{"id":9590,"date":"2023-10-11T19:46:41","date_gmt":"2023-10-12T02:46:41","guid":{"rendered":"https:\/\/mattfife.com\/?p=9590"},"modified":"2023-10-01T19:54:07","modified_gmt":"2023-10-02T02:54:07","slug":"nvidia-uses-ai-for-place-and-route-on-its-chips","status":"publish","type":"post","link":"https:\/\/mattfife.com\/?p=9590","title":{"rendered":"nVidia uses AI for place and route on it’s chips"},"content":{"rendered":"\n

nVidia just published a\u00a0paper<\/a>\u00a0and\u00a0blog post<\/a>\u00a0revealing how its AutoDMP system can accelerate modern chip floor-planning using GPU-accelerated AI\/ML optimization, resulting in a 30X speedup over previous methods.\u00a0Hopefully it doesn’t get the treatment the Google AI place-and-route solution got<\/a>.<\/p>\n\n\n

\n
\"\"<\/figure>\n<\/div>\n\n\n

AutoDMP is short for Automated DREAMPlace-based Macro Placement. It is designed to plug into an Electronic Design Automation (EDA) system used by chip designers, to accelerate and optimize the time-consuming process of finding optimal placements for the building blocks of processors. In one of Nvidia\u2019s examples of AutoDMP at work, the tool leveraged its AI on the problem of determining an optimal layout of 256 RSIC-V cores with 2.7 million standard cells and 320 memory macros. AutoDMP took 3.5 hours to come up with an optimal layout on a single\u00a0Nvidia DGX Station A100.\u00a0<\/p>\n\n\n\n

Initial metrics shows it does an amazing job – in a fraction of the time. Definitely worth the read.<\/p>\n\n\n\n

AutoDMP is open source, with the code published\u00a0on GitHub<\/a>. Below is a link to an article about Cadence’s Cerebrus AI place-and-route solution.<\/p>\n\n\n\n

Article:<\/p>\n\n\n\n