Summary-based inference of quantitative bounds of live heap objects

VICTOR A. BRABERMAN; DIEGO GARBERVETSKY; SAMUEL HYM; SERGIO YOVINE

SCIENCE OF COMPUTER PROGRAMMING

ELSEVIER SCIENCE BV

Lugar: Amsterdam; Año: 2014

0167-6423

This article presents a symbolic static analysis for computing parametric upper bounds of the number of simultaneously live objects of sequential Java-like programs. Inferring the peak amount of irreclaimable objects is the cornerstone for analyzing potential heap-memory consumption of stand-alone applications or libraries. The analysis builds method-level summaries quantifying the peak number of live objects and the number of escaping objects. Summaries are built by resorting to summaries of their callees. The usability, scalability and precision of the technique is validated by successfully predicting the object heap usage of a medium-size, real-life application which is significantly larger than other previously reported case-studies.