RUS  ENG
Full version
JOURNALS // Prikladnaya Diskretnaya Matematika. Supplement // Archive

Prikl. Diskr. Mat. Suppl., 2013 Issue 6, Pages 111–112 (Mi pdma75)

Computational methods in discrete mathematics

GPU-based implementation of DPLL algorithm with limited non-chronological backtracking

V. G. Bulavintsev, A. A. Semenov

Institute of System Dynamics and Control Theory, Siberian Branch of the Russian Academy of Sciences, Irkutsk

Abstract: A new GPU-based SAT solver named ngsat is presented. The solver employs DPLL algorithm with limited version of non-chronological backtracking without Clause Learning. Some new techniques are developed and applied to increase the effectiveness of DPLL algorithm on SIMD. Ngsat's performance is demonstrated in application to the problems of search for pairs of orthogonal Latin squares.

Keywords: GPU, DPLL algorithm, SAT, parallel computer architectures, CUDA, SIMD.

UDC: 519.7



© Steklov Math. Inst. of RAS, 2025