I. Introduction
Application mapping on multiprocessor-systems-on-chip (MPSoCs) has been identified as one of the most important problems in embedded systems design [1]–[3]. Under dynamic workloads, run-time mapping is required in order to optimize for performance and energy consumption toward fulfilling the end user demands. The run-time mapping can be accomplished either with or without previously analyzed results. For mapping without previously analyzed results, i.e., on-the-fly processing, a large body of research exists [4]–[7]. Efficient heuristics have been used to assign new arriving tasks on the system resources. These heuristics may not guarantee schedulability and a high-quality mapping due to limited processing power at run-time.