Computer Science Colloquia
April 18, 2012
Michael Boyer
Advisor: Kevin Skadron
Attending Faculty: Sudhanva Gurumurthi, Chair; Jack Davidson, Mircea Stan, and Westley Weimer
11:00 AM, Rice Hall, Rm. 242
Ph.D. Proposal Presentation
Dynamic Rate Matching in Heterogeneous Systems
ABSTRACT
GPUs have attracted enormous interest over the past decade due to substantial increases in both performance and programmability. Their ubiquity has helped usher in an era of heterogeneous computing, in which computational devices of different capabilities are combined into a single system. Due to the potentially wide performance gap between different devices, operations in which multiple devices work in concert require effective rate matching techniques to ensure efficient execution. And because the throughput of each device is application dependent and can fluctuate at runtime, the rate matching must occur dynamically. We propose to apply dynamic rate matching to solve two interesting research problems in the context of heterogeneous systems: load balancing and improving energy efficiency.