Creating an Energy-Efficient Improved-Resolution Weather Model Using Inexact Computing

Abstract

Scientists and engineers constantly seek better accuracy and improved resolution from supercomputer simulations.Unfortunately, the high energy cost associated with running supercomputers has become a significant obstacle to attaining the sought-after resolution improvement. Since energy costs will continue to increase, improved resolution will require better, more energy-efficient computational methods.Our proposed technical approach for overcoming the energy-cost barrier is based on a field of study known as inexact computing. We acknowledge that increasing resolution by using less exact computation is counterintuitive. We have discovered, however, by dividing a computation in to a savings phase followed by an accuracy-improvement phase, we can indeed achieve greater accuracy/higher resolution in an energy-efficient manner.In this proposal, we will apply our techniques to the problem of -Numerical Weather Prediction-. Furthermore, the code vehicle we have chosen for our study is a productionquality weather model known as openIFS.

Document Details

Document Type
DoD Grant Award
Publication Date
Jan 04, 2017
Source ID
N000141712178

Entities

People

  • Krishna Palem

Organizations

  • Office of Naval Research
  • Rice University
  • United States Navy

Tags

Readers

  • Applied Combinatorial Optimization and Logic Circuit Design.
  • Energy Conservation and Renewable Energy Engineering.
  • Parallel and Distributed Computing.