AI for EDA
Search documents
论芯率先跑进AI for EDA产线:读芯片协议文档速度25倍,揪出respin级bug
量子位· 2026-03-29 00:51
Core Viewpoint - The article emphasizes that while the complexity of chip design doubles every two years, the efficiency of reading documentation has remained stagnant, leading to significant challenges in the verification process. The company Lunxin Technology has developed a system that automates the generation of verification code from chip protocol documents, addressing a critical gap in the EDA (Electronic Design Automation) industry [1][5][7]. Group 1: Challenges in Chip Design - The complexity of chip design increases significantly, with verification engineers often spending weeks or months reading extensive protocol specifications before writing code [2][4]. - Any oversight in this process can lead to costly respins, resulting in millions of dollars lost and extended project timelines [5]. Group 2: Lunxin Technology's Solution - Lunxin's system automates the process of generating usable verification code from chip protocol documents, significantly improving efficiency [6][7]. - In real customer projects, Lunxin's system has demonstrated the ability to identify critical bugs and timing violations at a speed 25 times faster than experienced engineers [11]. Group 3: Founders' Background and Vision - The founder, He Zhuolun, has nearly a decade of experience in the intersection of AI and EDA, which informs the company's approach to bridging the gap between academic research and practical engineering needs [9][13]. - The co-founder, Pu Yuan, has a strong academic background in EDA and has focused on the productization of technology rather than continuing in academia [14][16]. Group 4: Technical Approach - Lunxin's approach involves creating a knowledge graph from the specifications, which allows for automatic parsing and organization of information, identifying conflicts and inconsistencies that are often missed by human engineers [19]. - The system utilizes a large language model as a reasoning engine to generate necessary outputs based on the parsed knowledge graph, effectively linking documentation to executable verification code [20][22]. Group 5: Industry Positioning and Future Goals - The article highlights that the true measure of a company's position in the AI for EDA space is not just narrative but the ability to implement technology in customer production lines and achieve verifiable results [23]. - Lunxin aims to establish trust by proving its effectiveness in the most challenging scenarios before expanding its platform capabilities [24][25]. - The ultimate goal is to create an AI-native platform that automates and systematizes every aspect of the chip design process, gradually integrating AI capabilities into various stages of design and verification [25][26].
IDAS 2025 设计自动化产业峰会:巅峰聚首!80+顶尖行业专家领衔,即刻报名锁定参会资格!
半导体芯闻· 2025-08-21 10:26
Core Viewpoint - The article discusses the upcoming Intelligent Design Automation Summit 2025 (IDAS 2025), focusing on the rapid restructuring of the global semiconductor industry and the need for diverse supply chains and EDA technology advancements [3][4]. Event Overview - IDAS 2025 will take place on September 15-16, 2025, at the Hangzhou International Expo Center, aiming to gather key players from the semiconductor industry, including leading companies, universities, and research institutions [3][4]. - The summit will feature a main forum, 12 specialized forums, and related enterprise user conferences, expecting over 500 leading semiconductor companies and 2,500 professional attendees [4]. Themes and Topics - The summit will cover a wide range of topics from circuit-level to system-level design, including analog and digital circuit design, chip manufacturing, and packaging [3][4]. - Key discussion areas will include AI for EDA, 3D IC, and various technological roadmaps, with a focus on sharing the latest research findings and analyzing macro trends in the industry [4]. Agenda Highlights - The agenda includes an opening ceremony and main forum on September 15, followed by specialized forums covering topics such as custom design, memory design, and advanced packaging [11][12]. - On September 16, the main forum will address solutions for improving chip lifecycle yield, alongside various specialized discussions on AI technology applications in semiconductors [13][15]. Participation and Registration - Registration for the summit is currently open, with multiple ways to sign up, including through the EDA² official website and the "EDA Square" WeChat account [6][7][9].