一种面向星载计算机的功能级功耗估计方法

An Approach for Functional Level Power Estimation of On-board Computer

在星载计算机的软硬件协同设计过程中需要解决处理器功耗估计问题.现有的工具和方法主要面向ARM或DSP处理器,缺乏可用于LEON3处理器的软件功耗估计方法.针对这一不足,基于功能级功耗分析方法进行功耗估计,重点解决LEON3处理器的功耗建模问题.提出了从C代码中提取模型参数的方法,并且对序贯最小优化支持向量回归算法进行了改进,提高模型的精度.在数值试验部分,通过常用的C语言算法试验对估计方法进行了验证.结果表明功耗估计误差小于3%,能够满足星载计算机软硬件协同设计的需要.

During the process of hardware/software co-design of on-board computer, power estimation of processor is an essential problem. The existing methods involving power estimation of processor mainly aim at ARM and DSP processors without the ability of dealing with LEON3 processor. To cope with this problem, based on the functional lever power analysis, a method is proposed to extract parameters from C code. At the same time, the sequential minimal optimization algorithm of support vector regression is im- proved to enhance the model precision. In the experiment section, the predicting precision of the pro- posed method is verified by some common algorithms programmed in C language. The results show that the predicting error is less than 3% , and the precision can meet the requirement for hardware/software co-design of on-board computer.