While studying aerospace engineering, Brian Gunney became fascinated with the field of computational fluid dynamics because he thought it could be critical in solving many problems he considered unsolvable. After receiving the Department of Energy’s Computational Science Graduate Fellowship, he was enticed even further into the computational aspects of his work. Now, as a computer scientist and programmer, Brian’s days are devoted to improving the mathematical and computational underpinnings of the simulations used to solve engineering and science problems.

“I develop and test new algorithms and codes to support adaptive mesh refinement, which is the use of dynamic, variable-resolution meshes in simulations,” Brian says. He has developed a new, more scalable way of managing adaptive meshes on massively parallel computers. Because the meshes and the job of mesh management grow concurrently with computers and their simulations, coordinating all the processors during mesh adaptation evolved into a classic non-scalable approach. Brian’s approach distributes the mesh management data and work across the processors, similar to how the computational work is distributed, and dramatically reduces the mesh adaption cost.

“Because we don’t know where the key features will appear or move, each processor had to build a global picture of the mesh by acquiring information from all other processors,” Brian explains. “The global information was expensive to communicate, store, and process. When the mesh is very large, it fails to even fit into memory. We have to know what the mesh we’re adapting looks like, but no processor ever sees the global picture.”

LLNL’s SAMRAI framework uses the new approach, and it is in turn used by other codes to break new ground in dynamically adaptive simulations. Simulations using SAMRAI on one million processors are ten thousand times bigger than the point at which the scalability problem was first noticed.

Away from the Laboratory, Brian spends his free time volunteering with AFS (originally called American Field Service), a nonprofit organization that has been leading international high school student exchanges for nearly 70 years. Brian serves as a liaison, the front line of communication with AFS. He also trains other liaisons and helps out on the more challenging cases. Volunteers can use their talents and interests to support AFS. Because Brian loves the outdoors, he organizes backpacking, camping, and hiking trips to expose visiting students to the natural wonders of Northern California. In addition, he co-leads wilderness backpacking trips each summer for the San Francisco Bay chapter of the Sierra Club.

Brian values achieving tangible results, both in his free-time pursuits and in the workplace. “I like breaking new ground in science and technology and seeing my work make a difference toward solving real-world problems," he says. Plus, there are other perks to his job: "I get to work with some of the smartest and most dedicated people I’ve ever met. The supercomputers are pretty fast, too.”