invcot (example 3.9)

Percentage Accurate: 6.4% → 99.9%
Time: 15.0s
Alternatives: 5
Speedup: 35.7×

Specification

?
\[-0.026 < x \land x < 0.026\]
\[\begin{array}{l} \\ \frac{1}{x} - \frac{1}{\tan x} \end{array} \]
(FPCore (x) :precision binary64 (- (/ 1.0 x) (/ 1.0 (tan x))))
double code(double x) {
	return (1.0 / x) - (1.0 / tan(x));
}
real(8) function code(x)
    real(8), intent (in) :: x
    code = (1.0d0 / x) - (1.0d0 / tan(x))
end function
public static double code(double x) {
	return (1.0 / x) - (1.0 / Math.tan(x));
}
def code(x):
	return (1.0 / x) - (1.0 / math.tan(x))
function code(x)
	return Float64(Float64(1.0 / x) - Float64(1.0 / tan(x)))
end
function tmp = code(x)
	tmp = (1.0 / x) - (1.0 / tan(x));
end
code[x_] := N[(N[(1.0 / x), $MachinePrecision] - N[(1.0 / N[Tan[x], $MachinePrecision]), $MachinePrecision]), $MachinePrecision]
\begin{array}{l}

\\
\frac{1}{x} - \frac{1}{\tan x}
\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 5 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: 6.4% accurate, 1.0× speedup?

\[\begin{array}{l} \\ \frac{1}{x} - \frac{1}{\tan x} \end{array} \]
(FPCore (x) :precision binary64 (- (/ 1.0 x) (/ 1.0 (tan x))))
double code(double x) {
	return (1.0 / x) - (1.0 / tan(x));
}
real(8) function code(x)
    real(8), intent (in) :: x
    code = (1.0d0 / x) - (1.0d0 / tan(x))
end function
public static double code(double x) {
	return (1.0 / x) - (1.0 / Math.tan(x));
}
def code(x):
	return (1.0 / x) - (1.0 / math.tan(x))
function code(x)
	return Float64(Float64(1.0 / x) - Float64(1.0 / tan(x)))
end
function tmp = code(x)
	tmp = (1.0 / x) - (1.0 / tan(x));
end
code[x_] := N[(N[(1.0 / x), $MachinePrecision] - N[(1.0 / N[Tan[x], $MachinePrecision]), $MachinePrecision]), $MachinePrecision]
\begin{array}{l}

\\
\frac{1}{x} - \frac{1}{\tan x}
\end{array}

Alternative 1: 99.9% accurate, 0.3× speedup?

\[\begin{array}{l} \\ \frac{x}{\frac{\mathsf{fma}\left(x, x \cdot 0.022222222222222223, -0.3333333333333333\right)}{\mathsf{fma}\left({x}^{4}, 0.0004938271604938272, -0.1111111111111111\right)}} \end{array} \]
(FPCore (x)
 :precision binary64
 (/
  x
  (/
   (fma x (* x 0.022222222222222223) -0.3333333333333333)
   (fma (pow x 4.0) 0.0004938271604938272 -0.1111111111111111))))
double code(double x) {
	return x / (fma(x, (x * 0.022222222222222223), -0.3333333333333333) / fma(pow(x, 4.0), 0.0004938271604938272, -0.1111111111111111));
}
function code(x)
	return Float64(x / Float64(fma(x, Float64(x * 0.022222222222222223), -0.3333333333333333) / fma((x ^ 4.0), 0.0004938271604938272, -0.1111111111111111)))
end
code[x_] := N[(x / N[(N[(x * N[(x * 0.022222222222222223), $MachinePrecision] + -0.3333333333333333), $MachinePrecision] / N[(N[Power[x, 4.0], $MachinePrecision] * 0.0004938271604938272 + -0.1111111111111111), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]
\begin{array}{l}

\\
\frac{x}{\frac{\mathsf{fma}\left(x, x \cdot 0.022222222222222223, -0.3333333333333333\right)}{\mathsf{fma}\left({x}^{4}, 0.0004938271604938272, -0.1111111111111111\right)}}
\end{array}
Derivation
  1. Initial program 6.7%

    \[\frac{1}{x} - \frac{1}{\tan x} \]
  2. Taylor expanded in x around 0 99.4%

    \[\leadsto \color{blue}{0.022222222222222223 \cdot {x}^{3} + 0.3333333333333333 \cdot x} \]
  3. Step-by-step derivation
    1. unpow399.4%

      \[\leadsto 0.022222222222222223 \cdot \color{blue}{\left(\left(x \cdot x\right) \cdot x\right)} + 0.3333333333333333 \cdot x \]
  4. Applied egg-rr99.4%

    \[\leadsto 0.022222222222222223 \cdot \color{blue}{\left(\left(x \cdot x\right) \cdot x\right)} + 0.3333333333333333 \cdot x \]
  5. Step-by-step derivation
    1. associate-*r*99.4%

      \[\leadsto \color{blue}{\left(0.022222222222222223 \cdot \left(x \cdot x\right)\right) \cdot x} + 0.3333333333333333 \cdot x \]
    2. distribute-rgt-out99.4%

      \[\leadsto \color{blue}{x \cdot \left(0.022222222222222223 \cdot \left(x \cdot x\right) + 0.3333333333333333\right)} \]
  6. Applied egg-rr99.4%

    \[\leadsto \color{blue}{x \cdot \left(0.022222222222222223 \cdot \left(x \cdot x\right) + 0.3333333333333333\right)} \]
  7. Step-by-step derivation
    1. flip-+99.4%

      \[\leadsto x \cdot \color{blue}{\frac{\left(0.022222222222222223 \cdot \left(x \cdot x\right)\right) \cdot \left(0.022222222222222223 \cdot \left(x \cdot x\right)\right) - 0.3333333333333333 \cdot 0.3333333333333333}{0.022222222222222223 \cdot \left(x \cdot x\right) - 0.3333333333333333}} \]
    2. associate-*r/99.5%

      \[\leadsto \color{blue}{\frac{x \cdot \left(\left(0.022222222222222223 \cdot \left(x \cdot x\right)\right) \cdot \left(0.022222222222222223 \cdot \left(x \cdot x\right)\right) - 0.3333333333333333 \cdot 0.3333333333333333\right)}{0.022222222222222223 \cdot \left(x \cdot x\right) - 0.3333333333333333}} \]
    3. *-commutative99.5%

      \[\leadsto \frac{x \cdot \left(\color{blue}{\left(\left(x \cdot x\right) \cdot 0.022222222222222223\right)} \cdot \left(0.022222222222222223 \cdot \left(x \cdot x\right)\right) - 0.3333333333333333 \cdot 0.3333333333333333\right)}{0.022222222222222223 \cdot \left(x \cdot x\right) - 0.3333333333333333} \]
    4. *-commutative99.5%

      \[\leadsto \frac{x \cdot \left(\left(\left(x \cdot x\right) \cdot 0.022222222222222223\right) \cdot \color{blue}{\left(\left(x \cdot x\right) \cdot 0.022222222222222223\right)} - 0.3333333333333333 \cdot 0.3333333333333333\right)}{0.022222222222222223 \cdot \left(x \cdot x\right) - 0.3333333333333333} \]
    5. swap-sqr99.5%

      \[\leadsto \frac{x \cdot \left(\color{blue}{\left(\left(x \cdot x\right) \cdot \left(x \cdot x\right)\right) \cdot \left(0.022222222222222223 \cdot 0.022222222222222223\right)} - 0.3333333333333333 \cdot 0.3333333333333333\right)}{0.022222222222222223 \cdot \left(x \cdot x\right) - 0.3333333333333333} \]
    6. fma-neg99.5%

      \[\leadsto \frac{x \cdot \color{blue}{\mathsf{fma}\left(\left(x \cdot x\right) \cdot \left(x \cdot x\right), 0.022222222222222223 \cdot 0.022222222222222223, -0.3333333333333333 \cdot 0.3333333333333333\right)}}{0.022222222222222223 \cdot \left(x \cdot x\right) - 0.3333333333333333} \]
    7. pow299.5%

      \[\leadsto \frac{x \cdot \mathsf{fma}\left(\color{blue}{{x}^{2}} \cdot \left(x \cdot x\right), 0.022222222222222223 \cdot 0.022222222222222223, -0.3333333333333333 \cdot 0.3333333333333333\right)}{0.022222222222222223 \cdot \left(x \cdot x\right) - 0.3333333333333333} \]
    8. pow299.5%

      \[\leadsto \frac{x \cdot \mathsf{fma}\left({x}^{2} \cdot \color{blue}{{x}^{2}}, 0.022222222222222223 \cdot 0.022222222222222223, -0.3333333333333333 \cdot 0.3333333333333333\right)}{0.022222222222222223 \cdot \left(x \cdot x\right) - 0.3333333333333333} \]
    9. pow-prod-up99.5%

      \[\leadsto \frac{x \cdot \mathsf{fma}\left(\color{blue}{{x}^{\left(2 + 2\right)}}, 0.022222222222222223 \cdot 0.022222222222222223, -0.3333333333333333 \cdot 0.3333333333333333\right)}{0.022222222222222223 \cdot \left(x \cdot x\right) - 0.3333333333333333} \]
    10. metadata-eval99.5%

      \[\leadsto \frac{x \cdot \mathsf{fma}\left({x}^{\color{blue}{4}}, 0.022222222222222223 \cdot 0.022222222222222223, -0.3333333333333333 \cdot 0.3333333333333333\right)}{0.022222222222222223 \cdot \left(x \cdot x\right) - 0.3333333333333333} \]
    11. metadata-eval99.5%

      \[\leadsto \frac{x \cdot \mathsf{fma}\left({x}^{4}, \color{blue}{0.0004938271604938272}, -0.3333333333333333 \cdot 0.3333333333333333\right)}{0.022222222222222223 \cdot \left(x \cdot x\right) - 0.3333333333333333} \]
    12. metadata-eval99.5%

      \[\leadsto \frac{x \cdot \mathsf{fma}\left({x}^{4}, 0.0004938271604938272, -\color{blue}{0.1111111111111111}\right)}{0.022222222222222223 \cdot \left(x \cdot x\right) - 0.3333333333333333} \]
    13. metadata-eval99.5%

      \[\leadsto \frac{x \cdot \mathsf{fma}\left({x}^{4}, 0.0004938271604938272, \color{blue}{-0.1111111111111111}\right)}{0.022222222222222223 \cdot \left(x \cdot x\right) - 0.3333333333333333} \]
    14. associate-*r*99.5%

      \[\leadsto \frac{x \cdot \mathsf{fma}\left({x}^{4}, 0.0004938271604938272, -0.1111111111111111\right)}{\color{blue}{\left(0.022222222222222223 \cdot x\right) \cdot x} - 0.3333333333333333} \]
    15. *-commutative99.5%

      \[\leadsto \frac{x \cdot \mathsf{fma}\left({x}^{4}, 0.0004938271604938272, -0.1111111111111111\right)}{\color{blue}{x \cdot \left(0.022222222222222223 \cdot x\right)} - 0.3333333333333333} \]
    16. fma-neg99.5%

      \[\leadsto \frac{x \cdot \mathsf{fma}\left({x}^{4}, 0.0004938271604938272, -0.1111111111111111\right)}{\color{blue}{\mathsf{fma}\left(x, 0.022222222222222223 \cdot x, -0.3333333333333333\right)}} \]
    17. *-commutative99.5%

      \[\leadsto \frac{x \cdot \mathsf{fma}\left({x}^{4}, 0.0004938271604938272, -0.1111111111111111\right)}{\mathsf{fma}\left(x, \color{blue}{x \cdot 0.022222222222222223}, -0.3333333333333333\right)} \]
    18. metadata-eval99.5%

      \[\leadsto \frac{x \cdot \mathsf{fma}\left({x}^{4}, 0.0004938271604938272, -0.1111111111111111\right)}{\mathsf{fma}\left(x, x \cdot 0.022222222222222223, \color{blue}{-0.3333333333333333}\right)} \]
  8. Applied egg-rr99.5%

    \[\leadsto \color{blue}{\frac{x \cdot \mathsf{fma}\left({x}^{4}, 0.0004938271604938272, -0.1111111111111111\right)}{\mathsf{fma}\left(x, x \cdot 0.022222222222222223, -0.3333333333333333\right)}} \]
  9. Step-by-step derivation
    1. associate-/l*100.0%

      \[\leadsto \color{blue}{\frac{x}{\frac{\mathsf{fma}\left(x, x \cdot 0.022222222222222223, -0.3333333333333333\right)}{\mathsf{fma}\left({x}^{4}, 0.0004938271604938272, -0.1111111111111111\right)}}} \]
  10. Simplified100.0%

    \[\leadsto \color{blue}{\frac{x}{\frac{\mathsf{fma}\left(x, x \cdot 0.022222222222222223, -0.3333333333333333\right)}{\mathsf{fma}\left({x}^{4}, 0.0004938271604938272, -0.1111111111111111\right)}}} \]
  11. Final simplification100.0%

    \[\leadsto \frac{x}{\frac{\mathsf{fma}\left(x, x \cdot 0.022222222222222223, -0.3333333333333333\right)}{\mathsf{fma}\left({x}^{4}, 0.0004938271604938272, -0.1111111111111111\right)}} \]

Alternative 2: 99.3% accurate, 9.7× speedup?

\[\begin{array}{l} \\ \frac{1}{x \cdot -0.2 + 3 \cdot \frac{1}{x}} \end{array} \]
(FPCore (x) :precision binary64 (/ 1.0 (+ (* x -0.2) (* 3.0 (/ 1.0 x)))))
double code(double x) {
	return 1.0 / ((x * -0.2) + (3.0 * (1.0 / x)));
}
real(8) function code(x)
    real(8), intent (in) :: x
    code = 1.0d0 / ((x * (-0.2d0)) + (3.0d0 * (1.0d0 / x)))
end function
public static double code(double x) {
	return 1.0 / ((x * -0.2) + (3.0 * (1.0 / x)));
}
def code(x):
	return 1.0 / ((x * -0.2) + (3.0 * (1.0 / x)))
function code(x)
	return Float64(1.0 / Float64(Float64(x * -0.2) + Float64(3.0 * Float64(1.0 / x))))
end
function tmp = code(x)
	tmp = 1.0 / ((x * -0.2) + (3.0 * (1.0 / x)));
end
code[x_] := N[(1.0 / N[(N[(x * -0.2), $MachinePrecision] + N[(3.0 * N[(1.0 / x), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]
\begin{array}{l}

\\
\frac{1}{x \cdot -0.2 + 3 \cdot \frac{1}{x}}
\end{array}
Derivation
  1. Initial program 6.7%

    \[\frac{1}{x} - \frac{1}{\tan x} \]
  2. Taylor expanded in x around 0 99.4%

    \[\leadsto \color{blue}{0.022222222222222223 \cdot {x}^{3} + 0.3333333333333333 \cdot x} \]
  3. Step-by-step derivation
    1. unpow399.4%

      \[\leadsto 0.022222222222222223 \cdot \color{blue}{\left(\left(x \cdot x\right) \cdot x\right)} + 0.3333333333333333 \cdot x \]
  4. Applied egg-rr99.4%

    \[\leadsto 0.022222222222222223 \cdot \color{blue}{\left(\left(x \cdot x\right) \cdot x\right)} + 0.3333333333333333 \cdot x \]
  5. Step-by-step derivation
    1. pow399.4%

      \[\leadsto 0.022222222222222223 \cdot \color{blue}{{x}^{3}} + 0.3333333333333333 \cdot x \]
    2. +-commutative99.4%

      \[\leadsto \color{blue}{0.3333333333333333 \cdot x + 0.022222222222222223 \cdot {x}^{3}} \]
    3. fma-def99.4%

      \[\leadsto \color{blue}{\mathsf{fma}\left(0.3333333333333333, x, 0.022222222222222223 \cdot {x}^{3}\right)} \]
    4. add-sqr-sqrt49.1%

      \[\leadsto \mathsf{fma}\left(0.3333333333333333, \color{blue}{\sqrt{x} \cdot \sqrt{x}}, 0.022222222222222223 \cdot {x}^{3}\right) \]
    5. sqrt-prod26.5%

      \[\leadsto \mathsf{fma}\left(0.3333333333333333, \color{blue}{\sqrt{x \cdot x}}, 0.022222222222222223 \cdot {x}^{3}\right) \]
    6. fma-def26.5%

      \[\leadsto \color{blue}{0.3333333333333333 \cdot \sqrt{x \cdot x} + 0.022222222222222223 \cdot {x}^{3}} \]
    7. metadata-eval26.5%

      \[\leadsto \color{blue}{\sqrt{0.1111111111111111}} \cdot \sqrt{x \cdot x} + 0.022222222222222223 \cdot {x}^{3} \]
    8. sqrt-prod26.5%

      \[\leadsto \color{blue}{\sqrt{0.1111111111111111 \cdot \left(x \cdot x\right)}} + 0.022222222222222223 \cdot {x}^{3} \]
    9. +-commutative26.5%

      \[\leadsto \color{blue}{0.022222222222222223 \cdot {x}^{3} + \sqrt{0.1111111111111111 \cdot \left(x \cdot x\right)}} \]
    10. fma-def26.5%

      \[\leadsto \color{blue}{\mathsf{fma}\left(0.022222222222222223, {x}^{3}, \sqrt{0.1111111111111111 \cdot \left(x \cdot x\right)}\right)} \]
    11. metadata-eval26.5%

      \[\leadsto \mathsf{fma}\left(0.022222222222222223, {x}^{3}, \sqrt{\color{blue}{\left(-0.3333333333333333 \cdot -0.3333333333333333\right)} \cdot \left(x \cdot x\right)}\right) \]
    12. swap-sqr26.5%

      \[\leadsto \mathsf{fma}\left(0.022222222222222223, {x}^{3}, \sqrt{\color{blue}{\left(-0.3333333333333333 \cdot x\right) \cdot \left(-0.3333333333333333 \cdot x\right)}}\right) \]
    13. sqrt-unprod2.0%

      \[\leadsto \mathsf{fma}\left(0.022222222222222223, {x}^{3}, \color{blue}{\sqrt{-0.3333333333333333 \cdot x} \cdot \sqrt{-0.3333333333333333 \cdot x}}\right) \]
    14. add-sqr-sqrt4.0%

      \[\leadsto \mathsf{fma}\left(0.022222222222222223, {x}^{3}, \color{blue}{-0.3333333333333333 \cdot x}\right) \]
    15. /-rgt-identity4.0%

      \[\leadsto \color{blue}{\frac{\mathsf{fma}\left(0.022222222222222223, {x}^{3}, -0.3333333333333333 \cdot x\right)}{1}} \]
    16. clear-num4.0%

      \[\leadsto \color{blue}{\frac{1}{\frac{1}{\mathsf{fma}\left(0.022222222222222223, {x}^{3}, -0.3333333333333333 \cdot x\right)}}} \]
    17. add-sqr-sqrt2.0%

      \[\leadsto \frac{1}{\frac{1}{\mathsf{fma}\left(0.022222222222222223, {x}^{3}, \color{blue}{\sqrt{-0.3333333333333333 \cdot x} \cdot \sqrt{-0.3333333333333333 \cdot x}}\right)}} \]
    18. sqrt-unprod26.5%

      \[\leadsto \frac{1}{\frac{1}{\mathsf{fma}\left(0.022222222222222223, {x}^{3}, \color{blue}{\sqrt{\left(-0.3333333333333333 \cdot x\right) \cdot \left(-0.3333333333333333 \cdot x\right)}}\right)}} \]
    19. swap-sqr26.5%

      \[\leadsto \frac{1}{\frac{1}{\mathsf{fma}\left(0.022222222222222223, {x}^{3}, \sqrt{\color{blue}{\left(-0.3333333333333333 \cdot -0.3333333333333333\right) \cdot \left(x \cdot x\right)}}\right)}} \]
    20. metadata-eval26.5%

      \[\leadsto \frac{1}{\frac{1}{\mathsf{fma}\left(0.022222222222222223, {x}^{3}, \sqrt{\color{blue}{0.1111111111111111} \cdot \left(x \cdot x\right)}\right)}} \]
    21. fma-def26.5%

      \[\leadsto \frac{1}{\frac{1}{\color{blue}{0.022222222222222223 \cdot {x}^{3} + \sqrt{0.1111111111111111 \cdot \left(x \cdot x\right)}}}} \]
    22. +-commutative26.5%

      \[\leadsto \frac{1}{\frac{1}{\color{blue}{\sqrt{0.1111111111111111 \cdot \left(x \cdot x\right)} + 0.022222222222222223 \cdot {x}^{3}}}} \]
  6. Applied egg-rr99.3%

    \[\leadsto \color{blue}{\frac{1}{\frac{1}{\mathsf{fma}\left(0.022222222222222223, {x}^{3}, x \cdot 0.3333333333333333\right)}}} \]
  7. Taylor expanded in x around 0 99.5%

    \[\leadsto \frac{1}{\color{blue}{-0.2 \cdot x + 3 \cdot \frac{1}{x}}} \]
  8. Final simplification99.5%

    \[\leadsto \frac{1}{x \cdot -0.2 + 3 \cdot \frac{1}{x}} \]

Alternative 3: 99.4% accurate, 11.9× speedup?

\[\begin{array}{l} \\ x \cdot \left(0.022222222222222223 \cdot \left(x \cdot x\right) + 0.3333333333333333\right) \end{array} \]
(FPCore (x)
 :precision binary64
 (* x (+ (* 0.022222222222222223 (* x x)) 0.3333333333333333)))
double code(double x) {
	return x * ((0.022222222222222223 * (x * x)) + 0.3333333333333333);
}
real(8) function code(x)
    real(8), intent (in) :: x
    code = x * ((0.022222222222222223d0 * (x * x)) + 0.3333333333333333d0)
end function
public static double code(double x) {
	return x * ((0.022222222222222223 * (x * x)) + 0.3333333333333333);
}
def code(x):
	return x * ((0.022222222222222223 * (x * x)) + 0.3333333333333333)
function code(x)
	return Float64(x * Float64(Float64(0.022222222222222223 * Float64(x * x)) + 0.3333333333333333))
end
function tmp = code(x)
	tmp = x * ((0.022222222222222223 * (x * x)) + 0.3333333333333333);
end
code[x_] := N[(x * N[(N[(0.022222222222222223 * N[(x * x), $MachinePrecision]), $MachinePrecision] + 0.3333333333333333), $MachinePrecision]), $MachinePrecision]
\begin{array}{l}

\\
x \cdot \left(0.022222222222222223 \cdot \left(x \cdot x\right) + 0.3333333333333333\right)
\end{array}
Derivation
  1. Initial program 6.7%

    \[\frac{1}{x} - \frac{1}{\tan x} \]
  2. Taylor expanded in x around 0 99.4%

    \[\leadsto \color{blue}{0.022222222222222223 \cdot {x}^{3} + 0.3333333333333333 \cdot x} \]
  3. Step-by-step derivation
    1. unpow399.4%

      \[\leadsto 0.022222222222222223 \cdot \color{blue}{\left(\left(x \cdot x\right) \cdot x\right)} + 0.3333333333333333 \cdot x \]
  4. Applied egg-rr99.4%

    \[\leadsto 0.022222222222222223 \cdot \color{blue}{\left(\left(x \cdot x\right) \cdot x\right)} + 0.3333333333333333 \cdot x \]
  5. Step-by-step derivation
    1. associate-*r*99.4%

      \[\leadsto \color{blue}{\left(0.022222222222222223 \cdot \left(x \cdot x\right)\right) \cdot x} + 0.3333333333333333 \cdot x \]
    2. distribute-rgt-out99.4%

      \[\leadsto \color{blue}{x \cdot \left(0.022222222222222223 \cdot \left(x \cdot x\right) + 0.3333333333333333\right)} \]
  6. Applied egg-rr99.4%

    \[\leadsto \color{blue}{x \cdot \left(0.022222222222222223 \cdot \left(x \cdot x\right) + 0.3333333333333333\right)} \]
  7. Final simplification99.4%

    \[\leadsto x \cdot \left(0.022222222222222223 \cdot \left(x \cdot x\right) + 0.3333333333333333\right) \]

Alternative 4: 99.1% accurate, 21.4× speedup?

\[\begin{array}{l} \\ \frac{1}{\frac{3}{x}} \end{array} \]
(FPCore (x) :precision binary64 (/ 1.0 (/ 3.0 x)))
double code(double x) {
	return 1.0 / (3.0 / x);
}
real(8) function code(x)
    real(8), intent (in) :: x
    code = 1.0d0 / (3.0d0 / x)
end function
public static double code(double x) {
	return 1.0 / (3.0 / x);
}
def code(x):
	return 1.0 / (3.0 / x)
function code(x)
	return Float64(1.0 / Float64(3.0 / x))
end
function tmp = code(x)
	tmp = 1.0 / (3.0 / x);
end
code[x_] := N[(1.0 / N[(3.0 / x), $MachinePrecision]), $MachinePrecision]
\begin{array}{l}

\\
\frac{1}{\frac{3}{x}}
\end{array}
Derivation
  1. Initial program 6.7%

    \[\frac{1}{x} - \frac{1}{\tan x} \]
  2. Taylor expanded in x around 0 99.4%

    \[\leadsto \color{blue}{0.022222222222222223 \cdot {x}^{3} + 0.3333333333333333 \cdot x} \]
  3. Step-by-step derivation
    1. unpow399.4%

      \[\leadsto 0.022222222222222223 \cdot \color{blue}{\left(\left(x \cdot x\right) \cdot x\right)} + 0.3333333333333333 \cdot x \]
  4. Applied egg-rr99.4%

    \[\leadsto 0.022222222222222223 \cdot \color{blue}{\left(\left(x \cdot x\right) \cdot x\right)} + 0.3333333333333333 \cdot x \]
  5. Step-by-step derivation
    1. pow399.4%

      \[\leadsto 0.022222222222222223 \cdot \color{blue}{{x}^{3}} + 0.3333333333333333 \cdot x \]
    2. +-commutative99.4%

      \[\leadsto \color{blue}{0.3333333333333333 \cdot x + 0.022222222222222223 \cdot {x}^{3}} \]
    3. fma-def99.4%

      \[\leadsto \color{blue}{\mathsf{fma}\left(0.3333333333333333, x, 0.022222222222222223 \cdot {x}^{3}\right)} \]
    4. add-sqr-sqrt49.1%

      \[\leadsto \mathsf{fma}\left(0.3333333333333333, \color{blue}{\sqrt{x} \cdot \sqrt{x}}, 0.022222222222222223 \cdot {x}^{3}\right) \]
    5. sqrt-prod26.5%

      \[\leadsto \mathsf{fma}\left(0.3333333333333333, \color{blue}{\sqrt{x \cdot x}}, 0.022222222222222223 \cdot {x}^{3}\right) \]
    6. fma-def26.5%

      \[\leadsto \color{blue}{0.3333333333333333 \cdot \sqrt{x \cdot x} + 0.022222222222222223 \cdot {x}^{3}} \]
    7. metadata-eval26.5%

      \[\leadsto \color{blue}{\sqrt{0.1111111111111111}} \cdot \sqrt{x \cdot x} + 0.022222222222222223 \cdot {x}^{3} \]
    8. sqrt-prod26.5%

      \[\leadsto \color{blue}{\sqrt{0.1111111111111111 \cdot \left(x \cdot x\right)}} + 0.022222222222222223 \cdot {x}^{3} \]
    9. +-commutative26.5%

      \[\leadsto \color{blue}{0.022222222222222223 \cdot {x}^{3} + \sqrt{0.1111111111111111 \cdot \left(x \cdot x\right)}} \]
    10. fma-def26.5%

      \[\leadsto \color{blue}{\mathsf{fma}\left(0.022222222222222223, {x}^{3}, \sqrt{0.1111111111111111 \cdot \left(x \cdot x\right)}\right)} \]
    11. metadata-eval26.5%

      \[\leadsto \mathsf{fma}\left(0.022222222222222223, {x}^{3}, \sqrt{\color{blue}{\left(-0.3333333333333333 \cdot -0.3333333333333333\right)} \cdot \left(x \cdot x\right)}\right) \]
    12. swap-sqr26.5%

      \[\leadsto \mathsf{fma}\left(0.022222222222222223, {x}^{3}, \sqrt{\color{blue}{\left(-0.3333333333333333 \cdot x\right) \cdot \left(-0.3333333333333333 \cdot x\right)}}\right) \]
    13. sqrt-unprod2.0%

      \[\leadsto \mathsf{fma}\left(0.022222222222222223, {x}^{3}, \color{blue}{\sqrt{-0.3333333333333333 \cdot x} \cdot \sqrt{-0.3333333333333333 \cdot x}}\right) \]
    14. add-sqr-sqrt4.0%

      \[\leadsto \mathsf{fma}\left(0.022222222222222223, {x}^{3}, \color{blue}{-0.3333333333333333 \cdot x}\right) \]
    15. /-rgt-identity4.0%

      \[\leadsto \color{blue}{\frac{\mathsf{fma}\left(0.022222222222222223, {x}^{3}, -0.3333333333333333 \cdot x\right)}{1}} \]
    16. clear-num4.0%

      \[\leadsto \color{blue}{\frac{1}{\frac{1}{\mathsf{fma}\left(0.022222222222222223, {x}^{3}, -0.3333333333333333 \cdot x\right)}}} \]
    17. add-sqr-sqrt2.0%

      \[\leadsto \frac{1}{\frac{1}{\mathsf{fma}\left(0.022222222222222223, {x}^{3}, \color{blue}{\sqrt{-0.3333333333333333 \cdot x} \cdot \sqrt{-0.3333333333333333 \cdot x}}\right)}} \]
    18. sqrt-unprod26.5%

      \[\leadsto \frac{1}{\frac{1}{\mathsf{fma}\left(0.022222222222222223, {x}^{3}, \color{blue}{\sqrt{\left(-0.3333333333333333 \cdot x\right) \cdot \left(-0.3333333333333333 \cdot x\right)}}\right)}} \]
    19. swap-sqr26.5%

      \[\leadsto \frac{1}{\frac{1}{\mathsf{fma}\left(0.022222222222222223, {x}^{3}, \sqrt{\color{blue}{\left(-0.3333333333333333 \cdot -0.3333333333333333\right) \cdot \left(x \cdot x\right)}}\right)}} \]
    20. metadata-eval26.5%

      \[\leadsto \frac{1}{\frac{1}{\mathsf{fma}\left(0.022222222222222223, {x}^{3}, \sqrt{\color{blue}{0.1111111111111111} \cdot \left(x \cdot x\right)}\right)}} \]
    21. fma-def26.5%

      \[\leadsto \frac{1}{\frac{1}{\color{blue}{0.022222222222222223 \cdot {x}^{3} + \sqrt{0.1111111111111111 \cdot \left(x \cdot x\right)}}}} \]
    22. +-commutative26.5%

      \[\leadsto \frac{1}{\frac{1}{\color{blue}{\sqrt{0.1111111111111111 \cdot \left(x \cdot x\right)} + 0.022222222222222223 \cdot {x}^{3}}}} \]
  6. Applied egg-rr99.3%

    \[\leadsto \color{blue}{\frac{1}{\frac{1}{\mathsf{fma}\left(0.022222222222222223, {x}^{3}, x \cdot 0.3333333333333333\right)}}} \]
  7. Taylor expanded in x around 0 98.9%

    \[\leadsto \frac{1}{\color{blue}{\frac{3}{x}}} \]
  8. Final simplification98.9%

    \[\leadsto \frac{1}{\frac{3}{x}} \]

Alternative 5: 99.0% accurate, 35.7× speedup?

\[\begin{array}{l} \\ x \cdot 0.3333333333333333 \end{array} \]
(FPCore (x) :precision binary64 (* x 0.3333333333333333))
double code(double x) {
	return x * 0.3333333333333333;
}
real(8) function code(x)
    real(8), intent (in) :: x
    code = x * 0.3333333333333333d0
end function
public static double code(double x) {
	return x * 0.3333333333333333;
}
def code(x):
	return x * 0.3333333333333333
function code(x)
	return Float64(x * 0.3333333333333333)
end
function tmp = code(x)
	tmp = x * 0.3333333333333333;
end
code[x_] := N[(x * 0.3333333333333333), $MachinePrecision]
\begin{array}{l}

\\
x \cdot 0.3333333333333333
\end{array}
Derivation
  1. Initial program 6.7%

    \[\frac{1}{x} - \frac{1}{\tan x} \]
  2. Taylor expanded in x around 0 98.8%

    \[\leadsto \color{blue}{0.3333333333333333 \cdot x} \]
  3. Final simplification98.8%

    \[\leadsto x \cdot 0.3333333333333333 \]

Developer target: 99.9% accurate, 0.5× speedup?

\[\begin{array}{l} \\ \begin{array}{l} \mathbf{if}\;\left|x\right| < 0.026:\\ \;\;\;\;\frac{x}{3} \cdot \left(1 + \frac{x \cdot x}{15}\right)\\ \mathbf{else}:\\ \;\;\;\;\frac{1}{x} - \frac{1}{\tan x}\\ \end{array} \end{array} \]
(FPCore (x)
 :precision binary64
 (if (< (fabs x) 0.026)
   (* (/ x 3.0) (+ 1.0 (/ (* x x) 15.0)))
   (- (/ 1.0 x) (/ 1.0 (tan x)))))
double code(double x) {
	double tmp;
	if (fabs(x) < 0.026) {
		tmp = (x / 3.0) * (1.0 + ((x * x) / 15.0));
	} else {
		tmp = (1.0 / x) - (1.0 / tan(x));
	}
	return tmp;
}
real(8) function code(x)
    real(8), intent (in) :: x
    real(8) :: tmp
    if (abs(x) < 0.026d0) then
        tmp = (x / 3.0d0) * (1.0d0 + ((x * x) / 15.0d0))
    else
        tmp = (1.0d0 / x) - (1.0d0 / tan(x))
    end if
    code = tmp
end function
public static double code(double x) {
	double tmp;
	if (Math.abs(x) < 0.026) {
		tmp = (x / 3.0) * (1.0 + ((x * x) / 15.0));
	} else {
		tmp = (1.0 / x) - (1.0 / Math.tan(x));
	}
	return tmp;
}
def code(x):
	tmp = 0
	if math.fabs(x) < 0.026:
		tmp = (x / 3.0) * (1.0 + ((x * x) / 15.0))
	else:
		tmp = (1.0 / x) - (1.0 / math.tan(x))
	return tmp
function code(x)
	tmp = 0.0
	if (abs(x) < 0.026)
		tmp = Float64(Float64(x / 3.0) * Float64(1.0 + Float64(Float64(x * x) / 15.0)));
	else
		tmp = Float64(Float64(1.0 / x) - Float64(1.0 / tan(x)));
	end
	return tmp
end
function tmp_2 = code(x)
	tmp = 0.0;
	if (abs(x) < 0.026)
		tmp = (x / 3.0) * (1.0 + ((x * x) / 15.0));
	else
		tmp = (1.0 / x) - (1.0 / tan(x));
	end
	tmp_2 = tmp;
end
code[x_] := If[Less[N[Abs[x], $MachinePrecision], 0.026], N[(N[(x / 3.0), $MachinePrecision] * N[(1.0 + N[(N[(x * x), $MachinePrecision] / 15.0), $MachinePrecision]), $MachinePrecision]), $MachinePrecision], N[(N[(1.0 / x), $MachinePrecision] - N[(1.0 / N[Tan[x], $MachinePrecision]), $MachinePrecision]), $MachinePrecision]]
\begin{array}{l}

\\
\begin{array}{l}
\mathbf{if}\;\left|x\right| < 0.026:\\
\;\;\;\;\frac{x}{3} \cdot \left(1 + \frac{x \cdot x}{15}\right)\\

\mathbf{else}:\\
\;\;\;\;\frac{1}{x} - \frac{1}{\tan x}\\


\end{array}
\end{array}

Reproduce

?
herbie shell --seed 2023297 
(FPCore (x)
  :name "invcot (example 3.9)"
  :precision binary64
  :pre (and (< -0.026 x) (< x 0.026))

  :herbie-target
  (if (< (fabs x) 0.026) (* (/ x 3.0) (+ 1.0 (/ (* x x) 15.0))) (- (/ 1.0 x) (/ 1.0 (tan x))))

  (- (/ 1.0 x) (/ 1.0 (tan x))))