Department Chair and Professor of Computer Science
"Sequentiality is an illusion."
Areas Of Interest
Computer architecture, especially: multi-core and multi-threaded chip architectures, CPU/GPU convergence, and novel processor organizations; architectures for managing power, temperature, and reliability; applications of control theory to computer architecture; and architectural modeling and simulation methodology.
Kevin Skadron received his Ph.D. in Computer Science from Princeton University in 1999. He joined the University of Virginia as Assistant Professor in 1999, was promoted to Associate Professor in 2005, and to Professor in 2010 and Department Chair in 2012. He spent the 2007-08 academic year on sabbatical at NVIDIA Research. Skadron was the 2011 recipient of the ACM SIGARCH Maurice Wilkes award and is a Fellow of the IEEE and a Distinguished Scientist of the ACM. He received an NSF CAREER Award in 2002, a Seven Society Teaching Award in 2002, University of Virginia Teaching Fellowship in 2003, and a UVa Excellence in Science and Technology award. Among other professional activities, he is founding associate editor-in-chief of IEEE Computer Architecture Letters and became editor-in-chief in 2010. He also serves on the editorial board of IEEE Micro and on the technical advisory board of Gradient-DA. From 2007-11 he was secretary-treasurer of ACM's SIGARCH and has also served as technical program co-chair of PACT 2006, general co-chair for PACT 2002 and MICRO-37, and co-organizer of the Workshop on Temperature Aware Computer Systems. Skadron has also presented tutorials at a number of conferences, most recently at ASPLOS 2010, ASPLOS 2008, Semi-Therm 2009, and Cool Chips 2008. He has graduated six masters and eleven PhD advisees and co-advisees, and is author or co-author of over one hundred refereed papers and book chapters.
Our research currently focuses on how to design multicore architectures in the presence of severe physical constraints, especially thermal, power delivery, process variations, and wear-out. We are chiefly focusing on these issues in the context of asymmetric and heterogeneous designs, which provide the best balance between high single-thread performance and high throughput for parallel tasks. Support for asymmetry is also becoming essential as process variations create performance and power asymmetry even in organizations that were originally designed to be symmetric. We are also one of the first groups to explore the use of graphics processors (GPUs) for general-purpose computing and the first to develop an architectural simulation infrastructure-Qsilver-for performance, power, and thermal studies. In addition to exploring the implications of heterogeneous organizations combining CPUs, GPUs, and other processor types, we are exploring how the massive parallelism of the GPU and its novel SIMD and memory organization can most effectively be used. To address these questions, we are pursuing a variety of investigations, as well as developing the Rodinia benchmark suite. Rodinia provides both optimized GPU and multicore-CPU implementations of a diverse set of applications. Skadron's group also continues its work on thermal modeling, with the HotSpot and ArchFP tools. Skadron's work has been funded by the NSF, SRC, ARO, AMD, NVIDIA, IBM, and Intel.
For an up-to-date list of selected publications, see Skadron's home page: www.cs.virginia.edu/~skadron