By Eric Shaw
Office of the Vice President for Research
The Michigan Institute for Computational Discovery and Engineering (MICDE) welcomed distinguished scientists to discuss emerging and future paradigms for High Performance Computing during its 2023 symposium, held on Friday, March 24.
“Computational advancements have reached a pivotal juncture, empowering researchers to make progress in scientific discoveries and engineering innovations. This is enabled by the confluence of algorithms, software and hardware, and it is critical for experts across disciplines to work together to continue the pace of progress and achieve desired outcomes,” said Karthik Duraisamy, Director of Michigan Institute for Computational Discovery and Engineering.
The symposium featured a wide range of topics related to high performance computing and other computational science-related issues, demonstrating the multidisciplinary nature of the field. These leading-edge developments highlight the vast potential for the field to address some of the most significant challenges facing society today. From improving weather and climate forecasting to advancing materials science and drug discovery to understanding nuclear fusion, the power of high-performance computing is truly remarkable.
Dr. Fariba Fahroo, program officer at the Air Force Office of Scientific Research (AFOSR), spoke on challenges, opportunities, and national needs in computational science. During the talk, she introduced AFOSR and the computational mathematics program she manages. She also discussed the challenges and new directions emerging in computational mathematics as a field bridging areas in applied mathematics and computational science. Dr. Fahroo shared her insights on large-scale projects in machine learning for modeling physical systems, rare events, data assimilation, and reduced order modeling. The talk highlighted the importance of basic research programs in various areas of computational math and control theory, such as multiscale modeling and computation, design under uncertainty, distributed, multi-agent control and estimation, and computational control theory.
Stanford professor, Alex Aiken, presented Legion, a programming model, and runtime system that is designed to handle the increasingly complex and hierarchical nature of modern machines. Aiken discussed the design of Legion, its rationale, and recent work in developing these libraries before highlighting the importance of considering data movement in parallel programming and the potential for Legion to improve the efficiency and productivity of programming for modern machines. Aiken also highlighted usability – particularly demonstrating how simple codes written for course projects scale to a massive number of GPU nodes.
Princeton University professor of astrophysical sciences, Amitava Bhattacharjee, presented his research on the High-Fidelity Whole Device Model of Magnetically Confined Fusion Plasma (WDMApp) as part of the Department of Energy’s Exascale Computing Project (ECP). Bhattacharjee explained that WDMApp is a ten-year project that involves plasma physicists, applied mathematicians, and computer scientists to simulate whole device burning plasmas applicable to an advanced tokamak regime. Bhattacharjee explained that the most crucial step of the project was coupling two existing, well-established, extreme-scale gyrokinetic codes, the GENE continuum code, and the XGC particle-in-cell (PIC) code, to develop novel algorithms for both GENE-XGC and GEM-XGC coupling. The WDMApp codes (GENE, GEM, and XGC) were optimized, leveraging the ECP Co-Design and Software Technologies projects for portability and performance.
Dr. Patty Lee, the chief scientist of hardware technology development at Quantinuum, presented the current capabilities of quantum computing hardware and discussed the scientific and industrial applications that have been run on the hardware. Dr. Lee also provided insights into the software development toolkits available to support the quantum programmer community and the outlook for achieving quantum advantage in the near term. She highlighted the exponential improvement in the computational capability of state-of-the-art quantum computing hardware compared to classical computers.
Dr. Christiane Jablonowski, a professor in the Department of Climate and Space Sciences and Engineering at the University of Michigan, gave a talk on “Computational Frontiers in Weather and Climate Modeling”. She reviewed the state-of-the-art weather and climate modeling approaches at NOAA, the National Center for Atmospheric Research (NCAR), and the Department of Energy, and discussed the emerging computational frontiers. The talk focused on high-resolution weather and climate modeling trends, the ‘digital twin’ concept, and emerging computational paradigms.
The much-anticipated exascale computing era is here, with the arrival of the Frontier system at Oak Ridge National Laboratory in the US. The US Department of Energy’s Exascale Computing Project (ECP) is poised to take full advantage of Frontier’s capabilities in tackling problems of national and international interest. According to Doug Kothe, the Director of the US Department of Energy Exascale Computing Project, and the day’s final speaker, “When we collaborate, we get the most powerful tools and discoveries.” The ECP’s mission is to deliver on targeted exascale systems such as Frontier, which are capable of addressing high-priority strategic problems of national interest that are intractable with at least 50 times the computational power of the HPC systems available in 2016, yet at a very high efficiency. The ECP’s software technology effort is developing an expanded and vertically integrated software stack that includes advanced mathematical libraries, extreme-scale programming environments, development tools, visualization libraries, and the software infrastructure to support large-scale data management and data science for science and security applications.
In addition to the inspiring talks, symposium attendees also had the opportunity to engage in a lively panel discussion with the day’s speakers. Moderated by Venkat Raman, Professor of Aerospace Engineering and Mechanical Engineering at the University of Michigan, the panelists – Fariba Fahroo, Doug Kothe, Amitava Bhattacharjee, Patty Lee, Christiane Jablonowski, and Alex Aiken – tackled some of the most pressing issues and challenges in the field of high performance computing and computational discovery. The audience was able to ask questions and participate in the discussion, making it an engaging experience. It was a fitting end to an informative and thought-provoking day at the MICDE Symposium.
The 2023 MICDE Symposium featured a poster competition where 58 participants showcased their research. The competition winners were announced on Friday, March 24. Tommy Waltmann secured the first place for his work on “Fast and Efficient Particle Trajectory Analysis with the freud Library.” Doruk Aksoy won the second place for “An Incremental Tensor Train Decomposition for High-Dimensional Data Streams,” and Archana Sridhar and Parameshwaran Pasupathy shared the third place for their respective works on “Simulation and modeling of particle-laden compressible flows” and “A Fractional Viscoelastic Model of the Axon in Brain White Matter.” The fourth place was shared among Keith Phuthi, Srinivasan Arunachalam, Kyle Bushick, and Vishal Subramanian, for their works on various topics related to simulation and modeling.
“It has truly been an honor hosting these distinguished speakers, and to attend the poster session which highlighted the incredible depth and breadth of research in computational science at the University of Michigan. It is a testament to our pursuit of knowledge and innovation and a reminder of the direct impact computational science has on science and society,” Duraisamy said.