Allocation and Scheduling for MPSoCs via decomposition and no-good generation

This paper proposes a decomposition approach to the allocation and scheduling of a multi-task application on a multiprocessor system-on-chip (MPSoCs). Based on our methodology, the target application running on top of the hardware platform is pre-characterized and abstracted as a task graph, with specification of computation, storage and communication requirements. The objective function of the overall problem is the minimization of communication cost. Once the communication cost has been minimized, among feasible schedules we prefer those having minimum makespan.