Fast Escape Analysis and Stack Allocation for Object-Based Programs

Abstract

A fast and scalable interprocedural escape analysis algorithm is presented. The analysis computes a description of a subset of created objects whose lifetime is bounded by the lifetime of a runtime stack frame. The analysis results can be used for many purposes, including stack allocation of objects, thread synchronization elimination, deadstore removal, code motion, and iterator reduction. A method to use the analysis results for transforming a program to allocate some objects on the runtime stack is also presented. For non-trivial programs, typically 10%-20% of all allocated objects are placed on the runtime stack after the transformation.

A Phase-Adaptive Garbage Collector Using Dynamic Heap Partitioning and Opportunistic Collection

“…However, unlike escape analysis [8], [9] finding objects whose lifetime is bound to the lifetime of its creation frame, we find the stack frames to which the lifetimes of many long-lived objects are bound. The detailed explanation is given in Sect.…”

“…For this purpose, we consider the object escape-pattern but exploit it differently than escape analysis [8], [9]. Basically, an object escapes its creation stack frame when it becomes reachable from other stack frame.…”

“…Escape analysis [8], [9] finds the objects not escaping their creation stack frames. Those objects are allocated on the stack rather than the heap and automatically reclaimed at the end of methods.…”

A Practical Escape and Effect Analysis for Building Lightweight Method Summaries

“…Most analyses use infinite depth and field branching (k = b = ∞). Exceptions include the analysis of Gay and Steensgaard [11] where objects escape when they are stored in the heap, i.e. k = 1; the analysis of Bodga and Hoelzle [12], with k = 2; and the analysis of Blanchet [10] which identifies objects by their heap depth, hence b = 0.…”

“…Escape analyses have been traditionally used to identify objects that do not escape their method or thread scopes, thus enabling stack allocation or synchronization elimination optimizations [11,9,12,10]. Pointer analyses have been mainly concerned with building points-to sets, or resolving alias queries [21,16,17,18].…”