3

GitHub - sarsko/CreuSAT: CreuSAT - A formally verified SAT solver written in Rus...

 1 year ago
source link: https://github.com/sarsko/CreuSAT
Go to the source link to view the article. You can view the picture content, updated content and better typesetting reading experience. If the link is broken, please click the button below to view the snapshot at that time.

CreuSAT

What is this?

A SAT solver which is written in Rust. It is formally verified using Creusot

What does that mean?

It means that it solves the Boolean satisfiability problem (also known as SAT) and that if it states that the formula is satisfiable (SAT), then we know (read: it is proven) that the formula is SAT, and if it states that the formula is unsatisfiable (UNSAT), then we know (read: it is proven) that the formula is UNSAT. Also, the solver is statically proven to be free of runtime panics, which means that it cannot crash.

Ah, nice. What features does it have?

It currently has the following features:

  • Clause analysis with clause learning
  • Unit propagation
  • Two watched literals (2WL) with blocking literals and circular search
  • The variable move-to-front (VMTF) decision heuristic
  • Phase saving
  • Backtracking of the trail to asserting level
  • Exponential moving averages (EMA) based restarts
  • Clause deletion (without garbage collection)

How do I run it?

Firstly you'll need to get Rust

Then afterwards, the solver can be built with:

cargo build

and tested with:

cargo test

and run with

cargo run -- --file [PATH_TO_FILE]

where the provided file must be a correctly formatted DIMACS CNF file.

Remember do add the --release as in cargo test --release [TEST_TO_RUN], otherwise it will be built in debug mode, which is slow.

How do I prove the solver?

I would recommend following the instructions in the Creusot directory for instructions on how to get Why3 and Creusot up and running.

To prove it you'll need to do the following:

  1. Install Rust
  2. Install Creusot. Clone it, and then run cargo install --path creusot.
  3. Install Why3. I recommend following the guide in the Creusot repository.

CreuSAT is using Cargo Make to make building easier. It can be installed by running:

cargo install --force cargo-make

After installing Cargo Make, simply run:

cargo make prove-CreuSAT

And hopefully the Why3 IDE will appear. If not, then most likely something is not installed or pathed correctly, or I have given the wrong instructions (sorry).

If the Why3 IDE appears, then it should work to press 3 and wait a bit. If you are doing the proof from scratch, then you will need to wait a while.

The following cargo make commands are supported:

  • prove-CreuSAT/p : Generate the MLCFG for CreuSAT and run the Why3 IDE.
  • prove-Robinson : Generate the MLCFG for Friday and run the Why3 IDE.
  • prove-Friday : Generate the MLCFG for Friday and run the Why3 IDE.
  • clean : Cleans all generated CFG and Why3 session files.
    • clean-CreuSAT : Clean just the CreuSAT files.
    • clean-Robinson : Clean just the Robinson files.
    • clean-Friday : Clean just the Friday files.
  • StarExec : Generate a creusat.zip file ready to be uploaded to the StarExec clusters
  • StarExec-JigSAT : Generate a jigsat.zip file ready to be uploaded to the StarExec clusters

Creusot seems really cool! How can I learn it?

There are a bunch of tests in the Creusot directory which I recommend looking at.

You could also check out Friday and Robinson for a couple of verified solvers which are both easier to grok algorithmically and proof-wise.

Overview of the repository

Overview of the repository:
/CreuSAT - The source code for CreuSAT
/Friday - A fully verified and super naive SAT solver.
/JigSAT - An unverified solver based on CreuSAT. Used for experimenting.
/Robinson - A fully verified DPLL-based solver.
/mlcfgs - Output directory for generated mlcfg + Why3 session files.
/prelude - Copy of prelude from the Creusot directory. Included here to make cargo make happy.
/tests - Directory for tests. \


About Joyk


Aggregate valuable and interesting links.
Joyk means Joy of geeK