Collective Program Analysis

By: Ganesha Upadhyaya and Hridesh Rajan

Abstract

Popularity of data-driven software engineering has led to an increasing demand on the infrastructures to support efficient execution of tasks that require deeper source code analysis. While task optimization and parallelization are the adopted solutions, other research directions are less explored. We present collective program analysis (CPA), a technique for scaling large scale source code analysis by leveraging analysis specific similarity. Analysis specific similarity is about, whether two or more programs can be considered similar for a given analysis. The key idea of collective program analysis is to cluster programs based on analysis specific similarity, such that running the analysis on one candidate in each cluster is sufficient to produce the result for others. For determining the analysis specific similarity and for clustering analysis-equivalent programs, we use a sparse representation and a canonical labeling scheme. A sparse representation contains only the parts that are relevant for the analysis and the canonical labeling helps with finding isomorphic sparse representations. In a nutshell, two or more programs with same sparse representation must behave similarly for the given analysis. Our evaluation shows that for a variety of source code analysis tasks when run on a large dataset of programs, our technique is able to achieve substantial reduction in the analysis times; on average 69% when compared to baseline and on average 36% when compared to a prior technique. We also show that there exists a large amount of analysis-equivalent programs in large datasets for variety of analysis.

ACM Reference

Upadhyaya, G. and Rajan, H. 2018. Collective Program Analysis. ICSE’18: The 40th International Conference on Software Engineering (May 2018).

BibTeX Reference

@inproceedings{collective2018,
  author = {Ganesha Upadhyaya and Hridesh Rajan},
  title = {Collective Program Analysis},
  booktitle = {ICSE'18: The 40th International Conference on Software Engineering},
  location = {Gothenberg, Sweden},
  month = {May 27-June 3, 2018},
  year = {2018},
  entrysubtype = {conference},
  abstract = {
    Popularity of data-driven software engineering has led to an increasing demand
    on the infrastructures to support efficient execution of tasks that require
    deeper source code analysis. While task optimization and parallelization are the
    adopted solutions, other research directions are less explored.
    We present collective program analysis (CPA), a technique for
    scaling large scale source code analysis by leveraging analysis specific similarity.
    Analysis specific similarity is about, whether two or more programs can be
    considered similar for a given analysis.
    The key idea of collective program analysis is to cluster programs based on
    analysis specific similarity, such that running the analysis on one candidate
    in each cluster is sufficient to produce the result for others.
    For determining the analysis specific similarity and for clustering
    analysis-equivalent programs, we use a sparse representation and a
    canonical labeling scheme.
    A sparse representation contains only the parts that are relevant for the
    analysis and the canonical labeling helps with finding isomorphic sparse representations.
    In a nutshell, two or more programs with same sparse representation must behave
    similarly for the given analysis.
    Our evaluation shows that for a variety of source code analysis tasks when run
    on a large dataset of programs, our technique is able to achieve substantial
    reduction in the analysis times; on average 69% when compared to baseline and
    on average 36% when compared to a prior technique.
    We also show that there exists a large amount of analysis-equivalent programs in
    large datasets for variety of analysis.
  }
}