Computational Scientist Ramesh Pankajakshan came to LLNL in 2016 directly from the University of Tennessee at Chattanooga. But unlike most recent hires from universities—students, grad students, or postdocs—Ramesh made a mid-career switch from research professor to professional researcher. At Chattanooga, Ramesh worked on development and applications of an in-house computational fluid dynamics (CFD) code called Tenasi. The applications ranged from drag reduction for Class 8 trucks to real time predictions of toxic plume propagation. Since it was a small group, in addition to directing student research, he also had to work on benchmarking, machine procurement, and deployment.
Ramesh became interested in high performance computing (HPC) while studying CFD. "CFD at the scale I was looking at—ships, submarines, trucks, small cities—was completely dependent on HPC. My first exposure to HPC was on the Cray XMP in 1993."
Livermore Computing (LC) offered new excitement at the cutting edge of HPC. Ramesh explains, "The GPGPU revolution is bringing about a sea change in the HPC and simulations fields. Unlike the vector revolution, which was the last time this happened, this will take a lot of effort on the software side to succeed. So it is exciting to be working on cutting edge hardware and software that can result in substantial real world impacts."
Ramesh recently finished porting a seismic code called SW4 to the CORAL machines. On a Lassen node, SW4 is now 28X faster than a CTS-1 node. This means that problems that were grand challenge problems that needed all of Cori can now be run multiple times a day on a third of Lassen. He is also using SW4 for performance regressions tests whenever the CORAL machines get a major upgrade.
Of his work at LC he says, "LC is involved in so many aspects of the HPC simulation space that there is no chance of your work day getting routine. If you get bogged working on a deployed machine you can always switch to something involving the next procurement or the early research into the one after that." Still, it's the real-world impact that keeps Ramesh engaged. He says, "Novelty is motivating only to the extent to which it can potentially impact the real world. So I would rather work on a promising evolutionary machine than on a radically new one that is not going anywhere."
As happy as Ramesh is, he does have one career regret: "There are too many interesting problems to solve, and selecting the ones to focus on is not always easy."