A Snapshot-Based Evaluation Method for Garbage Collection

The evaluation of garbage collection (GC) algorithms typically needs a complicated experimental phase, involving a considerable amount of engineering effort. This paper presents a new evaluation method to simplify this experimental phase. By the hypothesis that most of the GC algorithms are based on analyzing a set of object graphs, we propose an indirect evaluation approach that only focuses on the analysis of those graphs (called snapshots). Since the context of evaluation is shifted from a complicated runtime to a simpler one, the researchers can then more easily concentrate on devising algorithms. We also implement a prototype as a proof of concept, with the focus on cyclic reference counting (a classical GC issue), which, technically, involves two procedures: taking snapshots and simulation. Our results demonstrate the feasibility and effectiveness of the proposed method.