摘要
无论是可靠性工程还是软件可靠性中的可靠性模型,都难以描述硬件故障在程序中的传播问题.首先建立了计算数据流模型,并以无穷存储机器的指令集为例,说明可以为任意程序建立计算数据流图.在计算数据流模型的基础上,进一步建立了错误流模型.把计算过程中的错误分成物理错误和传播错误两种,通过分析这两种错误的本质和传播规律,给出了6条有关错误传播的规则和2条独立定律.根据这些规则和定律,能够计算出在程序运行过程中,任意时刻在任意位置上出现错误的概率.最后以一个简单的无穷存储机器程序为例,简要地展示了错误流模型描述硬件故障在程序中传播的能力.
Neither reliability models in reliability engineering nor in software reliability can be directly applied to describe the propagation of hardware errors in programs. This paper first sets up a computational data flow model, and then explains that a computational data flow graph for the program can be built, using the instruction set of URM （unlimited register machine） as an example. Upon the computational data flow model, the error flow model is set up. Errors are categorized into two kinds： Original errors and propagated errors. By analyzing the propagation rules of these two kinds of errors, 6 assumptions about error propagation are given, upon which the probabilities of errors at any time and at any place in a program can be calculated. At last, a sample of URM program is given to demonstrate the capability of the fault flow model.
出处
《软件学报》
EI
CSCD
北大核心
2007年第4期808-820,共13页
Journal of Software
基金
SupportedbytheNationalNaturalScienceFoundationofChina
theNationalScienceFundforDistinguishedYoungScholarsofChinaunderGrantNo.60621003(国家自然科学基金创新研究群体)
关键词
软件实现的硬件容错
可靠性
计算数据流模型
错误流模型
错误传播
SIHFT （software implemented hardware fault tolerance）
reliability
computational data flow model
error flow model
error propagation
