In an interview, three U.S. Department of Energy (DoE) scientists discuss providing scalability and performance portability to prepare for exascale computing.
The Oak Ridge Leadership Computing Facility's Tjerk Straatsma stresses performance portability for transferring research between different facilities, and he says sharing these programming strategies is vital as the DoE sets up the latest generation of supercomputers at the national labs.
The National Energy Research Scientific Computing Center's Katerina Antypas cites the need for applications to leverage three central technologies to realize good performance on exascale computers — longer vector units, high-bandwidth memory, and numerous low-powered cores.
The Argonne Leadership Computing Facility's Timothy Williams notes developers/computational scientists and experimentalists and theorists are becoming increasingly codependent. In this context, Straatsma says pioneering exascale research should benefit the scientific community via large computer-enabled applications and methods "that can be translated to other codes or other application domains and be used to make these applications run very well on these new architectures."
View Full Article
Abstracts Copyright © 2017 Information Inc., Bethesda, Maryland, USA