Test Knowledge Data Base

Verification of designs has become extremely complex, time consuming with increasing design complexities and size, and has become the bottleneck of chip development. Designers and verification engineers rely on advanced techniques, methodologies and flows to capture design intent, test/verification plans to develop test suites to ensure that new RTL changes are being verified and certified daily. Regularly scheduled regressions are also running to ensure that the designs are being tested at unit and system-level. This has led to an explosion in the number of tests, generated logs and reports, which in turn leads to increased infrastructure spending. This paper introduces an approach of using big data solutions (management and architecture) to handle massive volumes of data test sets. This enables design and verification teams to build their own big data platforms. Furthermore, once the data management is built, it also provides an avenue to apply new techniques like machine learning to gain insight into test suites that should help identify right set of tests, find duplicate tests, prioritize/grade tests with the main intention of improving quality/productivity and reducing IT costs. This paper describes an exemplar system called Test Knowledge DataBase (TKDB) developed at Synopsys to meet the above goals.