High Performance Computing for Large Eddy Simulations: Today and Tomorrow
Abstract
Turbulence, said to be the last unsolved problem in classical physics, is among the most important areas of research facing scientists and engineers today. The study of turbulence is essential to many of the Grand Challenges of Science identified by the Federal High Performance Computing Program. Grand Challenge problems such as understanding global environmental change, designing efficient combustion systems and developing controlled nuclear fusion all require the ability to simulate turbulence realistically. The economic impact of turbulence studies affects industries ranging from aerospace and automotive engineering to oil, medicine and electronics. Turbulence studies, including large eddy simulations, are also among the most computationally demanding research areas. The study of vehicle signature calls for sustained performance in the 10 GigaFLOPS range, and problems such as vehicle dynamics, ocean circulation, viscous fluid dynamics and climate modeling will require sustained performance of a TeraFLOPS (one trillion floating point operations/ second). Lacking these levels of performance, LES researchers are forced to parameterize their simulations, reducing a simulation's complexity to fit the available computational power.
Document Details
- Document Type
- Technical Report
- Publication Date
- Dec 01, 1990
- Accession Number
- ADA231220
Entities
People
- Justin Rattner
Organizations
- Intel Corporation