Utilizing PCM for Energy Optimization in Embedded Systems

Abstract

Due to its high density, bit alterability, and low standby power, phase change memory (PCM) is considered as a promising DRAM alternative. In embedded systems, especially battery-driven mobile devices, energy is one of the most important performance metrics. Therefore, it becomes an interesting problem of utilizing PCM for energy optimization in embedded systems. While recently there have been extensive studies on PCM, energy optimization with PCM in embedded systems has not been fully addressed. In this paper, we present a hybrid memory system architecture in which PCM is used to replace DRAM as much as possible so the system energy can be reduced by utilizing the lower standby power of PCM. However, to achieve this, system-level software optimization techniques are required in order to solve problems caused by the three disadvantages of PCM: namely, long write latency, large write energy and limited write endurance. We propose an optimal static data allocation scheme to solve a simplified problem, and discuss how to extend this to solve more complex problems. We also present emerging research issues in compiler optimization, real-time task scheduling and operating systems when utilizing PCM for energy optimization in embedded systems.