sinistersnare / aams

Abstract Machines Galore

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Abstract Machines ... Abstracted!

I wrote about what abstraction machines are on my blog here. This repo implements an abstract machine (think CESK machine). I then use concepts from the 'Abstracting Abstract Machines' paper To create a static-analysis of it. See the bibliography below for more info.

  • latex/: Formalizations of these machines in LaTeX computer-sciencese.
  • src/: Implementation of the formalizations in both Racket and Rust.
  • examples/: Example scheme programs that the machines should accept.

The formalizations of these machines are available in the latex/ directory.

I have both Racket and Rust implementations of the machine. See src/ for those.

In the examples/ directory are some basic scheme programs accepted by these machines.

Running

To run the Rust code

$ cargo run -- examples/basic.scm # to run a scheme file
$ cargo run # to start a REPL

For Racket, load the machine up in your Racket environment of choice, and use the evaluate function to run a program.

> (evaluate '(+ 3 4 (let ([x 6] [y 10]) (+ x y))))

I haphazardly implemented primitives, the racket version has more than the Rust.

Abstract Interpretation

Abstract Interpretation is a style of static-analysis that attempts to run the program in a decidably computable and sound way (No we did not have to solve the halting problem). By turning the infinite-state-machine of the interpreter into a finite-state-machine (by over-approximating values) we can safely analyze programs.

'Abstracting Abstract Machines' shows how we can use Abstract Machines to more easily create abstract interpreters. Thats what we are doing here, We first implement a concrete machine, and then we lift it into abstract-space to create an analysis.

Future Work / TODO

I also plan on implementing Delimited Continuations instead of full-shot, undelimited, escaping continuations. DCs are much more efficient and usable. Hopefully after I get the analysis finished, I can start on this!

I want to extend the machines semantics to cover an early pass of SinScheme. That way, I can do a static analysis of it, and perform real optimizations on it. This is a a stretch goal, but I think it's doable!

Bibliography

  • Abstracting Abstract Machines (AAM)
    • Paper that introduces the idea of making an Abstract Interpreter out of a direct-style abstract machine. Has a good intro section describing how to go from CEK to an abstract time-stamped CESK*.
    • D. V. Horn and M. Might, “Abstracting Abstract Machines”.
    • Available online here
  • Pushdown Control-Flow Analysis for Free (P4F)
    • Continues on AAM to describe a good analysis that you can do for cheap. The background section was very helpful in understanding how to better abstract an AM. Would recommend just the background section if you are still confused after AAM.
    • T. Gilray, S. Lyde, M. D. Adams, M. Might, and D. Van Horn, “Pushdown Control-Flow Analysis for Free,” Proceedings of the 43rd Annual ACM SIGPLAN-SIGACT Symposium on Principles of Programming Languages - POPL 2016, pp. 691–704, 2016, doi: 10.1145/2837614.2837631.
    • Available online here
  • Abstract Interpretation by Cousot & Cousot
    • The original paper on Abstract Interpretation. Pretty tough read, but helpful after a bunch of thorough readings!
    • P. Cousot and R. Cousot, “Abstract Interpretation: A Unified Lattice Model for Static Analysis of Programs by Construction or Approximation Of Fixpoints".
    • Available online here (search around for a PDF)
  • Matt Might's Dissertation
    • This paper doesnt do AAMs but talks about Abstract Interpretation, CFA, and a few other topics. Also a helpful paper!
    • M. Might, “Environment analysis of higher-order languages” 2007.
    • Available online here
  • Resolving and Exploiting the k-CFA Paradox (m-CFA paper)
    • This paper introduces the m-CFA analysis, and talks about how you can use flat closures to achieve linear-time CFA for higher-order functional-style (closure based) languages. The machines in this repo use flat-closures nowadays
    • Matthew Might, Yannis Smaragdakis, and David Van Horn, “Resolving and Exploiting the k-CFA Paradox,” 2010.
    • Available online here.

License

MIT license! Enjoy!

Let me know if you use/enjoy this project, I would be very happy to hear it.

About

Abstract Machines Galore

License:MIT License


Languages

Language:Racket 42.1%Language:Rust 40.7%Language:TeX 17.2%