Educating for Exascale

Some 25 graduate and post-graduate students recently spent four intense days preparing for the next generation of parallel supercomputers and exascale at the Parallel Computing in Molecular Sciences (ParCompMolSci) Summer School and Workshop hosted by Berkeley Lab.

Parallel Computing in Molecular Sciences (ParCompMolSci) Summer School and Workshop students pose outside the Brower Center in Berkeley. Select image to enlarge. (Photograph: Margie Wylie for Berkeley Lab)

Held August 6 - 9 at the Brower Center in downtown Berkeley, the event aimed to “prepare the next generation of computational molecular scientists to use new parallel hardware platforms, such as the [DOE] exascale computer arriving in 2021,” said Berkeley Lab Senior Scientist Bert de Jong, an organizer of the summer school and one of the scientists behind the DOE Exascale Computing Project’s NWChemEx effort. NWChemEx belongs to the less talked about, but equally necessary half of building exascale systems: software.

It was, in fact, NWChemEx principal investigators behind the summer school and workshop, said de Jong. “We recognized the need for training graduate students and postdocs in our field, and we partnered with our National Science Foundation counterpart MoISSI to make it happen,” he said. MoISSI is the Molecular Sciences Software Institute.

“Workshops like this are particularly valuable for helping you find out what it is you don’t already know,” said attendee Marjory Clement, a Virginia Tech graduate student. “You won’t come away being an expert in any particular thing being taught, but you learn about new directions you may want to take in your own work,” she said. Clement, for example, had no exposure to using GPUs (graphical processing units) until the workshop. The GPU is one technology that will be deployed in future exascale systems.

Learning new ways to wring the most out of even bigger and more capable systems is especially important to fields like biochemistry and pharmaceuticals that need to simulate large drug molecules and complex proteins. “We’re never going to have enough compute points. As soon as they come out with a bigger machine, we’re going to simulate the biggest molecule we can on that, and then we’re going to want something even bigger,” Clement said. This voracious appetite has made computational chemistry one of the driving applications for the nascent field of quantum computing, not coincidentally the subject of the gatherings’ final presentations.

Berkeley Lab’s Sam Williams and Khaled Ibrahim taught classes on intranode and hybrid programming models.

Other organizers of ParCompSci were Robert Harrison of the Institute of Advanced Computational Science (IACS) at Stony Brook University and Brookhaven National Laboratory, Ed Valeev of Virginia Tech, and Carlos Simmerling of the Laufer Center for Physical and Quantitative Biology at Stony Brook University. The DOE Exascale Computing Project and National Science Foundation’s MoISSI were major underwriters with additional funding from Berkeley Lab and IACS.

Parallel Computing in Molecular Sciences Summer School and Workshop participants, teachers and organizers pose outside the Brower Center. Select image to enlarge. (Photograph: Margie Wylie for Berkeley Lab)