Duane Merrill, Ph.D.

duane.merrill "at" gmail.com

Current Research

My current research concerns algorithms, design patterns, and programming idioms that we can use to construct faster and more efficient software for GPU architectures.

I'm specifically interested in the challenges inherent in managing dynamic parallelism on GPUs: i.e., constructing efficient solutions for problems that are generally regarded as being poor matches for the machine model because of irregular and data-dependent parallelism.