Resolution vs. cutting plane solution of inference problems: Some computational experience