Project Grant 2216903

This five-year, $899,840 Project Grant from the National Science Foundation's Computer and Information Science and Engineering program will fund research at The Ohio State University to develop a comprehensive framework for efficient, scalable, and performance-portable tensor applications. The University will contribute expertise in algorithms, software, and hardware to address challenges in sustaining improved performance and energy efficiency as transistor resources become bounded, and in increasing developer productivity and performance portability across heterogeneous hardware architectures. Key deliverables include a unified compiler optimization methodology for automated optimization of dense tensor computations; a multi-level blocking methodology to enhance scalability of sparse-tensor computations; techniques for hardware-software co-design of accelerators for dense- and sparse-tensor workloads; and methods to ensure correctness and floating-point accuracy across compiler transformations and design-space exploration. The University will also apply the developed tools and methodology to advance machine learning and scientific computing applications involving PDE solvers, quantum simulation, tensor networks, and large-scale image analysis.