ACM SIGBED Review - Special Issue on the 2nd International Workshop on Adaptive and Reconfigurable Embedded Systems (APRES'09)
| Hi-index | 0.00 |
The introduction of embedded systems equipped with FPGA having a GPP contained inside them (Reconfigurable SoC (RSoC)) create a lot of challenges to OS for resource management. In distributed RSoCs, different applications may run on different RSoCs with variant resource requirements. Due to the variety of applications, a continuous change in demands from OS services (e.g. expected response-time) may exist, also a continuous change in the availability of resources (power and area). These variations can be managed by enabling the OS services to adapt their execution paths (on FPGA and GPP) depending on the application needs and the availability of resources. In this paper, an algorithm for distributed RSoC systems is introduced that enables OS services to execute on both FPGA and GPP along with a dynamic runtime change in execution paths of these services when needed. The algorithm relies on dynamic programming which provides single-criteria optima by taking each constraint alone. In the second step the algorithm finds a multi-criteria solution by local exchange small parts depending on the single-criteria optima solutions. In total a polynomial time heuristic multi-criteria optimization at runtime is obtained.