Brief Bio

I am currently a PhD student and a Cunningham Fellow at Virginia Tech. I graduated Magna Cum Laude from the University of South Carolina in December 2008. I have been a member of the SCAPE Lab since Fall '04, first at USC, now at Virginia Tech. In my spare time, I rock climb religiously, mountain bike, backpack, and generally enjoy the outdoors. I also play the guitar, banjo, and harmonica, often with my bluegrass band the Soggy Biscuit Boys.

Research Interests

My primary research interest lies in high performance computing, especially large-scale scientific systems and codes. Here are some specific research interests that I have.

Performance

• The creation (both automagically and by hand) of highly accurate analytical models of parallel codes. Specifically, I'm interested in the problem of automatically creating analytical models that are machine-parameterized. This interest has been fueled in part by the work of the PAL group at Los Alamos and the work of the PERC group, as well as the studies I have performed on large-scale codes, such as POP and GYRO. I am currently involved in a project in this area.
• The use of hardware performance counters to analyze and predict performance patterns at a fundamental level. I am currently involved in projects in this area.
• Benchmarking and performance measurement.

Systems

• The creation of transparent methods of self-adaptation for systems to optimize over power, performance, thermal constraints, etc. By leveraging emergent monitoring facilities (e.g. hardware performance counters) and applying control-theoretic techniques, it may be possible to create these adaptive control systems with both low overhead and high accuracy. I find the work of Martonosi et al particularly engaging. I am currently involved in a project in this area.
• Heterogeneous parallel and distributed systems. Scheduling, transparent access, etc.

Misc.

• Real-time graph drawing. I'd like to create a system where even large-scale graphs can be drawn in real-time, adapting to changes from the user. I have a code I whipped together in a couple of nights that can accomodate graphs of hundreds (but not thousands) of nodes. It uses OpenGL and glut, and is available here with a couple of sample graphs.
• Genetic algorithms, especially their application to non-traditional problems, and the theoretical basis for their correctness.
• Computational combinatorics, graph-theoretic stuff especially. Fixed parameter tractability, approximation algorithms, parallelization, etc.