The reason expm1 exists is because that naive approach is numerically poor when x is small (then e^x is almost 1, and subtracting two almost equal values throws

The reason expm1 exists is because that naive approach is numerically poor when x is small (then e^x is almost 1, and subtracting two almost equal values throws away significant digits and gives you an inaccurate result).