The satisfiability (SAT) problem is to find an assignment of binary values to the variables which satisfy a given clausal normal form (CNF). Many practical application problems can be transformed to SAT problems, and many SAT solvers have been developed. SAT problem is, however, NP-complete and its computational cost is very high. In order to reduce the computational cost, preprocessors are widely used by SAT solvers. In this paper, we describe an approach for implementing a preprocessor (SatELite) on FPGA. In SatELite, the variables and clauses whose values can be uniquely determined from other variables and clauses are eliminated to reduce the search space of the given SAT problem. The algorithms used in SatELite have inherent parallelism, but the data size of the SAT problems is very large, and the performance of the system is limited by the throughput of the off-chip DRAM banks. In our implementation, several clauses are held on the FPGA, and are compared in parallel with a sequence of new clauses. The sequence is cached on the FPGA, and reused in order to hide the access delay to the DRAM banks. The speedup by our system depends on the problem size, however it becomes higher for larger problems.
[1]
Tsutomu Maruyama,et al.
An FPGA Solver for SAT-Encoded Formal Verification Problems
,
2011,
2011 21st International Conference on Field Programmable Logic and Applications.
[2]
Lakhdar Sais,et al.
Vivifying Propositional Clausal Formulae
,
2008,
ECAI.
[3]
Armin Biere,et al.
Effective Preprocessing in SAT Through Variable and Clause Elimination
,
2005,
SAT.
[4]
Tsutomu Maruyama,et al.
An Approach for Solving Large SAT Problems on FPGA
,
2010,
TRETS.
[5]
Srinivas Patil,et al.
FPGA-based hardware acceleration for Boolean satisfiability
,
2009,
TODE.