HMC Colloquium Series presents

Adventures in Garbage Collection

Melissa O'Neill, Andrew Hunter, and Zvi Effron
Pryne Auditorium
4:15 PM on Oct. 9, 2008





Abstract:
O'Neill: "An introduction to garbage collecting."



Andrew Hunter: "A simple method to analyze the age behavior of a functional language".

Heap data exhibits patterns, and exploiting these patterns is key to producing high-quality garbage collectors. In particular, theorized models and observed patterns of age-dependent behavior have spurred development on a number of promising technologies, most notably generation collection, but also such ideas as immortal areas, oldest-first collection, pretenuring, and phase-aware collection. However, little work has been done to analyze the age behavior of real programs, and several promising ideas for garbage collectors suffer for the lack of confirming data for their justifying models. Also, available data is largely taken from a homogenous set of system, with little representation, for example, of functional languages. My work this summer involved filling this hole. I'll describe the system I created to analyze the age behavior of a functional language (Haskell) and show a few of the kind of results this can give us.



Zvi Effron: "An efficient method to accurately determine death times of objects"