bug500, discussion (missed optimization)

Percentage Accurate: 53.0% → 98.1%
Time: 22.0s
Alternatives: 9
Speedup: 2.0×

Specification

?
\[\begin{array}{l} \\ \log \left(\frac{\sinh x}{x}\right) \end{array} \]
(FPCore (x) :precision binary64 (log (/ (sinh x) x)))
double code(double x) {
	return log((sinh(x) / x));
}
real(8) function code(x)
    real(8), intent (in) :: x
    code = log((sinh(x) / x))
end function
public static double code(double x) {
	return Math.log((Math.sinh(x) / x));
}
def code(x):
	return math.log((math.sinh(x) / x))
function code(x)
	return log(Float64(sinh(x) / x))
end
function tmp = code(x)
	tmp = log((sinh(x) / x));
end
code[x_] := N[Log[N[(N[Sinh[x], $MachinePrecision] / x), $MachinePrecision]], $MachinePrecision]
\begin{array}{l}

\\
\log \left(\frac{\sinh x}{x}\right)
\end{array}

Sampling outcomes in binary64 precision:

Local Percentage Accuracy vs ?

The average percentage accuracy by input value. Horizontal axis shows value of an input variable; the variable is choosen in the title. Vertical axis is accuracy; higher is better. Red represent the original program, while blue represents Herbie's suggestion. These can be toggled with buttons below the plot. The line is an average while dots represent individual samples.

Accuracy vs Speed?

Herbie found 9 alternatives:

AlternativeAccuracySpeedup
The accuracy (vertical axis) and speed (horizontal axis) of each alternatives. Up and to the right is better. The red square shows the initial program, and each blue circle shows an alternative.The line shows the best available speed-accuracy tradeoffs.

Initial Program: 53.0% accurate, 1.0× speedup?

\[\begin{array}{l} \\ \log \left(\frac{\sinh x}{x}\right) \end{array} \]
(FPCore (x) :precision binary64 (log (/ (sinh x) x)))
double code(double x) {
	return log((sinh(x) / x));
}
real(8) function code(x)
    real(8), intent (in) :: x
    code = log((sinh(x) / x))
end function
public static double code(double x) {
	return Math.log((Math.sinh(x) / x));
}
def code(x):
	return math.log((math.sinh(x) / x))
function code(x)
	return log(Float64(sinh(x) / x))
end
function tmp = code(x)
	tmp = log((sinh(x) / x));
end
code[x_] := N[Log[N[(N[Sinh[x], $MachinePrecision] / x), $MachinePrecision]], $MachinePrecision]
\begin{array}{l}

\\
\log \left(\frac{\sinh x}{x}\right)
\end{array}

Alternative 1: 98.1% accurate, 0.5× speedup?

\[\begin{array}{l} x_m = \left|x\right| \\ \begin{array}{l} \mathbf{if}\;x_m \leq 0.029:\\ \;\;\;\;-0.005555555555555556 \cdot {x_m}^{4} + \left(0.0003527336860670194 \cdot {x_m}^{6} + x_m \cdot \left(x_m \cdot 0.16666666666666666\right)\right)\\ \mathbf{elif}\;x_m \leq 700:\\ \;\;\;\;-\sqrt[3]{{\log \left(\frac{x_m}{\sinh x_m}\right)}^{3}}\\ \mathbf{else}:\\ \;\;\;\;\log 0.0001984126984126984 + \left(-6 \cdot \log \left(\frac{1}{x_m}\right) + 42 \cdot \frac{1}{{x_m}^{2}}\right)\\ \end{array} \end{array} \]
x_m = (fabs.f64 x)
(FPCore (x_m)
 :precision binary64
 (if (<= x_m 0.029)
   (+
    (* -0.005555555555555556 (pow x_m 4.0))
    (+
     (* 0.0003527336860670194 (pow x_m 6.0))
     (* x_m (* x_m 0.16666666666666666))))
   (if (<= x_m 700.0)
     (- (cbrt (pow (log (/ x_m (sinh x_m))) 3.0)))
     (+
      (log 0.0001984126984126984)
      (+ (* -6.0 (log (/ 1.0 x_m))) (* 42.0 (/ 1.0 (pow x_m 2.0))))))))
x_m = fabs(x);
double code(double x_m) {
	double tmp;
	if (x_m <= 0.029) {
		tmp = (-0.005555555555555556 * pow(x_m, 4.0)) + ((0.0003527336860670194 * pow(x_m, 6.0)) + (x_m * (x_m * 0.16666666666666666)));
	} else if (x_m <= 700.0) {
		tmp = -cbrt(pow(log((x_m / sinh(x_m))), 3.0));
	} else {
		tmp = log(0.0001984126984126984) + ((-6.0 * log((1.0 / x_m))) + (42.0 * (1.0 / pow(x_m, 2.0))));
	}
	return tmp;
}
x_m = Math.abs(x);
public static double code(double x_m) {
	double tmp;
	if (x_m <= 0.029) {
		tmp = (-0.005555555555555556 * Math.pow(x_m, 4.0)) + ((0.0003527336860670194 * Math.pow(x_m, 6.0)) + (x_m * (x_m * 0.16666666666666666)));
	} else if (x_m <= 700.0) {
		tmp = -Math.cbrt(Math.pow(Math.log((x_m / Math.sinh(x_m))), 3.0));
	} else {
		tmp = Math.log(0.0001984126984126984) + ((-6.0 * Math.log((1.0 / x_m))) + (42.0 * (1.0 / Math.pow(x_m, 2.0))));
	}
	return tmp;
}
x_m = abs(x)
function code(x_m)
	tmp = 0.0
	if (x_m <= 0.029)
		tmp = Float64(Float64(-0.005555555555555556 * (x_m ^ 4.0)) + Float64(Float64(0.0003527336860670194 * (x_m ^ 6.0)) + Float64(x_m * Float64(x_m * 0.16666666666666666))));
	elseif (x_m <= 700.0)
		tmp = Float64(-cbrt((log(Float64(x_m / sinh(x_m))) ^ 3.0)));
	else
		tmp = Float64(log(0.0001984126984126984) + Float64(Float64(-6.0 * log(Float64(1.0 / x_m))) + Float64(42.0 * Float64(1.0 / (x_m ^ 2.0)))));
	end
	return tmp
end
x_m = N[Abs[x], $MachinePrecision]
code[x$95$m_] := If[LessEqual[x$95$m, 0.029], N[(N[(-0.005555555555555556 * N[Power[x$95$m, 4.0], $MachinePrecision]), $MachinePrecision] + N[(N[(0.0003527336860670194 * N[Power[x$95$m, 6.0], $MachinePrecision]), $MachinePrecision] + N[(x$95$m * N[(x$95$m * 0.16666666666666666), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]), $MachinePrecision], If[LessEqual[x$95$m, 700.0], (-N[Power[N[Power[N[Log[N[(x$95$m / N[Sinh[x$95$m], $MachinePrecision]), $MachinePrecision]], $MachinePrecision], 3.0], $MachinePrecision], 1/3], $MachinePrecision]), N[(N[Log[0.0001984126984126984], $MachinePrecision] + N[(N[(-6.0 * N[Log[N[(1.0 / x$95$m), $MachinePrecision]], $MachinePrecision]), $MachinePrecision] + N[(42.0 * N[(1.0 / N[Power[x$95$m, 2.0], $MachinePrecision]), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]]]
\begin{array}{l}
x_m = \left|x\right|

\\
\begin{array}{l}
\mathbf{if}\;x_m \leq 0.029:\\
\;\;\;\;-0.005555555555555556 \cdot {x_m}^{4} + \left(0.0003527336860670194 \cdot {x_m}^{6} + x_m \cdot \left(x_m \cdot 0.16666666666666666\right)\right)\\

\mathbf{elif}\;x_m \leq 700:\\
\;\;\;\;-\sqrt[3]{{\log \left(\frac{x_m}{\sinh x_m}\right)}^{3}}\\

\mathbf{else}:\\
\;\;\;\;\log 0.0001984126984126984 + \left(-6 \cdot \log \left(\frac{1}{x_m}\right) + 42 \cdot \frac{1}{{x_m}^{2}}\right)\\


\end{array}
\end{array}
Derivation
  1. Split input into 3 regimes
  2. if x < 0.0290000000000000015

    1. Initial program 56.5%

      \[\log \left(\frac{\sinh x}{x}\right) \]
    2. Add Preprocessing
    3. Taylor expanded in x around 0 98.7%

      \[\leadsto \color{blue}{-0.005555555555555556 \cdot {x}^{4} + \left(0.0003527336860670194 \cdot {x}^{6} + 0.16666666666666666 \cdot {x}^{2}\right)} \]
    4. Step-by-step derivation
      1. add-sqr-sqrt98.4%

        \[\leadsto -0.005555555555555556 \cdot {x}^{4} + \left(0.0003527336860670194 \cdot {x}^{6} + \color{blue}{\sqrt{0.16666666666666666 \cdot {x}^{2}} \cdot \sqrt{0.16666666666666666 \cdot {x}^{2}}}\right) \]
      2. pow298.4%

        \[\leadsto -0.005555555555555556 \cdot {x}^{4} + \left(0.0003527336860670194 \cdot {x}^{6} + \color{blue}{{\left(\sqrt{0.16666666666666666 \cdot {x}^{2}}\right)}^{2}}\right) \]
      3. *-commutative98.4%

        \[\leadsto -0.005555555555555556 \cdot {x}^{4} + \left(0.0003527336860670194 \cdot {x}^{6} + {\left(\sqrt{\color{blue}{{x}^{2} \cdot 0.16666666666666666}}\right)}^{2}\right) \]
      4. sqrt-prod98.5%

        \[\leadsto -0.005555555555555556 \cdot {x}^{4} + \left(0.0003527336860670194 \cdot {x}^{6} + {\color{blue}{\left(\sqrt{{x}^{2}} \cdot \sqrt{0.16666666666666666}\right)}}^{2}\right) \]
      5. unpow298.5%

        \[\leadsto -0.005555555555555556 \cdot {x}^{4} + \left(0.0003527336860670194 \cdot {x}^{6} + {\left(\sqrt{\color{blue}{x \cdot x}} \cdot \sqrt{0.16666666666666666}\right)}^{2}\right) \]
      6. sqrt-prod46.0%

        \[\leadsto -0.005555555555555556 \cdot {x}^{4} + \left(0.0003527336860670194 \cdot {x}^{6} + {\left(\color{blue}{\left(\sqrt{x} \cdot \sqrt{x}\right)} \cdot \sqrt{0.16666666666666666}\right)}^{2}\right) \]
      7. add-sqr-sqrt98.5%

        \[\leadsto -0.005555555555555556 \cdot {x}^{4} + \left(0.0003527336860670194 \cdot {x}^{6} + {\left(\color{blue}{x} \cdot \sqrt{0.16666666666666666}\right)}^{2}\right) \]
    5. Applied egg-rr98.5%

      \[\leadsto -0.005555555555555556 \cdot {x}^{4} + \left(0.0003527336860670194 \cdot {x}^{6} + \color{blue}{{\left(x \cdot \sqrt{0.16666666666666666}\right)}^{2}}\right) \]
    6. Step-by-step derivation
      1. *-commutative98.5%

        \[\leadsto -0.005555555555555556 \cdot {x}^{4} + \left(0.0003527336860670194 \cdot {x}^{6} + {\color{blue}{\left(\sqrt{0.16666666666666666} \cdot x\right)}}^{2}\right) \]
      2. add-sqr-sqrt46.0%

        \[\leadsto -0.005555555555555556 \cdot {x}^{4} + \left(0.0003527336860670194 \cdot {x}^{6} + {\left(\sqrt{0.16666666666666666} \cdot \color{blue}{\left(\sqrt{x} \cdot \sqrt{x}\right)}\right)}^{2}\right) \]
      3. sqrt-prod98.5%

        \[\leadsto -0.005555555555555556 \cdot {x}^{4} + \left(0.0003527336860670194 \cdot {x}^{6} + {\left(\sqrt{0.16666666666666666} \cdot \color{blue}{\sqrt{x \cdot x}}\right)}^{2}\right) \]
      4. unpow298.5%

        \[\leadsto -0.005555555555555556 \cdot {x}^{4} + \left(0.0003527336860670194 \cdot {x}^{6} + {\left(\sqrt{0.16666666666666666} \cdot \sqrt{\color{blue}{{x}^{2}}}\right)}^{2}\right) \]
      5. sqrt-prod98.4%

        \[\leadsto -0.005555555555555556 \cdot {x}^{4} + \left(0.0003527336860670194 \cdot {x}^{6} + {\color{blue}{\left(\sqrt{0.16666666666666666 \cdot {x}^{2}}\right)}}^{2}\right) \]
      6. pow298.4%

        \[\leadsto -0.005555555555555556 \cdot {x}^{4} + \left(0.0003527336860670194 \cdot {x}^{6} + \color{blue}{\sqrt{0.16666666666666666 \cdot {x}^{2}} \cdot \sqrt{0.16666666666666666 \cdot {x}^{2}}}\right) \]
      7. add-sqr-sqrt98.7%

        \[\leadsto -0.005555555555555556 \cdot {x}^{4} + \left(0.0003527336860670194 \cdot {x}^{6} + \color{blue}{0.16666666666666666 \cdot {x}^{2}}\right) \]
      8. unpow298.7%

        \[\leadsto -0.005555555555555556 \cdot {x}^{4} + \left(0.0003527336860670194 \cdot {x}^{6} + 0.16666666666666666 \cdot \color{blue}{\left(x \cdot x\right)}\right) \]
      9. associate-*r*98.7%

        \[\leadsto -0.005555555555555556 \cdot {x}^{4} + \left(0.0003527336860670194 \cdot {x}^{6} + \color{blue}{\left(0.16666666666666666 \cdot x\right) \cdot x}\right) \]
    7. Applied egg-rr98.7%

      \[\leadsto -0.005555555555555556 \cdot {x}^{4} + \left(0.0003527336860670194 \cdot {x}^{6} + \color{blue}{\left(0.16666666666666666 \cdot x\right) \cdot x}\right) \]

    if 0.0290000000000000015 < x < 700

    1. Initial program 96.4%

      \[\log \left(\frac{\sinh x}{x}\right) \]
    2. Add Preprocessing
    3. Step-by-step derivation
      1. clear-num96.4%

        \[\leadsto \log \color{blue}{\left(\frac{1}{\frac{x}{\sinh x}}\right)} \]
      2. neg-log96.8%

        \[\leadsto \color{blue}{-\log \left(\frac{x}{\sinh x}\right)} \]
    4. Applied egg-rr96.8%

      \[\leadsto \color{blue}{-\log \left(\frac{x}{\sinh x}\right)} \]
    5. Step-by-step derivation
      1. add-cbrt-cube96.8%

        \[\leadsto -\color{blue}{\sqrt[3]{\left(\log \left(\frac{x}{\sinh x}\right) \cdot \log \left(\frac{x}{\sinh x}\right)\right) \cdot \log \left(\frac{x}{\sinh x}\right)}} \]
      2. pow396.8%

        \[\leadsto -\sqrt[3]{\color{blue}{{\log \left(\frac{x}{\sinh x}\right)}^{3}}} \]
    6. Applied egg-rr96.8%

      \[\leadsto -\color{blue}{\sqrt[3]{{\log \left(\frac{x}{\sinh x}\right)}^{3}}} \]

    if 700 < x

    1. Initial program 3.2%

      \[\log \left(\frac{\sinh x}{x}\right) \]
    2. Add Preprocessing
    3. Taylor expanded in x around 0 13.2%

      \[\leadsto \log \left(\frac{\color{blue}{x + \left(0.0001984126984126984 \cdot {x}^{7} + \left(0.008333333333333333 \cdot {x}^{5} + 0.16666666666666666 \cdot {x}^{3}\right)\right)}}{x}\right) \]
    4. Taylor expanded in x around inf 13.2%

      \[\leadsto \color{blue}{\log 0.0001984126984126984 + \left(-6 \cdot \log \left(\frac{1}{x}\right) + 42 \cdot \frac{1}{{x}^{2}}\right)} \]
  3. Recombined 3 regimes into one program.
  4. Final simplification97.7%

    \[\leadsto \begin{array}{l} \mathbf{if}\;x \leq 0.029:\\ \;\;\;\;-0.005555555555555556 \cdot {x}^{4} + \left(0.0003527336860670194 \cdot {x}^{6} + x \cdot \left(x \cdot 0.16666666666666666\right)\right)\\ \mathbf{elif}\;x \leq 700:\\ \;\;\;\;-\sqrt[3]{{\log \left(\frac{x}{\sinh x}\right)}^{3}}\\ \mathbf{else}:\\ \;\;\;\;\log 0.0001984126984126984 + \left(-6 \cdot \log \left(\frac{1}{x}\right) + 42 \cdot \frac{1}{{x}^{2}}\right)\\ \end{array} \]
  5. Add Preprocessing

Alternative 2: 97.9% accurate, 0.4× speedup?

\[\begin{array}{l} x_m = \left|x\right| \\ \begin{array}{l} t_0 := \frac{\sinh x_m}{x_m}\\ \mathbf{if}\;t_0 \leq 1.0002:\\ \;\;\;\;-0.005555555555555556 \cdot {x_m}^{4} + \left(0.0003527336860670194 \cdot {x_m}^{6} + x_m \cdot \left(x_m \cdot 0.16666666666666666\right)\right)\\ \mathbf{elif}\;t_0 \leq 5 \cdot 10^{+40}:\\ \;\;\;\;-\log \left(\frac{x_m}{\sinh x_m}\right)\\ \mathbf{else}:\\ \;\;\;\;\log 0.0001984126984126984 + \left(-6 \cdot \log \left(\frac{1}{x_m}\right) + 42 \cdot \frac{1}{{x_m}^{2}}\right)\\ \end{array} \end{array} \]
x_m = (fabs.f64 x)
(FPCore (x_m)
 :precision binary64
 (let* ((t_0 (/ (sinh x_m) x_m)))
   (if (<= t_0 1.0002)
     (+
      (* -0.005555555555555556 (pow x_m 4.0))
      (+
       (* 0.0003527336860670194 (pow x_m 6.0))
       (* x_m (* x_m 0.16666666666666666))))
     (if (<= t_0 5e+40)
       (- (log (/ x_m (sinh x_m))))
       (+
        (log 0.0001984126984126984)
        (+ (* -6.0 (log (/ 1.0 x_m))) (* 42.0 (/ 1.0 (pow x_m 2.0)))))))))
x_m = fabs(x);
double code(double x_m) {
	double t_0 = sinh(x_m) / x_m;
	double tmp;
	if (t_0 <= 1.0002) {
		tmp = (-0.005555555555555556 * pow(x_m, 4.0)) + ((0.0003527336860670194 * pow(x_m, 6.0)) + (x_m * (x_m * 0.16666666666666666)));
	} else if (t_0 <= 5e+40) {
		tmp = -log((x_m / sinh(x_m)));
	} else {
		tmp = log(0.0001984126984126984) + ((-6.0 * log((1.0 / x_m))) + (42.0 * (1.0 / pow(x_m, 2.0))));
	}
	return tmp;
}
x_m = abs(x)
real(8) function code(x_m)
    real(8), intent (in) :: x_m
    real(8) :: t_0
    real(8) :: tmp
    t_0 = sinh(x_m) / x_m
    if (t_0 <= 1.0002d0) then
        tmp = ((-0.005555555555555556d0) * (x_m ** 4.0d0)) + ((0.0003527336860670194d0 * (x_m ** 6.0d0)) + (x_m * (x_m * 0.16666666666666666d0)))
    else if (t_0 <= 5d+40) then
        tmp = -log((x_m / sinh(x_m)))
    else
        tmp = log(0.0001984126984126984d0) + (((-6.0d0) * log((1.0d0 / x_m))) + (42.0d0 * (1.0d0 / (x_m ** 2.0d0))))
    end if
    code = tmp
end function
x_m = Math.abs(x);
public static double code(double x_m) {
	double t_0 = Math.sinh(x_m) / x_m;
	double tmp;
	if (t_0 <= 1.0002) {
		tmp = (-0.005555555555555556 * Math.pow(x_m, 4.0)) + ((0.0003527336860670194 * Math.pow(x_m, 6.0)) + (x_m * (x_m * 0.16666666666666666)));
	} else if (t_0 <= 5e+40) {
		tmp = -Math.log((x_m / Math.sinh(x_m)));
	} else {
		tmp = Math.log(0.0001984126984126984) + ((-6.0 * Math.log((1.0 / x_m))) + (42.0 * (1.0 / Math.pow(x_m, 2.0))));
	}
	return tmp;
}
x_m = math.fabs(x)
def code(x_m):
	t_0 = math.sinh(x_m) / x_m
	tmp = 0
	if t_0 <= 1.0002:
		tmp = (-0.005555555555555556 * math.pow(x_m, 4.0)) + ((0.0003527336860670194 * math.pow(x_m, 6.0)) + (x_m * (x_m * 0.16666666666666666)))
	elif t_0 <= 5e+40:
		tmp = -math.log((x_m / math.sinh(x_m)))
	else:
		tmp = math.log(0.0001984126984126984) + ((-6.0 * math.log((1.0 / x_m))) + (42.0 * (1.0 / math.pow(x_m, 2.0))))
	return tmp
x_m = abs(x)
function code(x_m)
	t_0 = Float64(sinh(x_m) / x_m)
	tmp = 0.0
	if (t_0 <= 1.0002)
		tmp = Float64(Float64(-0.005555555555555556 * (x_m ^ 4.0)) + Float64(Float64(0.0003527336860670194 * (x_m ^ 6.0)) + Float64(x_m * Float64(x_m * 0.16666666666666666))));
	elseif (t_0 <= 5e+40)
		tmp = Float64(-log(Float64(x_m / sinh(x_m))));
	else
		tmp = Float64(log(0.0001984126984126984) + Float64(Float64(-6.0 * log(Float64(1.0 / x_m))) + Float64(42.0 * Float64(1.0 / (x_m ^ 2.0)))));
	end
	return tmp
end
x_m = abs(x);
function tmp_2 = code(x_m)
	t_0 = sinh(x_m) / x_m;
	tmp = 0.0;
	if (t_0 <= 1.0002)
		tmp = (-0.005555555555555556 * (x_m ^ 4.0)) + ((0.0003527336860670194 * (x_m ^ 6.0)) + (x_m * (x_m * 0.16666666666666666)));
	elseif (t_0 <= 5e+40)
		tmp = -log((x_m / sinh(x_m)));
	else
		tmp = log(0.0001984126984126984) + ((-6.0 * log((1.0 / x_m))) + (42.0 * (1.0 / (x_m ^ 2.0))));
	end
	tmp_2 = tmp;
end
x_m = N[Abs[x], $MachinePrecision]
code[x$95$m_] := Block[{t$95$0 = N[(N[Sinh[x$95$m], $MachinePrecision] / x$95$m), $MachinePrecision]}, If[LessEqual[t$95$0, 1.0002], N[(N[(-0.005555555555555556 * N[Power[x$95$m, 4.0], $MachinePrecision]), $MachinePrecision] + N[(N[(0.0003527336860670194 * N[Power[x$95$m, 6.0], $MachinePrecision]), $MachinePrecision] + N[(x$95$m * N[(x$95$m * 0.16666666666666666), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]), $MachinePrecision], If[LessEqual[t$95$0, 5e+40], (-N[Log[N[(x$95$m / N[Sinh[x$95$m], $MachinePrecision]), $MachinePrecision]], $MachinePrecision]), N[(N[Log[0.0001984126984126984], $MachinePrecision] + N[(N[(-6.0 * N[Log[N[(1.0 / x$95$m), $MachinePrecision]], $MachinePrecision]), $MachinePrecision] + N[(42.0 * N[(1.0 / N[Power[x$95$m, 2.0], $MachinePrecision]), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]]]]
\begin{array}{l}
x_m = \left|x\right|

\\
\begin{array}{l}
t_0 := \frac{\sinh x_m}{x_m}\\
\mathbf{if}\;t_0 \leq 1.0002:\\
\;\;\;\;-0.005555555555555556 \cdot {x_m}^{4} + \left(0.0003527336860670194 \cdot {x_m}^{6} + x_m \cdot \left(x_m \cdot 0.16666666666666666\right)\right)\\

\mathbf{elif}\;t_0 \leq 5 \cdot 10^{+40}:\\
\;\;\;\;-\log \left(\frac{x_m}{\sinh x_m}\right)\\

\mathbf{else}:\\
\;\;\;\;\log 0.0001984126984126984 + \left(-6 \cdot \log \left(\frac{1}{x_m}\right) + 42 \cdot \frac{1}{{x_m}^{2}}\right)\\


\end{array}
\end{array}
Derivation
  1. Split input into 3 regimes
  2. if (/.f64 (sinh.f64 x) x) < 1.0002

    1. Initial program 56.7%

      \[\log \left(\frac{\sinh x}{x}\right) \]
    2. Add Preprocessing
    3. Taylor expanded in x around 0 99.7%

      \[\leadsto \color{blue}{-0.005555555555555556 \cdot {x}^{4} + \left(0.0003527336860670194 \cdot {x}^{6} + 0.16666666666666666 \cdot {x}^{2}\right)} \]
    4. Step-by-step derivation
      1. add-sqr-sqrt99.5%

        \[\leadsto -0.005555555555555556 \cdot {x}^{4} + \left(0.0003527336860670194 \cdot {x}^{6} + \color{blue}{\sqrt{0.16666666666666666 \cdot {x}^{2}} \cdot \sqrt{0.16666666666666666 \cdot {x}^{2}}}\right) \]
      2. pow299.5%

        \[\leadsto -0.005555555555555556 \cdot {x}^{4} + \left(0.0003527336860670194 \cdot {x}^{6} + \color{blue}{{\left(\sqrt{0.16666666666666666 \cdot {x}^{2}}\right)}^{2}}\right) \]
      3. *-commutative99.5%

        \[\leadsto -0.005555555555555556 \cdot {x}^{4} + \left(0.0003527336860670194 \cdot {x}^{6} + {\left(\sqrt{\color{blue}{{x}^{2} \cdot 0.16666666666666666}}\right)}^{2}\right) \]
      4. sqrt-prod99.6%

        \[\leadsto -0.005555555555555556 \cdot {x}^{4} + \left(0.0003527336860670194 \cdot {x}^{6} + {\color{blue}{\left(\sqrt{{x}^{2}} \cdot \sqrt{0.16666666666666666}\right)}}^{2}\right) \]
      5. unpow299.6%

        \[\leadsto -0.005555555555555556 \cdot {x}^{4} + \left(0.0003527336860670194 \cdot {x}^{6} + {\left(\sqrt{\color{blue}{x \cdot x}} \cdot \sqrt{0.16666666666666666}\right)}^{2}\right) \]
      6. sqrt-prod46.6%

        \[\leadsto -0.005555555555555556 \cdot {x}^{4} + \left(0.0003527336860670194 \cdot {x}^{6} + {\left(\color{blue}{\left(\sqrt{x} \cdot \sqrt{x}\right)} \cdot \sqrt{0.16666666666666666}\right)}^{2}\right) \]
      7. add-sqr-sqrt99.6%

        \[\leadsto -0.005555555555555556 \cdot {x}^{4} + \left(0.0003527336860670194 \cdot {x}^{6} + {\left(\color{blue}{x} \cdot \sqrt{0.16666666666666666}\right)}^{2}\right) \]
    5. Applied egg-rr99.6%

      \[\leadsto -0.005555555555555556 \cdot {x}^{4} + \left(0.0003527336860670194 \cdot {x}^{6} + \color{blue}{{\left(x \cdot \sqrt{0.16666666666666666}\right)}^{2}}\right) \]
    6. Step-by-step derivation
      1. *-commutative99.6%

        \[\leadsto -0.005555555555555556 \cdot {x}^{4} + \left(0.0003527336860670194 \cdot {x}^{6} + {\color{blue}{\left(\sqrt{0.16666666666666666} \cdot x\right)}}^{2}\right) \]
      2. add-sqr-sqrt46.6%

        \[\leadsto -0.005555555555555556 \cdot {x}^{4} + \left(0.0003527336860670194 \cdot {x}^{6} + {\left(\sqrt{0.16666666666666666} \cdot \color{blue}{\left(\sqrt{x} \cdot \sqrt{x}\right)}\right)}^{2}\right) \]
      3. sqrt-prod99.6%

        \[\leadsto -0.005555555555555556 \cdot {x}^{4} + \left(0.0003527336860670194 \cdot {x}^{6} + {\left(\sqrt{0.16666666666666666} \cdot \color{blue}{\sqrt{x \cdot x}}\right)}^{2}\right) \]
      4. unpow299.6%

        \[\leadsto -0.005555555555555556 \cdot {x}^{4} + \left(0.0003527336860670194 \cdot {x}^{6} + {\left(\sqrt{0.16666666666666666} \cdot \sqrt{\color{blue}{{x}^{2}}}\right)}^{2}\right) \]
      5. sqrt-prod99.5%

        \[\leadsto -0.005555555555555556 \cdot {x}^{4} + \left(0.0003527336860670194 \cdot {x}^{6} + {\color{blue}{\left(\sqrt{0.16666666666666666 \cdot {x}^{2}}\right)}}^{2}\right) \]
      6. pow299.5%

        \[\leadsto -0.005555555555555556 \cdot {x}^{4} + \left(0.0003527336860670194 \cdot {x}^{6} + \color{blue}{\sqrt{0.16666666666666666 \cdot {x}^{2}} \cdot \sqrt{0.16666666666666666 \cdot {x}^{2}}}\right) \]
      7. add-sqr-sqrt99.7%

        \[\leadsto -0.005555555555555556 \cdot {x}^{4} + \left(0.0003527336860670194 \cdot {x}^{6} + \color{blue}{0.16666666666666666 \cdot {x}^{2}}\right) \]
      8. unpow299.7%

        \[\leadsto -0.005555555555555556 \cdot {x}^{4} + \left(0.0003527336860670194 \cdot {x}^{6} + 0.16666666666666666 \cdot \color{blue}{\left(x \cdot x\right)}\right) \]
      9. associate-*r*99.8%

        \[\leadsto -0.005555555555555556 \cdot {x}^{4} + \left(0.0003527336860670194 \cdot {x}^{6} + \color{blue}{\left(0.16666666666666666 \cdot x\right) \cdot x}\right) \]
    7. Applied egg-rr99.8%

      \[\leadsto -0.005555555555555556 \cdot {x}^{4} + \left(0.0003527336860670194 \cdot {x}^{6} + \color{blue}{\left(0.16666666666666666 \cdot x\right) \cdot x}\right) \]

    if 1.0002 < (/.f64 (sinh.f64 x) x) < 5.00000000000000003e40

    1. Initial program 97.0%

      \[\log \left(\frac{\sinh x}{x}\right) \]
    2. Add Preprocessing
    3. Step-by-step derivation
      1. clear-num97.0%

        \[\leadsto \log \color{blue}{\left(\frac{1}{\frac{x}{\sinh x}}\right)} \]
      2. neg-log97.3%

        \[\leadsto \color{blue}{-\log \left(\frac{x}{\sinh x}\right)} \]
    4. Applied egg-rr97.3%

      \[\leadsto \color{blue}{-\log \left(\frac{x}{\sinh x}\right)} \]

    if 5.00000000000000003e40 < (/.f64 (sinh.f64 x) x)

    1. Initial program 3.2%

      \[\log \left(\frac{\sinh x}{x}\right) \]
    2. Add Preprocessing
    3. Taylor expanded in x around 0 13.4%

      \[\leadsto \log \left(\frac{\color{blue}{x + \left(0.0001984126984126984 \cdot {x}^{7} + \left(0.008333333333333333 \cdot {x}^{5} + 0.16666666666666666 \cdot {x}^{3}\right)\right)}}{x}\right) \]
    4. Taylor expanded in x around inf 7.9%

      \[\leadsto \color{blue}{\log 0.0001984126984126984 + \left(-6 \cdot \log \left(\frac{1}{x}\right) + 42 \cdot \frac{1}{{x}^{2}}\right)} \]
  3. Recombined 3 regimes into one program.
  4. Final simplification97.9%

    \[\leadsto \begin{array}{l} \mathbf{if}\;\frac{\sinh x}{x} \leq 1.0002:\\ \;\;\;\;-0.005555555555555556 \cdot {x}^{4} + \left(0.0003527336860670194 \cdot {x}^{6} + x \cdot \left(x \cdot 0.16666666666666666\right)\right)\\ \mathbf{elif}\;\frac{\sinh x}{x} \leq 5 \cdot 10^{+40}:\\ \;\;\;\;-\log \left(\frac{x}{\sinh x}\right)\\ \mathbf{else}:\\ \;\;\;\;\log 0.0001984126984126984 + \left(-6 \cdot \log \left(\frac{1}{x}\right) + 42 \cdot \frac{1}{{x}^{2}}\right)\\ \end{array} \]
  5. Add Preprocessing

Alternative 3: 97.9% accurate, 0.5× speedup?

\[\begin{array}{l} x_m = \left|x\right| \\ \begin{array}{l} \mathbf{if}\;x_m \leq 0.082:\\ \;\;\;\;-0.005555555555555556 \cdot {x_m}^{4} + \left(-2.6455026455026456 \cdot 10^{-5} \cdot {x_m}^{8} + \left(0.0003527336860670194 \cdot {x_m}^{6} + 0.16666666666666666 \cdot {x_m}^{2}\right)\right)\\ \mathbf{else}:\\ \;\;\;\;-\log \left(\frac{x_m}{\sinh x_m}\right)\\ \end{array} \end{array} \]
x_m = (fabs.f64 x)
(FPCore (x_m)
 :precision binary64
 (if (<= x_m 0.082)
   (+
    (* -0.005555555555555556 (pow x_m 4.0))
    (+
     (* -2.6455026455026456e-5 (pow x_m 8.0))
     (+
      (* 0.0003527336860670194 (pow x_m 6.0))
      (* 0.16666666666666666 (pow x_m 2.0)))))
   (- (log (/ x_m (sinh x_m))))))
x_m = fabs(x);
double code(double x_m) {
	double tmp;
	if (x_m <= 0.082) {
		tmp = (-0.005555555555555556 * pow(x_m, 4.0)) + ((-2.6455026455026456e-5 * pow(x_m, 8.0)) + ((0.0003527336860670194 * pow(x_m, 6.0)) + (0.16666666666666666 * pow(x_m, 2.0))));
	} else {
		tmp = -log((x_m / sinh(x_m)));
	}
	return tmp;
}
x_m = abs(x)
real(8) function code(x_m)
    real(8), intent (in) :: x_m
    real(8) :: tmp
    if (x_m <= 0.082d0) then
        tmp = ((-0.005555555555555556d0) * (x_m ** 4.0d0)) + (((-2.6455026455026456d-5) * (x_m ** 8.0d0)) + ((0.0003527336860670194d0 * (x_m ** 6.0d0)) + (0.16666666666666666d0 * (x_m ** 2.0d0))))
    else
        tmp = -log((x_m / sinh(x_m)))
    end if
    code = tmp
end function
x_m = Math.abs(x);
public static double code(double x_m) {
	double tmp;
	if (x_m <= 0.082) {
		tmp = (-0.005555555555555556 * Math.pow(x_m, 4.0)) + ((-2.6455026455026456e-5 * Math.pow(x_m, 8.0)) + ((0.0003527336860670194 * Math.pow(x_m, 6.0)) + (0.16666666666666666 * Math.pow(x_m, 2.0))));
	} else {
		tmp = -Math.log((x_m / Math.sinh(x_m)));
	}
	return tmp;
}
x_m = math.fabs(x)
def code(x_m):
	tmp = 0
	if x_m <= 0.082:
		tmp = (-0.005555555555555556 * math.pow(x_m, 4.0)) + ((-2.6455026455026456e-5 * math.pow(x_m, 8.0)) + ((0.0003527336860670194 * math.pow(x_m, 6.0)) + (0.16666666666666666 * math.pow(x_m, 2.0))))
	else:
		tmp = -math.log((x_m / math.sinh(x_m)))
	return tmp
x_m = abs(x)
function code(x_m)
	tmp = 0.0
	if (x_m <= 0.082)
		tmp = Float64(Float64(-0.005555555555555556 * (x_m ^ 4.0)) + Float64(Float64(-2.6455026455026456e-5 * (x_m ^ 8.0)) + Float64(Float64(0.0003527336860670194 * (x_m ^ 6.0)) + Float64(0.16666666666666666 * (x_m ^ 2.0)))));
	else
		tmp = Float64(-log(Float64(x_m / sinh(x_m))));
	end
	return tmp
end
x_m = abs(x);
function tmp_2 = code(x_m)
	tmp = 0.0;
	if (x_m <= 0.082)
		tmp = (-0.005555555555555556 * (x_m ^ 4.0)) + ((-2.6455026455026456e-5 * (x_m ^ 8.0)) + ((0.0003527336860670194 * (x_m ^ 6.0)) + (0.16666666666666666 * (x_m ^ 2.0))));
	else
		tmp = -log((x_m / sinh(x_m)));
	end
	tmp_2 = tmp;
end
x_m = N[Abs[x], $MachinePrecision]
code[x$95$m_] := If[LessEqual[x$95$m, 0.082], N[(N[(-0.005555555555555556 * N[Power[x$95$m, 4.0], $MachinePrecision]), $MachinePrecision] + N[(N[(-2.6455026455026456e-5 * N[Power[x$95$m, 8.0], $MachinePrecision]), $MachinePrecision] + N[(N[(0.0003527336860670194 * N[Power[x$95$m, 6.0], $MachinePrecision]), $MachinePrecision] + N[(0.16666666666666666 * N[Power[x$95$m, 2.0], $MachinePrecision]), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]), $MachinePrecision], (-N[Log[N[(x$95$m / N[Sinh[x$95$m], $MachinePrecision]), $MachinePrecision]], $MachinePrecision])]
\begin{array}{l}
x_m = \left|x\right|

\\
\begin{array}{l}
\mathbf{if}\;x_m \leq 0.082:\\
\;\;\;\;-0.005555555555555556 \cdot {x_m}^{4} + \left(-2.6455026455026456 \cdot 10^{-5} \cdot {x_m}^{8} + \left(0.0003527336860670194 \cdot {x_m}^{6} + 0.16666666666666666 \cdot {x_m}^{2}\right)\right)\\

\mathbf{else}:\\
\;\;\;\;-\log \left(\frac{x_m}{\sinh x_m}\right)\\


\end{array}
\end{array}
Derivation
  1. Split input into 2 regimes
  2. if x < 0.0820000000000000034

    1. Initial program 56.6%

      \[\log \left(\frac{\sinh x}{x}\right) \]
    2. Add Preprocessing
    3. Taylor expanded in x around 0 98.6%

      \[\leadsto \color{blue}{-0.005555555555555556 \cdot {x}^{4} + \left(-2.6455026455026456 \cdot 10^{-5} \cdot {x}^{8} + \left(0.0003527336860670194 \cdot {x}^{6} + 0.16666666666666666 \cdot {x}^{2}\right)\right)} \]

    if 0.0820000000000000034 < x

    1. Initial program 58.1%

      \[\log \left(\frac{\sinh x}{x}\right) \]
    2. Add Preprocessing
    3. Step-by-step derivation
      1. clear-num58.1%

        \[\leadsto \log \color{blue}{\left(\frac{1}{\frac{x}{\sinh x}}\right)} \]
      2. neg-log58.3%

        \[\leadsto \color{blue}{-\log \left(\frac{x}{\sinh x}\right)} \]
    4. Applied egg-rr58.3%

      \[\leadsto \color{blue}{-\log \left(\frac{x}{\sinh x}\right)} \]
  3. Recombined 2 regimes into one program.
  4. Final simplification97.5%

    \[\leadsto \begin{array}{l} \mathbf{if}\;x \leq 0.082:\\ \;\;\;\;-0.005555555555555556 \cdot {x}^{4} + \left(-2.6455026455026456 \cdot 10^{-5} \cdot {x}^{8} + \left(0.0003527336860670194 \cdot {x}^{6} + 0.16666666666666666 \cdot {x}^{2}\right)\right)\\ \mathbf{else}:\\ \;\;\;\;-\log \left(\frac{x}{\sinh x}\right)\\ \end{array} \]
  5. Add Preprocessing

Alternative 4: 97.9% accurate, 0.5× speedup?

\[\begin{array}{l} x_m = \left|x\right| \\ \begin{array}{l} t_0 := \frac{\sinh x_m}{x_m}\\ \mathbf{if}\;t_0 \leq 1.0002:\\ \;\;\;\;-0.005555555555555556 \cdot {x_m}^{4} + \left(0.0003527336860670194 \cdot {x_m}^{6} + x_m \cdot \left(x_m \cdot 0.16666666666666666\right)\right)\\ \mathbf{elif}\;t_0 \leq 5 \cdot 10^{+40}:\\ \;\;\;\;-\log \left(\frac{x_m}{\sinh x_m}\right)\\ \mathbf{else}:\\ \;\;\;\;\log 0.0001984126984126984 + 6 \cdot \log x_m\\ \end{array} \end{array} \]
x_m = (fabs.f64 x)
(FPCore (x_m)
 :precision binary64
 (let* ((t_0 (/ (sinh x_m) x_m)))
   (if (<= t_0 1.0002)
     (+
      (* -0.005555555555555556 (pow x_m 4.0))
      (+
       (* 0.0003527336860670194 (pow x_m 6.0))
       (* x_m (* x_m 0.16666666666666666))))
     (if (<= t_0 5e+40)
       (- (log (/ x_m (sinh x_m))))
       (+ (log 0.0001984126984126984) (* 6.0 (log x_m)))))))
x_m = fabs(x);
double code(double x_m) {
	double t_0 = sinh(x_m) / x_m;
	double tmp;
	if (t_0 <= 1.0002) {
		tmp = (-0.005555555555555556 * pow(x_m, 4.0)) + ((0.0003527336860670194 * pow(x_m, 6.0)) + (x_m * (x_m * 0.16666666666666666)));
	} else if (t_0 <= 5e+40) {
		tmp = -log((x_m / sinh(x_m)));
	} else {
		tmp = log(0.0001984126984126984) + (6.0 * log(x_m));
	}
	return tmp;
}
x_m = abs(x)
real(8) function code(x_m)
    real(8), intent (in) :: x_m
    real(8) :: t_0
    real(8) :: tmp
    t_0 = sinh(x_m) / x_m
    if (t_0 <= 1.0002d0) then
        tmp = ((-0.005555555555555556d0) * (x_m ** 4.0d0)) + ((0.0003527336860670194d0 * (x_m ** 6.0d0)) + (x_m * (x_m * 0.16666666666666666d0)))
    else if (t_0 <= 5d+40) then
        tmp = -log((x_m / sinh(x_m)))
    else
        tmp = log(0.0001984126984126984d0) + (6.0d0 * log(x_m))
    end if
    code = tmp
end function
x_m = Math.abs(x);
public static double code(double x_m) {
	double t_0 = Math.sinh(x_m) / x_m;
	double tmp;
	if (t_0 <= 1.0002) {
		tmp = (-0.005555555555555556 * Math.pow(x_m, 4.0)) + ((0.0003527336860670194 * Math.pow(x_m, 6.0)) + (x_m * (x_m * 0.16666666666666666)));
	} else if (t_0 <= 5e+40) {
		tmp = -Math.log((x_m / Math.sinh(x_m)));
	} else {
		tmp = Math.log(0.0001984126984126984) + (6.0 * Math.log(x_m));
	}
	return tmp;
}
x_m = math.fabs(x)
def code(x_m):
	t_0 = math.sinh(x_m) / x_m
	tmp = 0
	if t_0 <= 1.0002:
		tmp = (-0.005555555555555556 * math.pow(x_m, 4.0)) + ((0.0003527336860670194 * math.pow(x_m, 6.0)) + (x_m * (x_m * 0.16666666666666666)))
	elif t_0 <= 5e+40:
		tmp = -math.log((x_m / math.sinh(x_m)))
	else:
		tmp = math.log(0.0001984126984126984) + (6.0 * math.log(x_m))
	return tmp
x_m = abs(x)
function code(x_m)
	t_0 = Float64(sinh(x_m) / x_m)
	tmp = 0.0
	if (t_0 <= 1.0002)
		tmp = Float64(Float64(-0.005555555555555556 * (x_m ^ 4.0)) + Float64(Float64(0.0003527336860670194 * (x_m ^ 6.0)) + Float64(x_m * Float64(x_m * 0.16666666666666666))));
	elseif (t_0 <= 5e+40)
		tmp = Float64(-log(Float64(x_m / sinh(x_m))));
	else
		tmp = Float64(log(0.0001984126984126984) + Float64(6.0 * log(x_m)));
	end
	return tmp
end
x_m = abs(x);
function tmp_2 = code(x_m)
	t_0 = sinh(x_m) / x_m;
	tmp = 0.0;
	if (t_0 <= 1.0002)
		tmp = (-0.005555555555555556 * (x_m ^ 4.0)) + ((0.0003527336860670194 * (x_m ^ 6.0)) + (x_m * (x_m * 0.16666666666666666)));
	elseif (t_0 <= 5e+40)
		tmp = -log((x_m / sinh(x_m)));
	else
		tmp = log(0.0001984126984126984) + (6.0 * log(x_m));
	end
	tmp_2 = tmp;
end
x_m = N[Abs[x], $MachinePrecision]
code[x$95$m_] := Block[{t$95$0 = N[(N[Sinh[x$95$m], $MachinePrecision] / x$95$m), $MachinePrecision]}, If[LessEqual[t$95$0, 1.0002], N[(N[(-0.005555555555555556 * N[Power[x$95$m, 4.0], $MachinePrecision]), $MachinePrecision] + N[(N[(0.0003527336860670194 * N[Power[x$95$m, 6.0], $MachinePrecision]), $MachinePrecision] + N[(x$95$m * N[(x$95$m * 0.16666666666666666), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]), $MachinePrecision], If[LessEqual[t$95$0, 5e+40], (-N[Log[N[(x$95$m / N[Sinh[x$95$m], $MachinePrecision]), $MachinePrecision]], $MachinePrecision]), N[(N[Log[0.0001984126984126984], $MachinePrecision] + N[(6.0 * N[Log[x$95$m], $MachinePrecision]), $MachinePrecision]), $MachinePrecision]]]]
\begin{array}{l}
x_m = \left|x\right|

\\
\begin{array}{l}
t_0 := \frac{\sinh x_m}{x_m}\\
\mathbf{if}\;t_0 \leq 1.0002:\\
\;\;\;\;-0.005555555555555556 \cdot {x_m}^{4} + \left(0.0003527336860670194 \cdot {x_m}^{6} + x_m \cdot \left(x_m \cdot 0.16666666666666666\right)\right)\\

\mathbf{elif}\;t_0 \leq 5 \cdot 10^{+40}:\\
\;\;\;\;-\log \left(\frac{x_m}{\sinh x_m}\right)\\

\mathbf{else}:\\
\;\;\;\;\log 0.0001984126984126984 + 6 \cdot \log x_m\\


\end{array}
\end{array}
Derivation
  1. Split input into 3 regimes
  2. if (/.f64 (sinh.f64 x) x) < 1.0002

    1. Initial program 56.7%

      \[\log \left(\frac{\sinh x}{x}\right) \]
    2. Add Preprocessing
    3. Taylor expanded in x around 0 99.7%

      \[\leadsto \color{blue}{-0.005555555555555556 \cdot {x}^{4} + \left(0.0003527336860670194 \cdot {x}^{6} + 0.16666666666666666 \cdot {x}^{2}\right)} \]
    4. Step-by-step derivation
      1. add-sqr-sqrt99.5%

        \[\leadsto -0.005555555555555556 \cdot {x}^{4} + \left(0.0003527336860670194 \cdot {x}^{6} + \color{blue}{\sqrt{0.16666666666666666 \cdot {x}^{2}} \cdot \sqrt{0.16666666666666666 \cdot {x}^{2}}}\right) \]
      2. pow299.5%

        \[\leadsto -0.005555555555555556 \cdot {x}^{4} + \left(0.0003527336860670194 \cdot {x}^{6} + \color{blue}{{\left(\sqrt{0.16666666666666666 \cdot {x}^{2}}\right)}^{2}}\right) \]
      3. *-commutative99.5%

        \[\leadsto -0.005555555555555556 \cdot {x}^{4} + \left(0.0003527336860670194 \cdot {x}^{6} + {\left(\sqrt{\color{blue}{{x}^{2} \cdot 0.16666666666666666}}\right)}^{2}\right) \]
      4. sqrt-prod99.6%

        \[\leadsto -0.005555555555555556 \cdot {x}^{4} + \left(0.0003527336860670194 \cdot {x}^{6} + {\color{blue}{\left(\sqrt{{x}^{2}} \cdot \sqrt{0.16666666666666666}\right)}}^{2}\right) \]
      5. unpow299.6%

        \[\leadsto -0.005555555555555556 \cdot {x}^{4} + \left(0.0003527336860670194 \cdot {x}^{6} + {\left(\sqrt{\color{blue}{x \cdot x}} \cdot \sqrt{0.16666666666666666}\right)}^{2}\right) \]
      6. sqrt-prod46.6%

        \[\leadsto -0.005555555555555556 \cdot {x}^{4} + \left(0.0003527336860670194 \cdot {x}^{6} + {\left(\color{blue}{\left(\sqrt{x} \cdot \sqrt{x}\right)} \cdot \sqrt{0.16666666666666666}\right)}^{2}\right) \]
      7. add-sqr-sqrt99.6%

        \[\leadsto -0.005555555555555556 \cdot {x}^{4} + \left(0.0003527336860670194 \cdot {x}^{6} + {\left(\color{blue}{x} \cdot \sqrt{0.16666666666666666}\right)}^{2}\right) \]
    5. Applied egg-rr99.6%

      \[\leadsto -0.005555555555555556 \cdot {x}^{4} + \left(0.0003527336860670194 \cdot {x}^{6} + \color{blue}{{\left(x \cdot \sqrt{0.16666666666666666}\right)}^{2}}\right) \]
    6. Step-by-step derivation
      1. *-commutative99.6%

        \[\leadsto -0.005555555555555556 \cdot {x}^{4} + \left(0.0003527336860670194 \cdot {x}^{6} + {\color{blue}{\left(\sqrt{0.16666666666666666} \cdot x\right)}}^{2}\right) \]
      2. add-sqr-sqrt46.6%

        \[\leadsto -0.005555555555555556 \cdot {x}^{4} + \left(0.0003527336860670194 \cdot {x}^{6} + {\left(\sqrt{0.16666666666666666} \cdot \color{blue}{\left(\sqrt{x} \cdot \sqrt{x}\right)}\right)}^{2}\right) \]
      3. sqrt-prod99.6%

        \[\leadsto -0.005555555555555556 \cdot {x}^{4} + \left(0.0003527336860670194 \cdot {x}^{6} + {\left(\sqrt{0.16666666666666666} \cdot \color{blue}{\sqrt{x \cdot x}}\right)}^{2}\right) \]
      4. unpow299.6%

        \[\leadsto -0.005555555555555556 \cdot {x}^{4} + \left(0.0003527336860670194 \cdot {x}^{6} + {\left(\sqrt{0.16666666666666666} \cdot \sqrt{\color{blue}{{x}^{2}}}\right)}^{2}\right) \]
      5. sqrt-prod99.5%

        \[\leadsto -0.005555555555555556 \cdot {x}^{4} + \left(0.0003527336860670194 \cdot {x}^{6} + {\color{blue}{\left(\sqrt{0.16666666666666666 \cdot {x}^{2}}\right)}}^{2}\right) \]
      6. pow299.5%

        \[\leadsto -0.005555555555555556 \cdot {x}^{4} + \left(0.0003527336860670194 \cdot {x}^{6} + \color{blue}{\sqrt{0.16666666666666666 \cdot {x}^{2}} \cdot \sqrt{0.16666666666666666 \cdot {x}^{2}}}\right) \]
      7. add-sqr-sqrt99.7%

        \[\leadsto -0.005555555555555556 \cdot {x}^{4} + \left(0.0003527336860670194 \cdot {x}^{6} + \color{blue}{0.16666666666666666 \cdot {x}^{2}}\right) \]
      8. unpow299.7%

        \[\leadsto -0.005555555555555556 \cdot {x}^{4} + \left(0.0003527336860670194 \cdot {x}^{6} + 0.16666666666666666 \cdot \color{blue}{\left(x \cdot x\right)}\right) \]
      9. associate-*r*99.8%

        \[\leadsto -0.005555555555555556 \cdot {x}^{4} + \left(0.0003527336860670194 \cdot {x}^{6} + \color{blue}{\left(0.16666666666666666 \cdot x\right) \cdot x}\right) \]
    7. Applied egg-rr99.8%

      \[\leadsto -0.005555555555555556 \cdot {x}^{4} + \left(0.0003527336860670194 \cdot {x}^{6} + \color{blue}{\left(0.16666666666666666 \cdot x\right) \cdot x}\right) \]

    if 1.0002 < (/.f64 (sinh.f64 x) x) < 5.00000000000000003e40

    1. Initial program 97.0%

      \[\log \left(\frac{\sinh x}{x}\right) \]
    2. Add Preprocessing
    3. Step-by-step derivation
      1. clear-num97.0%

        \[\leadsto \log \color{blue}{\left(\frac{1}{\frac{x}{\sinh x}}\right)} \]
      2. neg-log97.3%

        \[\leadsto \color{blue}{-\log \left(\frac{x}{\sinh x}\right)} \]
    4. Applied egg-rr97.3%

      \[\leadsto \color{blue}{-\log \left(\frac{x}{\sinh x}\right)} \]

    if 5.00000000000000003e40 < (/.f64 (sinh.f64 x) x)

    1. Initial program 3.2%

      \[\log \left(\frac{\sinh x}{x}\right) \]
    2. Add Preprocessing
    3. Taylor expanded in x around 0 13.4%

      \[\leadsto \log \left(\frac{\color{blue}{x + \left(0.0001984126984126984 \cdot {x}^{7} + \left(0.008333333333333333 \cdot {x}^{5} + 0.16666666666666666 \cdot {x}^{3}\right)\right)}}{x}\right) \]
    4. Taylor expanded in x around inf 7.9%

      \[\leadsto \color{blue}{\log 0.0001984126984126984 + -6 \cdot \log \left(\frac{1}{x}\right)} \]
    5. Step-by-step derivation
      1. log-rec7.9%

        \[\leadsto \log 0.0001984126984126984 + -6 \cdot \color{blue}{\left(-\log x\right)} \]
      2. neg-mul-17.9%

        \[\leadsto \log 0.0001984126984126984 + -6 \cdot \color{blue}{\left(-1 \cdot \log x\right)} \]
      3. associate-*r*7.9%

        \[\leadsto \log 0.0001984126984126984 + \color{blue}{\left(-6 \cdot -1\right) \cdot \log x} \]
      4. metadata-eval7.9%

        \[\leadsto \log 0.0001984126984126984 + \color{blue}{6} \cdot \log x \]
    6. Simplified7.9%

      \[\leadsto \color{blue}{\log 0.0001984126984126984 + 6 \cdot \log x} \]
  3. Recombined 3 regimes into one program.
  4. Final simplification97.9%

    \[\leadsto \begin{array}{l} \mathbf{if}\;\frac{\sinh x}{x} \leq 1.0002:\\ \;\;\;\;-0.005555555555555556 \cdot {x}^{4} + \left(0.0003527336860670194 \cdot {x}^{6} + x \cdot \left(x \cdot 0.16666666666666666\right)\right)\\ \mathbf{elif}\;\frac{\sinh x}{x} \leq 5 \cdot 10^{+40}:\\ \;\;\;\;-\log \left(\frac{x}{\sinh x}\right)\\ \mathbf{else}:\\ \;\;\;\;\log 0.0001984126984126984 + 6 \cdot \log x\\ \end{array} \]
  5. Add Preprocessing

Alternative 5: 97.9% accurate, 0.6× speedup?

\[\begin{array}{l} x_m = \left|x\right| \\ \begin{array}{l} \mathbf{if}\;\frac{\sinh x_m}{x_m} \leq 1.00005:\\ \;\;\;\;\mathsf{fma}\left(x_m \cdot 0.16666666666666666, x_m, -0.005555555555555556 \cdot {x_m}^{4}\right)\\ \mathbf{else}:\\ \;\;\;\;-\log \left(\frac{x_m}{\sinh x_m}\right)\\ \end{array} \end{array} \]
x_m = (fabs.f64 x)
(FPCore (x_m)
 :precision binary64
 (if (<= (/ (sinh x_m) x_m) 1.00005)
   (fma
    (* x_m 0.16666666666666666)
    x_m
    (* -0.005555555555555556 (pow x_m 4.0)))
   (- (log (/ x_m (sinh x_m))))))
x_m = fabs(x);
double code(double x_m) {
	double tmp;
	if ((sinh(x_m) / x_m) <= 1.00005) {
		tmp = fma((x_m * 0.16666666666666666), x_m, (-0.005555555555555556 * pow(x_m, 4.0)));
	} else {
		tmp = -log((x_m / sinh(x_m)));
	}
	return tmp;
}
x_m = abs(x)
function code(x_m)
	tmp = 0.0
	if (Float64(sinh(x_m) / x_m) <= 1.00005)
		tmp = fma(Float64(x_m * 0.16666666666666666), x_m, Float64(-0.005555555555555556 * (x_m ^ 4.0)));
	else
		tmp = Float64(-log(Float64(x_m / sinh(x_m))));
	end
	return tmp
end
x_m = N[Abs[x], $MachinePrecision]
code[x$95$m_] := If[LessEqual[N[(N[Sinh[x$95$m], $MachinePrecision] / x$95$m), $MachinePrecision], 1.00005], N[(N[(x$95$m * 0.16666666666666666), $MachinePrecision] * x$95$m + N[(-0.005555555555555556 * N[Power[x$95$m, 4.0], $MachinePrecision]), $MachinePrecision]), $MachinePrecision], (-N[Log[N[(x$95$m / N[Sinh[x$95$m], $MachinePrecision]), $MachinePrecision]], $MachinePrecision])]
\begin{array}{l}
x_m = \left|x\right|

\\
\begin{array}{l}
\mathbf{if}\;\frac{\sinh x_m}{x_m} \leq 1.00005:\\
\;\;\;\;\mathsf{fma}\left(x_m \cdot 0.16666666666666666, x_m, -0.005555555555555556 \cdot {x_m}^{4}\right)\\

\mathbf{else}:\\
\;\;\;\;-\log \left(\frac{x_m}{\sinh x_m}\right)\\


\end{array}
\end{array}
Derivation
  1. Split input into 2 regimes
  2. if (/.f64 (sinh.f64 x) x) < 1.00005000000000011

    1. Initial program 56.6%

      \[\log \left(\frac{\sinh x}{x}\right) \]
    2. Add Preprocessing
    3. Taylor expanded in x around 0 99.7%

      \[\leadsto \color{blue}{-0.005555555555555556 \cdot {x}^{4} + 0.16666666666666666 \cdot {x}^{2}} \]
    4. Step-by-step derivation
      1. +-commutative99.7%

        \[\leadsto \color{blue}{0.16666666666666666 \cdot {x}^{2} + -0.005555555555555556 \cdot {x}^{4}} \]
      2. unpow299.7%

        \[\leadsto 0.16666666666666666 \cdot \color{blue}{\left(x \cdot x\right)} + -0.005555555555555556 \cdot {x}^{4} \]
      3. associate-*r*99.8%

        \[\leadsto \color{blue}{\left(0.16666666666666666 \cdot x\right) \cdot x} + -0.005555555555555556 \cdot {x}^{4} \]
      4. fma-def99.8%

        \[\leadsto \color{blue}{\mathsf{fma}\left(0.16666666666666666 \cdot x, x, -0.005555555555555556 \cdot {x}^{4}\right)} \]
    5. Applied egg-rr99.8%

      \[\leadsto \color{blue}{\mathsf{fma}\left(0.16666666666666666 \cdot x, x, -0.005555555555555556 \cdot {x}^{4}\right)} \]

    if 1.00005000000000011 < (/.f64 (sinh.f64 x) x)

    1. Initial program 56.6%

      \[\log \left(\frac{\sinh x}{x}\right) \]
    2. Add Preprocessing
    3. Step-by-step derivation
      1. clear-num56.6%

        \[\leadsto \log \color{blue}{\left(\frac{1}{\frac{x}{\sinh x}}\right)} \]
      2. neg-log56.9%

        \[\leadsto \color{blue}{-\log \left(\frac{x}{\sinh x}\right)} \]
    4. Applied egg-rr56.9%

      \[\leadsto \color{blue}{-\log \left(\frac{x}{\sinh x}\right)} \]
  3. Recombined 2 regimes into one program.
  4. Final simplification97.8%

    \[\leadsto \begin{array}{l} \mathbf{if}\;\frac{\sinh x}{x} \leq 1.00005:\\ \;\;\;\;\mathsf{fma}\left(x \cdot 0.16666666666666666, x, -0.005555555555555556 \cdot {x}^{4}\right)\\ \mathbf{else}:\\ \;\;\;\;-\log \left(\frac{x}{\sinh x}\right)\\ \end{array} \]
  5. Add Preprocessing

Alternative 6: 97.5% accurate, 0.7× speedup?

\[\begin{array}{l} x_m = \left|x\right| \\ \begin{array}{l} \mathbf{if}\;\frac{\sinh x_m}{x_m} \leq 1.0000001:\\ \;\;\;\;0.16666666666666666 \cdot {x_m}^{2}\\ \mathbf{else}:\\ \;\;\;\;-\log \left(\frac{x_m}{\sinh x_m}\right)\\ \end{array} \end{array} \]
x_m = (fabs.f64 x)
(FPCore (x_m)
 :precision binary64
 (if (<= (/ (sinh x_m) x_m) 1.0000001)
   (* 0.16666666666666666 (pow x_m 2.0))
   (- (log (/ x_m (sinh x_m))))))
x_m = fabs(x);
double code(double x_m) {
	double tmp;
	if ((sinh(x_m) / x_m) <= 1.0000001) {
		tmp = 0.16666666666666666 * pow(x_m, 2.0);
	} else {
		tmp = -log((x_m / sinh(x_m)));
	}
	return tmp;
}
x_m = abs(x)
real(8) function code(x_m)
    real(8), intent (in) :: x_m
    real(8) :: tmp
    if ((sinh(x_m) / x_m) <= 1.0000001d0) then
        tmp = 0.16666666666666666d0 * (x_m ** 2.0d0)
    else
        tmp = -log((x_m / sinh(x_m)))
    end if
    code = tmp
end function
x_m = Math.abs(x);
public static double code(double x_m) {
	double tmp;
	if ((Math.sinh(x_m) / x_m) <= 1.0000001) {
		tmp = 0.16666666666666666 * Math.pow(x_m, 2.0);
	} else {
		tmp = -Math.log((x_m / Math.sinh(x_m)));
	}
	return tmp;
}
x_m = math.fabs(x)
def code(x_m):
	tmp = 0
	if (math.sinh(x_m) / x_m) <= 1.0000001:
		tmp = 0.16666666666666666 * math.pow(x_m, 2.0)
	else:
		tmp = -math.log((x_m / math.sinh(x_m)))
	return tmp
x_m = abs(x)
function code(x_m)
	tmp = 0.0
	if (Float64(sinh(x_m) / x_m) <= 1.0000001)
		tmp = Float64(0.16666666666666666 * (x_m ^ 2.0));
	else
		tmp = Float64(-log(Float64(x_m / sinh(x_m))));
	end
	return tmp
end
x_m = abs(x);
function tmp_2 = code(x_m)
	tmp = 0.0;
	if ((sinh(x_m) / x_m) <= 1.0000001)
		tmp = 0.16666666666666666 * (x_m ^ 2.0);
	else
		tmp = -log((x_m / sinh(x_m)));
	end
	tmp_2 = tmp;
end
x_m = N[Abs[x], $MachinePrecision]
code[x$95$m_] := If[LessEqual[N[(N[Sinh[x$95$m], $MachinePrecision] / x$95$m), $MachinePrecision], 1.0000001], N[(0.16666666666666666 * N[Power[x$95$m, 2.0], $MachinePrecision]), $MachinePrecision], (-N[Log[N[(x$95$m / N[Sinh[x$95$m], $MachinePrecision]), $MachinePrecision]], $MachinePrecision])]
\begin{array}{l}
x_m = \left|x\right|

\\
\begin{array}{l}
\mathbf{if}\;\frac{\sinh x_m}{x_m} \leq 1.0000001:\\
\;\;\;\;0.16666666666666666 \cdot {x_m}^{2}\\

\mathbf{else}:\\
\;\;\;\;-\log \left(\frac{x_m}{\sinh x_m}\right)\\


\end{array}
\end{array}
Derivation
  1. Split input into 2 regimes
  2. if (/.f64 (sinh.f64 x) x) < 1.00000010000000006

    1. Initial program 56.5%

      \[\log \left(\frac{\sinh x}{x}\right) \]
    2. Add Preprocessing
    3. Taylor expanded in x around 0 99.7%

      \[\leadsto \color{blue}{0.16666666666666666 \cdot {x}^{2}} \]

    if 1.00000010000000006 < (/.f64 (sinh.f64 x) x)

    1. Initial program 58.1%

      \[\log \left(\frac{\sinh x}{x}\right) \]
    2. Add Preprocessing
    3. Step-by-step derivation
      1. clear-num58.1%

        \[\leadsto \log \color{blue}{\left(\frac{1}{\frac{x}{\sinh x}}\right)} \]
      2. neg-log59.1%

        \[\leadsto \color{blue}{-\log \left(\frac{x}{\sinh x}\right)} \]
    4. Applied egg-rr59.1%

      \[\leadsto \color{blue}{-\log \left(\frac{x}{\sinh x}\right)} \]
  3. Recombined 2 regimes into one program.
  4. Final simplification97.5%

    \[\leadsto \begin{array}{l} \mathbf{if}\;\frac{\sinh x}{x} \leq 1.0000001:\\ \;\;\;\;0.16666666666666666 \cdot {x}^{2}\\ \mathbf{else}:\\ \;\;\;\;-\log \left(\frac{x}{\sinh x}\right)\\ \end{array} \]
  5. Add Preprocessing

Alternative 7: 97.5% accurate, 0.7× speedup?

\[\begin{array}{l} x_m = \left|x\right| \\ \begin{array}{l} t_0 := \frac{\sinh x_m}{x_m}\\ \mathbf{if}\;t_0 \leq 1.0000001:\\ \;\;\;\;0.16666666666666666 \cdot {x_m}^{2}\\ \mathbf{else}:\\ \;\;\;\;\log t_0\\ \end{array} \end{array} \]
x_m = (fabs.f64 x)
(FPCore (x_m)
 :precision binary64
 (let* ((t_0 (/ (sinh x_m) x_m)))
   (if (<= t_0 1.0000001) (* 0.16666666666666666 (pow x_m 2.0)) (log t_0))))
x_m = fabs(x);
double code(double x_m) {
	double t_0 = sinh(x_m) / x_m;
	double tmp;
	if (t_0 <= 1.0000001) {
		tmp = 0.16666666666666666 * pow(x_m, 2.0);
	} else {
		tmp = log(t_0);
	}
	return tmp;
}
x_m = abs(x)
real(8) function code(x_m)
    real(8), intent (in) :: x_m
    real(8) :: t_0
    real(8) :: tmp
    t_0 = sinh(x_m) / x_m
    if (t_0 <= 1.0000001d0) then
        tmp = 0.16666666666666666d0 * (x_m ** 2.0d0)
    else
        tmp = log(t_0)
    end if
    code = tmp
end function
x_m = Math.abs(x);
public static double code(double x_m) {
	double t_0 = Math.sinh(x_m) / x_m;
	double tmp;
	if (t_0 <= 1.0000001) {
		tmp = 0.16666666666666666 * Math.pow(x_m, 2.0);
	} else {
		tmp = Math.log(t_0);
	}
	return tmp;
}
x_m = math.fabs(x)
def code(x_m):
	t_0 = math.sinh(x_m) / x_m
	tmp = 0
	if t_0 <= 1.0000001:
		tmp = 0.16666666666666666 * math.pow(x_m, 2.0)
	else:
		tmp = math.log(t_0)
	return tmp
x_m = abs(x)
function code(x_m)
	t_0 = Float64(sinh(x_m) / x_m)
	tmp = 0.0
	if (t_0 <= 1.0000001)
		tmp = Float64(0.16666666666666666 * (x_m ^ 2.0));
	else
		tmp = log(t_0);
	end
	return tmp
end
x_m = abs(x);
function tmp_2 = code(x_m)
	t_0 = sinh(x_m) / x_m;
	tmp = 0.0;
	if (t_0 <= 1.0000001)
		tmp = 0.16666666666666666 * (x_m ^ 2.0);
	else
		tmp = log(t_0);
	end
	tmp_2 = tmp;
end
x_m = N[Abs[x], $MachinePrecision]
code[x$95$m_] := Block[{t$95$0 = N[(N[Sinh[x$95$m], $MachinePrecision] / x$95$m), $MachinePrecision]}, If[LessEqual[t$95$0, 1.0000001], N[(0.16666666666666666 * N[Power[x$95$m, 2.0], $MachinePrecision]), $MachinePrecision], N[Log[t$95$0], $MachinePrecision]]]
\begin{array}{l}
x_m = \left|x\right|

\\
\begin{array}{l}
t_0 := \frac{\sinh x_m}{x_m}\\
\mathbf{if}\;t_0 \leq 1.0000001:\\
\;\;\;\;0.16666666666666666 \cdot {x_m}^{2}\\

\mathbf{else}:\\
\;\;\;\;\log t_0\\


\end{array}
\end{array}
Derivation
  1. Split input into 2 regimes
  2. if (/.f64 (sinh.f64 x) x) < 1.00000010000000006

    1. Initial program 56.5%

      \[\log \left(\frac{\sinh x}{x}\right) \]
    2. Add Preprocessing
    3. Taylor expanded in x around 0 99.7%

      \[\leadsto \color{blue}{0.16666666666666666 \cdot {x}^{2}} \]

    if 1.00000010000000006 < (/.f64 (sinh.f64 x) x)

    1. Initial program 58.1%

      \[\log \left(\frac{\sinh x}{x}\right) \]
    2. Add Preprocessing
  3. Recombined 2 regimes into one program.
  4. Final simplification97.4%

    \[\leadsto \begin{array}{l} \mathbf{if}\;\frac{\sinh x}{x} \leq 1.0000001:\\ \;\;\;\;0.16666666666666666 \cdot {x}^{2}\\ \mathbf{else}:\\ \;\;\;\;\log \left(\frac{\sinh x}{x}\right)\\ \end{array} \]
  5. Add Preprocessing

Alternative 8: 96.7% accurate, 2.0× speedup?

\[\begin{array}{l} x_m = \left|x\right| \\ 0.16666666666666666 \cdot {x_m}^{2} \end{array} \]
x_m = (fabs.f64 x)
(FPCore (x_m) :precision binary64 (* 0.16666666666666666 (pow x_m 2.0)))
x_m = fabs(x);
double code(double x_m) {
	return 0.16666666666666666 * pow(x_m, 2.0);
}
x_m = abs(x)
real(8) function code(x_m)
    real(8), intent (in) :: x_m
    code = 0.16666666666666666d0 * (x_m ** 2.0d0)
end function
x_m = Math.abs(x);
public static double code(double x_m) {
	return 0.16666666666666666 * Math.pow(x_m, 2.0);
}
x_m = math.fabs(x)
def code(x_m):
	return 0.16666666666666666 * math.pow(x_m, 2.0)
x_m = abs(x)
function code(x_m)
	return Float64(0.16666666666666666 * (x_m ^ 2.0))
end
x_m = abs(x);
function tmp = code(x_m)
	tmp = 0.16666666666666666 * (x_m ^ 2.0);
end
x_m = N[Abs[x], $MachinePrecision]
code[x$95$m_] := N[(0.16666666666666666 * N[Power[x$95$m, 2.0], $MachinePrecision]), $MachinePrecision]
\begin{array}{l}
x_m = \left|x\right|

\\
0.16666666666666666 \cdot {x_m}^{2}
\end{array}
Derivation
  1. Initial program 56.6%

    \[\log \left(\frac{\sinh x}{x}\right) \]
  2. Add Preprocessing
  3. Taylor expanded in x around 0 95.6%

    \[\leadsto \color{blue}{0.16666666666666666 \cdot {x}^{2}} \]
  4. Final simplification95.6%

    \[\leadsto 0.16666666666666666 \cdot {x}^{2} \]
  5. Add Preprocessing

Alternative 9: 50.9% accurate, 203.0× speedup?

\[\begin{array}{l} x_m = \left|x\right| \\ 0 \end{array} \]
x_m = (fabs.f64 x)
(FPCore (x_m) :precision binary64 0.0)
x_m = fabs(x);
double code(double x_m) {
	return 0.0;
}
x_m = abs(x)
real(8) function code(x_m)
    real(8), intent (in) :: x_m
    code = 0.0d0
end function
x_m = Math.abs(x);
public static double code(double x_m) {
	return 0.0;
}
x_m = math.fabs(x)
def code(x_m):
	return 0.0
x_m = abs(x)
function code(x_m)
	return 0.0
end
x_m = abs(x);
function tmp = code(x_m)
	tmp = 0.0;
end
x_m = N[Abs[x], $MachinePrecision]
code[x$95$m_] := 0.0
\begin{array}{l}
x_m = \left|x\right|

\\
0
\end{array}
Derivation
  1. Initial program 56.6%

    \[\log \left(\frac{\sinh x}{x}\right) \]
  2. Add Preprocessing
  3. Taylor expanded in x around 0 53.3%

    \[\leadsto \log \color{blue}{1} \]
  4. Final simplification53.3%

    \[\leadsto 0 \]
  5. Add Preprocessing

Developer target: 97.9% accurate, 0.5× speedup?

\[\begin{array}{l} \\ \begin{array}{l} \mathbf{if}\;\left|x\right| < 0.085:\\ \;\;\;\;\left(x \cdot x\right) \cdot \mathsf{fma}\left(\mathsf{fma}\left(\mathsf{fma}\left(-2.6455026455026456 \cdot 10^{-5}, x \cdot x, 0.0003527336860670194\right), x \cdot x, -0.005555555555555556\right), x \cdot x, 0.16666666666666666\right)\\ \mathbf{else}:\\ \;\;\;\;\log \left(\frac{\sinh x}{x}\right)\\ \end{array} \end{array} \]
(FPCore (x)
 :precision binary64
 (if (< (fabs x) 0.085)
   (*
    (* x x)
    (fma
     (fma
      (fma -2.6455026455026456e-5 (* x x) 0.0003527336860670194)
      (* x x)
      -0.005555555555555556)
     (* x x)
     0.16666666666666666))
   (log (/ (sinh x) x))))
double code(double x) {
	double tmp;
	if (fabs(x) < 0.085) {
		tmp = (x * x) * fma(fma(fma(-2.6455026455026456e-5, (x * x), 0.0003527336860670194), (x * x), -0.005555555555555556), (x * x), 0.16666666666666666);
	} else {
		tmp = log((sinh(x) / x));
	}
	return tmp;
}
function code(x)
	tmp = 0.0
	if (abs(x) < 0.085)
		tmp = Float64(Float64(x * x) * fma(fma(fma(-2.6455026455026456e-5, Float64(x * x), 0.0003527336860670194), Float64(x * x), -0.005555555555555556), Float64(x * x), 0.16666666666666666));
	else
		tmp = log(Float64(sinh(x) / x));
	end
	return tmp
end
code[x_] := If[Less[N[Abs[x], $MachinePrecision], 0.085], N[(N[(x * x), $MachinePrecision] * N[(N[(N[(-2.6455026455026456e-5 * N[(x * x), $MachinePrecision] + 0.0003527336860670194), $MachinePrecision] * N[(x * x), $MachinePrecision] + -0.005555555555555556), $MachinePrecision] * N[(x * x), $MachinePrecision] + 0.16666666666666666), $MachinePrecision]), $MachinePrecision], N[Log[N[(N[Sinh[x], $MachinePrecision] / x), $MachinePrecision]], $MachinePrecision]]
\begin{array}{l}

\\
\begin{array}{l}
\mathbf{if}\;\left|x\right| < 0.085:\\
\;\;\;\;\left(x \cdot x\right) \cdot \mathsf{fma}\left(\mathsf{fma}\left(\mathsf{fma}\left(-2.6455026455026456 \cdot 10^{-5}, x \cdot x, 0.0003527336860670194\right), x \cdot x, -0.005555555555555556\right), x \cdot x, 0.16666666666666666\right)\\

\mathbf{else}:\\
\;\;\;\;\log \left(\frac{\sinh x}{x}\right)\\


\end{array}
\end{array}

Reproduce

?
herbie shell --seed 2024019 
(FPCore (x)
  :name "bug500, discussion (missed optimization)"
  :precision binary64

  :herbie-target
  (if (< (fabs x) 0.085) (* (* x x) (fma (fma (fma -2.6455026455026456e-5 (* x x) 0.0003527336860670194) (* x x) -0.005555555555555556) (* x x) 0.16666666666666666)) (log (/ (sinh x) x)))

  (log (/ (sinh x) x)))