A Portable Infrastructure Supporting Global Scheduling of Embedded Real-Time Applications on Asymmetric MPSoCs

Multiprocessor systems-on-chip (MPSoCs) open notable perspectives in the design of highly-targeted embedded solutions for real-time multitasking applications. The heterogeneity of available platforms, however, often hinders plain applicability of efficient scheduling policies, particularly in the case of loosely coupled architectures which do not provide any support for inter-processor tasks migration. In this paper we present a portable software infrastructure addressing the global scheduling of periodic real-time tasks on such platforms. We show that global scheduling policies, under the restrictedmigration model, are applicable also on asymmetric multiprocessing systems and experimentally evaluate the validity of the approach using different FPGAbased configurations that recall manifold architectures of commercial MPSoCs.