Synopsys is bringing artificial intelligence (AI) to its PrimeTime sign-off tool. It claims doing so can speed up the tool's power recovery algorithms fivefold when they are applied to leading-edge process nodes.
This enhancement of PrimeTime is one result of a multiyear strategy for Synopsys Design Group to make greater use of AI techniques, work that is being done with industry partners.
"The evolution of machine-learning technologies in recent years has been remarkable. As a long-time user of Synopsys tools, we were very interested in the new AI-enhanced signoff-driven power recovery in PrimeTime," said Hideyuki Okabe, director of the Digital Design Technology Department, Shared R&D Division 2, Broad-based Solution Business Unit at Renesas Electronics. "In our testing for a large SoC, we observed more than fourfold power ECO speed-up over our existing production flow, while achieving the same timing and power QoR. PrimeTime's new machine-learning technology allows us to complete ECO runs in hours versus the days it took before, and reuse learnings across blocks, hierarchies, and even design styles."
As processes become more complex, and designs more challenging, AI holds out the promise of being able to continuously learn and improve in customer environments. AI-enhanced tools should also boost designer productivity by speeding up computationally intensive analyses, predicting results to drive better decision-making, and applying past learning to guide debug strategies.
"Machine-learning has emerged as a powerful technology for addressing high-complexity, high-cost challenges, extending an avenue to innovation for our customers with disruptive potential," said Sassine Ghazi, co-general manager and corporate staff, Design Group at Synopsys. "Working closely with our development partners, we have seen many areas where AI technologies are delivering impressive results."