Co-design at Lawrence Livermore National Lab
The Exascale Co-Design Center for Materials in Extreme Environments (ExMatEx) was established in August 2011 to address the goal set by the Department of Energy to build an exascale supercomputer capable of performing 1018 floating point operations per second by 2020. Such a simulation capability will play a key role in solving many of the nation’s most pressing problems, including producing clean energy, extending nuclear reactor lifetimes, and certifying the aging nuclear stockpile.
The purpose of ExMatEx is to develop a materials modeling capability for exascale computing. Creating systems with 1,000 times the performance of today’s fastest machines requires extraordinary innovation across the entire computing ecosystem simultaneously. By coordinating the development of hardware, middleware (software stack), programming models, and algorithms, ExMatEx will ensure that all of these elements can be used together on future platforms to achieve efficient multiphysics simulations of materials in extreme mechanical and radiation environments.
ExMatEx participants include researchers from Lawrence Livermore, Los Alamos, Sandia, and Oak Ridge national laboratories; Stanford University; and the California Institute of Technology. The project’s target materials science applications are multiscale/multiphysics problems that combine techniques such as classical molecular dynamics, mesoscale modeling (phase field and dislocation dynamics), and continuum modeling.
ExMatEx focuses effort in four primary areas:
- Scale-bridging algorithms: ExMatEx employs uncertainty quantification-driven adaptive physics refinement, in which coarse-scale simulations dynamically spawn tightly coupled and self-consistent fine-scale simulations as needed.
- Programming models: ExMatEx uses a task-based, scale-bridging programming approach that leverages the extensive concurrency and heterogeneity expected at exascale while enabling novel data models, power management, and fault tolerance strategies within applications. Models and approaches will be broadly applicable to a variety of multiscale, multiphysics applications.
- Proxy applications: The project relies on proxy applications to communicate workload requirements to hardware architects and system software developers. Proxy applications also serve as a test bed for new algorithms and programming models.
- Analysis and optimization: ExMatEx researchers are developing the GREMLIN framework, which provides a flexible and extensible framework to emulate expected exascale architecture characteristics and constraints on today’s machines. Currently, the project is targeting power and thermal constraints, resource limitations in the memory system and the network, soft and hard faults, and the impact of noise.
ExMatEx Proxy Apps can be found at: github.com/exmatex
Are you a vendor or researcher interested in exploring co-design with LLNL?
For additional information, please contact our team.