Federal Grant Award Summary

The University of California, San Diego received a $215,035 Project Grant award from the National Science Foundation's Division of Information and Intelligent Systems under the Computer and Information Science and Engineering (CISE) program (CFDA 47.070) for collaborative research on compiler infrastructure redesign. The award, effective May 1, 2026 through April 30, 2029, supports the development of synthesis-based compiler construction methods that enable scientists to express and statically enforce complex physical and mathematical constraints in scientific computing. The research focuses on three primary technical objectives: synthesizing domain-specific and program-specific static analyses through stochastic search and satisfiability modulo theories (SMT)-based verification; creating semantic MLIR (Multi-Level Intermediate Representation) dialects for scientific computing domains including floating-point numbers and tensors; and developing project-specific correctness verification techniques.

This project addresses a critical gap in compiler infrastructure by enhancing LLVM (Low-Level Virtual Machine)—a widely-used compiler backbone powering languages including Julia, Python, and Fortran—to better serve the scientific computing community. The deliverables include integrated correct static analyses that will be incorporated into production compiler systems, thereby enabling scientists to automatically generate compiler components rather than manually engineer them. By lowering the burden of proof for mathematical correctness in scientific code compilation, this research advance is intended to enhance computational reliability and productivity across a broad community of scientists and engineers reliant on high-performance scientific computing infrastructure.