Peter F. Sweeney

I consider myself an experimental computer scientist. I build software prototypes that explore cutting-edge software engineering technology and analyze that technology's behavior. The technologies that I am most interested in are the analysis and optimization of object-oriented languages, and software distribution.

One interest of mine is determining the sweet spot between the precision of software analysis and the time required to perform the analysis.

• Adaptive optimization techniques, which apply optimizations at program execution time, require selective use of optimization to avoid excessive overhead. Run-time measurements provide the feedback that can be used to identify likely candidates for optimization. These measurements must be fast also to avoid excessive overhead, and are, therefore, excellent candidates to trade off measurement precision for speed in gathering the measurement data. I am involved in building a prototype adaptive optimization system, which is being developed within the Jalapeño project.
• In our work on detecting dead data members in C++ applications, Frank Tip and I developed a conservative, yet efficient, algorithm to detect dead data members. The algorithm's evaluation on a set of benchmarks demonstrated its effectiveness. The algorithm and the evaluation results were published in PLDI'98.

• I co-developed JAX, Java application extractor, a tool that reduces the size of Java code. JAX has incorporated a number of previously techiques such as eliminating unreachable methods, eliminating dead data members, and transforming class hierarchies to reduce the size of Java code.
• With Michael Burke, I have quantified the space overhead for object-oriented features. And with Yossi Gil, I have developed new techniques to reduce the space and time overhead for multiple inheritance.

• In general, I am interested in the problem of how to package software in one computing environment so that it may be successfully deployed in another computing environment over the internet. JAX is a specific approach for extracting Java code and I am interested in exploiting this technology for software distribution.

Since graduating from Columbia as an undergraduate, I have been working at the IBM T.J. Watson Research Center. It is a great place to work!