What is the Co-Design Summer School?
The Los Alamos National Laboratory Co-Design Summer School was created to train future scientists to work on the kinds of interdisciplinary teams that are demanded by today’s scientific challenges. Launched in 2011, the summer school recruits top candidates in a range of fields spanning domain sciences, applied mathematics, computational and computer sciences, and computer architecture. Participants work together to solve a focused problem that is designed to build the skills needed to tackle the grand challenges of the future. Foremost among the skills on which we focus is the ability of students to work across disciplines with other team members, while employing their own unique expertise. This is the heart of Co-Design.
Past summer school challenges have included problems in kinetic theory (Boltzmann Transport Equation), molecular dynamics, hydrodynamics (Adaptive Mesh Refinement), quantum molecular dynamics, astrophysics (core-collapse supernovae and neutron star mergers), and tabulated equations of state. The summer school is hosted by the Applied Computer Science Group (CCS-7), led by Christoph Junghans.
What is Co-Design?
Co-Design is the social and technical equivalent of a multiple-constraint optimization problem. The rapid evolution of computing architectures and the expanding space between specializations in domain science and computer architecture means that it is virtually impossible for a single individual to cover all of the skills needed to solve current-day computational science challenges. Co-Design bridges this space through interactions between members of an interdisciplinary team. With the right amount of overlap, team members can communicate with each other effectively to solve a problem.
2022 Co-Design Summer School Focus:
Hardware/Software co-design for automated hardware pipelining.
The last decade has seen computer systems built from combinations of different processor types to improve performance, meet energy efficiency constraints, and provide unique capabilities for targeted workloads. At the same time, powerful open-source software and hardware projects have opened the door to additional opportunities for exploring the design and construction of tightly coupled software and hardware solutions that can tailor to unique, application-specific requirements. Several open projects span these areas, including the LLVM Compiler Infrastructure (https://llvm.org), the RISC-V instruction set architecture (https://riscv.org), and the Rocket Chip Generator (https://www2.eecs.berkeley.edu/Pubs/TechRpts/2016/EECS-2016-17.html). Co-design projects like this require a team of members across applications, programming language and compiler design, and system and processor architectures. In general, co-design plays a fundamental role in exploring the pros and cons of the various design choices that software developers and hardware architects can make across a combined software and hardware ecosystem.
The goal for the 2022 Co-Design Summer School will be to explore a subset of components across such a design space. Given the school’s duration, the school will restrict the scope to provide a suitable amount of time to execute multiple co-design phases, evaluate final results, and address shortcomings. The desired team will include a set of computational scientists, software toolchain designers, and hardware/processor architects. The final determination of the project scope will be formulated based on the makeup and experience of the chosen team, the assigned mentors, and the overall schedule and duration of the school.