China Launches the "Yisuan Ark" Full-Stack Platform to Enhance the Capability of Domestic Scientific Computing Ecosystem

The Computer Network Information Center of the Chinese Academy of Sciences, the University of Science and Technology of China, the Institute of Mechanics of the Chinese Academy of Sciences, and Sugon and other institutions announced the launch of the domestic computing system software ecosystem fullstack platform "YiSuan FangZhou" in Beijing on June 29. Centered on three major directions—algorithm supply, code migration, and intelligent applications—the platform provides an integrated adaptation and deployment solution for domestic GPU computing environments. According to the introduction, the platform's underlying "JiuYan Shu" algorithm library integrates 16 highperformance solvers, covering scenarios such as linear algebra, parallel computing, fluid simulation, biological computing, and deep learning, and has been adapted for domestic GPU architectures. The R&D team said that some core modules achieved significant performance improvements in testing. The platform also introduces a large codeconversion model to help developers migrate CUDA code and key modules in scientific computing software to domestic GPU software environments, simplifying the adaptation, compilation, and execution process. In addition, an automated simulation agent can support parameter configuration, automatic solving, and fault diagnosis in engineering scenarios such as fluid simulation, and can complete some tasks through natural language instructions. The developers said that this fullstack platform aims to promote collaboration among algorithms, code, and applications, unleash the potential of domestic computing power, serve scientific research scenarios such as AI for Science, and promote the continued development of the domestic scientific computing software ecosystem.