Exascale design space exploration and co-design

• Published in: Future generation computer systems 2014-01, Vol.30, p.46-58
• Main Authors: Dosanjh, S.S., Barrett, R.F., Doerfler, D.W., Hammond, S.D., Hemmert, K.S., Heroux, M.A., Lin, P.T., Pedretti, K.T., Rodrigues, A.F., Trucano, T.G., Luitjens, J.P.
• Format: Article
• Language: English
• Subjects: Co-design; Exascale preparation; High performance computing; Scientific computing
• ISSN: 0167-739X; 1872-7115
• DOI: 10.1016/j.future.2013.04.018

The co-design of architectures and algorithms has been postulated as a strategy for achieving Exascale computing in this decade. Exascale design space exploration is prohibitively expensive, at least partially due to the size and complexity of scientific applications of interest. Application codes can contain millions of lines and involve many libraries. Mini-applications, which attempt to capture some key performance issues, can potentially reduce the order of the exploration by a factor of a thousand. However, we need to carefully understand how representative mini-applications are of the full application code. This paper describes a methodology for this comparison and applies it to a particularly challenging mini-application. A multi-faceted methodology for design space exploration is also described that includes measurements on advanced architecture testbeds, experiments that use supercomputers and system software to emulate future hardware, and hardware/software co-simulation tools to predict the behavior of applications on hardware that does not yet exist. •A codesign-based methodology is described for exploring the exascale design space.•Codesign requires a multi-faceted approach.•Architecture testbeds are being used to study performance issues of key algorithms.•Network bandwidth degradation studies help define requirements for future systems.•The Structural Simulation Toolkit is described, with some example use cases.