In C language when we do an operation involving operands of different sizes, the smaller one is changed to the larger. So, here even though b is of type float, since addition is done with a double, it also gets converted to a double. So, minimum size required of the register is $8$ bytes.