Chen Ding Research Group

University of Rochester

Can mathematical locality measures predict and minimize data movement across cache hierarchies to improve program performance and energy efficiency? The research develops formal locality theory and reuse-distance models to quantify data movement and working-set behavior across memory levels. The group designs compiler and runtime techniques, including cache hints and programmable caches, that translate locality metrics into code and cache-control optimizations. Researchers analyze data movement complexity and develop tools for benchmarking and modeling CPU/GPU cache behavior to guide optimizations. Work also addresses parallel-memory allocation and locality-aware runtime systems to reduce contention and improve scalable performance.