\frac{e^{x}}{e^{x} - 1}\begin{array}{l}
\mathbf{if}\;e^{x} \le 0.991275043797484545:\\
\;\;\;\;\frac{e^{x}}{\frac{\mathsf{fma}\left(-1, 1, e^{x + x}\right)}{e^{x} + 1}}\\
\mathbf{else}:\\
\;\;\;\;\mathsf{fma}\left(\frac{1}{12}, x, \frac{1}{x}\right) + \frac{1}{2}\\
\end{array}double code(double x) {
return (exp(x) / (exp(x) - 1.0));
}
double code(double x) {
double temp;
if ((exp(x) <= 0.9912750437974845)) {
temp = (exp(x) / (fma(-1.0, 1.0, exp((x + x))) / (exp(x) + 1.0)));
} else {
temp = (fma(0.08333333333333333, x, (1.0 / x)) + 0.5);
}
return temp;
}




Bits error versus x
Results
| Original | 41.0 |
|---|---|
| Target | 40.6 |
| Herbie | 0.6 |
if (exp x) < 0.9912750437974845Initial program 0.0
rmApplied flip--0.0
Simplified0.0
if 0.9912750437974845 < (exp x) Initial program 61.8
Taylor expanded around 0 1.0
Simplified1.0
Final simplification0.6
herbie shell --seed 2020056 +o rules:numerics
(FPCore (x)
:name "expq2 (section 3.11)"
:precision binary64
:herbie-target
(/ 1 (- 1 (exp (- x))))
(/ (exp x) (- (exp x) 1)))