egg: e-graphs good

The egg project uses e-graphs to provide a new way to build program optimizers and synthesizers.

The core egg library provides high-performance, flexible e-graphs implemented in Rust. It is packaged on crates.io and documented on docs.rs, including a tutorial that provides an introduction to e-graphs and their use cases.

The newer egglog system provides an alternative approach to equality saturation based on Datalog. It features a language-based design, incremental execution, and composable analyses. Check out the paper and the web demo.

@article{2021-egg,
    author = {Willsey, Max and Nandi, Chandrakana and Wang, Yisu Remy and Flatt, Oliver and Tatlock, Zachary and Panchekha, Pavel},
    title = {egg: Fast and Extensible Equality Saturation},
    year = {2021},
    issue_date = {January 2021},
    publisher = {Association for Computing Machinery},
    address = {New York, NY, USA},
    volume = {5},
    number = {POPL},
    url = {https://doi.org/10.1145/3434304},
    doi = {10.1145/3434304},
    abstract = {An e-graph efficiently represents a congruence relation over many expressions. Although they were originally developed in the late 1970s for use in automated theorem provers, a more recent technique known as equality saturation repurposes e-graphs to implement state-of-the-art, rewrite-driven compiler optimizations and program synthesizers. However, e-graphs remain unspecialized for this newer use case. Equality saturation workloads exhibit distinct characteristics and often require ad-hoc e-graph extensions to incorporate transformations beyond purely syntactic rewrites.  This work contributes two techniques that make e-graphs fast and extensible, specializing them to equality saturation. A new amortized invariant restoration technique called rebuilding takes advantage of equality saturation's distinct workload, providing asymptotic speedups over current techniques in practice. A general mechanism called e-class analyses integrates domain-specific analyses into the e-graph, reducing the need for ad hoc manipulation.  We implemented these techniques in a new open-source library called egg. Our case studies on three previously published applications of equality saturation highlight how egg's performance and flexibility enable state-of-the-art results across diverse domains.},
    journal = {Proc. ACM Program. Lang.},
    month = jan,
    articleno = {23},
    numpages = {29},
    keywords = {equality saturation, e-graphs}
  }

EGRAPHS Community

The egg project is part of the larger EGRAPHS Community, which aims to bring together researchers and practitioners interested in e-graphs and related techniques. Check out the EGRAPHS Community resources, including:

• The annual EGRAPHS Workshop at PLDI,
• The monthly EGRAPHS Seminar,
• The EGRAPHS Zulip for discussion and announcements about egg, egglog, and other projects!

News

For updates on egg itself, see the changelog.

• Our work connecting egg to Datalog will appear at PLDI 2023.
• Our paper Small Proofs from Congruence Closure is accepted to FMCAD 2022! See the recorded talk here.
• The inaugural EGRAPHS workshop at PLDI 2022 was a huge success!
• Ruler won a Distinguished Paper Award at OOPSLA 2021! Check out the talk here.
• Our paper on using equality saturation to automatically infer rewrite rules is accepted at OOPSLA 2021!
• Max wrote a post on the SIGPLAN blog about egg!
• Check out egg's POPL 2021 videos on YouTube (5 min and 30 min) or Clowdr.
• The egg paper won Distinguished Paper at POPL 2021.
• Max gave a 10 minute talk about egg as part of the UW CSE Colloquium series.
• The egg paper will appear at POPL 2021! Check out the preprint.

Projects using egg

Are you using egg in your project? Open a pull request to add your project to this list!

See who's using egg on GitHub and Google Scholar.

• Ruler automatically infers rewrite rules using equality saturation. OOPSLA 2021
• Diospyros automatically vectorizes digital signal processing code. ASPLOS 2021
• Tensat optimizes deep learning compute graphs both better and faster (up to 50x) than the state of the art. MLSys 2021
• Herbie improves the accuracy of floating point expressions. The egg-herbie library made parts of Herbie over 3000x faster! PLDI 2015
• Szalinski shrinks 3D CAD programs to make them more editable. PLDI 2020
• SPORES optimized linear algebra expressions up to 5x better than state-of-the-art. VLDB 2020
• Glenside explores the design space of hardware accelerators for a given deep learning program. MAPS 2021
• RisingLight is an educational OLAP database system.
• The folks at Intel have built a tool for Automating Constraint-Aware Datapath Optimization using egg. DAC 2023

Projects inspired by egg

• Metatheory.jl uses e-graphs and equality saturation for algebraic reasoning in Julia.
• Zachary DeVito made an awesome Colab exploring e-graphs.
• Philip Zucker has two blog posts about e-graphs and e-matching in Julia.

View or edit this site on GitHub.