FaLoFiN:基于不确定性过滤的硬件故障局部化方法  

FaLoFiN:a Hardware Fault Localization Approach via Filtering Nondeterminacy

作　　者：郑衍松[1,2] 佟冬[1] 李皓[1] 王克义[1] 程旭[1] 

机构地区：[1]北京大学微处理器研发中心,北京100871 [2]北京大学深圳研究生院,深圳518055

出　　处：《计算机辅助设计与图形学学报》2010年第7期1207-1214,共8页Journal of Computer-Aided Design & Computer Graphics

基　　金：国家"八六三"高技术研究发展计划(2006AA010202);国际科技合作计划(2008DFB10010)

摘　　要：运行已有大型商业软件进行系统级验证是获得高质量SoC设计的有效手段,然而一旦硬件故障导致该类软件运行失效,故障局部化过程将十分困难.为此提出一种基于不确定性过滤的硬件故障局部化方法FaLoFiN.首先提取不确定性函数调用集;然后分别从参考运行和故障运行中过滤不确定性函数调用集;最后通过比较、定位异常函数调用,实现与硬件故障相关代码范围的局部化.实验结果表明,FaLoFiN方法具有较高的硬件故障局部化准确度,能有效地将PKUnity-86系统中的真实硬件故障局部化.System level verification by running existing commodity software stacks is essential for achieving a high quality SoC design. However, it is hard to find the root cause of hardware faults when running large commodity software stacks. A hardware fault localization approach by filtering out nondeterministic behaviors is proposed. This approach is called FaLoFiN. The approach first extracts the nondeterministic function call set. Then it filters out this nondeterministic function call set from a reference run and a faulty run respectively. Finally it localizes the code scope related to hardware faults through comparing the filtered function call sets, and locating the abnormal function call. The experimental results show that FaLoFiN can localize injected hardware faults with high accuracy. Moreover, FaLoFiN localized real hardware faults in the PKUnity -86 system effectively.

关 键 词：硬件故障 故障局部化 不确定性过滤 异常函数调用 

分 类 号：TP306.3[自动化与计算机技术—计算机系统结构]