基于聚类的异步时钟SoC测试  

Clustering-Based Asynchronous Clock Periods SoC Testing

作　　者：凌立[1] 江建慧[1] 张颖[1] 王真[2] 

机构地区：[1]同济大学软件学院,上海201804 [2]上海电力学院计算机科学与技术学院,上海200090

出　　处：《电子学报》2017年第3期740-746,共7页Acta Electronica Sinica

基　　金：国家自然科学基金(No.61432017)

摘　　要：为进一步减少片上系统(System-on-Chip,SoC)测试耗时、降低测试成本,本文结合异步时钟测试机制,提出一种基于聚类的测试调度方法.该方法利用了SoC各测试的特征以及异步时钟测试的特点,对测试数据进行预处理.在ITC’02基准SoC集上,将本文方法与未采用异步时钟机制以及基于混合整型线性规划模型求解的方法进行对比.结果表明,本文的方法分别能平均减少测试耗时20.39%和5.53%,提升了调度算法的优化效率.并且在功耗约束较强时,最终调度结果与耗时下界仅相差0.9%.The clustering-based SoC test scheduling algorithm combined with the asynchronous clock periods testing is proposed to further reduce the SoC（ System-on-Chip） test application time（ TAT） and test cost. The scheduling algorithm pre-processes the test data by exploiting the characteristics of the tests. After conducting experiments on the ITC ＇02 SoC benchmark,we find out that the proposed scheduling method based on clustering can reduce TAT by 20. 39% and 5. 53% on average,when comparing with a synchronous clock testing method and an asynchronous method based on the MILP model,respectively. Besides,when the power constraint is tight,there is only a difference of 0. 9% between the scheduling result and the lower bound.

关 键 词：SoC测试调度 异步时钟 混合整型线性规划模型 聚类 

分 类 号：TP391.7[自动化与计算机技术—计算机应用技术]