Probabilistic bug-masking analysis for post-silicon tests in microprocessor verification

Abstract

Post-silicon validation has become essential in catching hard-todetect, rarely-occurring bugs that have slipped through pre-silicon verification. Post-silicon validation flows, however, are challenged by limited signal observability, which impacts their ability of diagnosing and detecting bugs. Indeed, bug manifestations during the execution of constrained-random tests may be masked and be unobservable from the test's outputs. The ability to evaluate the bug-masking rate of a test provides great value in generating and/or selecting effective tests for high coverage regressions. To this end, we propose an efficient, static bug-masking analysis solution, called BugMAPI. BugMAPI tracks the information flow in a test program, and it estimates the probability that bugs go undetected by the checking mechanisms in place in the post-silicon platform. To achieve this goal, we leverage static code analysis and a novel, lightweight, probability estimation algorithm. We evaluated BugMAPI on a range of industrial constrained-random tests and a range of bug injection models, and we found that it can estimate bugmasking rates with an accuracy of 77% in 3 orders-of-magnitude less time, compared to an ideal dynamic analysis solution.

Ristretto: Random Test Generator for Multicore Microprocessor Cache Coherence Verification

“…In [11], a software tool is proposed which makes possible to assess the pool of potentially unrevealed bugs. Moreover, it allows the user to select effective tests with the smallest masking effect for the high coverage regression.…”

Random Test Generator for Multicore Microprocessor Cache Coherence Verification (Ristretto)

“…Оценка вероятности пропуска ошибок в тестах позволяет выбирать эффективные тесты для регрессии, а также экспериментально подобрать сбалансированное количество проверок на один тест таким образом, чтобы не упростить тестовые ситуации и сохранить вероятность нахождения трудно обнаруживаемых и редко проявляющихся ошибок. В работе [11] предлагается программный инструмент, который позволяет проводить анализ числа потенциально пропущенных ошибок. С его помощью можно отбирать тесты для базы регрессии, код которых минимально маски́рует ошибки.…”

“…In Proceedings of the 44th Annual IEEE/ACM International Symposium on Microarchitecture (MICRO-44). ACM, New York, NY, USA, 386-397 [11]…”