Both teams, called SciDAC Institutes, are part of the Scientific Discovery through Advanced Computing program.
Rob Ross, a senior computer scientist in Argonne’s Mathematics and Computer Science (MCS) Division, will lead “RAPIDS2: A SciDAC Institute for Computer Science, Data, and Artificial Intelligence.” A collaboration of six universities, six national labs and a software research and development company, RAPIDS2 will focus on helping application developers address challenges arising with the increasing deluge of data, new technologies and extreme-scale computing.
Ross, who has been director of the SciDAC-4 RAPIDS Institute for Computer Science and Data, said he was looking forward to bringing together DOE scientists, computer scientists and applied mathematicians, as well as staff at DOE facilities, in addressing these challenges.
“AI offers an exciting new approach for tackling complex problems spanning from chemistry and materials science to biology and cosmology,” Ross said. “New methods for applying AI techniques have the potential to accelerate science in a variety of domains in which large-scale high-performance computing also plays a key role.”
Todd Munson, a senior computational scientist in Argonne’s MCS Division, is the Argonne lead for the other new SciDAC Institute, “Frameworks, Algorithms, and Scalable Technologies for Mathematics (FASTMath).” Led by Lawrence Berkeley National Laboratory and involving five national labs and four universities, the FASTMath Institute will focus on the development of new mathematical techniques that can exploit the power of extreme-scale computers.
Munson has been the Argonne lead and numerical optimization area lead of the DOE SciDAC-4 FASTMath Institute. Not only has he developed new optimization solvers, but he has worked with the partner institutions to create an integrated optimization plan for the institute.
“Several challenging issues must be resolved to achieve efficient and scalable performance on emerging exascale architectures,” Munson said. “For example, what algorithms provide the best performance under power or energy constraints? What numerical methods will perform best when using GPUs? The new funding will enable our FASTMath team to explore novel approaches and devise numerical tools to address these issues.”
Total funding for the two SciDAC Institutes is planned for $57.5 million over five years.
For the press announcement from DOE, see https://www.energy.gov/articles/department-energy-provide-575-million-science-computing-teams. A list of lead and partner institutions for the two teams can be found on the homepage of the DOE Office of Science, Office of Advanced Scientific Computing Research, under the heading, “What’s New.”