Federal Project Grant Award Summary

The National Science Foundation (NSF) Division of Information and Intelligent Systems awarded $507,378 to the University of California, San Diego under the Computer and Information Science and Engineering program (CFDA 47.070) for a five-year CAREER grant effective July 1, 2026 through June 30, 2031. This project will develop systems and frameworks to enable efficient execution of finely interleaved programs that dynamically offload computation between general-purpose processors and specialized hardware accelerators at microsecond timescales.

The research deliverables encompass three primary thrusts: (1) characterization and modeling of how existing applications interleave operations amenable to hardware acceleration with application-specific logic, identifying optimization opportunities and control/data flow patterns; (2) design and implementation of a runtime system that executes finely interleaved programs efficiently by enabling direct chaining of accelerated operations while bypassing the processor where feasible; and (3) development of a programming model and compiler framework that facilitate fine-grained hardware offloading with minimal application code rewriting. Collectively, these deliverables will enable a new computing paradigm addressing current challenges in leveraging diverse accelerators—such as those used in artificial intelligence applications—by making it practical for developers to write and execute programs that exploit heterogeneous computing architectures at scale.