Never miss a story from Verification. Subscribe for in-depth analysis and articles.
When it comes to projects that involve cover groups and properties, and tests that utilize randomization, it can be challenging to ensure that the desired coverage goals are achieved.
Coverage merging involves combining the coverage data generated from multiple simulation runs to create a comprehensive coverage report. Achieving coverage closure can be time-consuming and requires careful management of the simulation environment. Additionally, identifying which coverage goals were not met and optimizing the testbench to achieve closure is a complex process.
The Xcelium Machine Learning (ML) App addresses these challenges by automating the simulation process and controlling the randomization of tests. It uses machine learning algorithms to learn from previous regression sessions and automatically optimizes the randomization for each simulation run. This reduces the number of simulation runs required to achieve coverage closure, reducing the time and resources required for the simulation process.
One of the unique features of the Xcelium ML is its ability to predict which coverage goals are most likely to be met in future simulations, helping users optimize the regression environment and achieve coverage closure more efficiently. It also allows users to generate a comprehensive coverage report that includes all coverage goals achieved in each simulation run, simplifying the process of coverage merging.
If you have a project with cover groups and properties, and you are running regression several times to merge them to get the desired coverage, Xcelium ML can help you ease the process, and this is typically where the app works best. What you would do in such a case is train Xcelium ML on one or two sessions of your regression. And then run Xcelium ML "N" times to saturate the coverage.
You should generally expect Xcelium ML to converge 2-3X faster than your full regression. Our recommendation would be to train the ML regression at the knee of your coverage curve. So, if that happens in one regression that is what you would train on, if it takes five sessions to get there then that is what you would use.
In conclusion, Xcelium ML is a powerful tool for automating simulations and controlling randomization for Cover Groups and Properties Projects.
Refer to our RAK example available on support.cadence.com