Numeric.Log:$clog1p from log-domain-0.10.2.1, B

Percentage Accurate: 99.7% → 99.5%

Time: 5.2s

Precision: binary64

Cost: 6852

Math

\[\frac{x}{1 + \sqrt{x + 1}} \]

↓

\[\begin{array}{l} \mathbf{if}\;x \leq 1.1 \cdot 10^{-5}:\\ \;\;\;\;\frac{x}{x \cdot 0.5 + 2}\\ \mathbf{else}:\\ \;\;\;\;\sqrt{x + 1} + -1\\ \end{array} \]

FPCore

(FPCore (x) :precision binary64 (/ x (+ 1.0 (sqrt (+ x 1.0)))))

↓

(FPCore (x)
 :precision binary64
 (if (<= x 1.1e-5) (/ x (+ (* x 0.5) 2.0)) (+ (sqrt (+ x 1.0)) -1.0)))

C

double code(double x) {
	return x / (1.0 + sqrt((x + 1.0)));
}

↓

double code(double x) {
	double tmp;
	if (x <= 1.1e-5) {
		tmp = x / ((x * 0.5) + 2.0);
	} else {
		tmp = sqrt((x + 1.0)) + -1.0;
	}
	return tmp;
}

Fortran

real(8) function code(x)
    real(8), intent (in) :: x
    code = x / (1.0d0 + sqrt((x + 1.0d0)))
end function

↓

real(8) function code(x)
    real(8), intent (in) :: x
    real(8) :: tmp
    if (x <= 1.1d-5) then
        tmp = x / ((x * 0.5d0) + 2.0d0)
    else
        tmp = sqrt((x + 1.0d0)) + (-1.0d0)
    end if
    code = tmp
end function

Java

public static double code(double x) {
	return x / (1.0 + Math.sqrt((x + 1.0)));
}

↓

public static double code(double x) {
	double tmp;
	if (x <= 1.1e-5) {
		tmp = x / ((x * 0.5) + 2.0);
	} else {
		tmp = Math.sqrt((x + 1.0)) + -1.0;
	}
	return tmp;
}

Python

def code(x):
	return x / (1.0 + math.sqrt((x + 1.0)))

↓

def code(x):
	tmp = 0
	if x <= 1.1e-5:
		tmp = x / ((x * 0.5) + 2.0)
	else:
		tmp = math.sqrt((x + 1.0)) + -1.0
	return tmp

Julia

function code(x)
	return Float64(x / Float64(1.0 + sqrt(Float64(x + 1.0))))
end

↓

function code(x)
	tmp = 0.0
	if (x <= 1.1e-5)
		tmp = Float64(x / Float64(Float64(x * 0.5) + 2.0));
	else
		tmp = Float64(sqrt(Float64(x + 1.0)) + -1.0);
	end
	return tmp
end

MATLAB

function tmp = code(x)
	tmp = x / (1.0 + sqrt((x + 1.0)));
end

↓

function tmp_2 = code(x)
	tmp = 0.0;
	if (x <= 1.1e-5)
		tmp = x / ((x * 0.5) + 2.0);
	else
		tmp = sqrt((x + 1.0)) + -1.0;
	end
	tmp_2 = tmp;
end

Wolfram

code[x_] := N[(x / N[(1.0 + N[Sqrt[N[(x + 1.0), $MachinePrecision]], $MachinePrecision]), $MachinePrecision]), $MachinePrecision]

↓

code[x_] := If[LessEqual[x, 1.1e-5], N[(x / N[(N[(x * 0.5), $MachinePrecision] + 2.0), $MachinePrecision]), $MachinePrecision], N[(N[Sqrt[N[(x + 1.0), $MachinePrecision]], $MachinePrecision] + -1.0), $MachinePrecision]]

Derivation

1. Split input into 2 regimes

2. if x < 1.1e-5

   1. Initial program 100.0%

      \[\frac{x}{1 + \sqrt{x + 1}} \]

   2. Taylor expanded in x around 0 99.9%

      \[\leadsto \frac{x}{\color{blue}{0.5 \cdot x + 2}} \]

   if 1.1e-5 < x

   1. Initial program 99.4%

      \[\frac{x}{1 + \sqrt{x + 1}} \]

   2. Applied egg-rr 99.9%

      \[\leadsto \color{blue}{\frac{x}{-x} \cdot \left(1 - \sqrt{x + 1}\right)} \]
      Step-by-step derivation

      [Start] 99.4%	\[ \frac{x}{1 + \sqrt{x + 1}} \]
      flip-+ [=>] 99.4%	\[ \frac{x}{\color{blue}{\frac{1 \cdot 1 - \sqrt{x + 1} \cdot \sqrt{x + 1}}{1 - \sqrt{x + 1}}}} \]
      metadata-eval [=>] 99.4%	\[ \frac{x}{\frac{\color{blue}{1} - \sqrt{x + 1} \cdot \sqrt{x + 1}}{1 - \sqrt{x + 1}}} \]
      add-sqr-sqrt [<=] 100.0%	\[ \frac{x}{\frac{1 - \color{blue}{\left(x + 1\right)}}{1 - \sqrt{x + 1}}} \]
      +-commutative [=>] 100.0%	\[ \frac{x}{\frac{1 - \color{blue}{\left(1 + x\right)}}{1 - \sqrt{x + 1}}} \]
      associate--r+ [=>] 99.9%	\[ \frac{x}{\frac{\color{blue}{\left(1 - 1\right) - x}}{1 - \sqrt{x + 1}}} \]
      metadata-eval [=>] 99.9%	\[ \frac{x}{\frac{\color{blue}{0} - x}{1 - \sqrt{x + 1}}} \]
      neg-sub0 [<=] 99.9%	\[ \frac{x}{\frac{\color{blue}{-x}}{1 - \sqrt{x + 1}}} \]
      associate-/r/ [=>] 99.9%	\[ \color{blue}{\frac{x}{-x} \cdot \left(1 - \sqrt{x + 1}\right)} \]

   3. Simplified 99.9%

      \[\leadsto \color{blue}{\sqrt{x + 1} + -1} \]
      Step-by-step derivation

      [Start] 99.9%	\[ \frac{x}{-x} \cdot \left(1 - \sqrt{x + 1}\right) \]
      sub-neg [=>] 99.9%	\[ \frac{x}{-x} \cdot \color{blue}{\left(1 + \left(-\sqrt{x + 1}\right)\right)} \]
      +-commutative [=>] 99.9%	\[ \frac{x}{-x} \cdot \color{blue}{\left(\left(-\sqrt{x + 1}\right) + 1\right)} \]
      remove-double-neg [<=] 99.9%	\[ \frac{\color{blue}{-\left(-x\right)}}{-x} \cdot \left(\left(-\sqrt{x + 1}\right) + 1\right) \]
      distribute-frac-neg [=>] 99.9%	\[ \color{blue}{\left(-\frac{-x}{-x}\right)} \cdot \left(\left(-\sqrt{x + 1}\right) + 1\right) \]
      *-inverses [=>] 99.9%	\[ \left(-\color{blue}{1}\right) \cdot \left(\left(-\sqrt{x + 1}\right) + 1\right) \]
      metadata-eval [=>] 99.9%	\[ \color{blue}{-1} \cdot \left(\left(-\sqrt{x + 1}\right) + 1\right) \]
      distribute-lft-in [=>] 99.9%	\[ \color{blue}{-1 \cdot \left(-\sqrt{x + 1}\right) + -1 \cdot 1} \]
      neg-mul-1 [<=] 99.9%	\[ \color{blue}{\left(-\left(-\sqrt{x + 1}\right)\right)} + -1 \cdot 1 \]
      remove-double-neg [=>] 99.9%	\[ \color{blue}{\sqrt{x + 1}} + -1 \cdot 1 \]
      metadata-eval [=>] 99.9%	\[ \sqrt{x + 1} + \color{blue}{-1} \]

3. Recombined 2 regimes into one program.

4. Final simplification 99.9%

   \[\leadsto \begin{array}{l} \mathbf{if}\;x \leq 1.1 \cdot 10^{-5}:\\ \;\;\;\;\frac{x}{x \cdot 0.5 + 2}\\ \mathbf{else}:\\ \;\;\;\;\sqrt{x + 1} + -1\\ \end{array} \]

Alternatives

Alternative 1
Accuracy	99.5%
Cost	6852

\[\begin{array}{l} \mathbf{if}\;x \leq 1.1 \cdot 10^{-5}:\\ \;\;\;\;\frac{x}{x \cdot 0.5 + 2}\\ \mathbf{else}:\\ \;\;\;\;\sqrt{x + 1} + -1\\ \end{array} \]

Alternative 2
Accuracy	99.7%
Cost	6848

\[\frac{x}{1 + \sqrt{x + 1}} \]

Alternative 3
Accuracy	67.8%
Cost	448

\[\frac{x}{x \cdot 0.5 + 2} \]

Alternative 4
Accuracy	67.1%
Cost	192

\[\frac{x}{2} \]

Alternative 5
Accuracy	4.8%
Cost	64

\[2 \]

Reproduce

herbie shell --seed 2023263 
(FPCore (x)
  :name "Numeric.Log:$clog1p from log-domain-0.10.2.1, B"
  :precision binary64
  (/ x (+ 1.0 (sqrt (+ x 1.0)))))