Employing Artificial Intelligence for the Accurate, Automated Prediction of Regression Failures to Accelerate Semiconductor Simulation and Emulation

 

 

LUND, SWEDEN, February 18th, 2021 ––Verifyter, the pioneer of automated regression debug technology, today released its latest version of PinDown-ML that leverages next-generation Artificial Intelligence (AI) algorithms to enhance the automatic identification of regression bugs during Integrated Circuit (IC) verification. PinDown-ML has been proven across multiple production designs to accelerate this process by four times (4X).

Verifyter’s new PinDown-ML product line will be on display at the DVCon US 2021 conference from March 1st to 4th, 2021.

“Regression debug remains one of the most onerous, error-prone and time-consuming tasks in modern IC verification,” remarks Lars-Eric Lundgren, Chairman of the Board, Verifyter and former CEO of HARDI (now part of Synopsys). “Verifyter has harnessed the full power of machine learning and AI to dramatically accelerate this process, shrinking the IC verification schedule while freeing up engineering bandwidth.”

The verification of ICs involves running test content on design code using large simulation installations or expensive emulation systems using a regression-style verification methodology. In regression verification, the design code and associated tests are updated regularly with additional code segments, or “commits,” and the verification process is re-executed, often every night and sometimes several times per day.

Some tests will fail causing the engineering team to manually sift through a large quantity of data to understand which commits caused these failures, and why. This lengthy, error-prone, and laborious activity occurs on a daily basis, and can take hours to complete. Sometimes parts of the verification process must be rerun to identify bug locations and mistakes can be made in identifying bad commits, further elongating this activity.

PinDown-ML performs this task automatically. Using Machine Learning (ML) it predicts the probability of specific commits causing failures prior to the regression runs. After the run it examines all regression failures and diagnoses which commits caused these failures. It will also rerun failing tests without the offending commits and demonstrate a passing regression. Bad commits are passed directly to the responsible engineer for repair. This technique eliminates the team debug process post-regression and has been shown to shorten the regression debug process by 4X in production environments. It also increases the accuracy of the process, eliminating wasted time and resources.

PinDown-ML has been enhanced with AI algorithms to increase its ability to predict failures, further accelerating this process. Its ML platform “learns online” where potential bugs occur over time, increasing this efficiency. It also may be “trained” using domain expertise to further improve the capability.

 

Verifyter at DVCon US 2021

Oscar Werneman, software engineer at Verifyter, will present a paper on March 3rd, 2021 at 10AM PST entitled “Supporting Root Cause Analysis of Inaccurate Bug Prediction Based on Machine Learning – Lessons Learned When Interweaving Training Data and Source Code”. More information may be found at the DVCon website.

Attendees can visit the Verifyter booth at the virtual conference to see a demonstration of PinDown-ML in action. Daniel Hansson, Chief Executive Officer at Verifyter, will hold two presentations at the virtual Verifyter booth:

“How does an Automatic Debugger work?”, March 2nd, 2021 at 2PM PST.

“Automatic Debug using ML”, March 4th, 2021 at 12.30PM PST

Break-out meetings may be scheduled for a more in-depth discussion on any aspect of the Verifyter technology.

About Verifyter

Verifyter is the pioneer of automated regression debug solutions. PinDown, its regression failure debug technology, makes use of AI algorithms to predict issues in large scale regression verification processes, thus automating one of the most time-consuming aspects of modern IC verification. Operating on both simulation and emulation verification farms, it commonly accelerates this process by four times over traditional manual methods. Based in Lund, Sweden and operating worldwide, Verifyter is privately held and counts among its customers many of the leading semiconductor companies. Visit http://www.verifyter.com/to learn more. 

 

Engage with Verifyter at:

Twitter: @verifyter

LinkedIn: linkedin.com/company/verifyter

PinDown and PinDown-ML are registered trademarks of Verifyter. Verifyter acknowledges trademarks or registered trademarks of other organizations for their respective products.

 

For more information, contact:

Daniel Hansson

+46 72 651 75 76

This email address is being protected from spambots. You need JavaScript enabled to view it.