We have built a CP system that allows one to turn on caching.
It will store a representation of each subproblem encountered
during the search, and lookup to see if a new subproblem has
already been encountered and in that case fail. For problems
where subproblem equivalence exists this can speed up search
Our method is reported here.
The benchmarks are here in MiniZinc format
G. Chu, M. Garcia de la Banda, and P.J. Stuckey.
Automatically Exploiting Subproblem Equivalence in Constraint Programming.
- Tar File