首页> 外文期刊>International Journal of Image, Graphics and Signal Processing >A CLB Priority based Power Gating Technique in Field Programmable Gate Arrays
【24h】

A CLB Priority based Power Gating Technique in Field Programmable Gate Arrays

机译:现场可编程门阵列中基于CLB优先级的功率门控技术

获取原文
           

摘要

In this work, an autonomous technique of power gating is introduced at coarse level in Field Programmable Gate Array (FPGA) architecture to minimize leakage power. One of the major disadvantages of FPGA is the unnecessary power dissipation associated with the unused logic/inactive blocks. These inactive blocks in a FPGA are automatically cut-off from the power supply in this approach, based on a CLB priority algorithm. Our method focuses on introducing gating into both the logic blocks and routing resources of an FPGA at the same time, contrary to previous approaches. The proposed technique divides the FPGA fabric into clusters of CLBs and associated routing resources and introduces power gating separately for each cluster during runtime. The FPGA prototype has been developed in Cadence virtuoso spectrum at 45 nm technology and the layout of the proposed power gated FPGA is developed also. Simulation has been carried out for a ‘4 CLB’ prototype and results in a maximum of 55 % power reduction. The area overhead is 1.85 % for the ‘4 CLB’ FPGA prototype and tends to reduce with the increase in number of CLBs. The area overhead of a ‘128 CLB’ FPGA prototype is only 0.058 %, considering 4 sleep transistors. As an extension to the proposed gating in ‘4 CLB’ prototype, two techniques for an ‘8 CLB’ prototype are also evaluated in this paper, each having its own advantages. Due to the wake up time associated with power gated blocks, delay tends to increase. The wake-up time however, reduces with the increase in sleep transistor width.
机译:在这项工作中,在现场可编程门阵列(FPGA)架构中引入了一种粗略的电源门控自主技术,以最大程度地降低泄漏功率。 FPGA的主要缺点之一是与未使用的逻辑/非活动模块相关的不必要的功耗。基于CLB优先级算法,通过这种方法,FPGA中的这些非活动模块会自动切断电源。与以前的方法相反,我们的方法着重于将门控同时引入到FPGA的逻辑块和路由资源中。所提出的技术将FPGA架构划分为CLB和相关路由资源的集群,并在运行时为每个集群分别引入功率门控。 FPGA原型已在Cadence的45纳米技术谱中进行了开发,并且还开发了所建议的功率门控FPGA的布局。对“ 4 CLB”原型进行了仿真,结果最大降低了55%的功耗。 “ 4 CLB” FPGA原型的面积开销为1.85%,并且随着CLB数量的增加而趋于减少。考虑到4个睡眠晶体管,“ 128 CLB” FPGA原型的面积开销仅为0.058%。作为对“ 4 CLB”原型中提议的门的扩展,本文还评估了“ 8 CLB”原型中的两种技术,每种技术都有自己的优势。由于与功率门控模块相关的唤醒时间,延迟趋于增加。然而,唤醒时间随着睡眠晶体管宽度的增加而减少。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号