Power Ecosystem Unveils Domestic YiSuan FangZhou Full-Stack Platform

The convergence of artificial intelligence and scientific computing has reached a pivotal milestone. On June 29, the full-stack platform "Yisan Fangzhou," designed for China's domestic computing software ecosystem, was officially unveiled in Beijing. Developed collaboratively by the Center for Computer Network Information at the Chinese Academy of Sciences, the University of Science and Technology of China, the Institute of Mechanics at the Chinese Academy of Sciences, and Inspur, this platform delivers a robust, efficient, and practical software ecosystem tailored for domestic GPU computing infrastructure.
Despite recent progress, domestic heterogeneous computing environments continue to face significant hurdles in deploying scientific computing software. To bridge the gap between algorithms, code, and real-world applications, "Yisan Fangzhou" establishes a comprehensive technical pipeline spanning from foundational algorithms to intelligent applications.
At the core of its computing capabilities, the platform features the "Jiu Yan Shu" algorithm library, which integrates 16 high-performance solvers for key scenarios including linear algebra, fluid dynamics, bio-computation, and deep learning. These algorithms have been deeply adapted and optimized for domestic GPU architectures, delivering substantial performance gains in critical modules and offering researchers a powerful computational engine.
Code migration has long been a major bottleneck in adopting domestic computing resources. Leveraging the platform's advanced code conversion large language model, developers can rapidly migrate existing CUDA code to domestic environments. The system supports intelligent analysis and automated compilation, significantly reducing the time and effort required for code refactoring and enabling existing research workflows to run seamlessly on domestic platforms.
Furthermore, the platform incorporates an automated simulation intelligent agent that manages tasks such as parameter configuration, solving, and fault diagnosis through a multi-agent architecture. By simply providing natural language instructions, users can efficiently execute complex simulations, transforming scientific tools from highly specialized instruments into user-friendly solutions.
The launch of "Yisan Fangzhou" signifies that China's scientific computing software ecosystem has established a complete, closed-loop framework featuring robust algorithmic kernels, interconnected ecosystems, and practical business applications. Moving forward, the development team will continue to drive platform iterations, integrate closely with various computing hubs and research scenarios, and accelerate the adoption of AI-driven scientific research (AI4S) paradigms across China.
Related article
Mihoyo’s AI Companion BSide: Olivia Lin Shuts Down Weeks After Launch; Offline Version Available Today
miHoYo has confirmed that its desktop companion application, "BSide: Olivia Lin," will officially shut down on August 31. This closure comes less than a month after the software entered early access on Steam on July 13. To ensure users can retain cer
Snap Splits Generative AI Video Team to Launch Dotmo, Easing High R&D Costs
Snap, parent company of Snapchat, has officially spun off its internal generative AI video team into a new independent entity named Dotmo. This strategic decision addresses the high costs associated with in-house generative AI development, aiming to
Google’s Gemini 3.5 Pro Hit by Delays, Staff Exodus, and Military Ties
Axios reports that six current and former Google DeepMind staff members cite low morale as a key factor behind delays in launching Gemini 3.5 Pro. Bloomberg previously noted the model is months behind schedule, while recent smaller releases have draw
Related Special Topic Recommendations
Comments (0)
0/500

The convergence of artificial intelligence and scientific computing has reached a pivotal milestone. On June 29, the full-stack platform "Yisan Fangzhou," designed for China's domestic computing software ecosystem, was officially unveiled in Beijing. Developed collaboratively by the Center for Computer Network Information at the Chinese Academy of Sciences, the University of Science and Technology of China, the Institute of Mechanics at the Chinese Academy of Sciences, and Inspur, this platform delivers a robust, efficient, and practical software ecosystem tailored for domestic GPU computing infrastructure.
Despite recent progress, domestic heterogeneous computing environments continue to face significant hurdles in deploying scientific computing software. To bridge the gap between algorithms, code, and real-world applications, "Yisan Fangzhou" establishes a comprehensive technical pipeline spanning from foundational algorithms to intelligent applications.
At the core of its computing capabilities, the platform features the "Jiu Yan Shu" algorithm library, which integrates 16 high-performance solvers for key scenarios including linear algebra, fluid dynamics, bio-computation, and deep learning. These algorithms have been deeply adapted and optimized for domestic GPU architectures, delivering substantial performance gains in critical modules and offering researchers a powerful computational engine.
Code migration has long been a major bottleneck in adopting domestic computing resources. Leveraging the platform's advanced code conversion large language model, developers can rapidly migrate existing CUDA code to domestic environments. The system supports intelligent analysis and automated compilation, significantly reducing the time and effort required for code refactoring and enabling existing research workflows to run seamlessly on domestic platforms.
Furthermore, the platform incorporates an automated simulation intelligent agent that manages tasks such as parameter configuration, solving, and fault diagnosis through a multi-agent architecture. By simply providing natural language instructions, users can efficiently execute complex simulations, transforming scientific tools from highly specialized instruments into user-friendly solutions.
The launch of "Yisan Fangzhou" signifies that China's scientific computing software ecosystem has established a complete, closed-loop framework featuring robust algorithmic kernels, interconnected ecosystems, and practical business applications. Moving forward, the development team will continue to drive platform iterations, integrate closely with various computing hubs and research scenarios, and accelerate the adoption of AI-driven scientific research (AI4S) paradigms across China.
Mihoyo’s AI Companion BSide: Olivia Lin Shuts Down Weeks After Launch; Offline Version Available Today
miHoYo has confirmed that its desktop companion application, "BSide: Olivia Lin," will officially shut down on August 31. This closure comes less than a month after the software entered early access on Steam on July 13. To ensure users can retain cer
Snap Splits Generative AI Video Team to Launch Dotmo, Easing High R&D Costs
Snap, parent company of Snapchat, has officially spun off its internal generative AI video team into a new independent entity named Dotmo. This strategic decision addresses the high costs associated with in-house generative AI development, aiming to
Google’s Gemini 3.5 Pro Hit by Delays, Staff Exodus, and Military Ties
Axios reports that six current and former Google DeepMind staff members cite low morale as a key factor behind delays in launching Gemini 3.5 Pro. Bloomberg previously noted the model is months behind schedule, while recent smaller releases have draw





Home






