exactExp

function exactExp(x: number): number

import { exactExp } from '@driftengine/physics';

Parameters

In depth

Reduce by a multiple of ln 2 into [-ln2/2, ln2/2], sum the Taylor series there, and scale by a power of two. A degree-14 series and not a shorter minimax, because on that interval the fifteenth term is about 4e-20 — below the precision of a double — so the series is exact to the representation and there is nothing a fitted polynomial would buy.