Careers | Phone Book | A - Z Index

CAL: Computer Architecture Laboratory for Design Space Exploration

Computer Architecture Laboratory for Design Space Exploration

CALLogoOnly

DOE mission needs in energy, national security and science require a thousand-fold increase in supercomputing performance during the next decade, under stringent power and reliability constraints. Architectures and applications must change at a dramatically increased pace to overcome Exascale challenges in energy efficiency, concurrency, performance, data locality, programmability, resilience and scalability. The current incremental approach to HPC system design is not sustainable. The DOE recognizes that a highly integrated hardware-software codesign process is essential to foster the rapid pace of innovation needed to create Exascale Computing systems that meet DOE mission needs. Therefore, for Office of Science and NNSA to be successful in their programmatic missions in constrained budget environments, it is essential that the most efficient systems possible be designed and built, or the historical improvements in simulation capability will stall. The entire computing industry faces many of the same challenges, and it is anticipated that the advances driven by Exascale technology will positively impact all scales of computing, from single nodes to racks to supercomputers.

The Computer Architecture Laboratory (CAL) will advance Exascale Design Space Exploration to develop energy efficient and effective processor and memory architecture R&D for DOE’s Exascale program. CAL will coordinate hardware architecture R&D activities across the DOE. CAL is a joint NNSA/SC activity involving Sandia National Laboratories (CAL-Sandia) and Lawrence Berkeley National Laboratory (CAL-Berkeley), which will provide the following essential services to the DOE.

Modeling and Simulation Infrastructure: CAL will assemble a common set of tools to establish a unified capability for quantitative analysis of the trade-space for hardware and algorithmic/software design. This ensures a uniformity and credibility of DOE analysis through a systematic methodology for performing design space exploration.

Proving Ground for Industry Technology Options: CAL will provide computer architecture staff with expertise in the processor, memory and advanced component technology areas to function as technical liaisons between the PathForward awardees or other industry technology development collaborators and DOE. These CAL staff will collaborate with awardees to provide reasonable drivers/ constraints for design evaluation and will be responsible for monitoring and summarizing for DOE.

Nexus for Codesign Interaction: CAL will act as a liaison among the codesign centers and commercial computing technology providers to facilitate the codesign process. The center will be able to identify cross-cutting concerns to reduce duplication of effort and manage the complex interactions between M centers and N technology development collaborators.

Computer Architecture Research and Exploration: CAL will explore novel hardware/ architecture concepts in cooperation with industry hardware technology developers and the DOE research community. CAL will lead the definition, development and evolution of proxy architectures to facilitate non-proprietary exploration of advanced architecture concepts. These tools will be used on behalf of the codesign centers and DOE X-Stack research projects to support the DOE mission.

The public-private partnership between industry and the DOE ensures development of technology that reduces economic and manufacturing barriers to constructing exascale systems, but also ensures the technologies developed through this cooperative design process have broad market impact for low-power embedded, cloud/datacenter, and midrange HPC applications. This ensures the DOE investment forms the center of a sustainable software/hardware ecosystem that is supported by applications across the broader IT industry.


Staff

  • Directors: James Ang, Principal Investigator (Sandia) and John Shalf, Principal Investigator (LBNL)
  • Steering: Sudip Dosanjh (LBNL) and Robert Leland (Sandia)
  • Novel Computer Architecture/ Technologies: Arun Rodrigues (Sandia) and Dan Burke (LBNL)
  • Industry/CDC Nexus: Robert Hoekstra (Sandia) and Erich Strohmaier (LBNL)
  • Design Space Exploration support for FastForward and DesignForward Industry Collaborators: James Ang (Sandia) and Nicholas Wright (LBNL)

Related Projects

CoDEx: https://sites.google.com/a/lbl.gov/codex/
SST: http://code.google.com/p/sst-simulator
Mantevo Proxy Apps: http://www.mantevo.org