Interval operations involving NaNs (Q1916991)

Besides a basic set of floating point numbers, also \(\pm\infty\), two distinguished values \(+0\) and \(-0\) and NaNs (Not a Number) are supported by the IEEE standard for floating point arithmetic. Further, certain exceptional situations that can occur during the numerical computations are suggested to deliver a default result, specified for each possible exception (IEEE non-trapping mode). Now, the paper investigates an interval arithmetic involving NaNs and the values \(\pm 0\) and proposes a class of interval arithmetic exception and their treatment in the non-trapping mode.