Kevin Skadron Research Group

University of Virginia (Main campus)

How can memory be augmented to perform computation directly to reduce data movement and improve throughput for data-intensive workloads? His group develops new processing-in-memory architectures and evaluates them (e.g., Fulcrum, Gearbox, Sieve, DRAM-CAM, DRAM-AP) to explore performance and design trade-offs. The researchers design new automata and regular-expression processing hardware and runtimes through the UVA Center for Automata Processing (CAP) to enable high-throughput pattern matching and monitoring. The team builds programming models, APIs, and compiler support to simplify mapping applications to heterogeneous and PIM systems. They also create simulation frameworks and benchmark suites, including the PIMeval simulation framework and PIMbench benchmark suite, for quantitative evaluation.