I am Meifeng Lin from Brookhaven National Laboratory. I am a computational scientist and the group leader of high-performance computing at the Computational Science Initiative. I have been working on the exascale application development project called “Lattice QCD.”
Lattice QCD is a numerical framework that allows us to simulate the strong force that binds the tiny particles called quarks and gluons together to form the visible matter. It does this by discretizing the four-dimensional space-time into regular grids, like a lattice—hence, the name lattice QCD. It is very computationally demanding, especially when we make the spacing between the grids smaller and smaller to mimic the continuous space-time and use simulation parameters that are closer to what nature hands us. The upcoming exascale computers, such as Aurora and Frontier, will allow us to perform lattice QCD simulations with much higher precision, and we can study much more complicated systems to understand better the strong force and potentially discover new laws of physics.
To efficiently use the exascale computers, we need to modify or redesign our software such that it can harness all the computing powers these new computer architectures have to offer. What I have been focused on is investigating ways to write performance-portable software, so that the same code can run on different architectures with reasonable performance. The ECP software ecosystem offers many tools at our disposal, and I have enjoyed utilizing different software technologies that support performance portability, which I can also apply to other applications beyond lattice QCD. To me, the most exciting thing is the possibility for a wide range of applications to run on the exascale computers outside of current ECP-supported projects.