VandenBroeck and Keller, Equation (20)

Percentage Accurate: 7.0% → 98.9%

Time: 30.3s

Alternatives: 8

Speedup: 4.9×

• could not determine a ground truth

  1. f = 2.00000000000000007e154

Specification

Math

\[\begin{array}{l} \\ \begin{array}{l} t_0 := \frac{\pi}{4} \cdot f\\ t_1 := e^{t\_0}\\ t_2 := e^{-t\_0}\\ -\frac{1}{\frac{\pi}{4}} \cdot \log \left(\frac{t\_1 + t\_2}{t\_1 - t\_2}\right) \end{array} \end{array} \]

FPCore

(FPCore (f)
 :precision binary64
 (let* ((t_0 (* (/ PI 4.0) f)) (t_1 (exp t_0)) (t_2 (exp (- t_0))))
   (- (* (/ 1.0 (/ PI 4.0)) (log (/ (+ t_1 t_2) (- t_1 t_2)))))))

C

double code(double f) {
	double t_0 = (((double) M_PI) / 4.0) * f;
	double t_1 = exp(t_0);
	double t_2 = exp(-t_0);
	return -((1.0 / (((double) M_PI) / 4.0)) * log(((t_1 + t_2) / (t_1 - t_2))));
}

Java

public static double code(double f) {
	double t_0 = (Math.PI / 4.0) * f;
	double t_1 = Math.exp(t_0);
	double t_2 = Math.exp(-t_0);
	return -((1.0 / (Math.PI / 4.0)) * Math.log(((t_1 + t_2) / (t_1 - t_2))));
}

Python

def code(f):
	t_0 = (math.pi / 4.0) * f
	t_1 = math.exp(t_0)
	t_2 = math.exp(-t_0)
	return -((1.0 / (math.pi / 4.0)) * math.log(((t_1 + t_2) / (t_1 - t_2))))

Julia

function code(f)
	t_0 = Float64(Float64(pi / 4.0) * f)
	t_1 = exp(t_0)
	t_2 = exp(Float64(-t_0))
	return Float64(-Float64(Float64(1.0 / Float64(pi / 4.0)) * log(Float64(Float64(t_1 + t_2) / Float64(t_1 - t_2)))))
end

MATLAB

function tmp = code(f)
	t_0 = (pi / 4.0) * f;
	t_1 = exp(t_0);
	t_2 = exp(-t_0);
	tmp = -((1.0 / (pi / 4.0)) * log(((t_1 + t_2) / (t_1 - t_2))));
end

Wolfram

code[f_] := Block[{t$95$0 = N[(N[(Pi / 4.0), $MachinePrecision] * f), $MachinePrecision]}, Block[{t$95$1 = N[Exp[t$95$0], $MachinePrecision]}, Block[{t$95$2 = N[Exp[(-t$95$0)], $MachinePrecision]}, (-N[(N[(1.0 / N[(Pi / 4.0), $MachinePrecision]), $MachinePrecision] * N[Log[N[(N[(t$95$1 + t$95$2), $MachinePrecision] / N[(t$95$1 - t$95$2), $MachinePrecision]), $MachinePrecision]], $MachinePrecision]), $MachinePrecision])]]]

Initial Program: 7.0% accurate, 1.0× speedup

Math

\[\begin{array}{l} \\ \begin{array}{l} t_0 := \frac{\pi}{4} \cdot f\\ t_1 := e^{t\_0}\\ t_2 := e^{-t\_0}\\ -\frac{1}{\frac{\pi}{4}} \cdot \log \left(\frac{t\_1 + t\_2}{t\_1 - t\_2}\right) \end{array} \end{array} \]

FPCore

(FPCore (f)
 :precision binary64
 (let* ((t_0 (* (/ PI 4.0) f)) (t_1 (exp t_0)) (t_2 (exp (- t_0))))
   (- (* (/ 1.0 (/ PI 4.0)) (log (/ (+ t_1 t_2) (- t_1 t_2)))))))

C

double code(double f) {
	double t_0 = (((double) M_PI) / 4.0) * f;
	double t_1 = exp(t_0);
	double t_2 = exp(-t_0);
	return -((1.0 / (((double) M_PI) / 4.0)) * log(((t_1 + t_2) / (t_1 - t_2))));
}

Java

public static double code(double f) {
	double t_0 = (Math.PI / 4.0) * f;
	double t_1 = Math.exp(t_0);
	double t_2 = Math.exp(-t_0);
	return -((1.0 / (Math.PI / 4.0)) * Math.log(((t_1 + t_2) / (t_1 - t_2))));
}

Python

def code(f):
	t_0 = (math.pi / 4.0) * f
	t_1 = math.exp(t_0)
	t_2 = math.exp(-t_0)
	return -((1.0 / (math.pi / 4.0)) * math.log(((t_1 + t_2) / (t_1 - t_2))))

Julia

function code(f)
	t_0 = Float64(Float64(pi / 4.0) * f)
	t_1 = exp(t_0)
	t_2 = exp(Float64(-t_0))
	return Float64(-Float64(Float64(1.0 / Float64(pi / 4.0)) * log(Float64(Float64(t_1 + t_2) / Float64(t_1 - t_2)))))
end

MATLAB

function tmp = code(f)
	t_0 = (pi / 4.0) * f;
	t_1 = exp(t_0);
	t_2 = exp(-t_0);
	tmp = -((1.0 / (pi / 4.0)) * log(((t_1 + t_2) / (t_1 - t_2))));
end

Wolfram

code[f_] := Block[{t$95$0 = N[(N[(Pi / 4.0), $MachinePrecision] * f), $MachinePrecision]}, Block[{t$95$1 = N[Exp[t$95$0], $MachinePrecision]}, Block[{t$95$2 = N[Exp[(-t$95$0)], $MachinePrecision]}, (-N[(N[(1.0 / N[(Pi / 4.0), $MachinePrecision]), $MachinePrecision] * N[Log[N[(N[(t$95$1 + t$95$2), $MachinePrecision] / N[(t$95$1 - t$95$2), $MachinePrecision]), $MachinePrecision]], $MachinePrecision]), $MachinePrecision])]]]

Alternative 1: 98.9% accurate, 1.7× speedup

Math

\[\begin{array}{l} \\ \frac{\log \left(\frac{1}{\mathsf{expm1}\left(\pi \cdot \left(f \cdot 0.5\right)\right)} + \frac{-1}{\mathsf{expm1}\left(\pi \cdot \left(f \cdot -0.5\right)\right)}\right) \cdot -4}{\pi} \end{array} \]

FPCore

(FPCore (f)
 :precision binary64
 (/
  (*
   (log
    (+ (/ 1.0 (expm1 (* PI (* f 0.5)))) (/ -1.0 (expm1 (* PI (* f -0.5))))))
   -4.0)
  PI))

C

double code(double f) {
	return (log(((1.0 / expm1((((double) M_PI) * (f * 0.5)))) + (-1.0 / expm1((((double) M_PI) * (f * -0.5)))))) * -4.0) / ((double) M_PI);
}

Java

public static double code(double f) {
	return (Math.log(((1.0 / Math.expm1((Math.PI * (f * 0.5)))) + (-1.0 / Math.expm1((Math.PI * (f * -0.5)))))) * -4.0) / Math.PI;
}

Python

def code(f):
	return (math.log(((1.0 / math.expm1((math.pi * (f * 0.5)))) + (-1.0 / math.expm1((math.pi * (f * -0.5)))))) * -4.0) / math.pi

Julia

function code(f)
	return Float64(Float64(log(Float64(Float64(1.0 / expm1(Float64(pi * Float64(f * 0.5)))) + Float64(-1.0 / expm1(Float64(pi * Float64(f * -0.5)))))) * -4.0) / pi)
end

Wolfram

code[f_] := N[(N[(N[Log[N[(N[(1.0 / N[(Exp[N[(Pi * N[(f * 0.5), $MachinePrecision]), $MachinePrecision]] - 1), $MachinePrecision]), $MachinePrecision] + N[(-1.0 / N[(Exp[N[(Pi * N[(f * -0.5), $MachinePrecision]), $MachinePrecision]] - 1), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]], $MachinePrecision] * -4.0), $MachinePrecision] / Pi), $MachinePrecision]

Derivation

1. Initial program 5.8%

   \[-\frac{1}{\frac{\pi}{4}} \cdot \log \left(\frac{e^{\frac{\pi}{4} \cdot f} + e^{-\frac{\pi}{4} \cdot f}}{e^{\frac{\pi}{4} \cdot f} - e^{-\frac{\pi}{4} \cdot f}}\right) \]

3. Simplified 98.1%

   \[\leadsto \color{blue}{\log \left(\frac{1}{\mathsf{expm1}\left(\pi \cdot \left(f \cdot 0.5\right)\right)} + \frac{-1}{\mathsf{expm1}\left(\pi \cdot \left(f \cdot -0.5\right)\right)}\right) \cdot \frac{-4}{\pi}} \]
4. Add Preprocessing

5. Taylor expanded in f around inf 4.5%

   \[\leadsto \color{blue}{-4 \cdot \frac{\log \left(\frac{1}{e^{0.5 \cdot \left(f \cdot \pi\right)} - 1} - \frac{1}{e^{-0.5 \cdot \left(f \cdot \pi\right)} - 1}\right)}{\pi}} \]
6. Step-by-step derivation

   1. *-commutative 4.5%

      \[\leadsto \color{blue}{\frac{\log \left(\frac{1}{e^{0.5 \cdot \left(f \cdot \pi\right)} - 1} - \frac{1}{e^{-0.5 \cdot \left(f \cdot \pi\right)} - 1}\right)}{\pi} \cdot -4} \]

   2. associate-*l/ 4.5%

      \[\leadsto \color{blue}{\frac{\log \left(\frac{1}{e^{0.5 \cdot \left(f \cdot \pi\right)} - 1} - \frac{1}{e^{-0.5 \cdot \left(f \cdot \pi\right)} - 1}\right) \cdot -4}{\pi}} \]

7. Simplified 98.4%

   \[\leadsto \color{blue}{\frac{\log \left(\frac{1}{\mathsf{expm1}\left(\pi \cdot \left(f \cdot 0.5\right)\right)} + \frac{-1}{\mathsf{expm1}\left(\pi \cdot \left(f \cdot -0.5\right)\right)}\right) \cdot -4}{\pi}} \]
8. Add Preprocessing

Alternative 2: 98.7% accurate, 1.7× speedup

Math

\[\begin{array}{l} \\ \log \left(\frac{1}{\mathsf{expm1}\left(\pi \cdot \left(f \cdot 0.5\right)\right)} + \frac{-1}{\mathsf{expm1}\left(\pi \cdot \left(f \cdot -0.5\right)\right)}\right) \cdot \frac{-4}{\pi} \end{array} \]

FPCore

(FPCore (f)
 :precision binary64
 (*
  (log (+ (/ 1.0 (expm1 (* PI (* f 0.5)))) (/ -1.0 (expm1 (* PI (* f -0.5))))))
  (/ -4.0 PI)))

C

double code(double f) {
	return log(((1.0 / expm1((((double) M_PI) * (f * 0.5)))) + (-1.0 / expm1((((double) M_PI) * (f * -0.5)))))) * (-4.0 / ((double) M_PI));
}

Java

public static double code(double f) {
	return Math.log(((1.0 / Math.expm1((Math.PI * (f * 0.5)))) + (-1.0 / Math.expm1((Math.PI * (f * -0.5)))))) * (-4.0 / Math.PI);
}

Python

def code(f):
	return math.log(((1.0 / math.expm1((math.pi * (f * 0.5)))) + (-1.0 / math.expm1((math.pi * (f * -0.5)))))) * (-4.0 / math.pi)

Julia

function code(f)
	return Float64(log(Float64(Float64(1.0 / expm1(Float64(pi * Float64(f * 0.5)))) + Float64(-1.0 / expm1(Float64(pi * Float64(f * -0.5)))))) * Float64(-4.0 / pi))
end

Wolfram

code[f_] := N[(N[Log[N[(N[(1.0 / N[(Exp[N[(Pi * N[(f * 0.5), $MachinePrecision]), $MachinePrecision]] - 1), $MachinePrecision]), $MachinePrecision] + N[(-1.0 / N[(Exp[N[(Pi * N[(f * -0.5), $MachinePrecision]), $MachinePrecision]] - 1), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]], $MachinePrecision] * N[(-4.0 / Pi), $MachinePrecision]), $MachinePrecision]

Derivation

1. Initial program 5.8%

   \[-\frac{1}{\frac{\pi}{4}} \cdot \log \left(\frac{e^{\frac{\pi}{4} \cdot f} + e^{-\frac{\pi}{4} \cdot f}}{e^{\frac{\pi}{4} \cdot f} - e^{-\frac{\pi}{4} \cdot f}}\right) \]

3. Simplified 98.1%

   \[\leadsto \color{blue}{\log \left(\frac{1}{\mathsf{expm1}\left(\pi \cdot \left(f \cdot 0.5\right)\right)} + \frac{-1}{\mathsf{expm1}\left(\pi \cdot \left(f \cdot -0.5\right)\right)}\right) \cdot \frac{-4}{\pi}} \]
4. Add Preprocessing
5. Add Preprocessing

Alternative 3: 96.4% accurate, 2.3× speedup

Math

\[\begin{array}{l} \\ \frac{-4}{\pi} \cdot \log \left(\frac{-1}{\mathsf{expm1}\left(\pi \cdot \left(f \cdot -0.5\right)\right)} + \frac{2 \cdot \frac{1}{\pi} - f \cdot \left(0.5 + f \cdot \left(\pi \cdot -0.041666666666666664\right)\right)}{f}\right) \end{array} \]

FPCore

(FPCore (f)
 :precision binary64
 (*
  (/ -4.0 PI)
  (log
   (+
    (/ -1.0 (expm1 (* PI (* f -0.5))))
    (/
     (- (* 2.0 (/ 1.0 PI)) (* f (+ 0.5 (* f (* PI -0.041666666666666664)))))
     f)))))

C

double code(double f) {
	return (-4.0 / ((double) M_PI)) * log(((-1.0 / expm1((((double) M_PI) * (f * -0.5)))) + (((2.0 * (1.0 / ((double) M_PI))) - (f * (0.5 + (f * (((double) M_PI) * -0.041666666666666664))))) / f)));
}

Java

public static double code(double f) {
	return (-4.0 / Math.PI) * Math.log(((-1.0 / Math.expm1((Math.PI * (f * -0.5)))) + (((2.0 * (1.0 / Math.PI)) - (f * (0.5 + (f * (Math.PI * -0.041666666666666664))))) / f)));
}

Python

def code(f):
	return (-4.0 / math.pi) * math.log(((-1.0 / math.expm1((math.pi * (f * -0.5)))) + (((2.0 * (1.0 / math.pi)) - (f * (0.5 + (f * (math.pi * -0.041666666666666664))))) / f)))

Julia

function code(f)
	return Float64(Float64(-4.0 / pi) * log(Float64(Float64(-1.0 / expm1(Float64(pi * Float64(f * -0.5)))) + Float64(Float64(Float64(2.0 * Float64(1.0 / pi)) - Float64(f * Float64(0.5 + Float64(f * Float64(pi * -0.041666666666666664))))) / f))))
end

Wolfram

code[f_] := N[(N[(-4.0 / Pi), $MachinePrecision] * N[Log[N[(N[(-1.0 / N[(Exp[N[(Pi * N[(f * -0.5), $MachinePrecision]), $MachinePrecision]] - 1), $MachinePrecision]), $MachinePrecision] + N[(N[(N[(2.0 * N[(1.0 / Pi), $MachinePrecision]), $MachinePrecision] - N[(f * N[(0.5 + N[(f * N[(Pi * -0.041666666666666664), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]), $MachinePrecision] / f), $MachinePrecision]), $MachinePrecision]], $MachinePrecision]), $MachinePrecision]

Derivation

1. Initial program 5.8%

   \[-\frac{1}{\frac{\pi}{4}} \cdot \log \left(\frac{e^{\frac{\pi}{4} \cdot f} + e^{-\frac{\pi}{4} \cdot f}}{e^{\frac{\pi}{4} \cdot f} - e^{-\frac{\pi}{4} \cdot f}}\right) \]

3. Simplified 98.1%

   \[\leadsto \color{blue}{\log \left(\frac{1}{\mathsf{expm1}\left(\pi \cdot \left(f \cdot 0.5\right)\right)} + \frac{-1}{\mathsf{expm1}\left(\pi \cdot \left(f \cdot -0.5\right)\right)}\right) \cdot \frac{-4}{\pi}} \]
4. Add Preprocessing

5. Taylor expanded in f around 0 96.2%

   \[\leadsto \log \left(\color{blue}{\frac{f \cdot \left(-1 \cdot \left(f \cdot \left(-0.125 \cdot \pi + 0.08333333333333333 \cdot \pi\right)\right) - 0.5\right) + 2 \cdot \frac{1}{\pi}}{f}} + \frac{-1}{\mathsf{expm1}\left(\pi \cdot \left(f \cdot -0.5\right)\right)}\right) \cdot \frac{-4}{\pi} \]
6. Step-by-step derivation

   1. *-commutative 96.2%

      \[\leadsto \log \left(\frac{f \cdot \left(-1 \cdot \color{blue}{\left(\left(-0.125 \cdot \pi + 0.08333333333333333 \cdot \pi\right) \cdot f\right)} - 0.5\right) + 2 \cdot \frac{1}{\pi}}{f} + \frac{-1}{\mathsf{expm1}\left(\pi \cdot \left(f \cdot -0.5\right)\right)}\right) \cdot \frac{-4}{\pi} \]

   2. distribute-rgt-out 96.2%

      \[\leadsto \log \left(\frac{f \cdot \left(-1 \cdot \left(\color{blue}{\left(\pi \cdot \left(-0.125 + 0.08333333333333333\right)\right)} \cdot f\right) - 0.5\right) + 2 \cdot \frac{1}{\pi}}{f} + \frac{-1}{\mathsf{expm1}\left(\pi \cdot \left(f \cdot -0.5\right)\right)}\right) \cdot \frac{-4}{\pi} \]

   3. metadata-eval 96.2%

      \[\leadsto \log \left(\frac{f \cdot \left(-1 \cdot \left(\left(\pi \cdot \color{blue}{-0.041666666666666664}\right) \cdot f\right) - 0.5\right) + 2 \cdot \frac{1}{\pi}}{f} + \frac{-1}{\mathsf{expm1}\left(\pi \cdot \left(f \cdot -0.5\right)\right)}\right) \cdot \frac{-4}{\pi} \]

7. Applied egg-rr 96.2%

   \[\leadsto \log \left(\frac{f \cdot \left(-1 \cdot \color{blue}{\left(\left(\pi \cdot -0.041666666666666664\right) \cdot f\right)} - 0.5\right) + 2 \cdot \frac{1}{\pi}}{f} + \frac{-1}{\mathsf{expm1}\left(\pi \cdot \left(f \cdot -0.5\right)\right)}\right) \cdot \frac{-4}{\pi} \]

8. Final simplification 96.2%

   \[\leadsto \frac{-4}{\pi} \cdot \log \left(\frac{-1}{\mathsf{expm1}\left(\pi \cdot \left(f \cdot -0.5\right)\right)} + \frac{2 \cdot \frac{1}{\pi} - f \cdot \left(0.5 + f \cdot \left(\pi \cdot -0.041666666666666664\right)\right)}{f}\right) \]
9. Add Preprocessing

Alternative 4: 96.4% accurate, 3.6× speedup

Math

\[\begin{array}{l} \\ \begin{array}{l} t_0 := f \cdot \left(\pi \cdot -0.125 + \pi \cdot 0.08333333333333333\right)\\ t_1 := 2 \cdot \frac{1}{\pi}\\ \frac{-4}{\pi} \cdot \log \left(\frac{t\_1 - f \cdot \left(0.5 + t\_0\right)}{f} + \frac{t\_1 + f \cdot \left(0.5 - t\_0\right)}{f}\right) \end{array} \end{array} \]

FPCore

(FPCore (f)
 :precision binary64
 (let* ((t_0 (* f (+ (* PI -0.125) (* PI 0.08333333333333333))))
        (t_1 (* 2.0 (/ 1.0 PI))))
   (*
    (/ -4.0 PI)
    (log
     (+ (/ (- t_1 (* f (+ 0.5 t_0))) f) (/ (+ t_1 (* f (- 0.5 t_0))) f))))))

C

double code(double f) {
	double t_0 = f * ((((double) M_PI) * -0.125) + (((double) M_PI) * 0.08333333333333333));
	double t_1 = 2.0 * (1.0 / ((double) M_PI));
	return (-4.0 / ((double) M_PI)) * log((((t_1 - (f * (0.5 + t_0))) / f) + ((t_1 + (f * (0.5 - t_0))) / f)));
}

Java

public static double code(double f) {
	double t_0 = f * ((Math.PI * -0.125) + (Math.PI * 0.08333333333333333));
	double t_1 = 2.0 * (1.0 / Math.PI);
	return (-4.0 / Math.PI) * Math.log((((t_1 - (f * (0.5 + t_0))) / f) + ((t_1 + (f * (0.5 - t_0))) / f)));
}

Python

def code(f):
	t_0 = f * ((math.pi * -0.125) + (math.pi * 0.08333333333333333))
	t_1 = 2.0 * (1.0 / math.pi)
	return (-4.0 / math.pi) * math.log((((t_1 - (f * (0.5 + t_0))) / f) + ((t_1 + (f * (0.5 - t_0))) / f)))

Julia

function code(f)
	t_0 = Float64(f * Float64(Float64(pi * -0.125) + Float64(pi * 0.08333333333333333)))
	t_1 = Float64(2.0 * Float64(1.0 / pi))
	return Float64(Float64(-4.0 / pi) * log(Float64(Float64(Float64(t_1 - Float64(f * Float64(0.5 + t_0))) / f) + Float64(Float64(t_1 + Float64(f * Float64(0.5 - t_0))) / f))))
end

MATLAB

function tmp = code(f)
	t_0 = f * ((pi * -0.125) + (pi * 0.08333333333333333));
	t_1 = 2.0 * (1.0 / pi);
	tmp = (-4.0 / pi) * log((((t_1 - (f * (0.5 + t_0))) / f) + ((t_1 + (f * (0.5 - t_0))) / f)));
end

Wolfram

code[f_] := Block[{t$95$0 = N[(f * N[(N[(Pi * -0.125), $MachinePrecision] + N[(Pi * 0.08333333333333333), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]}, Block[{t$95$1 = N[(2.0 * N[(1.0 / Pi), $MachinePrecision]), $MachinePrecision]}, N[(N[(-4.0 / Pi), $MachinePrecision] * N[Log[N[(N[(N[(t$95$1 - N[(f * N[(0.5 + t$95$0), $MachinePrecision]), $MachinePrecision]), $MachinePrecision] / f), $MachinePrecision] + N[(N[(t$95$1 + N[(f * N[(0.5 - t$95$0), $MachinePrecision]), $MachinePrecision]), $MachinePrecision] / f), $MachinePrecision]), $MachinePrecision]], $MachinePrecision]), $MachinePrecision]]]

Derivation

1. Initial program 5.8%

   \[-\frac{1}{\frac{\pi}{4}} \cdot \log \left(\frac{e^{\frac{\pi}{4} \cdot f} + e^{-\frac{\pi}{4} \cdot f}}{e^{\frac{\pi}{4} \cdot f} - e^{-\frac{\pi}{4} \cdot f}}\right) \]

3. Simplified 98.1%

   \[\leadsto \color{blue}{\log \left(\frac{1}{\mathsf{expm1}\left(\pi \cdot \left(f \cdot 0.5\right)\right)} + \frac{-1}{\mathsf{expm1}\left(\pi \cdot \left(f \cdot -0.5\right)\right)}\right) \cdot \frac{-4}{\pi}} \]
4. Add Preprocessing

5. Taylor expanded in f around 0 96.2%

   \[\leadsto \log \left(\color{blue}{\frac{f \cdot \left(-1 \cdot \left(f \cdot \left(-0.125 \cdot \pi + 0.08333333333333333 \cdot \pi\right)\right) - 0.5\right) + 2 \cdot \frac{1}{\pi}}{f}} + \frac{-1}{\mathsf{expm1}\left(\pi \cdot \left(f \cdot -0.5\right)\right)}\right) \cdot \frac{-4}{\pi} \]

6. Taylor expanded in f around 0 96.2%

   \[\leadsto \log \left(\frac{f \cdot \left(-1 \cdot \left(f \cdot \left(-0.125 \cdot \pi + 0.08333333333333333 \cdot \pi\right)\right) - 0.5\right) + 2 \cdot \frac{1}{\pi}}{f} + \color{blue}{\frac{f \cdot \left(0.5 + -1 \cdot \left(f \cdot \left(-0.125 \cdot \pi + 0.08333333333333333 \cdot \pi\right)\right)\right) + 2 \cdot \frac{1}{\pi}}{f}}\right) \cdot \frac{-4}{\pi} \]

7. Final simplification 96.2%

   \[\leadsto \frac{-4}{\pi} \cdot \log \left(\frac{2 \cdot \frac{1}{\pi} - f \cdot \left(0.5 + f \cdot \left(\pi \cdot -0.125 + \pi \cdot 0.08333333333333333\right)\right)}{f} + \frac{2 \cdot \frac{1}{\pi} + f \cdot \left(0.5 - f \cdot \left(\pi \cdot -0.125 + \pi \cdot 0.08333333333333333\right)\right)}{f}\right) \]
8. Add Preprocessing

Alternative 5: 95.9% accurate, 4.7× speedup

Math

\[\begin{array}{l} \\ -4 \cdot \left(-1 + \left(1 + \frac{\log \left(\frac{4}{\pi \cdot f}\right)}{\pi}\right)\right) \end{array} \]

FPCore

(FPCore (f)
 :precision binary64
 (* -4.0 (+ -1.0 (+ 1.0 (/ (log (/ 4.0 (* PI f))) PI)))))

C

double code(double f) {
	return -4.0 * (-1.0 + (1.0 + (log((4.0 / (((double) M_PI) * f))) / ((double) M_PI))));
}

Java

public static double code(double f) {
	return -4.0 * (-1.0 + (1.0 + (Math.log((4.0 / (Math.PI * f))) / Math.PI)));
}

Python

def code(f):
	return -4.0 * (-1.0 + (1.0 + (math.log((4.0 / (math.pi * f))) / math.pi)))

Julia

function code(f)
	return Float64(-4.0 * Float64(-1.0 + Float64(1.0 + Float64(log(Float64(4.0 / Float64(pi * f))) / pi))))
end

MATLAB

function tmp = code(f)
	tmp = -4.0 * (-1.0 + (1.0 + (log((4.0 / (pi * f))) / pi)));
end

Wolfram

code[f_] := N[(-4.0 * N[(-1.0 + N[(1.0 + N[(N[Log[N[(4.0 / N[(Pi * f), $MachinePrecision]), $MachinePrecision]], $MachinePrecision] / Pi), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]

Derivation

1. Initial program 5.8%

   \[-\frac{1}{\frac{\pi}{4}} \cdot \log \left(\frac{e^{\frac{\pi}{4} \cdot f} + e^{-\frac{\pi}{4} \cdot f}}{e^{\frac{\pi}{4} \cdot f} - e^{-\frac{\pi}{4} \cdot f}}\right) \]

3. Simplified 98.1%

   \[\leadsto \color{blue}{\log \left(\frac{1}{\mathsf{expm1}\left(\pi \cdot \left(f \cdot 0.5\right)\right)} + \frac{-1}{\mathsf{expm1}\left(\pi \cdot \left(f \cdot -0.5\right)\right)}\right) \cdot \frac{-4}{\pi}} \]
4. Add Preprocessing

5. Taylor expanded in f around 0 96.0%

   \[\leadsto \color{blue}{-4 \cdot \frac{\log \left(\frac{4}{\pi}\right) + -1 \cdot \log f}{\pi}} \]
6. Step-by-step derivation

   1. mul-1-neg 96.0%

      \[\leadsto -4 \cdot \frac{\log \left(\frac{4}{\pi}\right) + \color{blue}{\left(-\log f\right)}}{\pi} \]

   2. unsub-neg 96.0%

      \[\leadsto -4 \cdot \frac{\color{blue}{\log \left(\frac{4}{\pi}\right) - \log f}}{\pi} \]

7. Simplified 96.0%

   \[\leadsto \color{blue}{-4 \cdot \frac{\log \left(\frac{4}{\pi}\right) - \log f}{\pi}} \]
8. Step-by-step derivation

   1. diff-log 96.0%

      \[\leadsto -4 \cdot \frac{\color{blue}{\log \left(\frac{\frac{4}{\pi}}{f}\right)}}{\pi} \]

9. Applied egg-rr 96.0%

   \[\leadsto -4 \cdot \color{blue}{\frac{\log \left(\frac{\frac{4}{\pi}}{f}\right)}{\pi}} \]
10. Step-by-step derivation

    1. expm1-log1p-u 94.6%

       \[\leadsto -4 \cdot \color{blue}{\mathsf{expm1}\left(\mathsf{log1p}\left(\frac{\log \left(\frac{\frac{4}{\pi}}{f}\right)}{\pi}\right)\right)} \]

    2. expm1-undefine 94.7%

       \[\leadsto -4 \cdot \color{blue}{\left(e^{\mathsf{log1p}\left(\frac{\log \left(\frac{\frac{4}{\pi}}{f}\right)}{\pi}\right)} - 1\right)} \]

    3. log1p-undefine 94.6%

       \[\leadsto -4 \cdot \left(e^{\color{blue}{\log \left(1 + \frac{\log \left(\frac{\frac{4}{\pi}}{f}\right)}{\pi}\right)}} - 1\right) \]

    4. rem-exp-log 96.0%

       \[\leadsto -4 \cdot \left(\color{blue}{\left(1 + \frac{\log \left(\frac{\frac{4}{\pi}}{f}\right)}{\pi}\right)} - 1\right) \]

    5. +-commutative 96.0%

       \[\leadsto -4 \cdot \left(\color{blue}{\left(\frac{\log \left(\frac{\frac{4}{\pi}}{f}\right)}{\pi} + 1\right)} - 1\right) \]

    6. associate-/r* 96.0%

       \[\leadsto -4 \cdot \left(\left(\frac{\log \color{blue}{\left(\frac{4}{\pi \cdot f}\right)}}{\pi} + 1\right) - 1\right) \]

11. Applied egg-rr 96.0%

    \[\leadsto -4 \cdot \color{blue}{\left(\left(\frac{\log \left(\frac{4}{\pi \cdot f}\right)}{\pi} + 1\right) - 1\right)} \]

12. Final simplification 96.0%

    \[\leadsto -4 \cdot \left(-1 + \left(1 + \frac{\log \left(\frac{4}{\pi \cdot f}\right)}{\pi}\right)\right) \]
13. Add Preprocessing

Alternative 6: 95.9% accurate, 4.8× speedup

Math

\[\begin{array}{l} \\ -4 \cdot \frac{\log \left(f \cdot \left(\pi \cdot 0.25\right)\right)}{-\pi} \end{array} \]

FPCore

(FPCore (f) :precision binary64 (* -4.0 (/ (log (* f (* PI 0.25))) (- PI))))

C

double code(double f) {
	return -4.0 * (log((f * (((double) M_PI) * 0.25))) / -((double) M_PI));
}

Java

public static double code(double f) {
	return -4.0 * (Math.log((f * (Math.PI * 0.25))) / -Math.PI);
}

Python

def code(f):
	return -4.0 * (math.log((f * (math.pi * 0.25))) / -math.pi)

Julia

function code(f)
	return Float64(-4.0 * Float64(log(Float64(f * Float64(pi * 0.25))) / Float64(-pi)))
end

MATLAB

function tmp = code(f)
	tmp = -4.0 * (log((f * (pi * 0.25))) / -pi);
end

Wolfram

code[f_] := N[(-4.0 * N[(N[Log[N[(f * N[(Pi * 0.25), $MachinePrecision]), $MachinePrecision]], $MachinePrecision] / (-Pi)), $MachinePrecision]), $MachinePrecision]

Derivation

1. Initial program 5.8%

   \[-\frac{1}{\frac{\pi}{4}} \cdot \log \left(\frac{e^{\frac{\pi}{4} \cdot f} + e^{-\frac{\pi}{4} \cdot f}}{e^{\frac{\pi}{4} \cdot f} - e^{-\frac{\pi}{4} \cdot f}}\right) \]

3. Simplified 98.1%

   \[\leadsto \color{blue}{\log \left(\frac{1}{\mathsf{expm1}\left(\pi \cdot \left(f \cdot 0.5\right)\right)} + \frac{-1}{\mathsf{expm1}\left(\pi \cdot \left(f \cdot -0.5\right)\right)}\right) \cdot \frac{-4}{\pi}} \]
4. Add Preprocessing

5. Taylor expanded in f around 0 96.0%

   \[\leadsto \color{blue}{-4 \cdot \frac{\log \left(\frac{4}{\pi}\right) + -1 \cdot \log f}{\pi}} \]
6. Step-by-step derivation

   1. mul-1-neg 96.0%

      \[\leadsto -4 \cdot \frac{\log \left(\frac{4}{\pi}\right) + \color{blue}{\left(-\log f\right)}}{\pi} \]

   2. unsub-neg 96.0%

      \[\leadsto -4 \cdot \frac{\color{blue}{\log \left(\frac{4}{\pi}\right) - \log f}}{\pi} \]

7. Simplified 96.0%

   \[\leadsto \color{blue}{-4 \cdot \frac{\log \left(\frac{4}{\pi}\right) - \log f}{\pi}} \]
8. Step-by-step derivation

   1. add-log-exp 96.0%

      \[\leadsto -4 \cdot \color{blue}{\log \left(e^{\frac{\log \left(\frac{4}{\pi}\right) - \log f}{\pi}}\right)} \]

   2. div-inv 95.9%

      \[\leadsto -4 \cdot \log \left(e^{\color{blue}{\left(\log \left(\frac{4}{\pi}\right) - \log f\right) \cdot \frac{1}{\pi}}}\right) \]

   3. diff-log 95.8%

      \[\leadsto -4 \cdot \log \left(e^{\color{blue}{\log \left(\frac{\frac{4}{\pi}}{f}\right)} \cdot \frac{1}{\pi}}\right) \]

   4. exp-to-pow 95.8%

      \[\leadsto -4 \cdot \log \color{blue}{\left({\left(\frac{\frac{4}{\pi}}{f}\right)}^{\left(\frac{1}{\pi}\right)}\right)} \]

9. Applied egg-rr 95.8%

   \[\leadsto -4 \cdot \color{blue}{\log \left({\left(\frac{\frac{4}{\pi}}{f}\right)}^{\left(\frac{1}{\pi}\right)}\right)} \]
10. Step-by-step derivation

    1. pow-to-exp 95.8%

       \[\leadsto -4 \cdot \log \color{blue}{\left(e^{\log \left(\frac{\frac{4}{\pi}}{f}\right) \cdot \frac{1}{\pi}}\right)} \]

    2. div-inv 96.0%

       \[\leadsto -4 \cdot \log \left(e^{\color{blue}{\frac{\log \left(\frac{\frac{4}{\pi}}{f}\right)}{\pi}}}\right) \]

    3. frac-2neg 96.0%

       \[\leadsto -4 \cdot \log \left(e^{\color{blue}{\frac{-\log \left(\frac{\frac{4}{\pi}}{f}\right)}{-\pi}}}\right) \]

    4. add-log-exp 96.0%

       \[\leadsto -4 \cdot \color{blue}{\frac{-\log \left(\frac{\frac{4}{\pi}}{f}\right)}{-\pi}} \]

    5. neg-log 96.0%

       \[\leadsto -4 \cdot \frac{\color{blue}{\log \left(\frac{1}{\frac{\frac{4}{\pi}}{f}}\right)}}{-\pi} \]

    6. clear-num 96.0%

       \[\leadsto -4 \cdot \frac{\log \color{blue}{\left(\frac{f}{\frac{4}{\pi}}\right)}}{-\pi} \]

    7. div-inv 96.0%

       \[\leadsto -4 \cdot \frac{\log \color{blue}{\left(f \cdot \frac{1}{\frac{4}{\pi}}\right)}}{-\pi} \]

    8. clear-num 96.0%

       \[\leadsto -4 \cdot \frac{\log \left(f \cdot \color{blue}{\frac{\pi}{4}}\right)}{-\pi} \]

    9. div-inv 96.0%

       \[\leadsto -4 \cdot \frac{\log \left(f \cdot \color{blue}{\left(\pi \cdot \frac{1}{4}\right)}\right)}{-\pi} \]

    10. metadata-eval 96.0%

        \[\leadsto -4 \cdot \frac{\log \left(f \cdot \left(\pi \cdot \color{blue}{0.25}\right)\right)}{-\pi} \]

11. Applied egg-rr 96.0%

    \[\leadsto -4 \cdot \color{blue}{\frac{\log \left(f \cdot \left(\pi \cdot 0.25\right)\right)}{-\pi}} \]
12. Add Preprocessing

Alternative 7: 95.9% accurate, 4.9× speedup

Math

\[\begin{array}{l} \\ -4 \cdot \frac{\log \left(\frac{\frac{4}{\pi}}{f}\right)}{\pi} \end{array} \]

FPCore

(FPCore (f) :precision binary64 (* -4.0 (/ (log (/ (/ 4.0 PI) f)) PI)))

C

double code(double f) {
	return -4.0 * (log(((4.0 / ((double) M_PI)) / f)) / ((double) M_PI));
}

Java

public static double code(double f) {
	return -4.0 * (Math.log(((4.0 / Math.PI) / f)) / Math.PI);
}

Python

def code(f):
	return -4.0 * (math.log(((4.0 / math.pi) / f)) / math.pi)

Julia

function code(f)
	return Float64(-4.0 * Float64(log(Float64(Float64(4.0 / pi) / f)) / pi))
end

MATLAB

function tmp = code(f)
	tmp = -4.0 * (log(((4.0 / pi) / f)) / pi);
end

Wolfram

code[f_] := N[(-4.0 * N[(N[Log[N[(N[(4.0 / Pi), $MachinePrecision] / f), $MachinePrecision]], $MachinePrecision] / Pi), $MachinePrecision]), $MachinePrecision]

Derivation

1. Initial program 5.8%

   \[-\frac{1}{\frac{\pi}{4}} \cdot \log \left(\frac{e^{\frac{\pi}{4} \cdot f} + e^{-\frac{\pi}{4} \cdot f}}{e^{\frac{\pi}{4} \cdot f} - e^{-\frac{\pi}{4} \cdot f}}\right) \]

3. Simplified 98.1%

   \[\leadsto \color{blue}{\log \left(\frac{1}{\mathsf{expm1}\left(\pi \cdot \left(f \cdot 0.5\right)\right)} + \frac{-1}{\mathsf{expm1}\left(\pi \cdot \left(f \cdot -0.5\right)\right)}\right) \cdot \frac{-4}{\pi}} \]
4. Add Preprocessing

5. Taylor expanded in f around 0 96.0%

   \[\leadsto \color{blue}{-4 \cdot \frac{\log \left(\frac{4}{\pi}\right) + -1 \cdot \log f}{\pi}} \]
6. Step-by-step derivation

   1. mul-1-neg 96.0%

      \[\leadsto -4 \cdot \frac{\log \left(\frac{4}{\pi}\right) + \color{blue}{\left(-\log f\right)}}{\pi} \]

   2. unsub-neg 96.0%

      \[\leadsto -4 \cdot \frac{\color{blue}{\log \left(\frac{4}{\pi}\right) - \log f}}{\pi} \]

7. Simplified 96.0%

   \[\leadsto \color{blue}{-4 \cdot \frac{\log \left(\frac{4}{\pi}\right) - \log f}{\pi}} \]
8. Step-by-step derivation

   1. diff-log 96.0%

      \[\leadsto -4 \cdot \frac{\color{blue}{\log \left(\frac{\frac{4}{\pi}}{f}\right)}}{\pi} \]

9. Applied egg-rr 96.0%

   \[\leadsto -4 \cdot \color{blue}{\frac{\log \left(\frac{\frac{4}{\pi}}{f}\right)}{\pi}} \]
10. Add Preprocessing

Alternative 8: 4.9% accurate, 106.4× speedup

Math

\[\begin{array}{l} \\ \frac{-4}{\pi} \cdot 0 \end{array} \]

FPCore

(FPCore (f) :precision binary64 (* (/ -4.0 PI) 0.0))

C

double code(double f) {
	return (-4.0 / ((double) M_PI)) * 0.0;
}

Java

public static double code(double f) {
	return (-4.0 / Math.PI) * 0.0;
}

Python

def code(f):
	return (-4.0 / math.pi) * 0.0

Julia

function code(f)
	return Float64(Float64(-4.0 / pi) * 0.0)
end

MATLAB

function tmp = code(f)
	tmp = (-4.0 / pi) * 0.0;
end

Wolfram

code[f_] := N[(N[(-4.0 / Pi), $MachinePrecision] * 0.0), $MachinePrecision]

Derivation

1. Initial program 5.8%

   \[-\frac{1}{\frac{\pi}{4}} \cdot \log \left(\frac{e^{\frac{\pi}{4} \cdot f} + e^{-\frac{\pi}{4} \cdot f}}{e^{\frac{\pi}{4} \cdot f} - e^{-\frac{\pi}{4} \cdot f}}\right) \]

3. Simplified 98.1%

   \[\leadsto \color{blue}{\log \left(\frac{1}{\mathsf{expm1}\left(\pi \cdot \left(f \cdot 0.5\right)\right)} + \frac{-1}{\mathsf{expm1}\left(\pi \cdot \left(f \cdot -0.5\right)\right)}\right) \cdot \frac{-4}{\pi}} \]
4. Add Preprocessing

6. Applied egg-rr 94.7%

   \[\leadsto \color{blue}{{\left(\sqrt{\log \left(\frac{1}{\mathsf{expm1}\left(\pi \cdot \left(f \cdot 0.5\right)\right)} + \frac{1}{\mathsf{expm1}\left(\pi \cdot \left(f \cdot 0.5\right)\right)}\right)}\right)}^{2}} \cdot \frac{-4}{\pi} \]
7. Step-by-step derivation

   1. unpow2 94.7%

      \[\leadsto \color{blue}{\left(\sqrt{\log \left(\frac{1}{\mathsf{expm1}\left(\pi \cdot \left(f \cdot 0.5\right)\right)} + \frac{1}{\mathsf{expm1}\left(\pi \cdot \left(f \cdot 0.5\right)\right)}\right)} \cdot \sqrt{\log \left(\frac{1}{\mathsf{expm1}\left(\pi \cdot \left(f \cdot 0.5\right)\right)} + \frac{1}{\mathsf{expm1}\left(\pi \cdot \left(f \cdot 0.5\right)\right)}\right)}\right)} \cdot \frac{-4}{\pi} \]

   2. add-sqr-sqrt 95.1%

      \[\leadsto \color{blue}{\log \left(\frac{1}{\mathsf{expm1}\left(\pi \cdot \left(f \cdot 0.5\right)\right)} + \frac{1}{\mathsf{expm1}\left(\pi \cdot \left(f \cdot 0.5\right)\right)}\right)} \cdot \frac{-4}{\pi} \]

   3. flip-+ 0.0%

      \[\leadsto \log \color{blue}{\left(\frac{\frac{1}{\mathsf{expm1}\left(\pi \cdot \left(f \cdot 0.5\right)\right)} \cdot \frac{1}{\mathsf{expm1}\left(\pi \cdot \left(f \cdot 0.5\right)\right)} - \frac{1}{\mathsf{expm1}\left(\pi \cdot \left(f \cdot 0.5\right)\right)} \cdot \frac{1}{\mathsf{expm1}\left(\pi \cdot \left(f \cdot 0.5\right)\right)}}{\frac{1}{\mathsf{expm1}\left(\pi \cdot \left(f \cdot 0.5\right)\right)} - \frac{1}{\mathsf{expm1}\left(\pi \cdot \left(f \cdot 0.5\right)\right)}}\right)} \cdot \frac{-4}{\pi} \]

   4. log-div 0.0%

      \[\leadsto \color{blue}{\left(\log \left(\frac{1}{\mathsf{expm1}\left(\pi \cdot \left(f \cdot 0.5\right)\right)} \cdot \frac{1}{\mathsf{expm1}\left(\pi \cdot \left(f \cdot 0.5\right)\right)} - \frac{1}{\mathsf{expm1}\left(\pi \cdot \left(f \cdot 0.5\right)\right)} \cdot \frac{1}{\mathsf{expm1}\left(\pi \cdot \left(f \cdot 0.5\right)\right)}\right) - \log \left(\frac{1}{\mathsf{expm1}\left(\pi \cdot \left(f \cdot 0.5\right)\right)} - \frac{1}{\mathsf{expm1}\left(\pi \cdot \left(f \cdot 0.5\right)\right)}\right)\right)} \cdot \frac{-4}{\pi} \]

8. Applied egg-rr 0.0%

   \[\leadsto \color{blue}{\left(\log \left({\left(\mathsf{expm1}\left(\pi \cdot \left(f \cdot 0.5\right)\right)\right)}^{-2} - {\left(\mathsf{expm1}\left(\pi \cdot \left(f \cdot 0.5\right)\right)\right)}^{-2}\right) - \log \left(\frac{1}{\mathsf{expm1}\left(\pi \cdot \left(f \cdot 0.5\right)\right)} - \frac{1}{\mathsf{expm1}\left(\pi \cdot \left(f \cdot 0.5\right)\right)}\right)\right)} \cdot \frac{-4}{\pi} \]
9. Step-by-step derivation

   1. +-inverses 0.0%

      \[\leadsto \left(\log \left({\left(\mathsf{expm1}\left(\pi \cdot \left(f \cdot 0.5\right)\right)\right)}^{-2} - {\left(\mathsf{expm1}\left(\pi \cdot \left(f \cdot 0.5\right)\right)\right)}^{-2}\right) - \log \color{blue}{0}\right) \cdot \frac{-4}{\pi} \]

   2. +-inverses 0.0%

      \[\leadsto \left(\log \left({\left(\mathsf{expm1}\left(\pi \cdot \left(f \cdot 0.5\right)\right)\right)}^{-2} - {\left(\mathsf{expm1}\left(\pi \cdot \left(f \cdot 0.5\right)\right)\right)}^{-2}\right) - \log \color{blue}{\left({\left(\mathsf{expm1}\left(\pi \cdot \left(f \cdot 0.5\right)\right)\right)}^{-2} - {\left(\mathsf{expm1}\left(\pi \cdot \left(f \cdot 0.5\right)\right)\right)}^{-2}\right)}\right) \cdot \frac{-4}{\pi} \]

   3. +-inverses 4.6%

      \[\leadsto \color{blue}{0} \cdot \frac{-4}{\pi} \]

10. Simplified 4.6%

    \[\leadsto \color{blue}{0} \cdot \frac{-4}{\pi} \]

11. Final simplification 4.6%

    \[\leadsto \frac{-4}{\pi} \cdot 0 \]
12. Add Preprocessing

Reproduce

herbie shell --seed 2024096 
(FPCore (f)
  :name "VandenBroeck and Keller, Equation (20)"
  :precision binary64
  (- (* (/ 1.0 (/ PI 4.0)) (log (/ (+ (exp (* (/ PI 4.0) f)) (exp (- (* (/ PI 4.0) f)))) (- (exp (* (/ PI 4.0) f)) (exp (- (* (/ PI 4.0) f)))))))))