September 2007
Return to
this issue’s stories
Meeting Predictivity Milestones with High-Performance Computing
High-performance computing (HPC) is needed to support the Predictivity milestones for certification. In particular, HPC can be used to eliminate knobs by solving the many microphysics issues in relevant regimes and over the relevant range of spatial scales. This is called Science@Scale. This strategy developed at Los Alamos involves multi-scale computing in which they couple multiphysics codes with unit-physics codes to obtain more physics-based design tools. For certification, the multiphysics codes must describe a variety of phenomena and, thus, must employ many approximate sub-grid physics models. In contrast, the unit-physics codes are nearly ab initio because they focus on specific issues and utilize few approximations. These include simulations such as molecular dynamics (MD) and particle-in-cell (PIC), which can be used to calculate basic material properties and transport rates under the realistic conditions set by the multiphysics codes. In order to achieve this multi-scale computing, they estimate needing petaFLOPS computing for a NIF-scale object and more for larger applications. This may require an extrapolation of HPC to heterogeneous architectures, such as Roadrunner (sited at Los Alamos National Laboratory), that employ a combination of conventional processors (Cores) and single-instruction, vector processors (Cells). Researchers at Los Alamos have successfully converted several unit-physics codes to the Roadrunner architecture, including PIC, MD, and hydrodynamics codes. For more information, contact Guy Dimonte at dimonte@lanl.gov.
Multi-scale computing will allow near ab initio simulation of a NIF-scale implosion.
DOE Privacy Disclaimer | Sandia Privacy Disclaimer
Last Modified September 27, 2007
Developed and maintained by Sandia National Laboratories for NA.114