Compiler Schemes for Server-assisted Energy-efficient Mobile Computing on Handheld Devices

点击翻译按钮获取中文摘要

Handheld computing devices with wireless network connections have a great potential to be powerful mobile tools to access information and software resources in the networked world. Unfortunately, the computing power and connection time of current handheld devices are severely constrained by the short battery life. Moreover, the battery issue also places constraints on various design aspects including program memory and secondary storage, CPU, and operating system. With connection to computing servers through wireless networks, there exist many opportunities for handheld devices to save battery by carefully utilizing the power of the servers. This project explores compiler and run-time techniques to assist the partitioning of application programs into server tasks and client tasks. The optimal partition depends on available network bandwidth and on the nature of applications. In order for handheld devices to optimally exploit servers for a wide range of applications, careful tradeoff is made between energy consumed by network communication versus computation. The proposed compiler techniques and run-time support include compile-time and run-time analysis of work and communication requirement, exploitation of low-power networking states on handhelds, run-time decision-making, and a light-weight software coherence mechanism.