Alumnus Tobin Isaac on Gordon Bell Prize-winning Team

Date

A research team that includes Department of Energy Computational Science Graduate Fellowship (DOE CSGF) alumnus Tobin Isaac has won the prestigious ACM Gordon Bell Prize for outstanding achievement in high-performance computing (HPC).

Amanda Randles, a DOE CSGF alumna who’s now an assistant professor of biomedical engineering at Duke University, is lead author on another project that was an award finalist.

The honor was announced Thursday, November 19 at SC15, the annual supercomputing conference in Austin, Texas. The $10,000 prize, overseen by the Association for Computing Machinery, recognizes innovation in HPC applications for science, engineering and large-scale data analytics. Prizes highlight peak HPC performance and scalability, a code’s capacity to run efficiently on an increasing number of processors.

Isaac and fellows researchers from the University of Texas at Austin, IBM Research, New York University’s Courant Institute of Mathematical Sciences and the California Institute of Technology won for a detailed simulation of Earth’s interior. Their code achieved a world record 97 percent scalability. Most of the calculations were done on Lawrence Livermore National Laboratory’s Sequoia, an IBM BlueGene/Q supercomputer.

The code is designed to simulate mantle convection, a complex, constantly changing system that drives the movement of Earth’s tectonic plates, influencing things like earthquakes, volcanoes and mountain building. Accurate models can help scientists predict and understand these phenomena, but it’s an incredibly hard problem. Simulations must account for trillions of contributing factors.

The team used innovative algorithms and mathematical approaches to accurately predict the velocity, width, depth and motions of the Earth’s plates while also simulating mantle flow, an IBM release says. The team completed a full simulation in less than 24 hours.

Isaac, who earned his Ph.D. from UT-Austin, now is a postdoctoral researcher at the University of Chicago. His doctoral supervisor, Omar Ghattas, is a coauthor on the winning paper.

Randles’ project, “Massively Parallel Models of the Human Circulatory System,” emulates blood flow across a range of vessel sizes. Such models can help diagnose and treat hypertension and other maladies. The team’s three-dimensional, high-resolution model scaled to more than 1.5 million processor cores on Sequoia.