Brief Research Descriptions

Using Genetic Programming to Automatically Tune Compiler Optimizations: Continuing work from my internship with Microsoft, I am currently working on using genetic programming to automatically optimize compiler optimizations. Rather than tuning parameters to current optmizations, I am actually creating new, more effective heuristics to be used in the optmizations to improve program performance and decrease code size.

Dynamic Code Alignment: I previously researched ways to dynamically realign code in the instruction cache to take advantage of microarchitectural features, thereby improving performance.

Atom Performance Issues: I previously worked on characterizing performance on the Intel Atom processor.

Workshops

Michelle McDaniel and Kim Hazelwood. Runtime adaptation: a case for reactive code alignment. International Workshop on Adaptive Self-Tuning Computing Systems for the Exaflop Era (EXADAPT), March 2012. [PDF] [SLIDES]

Technical Reports and Posters

Michelle McDaniel and Kim Hazelwood. Performance characterization of mobile-class nodes: why fewer bits is better. International Symposium on Performance Analysis of Systems and Software (ISPASS), pp. 131-132, April 2011. (poster) [PDF] [POSTER]

Ben Kreuter, Ryan Layer, Michelle McDaniel, Gabriel Robins, and Kevin Skadron. Accelerating genomic analyses with parallel sliding windows. University of Virginia Technical Report CS-2010-14. November 2010. [PDF]

Non-Peer Reviewed

Michelle McDaniel. Assessing the Opportunities for Reactive Code Alignment. Master's Project, March 2012. [PDF] [SLIDES]