Project Grant 2217154

This five-year, $7.3 million project grant from the National Science Foundation's Computer and Information Science and Engineering program aims to develop a comprehensive framework for efficient, scalable, and performance-portable tensor applications. The University of Utah will receive funding to collaborate with other researchers on advancing tensor computations, which are fundamental to large-scale parallel software applications in scientific computing and machine learning. Key deliverables include improved performance and energy efficiency of hardware architectures through algorithm-architecture co-design; increased developer productivity and performance portability across target platforms; and advances in scalable machine learning and scientific computing applications that use tensors. The researchers will work on compiler optimizations for automated optimization of dense tensor computations, scalability methods for sparse tensors, algorithm-architecture co-design of accelerators, ensuring correctness and accuracy, and applying the methodology and tools to cutting-edge applications. The award seeks to address challenges in sustaining improved computing performance and efficiency within bounded hardware resources and increasing customization and heterogeneity.