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

Percentage Accurate: 99.7% → 99.7%

Time: 6.1s

Alternatives: 10

Speedup: 1.0×

Specification

Math

\[\begin{array}{l} \\ \frac{x}{1 + \sqrt{x + 1}} \end{array} \]

FPCore

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

C

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

Fortran

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

Java

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

Python

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

Julia

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

MATLAB

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

Wolfram

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

Initial Program: 99.7% accurate, 1.0× speedup

Math

\[\begin{array}{l} \\ \frac{x}{1 + \sqrt{x + 1}} \end{array} \]

FPCore

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

C

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

Fortran

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

Java

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

Python

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

Julia

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

MATLAB

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

Wolfram

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

Alternative 1: 99.7% accurate, 1.0× speedup

Math

\[\begin{array}{l} \\ \frac{x}{1 + \sqrt{x + 1}} \end{array} \]

FPCore

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

C

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

Fortran

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

Java

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

Python

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

Julia

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

MATLAB

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

Wolfram

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

Derivation

1. Initial program 99.7%

   \[\frac{x}{1 + \sqrt{x + 1}} \]
2. Add Preprocessing
3. Add Preprocessing

Alternative 2: 98.9% accurate, 0.9× speedup

Math

\[\begin{array}{l} \\ \begin{array}{l} \mathbf{if}\;x \leq 1.95:\\ \;\;\;\;\mathsf{fma}\left(\mathsf{fma}\left(-0.0390625, x, 0.0625\right) \cdot x - 0.125, x, 0.5\right) \cdot x\\ \mathbf{else}:\\ \;\;\;\;\sqrt{x} - 1\\ \end{array} \end{array} \]

FPCore

(FPCore (x)
 :precision binary64
 (if (<= x 1.95)
   (* (fma (- (* (fma -0.0390625 x 0.0625) x) 0.125) x 0.5) x)
   (- (sqrt x) 1.0)))

C

double code(double x) {
	double tmp;
	if (x <= 1.95) {
		tmp = fma(((fma(-0.0390625, x, 0.0625) * x) - 0.125), x, 0.5) * x;
	} else {
		tmp = sqrt(x) - 1.0;
	}
	return tmp;
}

Julia

function code(x)
	tmp = 0.0
	if (x <= 1.95)
		tmp = Float64(fma(Float64(Float64(fma(-0.0390625, x, 0.0625) * x) - 0.125), x, 0.5) * x);
	else
		tmp = Float64(sqrt(x) - 1.0);
	end
	return tmp
end

Wolfram

code[x_] := If[LessEqual[x, 1.95], N[(N[(N[(N[(N[(-0.0390625 * x + 0.0625), $MachinePrecision] * x), $MachinePrecision] - 0.125), $MachinePrecision] * x + 0.5), $MachinePrecision] * x), $MachinePrecision], N[(N[Sqrt[x], $MachinePrecision] - 1.0), $MachinePrecision]]

Derivation

1. Split input into 2 regimes

2. if x < 1.94999999999999996

   1. Initial program 100.0%

      \[\frac{x}{1 + \sqrt{x + 1}} \]
   2. Add Preprocessing

   3. Taylor expanded in x around 0

      \[\leadsto \color{blue}{x \cdot \left(\frac{1}{2} + x \cdot \left(x \cdot \left(\frac{1}{16} + \frac{-5}{128} \cdot x\right) - \frac{1}{8}\right)\right)} \]
   4. Step-by-step derivation

      1. *-commutative N/A

         \[\leadsto \color{blue}{\left(\frac{1}{2} + x \cdot \left(x \cdot \left(\frac{1}{16} + \frac{-5}{128} \cdot x\right) - \frac{1}{8}\right)\right) \cdot x} \]

      2. lower-*.f64 N/A

         \[\leadsto \color{blue}{\left(\frac{1}{2} + x \cdot \left(x \cdot \left(\frac{1}{16} + \frac{-5}{128} \cdot x\right) - \frac{1}{8}\right)\right) \cdot x} \]

      3. +-commutative N/A

         \[\leadsto \color{blue}{\left(x \cdot \left(x \cdot \left(\frac{1}{16} + \frac{-5}{128} \cdot x\right) - \frac{1}{8}\right) + \frac{1}{2}\right)} \cdot x \]

      4. *-commutative N/A

         \[\leadsto \left(\color{blue}{\left(x \cdot \left(\frac{1}{16} + \frac{-5}{128} \cdot x\right) - \frac{1}{8}\right) \cdot x} + \frac{1}{2}\right) \cdot x \]

      5. lower-fma.f64 N/A

         \[\leadsto \color{blue}{\mathsf{fma}\left(x \cdot \left(\frac{1}{16} + \frac{-5}{128} \cdot x\right) - \frac{1}{8}, x, \frac{1}{2}\right)} \cdot x \]

      6. lower--.f64 N/A

         \[\leadsto \mathsf{fma}\left(\color{blue}{x \cdot \left(\frac{1}{16} + \frac{-5}{128} \cdot x\right) - \frac{1}{8}}, x, \frac{1}{2}\right) \cdot x \]

      7. *-commutative N/A

         \[\leadsto \mathsf{fma}\left(\color{blue}{\left(\frac{1}{16} + \frac{-5}{128} \cdot x\right) \cdot x} - \frac{1}{8}, x, \frac{1}{2}\right) \cdot x \]

      8. lower-*.f64 N/A

         \[\leadsto \mathsf{fma}\left(\color{blue}{\left(\frac{1}{16} + \frac{-5}{128} \cdot x\right) \cdot x} - \frac{1}{8}, x, \frac{1}{2}\right) \cdot x \]

      9. +-commutative N/A

         \[\leadsto \mathsf{fma}\left(\color{blue}{\left(\frac{-5}{128} \cdot x + \frac{1}{16}\right)} \cdot x - \frac{1}{8}, x, \frac{1}{2}\right) \cdot x \]

      10. lower-fma.f64 99.8

          \[\leadsto \mathsf{fma}\left(\color{blue}{\mathsf{fma}\left(-0.0390625, x, 0.0625\right)} \cdot x - 0.125, x, 0.5\right) \cdot x \]

   5. Applied rewrites 99.8%

      \[\leadsto \color{blue}{\mathsf{fma}\left(\mathsf{fma}\left(-0.0390625, x, 0.0625\right) \cdot x - 0.125, x, 0.5\right) \cdot x} \]

   if 1.94999999999999996 < x

   1. Initial program 99.3%

      \[\frac{x}{1 + \sqrt{x + 1}} \]
   2. Add Preprocessing

   3. Taylor expanded in x around inf

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

      1. lower--.f64 N/A

         \[\leadsto \color{blue}{\sqrt{x} - 1} \]

      2. lower-sqrt.f64 97.1

         \[\leadsto \color{blue}{\sqrt{x}} - 1 \]

   5. Applied rewrites 97.1%

      \[\leadsto \color{blue}{\sqrt{x} - 1} \]
3. Recombined 2 regimes into one program.
4. Add Preprocessing

Alternative 3: 98.8% accurate, 0.9× speedup

Math

\[\begin{array}{l} \\ \begin{array}{l} \mathbf{if}\;x \leq 3.5:\\ \;\;\;\;\frac{x}{\mathsf{fma}\left(\mathsf{fma}\left(-0.125, x, 0.5\right), x, 2\right)}\\ \mathbf{else}:\\ \;\;\;\;\sqrt{x} - 1\\ \end{array} \end{array} \]

FPCore

(FPCore (x)
 :precision binary64
 (if (<= x 3.5) (/ x (fma (fma -0.125 x 0.5) x 2.0)) (- (sqrt x) 1.0)))

C

double code(double x) {
	double tmp;
	if (x <= 3.5) {
		tmp = x / fma(fma(-0.125, x, 0.5), x, 2.0);
	} else {
		tmp = sqrt(x) - 1.0;
	}
	return tmp;
}

Julia

function code(x)
	tmp = 0.0
	if (x <= 3.5)
		tmp = Float64(x / fma(fma(-0.125, x, 0.5), x, 2.0));
	else
		tmp = Float64(sqrt(x) - 1.0);
	end
	return tmp
end

Wolfram

code[x_] := If[LessEqual[x, 3.5], N[(x / N[(N[(-0.125 * x + 0.5), $MachinePrecision] * x + 2.0), $MachinePrecision]), $MachinePrecision], N[(N[Sqrt[x], $MachinePrecision] - 1.0), $MachinePrecision]]

Derivation

1. Split input into 2 regimes

2. if x < 3.5

   1. Initial program 100.0%

      \[\frac{x}{1 + \sqrt{x + 1}} \]
   2. Add Preprocessing

   3. Taylor expanded in x around 0

      \[\leadsto \frac{x}{\color{blue}{2 + x \cdot \left(\frac{1}{2} + \frac{-1}{8} \cdot x\right)}} \]
   4. Step-by-step derivation

      1. +-commutative N/A

         \[\leadsto \frac{x}{\color{blue}{x \cdot \left(\frac{1}{2} + \frac{-1}{8} \cdot x\right) + 2}} \]

      2. *-commutative N/A

         \[\leadsto \frac{x}{\color{blue}{\left(\frac{1}{2} + \frac{-1}{8} \cdot x\right) \cdot x} + 2} \]

      3. lower-fma.f64 N/A

         \[\leadsto \frac{x}{\color{blue}{\mathsf{fma}\left(\frac{1}{2} + \frac{-1}{8} \cdot x, x, 2\right)}} \]

      4. +-commutative N/A

         \[\leadsto \frac{x}{\mathsf{fma}\left(\color{blue}{\frac{-1}{8} \cdot x + \frac{1}{2}}, x, 2\right)} \]

      5. lower-fma.f64 99.7

         \[\leadsto \frac{x}{\mathsf{fma}\left(\color{blue}{\mathsf{fma}\left(-0.125, x, 0.5\right)}, x, 2\right)} \]

   5. Applied rewrites 99.7%

      \[\leadsto \frac{x}{\color{blue}{\mathsf{fma}\left(\mathsf{fma}\left(-0.125, x, 0.5\right), x, 2\right)}} \]

   if 3.5 < x

   1. Initial program 99.3%

      \[\frac{x}{1 + \sqrt{x + 1}} \]
   2. Add Preprocessing

   3. Taylor expanded in x around inf

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

      1. lower--.f64 N/A

         \[\leadsto \color{blue}{\sqrt{x} - 1} \]

      2. lower-sqrt.f64 97.1

         \[\leadsto \color{blue}{\sqrt{x}} - 1 \]

   5. Applied rewrites 97.1%

      \[\leadsto \color{blue}{\sqrt{x} - 1} \]
3. Recombined 2 regimes into one program.
4. Add Preprocessing

Alternative 4: 98.8% accurate, 1.1× speedup

Math

\[\begin{array}{l} \\ \begin{array}{l} \mathbf{if}\;x \leq 2.3:\\ \;\;\;\;\mathsf{fma}\left(0.0625 \cdot x - 0.125, x, 0.5\right) \cdot x\\ \mathbf{else}:\\ \;\;\;\;\sqrt{x} - 1\\ \end{array} \end{array} \]

FPCore

(FPCore (x)
 :precision binary64
 (if (<= x 2.3) (* (fma (- (* 0.0625 x) 0.125) x 0.5) x) (- (sqrt x) 1.0)))

C

double code(double x) {
	double tmp;
	if (x <= 2.3) {
		tmp = fma(((0.0625 * x) - 0.125), x, 0.5) * x;
	} else {
		tmp = sqrt(x) - 1.0;
	}
	return tmp;
}

Julia

function code(x)
	tmp = 0.0
	if (x <= 2.3)
		tmp = Float64(fma(Float64(Float64(0.0625 * x) - 0.125), x, 0.5) * x);
	else
		tmp = Float64(sqrt(x) - 1.0);
	end
	return tmp
end

Wolfram

code[x_] := If[LessEqual[x, 2.3], N[(N[(N[(N[(0.0625 * x), $MachinePrecision] - 0.125), $MachinePrecision] * x + 0.5), $MachinePrecision] * x), $MachinePrecision], N[(N[Sqrt[x], $MachinePrecision] - 1.0), $MachinePrecision]]

Derivation

1. Split input into 2 regimes

2. if x < 2.2999999999999998

   1. Initial program 100.0%

      \[\frac{x}{1 + \sqrt{x + 1}} \]
   2. Add Preprocessing

   3. Taylor expanded in x around 0

      \[\leadsto \color{blue}{x \cdot \left(\frac{1}{2} + x \cdot \left(\frac{1}{16} \cdot x - \frac{1}{8}\right)\right)} \]
   4. Step-by-step derivation

      1. *-commutative N/A

         \[\leadsto \color{blue}{\left(\frac{1}{2} + x \cdot \left(\frac{1}{16} \cdot x - \frac{1}{8}\right)\right) \cdot x} \]

      2. lower-*.f64 N/A

         \[\leadsto \color{blue}{\left(\frac{1}{2} + x \cdot \left(\frac{1}{16} \cdot x - \frac{1}{8}\right)\right) \cdot x} \]

      3. +-commutative N/A

         \[\leadsto \color{blue}{\left(x \cdot \left(\frac{1}{16} \cdot x - \frac{1}{8}\right) + \frac{1}{2}\right)} \cdot x \]

      4. *-commutative N/A

         \[\leadsto \left(\color{blue}{\left(\frac{1}{16} \cdot x - \frac{1}{8}\right) \cdot x} + \frac{1}{2}\right) \cdot x \]

      5. lower-fma.f64 N/A

         \[\leadsto \color{blue}{\mathsf{fma}\left(\frac{1}{16} \cdot x - \frac{1}{8}, x, \frac{1}{2}\right)} \cdot x \]

      6. lower--.f64 N/A

         \[\leadsto \mathsf{fma}\left(\color{blue}{\frac{1}{16} \cdot x - \frac{1}{8}}, x, \frac{1}{2}\right) \cdot x \]

      7. lower-*.f64 99.7

         \[\leadsto \mathsf{fma}\left(\color{blue}{0.0625 \cdot x} - 0.125, x, 0.5\right) \cdot x \]

   5. Applied rewrites 99.7%

      \[\leadsto \color{blue}{\mathsf{fma}\left(0.0625 \cdot x - 0.125, x, 0.5\right) \cdot x} \]

   if 2.2999999999999998 < x

   1. Initial program 99.3%

      \[\frac{x}{1 + \sqrt{x + 1}} \]
   2. Add Preprocessing

   3. Taylor expanded in x around inf

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

      1. lower--.f64 N/A

         \[\leadsto \color{blue}{\sqrt{x} - 1} \]

      2. lower-sqrt.f64 97.1

         \[\leadsto \color{blue}{\sqrt{x}} - 1 \]

   5. Applied rewrites 97.1%

      \[\leadsto \color{blue}{\sqrt{x} - 1} \]
3. Recombined 2 regimes into one program.
4. Add Preprocessing

Alternative 5: 98.6% accurate, 1.2× speedup

Math

\[\begin{array}{l} \\ \begin{array}{l} \mathbf{if}\;x \leq 4:\\ \;\;\;\;\frac{x}{\mathsf{fma}\left(0.5, x, 2\right)}\\ \mathbf{else}:\\ \;\;\;\;\sqrt{x} - 1\\ \end{array} \end{array} \]

FPCore

(FPCore (x)
 :precision binary64
 (if (<= x 4.0) (/ x (fma 0.5 x 2.0)) (- (sqrt x) 1.0)))

C

double code(double x) {
	double tmp;
	if (x <= 4.0) {
		tmp = x / fma(0.5, x, 2.0);
	} else {
		tmp = sqrt(x) - 1.0;
	}
	return tmp;
}

Julia

function code(x)
	tmp = 0.0
	if (x <= 4.0)
		tmp = Float64(x / fma(0.5, x, 2.0));
	else
		tmp = Float64(sqrt(x) - 1.0);
	end
	return tmp
end

Wolfram

code[x_] := If[LessEqual[x, 4.0], N[(x / N[(0.5 * x + 2.0), $MachinePrecision]), $MachinePrecision], N[(N[Sqrt[x], $MachinePrecision] - 1.0), $MachinePrecision]]

Derivation

1. Split input into 2 regimes

2. if x < 4

   1. Initial program 100.0%

      \[\frac{x}{1 + \sqrt{x + 1}} \]
   2. Add Preprocessing

   3. Taylor expanded in x around 0

      \[\leadsto \frac{x}{\color{blue}{2 + \frac{1}{2} \cdot x}} \]
   4. Step-by-step derivation

      1. +-commutative N/A

         \[\leadsto \frac{x}{\color{blue}{\frac{1}{2} \cdot x + 2}} \]

      2. lower-fma.f64 99.4

         \[\leadsto \frac{x}{\color{blue}{\mathsf{fma}\left(0.5, x, 2\right)}} \]

   5. Applied rewrites 99.4%

      \[\leadsto \frac{x}{\color{blue}{\mathsf{fma}\left(0.5, x, 2\right)}} \]

   if 4 < x

   1. Initial program 99.3%

      \[\frac{x}{1 + \sqrt{x + 1}} \]
   2. Add Preprocessing

   3. Taylor expanded in x around inf

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

      1. lower--.f64 N/A

         \[\leadsto \color{blue}{\sqrt{x} - 1} \]

      2. lower-sqrt.f64 97.1

         \[\leadsto \color{blue}{\sqrt{x}} - 1 \]

   5. Applied rewrites 97.1%

      \[\leadsto \color{blue}{\sqrt{x} - 1} \]
3. Recombined 2 regimes into one program.
4. Add Preprocessing

Alternative 6: 98.6% accurate, 1.2× speedup

Math

\[\begin{array}{l} \\ \begin{array}{l} \mathbf{if}\;x \leq 2.3:\\ \;\;\;\;\mathsf{fma}\left(-0.125 \cdot x, x, 0.5 \cdot x\right)\\ \mathbf{else}:\\ \;\;\;\;\sqrt{x} - 1\\ \end{array} \end{array} \]

FPCore

(FPCore (x)
 :precision binary64
 (if (<= x 2.3) (fma (* -0.125 x) x (* 0.5 x)) (- (sqrt x) 1.0)))

C

double code(double x) {
	double tmp;
	if (x <= 2.3) {
		tmp = fma((-0.125 * x), x, (0.5 * x));
	} else {
		tmp = sqrt(x) - 1.0;
	}
	return tmp;
}

Julia

function code(x)
	tmp = 0.0
	if (x <= 2.3)
		tmp = fma(Float64(-0.125 * x), x, Float64(0.5 * x));
	else
		tmp = Float64(sqrt(x) - 1.0);
	end
	return tmp
end

Wolfram

code[x_] := If[LessEqual[x, 2.3], N[(N[(-0.125 * x), $MachinePrecision] * x + N[(0.5 * x), $MachinePrecision]), $MachinePrecision], N[(N[Sqrt[x], $MachinePrecision] - 1.0), $MachinePrecision]]

Derivation

1. Split input into 2 regimes

2. if x < 2.2999999999999998

   1. Initial program 100.0%

      \[\frac{x}{1 + \sqrt{x + 1}} \]
   2. Add Preprocessing

   3. Taylor expanded in x around 0

      \[\leadsto \color{blue}{x \cdot \left(\frac{1}{2} + \frac{-1}{8} \cdot x\right)} \]
   4. Step-by-step derivation

      1. *-commutative N/A

         \[\leadsto \color{blue}{\left(\frac{1}{2} + \frac{-1}{8} \cdot x\right) \cdot x} \]

      2. lower-*.f64 N/A

         \[\leadsto \color{blue}{\left(\frac{1}{2} + \frac{-1}{8} \cdot x\right) \cdot x} \]

      3. +-commutative N/A

         \[\leadsto \color{blue}{\left(\frac{-1}{8} \cdot x + \frac{1}{2}\right)} \cdot x \]

      4. lower-fma.f64 99.4

         \[\leadsto \color{blue}{\mathsf{fma}\left(-0.125, x, 0.5\right)} \cdot x \]

   5. Applied rewrites 99.4%

      \[\leadsto \color{blue}{\mathsf{fma}\left(-0.125, x, 0.5\right) \cdot x} \]
   6. Step-by-step derivation

   7. Applied rewrites 99.4%

      \[\leadsto \mathsf{fma}\left(-0.125 \cdot x, \color{blue}{x}, 0.5 \cdot x\right) \]

   if 2.2999999999999998 < x

   1. Initial program 99.3%

      \[\frac{x}{1 + \sqrt{x + 1}} \]
   2. Add Preprocessing

   3. Taylor expanded in x around inf

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

      1. lower--.f64 N/A

         \[\leadsto \color{blue}{\sqrt{x} - 1} \]

      2. lower-sqrt.f64 97.1

         \[\leadsto \color{blue}{\sqrt{x}} - 1 \]

   5. Applied rewrites 97.1%

      \[\leadsto \color{blue}{\sqrt{x} - 1} \]
3. Recombined 2 regimes into one program.
4. Add Preprocessing

Alternative 7: 98.6% accurate, 1.4× speedup

Math

\[\begin{array}{l} \\ \begin{array}{l} \mathbf{if}\;x \leq 2.3:\\ \;\;\;\;\mathsf{fma}\left(-0.125, x, 0.5\right) \cdot x\\ \mathbf{else}:\\ \;\;\;\;\sqrt{x} - 1\\ \end{array} \end{array} \]

FPCore

(FPCore (x)
 :precision binary64
 (if (<= x 2.3) (* (fma -0.125 x 0.5) x) (- (sqrt x) 1.0)))

C

double code(double x) {
	double tmp;
	if (x <= 2.3) {
		tmp = fma(-0.125, x, 0.5) * x;
	} else {
		tmp = sqrt(x) - 1.0;
	}
	return tmp;
}

Julia

function code(x)
	tmp = 0.0
	if (x <= 2.3)
		tmp = Float64(fma(-0.125, x, 0.5) * x);
	else
		tmp = Float64(sqrt(x) - 1.0);
	end
	return tmp
end

Wolfram

code[x_] := If[LessEqual[x, 2.3], N[(N[(-0.125 * x + 0.5), $MachinePrecision] * x), $MachinePrecision], N[(N[Sqrt[x], $MachinePrecision] - 1.0), $MachinePrecision]]

Derivation

1. Split input into 2 regimes

2. if x < 2.2999999999999998

   1. Initial program 100.0%

      \[\frac{x}{1 + \sqrt{x + 1}} \]
   2. Add Preprocessing

   3. Taylor expanded in x around 0

      \[\leadsto \color{blue}{x \cdot \left(\frac{1}{2} + \frac{-1}{8} \cdot x\right)} \]
   4. Step-by-step derivation

      1. *-commutative N/A

         \[\leadsto \color{blue}{\left(\frac{1}{2} + \frac{-1}{8} \cdot x\right) \cdot x} \]

      2. lower-*.f64 N/A

         \[\leadsto \color{blue}{\left(\frac{1}{2} + \frac{-1}{8} \cdot x\right) \cdot x} \]

      3. +-commutative N/A

         \[\leadsto \color{blue}{\left(\frac{-1}{8} \cdot x + \frac{1}{2}\right)} \cdot x \]

      4. lower-fma.f64 99.4

         \[\leadsto \color{blue}{\mathsf{fma}\left(-0.125, x, 0.5\right)} \cdot x \]

   5. Applied rewrites 99.4%

      \[\leadsto \color{blue}{\mathsf{fma}\left(-0.125, x, 0.5\right) \cdot x} \]

   if 2.2999999999999998 < x

   1. Initial program 99.3%

      \[\frac{x}{1 + \sqrt{x + 1}} \]
   2. Add Preprocessing

   3. Taylor expanded in x around inf

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

      1. lower--.f64 N/A

         \[\leadsto \color{blue}{\sqrt{x} - 1} \]

      2. lower-sqrt.f64 97.1

         \[\leadsto \color{blue}{\sqrt{x}} - 1 \]

   5. Applied rewrites 97.1%

      \[\leadsto \color{blue}{\sqrt{x} - 1} \]
3. Recombined 2 regimes into one program.
4. Add Preprocessing

Alternative 8: 97.9% accurate, 1.6× speedup

Math

\[\begin{array}{l} \\ \begin{array}{l} \mathbf{if}\;x \leq 2.5:\\ \;\;\;\;\mathsf{fma}\left(-0.125, x, 0.5\right) \cdot x\\ \mathbf{else}:\\ \;\;\;\;\sqrt{x}\\ \end{array} \end{array} \]

FPCore

(FPCore (x)
 :precision binary64
 (if (<= x 2.5) (* (fma -0.125 x 0.5) x) (sqrt x)))

C

double code(double x) {
	double tmp;
	if (x <= 2.5) {
		tmp = fma(-0.125, x, 0.5) * x;
	} else {
		tmp = sqrt(x);
	}
	return tmp;
}

Julia

function code(x)
	tmp = 0.0
	if (x <= 2.5)
		tmp = Float64(fma(-0.125, x, 0.5) * x);
	else
		tmp = sqrt(x);
	end
	return tmp
end

Wolfram

code[x_] := If[LessEqual[x, 2.5], N[(N[(-0.125 * x + 0.5), $MachinePrecision] * x), $MachinePrecision], N[Sqrt[x], $MachinePrecision]]

Derivation

1. Split input into 2 regimes

2. if x < 2.5

   1. Initial program 100.0%

      \[\frac{x}{1 + \sqrt{x + 1}} \]
   2. Add Preprocessing

   3. Taylor expanded in x around 0

      \[\leadsto \color{blue}{x \cdot \left(\frac{1}{2} + \frac{-1}{8} \cdot x\right)} \]
   4. Step-by-step derivation

      1. *-commutative N/A

         \[\leadsto \color{blue}{\left(\frac{1}{2} + \frac{-1}{8} \cdot x\right) \cdot x} \]

      2. lower-*.f64 N/A

         \[\leadsto \color{blue}{\left(\frac{1}{2} + \frac{-1}{8} \cdot x\right) \cdot x} \]

      3. +-commutative N/A

         \[\leadsto \color{blue}{\left(\frac{-1}{8} \cdot x + \frac{1}{2}\right)} \cdot x \]

      4. lower-fma.f64 99.4

         \[\leadsto \color{blue}{\mathsf{fma}\left(-0.125, x, 0.5\right)} \cdot x \]

   5. Applied rewrites 99.4%

      \[\leadsto \color{blue}{\mathsf{fma}\left(-0.125, x, 0.5\right) \cdot x} \]

   if 2.5 < x

   1. Initial program 99.3%

      \[\frac{x}{1 + \sqrt{x + 1}} \]
   2. Add Preprocessing

   3. Taylor expanded in x around inf

      \[\leadsto \color{blue}{\sqrt{x}} \]
   4. Step-by-step derivation

      1. lower-sqrt.f64 94.3

         \[\leadsto \color{blue}{\sqrt{x}} \]

   5. Applied rewrites 94.3%

      \[\leadsto \color{blue}{\sqrt{x}} \]
3. Recombined 2 regimes into one program.
4. Add Preprocessing

Alternative 9: 97.2% accurate, 1.6× speedup

Math

\[\begin{array}{l} \\ \begin{array}{l} \mathbf{if}\;x \leq 4:\\ \;\;\;\;0.5 \cdot x\\ \mathbf{else}:\\ \;\;\;\;\sqrt{x}\\ \end{array} \end{array} \]

FPCore

(FPCore (x) :precision binary64 (if (<= x 4.0) (* 0.5 x) (sqrt x)))

C

double code(double x) {
	double tmp;
	if (x <= 4.0) {
		tmp = 0.5 * x;
	} else {
		tmp = sqrt(x);
	}
	return tmp;
}

Fortran

real(8) function code(x)
    real(8), intent (in) :: x
    real(8) :: tmp
    if (x <= 4.0d0) then
        tmp = 0.5d0 * x
    else
        tmp = sqrt(x)
    end if
    code = tmp
end function

Java

public static double code(double x) {
	double tmp;
	if (x <= 4.0) {
		tmp = 0.5 * x;
	} else {
		tmp = Math.sqrt(x);
	}
	return tmp;
}

Python

def code(x):
	tmp = 0
	if x <= 4.0:
		tmp = 0.5 * x
	else:
		tmp = math.sqrt(x)
	return tmp

Julia

function code(x)
	tmp = 0.0
	if (x <= 4.0)
		tmp = Float64(0.5 * x);
	else
		tmp = sqrt(x);
	end
	return tmp
end

MATLAB

function tmp_2 = code(x)
	tmp = 0.0;
	if (x <= 4.0)
		tmp = 0.5 * x;
	else
		tmp = sqrt(x);
	end
	tmp_2 = tmp;
end

Wolfram

code[x_] := If[LessEqual[x, 4.0], N[(0.5 * x), $MachinePrecision], N[Sqrt[x], $MachinePrecision]]

Derivation

1. Split input into 2 regimes

2. if x < 4

   1. Initial program 100.0%

      \[\frac{x}{1 + \sqrt{x + 1}} \]
   2. Add Preprocessing

   3. Taylor expanded in x around 0

      \[\leadsto \color{blue}{\frac{1}{2} \cdot x} \]
   4. Step-by-step derivation

      1. lower-*.f64 98.0

         \[\leadsto \color{blue}{0.5 \cdot x} \]

   5. Applied rewrites 98.0%

      \[\leadsto \color{blue}{0.5 \cdot x} \]

   if 4 < x

   1. Initial program 99.3%

      \[\frac{x}{1 + \sqrt{x + 1}} \]
   2. Add Preprocessing

   3. Taylor expanded in x around inf

      \[\leadsto \color{blue}{\sqrt{x}} \]
   4. Step-by-step derivation

      1. lower-sqrt.f64 94.3

         \[\leadsto \color{blue}{\sqrt{x}} \]

   5. Applied rewrites 94.3%

      \[\leadsto \color{blue}{\sqrt{x}} \]
3. Recombined 2 regimes into one program.
4. Add Preprocessing

Alternative 10: 67.5% accurate, 4.7× speedup

Math

\[\begin{array}{l} \\ 0.5 \cdot x \end{array} \]

FPCore

(FPCore (x) :precision binary64 (* 0.5 x))

C

double code(double x) {
	return 0.5 * x;
}

Fortran

real(8) function code(x)
    real(8), intent (in) :: x
    code = 0.5d0 * x
end function

Java

public static double code(double x) {
	return 0.5 * x;
}

Python

def code(x):
	return 0.5 * x

Julia

function code(x)
	return Float64(0.5 * x)
end

MATLAB

function tmp = code(x)
	tmp = 0.5 * x;
end

Wolfram

code[x_] := N[(0.5 * x), $MachinePrecision]

Derivation

1. Initial program 99.7%

   \[\frac{x}{1 + \sqrt{x + 1}} \]
2. Add Preprocessing

3. Taylor expanded in x around 0

   \[\leadsto \color{blue}{\frac{1}{2} \cdot x} \]
4. Step-by-step derivation

   1. lower-*.f64 62.1

      \[\leadsto \color{blue}{0.5 \cdot x} \]

5. Applied rewrites 62.1%

   \[\leadsto \color{blue}{0.5 \cdot x} \]
6. Add Preprocessing

Reproduce

herbie shell --seed 2024337 
(FPCore (x)
  :name "Numeric.Log:$clog1p from log-domain-0.10.2.1, B"
  :precision binary64
  (/ x (+ 1.0 (sqrt (+ x 1.0)))))