Abstract
The future of scientific computing hinges on the ability to effectively support converged workloads — workloads that incorporate elements of traditional High-Performance Computing, AI/ML, and data analytics. Historically, these computational paradigms have been isolated into distinct and divergent silos, with separate programming languages, programming models, and even hardware architectures supporting each. However, this artificial separation in now resulting in roadblocks to scientific progress as computational scientists increasingly desire to merge these paradigms into single tightly-integrated workflows. At Pacific Northwest National Laboratory, we have been working to bring principles of codesign together to create hardware and software computing systems that will effectively support this new era of scientific computing. In this talk, I will describe some of these efforts and point to how we envision the future of computing within the US Department of Energy.
Bio
Kevin J. Barker is a Chief Scientist and Group Lead for the Physical and Computational Sciences Directorate’s High-Performance Computing (HPC) Group with over 16 years’ national laboratory experience leading impactful Computer Science research. Research interests include emerging and novel computing architectures; hardware/software codesign to support future scientific computing workflows; large-scale runtime software; and performance modeling, prediction, and analysis. Dr. Barker serves Principal Investigator for the DOE Office of Advanced Scientific Computing Research (ASCR) funded Center for Advanced Technology Evaluation (CENATE), which aims to understand the impact of future computing technologies on scientific workloads of interest, and as Chief Scientist for the laboratory-funded Data-Model Convergence (DMC) Research Initiative.
