Abstract

In previous papers we showed that from a suitable termination order (called ranking) one can automatically compute the worst-case time complexity of a CHR constraint simplification rule program from its program text. We combined the worst-case derivation length of a query predicted from its ranking with a worst-case estimate of the number and cost of rule application attempts and the cost of rule applications to obtain the desired meta-theorem. Here we generalize the approach presented in these papers and use it to analyse several non-trivial rule-based constraint solver programs. These results also hold for naive CHR implementations. We also present empirical evidence through test runs that the actual run-time of a state-of-the-art CHR implementation is much better due to optimizations like indexing.

Citations

247	The complexity of some polynomial network consistency algorithms for constraint satisfaction problems - Mackworth, Freuder - 1985
176	A generic arc-consistency algorithm and its specializations - Hentenryck, Deville, et al. - 1992
92	Programming with Constraints - Marriott, Stuckey - 1998
55	On the complexity analysis of static analyses - McAllester
28	A Prolog Constraint Handling Rules compiler and runtime system - Holzbaur, Frühwirth
1	for Sicstus Prolog, TR-98-01, "Osterreichisches Forschungsinstitut f"ur - Manual - 1998
1	Special Issue on Description Logics - Patel-Schneider, Rousset, et al. - 1999
1	Ludwig-Maximilians-Universit"at M"unchen - Saft, Abdennadher, et al. - 2002