Exploring a wide range of communication subsystems from on-chip through data center scale.

High-performance parallel and distributed systems demand equally high-performance networks.  Our researchers look at the unique networking demands from GPU-enabled computational systems.  Working with our product and engineering teams, we propose innovative solutions to communication problems.Recent publications include Relaxations for High-Performance Message Passing on Massively Parallel SIMT Processors and Network Endpoint Congestion Control for Fine-Grained Communication.