Exascale Day Audio: Sherry Li

Impact Area: Mathematical Libraries/Algorithm Improvement

Sherry Li, Lawrence Berkeley National Laboratory

Transcript

Hi. I’m Sherry Li. I am a senior scientist in the Applied Math and Computational Research Division at Lawrence Berkeley National Laboratory. In ECP, I lead the mathematical libraries area in Software Technology. I am also principal investigator of an ECP project, delivering SuperLU and STRUMPACK sparse direct solvers and preconditioners.

Scientists and engineers develop mathematical models to study various physical phenomena at multiple time and spatial scales. For example, climate modeling needs a higher resolution for accurate study of climate changes at different locations. This requires better algorithms, and this is the role of mathematical libraries. They enable scientists to reuse algorithms and data structures developed by experts and cope with the complexity and changes of the underlying computer architectures. The solvers developed in the math libraries are the core computing engines for many of ECP’s modeling and simulation codes as well as the data analytics codes.

We are excited about exascale computing because it gives us a great opportunity to conduct cutting-edge, breakthrough research in both algorithm and software, pushing the scaling boundaries to an unprecedented level. We are extremely excited about the collaboration opportunity for the DOE’s [US Department of Energy’s] leading mathematical software developers, the domain scientists, and the computer vendors to work together in order to fully harness exascale computing power.

Collectively, the math libraries provide advanced solution methods in discretization, dense and sparse linear algebra, nonlinear and time stepping, optimization, sensitivity analysis, and uncertainty quantification. We envision that the math libraries will contribute to a larger software ecosystem deployed by a very broad HPC [high-performance computing] community beyond ECP.