TIME: 3:30 PM

LOCATION: GMCS 314

SPEAKER: Damian Rouson, Lawrence Berkeley National Laboratory

ABSTRACT: The advent of AI-generated code offers an opportunity to shift research software engineers’ activity away from fine-grained algorithm implementations and towards high-level problem specifications, code review, and software verification. This talk will demonstrate how Berkeley Lab’s Formal open-source the software framework addresses each of these concerns for partial differential equation solvers. Developed as part of a hardware/software co-design project aiming to accelerate stencil computations, Formal defines software abstractions that mimic vector calculus concepts. Formal supports these software abstractions with a discrete calculus defined by the mimetic discretizations of Corbino & Castillo (2020) and the staggered-grid interpolation operators of Dumett & Castillo (2022). The talk will demonstrate how Formal supports a level of expressiveness that eases problem specification and aids human comprehension. The talk will describe Formal language-based approach to performance portability with plans for running on heterogeneous hardware in shared or distributed memory solely using language features with no explicit reference to parallel programming models such as MPI, OpenMP, OpenACC, or CUDA. The talk will demonstrate how Formal aims to lower barriers to the formal verification of scientific programs by leveraging a powerful new type-safe generic programming technology to convert unit tests into correctness proofs.

BIO: Damian Rouson is a Senior Scientist and the Group Lead for the Computer Languages and Systems Software (CLaSS) Group at Berkeley Lab and an Adjunct Faculty at San Diego State University. He researches programming paradigms, patterns, and languages for high-performance computing. He leads the development of several open-source software packages, including the Julienne correctness-checking framework, the Caffeine parallel runtime library, the Fiats deep learning library, and the Formal embedded domain-specific language (DSL). He founded Archaeologic Inc. and Sourcery Institute. He holds a B.S. from Howard University and a M.S. and Ph.D. from Stanford University, all in mechanical engineering. He is a licensed Professional Engineer (P.E.) in the State of California. He co-authored the textbook Scientific Software Design: The Object-Oriented Way (Cambridge University Press, 2011) and has taught related short courses and graduate courses in approximately 40 venues. He has held faculty, staff, and visiting positions at the City University of New York; the University of Maryland, College Park; the University of Cyprus; the University of Bergen; and Stanford University. He has held research and technical management positions at the U.S. Naval Research Laboratory and Sandia National Laboratories.

HOST: Jose Castillo