Supercharging Programming Through Compiler Technology

Explore a seminar on advanced compiler technology and its impact on scientific computing and machine learning. Delve into William Moses' presentation from the University of Illinois Urbana-Champaign, focusing on high-level abstractions within general-purpose compilers. Discover how these innovations enable the automatic generation of fast, portable, and composable programs. Examine the application of this approach in two critical domains: automatic differentiation and parallelism. Learn about Enzyme, a compiler that automatically generates derivatives from existing computer programs without modifying the original application. Understand how this technology combines differentiation with program optimization, resulting in significantly faster code. Gain insights into the potential future applications of this domain-agnostic compiler approach, including its use in probabilistic programming. Sponsored by the MFEM project, this FEM@LLNL Seminar Series talk offers valuable knowledge for those interested in finite element research and applications.