GPU Computing & Irregular Parallelism

Colloq: Speaker: 
John Owens
Colloq: Speaker Institution: 
UC Davis
Colloq: Date and Time: 
Mon, 2010-07-12 10:00
Colloq: Location: 
Bldg. 5600, Room J302
Colloq: Host: 
Jeffrey Vetter
Colloq: Host Email: 
vetter@ornl.gov
Colloq: Abstract: 
The computational power of GPUs, coupled with increasing programmability, is making the GPU an increasingly compelling platform for high-performance computing. In this talk I'll give a brief overview of GPU computing and talk about the current research challenges that our group is tackling, with particular attention to supporting irregular parallelism on the GPU. While GPUs are particularly good at regular, structured codes, it's still a large challenge to efficiently support more irregular codes. I'll talk about our recent work in task queuing, hash tables, and fragment compositing, and conclude my talk with a discussion of the research problems I'd like to address going forward.
Colloq: Speaker Bio: 
John Owens is an associate professor of electrical and computer engineering at UC Davis, where he joined the faculty in 2003. His group pursues research problems in GPU computing in both GPU fundamentals (data structures, algorithms, and multi-GPU computing) and applications (including computer vision, GPU-based embedded systems, medical imaging, speech recognition, protein folding, computational fluid dynamics, and visualization). John is a PI of the SciDAC Institute for Ultrascale Visualization and received the DOE Early Career Principal Investigator Award in 2004. He graduated from Stanford with a Ph.D. in electrical engineering in 2002 and from Berkeley with a B.S. in EECS in 1995.