This question asks us to consider the stability of the given expression. Is there a specific rounding we should be using (e.g. fixed-point decimal dd.ddd or floating point .ddd ×10^d)?
Consider only floating-point and it does not matter how many digits.
You can answer in terms of ε_mach.
Just to add that we will always consider floating-point systems, unless otherwise stated.