DaCapo Project Paper Named Most Influential of OOPSLA 2006

Writer(s): Kristyn Childres

A paper that had provenance at Purdue more than a decade ago was recently named the Most Influential Paper of OOPSLA 2006 (the 21st annual ACM SIGPLAN Conference on Object-Oriented Programming Systems, Languages and Applications).

“The DaCapo benchmarks: Java benchmarking development and analysis” was one of the major outcomes of the NSF-sponsored DaCapo project, a collaboration between Purdue and a number of institutions, including UMass Amherst, UT Austin, the University of Colorado and the University of New Mexico. Associate Professor Antony Hosking was the principal investigator for Purdue, and Department of Computer Science alumni Tom VanDrunen (Ph.D. ’04), now an associate professor at Wheaton College, was also a co-author.

Since benchmarks drive computer science research and industry product development, which ones researchers use and how they evaluate them are key questions for the computer science community. This paper put forward a model for benchmarking selection and evaluation. It introduced the DaCapo benchmarks, a set of open source, client-side Java benchmarks, as well as new value, time-series, and statistical metrics for static and dynamic properties such as code complexity, code size, heap composition and pointer mutations.

According to the SIGPLAN committee, “The DaCapo benchmarks have become the de facto standard Java benchmarking suite, quickly replacing older suites, such as the SPEC JVM benchmarks. The DaCapo paper has set a standard that substantially raised the level of Java experiments in the overall literature of the past decade.”