A modular static analysis decomposes a program's analysis into analyses of its parts, or components. An intercomponent analysis instructs an intra-component analysis to analyse each component independently of the others. Additional analyses are scheduled for newly discovered components, and for dependent components that need to account for newly discovered component information. Modular static analyses are scalable, can be tuned to a high precision, and support the analysis of programs that are highly dynamic, featuring e.g., higher-order functions or dynamically allocated processes.In this paper, we present the engineering aspects of MAF, a static analysis framework for implementing modular analyses for higher-order languages. For any such modular analysis, the framework provides a reusable inter-component analysis and it suffices to implement its intra-component analysis. The intracomponent analysis can be composed from several interdependent and reusable Scala traits. This design facilitates changing the analysed language, as well as the analysis precision with minimal effort. We illustrate the use of MAF through its instantiation for several different analyses of Scheme programs.

One way to speed up static program analysis is to make use of today's multi-core CPUs by parallelising the analysis. Existing work on parallel analysis usually targets traditional data-How analyses for static, first-order languages such as C. Less attention has been given so far to the parallelisation of more general analyses that can also target dynamic, higherorder languages such as JavaScript. These are significantly more challenging to parallelise, as dependencies between analysis results are only discovered during the analysis itself. State-ofthe-art parallel analyses for such languages are therefore usually limited, both in their applicability and performance gains. In this work, we propose the parallelisation of modular analyses. Modular analyses compute different parts of the analysis in isolation of one another, and therefore offer inherent opportunities for parallelisation that have not been explored so far. In addition, they can be used to develop a general class of analysers for dynamic, higher-order languages. We present a parallel variant of the worklist algorithm that is used to drive such modular analyses. To further speed up its convergence, we show how this algorithm can exploit the monotonicity of the analysis. Existing modular analyses can be parallelised without additional effort by instead employing this parallel worklist algorithm. We demonstrate this for MODF, an inter-procedural modular analysis, and for MODCONC, an inter-process modular analysis. For MODCONC, we reveal an additional opportunity to exploit even more parallelism in the analysis. Our parallel worklist algorithm is implemented and integrated into MAF, a framework for modular program analysis. Using a set of Scheme benchmarks for MODF, we usually observe speedups between 3× and 8× when using 4 workers, and speedups between 8× and 32× when using 16 workers. For MODCONC, we achieve a maximum speedup of 15×.

Static analyses are used to gain more confidence in changes made by developers. To be of most use, such analyses must deliver feedback fast. Therefore, incremental static analyses update previous results rather than entirely recompute them. This reduces the analysis time upon a program change, and makes the analysis well-suited for environments where the code base is frequently updated, such as in IDEs and CI pipelines.In this work, we present a general approach to render a modular static analysis for highly dynamic programs incremental, by exploiting dependencies between intermediate analysis results. Modular analyses divide a program in interdependent parts that are analysed in isolation. The dependencies between these parts stem, for example, from the use of shared variables within the program. Our incrementalisation approach leverages the modularity of the analysis together with the dependencies that it reifies to compute and bound the impact of changes. This way, only the affected parts of the result need to be reanalysed, and unnecessary recomputations are avoided.We apply our approach to both a function-modular and a thread-modular analysis and evaluate it by comparing an incremental update of an existing result to a full reanalysis. We find reductions of the analysis time from 6% to 99% on 14 out of 16 benchmark programs, and on most programs the impact on precision is limited. On 7 of the programs, reanalysis time is reduced by more than 75%, showing that our approach results in fast incremental updates.

Abstract garbage collection is the application of garbage collection to an abstract interpreter. Existing work has shown that abstract garbage collection can improve both the interpreter’s precision and performance. Current approaches rely on heuristics to decide when to apply abstract garbage collection. Garbage will build up and impact precision and performance when the collection is applied infrequently, while too frequent applications will bring about their own performance overhead. A balance between these tradeoffs is often difficult to strike. We propose a new approach to cope with the buildup of garbage in the results of an abstract interpreter. Our approach is able to eliminate all garbage, therefore obtaining the maximum precision and performance benefits of abstract garbage collection. At the same time, our approach does not require frequent heap traversals, and therefore adds little to the interpreters’s running time. The core of our approach uses reference counting to detect and eliminate garbage as soon as it arises. However, reference counting cannot deal with cycles, and we show that cycles are much more common in an abstract interpreter than in its concrete counterpart. To alleviate this problem, our approach detects cycles and employs reference counting at the level of strongly connected components. While this technique in general works for any system that uses reference counting, we argue that it works particularly well for an abstract interpreter. In fact, we show formally that for the continuation store, where most of the cycles occur, the cycle detection technique only requires O(1) amortized operations per continuation push. We present our approach formally, and provide a proof-of-concept implementation in the Scala-AM framework. We empirically show our approach achieves both the optimal precision and significantly better performance compared to existing approaches to abstract garbage collection.

This paper presents the implementation of an efficient interpreter for a Scheme-like language using manually written asm.js code. The asm.js specification defines an optimizable subset of JavaScript that has already served well as a compilation target for web applications where performance is critical. However, its usage as a human-writable language that can be integrated into existing projects to improve performance has remained largely unexplored. We therefore apply this strategy to optimize the implementation of an interpreter. We also discuss the feasibility of this approach, as writing asm.js by hand is generally not its recommended use-case. We therefore present a macro system to solve the challenges we encounter. The resulting interpreter is compared to the original C implementation and its compiled equivalent in asm.js. This way, we evaluate whether manual integration with asm.js provides the necessary performance to bring larger applications and runtimes to the web. We also refactor our implementation to assess how more JavaScript code can cohabit with asm.js code, improving maintainability of the implementation while preserving near-native performance. In the case of our interpreter, this improved maintainability enables adding more complex optimizations. We investigate the addition of function inlining , for which we validate the performance gain.