previous_group previous up next next_group
Previous: 4. Conclusion Up: Benchmarking polyhedral libraries Next: 1. Dual Conversion versus


6. Results for Dual Conversion

We have several implementations for the dual conversion operator, which are mostly based on the Chernikova algorithm. The conversion is dual, thus we focus only in converting a constraint system to its generating system. Since we do not have databases for the dual conversion operator, we have chosen to use the convex hull sampled databases, since this operator is strongly based on the dual conversion operator (see 6_subsec:polyhedral_databases for POLYBENCH's polyhedral databases).

We experienced problems[*] in adapting the LRS library, to the POLYBENCH framework, thus the results for LRS are not yet available. New POLKA is not chosen since its convex hull implementation is strongly related to the dual conversion operator. Therefore, in this section, we only present the comparison between two implementations: $C^3$ dual conversion that use POLYLIB implementation and CDD.



Subsections

Nguyen Que Duong
2006-09-16