a search for your question "all the floating point operations that can produce NaN" gave useful results for me.