The Computer Languages & Systems Software (CLaSS) Group advances the frontiers of high-performance computing by researching, designing, and building innovative software tools, libraries, and programming models. Our work focuses on parallel programming models, compiler and runtime development, and high-performance communication libraries—delivering the performance, scalability, and portability required by today’s most demanding scientific and engineering applications.
With deep expertise in performance optimization and the development of foundational technologies like GASNet, UPC, and UPC++, CLaSS empowers scientists, engineers, and developers to fully leverage modern supercomputers. We have a proven track record of leadership in the HPC community and are committed to open collaboration with partners across academia, industry, and government.
Programming Models, Compilers, and Runtimes
We research, design, and implement advanced parallel programming models such as UPC and UPC++, along with the compilers and runtime systems that enable efficient and scalable execution on modern HPC systems. A key focus is on performance optimization, ensuring that applications and tools fully exploit the computational capabilities of leading-edge supercomputers.
High-Performance Communication Libraries
We lead the development of network-independent communication middleware like GASNet, delivering low-latency, high-throughput communication essential for parallel applications. This expertise includes optimizing these libraries for maximum performance and portability across diverse HPC architectures.
Cross-Disciplinary Applications
We apply our technologies to solve large-scale scientific and engineering problems, demonstrating impact in fields such as genomics, epidemiology, and numerical linear algebra by enabling unprecedented scalability and efficiency in computational research.
Community and Standards Leadership
We actively contribute to language standardization, open-source software, and collaborative research, fostering innovation and interoperability within the global HPC community.
GASNet
UPC++
Berkeley UPC
Fortran at Berkeley Lab
Caffeine
Fiats: Functional Inference and Training for Surrogates
Julienne
Assert
Pagoda Paves the Way for Fast, Scalable Applications on Modern Supercomputers
Damian Rouson Named 2024 Developer of the Year by Berkeley Lab
Berkeley Lab’s Networking Middleware GASNet Turns 20