Eliminating abstraction overhead of Java stream pipelines using ahead-of-time program optimization

Abstract

Java 8 introduced streams that allow developers to work with collections of data using functional-style operations. Streams are often used in pipelines of operations for processing the data elements, which leads to concise and elegant program code. However, the declarative data processing style comes at a cost. Compared to processing the data with traditional imperative language mechanisms, constructing stream pipelines requires extra heap objects and virtual method calls, which often results in significant run-time overheads. In this work we investigate how to mitigate these overheads to enable processing data in the declarative style without sacrificing performance. We argue that ahead-of-time bytecode-to-bytecode transformation is a suitable approach to optimization of stream pipelines, and we present a static analysis that is designed to guide such transformations. Experimental results show a significant performance gain, and that the technique works for realistic stream pipelines. For 10 of 11 micro-benchmarks, the optimizer is able to produce bytecode that is as effective as hand-written imperative-style code. Additionally, 77% of 6879 stream pipelines found in real-world Java programs are optimized successfully.

Floo

“…Floo is inspired by a larger body of prior work that optimize Java programs, but instead targets the Android framework (not general Java), and integrates new techniques to tackle the computation patterns and resource restrictions in mobile apps and phones. Beyond memoization [5,25], other efforts automatically parallelize Java programs to take advantage of multi-core platforms [9,15,18,51,53,67]. Though conceptually complementary to Floo, we note that our results highlight substantial state sharing across invocations in app interaction handling, thereby limiting the potential for safe parallelism ( §3).…”

Large‐scale characterization of Java streams

“…Moreover, we find problematic stream processing that may have an impact on application performance (see Section 5.3). Academic material (e.g., books and tutorials) could more extensively discuss about stream-related performance issues (e.g., abstraction 25,80,81 and parallelization overheads 26,53 due to the use of streams), such that developers better leverage the Stream API and are particularly well-informed while deciding on whether involving streams in the implementation of performance-critical functionalities.…”

“…Other authors have focused on removing streams at a bytecode level, as stream‐related overheads can sometimes be avoided by using semantically‐equivalent imperative code. Møller et al 25 present StreamLiner, a proof‐of‐concept bytecode‐to‐bytecode tool to transform sequential streams into more efficient imperative code. Due to limitations inherent to the static analysis techniques used, StreamLiner is not able to analyze streams whose creation and execution take place in different methods.…”

“…They find more than 500 Java and Scala projects containing unit tests that can be considered as good candidate workloads for benchmarking purposes. Villazón et al 20 conduct a large-scale study to analyze units tests available in open-source projects written in JavaScript (109,286), Java (25,918), and Scala (4076). They study the adoption of new constructs and find common bad coding practices in JavaScript.…”

“…We use a fully automatic approach that avoids manual intervention and which eases a safe, containerized, and distributed deployment of the analysis. In comparison to related work addressing the analysis and optimization of streams, [21][22][23][24][25][26][27] we introduce a profiler able to detect all forms of stream creation and execution available in the Stream API. Moreover, our study is the first to identify and profile multiple runtime metrics that are both specific to streams and suitable to be massively collected in the wild.…”