高级检索

    应用输入向量控制技术降低漏电功耗的快速算法

    A Fast Algorithm for Leakage Power Reduction by Input Vector Control

    • 摘要: 随着工艺的发展,为保证电路的性能和噪声容限必须降低阈值电压,这将导致漏电流呈指数增长,漏电功耗因而将逐渐超过动态功耗占据主导地位. CMOS的堆栈效应导致电路在不同向量下的静态功耗不同,因此在电路进入睡眠状态时使用输入向量控制技术是一种低功耗设计的有效方法,如何快速找到一个可降低电路漏电功耗的向量就成了问题的关键.介绍了一种在给定向量集合中查找低功耗向量的快速算法——基于概率传递的标记算法,并为此开发了一个事件驱动的门级组合电路仿真器.通过对ISCAS和龙芯处理器电路的实验结果表明,该算法同传统方法比较可以提高性能3.4倍,误差率仅约0.14%.

       

      Abstract: As IC technologies evolve, leakage current is increased exponentially with transistor threshold voltages reduced to maintain adequate performance and noise margins. As a result, leakage power will dominate total power dissipation soon. The stack effect of CMOS transistors makes leakage power different when different input vectors are applied. Input vector control is an effective method when a circuit enters sleep mode. It becomes critical to find a vector that minimizes leakage power to be statically applied to the primary inputs of a circuit. In order to evaluate various algorithms, a gate-level combinational circuit simulator is developed. An algorithm based on probabilities propagation is presented. Experimental results on ISCAS85 benchmark and Godson CPU show that this algorithm can accelerate calculation over triply with acceptable accuracy (error rate about 0.14%).

       

    /

    返回文章
    返回