Skip to content
Supporting scripts/data to the paper: Factoring semi-primes with (quantum) SAT-solvers
Mathematica Python Haskell Other
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
instances
nmr
solve
stats
.gitignore
Makefile
README.md
eqlen.py
semiprimes.sage
semiprimes.txt
semiprimes_100.txt
semiprimes_eqlen.txt
semiprimes_selection.txt

README.md

Factoring semi-primes with SAT solvers

This repository contains scripts and data complementary to the paper Factoring semi-primes with (quantum) SAT-solvers.

Most of the scripts are pretty small and self-explanatory. In addition most directories contain Makefiles that should give a clear indication of how the scripts should be called. Generate semiprimes.txt first, then generate the SAT-instances and then let the solvers solve those instances. Make sure you include a (symbolic) link to the solvers in the corresponding directories.

The solve directory has several subdirectories. The numeric directory measures the runtime of factoring with number-theoretical methods. The sat directory measures runtime of factoring with SAT solvers. The sat_min directory takes a shortcut when you want to determine the minimum runtime: try many seeds and stop any solver that runs longer than the currently minimum runtime for the current instance. The sat_multi directory collapses multiple instances into one by generating a single multiplication circuit, then fanning out the resulting semiprime, checking the output against many semi- primes and combining the results with a massive or-gate. Thus if there is one easy number to factor, we may hope that the SAT solver is able to find it and focus on factoring that number. (Although from the results we conclude it is not able to find such a number.) The zoom directory inspects some interesting distributions.

The Karatsuba instances were generated with ToughSat. Reproducing the instances will overload their service, so I did not attach the code to do so.

Running the benchmarks can take a long time, so for reproducibility of the statistics the instances and solutions have been included in .7z files.

Lastly, the stats directory contains measurements for for computing the community structure and other metrics of the SAT instances.

You can’t perform that action at this time.