So it’s less efficient than just implementing a NAND gate directly ?