Column-Based Precompiled Configurating Techniques for FPGA Fault Tolerance
Column-Based Precompiled Configurating Techniques for FPGA Fault Tolerance
复制标题
用于 FPGA 容错的基于列的预编译配置技术
DOI：
--
复制
发表时间：
2001
期刊：
影响因子：
--
通讯作者：
E. McCluskey
中科院分区：
文献类型：
--
作者：
Wei;E. McCluskey
在最近的现场可编程门阵列（FPGA）丰富的可配置逻辑元件和路由资源提供了一个成本效益的方法，容忍系统中的永久性故障。一旦发生永久性故障，FPGA可以通过用相同硬件中先前未使用的资源替换故障部分来重新配置。在本文中，我们提出了两个基于列的预编译配置技术，容忍永久性故障的FPGA为基础的系统。通过在设计阶段编译备选配置版本，这些方法确保了快速的重新配置，从而极大地提高了系统的可用性。此外，在不同的配置版本之间创建有意的相似性，以便通过差分编码和游程编码将预编译配置引起的存储开销降低几个数量级。实验和分析结果表明，我们的方法实现了显着的可靠性提高与小的配置存储开销。
The abundance of configurable logic elements and routing resources in recent Field-Programmable Gate Arrays (FPGAs) provides a cost-effective method for tolerating permanent faults in the system. Once a permanent fault occurs, the FPGA can be reconfigured by replacing the faulty part with previously unused resources in the same hardware. In this paper, we present two column-based precompiled configuration techniques for tolerating permanent faults in FPGA-based systems. By compiling alternative configuration versions in the design phase, these approaches ensure fast reconfiguration, and thus a tremendous increase in system availability. In addition, intentional similarities are created among different configuration versions so that the storage overhead due to precompiled configurations is reduced by orders of magnitude through differential coding and runlength coding. Experimental and analytical results show that our approaches achieve significant dependability improvement with small configuration storage overhead.