Coordinated Optimization of the Performance of Processors and Memory in a Heterogeneous System under Energy Constraints

摘要

As the power consumed by systems continues to increase, reducing their power consumption is one of the goals of system optimization. However, power consumption is also gradually becoming one of the important constraints that affects the design and implementation of the system. At present, research on optimization with energy constraints is mainly focused on optimizing the energy of the processors, but the energy consumed by off-chip memory accounts for a large proportion of the total consumed by the whole system. In this research, we investigate the coordinated optimization of the performance of the processors and memory for a heterogeneous system subject to energy constraints. A communication-computation pipeline for parallel programs is constructed to optimize program performance by simultaneously scaling the voltage and frequency of the processors and memory using a task partition strategy. Our experimental results indicate that our method of optimizing system performance by coordinating task partition and frequency scaling under energy constraints can result in a satisfying execution time and processor profit and meet the given energy constraints.