【摘要】 为了提高模拟验证中的可观测性覆盖率,建立了一个逻辑约束编程(CLP)模型,定义了变量之间单步传播的条件.在此基础上,借助通用CLP系统自动地完成路径搜索、冲突发现、回溯和约束求解.实验结果表明,基于CLP的可观测性分析技术一方面可以生成更有效的测试向量,使得特定语句上的错误能够被传播到输出,加快模拟验证发现bug的过程;另一方面可以识别出不可观测的语句,避免盲目追求高覆盖率,节约模拟验证的资源.更多还原

【Abstract】 Traditional RTL test generation algorithms mainly focus on exciting a specific type of error,but consider little about the observability of errors. In this paper,we address the problem of observability analysis for assignment statements in RTL designs. The conditions of observing a variable are described as deduction rules,considering both control and data flow. Based on these deduction rules,constrain logic programming system will complete path searching and constraints solving automatically. This method can not only detect error in a specific statement,but also find function irrelevant code. The experimental results show that tests generated by our method can improve the observability-based code and branch coverage and speed up the error detection with little overhead.更多还原