Evaluating CDCL Restart Schemes

17 pages•Published: March 15, 2019

Abstract

Modern CDCL (conflict-driven clause learning) SAT solvers are used for many practical applications. One of the key ingredients of state-of-the-art CDCL solvers are efficient restart schemes. The main contribution of this work is an extensive empirical evaluation of various restart strategies. We show that optimal static restart intervals are not only correlated with the satisfiability status of a certain instance, but also with the more specific problem class of the given benchmark. We further compare uniform restart intervals with the performance of non-uniform restart schemes, such as Luby restarts. Finally, we revisit the dynamic restart strategy used in Glucose and propose a new variant thereof, which is based on the concept of exponential moving averages. The resulting implementation in Lingeling improves state-of-the-art performance in SAT solving.

In: Daniel Le Berre and Matti Järvisalo (editors). Proceedings of Pragmatics of SAT 2015 and 2018, vol 59, pages 1-17.

Links:	https://easychair.org/publications/paper/RdBL
	https://doi.org/10.29007/89dw

BibTeX entry

@inproceedings{POS-18:Evaluating_CDCL_Restart_Schemes,
  author    = {Armin Biere and Andreas Fröhlich},
  title     = {Evaluating CDCL Restart Schemes},
  booktitle = {Proceedings of Pragmatics of SAT 2015 and 2018},
  editor    = {Daniel Le Berre and Matti Järvisalo},
  series    = {EPiC Series in Computing},
  volume    = {59},
  publisher = {EasyChair},
  bibsource = {EasyChair, https://easychair.org},
  issn      = {2398-7340},
  url       = {/publications/paper/RdBL},
  doi       = {10.29007/89dw},
  pages     = {1-17},
  year      = {2019}}

Download PDF Open PDF in browser