Arm has used machine-learning tools supplied by the Solido group at Siemens Digital Industries Software to speed up IP validation runtime a thousand-fold compared to conventional statistical methods.
The IP supplier used Graviton2 processors hosted by Amazon Web Services to further improve turnaround time. By migrating its Solido Variation Designer instances to multiple Neoverse-based processors in the cloud, Arm was able to cut compute costs by another 24 per cent and total CPU time by 12 percent. The turnaround time for the complete verification run reduced by 6 per cent.
“Improving compute cost and time-to-market is a constant priority for Arm,” said Philippe Moyer, vice president of design enablement at Arm. “Solido Variation Designer helps Arm to deliver these benefits to customers by providing greater verification accuracy and coverage in a fraction of the runtime.”
Solido Variation Designer offers a suite of tools for variation-aware design and verification. It uses machine learning methods to deliver accurate results while using orders-of-magnitude fewer simulations.
“Arm’s acceleration of IP validation cycles using Solido Variation Designer demonstrates the huge dividends that our investments in ML-powered technology are paying for our customers,” said Ravi Subramanian, senior vice president of IC verification at Siemens. “We are pleased to have helped engineering teams at Arm to accelerate IP verification for their next generation of semiconductor products. In addition, this successful customer deployment on AWS Graviton2 processors demonstrates our commitment to working with the Arm ecosystem and cloud partners to deliver scalable solutions to mutual customers.”