Improving test pattern generation in presence of unknown values beyond restricted symbolic logic

Test generation algorithms considering unknown (X) values are pessimistic if standard n-valued logic algebras are used. This results in an overestimation of the number of signals with X-values and an underestimation of the fault coverage. In contrast, algorithms based on quantified Boolean formula (QBF), are accurate in presence of X-values but have limits with respect to runtime, scalability and robustness. Recently, an algorithm based on restricted symbolic logic (RSL) has been presented which is more accurate than classical three-valued logic and faster than QBF. Nonetheless, this RSL-based approach is still pessimistic and is unable to detect all testable faults. Additionally, it does not allow the accurate identification of untestable faults. In this paper, we improve test pattern generation based on RSL in two directions in order to reduce the accuracy-gap to QBF further. First, we present techniques to go beyond the accuracy of RSL when generating test patterns. Second, we include a check which is able to accurately identify untestable faults. Experimental results show the high efficiency of the proposed method. It is able to classify almost all faults - either by generating a test pattern or proving untestability.