There is a large literature on errors in floating point calculations. You do not need complicated examples. Simple ones are given in papers by William Kahan, http://www.cs.berkeley.edu/~wkahan