LDRD: Holistic Approach to Energy Efficient Computing Architecture

John Shalf (PI)

The goal of this project is to develop new technology for energy-efficient computational science. The approach is to take a vertical slice through the space of applications, algorithms, software and hardware to perform a study on how to build the most energy-efficient system to solve a particular computation science problem. The initial target will be climate simulation, and the solution is likely to require higher degrees of parallelism in the climate algorithms and software, as well as new hardware designs based on low-power embedded technology. The result will be a proof-of-concept machine design, along with supporting analysis of how this compares to alternative designs, and new algorithms and software techniques as needed to make the case for feasibility of computing the science problem on the system.

The CoDEx project builds upon the research insights of this project.