In 2016, the Department of Energy’s Exascale Computing Project, or ECP, set out to develop advanced software for the arrival of exascale-class supercomputers capable of a quintillion (1018) or more calculations per second. That meant rethinking, reinventing and optimizing dozens of scientific applications and software tools to leverage exascale’s thousandfold increase in computing power. That time has arrived as the first DOE exascale computer — the Oak Ridge Leadership Computing Facility’s Frontier — opens to users around the world. “Exascale’s New Frontier” explores the applications and software technology for driving scientific discoveries in the exascale era.
The scientific challenge
Computational chemistry holds the potential to make the discovery of game-changing new prescription drugs, industrial compounds and chemical processes as easy as point-and-click. High-resolution models could help scientists predict how tweaking formulas and combining novel materials and molecules could deliver a desired result — from a cure for cancer to enhanced heterogeneous catalysis, the process used by catalytic converters to reduce nitrogen oxide pollution. But the computational demands of such simulations, which can include sorting through the properties of billions of candidate compounds and potential chemical interactions, have generally proved unwieldy for most supercomputers.
Why exascale?
The GAMESS project — a joint effort led by scientists at Ames National Laboratory, Iowa State, Old Dominion University, the University of Texas at El Paso, Georgia Tech, EP Analytics and Australian National University — seeks to bring these discoveries within reach. The team has spent years adapting GAMESS, an established and widely used suite of computational chemistry programs consisting of millions of lines of code, to tackle such challenges.
“GAMESS can be used for any kind of problem, from physics to chemistry to biology,” said Mark Gordon, a distinguished professor of chemistry at Iowa State University who served as the leader for the GAMESS project. “We wanted to achieve these kinds of simulations without making the shortcuts that can compromise accuracy, so we implemented new algorithms to take advantage of exascale and spread out the work over Frontier’s thousands of processors.”
As a test problem, the team sought to simulate the path of a heterogeneous catalysis reaction along a nanoparticle of mesoporous silica — porous sand or glass. The problem consisted of more than 15,500 total atoms.
“These kinds of processes are going to be necessary for the kind of carbon capture and conversion we hope can be used to remediate fossil fuel pollution,” Gordon said. “The number of atoms and electrons involved in these reactions has traditionally been a stumbling block, but GAMESS succeeded.”
Frontier success
The GAMESS codes ran across 5,400 of Frontier’s more than 9,000 compute nodes. The simulations, which would have overwhelmed most computers, completed in less than an hour.
“This would not have been possible without Frontier,” Gordon said. “The challenge problem showed we were able to use these nodes efficiently and effectively.”
What’s next?
Next steps include expanding the simulation to include more particles and applying the codes to heavy elements with exponentially larger numbers of atoms.
“Now that we’ve shown how effectively GAMESS can take advantage of exascale, we’re excited to see what else can be explored,” Gordon said.
Support for this research came from the Exascale Computing Project, a collaborative effort of the DOE Office of Science and the National Nuclear Security Administration, and from the DOE Office of Science’s Advanced Scientific Computing Research program. The OLCF is an Office of Science user facility.
UT-Battelle manages ORNL for DOE’s Office of Science, the single largest supporter of basic research in the physical sciences in the United States. DOE’s Office of Science is working to address some of the most pressing challenges of our time. For more information, visit https://energy.gov/science.