Testing latest pari + WASM + node.js... and it works?! Wow.
License: GPL3
ubuntu2004
Function: log1p
Section: transcendental
C-Name: glog1p
Prototype: Gp
Help: log1p(x): log(1+x)
Doc: return $\log(1+x)$, computed in a way that is also accurate
when the real part of $x$ is near $0$. This is the reciprocal function
of \kbd{expm1}$(x) = \exp(x)-1$.
\bprog
? default(realprecision, 10000); x = Pi*1e-100;
? (expm1(log1p(x)) - x) / x
%2 = -7.668242895059371866 E-10019
? (log1p(expm1(x)) - x) / x
%3 = -7.668242895059371866 E-10019
@eprog\noindent When $x$ is small, this function is both faster and more
accurate than $\log(1+x)$:
\bprog
? \p38
? x = 1e-20;
? localprec(100); c = log1p(x); \\ reference point
? a = log1p(x); abs((a - c)/c)
%6 = 0.E-38
? b = log(1+x); abs((b - c)/c) \\ slightly less accurate
%7 = 1.5930919111324522770 E-38
? for (i=1,10^5,log1p(x))
time = 81 ms.
? for (i=1,10^5,log(1+x))
time = 100 ms. \\ slower, too
@eprog