You are asking too much from the floating point representation. The only fractional values that can be represented with complete accuracy are those that are (negative) powers of two or the sum of...