Wednesday Keynote: Reinventing HPC with Specialized Architectures and New Applications Workflows

With the slowing of Moore's Law, the historical improvements in performance offered by successive generations of HPC systems is waning while costs for each new chip generation is growing. In the near term, the most practical path to continued performance growth will be architectural specialization in the form of many kinds of accelerators. New software implementations, and in many cases, new mathematical models and algorithmic approaches, are necessary to advance the science that can be done with these specialized architectures, but applications are just part of the HPC ecosystem, which is embedded in advanced scientific workflows. As application workflows become increasingly complex, including data analytics, AI, and HPC modeling and simulation – the demand for new programming models and tools will also grow. Each application and workflow may present unique requirements that will be difficult to address with a monolithic solution. Furthermore, AI's transformative role in these workflows, whether used in concert with traditional simulations (e.g., self-driving laboratories) or as surrogate models for traditional mechanistic models, brings ethical concerns that must be addressed, such as bias, explainability, and reproducibility. These emerging requirements will drive a new era of scientific and technological innovation. The scientific community is conscious of the underlying technological transformations and is eager to deploy these advances to accelerate scientific discovery but needs to learn how to integrate these new technologies with their existing research. Thus, they are turning to HPC expertise for innovative, pragmatic solutions. This keynote presentation, with a panel format moderated by Horst Simon and contributed by John Shalf and Rosa Badia, will focus on the emerging challenges and groundbreaking solutions at the forefront of reinventing HPC for discoveries.