一种基于可重定向编译器的功耗优化框架

当今,低功耗设计成为系统设计中的关键问题之一,而编译中的低功耗优化也成为系统设计中的一个重要环节.文章针对传统功耗优化缺乏通用性的缺点,提出一个基于可重定向编译器的功耗优化框架.该框架通过对编译生成的二进制目标码进行横向再调度来降低指令总线上的高低电位切换次数,从而达到降低系统功耗的目的,并且,基于xpADL的支持,为该框架提供不同的体系结构描述,可以生成针对不同体系结构的功耗优化代码.以IA-64体系结构为例,在其仿真器Ski上作了大量实验,实验表明,对于静态代码,该框架的优化可达25%左右,对于动态代码,该框架可以达到30%以上的优化.因此,该框架的优化是行之有效的,并且具有相当的可扩展性.

A Low Power Optimization Framework Based on Retargetable Compiler

The low power design becomes the key problem of the system, and compiler's participation in low power optimization of the system plays an important role. This paper proposes a low power optimization framework based on retargetable compiler, which aims at the shortcomings of traditional low power optimization in generalization. The framework reduces the number of voltage switch on instruction bus by rescheduling the generated binary code in horizontal,which leads to the reduction of system power. And based on the support of xpADL, the framework could generate low power optimization code for different architecture. With IA -64 as the target architecture, massive simulation experiments are carried out using the HP's IA - 64 simulator Ski,the results show that the optimization could reach up to 25% for the static codes, and more than 30% for the dynamic codes. Therefore, this low power optimization framework is extensible and efficient.