems-on-chip (MPSoCs) are becoming truly distributed systems at the micro-scale. From the application viewpoint, requirements for high performance and low power have increased at a breakneck speed in many embedded computing domains like wireless communication, audio and video processing. Applications in these areas are highly parallelizable and feature significant functional parallelism. In the past decades several programming models and tools aimed at easing the task of efficiently coding and mapping parallel applications. Most of them are tied to a specific platform or find far from optimal solutions due to the hardness of the allocation and scheduling problem. In this paper we presents (1) a Constraint Programming-based solver generating optimal scheduling and mapping decisions for such applications, optimized for maximal throughput, and (2) the framework used to validate the solver on a real multicore platform.