Object equivalence: revisiting object equality profiling (an experience report)

Abstract

Modern object-oriented programming languages greatly alleviate the memory management for programmers. Despite the efficiency of garbage collection and Just-In-Time program analyzes, memory still remains prone to be wasted. A bloated memory may have severe consequences, including frequent execution lags due to a high pressure on the garbage collector and suboptimal object dependencies. We found that dynamically monitoring object production sites and the equivalence of the produced objects is key to identify wasted memory consumption caused by redundant objects. We implemented optimizations for reducing the memory consumption of six applications, achieving a reduction over 40% in half of the applications without having any prior knowledge of these applications. Our results partially replicate the results obtained by Marinov and O'Callahan and explore new ways to identify redundant objects.

Deep Object Comparison for Interface-based Regression Testing of Software Components

“…There are several papers focused on the optimization of memory management in languages, which utilize garbage collection (i.e., automatic disposal of objects, which are no longer in use by the program). The main idea behind these works is that there are a number of equivalent objects in the memory during the execution of a program, which can be replaced by a single instance while preserving the same behavior of the entire program (see [7], [10], [11], [12], etc. ).…”

“…Similar approach is described in [10], although it is intended for a different programming language (Pharo). Again, the relevant heap activity is recorded during a run of the application under test and the post-mortem analysis is performed after the run.…”

“…Again, the relevant heap activity is recorded during a run of the application under test and the post-mortem analysis is performed after the run. The main difference is that, in [10], the objects are divided into several types and only certain types of objects, which are susceptible to redundancy (e.g., instances of the String or Point classes), are considered as optimization opportunities. Directed vertex and edge labeled graph is used for the representation of the internal structure of objects for the comparison purposes.…”

“…Directed vertex and edge labeled graph is used for the representation of the internal structure of objects for the comparison purposes. Based on it, a hash value is calculated for each considered object to speed up its comparison to other objects [10].…”

“…In [11], the conditions, which an object must satisfy to be considered for caching, are discussed. The paper is focused on immutability of objects, which is in some form often required by caching and similar techniques (including techniques described in [7] and [10]). It is pointed out that the border between construction and mutation of an object is not always clear [11].…”