As the fundamental hardware substrate of computing systems, chips undertake critical functionalities including instruction execution, data processing, and system resource management. Chips are ubiquitously deployed across personal computers, servers, smartphones, and IoT devices, serving as the cornerstone of modern digital economies. Chip design technology represents not only a strategically vital industry for national economic development but also a cutting-edge domain in scientific research, playing a pivotal role in advancing computer science and related disciplines. The design process constitutes an exceptionally complex systems engineering challenge, requiring deep cross-layer co-design between hardware and software architectures while simultaneously satisfying multifaceted design constraints encompassing functionality, performance, power efficiency, and area (PPA). These intrinsic characteristics have established processor chip design as one of the most technically demanding research frontiers in both academic and industrial spheres.
Chip Design consists of two parts: Automatic HDL Generation, which aims to train LLMs to generate HDL modules or fill the HDL code; and Automatic CPU Design, which fully automates the design of an entire CPU starting from I/Os.
Operating system aims to automatically generate optimized operating system configurations.
High-performance library aims to automatically optimize the performance of common operators (e.g., GEMM) and provides a complete automated optimization toolchain.
Compiler consists of two parts: Compiler Generation, which aims to leverage LLMs to complete automated compiler-generation tasks; and Tensor Program Transcompiler, which aims to automatically perform cross-platform tensor-program conversion.