King Abdullah University of Science and Technology (KAUST) has earned an unprecedented double finalist selection for both the Association for Computing Machinery (ACM) Gordon Bell Prize and ACM Gordon Bell Prize for Climate Modeling for two separate research projects exemplifying the power of interdisciplinary collaboration and global partners. 

Working with Massachusetts Institute of Technology (MIT), Saint Louis University and NVIDIA, KAUST researchers received a Gordon Bell Prize finalist selection for using NVIDIA graphic processing units (GPUs) to enhance genetic studies from the UK Biobank. This prestigious prize recognizes outstanding achievements in high-performance computing, particularly for science, engineering and large-scale data analytics applications. 

By using mixed-precision calculations, KAUST and collaborating researchers sped up a complex process called “Kernel Ridge Regression” (KRR), which is crucial for analyzing genetic interactions. This optimization resulted in performance up to 133 times faster than REGENIE — a leading software for solving similar problems. 

Meanwhile, KAUST researchers, in collaboration with the NSF National Center for Atmospheric Research, the University of Notre Dame, Saint Louis University and NVIDIA, received a Gordon Bell Prize for Climate Modeling finalist selection. They developed an exascale climate emulator to meet the rising computational and storage demands of high-resolution Earth System Models. This prize honors parallel computing solutions for addressing climate change. 

As demonstrated on multiple supercomputers, including KAUST's Shaheen III, the emulator boasts substantially higher spatial-temporal resolution compared to current global models. It utilizes advanced numerical techniques to optimize speed and efficiency, significantly enhancing climate modeling accuracy, supporting machine learning and providing critical data for predicting climate change and extreme weather events. 

“Our interdisciplinary team is paving the way for more accurate, accessible and sustainable climate projections, which, I believe, are fundamental for understanding and mitigating the impacts of climate change,” said Marc Genton, Al-Khawarizmi Distinguished Professor of Statistics at KAUST. He added that it is “very exciting” to be a selected finalist for this prestigious ACM prize. 

KAUST's four most recent Gordon Bell finalist selections, including in 2022, 2023 and the two in 2024, have involved enhancing efficiencies by leveraging data sparsity, solving problems to a specified accuracy with the minimum information required. By combining mixed precision and low-rank techniques, these methods reduce the energy and time needed to process dense linear algebra problems, advancing sustainable computing practices. 

Principal research scientist Hatem Ltaief and Professor David Keyes are co-authors on all four of KAUST's Gordon Bell finalist papers from the past three years. Their work is based on the Hierarchical Computations on Manycore Architectures (HiCMA) linear algebra framework developed in KAUST 

“Do linear algebra, see the world,” Keyes said while noting the wide variety of applications that HiCMA has served over the years — from seismic processing to wireless signal decoding, from adaptive optics for telescopes to aerodynamics, and from climate studies to genomics. 

This year’s two Gordon Bell awards will be presented during the International Conference for High Performance Computing, Networking, Storage and Analysis (SC24), which will be held in Atlanta, Georgia, Nov. 17-22, 2024.