Dynamic Adjustment of Test-Sequence Duration for Increasing the Functional Coverage

摘要

The importance of functional coverage during frontend verification is steadily increasing. Complete coverage statistics, possibly spanning from block- to top-level, are required as a proof of verification quality and project development status. In this work, we present a coverage-driven verification methodology that relies on coverage-directed stimulus generation, with the goal being to increase functional coverage and decrease test application time. The test application time given to each one of the available constrained-random test sequences is dynamically adjusted by a feedback-based mechanism that observes online the quality of each applied test. The higher the quality, the more cycles are assigned to this test for future trials. Misbehaving test sequences are automatically replaced by new ones, in order to spend verification cycles on other tests that actually improve functional coverage. The proposed methodology is successfully applied to the register renaming sub-system of a 2-way superscalar out-of-order RISC-V processor. The results demonstrate both increased functional coverage and reduced test application time, as compared to a purely random approach.