Discrete Computing Systems

Discrete Computing Systems

The Discrete Computing Systems (DCS) group proactively enables fundamental scientific and computational advances in high performance computing-based discrete event simulation and optimization, reversible scalable computing, discrete element methods, and their applications aimed at designing novel computational solutions to critical problems within the mission spaces of DOE and related agencies.

Discrete Computing Systems image

Research

The DCS group performs research and development in the areas of discrete event simulations, discrete/distinct element methods, and discrete optimization with applications to a wide range of design and analysis problems directly in the mission of Department of Energy and in other agencies such as Department of Defense. Basic and applied research is focused on new algorithms, scalable runtime systems, massively parallel solvers, novel physical models, and analyses on a range of hardware including supercomputers, cloud, virtual machines, and accelerators such as GPUs.  Applications include Internet simulations, electric grid simulations, virtual machine-based simulations, parallel program simulations, epidemiological simulations, and social behavioral simulations.

The group also advances the theory and implementation of reversible computing at multiple levels, from hardware circuits up to application-level algorithms and reversible models, aimed at increasing the scientific understanding about the intersection of reversible computing and reversibility in physics.  Reversible computing algorithmic principles are applied to emerging quantum and superconducting hardware by complementing ORNL research in physics/hardware.

The group also serves the new national need for incrementally transforming a very large base of critical application software that is limited by sequential execution.  It fills the widening gap between production-quality sequential simulations and niche supercomputer-based parallel simulations, and also advances the techniques for incremental parallelization of large and complex sequential codes, with a focus on non-traditional and commercial scientific applications.  Validation-preservation techniques are developed and refined in conjunction with increasing scale and speed, with relevance to industry, government, and military applications.

Contact

Kalyan R S Perumalla

Distinguished R&D Staff Member, Group Leader