This repository contains the tool chain for automatically inferring atomic sets that is described in the 2013 UIUC technical report Automated Inference of Atomic Sets for Safe Concurrent Execution by Peter Dinges, Minas Charalambides, and Gul Agha. It furthermore contains the evaluation data used in the report, as well as the custom workloads.
Atomic sets are a synchronization mechanism in which the programmer specifies the groups of data that must be accessed as a unit. The compiler can check this specification for consistency, detect deadlocks, and automatically add the primitives to prevent interleaved access. Atomic sets relieve the programmer from the burden of recognizing and pruning execution paths which lead to interleaved access, thereby reducing the potential for data races.
However, manually converting programs from lock-based synchronization
to atomic sets requires reasoning about the program's concurrency
structure, which can be a challenge even for small programs. Our
analysis eliminates the challenge by automating the reasoning. Our
implementation of the analysis allowed us to derive the atomic sets
for large code bases such as the Java collections framework in a
matter of minutes. The analysis is based on execution traces;
assuming all traces reflect intended behavior, our analysis allows
safe concurrency by preventing unobserved interleavings which may
harbor latent Heisenbugs.
The tool chain consists of a Java byte code instrumenter and an inference tool. The instrumenter uses WALA's Shrike library to insert calls to the field access tracing library into the input byte code. After instrumentation, the target program must be executed to generate field access traces. The traces are the input for the inference tool, which is a Python implementation of the algorithm described in the technical report.
- The
tools/directory contains the instrumenter (tools/tracing/), as well as the inference tool (tools/replay/). It furthermore contains a set of test cases used to ensure that everything works as expected. - The
applications/directory contains the fuzzing tools used to generate the execution traces for the Java collections framework and the JCurzez library variations. - The
data/directory contains the data used to evaluate the approach (data/aj_benchmarks/), as well as the expected outputs for the test cases.
The sub-directories of the data/aj_benchmarks/ directory contain the
data used to evaluate our tool. Each directory contains a shell
script run.sh that automates the process of instrumenting and
running the target application, and analyzing the results. The
commands in the shell scripts provide documentation on how to execute
each step.
The Atomic Set Inferencer tool chain may be used, copied, and modified
for education, research, and non-profit purposes as described in the
LICENSE file. Copyright (c) 2013 The University of Illinois Board
of Trustees. All Rights Reserved.
This work was supported in part by sponsorships from the Army Research Office under award number W911NF-09-1-0273, as well as the Air Force Research Laboratory and the Air Force Office of Scientific Research under agreement number FA8750-11-2-0084.