Compacting test vector sets via strategic use of implications