A Data-Driven GPU Performance Model

Colloq: Speaker: 
Dr. Orion Sky Lawlor
Colloq: Speaker Institution: 
University of Alaska Fairbanks- Dept. of Computer Science
Colloq: Date and Time: 
Fri, 2009-10-16 10:00
Colloq: Location: 
Building 5100, Room 128
Colloq: Host: 
Jeffrey Vetter
Colloq: Host Email: 
Colloq: Abstract: 
Graphics Processing Units (GPUs) can deliver astounding floating point performance, but current implementations are rather tricky to tune. In this talk, we will present a series of performance experiments that demonstrate how to achieve high performance for general-purpose programs on the GPU.
Colloq: Speaker Bio: 
Dr. Lawlor is an Assistant Professor of Computer Science at the University of Alaska at Fairbanks, where he builds robots, writes software, and runs a 10-node cluster of NVIDIA GTX 280s. Dr. Lawlor was not convinced GPU programming was worth learning about until 2005.