MiniSAT
Generated by GPT-5-miniExpansion Funnel Raw 73 → Dedup 0 → NER 0 → Enqueued 0
| MiniSAT | |
|---|---|
| Name | MiniSAT |
| Author | Niklas Eén and Niklas Sörensson |
| Initial release | 2003 |
| Programming language | C++ |
| License | BSD-like |
| Genre | SAT solver |
MiniSAT MiniSAT is a compact, open-source Boolean satisfiability (SAT) solver influential in automated reasoning, formal verification, and constraint solving communities. Developed as a minimalist reference implementation, it bridged research prototypes and industrial tools, shaping competitions, toolchains, and academic curricula. Its design influenced numerous successors and integrations across verification stacks and synthesis infrastructures.
Overview
MiniSAT introduced a small, high-performance SAT core used widely in competitions such as SAT Competition and integrated into projects associated with Intel, IBM, Google, Microsoft Research, Facebook, and Amazon Web Services. The solver’s codebase emphasized clarity and extensibility, inspiring tools in model checking like CBMC, nuXmv, and SPIN-related ecosystems. MiniSAT’s influence extended to theorem proving collaborations with teams at University of Cambridge, Stanford University, ETH Zurich, University of California, Berkeley, and Princeton University.
Architecture and Algorithms
MiniSAT implements the DPLL algorithm with modern enhancements: conflict-driven clause learning (CDCL), watched literals, non-chronological backtracking, and restarts. Its core scheduling and heuristics draw from research by Jonas Söderberg, Marek Suda, and predecessors related to Davis–Putnam–Logemann–Loveland algorithm lineage applied in projects at MIT and Carnegie Mellon University. Variable selection uses VSIDS-inspired heuristics similar to approaches from Eén and Sörensson’s peers at Royal Institute of Technology (KTH) and methodologies compared in benchmarks from DIMACS Challenge datasets curated by DIMACS Center. Clause learning and deletion policies were influenced by studies concurrent with work at CNRS and Max Planck Institute for Informatics research groups.
Implementations and Variants
Several implementations and forks extended MiniSAT’s architecture: parallel portfolios like those in P- MiniSAT and integrations into portfolio solvers used in TACAS-related toolchains; incremental solver interfaces used in Z3 and CVC4 projects; and domain-specific extensions in tools from NASA and Siemens. Variants added pre- and in-processing, decision heuristics from SATzilla research teams, and proof logging compatible with systems such as DRAT and LRAT proof frameworks developed at University of Oxford and Université Paris-Saclay. Academic derivatives influenced industrial-grade solvers at Landesinstitut and research prototypes at University of Tokyo and Tsinghua University.
Performance and Benchmarks
MiniSAT’s performance was benchmarked extensively in SAT Competition tracks, on library suites from DIMACS Challenge, and in regression tests used by groups at SRI International and Bell Labs. Comparative studies evaluated MiniSAT against solvers like Glucose, Lingeling, MapleSAT, and CryptoMiniSat, with results reported in workshops at IJCAI, SAT Workshop, and CADE. Performance analysis often referenced hardware platforms from Intel and AMD servers, and used datasets from formal verification cases at ARM Holdings and NVIDIA.
Applications
MiniSAT was embedded in verification and synthesis flows: bounded model checking in CBMC used by contributors at Imperial College London and KTH Royal Institute of Technology; software verification in tools linked to LLVM projects and GCC analysis plug-ins; hardware verification in flows from Cadence, Synopsys, and Mentor Graphics; and constraint solving in automated planning systems associated with DARPA and European Space Agency. Research applications included cryptanalysis case studies at École Normale Supérieure and scheduling tasks in projects with Siemens and Thales Group.
Development History and Licensing
MiniSAT was authored by Niklas Eén and Niklas Sörensson and released in 2003 with a permissive BSD-style license adopted by academic and industrial users including contributors from KTH Royal Institute of Technology, Chalmers University of Technology, and corporate partners at Ericsson. The project’s minimalist philosophy encouraged forks and research derivatives at institutions like University of Illinois Urbana-Champaign, University of Toronto, and Seoul National University. Licensing choices eased inclusion in proprietary toolchains at IBM and open-source stacks at Google, while governance and citation practices were discussed in conference panels at FLoC-era meetings and ICLR adjacent workshops.