Minimizing Energy Consumption for Embedded Multicore Systems Using Cache Configuration and Task Mapping

摘要

Caches are known for their effectiveness in alleviating the speed gap between processor and off-chip memory. But its energy consumption is a concern. In this paper, we proposed two approaches based on cache configuration(cache reconfiguration and cache partitioning)and task mapping that aim to optimize the energy consumption of caches on embedded multi-core systems. The first approach is optimal and based on integer linear programming (ILP), whereas the second approach is a genetic algorithm (GA) that is near-optimal, but scalable with low overhead. Experimental results demonstrate that our ILP based approach can achieve 11.1% energy saving on average compared to previous techniques, GA based approach can reduce 9.7% energy consumption on average.