Jmat.Real.lambertw, newton loop step

Percentage Accurate: 78.4% → 98.1%

Time: 12.8s

Alternatives: 13

Speedup: 331.0×

• could not determine a ground truth

  1. wj = -3.3659190363029203e+179
  2. x = -3.957738439385491e-244

Specification

Math

\[\begin{array}{l} \\ \begin{array}{l} t_0 := wj \cdot e^{wj}\\ wj - \frac{t\_0 - x}{e^{wj} + t\_0} \end{array} \end{array} \]

FPCore

(FPCore (wj x)
 :precision binary64
 (let* ((t_0 (* wj (exp wj)))) (- wj (/ (- t_0 x) (+ (exp wj) t_0)))))

C

double code(double wj, double x) {
	double t_0 = wj * exp(wj);
	return wj - ((t_0 - x) / (exp(wj) + t_0));
}

Fortran

module fmin_fmax_functions
    implicit none
    private
    public fmax
    public fmin

    interface fmax
        module procedure fmax88
        module procedure fmax44
        module procedure fmax84
        module procedure fmax48
    end interface
    interface fmin
        module procedure fmin88
        module procedure fmin44
        module procedure fmin84
        module procedure fmin48
    end interface
contains
    real(8) function fmax88(x, y) result (res)
        real(8), intent (in) :: x
        real(8), intent (in) :: y
        res = merge(y, merge(x, max(x, y), y /= y), x /= x)
    end function
    real(4) function fmax44(x, y) result (res)
        real(4), intent (in) :: x
        real(4), intent (in) :: y
        res = merge(y, merge(x, max(x, y), y /= y), x /= x)
    end function
    real(8) function fmax84(x, y) result(res)
        real(8), intent (in) :: x
        real(4), intent (in) :: y
        res = merge(dble(y), merge(x, max(x, dble(y)), y /= y), x /= x)
    end function
    real(8) function fmax48(x, y) result(res)
        real(4), intent (in) :: x
        real(8), intent (in) :: y
        res = merge(y, merge(dble(x), max(dble(x), y), y /= y), x /= x)
    end function
    real(8) function fmin88(x, y) result (res)
        real(8), intent (in) :: x
        real(8), intent (in) :: y
        res = merge(y, merge(x, min(x, y), y /= y), x /= x)
    end function
    real(4) function fmin44(x, y) result (res)
        real(4), intent (in) :: x
        real(4), intent (in) :: y
        res = merge(y, merge(x, min(x, y), y /= y), x /= x)
    end function
    real(8) function fmin84(x, y) result(res)
        real(8), intent (in) :: x
        real(4), intent (in) :: y
        res = merge(dble(y), merge(x, min(x, dble(y)), y /= y), x /= x)
    end function
    real(8) function fmin48(x, y) result(res)
        real(4), intent (in) :: x
        real(8), intent (in) :: y
        res = merge(y, merge(dble(x), min(dble(x), y), y /= y), x /= x)
    end function
end module

real(8) function code(wj, x)
use fmin_fmax_functions
    real(8), intent (in) :: wj
    real(8), intent (in) :: x
    real(8) :: t_0
    t_0 = wj * exp(wj)
    code = wj - ((t_0 - x) / (exp(wj) + t_0))
end function

Java

public static double code(double wj, double x) {
	double t_0 = wj * Math.exp(wj);
	return wj - ((t_0 - x) / (Math.exp(wj) + t_0));
}

Python

def code(wj, x):
	t_0 = wj * math.exp(wj)
	return wj - ((t_0 - x) / (math.exp(wj) + t_0))

Julia

function code(wj, x)
	t_0 = Float64(wj * exp(wj))
	return Float64(wj - Float64(Float64(t_0 - x) / Float64(exp(wj) + t_0)))
end

MATLAB

function tmp = code(wj, x)
	t_0 = wj * exp(wj);
	tmp = wj - ((t_0 - x) / (exp(wj) + t_0));
end

Wolfram

code[wj_, x_] := Block[{t$95$0 = N[(wj * N[Exp[wj], $MachinePrecision]), $MachinePrecision]}, N[(wj - N[(N[(t$95$0 - x), $MachinePrecision] / N[(N[Exp[wj], $MachinePrecision] + t$95$0), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]]

Initial Program: 78.4% accurate, 1.0× speedup

Math

\[\begin{array}{l} \\ \begin{array}{l} t_0 := wj \cdot e^{wj}\\ wj - \frac{t\_0 - x}{e^{wj} + t\_0} \end{array} \end{array} \]

FPCore

(FPCore (wj x)
 :precision binary64
 (let* ((t_0 (* wj (exp wj)))) (- wj (/ (- t_0 x) (+ (exp wj) t_0)))))

C

double code(double wj, double x) {
	double t_0 = wj * exp(wj);
	return wj - ((t_0 - x) / (exp(wj) + t_0));
}

Fortran

module fmin_fmax_functions
    implicit none
    private
    public fmax
    public fmin

    interface fmax
        module procedure fmax88
        module procedure fmax44
        module procedure fmax84
        module procedure fmax48
    end interface
    interface fmin
        module procedure fmin88
        module procedure fmin44
        module procedure fmin84
        module procedure fmin48
    end interface
contains
    real(8) function fmax88(x, y) result (res)
        real(8), intent (in) :: x
        real(8), intent (in) :: y
        res = merge(y, merge(x, max(x, y), y /= y), x /= x)
    end function
    real(4) function fmax44(x, y) result (res)
        real(4), intent (in) :: x
        real(4), intent (in) :: y
        res = merge(y, merge(x, max(x, y), y /= y), x /= x)
    end function
    real(8) function fmax84(x, y) result(res)
        real(8), intent (in) :: x
        real(4), intent (in) :: y
        res = merge(dble(y), merge(x, max(x, dble(y)), y /= y), x /= x)
    end function
    real(8) function fmax48(x, y) result(res)
        real(4), intent (in) :: x
        real(8), intent (in) :: y
        res = merge(y, merge(dble(x), max(dble(x), y), y /= y), x /= x)
    end function
    real(8) function fmin88(x, y) result (res)
        real(8), intent (in) :: x
        real(8), intent (in) :: y
        res = merge(y, merge(x, min(x, y), y /= y), x /= x)
    end function
    real(4) function fmin44(x, y) result (res)
        real(4), intent (in) :: x
        real(4), intent (in) :: y
        res = merge(y, merge(x, min(x, y), y /= y), x /= x)
    end function
    real(8) function fmin84(x, y) result(res)
        real(8), intent (in) :: x
        real(4), intent (in) :: y
        res = merge(dble(y), merge(x, min(x, dble(y)), y /= y), x /= x)
    end function
    real(8) function fmin48(x, y) result(res)
        real(4), intent (in) :: x
        real(8), intent (in) :: y
        res = merge(y, merge(dble(x), min(dble(x), y), y /= y), x /= x)
    end function
end module

real(8) function code(wj, x)
use fmin_fmax_functions
    real(8), intent (in) :: wj
    real(8), intent (in) :: x
    real(8) :: t_0
    t_0 = wj * exp(wj)
    code = wj - ((t_0 - x) / (exp(wj) + t_0))
end function

Java

public static double code(double wj, double x) {
	double t_0 = wj * Math.exp(wj);
	return wj - ((t_0 - x) / (Math.exp(wj) + t_0));
}

Python

def code(wj, x):
	t_0 = wj * math.exp(wj)
	return wj - ((t_0 - x) / (math.exp(wj) + t_0))

Julia

function code(wj, x)
	t_0 = Float64(wj * exp(wj))
	return Float64(wj - Float64(Float64(t_0 - x) / Float64(exp(wj) + t_0)))
end

MATLAB

function tmp = code(wj, x)
	t_0 = wj * exp(wj);
	tmp = wj - ((t_0 - x) / (exp(wj) + t_0));
end

Wolfram

code[wj_, x_] := Block[{t$95$0 = N[(wj * N[Exp[wj], $MachinePrecision]), $MachinePrecision]}, N[(wj - N[(N[(t$95$0 - x), $MachinePrecision] / N[(N[Exp[wj], $MachinePrecision] + t$95$0), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]]

Alternative 1: 98.1% accurate, 0.4× speedup

Math

\[\begin{array}{l} \\ \begin{array}{l} t_0 := wj \cdot e^{wj}\\ \mathbf{if}\;wj - \frac{t\_0 - x}{e^{wj} + t\_0} \leq 10^{+302}:\\ \;\;\;\;\mathsf{fma}\left(\mathsf{fma}\left(\frac{wj}{x} - \frac{\frac{e^{wj}}{x}}{e^{wj}}, wj \cdot wj, \frac{wj}{x}\right) \cdot wj, x, \frac{x}{\mathsf{fma}\left(e^{wj}, wj, e^{wj}\right)}\right)\\ \mathbf{else}:\\ \;\;\;\;\mathsf{fma}\left(\frac{wj}{x} - \frac{wj}{\mathsf{fma}\left(x, wj, x\right)} \cdot e^{0}, x, \frac{x}{e^{wj} \cdot wj}\right)\\ \end{array} \end{array} \]

FPCore

(FPCore (wj x)
 :precision binary64
 (let* ((t_0 (* wj (exp wj))))
   (if (<= (- wj (/ (- t_0 x) (+ (exp wj) t_0))) 1e+302)
     (fma
      (* (fma (- (/ wj x) (/ (/ (exp wj) x) (exp wj))) (* wj wj) (/ wj x)) wj)
      x
      (/ x (fma (exp wj) wj (exp wj))))
     (fma
      (- (/ wj x) (* (/ wj (fma x wj x)) (exp 0.0)))
      x
      (/ x (* (exp wj) wj))))))

C

double code(double wj, double x) {
	double t_0 = wj * exp(wj);
	double tmp;
	if ((wj - ((t_0 - x) / (exp(wj) + t_0))) <= 1e+302) {
		tmp = fma((fma(((wj / x) - ((exp(wj) / x) / exp(wj))), (wj * wj), (wj / x)) * wj), x, (x / fma(exp(wj), wj, exp(wj))));
	} else {
		tmp = fma(((wj / x) - ((wj / fma(x, wj, x)) * exp(0.0))), x, (x / (exp(wj) * wj)));
	}
	return tmp;
}

Julia

function code(wj, x)
	t_0 = Float64(wj * exp(wj))
	tmp = 0.0
	if (Float64(wj - Float64(Float64(t_0 - x) / Float64(exp(wj) + t_0))) <= 1e+302)
		tmp = fma(Float64(fma(Float64(Float64(wj / x) - Float64(Float64(exp(wj) / x) / exp(wj))), Float64(wj * wj), Float64(wj / x)) * wj), x, Float64(x / fma(exp(wj), wj, exp(wj))));
	else
		tmp = fma(Float64(Float64(wj / x) - Float64(Float64(wj / fma(x, wj, x)) * exp(0.0))), x, Float64(x / Float64(exp(wj) * wj)));
	end
	return tmp
end

Wolfram

code[wj_, x_] := Block[{t$95$0 = N[(wj * N[Exp[wj], $MachinePrecision]), $MachinePrecision]}, If[LessEqual[N[(wj - N[(N[(t$95$0 - x), $MachinePrecision] / N[(N[Exp[wj], $MachinePrecision] + t$95$0), $MachinePrecision]), $MachinePrecision]), $MachinePrecision], 1e+302], N[(N[(N[(N[(N[(wj / x), $MachinePrecision] - N[(N[(N[Exp[wj], $MachinePrecision] / x), $MachinePrecision] / N[Exp[wj], $MachinePrecision]), $MachinePrecision]), $MachinePrecision] * N[(wj * wj), $MachinePrecision] + N[(wj / x), $MachinePrecision]), $MachinePrecision] * wj), $MachinePrecision] * x + N[(x / N[(N[Exp[wj], $MachinePrecision] * wj + N[Exp[wj], $MachinePrecision]), $MachinePrecision]), $MachinePrecision]), $MachinePrecision], N[(N[(N[(wj / x), $MachinePrecision] - N[(N[(wj / N[(x * wj + x), $MachinePrecision]), $MachinePrecision] * N[Exp[0.0], $MachinePrecision]), $MachinePrecision]), $MachinePrecision] * x + N[(x / N[(N[Exp[wj], $MachinePrecision] * wj), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]]]

Derivation

1. Split input into 2 regimes

2. if (-.f64 wj (/.f64 (-.f64 (*.f64 wj (exp.f64 wj)) x) (+.f64 (exp.f64 wj) (*.f64 wj (exp.f64 wj))))) < 1.0000000000000001e302

   1. Initial program 79.2%

      \[wj - \frac{wj \cdot e^{wj} - x}{e^{wj} + wj \cdot e^{wj}} \]
   2. Add Preprocessing

   3. Taylor expanded in x around inf

      \[\leadsto \color{blue}{x \cdot \left(\left(\frac{1}{e^{wj} + wj \cdot e^{wj}} + \frac{wj}{x}\right) - \frac{wj \cdot e^{wj}}{x \cdot \left(e^{wj} + wj \cdot e^{wj}\right)}\right)} \]

   5. Applied rewrites 90.5%

      \[\leadsto \color{blue}{\mathsf{fma}\left(\frac{wj}{x} - \frac{wj}{\mathsf{fma}\left(x, wj, x\right)} \cdot e^{wj - wj}, x, \frac{x}{\mathsf{fma}\left(e^{wj}, wj, e^{wj}\right)}\right)} \]

   6. Taylor expanded in wj around 0

      \[\leadsto \mathsf{fma}\left(\frac{{wj}^{2}}{x}, x, \frac{x}{\mathsf{fma}\left(e^{wj}, wj, e^{wj}\right)}\right) \]
   7. Step-by-step derivation

   8. Applied rewrites 98.9%

      \[\leadsto \mathsf{fma}\left(\frac{wj \cdot wj}{x}, x, \frac{x}{\mathsf{fma}\left(e^{wj}, wj, e^{wj}\right)}\right) \]

   9. Taylor expanded in wj around 0

      \[\leadsto \mathsf{fma}\left({wj}^{2} \cdot \left(wj \cdot \left(\frac{wj}{x} - \frac{1}{x}\right) + \frac{1}{x}\right), x, \frac{x}{\mathsf{fma}\left(e^{wj}, wj, e^{wj}\right)}\right) \]

   11. Applied rewrites 99.8%

       \[\leadsto \mathsf{fma}\left(\mathsf{fma}\left(\left(wj \cdot wj\right) \cdot \left(\frac{wj}{x} - \frac{e^{wj}}{e^{wj} \cdot x}\right), wj, \frac{wj}{x} \cdot wj\right), x, \frac{x}{\mathsf{fma}\left(e^{wj}, wj, e^{wj}\right)}\right) \]

   13. Applied rewrites 99.8%

       \[\leadsto \color{blue}{\mathsf{fma}\left(\mathsf{fma}\left(\frac{wj}{x} - \frac{\frac{e^{wj}}{x}}{e^{wj}}, wj \cdot wj, \frac{wj}{x}\right) \cdot wj, x, \frac{x}{\mathsf{fma}\left(e^{wj}, wj, e^{wj}\right)}\right)} \]

   if 1.0000000000000001e302 < (-.f64 wj (/.f64 (-.f64 (*.f64 wj (exp.f64 wj)) x) (+.f64 (exp.f64 wj) (*.f64 wj (exp.f64 wj)))))

   1. Initial program 0.0%

      \[wj - \frac{wj \cdot e^{wj} - x}{e^{wj} + wj \cdot e^{wj}} \]
   2. Add Preprocessing

   3. Taylor expanded in x around inf

      \[\leadsto \color{blue}{x \cdot \left(\left(\frac{1}{e^{wj} + wj \cdot e^{wj}} + \frac{wj}{x}\right) - \frac{wj \cdot e^{wj}}{x \cdot \left(e^{wj} + wj \cdot e^{wj}\right)}\right)} \]

   5. Applied rewrites 82.4%

      \[\leadsto \color{blue}{\mathsf{fma}\left(\frac{wj}{x} - \frac{wj}{\mathsf{fma}\left(x, wj, x\right)} \cdot e^{wj - wj}, x, \frac{x}{\mathsf{fma}\left(e^{wj}, wj, e^{wj}\right)}\right)} \]

   6. Taylor expanded in wj around inf

      \[\leadsto \mathsf{fma}\left(\frac{wj}{x} - \frac{wj}{\mathsf{fma}\left(x, wj, x\right)} \cdot e^{wj - wj}, x, \frac{x}{wj \cdot e^{wj}}\right) \]
   7. Step-by-step derivation

   8. Applied rewrites 99.1%

      \[\leadsto \mathsf{fma}\left(\frac{wj}{x} - \frac{wj}{\mathsf{fma}\left(x, wj, x\right)} \cdot e^{wj - wj}, x, \frac{x}{e^{wj} \cdot wj}\right) \]
3. Recombined 2 regimes into one program.

4. Final simplification 99.8%

   \[\leadsto \begin{array}{l} \mathbf{if}\;wj - \frac{wj \cdot e^{wj} - x}{e^{wj} + wj \cdot e^{wj}} \leq 10^{+302}:\\ \;\;\;\;\mathsf{fma}\left(\mathsf{fma}\left(\frac{wj}{x} - \frac{\frac{e^{wj}}{x}}{e^{wj}}, wj \cdot wj, \frac{wj}{x}\right) \cdot wj, x, \frac{x}{\mathsf{fma}\left(e^{wj}, wj, e^{wj}\right)}\right)\\ \mathbf{else}:\\ \;\;\;\;\mathsf{fma}\left(\frac{wj}{x} - \frac{wj}{\mathsf{fma}\left(x, wj, x\right)} \cdot e^{0}, x, \frac{x}{e^{wj} \cdot wj}\right)\\ \end{array} \]
5. Add Preprocessing

Alternative 2: 98.1% accurate, 0.6× speedup

Math

\[\begin{array}{l} \\ \begin{array}{l} t_0 := wj \cdot e^{wj}\\ \mathbf{if}\;wj - \frac{t\_0 - x}{e^{wj} + t\_0} \leq 10^{+302}:\\ \;\;\;\;\mathsf{fma}\left(\frac{\mathsf{fma}\left(\mathsf{fma}\left(-wj, wj, wj\right), -wj, wj\right)}{x} \cdot wj, x, \frac{x}{\mathsf{fma}\left(e^{wj}, wj, e^{wj}\right)}\right)\\ \mathbf{else}:\\ \;\;\;\;\mathsf{fma}\left(\frac{wj}{x} - \frac{wj}{\mathsf{fma}\left(x, wj, x\right)} \cdot e^{0}, x, \frac{x}{e^{wj} \cdot wj}\right)\\ \end{array} \end{array} \]

FPCore

(FPCore (wj x)
 :precision binary64
 (let* ((t_0 (* wj (exp wj))))
   (if (<= (- wj (/ (- t_0 x) (+ (exp wj) t_0))) 1e+302)
     (fma
      (* (/ (fma (fma (- wj) wj wj) (- wj) wj) x) wj)
      x
      (/ x (fma (exp wj) wj (exp wj))))
     (fma
      (- (/ wj x) (* (/ wj (fma x wj x)) (exp 0.0)))
      x
      (/ x (* (exp wj) wj))))))

C

double code(double wj, double x) {
	double t_0 = wj * exp(wj);
	double tmp;
	if ((wj - ((t_0 - x) / (exp(wj) + t_0))) <= 1e+302) {
		tmp = fma(((fma(fma(-wj, wj, wj), -wj, wj) / x) * wj), x, (x / fma(exp(wj), wj, exp(wj))));
	} else {
		tmp = fma(((wj / x) - ((wj / fma(x, wj, x)) * exp(0.0))), x, (x / (exp(wj) * wj)));
	}
	return tmp;
}

Julia

function code(wj, x)
	t_0 = Float64(wj * exp(wj))
	tmp = 0.0
	if (Float64(wj - Float64(Float64(t_0 - x) / Float64(exp(wj) + t_0))) <= 1e+302)
		tmp = fma(Float64(Float64(fma(fma(Float64(-wj), wj, wj), Float64(-wj), wj) / x) * wj), x, Float64(x / fma(exp(wj), wj, exp(wj))));
	else
		tmp = fma(Float64(Float64(wj / x) - Float64(Float64(wj / fma(x, wj, x)) * exp(0.0))), x, Float64(x / Float64(exp(wj) * wj)));
	end
	return tmp
end

Wolfram

code[wj_, x_] := Block[{t$95$0 = N[(wj * N[Exp[wj], $MachinePrecision]), $MachinePrecision]}, If[LessEqual[N[(wj - N[(N[(t$95$0 - x), $MachinePrecision] / N[(N[Exp[wj], $MachinePrecision] + t$95$0), $MachinePrecision]), $MachinePrecision]), $MachinePrecision], 1e+302], N[(N[(N[(N[(N[((-wj) * wj + wj), $MachinePrecision] * (-wj) + wj), $MachinePrecision] / x), $MachinePrecision] * wj), $MachinePrecision] * x + N[(x / N[(N[Exp[wj], $MachinePrecision] * wj + N[Exp[wj], $MachinePrecision]), $MachinePrecision]), $MachinePrecision]), $MachinePrecision], N[(N[(N[(wj / x), $MachinePrecision] - N[(N[(wj / N[(x * wj + x), $MachinePrecision]), $MachinePrecision] * N[Exp[0.0], $MachinePrecision]), $MachinePrecision]), $MachinePrecision] * x + N[(x / N[(N[Exp[wj], $MachinePrecision] * wj), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]]]

Derivation

1. Split input into 2 regimes

2. if (-.f64 wj (/.f64 (-.f64 (*.f64 wj (exp.f64 wj)) x) (+.f64 (exp.f64 wj) (*.f64 wj (exp.f64 wj))))) < 1.0000000000000001e302

   1. Initial program 79.2%

      \[wj - \frac{wj \cdot e^{wj} - x}{e^{wj} + wj \cdot e^{wj}} \]
   2. Add Preprocessing

   3. Taylor expanded in x around inf

      \[\leadsto \color{blue}{x \cdot \left(\left(\frac{1}{e^{wj} + wj \cdot e^{wj}} + \frac{wj}{x}\right) - \frac{wj \cdot e^{wj}}{x \cdot \left(e^{wj} + wj \cdot e^{wj}\right)}\right)} \]

   5. Applied rewrites 90.5%

      \[\leadsto \color{blue}{\mathsf{fma}\left(\frac{wj}{x} - \frac{wj}{\mathsf{fma}\left(x, wj, x\right)} \cdot e^{wj - wj}, x, \frac{x}{\mathsf{fma}\left(e^{wj}, wj, e^{wj}\right)}\right)} \]

   6. Taylor expanded in wj around 0

      \[\leadsto \mathsf{fma}\left(\frac{{wj}^{2}}{x}, x, \frac{x}{\mathsf{fma}\left(e^{wj}, wj, e^{wj}\right)}\right) \]
   7. Step-by-step derivation

   8. Applied rewrites 98.9%

      \[\leadsto \mathsf{fma}\left(\frac{wj \cdot wj}{x}, x, \frac{x}{\mathsf{fma}\left(e^{wj}, wj, e^{wj}\right)}\right) \]

   9. Taylor expanded in wj around 0

      \[\leadsto \mathsf{fma}\left({wj}^{2} \cdot \left(wj \cdot \left(\frac{wj}{x} - \frac{1}{x}\right) + \frac{1}{x}\right), x, \frac{x}{\mathsf{fma}\left(e^{wj}, wj, e^{wj}\right)}\right) \]

   11. Applied rewrites 99.8%

       \[\leadsto \mathsf{fma}\left(\mathsf{fma}\left(\left(wj \cdot wj\right) \cdot \left(\frac{wj}{x} - \frac{e^{wj}}{e^{wj} \cdot x}\right), wj, \frac{wj}{x} \cdot wj\right), x, \frac{x}{\mathsf{fma}\left(e^{wj}, wj, e^{wj}\right)}\right) \]

   12. Taylor expanded in wj around 0

       \[\leadsto \mathsf{fma}\left({wj}^{2} \cdot \left(wj \cdot \left(\frac{wj}{x} - \frac{1}{x}\right) + \frac{1}{x}\right), x, \frac{x}{\mathsf{fma}\left(e^{wj}, wj, e^{wj}\right)}\right) \]

   14. Applied rewrites 99.8%

       \[\leadsto \mathsf{fma}\left(\frac{\mathsf{fma}\left(\mathsf{fma}\left(-wj, wj, wj\right), -wj, wj\right)}{x} \cdot wj, x, \frac{x}{\mathsf{fma}\left(e^{wj}, wj, e^{wj}\right)}\right) \]

   if 1.0000000000000001e302 < (-.f64 wj (/.f64 (-.f64 (*.f64 wj (exp.f64 wj)) x) (+.f64 (exp.f64 wj) (*.f64 wj (exp.f64 wj)))))

   1. Initial program 0.0%

      \[wj - \frac{wj \cdot e^{wj} - x}{e^{wj} + wj \cdot e^{wj}} \]
   2. Add Preprocessing

   3. Taylor expanded in x around inf

      \[\leadsto \color{blue}{x \cdot \left(\left(\frac{1}{e^{wj} + wj \cdot e^{wj}} + \frac{wj}{x}\right) - \frac{wj \cdot e^{wj}}{x \cdot \left(e^{wj} + wj \cdot e^{wj}\right)}\right)} \]

   5. Applied rewrites 82.4%

      \[\leadsto \color{blue}{\mathsf{fma}\left(\frac{wj}{x} - \frac{wj}{\mathsf{fma}\left(x, wj, x\right)} \cdot e^{wj - wj}, x, \frac{x}{\mathsf{fma}\left(e^{wj}, wj, e^{wj}\right)}\right)} \]

   6. Taylor expanded in wj around inf

      \[\leadsto \mathsf{fma}\left(\frac{wj}{x} - \frac{wj}{\mathsf{fma}\left(x, wj, x\right)} \cdot e^{wj - wj}, x, \frac{x}{wj \cdot e^{wj}}\right) \]
   7. Step-by-step derivation

   8. Applied rewrites 99.1%

      \[\leadsto \mathsf{fma}\left(\frac{wj}{x} - \frac{wj}{\mathsf{fma}\left(x, wj, x\right)} \cdot e^{wj - wj}, x, \frac{x}{e^{wj} \cdot wj}\right) \]
3. Recombined 2 regimes into one program.

4. Final simplification 99.7%

   \[\leadsto \begin{array}{l} \mathbf{if}\;wj - \frac{wj \cdot e^{wj} - x}{e^{wj} + wj \cdot e^{wj}} \leq 10^{+302}:\\ \;\;\;\;\mathsf{fma}\left(\frac{\mathsf{fma}\left(\mathsf{fma}\left(-wj, wj, wj\right), -wj, wj\right)}{x} \cdot wj, x, \frac{x}{\mathsf{fma}\left(e^{wj}, wj, e^{wj}\right)}\right)\\ \mathbf{else}:\\ \;\;\;\;\mathsf{fma}\left(\frac{wj}{x} - \frac{wj}{\mathsf{fma}\left(x, wj, x\right)} \cdot e^{0}, x, \frac{x}{e^{wj} \cdot wj}\right)\\ \end{array} \]
5. Add Preprocessing

Alternative 3: 98.6% accurate, 0.6× speedup

Math

\[\begin{array}{l} \\ \begin{array}{l} t_0 := \mathsf{fma}\left(e^{wj}, wj, e^{wj}\right)\\ t_1 := wj \cdot e^{wj}\\ \mathbf{if}\;wj - \frac{t\_1 - x}{e^{wj} + t\_1} \leq 2 \cdot 10^{-11}:\\ \;\;\;\;\mathsf{fma}\left(\frac{\mathsf{fma}\left(\mathsf{fma}\left(-wj, wj, wj\right), -wj, wj\right)}{x} \cdot wj, x, \frac{x}{t\_0}\right)\\ \mathbf{else}:\\ \;\;\;\;wj - \mathsf{fma}\left(\frac{wj}{\mathsf{fma}\left(x, wj, x\right)}, x, \frac{-x}{t\_0}\right)\\ \end{array} \end{array} \]

FPCore

(FPCore (wj x)
 :precision binary64
 (let* ((t_0 (fma (exp wj) wj (exp wj))) (t_1 (* wj (exp wj))))
   (if (<= (- wj (/ (- t_1 x) (+ (exp wj) t_1))) 2e-11)
     (fma (* (/ (fma (fma (- wj) wj wj) (- wj) wj) x) wj) x (/ x t_0))
     (- wj (fma (/ wj (fma x wj x)) x (/ (- x) t_0))))))

C

double code(double wj, double x) {
	double t_0 = fma(exp(wj), wj, exp(wj));
	double t_1 = wj * exp(wj);
	double tmp;
	if ((wj - ((t_1 - x) / (exp(wj) + t_1))) <= 2e-11) {
		tmp = fma(((fma(fma(-wj, wj, wj), -wj, wj) / x) * wj), x, (x / t_0));
	} else {
		tmp = wj - fma((wj / fma(x, wj, x)), x, (-x / t_0));
	}
	return tmp;
}

Julia

function code(wj, x)
	t_0 = fma(exp(wj), wj, exp(wj))
	t_1 = Float64(wj * exp(wj))
	tmp = 0.0
	if (Float64(wj - Float64(Float64(t_1 - x) / Float64(exp(wj) + t_1))) <= 2e-11)
		tmp = fma(Float64(Float64(fma(fma(Float64(-wj), wj, wj), Float64(-wj), wj) / x) * wj), x, Float64(x / t_0));
	else
		tmp = Float64(wj - fma(Float64(wj / fma(x, wj, x)), x, Float64(Float64(-x) / t_0)));
	end
	return tmp
end

Wolfram

code[wj_, x_] := Block[{t$95$0 = N[(N[Exp[wj], $MachinePrecision] * wj + N[Exp[wj], $MachinePrecision]), $MachinePrecision]}, Block[{t$95$1 = N[(wj * N[Exp[wj], $MachinePrecision]), $MachinePrecision]}, If[LessEqual[N[(wj - N[(N[(t$95$1 - x), $MachinePrecision] / N[(N[Exp[wj], $MachinePrecision] + t$95$1), $MachinePrecision]), $MachinePrecision]), $MachinePrecision], 2e-11], N[(N[(N[(N[(N[((-wj) * wj + wj), $MachinePrecision] * (-wj) + wj), $MachinePrecision] / x), $MachinePrecision] * wj), $MachinePrecision] * x + N[(x / t$95$0), $MachinePrecision]), $MachinePrecision], N[(wj - N[(N[(wj / N[(x * wj + x), $MachinePrecision]), $MachinePrecision] * x + N[((-x) / t$95$0), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]]]]

Derivation

1. Split input into 2 regimes

2. if (-.f64 wj (/.f64 (-.f64 (*.f64 wj (exp.f64 wj)) x) (+.f64 (exp.f64 wj) (*.f64 wj (exp.f64 wj))))) < 1.99999999999999988e-11

   1. Initial program 70.9%

      \[wj - \frac{wj \cdot e^{wj} - x}{e^{wj} + wj \cdot e^{wj}} \]
   2. Add Preprocessing

   3. Taylor expanded in x around inf

      \[\leadsto \color{blue}{x \cdot \left(\left(\frac{1}{e^{wj} + wj \cdot e^{wj}} + \frac{wj}{x}\right) - \frac{wj \cdot e^{wj}}{x \cdot \left(e^{wj} + wj \cdot e^{wj}\right)}\right)} \]

   5. Applied rewrites 86.8%

      \[\leadsto \color{blue}{\mathsf{fma}\left(\frac{wj}{x} - \frac{wj}{\mathsf{fma}\left(x, wj, x\right)} \cdot e^{wj - wj}, x, \frac{x}{\mathsf{fma}\left(e^{wj}, wj, e^{wj}\right)}\right)} \]

   6. Taylor expanded in wj around 0

      \[\leadsto \mathsf{fma}\left(\frac{{wj}^{2}}{x}, x, \frac{x}{\mathsf{fma}\left(e^{wj}, wj, e^{wj}\right)}\right) \]
   7. Step-by-step derivation

   8. Applied rewrites 98.9%

      \[\leadsto \mathsf{fma}\left(\frac{wj \cdot wj}{x}, x, \frac{x}{\mathsf{fma}\left(e^{wj}, wj, e^{wj}\right)}\right) \]

   9. Taylor expanded in wj around 0

      \[\leadsto \mathsf{fma}\left({wj}^{2} \cdot \left(wj \cdot \left(\frac{wj}{x} - \frac{1}{x}\right) + \frac{1}{x}\right), x, \frac{x}{\mathsf{fma}\left(e^{wj}, wj, e^{wj}\right)}\right) \]

   11. Applied rewrites 99.9%

       \[\leadsto \mathsf{fma}\left(\mathsf{fma}\left(\left(wj \cdot wj\right) \cdot \left(\frac{wj}{x} - \frac{e^{wj}}{e^{wj} \cdot x}\right), wj, \frac{wj}{x} \cdot wj\right), x, \frac{x}{\mathsf{fma}\left(e^{wj}, wj, e^{wj}\right)}\right) \]

   12. Taylor expanded in wj around 0

       \[\leadsto \mathsf{fma}\left({wj}^{2} \cdot \left(wj \cdot \left(\frac{wj}{x} - \frac{1}{x}\right) + \frac{1}{x}\right), x, \frac{x}{\mathsf{fma}\left(e^{wj}, wj, e^{wj}\right)}\right) \]

   14. Applied rewrites 99.9%

       \[\leadsto \mathsf{fma}\left(\frac{\mathsf{fma}\left(\mathsf{fma}\left(-wj, wj, wj\right), -wj, wj\right)}{x} \cdot wj, x, \frac{x}{\mathsf{fma}\left(e^{wj}, wj, e^{wj}\right)}\right) \]

   if 1.99999999999999988e-11 < (-.f64 wj (/.f64 (-.f64 (*.f64 wj (exp.f64 wj)) x) (+.f64 (exp.f64 wj) (*.f64 wj (exp.f64 wj)))))

   1. Initial program 92.2%

      \[wj - \frac{wj \cdot e^{wj} - x}{e^{wj} + wj \cdot e^{wj}} \]
   2. Add Preprocessing

   3. Taylor expanded in wj around 0

      \[\leadsto wj - \frac{wj \cdot e^{wj} - x}{e^{wj} + \color{blue}{wj}} \]
   4. Step-by-step derivation

   5. Applied rewrites 91.1%

      \[\leadsto wj - \frac{wj \cdot e^{wj} - x}{e^{wj} + \color{blue}{wj}} \]

   6. Taylor expanded in wj around 0

      \[\leadsto wj - \frac{\color{blue}{wj} - x}{e^{wj} + wj} \]
   7. Step-by-step derivation

   8. Applied rewrites 93.7%

      \[\leadsto wj - \frac{\color{blue}{wj} - x}{e^{wj} + wj} \]

   9. Taylor expanded in wj around 0

      \[\leadsto wj - \frac{\color{blue}{wj \cdot \left(1 + wj\right)} - x}{e^{wj} + wj} \]
   10. Step-by-step derivation

   11. Applied rewrites 93.9%

       \[\leadsto wj - \frac{\color{blue}{\mathsf{fma}\left(wj, wj, wj\right)} - x}{e^{wj} + wj} \]

   12. Taylor expanded in x around -inf

       \[\leadsto wj - \color{blue}{-1 \cdot \left(x \cdot \left(-1 \cdot \frac{wj \cdot e^{wj}}{x \cdot \left(e^{wj} + wj \cdot e^{wj}\right)} + \frac{1}{e^{wj} + wj \cdot e^{wj}}\right)\right)} \]

   14. Applied rewrites 98.6%

       \[\leadsto wj - \color{blue}{\mathsf{fma}\left(\frac{wj}{\mathsf{fma}\left(x, wj, x\right)}, x, \frac{-x}{\mathsf{fma}\left(e^{wj}, wj, e^{wj}\right)}\right)} \]
3. Recombined 2 regimes into one program.
4. Add Preprocessing

Alternative 4: 98.5% accurate, 0.6× speedup

Math

\[\begin{array}{l} \\ \begin{array}{l} t_0 := \mathsf{fma}\left(e^{wj}, wj, e^{wj}\right)\\ t_1 := wj \cdot e^{wj}\\ \mathbf{if}\;wj - \frac{t\_1 - x}{e^{wj} + t\_1} \leq 2.6 \cdot 10^{-14}:\\ \;\;\;\;\mathsf{fma}\left(\frac{\mathsf{fma}\left(-wj, wj, wj\right)}{x} \cdot wj, x, \frac{x}{t\_0}\right)\\ \mathbf{else}:\\ \;\;\;\;wj - \mathsf{fma}\left(\frac{wj}{\mathsf{fma}\left(x, wj, x\right)}, x, \frac{-x}{t\_0}\right)\\ \end{array} \end{array} \]

FPCore

(FPCore (wj x)
 :precision binary64
 (let* ((t_0 (fma (exp wj) wj (exp wj))) (t_1 (* wj (exp wj))))
   (if (<= (- wj (/ (- t_1 x) (+ (exp wj) t_1))) 2.6e-14)
     (fma (* (/ (fma (- wj) wj wj) x) wj) x (/ x t_0))
     (- wj (fma (/ wj (fma x wj x)) x (/ (- x) t_0))))))

C

double code(double wj, double x) {
	double t_0 = fma(exp(wj), wj, exp(wj));
	double t_1 = wj * exp(wj);
	double tmp;
	if ((wj - ((t_1 - x) / (exp(wj) + t_1))) <= 2.6e-14) {
		tmp = fma(((fma(-wj, wj, wj) / x) * wj), x, (x / t_0));
	} else {
		tmp = wj - fma((wj / fma(x, wj, x)), x, (-x / t_0));
	}
	return tmp;
}

Julia

function code(wj, x)
	t_0 = fma(exp(wj), wj, exp(wj))
	t_1 = Float64(wj * exp(wj))
	tmp = 0.0
	if (Float64(wj - Float64(Float64(t_1 - x) / Float64(exp(wj) + t_1))) <= 2.6e-14)
		tmp = fma(Float64(Float64(fma(Float64(-wj), wj, wj) / x) * wj), x, Float64(x / t_0));
	else
		tmp = Float64(wj - fma(Float64(wj / fma(x, wj, x)), x, Float64(Float64(-x) / t_0)));
	end
	return tmp
end

Wolfram

code[wj_, x_] := Block[{t$95$0 = N[(N[Exp[wj], $MachinePrecision] * wj + N[Exp[wj], $MachinePrecision]), $MachinePrecision]}, Block[{t$95$1 = N[(wj * N[Exp[wj], $MachinePrecision]), $MachinePrecision]}, If[LessEqual[N[(wj - N[(N[(t$95$1 - x), $MachinePrecision] / N[(N[Exp[wj], $MachinePrecision] + t$95$1), $MachinePrecision]), $MachinePrecision]), $MachinePrecision], 2.6e-14], N[(N[(N[(N[((-wj) * wj + wj), $MachinePrecision] / x), $MachinePrecision] * wj), $MachinePrecision] * x + N[(x / t$95$0), $MachinePrecision]), $MachinePrecision], N[(wj - N[(N[(wj / N[(x * wj + x), $MachinePrecision]), $MachinePrecision] * x + N[((-x) / t$95$0), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]]]]

Derivation

1. Split input into 2 regimes

2. if (-.f64 wj (/.f64 (-.f64 (*.f64 wj (exp.f64 wj)) x) (+.f64 (exp.f64 wj) (*.f64 wj (exp.f64 wj))))) < 2.59999999999999997e-14

   1. Initial program 70.7%

      \[wj - \frac{wj \cdot e^{wj} - x}{e^{wj} + wj \cdot e^{wj}} \]
   2. Add Preprocessing

   3. Taylor expanded in x around inf

      \[\leadsto \color{blue}{x \cdot \left(\left(\frac{1}{e^{wj} + wj \cdot e^{wj}} + \frac{wj}{x}\right) - \frac{wj \cdot e^{wj}}{x \cdot \left(e^{wj} + wj \cdot e^{wj}\right)}\right)} \]

   5. Applied rewrites 86.8%

      \[\leadsto \color{blue}{\mathsf{fma}\left(\frac{wj}{x} - \frac{wj}{\mathsf{fma}\left(x, wj, x\right)} \cdot e^{wj - wj}, x, \frac{x}{\mathsf{fma}\left(e^{wj}, wj, e^{wj}\right)}\right)} \]

   6. Taylor expanded in wj around 0

      \[\leadsto \mathsf{fma}\left(\frac{{wj}^{2}}{x}, x, \frac{x}{\mathsf{fma}\left(e^{wj}, wj, e^{wj}\right)}\right) \]
   7. Step-by-step derivation

   8. Applied rewrites 99.1%

      \[\leadsto \mathsf{fma}\left(\frac{wj \cdot wj}{x}, x, \frac{x}{\mathsf{fma}\left(e^{wj}, wj, e^{wj}\right)}\right) \]

   9. Taylor expanded in wj around 0

      \[\leadsto \mathsf{fma}\left({wj}^{2} \cdot \left(-1 \cdot \frac{wj}{x} + \frac{1}{x}\right), x, \frac{x}{\mathsf{fma}\left(e^{wj}, wj, e^{wj}\right)}\right) \]
   10. Step-by-step derivation

   11. Applied rewrites 99.8%

       \[\leadsto \mathsf{fma}\left(\frac{\mathsf{fma}\left(-wj, wj, wj\right)}{x} \cdot wj, x, \frac{x}{\mathsf{fma}\left(e^{wj}, wj, e^{wj}\right)}\right) \]

   if 2.59999999999999997e-14 < (-.f64 wj (/.f64 (-.f64 (*.f64 wj (exp.f64 wj)) x) (+.f64 (exp.f64 wj) (*.f64 wj (exp.f64 wj)))))

   1. Initial program 92.1%

      \[wj - \frac{wj \cdot e^{wj} - x}{e^{wj} + wj \cdot e^{wj}} \]
   2. Add Preprocessing

   3. Taylor expanded in wj around 0

      \[\leadsto wj - \frac{wj \cdot e^{wj} - x}{e^{wj} + \color{blue}{wj}} \]
   4. Step-by-step derivation

   5. Applied rewrites 90.6%

      \[\leadsto wj - \frac{wj \cdot e^{wj} - x}{e^{wj} + \color{blue}{wj}} \]

   6. Taylor expanded in wj around 0

      \[\leadsto wj - \frac{\color{blue}{wj} - x}{e^{wj} + wj} \]
   7. Step-by-step derivation

   8. Applied rewrites 92.9%

      \[\leadsto wj - \frac{\color{blue}{wj} - x}{e^{wj} + wj} \]

   9. Taylor expanded in wj around 0

      \[\leadsto wj - \frac{\color{blue}{wj \cdot \left(1 + wj\right)} - x}{e^{wj} + wj} \]
   10. Step-by-step derivation

   11. Applied rewrites 93.4%

       \[\leadsto wj - \frac{\color{blue}{\mathsf{fma}\left(wj, wj, wj\right)} - x}{e^{wj} + wj} \]

   12. Taylor expanded in x around -inf

       \[\leadsto wj - \color{blue}{-1 \cdot \left(x \cdot \left(-1 \cdot \frac{wj \cdot e^{wj}}{x \cdot \left(e^{wj} + wj \cdot e^{wj}\right)} + \frac{1}{e^{wj} + wj \cdot e^{wj}}\right)\right)} \]

   14. Applied rewrites 98.3%

       \[\leadsto wj - \color{blue}{\mathsf{fma}\left(\frac{wj}{\mathsf{fma}\left(x, wj, x\right)}, x, \frac{-x}{\mathsf{fma}\left(e^{wj}, wj, e^{wj}\right)}\right)} \]
3. Recombined 2 regimes into one program.
4. Add Preprocessing

Alternative 5: 98.3% accurate, 0.6× speedup

Math

\[\begin{array}{l} \\ \begin{array}{l} t_0 := \mathsf{fma}\left(e^{wj}, wj, e^{wj}\right)\\ t_1 := wj \cdot e^{wj}\\ \mathbf{if}\;wj - \frac{t\_1 - x}{e^{wj} + t\_1} \leq 10^{-19}:\\ \;\;\;\;\mathsf{fma}\left(\frac{wj \cdot wj}{x}, x, \frac{x}{t\_0}\right)\\ \mathbf{else}:\\ \;\;\;\;wj - \mathsf{fma}\left(\frac{wj}{\mathsf{fma}\left(x, wj, x\right)}, x, \frac{-x}{t\_0}\right)\\ \end{array} \end{array} \]

FPCore

(FPCore (wj x)
 :precision binary64
 (let* ((t_0 (fma (exp wj) wj (exp wj))) (t_1 (* wj (exp wj))))
   (if (<= (- wj (/ (- t_1 x) (+ (exp wj) t_1))) 1e-19)
     (fma (/ (* wj wj) x) x (/ x t_0))
     (- wj (fma (/ wj (fma x wj x)) x (/ (- x) t_0))))))

C

double code(double wj, double x) {
	double t_0 = fma(exp(wj), wj, exp(wj));
	double t_1 = wj * exp(wj);
	double tmp;
	if ((wj - ((t_1 - x) / (exp(wj) + t_1))) <= 1e-19) {
		tmp = fma(((wj * wj) / x), x, (x / t_0));
	} else {
		tmp = wj - fma((wj / fma(x, wj, x)), x, (-x / t_0));
	}
	return tmp;
}

Julia

function code(wj, x)
	t_0 = fma(exp(wj), wj, exp(wj))
	t_1 = Float64(wj * exp(wj))
	tmp = 0.0
	if (Float64(wj - Float64(Float64(t_1 - x) / Float64(exp(wj) + t_1))) <= 1e-19)
		tmp = fma(Float64(Float64(wj * wj) / x), x, Float64(x / t_0));
	else
		tmp = Float64(wj - fma(Float64(wj / fma(x, wj, x)), x, Float64(Float64(-x) / t_0)));
	end
	return tmp
end

Wolfram

code[wj_, x_] := Block[{t$95$0 = N[(N[Exp[wj], $MachinePrecision] * wj + N[Exp[wj], $MachinePrecision]), $MachinePrecision]}, Block[{t$95$1 = N[(wj * N[Exp[wj], $MachinePrecision]), $MachinePrecision]}, If[LessEqual[N[(wj - N[(N[(t$95$1 - x), $MachinePrecision] / N[(N[Exp[wj], $MachinePrecision] + t$95$1), $MachinePrecision]), $MachinePrecision]), $MachinePrecision], 1e-19], N[(N[(N[(wj * wj), $MachinePrecision] / x), $MachinePrecision] * x + N[(x / t$95$0), $MachinePrecision]), $MachinePrecision], N[(wj - N[(N[(wj / N[(x * wj + x), $MachinePrecision]), $MachinePrecision] * x + N[((-x) / t$95$0), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]]]]

Derivation

1. Split input into 2 regimes

2. if (-.f64 wj (/.f64 (-.f64 (*.f64 wj (exp.f64 wj)) x) (+.f64 (exp.f64 wj) (*.f64 wj (exp.f64 wj))))) < 9.9999999999999998e-20

   1. Initial program 70.4%

      \[wj - \frac{wj \cdot e^{wj} - x}{e^{wj} + wj \cdot e^{wj}} \]
   2. Add Preprocessing

   3. Taylor expanded in x around inf

      \[\leadsto \color{blue}{x \cdot \left(\left(\frac{1}{e^{wj} + wj \cdot e^{wj}} + \frac{wj}{x}\right) - \frac{wj \cdot e^{wj}}{x \cdot \left(e^{wj} + wj \cdot e^{wj}\right)}\right)} \]

   5. Applied rewrites 86.7%

      \[\leadsto \color{blue}{\mathsf{fma}\left(\frac{wj}{x} - \frac{wj}{\mathsf{fma}\left(x, wj, x\right)} \cdot e^{wj - wj}, x, \frac{x}{\mathsf{fma}\left(e^{wj}, wj, e^{wj}\right)}\right)} \]

   6. Taylor expanded in wj around 0

      \[\leadsto \mathsf{fma}\left(\frac{{wj}^{2}}{x}, x, \frac{x}{\mathsf{fma}\left(e^{wj}, wj, e^{wj}\right)}\right) \]
   7. Step-by-step derivation

   8. Applied rewrites 99.4%

      \[\leadsto \mathsf{fma}\left(\frac{wj \cdot wj}{x}, x, \frac{x}{\mathsf{fma}\left(e^{wj}, wj, e^{wj}\right)}\right) \]

   if 9.9999999999999998e-20 < (-.f64 wj (/.f64 (-.f64 (*.f64 wj (exp.f64 wj)) x) (+.f64 (exp.f64 wj) (*.f64 wj (exp.f64 wj)))))

   1. Initial program 92.0%

      \[wj - \frac{wj \cdot e^{wj} - x}{e^{wj} + wj \cdot e^{wj}} \]
   2. Add Preprocessing

   3. Taylor expanded in wj around 0

      \[\leadsto wj - \frac{wj \cdot e^{wj} - x}{e^{wj} + \color{blue}{wj}} \]
   4. Step-by-step derivation

   5. Applied rewrites 90.4%

      \[\leadsto wj - \frac{wj \cdot e^{wj} - x}{e^{wj} + \color{blue}{wj}} \]

   6. Taylor expanded in wj around 0

      \[\leadsto wj - \frac{\color{blue}{wj} - x}{e^{wj} + wj} \]
   7. Step-by-step derivation

   8. Applied rewrites 92.1%

      \[\leadsto wj - \frac{\color{blue}{wj} - x}{e^{wj} + wj} \]

   9. Taylor expanded in wj around 0

      \[\leadsto wj - \frac{\color{blue}{wj \cdot \left(1 + wj\right)} - x}{e^{wj} + wj} \]
   10. Step-by-step derivation

   11. Applied rewrites 93.1%

       \[\leadsto wj - \frac{\color{blue}{\mathsf{fma}\left(wj, wj, wj\right)} - x}{e^{wj} + wj} \]

   12. Taylor expanded in x around -inf

       \[\leadsto wj - \color{blue}{-1 \cdot \left(x \cdot \left(-1 \cdot \frac{wj \cdot e^{wj}}{x \cdot \left(e^{wj} + wj \cdot e^{wj}\right)} + \frac{1}{e^{wj} + wj \cdot e^{wj}}\right)\right)} \]

   14. Applied rewrites 98.0%

       \[\leadsto wj - \color{blue}{\mathsf{fma}\left(\frac{wj}{\mathsf{fma}\left(x, wj, x\right)}, x, \frac{-x}{\mathsf{fma}\left(e^{wj}, wj, e^{wj}\right)}\right)} \]
3. Recombined 2 regimes into one program.
4. Add Preprocessing

Alternative 6: 86.9% accurate, 0.6× speedup

Math

\[\begin{array}{l} \\ \begin{array}{l} t_0 := wj \cdot e^{wj}\\ \mathbf{if}\;wj - \frac{t\_0 - x}{e^{wj} + t\_0} \leq 2 \cdot 10^{-49}:\\ \;\;\;\;\frac{x}{\mathsf{fma}\left(e^{wj}, wj, e^{wj}\right)}\\ \mathbf{else}:\\ \;\;\;\;wj - \frac{\mathsf{fma}\left(wj, wj, wj\right) - x}{e^{wj} + wj}\\ \end{array} \end{array} \]

FPCore

(FPCore (wj x)
 :precision binary64
 (let* ((t_0 (* wj (exp wj))))
   (if (<= (- wj (/ (- t_0 x) (+ (exp wj) t_0))) 2e-49)
     (/ x (fma (exp wj) wj (exp wj)))
     (- wj (/ (- (fma wj wj wj) x) (+ (exp wj) wj))))))

C

double code(double wj, double x) {
	double t_0 = wj * exp(wj);
	double tmp;
	if ((wj - ((t_0 - x) / (exp(wj) + t_0))) <= 2e-49) {
		tmp = x / fma(exp(wj), wj, exp(wj));
	} else {
		tmp = wj - ((fma(wj, wj, wj) - x) / (exp(wj) + wj));
	}
	return tmp;
}

Julia

function code(wj, x)
	t_0 = Float64(wj * exp(wj))
	tmp = 0.0
	if (Float64(wj - Float64(Float64(t_0 - x) / Float64(exp(wj) + t_0))) <= 2e-49)
		tmp = Float64(x / fma(exp(wj), wj, exp(wj)));
	else
		tmp = Float64(wj - Float64(Float64(fma(wj, wj, wj) - x) / Float64(exp(wj) + wj)));
	end
	return tmp
end

Wolfram

code[wj_, x_] := Block[{t$95$0 = N[(wj * N[Exp[wj], $MachinePrecision]), $MachinePrecision]}, If[LessEqual[N[(wj - N[(N[(t$95$0 - x), $MachinePrecision] / N[(N[Exp[wj], $MachinePrecision] + t$95$0), $MachinePrecision]), $MachinePrecision]), $MachinePrecision], 2e-49], N[(x / N[(N[Exp[wj], $MachinePrecision] * wj + N[Exp[wj], $MachinePrecision]), $MachinePrecision]), $MachinePrecision], N[(wj - N[(N[(N[(wj * wj + wj), $MachinePrecision] - x), $MachinePrecision] / N[(N[Exp[wj], $MachinePrecision] + wj), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]]]

Derivation

1. Split input into 2 regimes

2. if (-.f64 wj (/.f64 (-.f64 (*.f64 wj (exp.f64 wj)) x) (+.f64 (exp.f64 wj) (*.f64 wj (exp.f64 wj))))) < 1.99999999999999987e-49

   1. Initial program 70.0%

      \[wj - \frac{wj \cdot e^{wj} - x}{e^{wj} + wj \cdot e^{wj}} \]
   2. Add Preprocessing

   3. Taylor expanded in x around inf

      \[\leadsto \color{blue}{\frac{x}{e^{wj} + wj \cdot e^{wj}}} \]
   4. Step-by-step derivation

   5. Applied rewrites 87.4%

      \[\leadsto \color{blue}{\frac{x}{\mathsf{fma}\left(e^{wj}, wj, e^{wj}\right)}} \]

   if 1.99999999999999987e-49 < (-.f64 wj (/.f64 (-.f64 (*.f64 wj (exp.f64 wj)) x) (+.f64 (exp.f64 wj) (*.f64 wj (exp.f64 wj)))))

   1. Initial program 90.0%

      \[wj - \frac{wj \cdot e^{wj} - x}{e^{wj} + wj \cdot e^{wj}} \]
   2. Add Preprocessing

   3. Taylor expanded in wj around 0

      \[\leadsto wj - \frac{wj \cdot e^{wj} - x}{e^{wj} + \color{blue}{wj}} \]
   4. Step-by-step derivation

   5. Applied rewrites 88.6%

      \[\leadsto wj - \frac{wj \cdot e^{wj} - x}{e^{wj} + \color{blue}{wj}} \]

   6. Taylor expanded in wj around 0

      \[\leadsto wj - \frac{\color{blue}{wj} - x}{e^{wj} + wj} \]
   7. Step-by-step derivation

   8. Applied rewrites 89.1%

      \[\leadsto wj - \frac{\color{blue}{wj} - x}{e^{wj} + wj} \]

   9. Taylor expanded in wj around 0

      \[\leadsto wj - \frac{\color{blue}{wj \cdot \left(1 + wj\right)} - x}{e^{wj} + wj} \]
   10. Step-by-step derivation

   11. Applied rewrites 91.0%

       \[\leadsto wj - \frac{\color{blue}{\mathsf{fma}\left(wj, wj, wj\right)} - x}{e^{wj} + wj} \]
3. Recombined 2 regimes into one program.
4. Add Preprocessing

Alternative 7: 86.3% accurate, 0.7× speedup

Math

\[\begin{array}{l} \\ \begin{array}{l} t_0 := wj \cdot e^{wj}\\ \mathbf{if}\;wj - \frac{t\_0 - x}{e^{wj} + t\_0} \leq 2 \cdot 10^{-49}:\\ \;\;\;\;\mathsf{fma}\left(x + x, -wj, x\right)\\ \mathbf{else}:\\ \;\;\;\;wj - \frac{\mathsf{fma}\left(wj, wj, wj\right) - x}{e^{wj} + wj}\\ \end{array} \end{array} \]

FPCore

(FPCore (wj x)
 :precision binary64
 (let* ((t_0 (* wj (exp wj))))
   (if (<= (- wj (/ (- t_0 x) (+ (exp wj) t_0))) 2e-49)
     (fma (+ x x) (- wj) x)
     (- wj (/ (- (fma wj wj wj) x) (+ (exp wj) wj))))))

C

double code(double wj, double x) {
	double t_0 = wj * exp(wj);
	double tmp;
	if ((wj - ((t_0 - x) / (exp(wj) + t_0))) <= 2e-49) {
		tmp = fma((x + x), -wj, x);
	} else {
		tmp = wj - ((fma(wj, wj, wj) - x) / (exp(wj) + wj));
	}
	return tmp;
}

Julia

function code(wj, x)
	t_0 = Float64(wj * exp(wj))
	tmp = 0.0
	if (Float64(wj - Float64(Float64(t_0 - x) / Float64(exp(wj) + t_0))) <= 2e-49)
		tmp = fma(Float64(x + x), Float64(-wj), x);
	else
		tmp = Float64(wj - Float64(Float64(fma(wj, wj, wj) - x) / Float64(exp(wj) + wj)));
	end
	return tmp
end

Wolfram

code[wj_, x_] := Block[{t$95$0 = N[(wj * N[Exp[wj], $MachinePrecision]), $MachinePrecision]}, If[LessEqual[N[(wj - N[(N[(t$95$0 - x), $MachinePrecision] / N[(N[Exp[wj], $MachinePrecision] + t$95$0), $MachinePrecision]), $MachinePrecision]), $MachinePrecision], 2e-49], N[(N[(x + x), $MachinePrecision] * (-wj) + x), $MachinePrecision], N[(wj - N[(N[(N[(wj * wj + wj), $MachinePrecision] - x), $MachinePrecision] / N[(N[Exp[wj], $MachinePrecision] + wj), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]]]

Derivation

1. Split input into 2 regimes

2. if (-.f64 wj (/.f64 (-.f64 (*.f64 wj (exp.f64 wj)) x) (+.f64 (exp.f64 wj) (*.f64 wj (exp.f64 wj))))) < 1.99999999999999987e-49

   1. Initial program 70.0%

      \[wj - \frac{wj \cdot e^{wj} - x}{e^{wj} + wj \cdot e^{wj}} \]
   2. Add Preprocessing

   3. Taylor expanded in wj around 0

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

   5. Applied rewrites 85.3%

      \[\leadsto \color{blue}{\mathsf{fma}\left(x + x, -wj, x\right)} \]

   if 1.99999999999999987e-49 < (-.f64 wj (/.f64 (-.f64 (*.f64 wj (exp.f64 wj)) x) (+.f64 (exp.f64 wj) (*.f64 wj (exp.f64 wj)))))

   1. Initial program 90.0%

      \[wj - \frac{wj \cdot e^{wj} - x}{e^{wj} + wj \cdot e^{wj}} \]
   2. Add Preprocessing

   3. Taylor expanded in wj around 0

      \[\leadsto wj - \frac{wj \cdot e^{wj} - x}{e^{wj} + \color{blue}{wj}} \]
   4. Step-by-step derivation

   5. Applied rewrites 88.6%

      \[\leadsto wj - \frac{wj \cdot e^{wj} - x}{e^{wj} + \color{blue}{wj}} \]

   6. Taylor expanded in wj around 0

      \[\leadsto wj - \frac{\color{blue}{wj} - x}{e^{wj} + wj} \]
   7. Step-by-step derivation

   8. Applied rewrites 89.1%

      \[\leadsto wj - \frac{\color{blue}{wj} - x}{e^{wj} + wj} \]

   9. Taylor expanded in wj around 0

      \[\leadsto wj - \frac{\color{blue}{wj \cdot \left(1 + wj\right)} - x}{e^{wj} + wj} \]
   10. Step-by-step derivation

   11. Applied rewrites 91.0%

       \[\leadsto wj - \frac{\color{blue}{\mathsf{fma}\left(wj, wj, wj\right)} - x}{e^{wj} + wj} \]
3. Recombined 2 regimes into one program.
4. Add Preprocessing

Alternative 8: 96.6% accurate, 1.4× speedup

Math

\[\begin{array}{l} \\ \mathsf{fma}\left(\frac{wj \cdot wj}{x}, x, \frac{x}{\mathsf{fma}\left(e^{wj}, wj, e^{wj}\right)}\right) \end{array} \]

FPCore

(FPCore (wj x)
 :precision binary64
 (fma (/ (* wj wj) x) x (/ x (fma (exp wj) wj (exp wj)))))

C

double code(double wj, double x) {
	return fma(((wj * wj) / x), x, (x / fma(exp(wj), wj, exp(wj))));
}

Julia

function code(wj, x)
	return fma(Float64(Float64(wj * wj) / x), x, Float64(x / fma(exp(wj), wj, exp(wj))))
end

Wolfram

code[wj_, x_] := N[(N[(N[(wj * wj), $MachinePrecision] / x), $MachinePrecision] * x + N[(x / N[(N[Exp[wj], $MachinePrecision] * wj + N[Exp[wj], $MachinePrecision]), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]

Derivation

1. Initial program 77.4%

   \[wj - \frac{wj \cdot e^{wj} - x}{e^{wj} + wj \cdot e^{wj}} \]
2. Add Preprocessing

3. Taylor expanded in x around inf

   \[\leadsto \color{blue}{x \cdot \left(\left(\frac{1}{e^{wj} + wj \cdot e^{wj}} + \frac{wj}{x}\right) - \frac{wj \cdot e^{wj}}{x \cdot \left(e^{wj} + wj \cdot e^{wj}\right)}\right)} \]

5. Applied rewrites 90.3%

   \[\leadsto \color{blue}{\mathsf{fma}\left(\frac{wj}{x} - \frac{wj}{\mathsf{fma}\left(x, wj, x\right)} \cdot e^{wj - wj}, x, \frac{x}{\mathsf{fma}\left(e^{wj}, wj, e^{wj}\right)}\right)} \]

6. Taylor expanded in wj around 0

   \[\leadsto \mathsf{fma}\left(\frac{{wj}^{2}}{x}, x, \frac{x}{\mathsf{fma}\left(e^{wj}, wj, e^{wj}\right)}\right) \]
7. Step-by-step derivation

8. Applied rewrites 96.8%

   \[\leadsto \mathsf{fma}\left(\frac{wj \cdot wj}{x}, x, \frac{x}{\mathsf{fma}\left(e^{wj}, wj, e^{wj}\right)}\right) \]
9. Add Preprocessing

Alternative 9: 96.6% accurate, 1.4× speedup

Math

\[\begin{array}{l} \\ \mathsf{fma}\left(\frac{wj}{x} \cdot wj, x, \frac{x}{\mathsf{fma}\left(e^{wj}, wj, e^{wj}\right)}\right) \end{array} \]

FPCore

(FPCore (wj x)
 :precision binary64
 (fma (* (/ wj x) wj) x (/ x (fma (exp wj) wj (exp wj)))))

C

double code(double wj, double x) {
	return fma(((wj / x) * wj), x, (x / fma(exp(wj), wj, exp(wj))));
}

Julia

function code(wj, x)
	return fma(Float64(Float64(wj / x) * wj), x, Float64(x / fma(exp(wj), wj, exp(wj))))
end

Wolfram

code[wj_, x_] := N[(N[(N[(wj / x), $MachinePrecision] * wj), $MachinePrecision] * x + N[(x / N[(N[Exp[wj], $MachinePrecision] * wj + N[Exp[wj], $MachinePrecision]), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]

Derivation

1. Initial program 77.4%

   \[wj - \frac{wj \cdot e^{wj} - x}{e^{wj} + wj \cdot e^{wj}} \]
2. Add Preprocessing

3. Taylor expanded in x around inf

   \[\leadsto \color{blue}{x \cdot \left(\left(\frac{1}{e^{wj} + wj \cdot e^{wj}} + \frac{wj}{x}\right) - \frac{wj \cdot e^{wj}}{x \cdot \left(e^{wj} + wj \cdot e^{wj}\right)}\right)} \]

5. Applied rewrites 90.3%

   \[\leadsto \color{blue}{\mathsf{fma}\left(\frac{wj}{x} - \frac{wj}{\mathsf{fma}\left(x, wj, x\right)} \cdot e^{wj - wj}, x, \frac{x}{\mathsf{fma}\left(e^{wj}, wj, e^{wj}\right)}\right)} \]

6. Taylor expanded in wj around 0

   \[\leadsto \mathsf{fma}\left(\frac{{wj}^{2}}{x}, x, \frac{x}{\mathsf{fma}\left(e^{wj}, wj, e^{wj}\right)}\right) \]
7. Step-by-step derivation

8. Applied rewrites 96.8%

   \[\leadsto \mathsf{fma}\left(\frac{wj \cdot wj}{x}, x, \frac{x}{\mathsf{fma}\left(e^{wj}, wj, e^{wj}\right)}\right) \]

9. Taylor expanded in wj around 0

   \[\leadsto \mathsf{fma}\left(\frac{{wj}^{2}}{x}, x, \frac{x}{\mathsf{fma}\left(e^{wj}, wj, e^{wj}\right)}\right) \]
10. Step-by-step derivation

11. Applied rewrites 96.8%

    \[\leadsto \mathsf{fma}\left(\frac{wj}{x} \cdot wj, x, \frac{x}{\mathsf{fma}\left(e^{wj}, wj, e^{wj}\right)}\right) \]
12. Add Preprocessing

Alternative 10: 84.6% accurate, 2.3× speedup

Math

\[\begin{array}{l} \\ \begin{array}{l} \mathbf{if}\;wj \leq 3.1 \cdot 10^{-22}:\\ \;\;\;\;\mathsf{fma}\left(x + x, -wj, x\right)\\ \mathbf{else}:\\ \;\;\;\;\mathsf{fma}\left(\frac{wj \cdot wj}{x}, x, \frac{x}{e^{wj} \cdot wj}\right)\\ \end{array} \end{array} \]

FPCore

(FPCore (wj x)
 :precision binary64
 (if (<= wj 3.1e-22)
   (fma (+ x x) (- wj) x)
   (fma (/ (* wj wj) x) x (/ x (* (exp wj) wj)))))

C

double code(double wj, double x) {
	double tmp;
	if (wj <= 3.1e-22) {
		tmp = fma((x + x), -wj, x);
	} else {
		tmp = fma(((wj * wj) / x), x, (x / (exp(wj) * wj)));
	}
	return tmp;
}

Julia

function code(wj, x)
	tmp = 0.0
	if (wj <= 3.1e-22)
		tmp = fma(Float64(x + x), Float64(-wj), x);
	else
		tmp = fma(Float64(Float64(wj * wj) / x), x, Float64(x / Float64(exp(wj) * wj)));
	end
	return tmp
end

Wolfram

code[wj_, x_] := If[LessEqual[wj, 3.1e-22], N[(N[(x + x), $MachinePrecision] * (-wj) + x), $MachinePrecision], N[(N[(N[(wj * wj), $MachinePrecision] / x), $MachinePrecision] * x + N[(x / N[(N[Exp[wj], $MachinePrecision] * wj), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]]

Derivation

1. Split input into 2 regimes

2. if wj < 3.10000000000000013e-22

   1. Initial program 80.4%

      \[wj - \frac{wj \cdot e^{wj} - x}{e^{wj} + wj \cdot e^{wj}} \]
   2. Add Preprocessing

   3. Taylor expanded in wj around 0

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

   5. Applied rewrites 90.8%

      \[\leadsto \color{blue}{\mathsf{fma}\left(x + x, -wj, x\right)} \]

   if 3.10000000000000013e-22 < wj

   1. Initial program 37.4%

      \[wj - \frac{wj \cdot e^{wj} - x}{e^{wj} + wj \cdot e^{wj}} \]
   2. Add Preprocessing

   3. Taylor expanded in x around inf

      \[\leadsto \color{blue}{x \cdot \left(\left(\frac{1}{e^{wj} + wj \cdot e^{wj}} + \frac{wj}{x}\right) - \frac{wj \cdot e^{wj}}{x \cdot \left(e^{wj} + wj \cdot e^{wj}\right)}\right)} \]

   5. Applied rewrites 65.3%

      \[\leadsto \color{blue}{\mathsf{fma}\left(\frac{wj}{x} - \frac{wj}{\mathsf{fma}\left(x, wj, x\right)} \cdot e^{wj - wj}, x, \frac{x}{\mathsf{fma}\left(e^{wj}, wj, e^{wj}\right)}\right)} \]

   6. Taylor expanded in wj around 0

      \[\leadsto \mathsf{fma}\left(\frac{{wj}^{2}}{x}, x, \frac{x}{\mathsf{fma}\left(e^{wj}, wj, e^{wj}\right)}\right) \]
   7. Step-by-step derivation

   8. Applied rewrites 64.5%

      \[\leadsto \mathsf{fma}\left(\frac{wj \cdot wj}{x}, x, \frac{x}{\mathsf{fma}\left(e^{wj}, wj, e^{wj}\right)}\right) \]

   9. Taylor expanded in wj around inf

      \[\leadsto \mathsf{fma}\left(\frac{wj \cdot wj}{x}, x, \frac{x}{wj \cdot e^{wj}}\right) \]
   10. Step-by-step derivation

   11. Applied rewrites 45.3%

       \[\leadsto \mathsf{fma}\left(\frac{wj \cdot wj}{x}, x, \frac{x}{e^{wj} \cdot wj}\right) \]
3. Recombined 2 regimes into one program.

4. Final simplification 87.6%

   \[\leadsto \begin{array}{l} \mathbf{if}\;wj \leq 3.1 \cdot 10^{-22}:\\ \;\;\;\;\mathsf{fma}\left(x + x, -wj, x\right)\\ \mathbf{else}:\\ \;\;\;\;\mathsf{fma}\left(\frac{wj \cdot wj}{x}, x, \frac{x}{e^{wj} \cdot wj}\right)\\ \end{array} \]
5. Add Preprocessing

Alternative 11: 84.9% accurate, 27.6× speedup

Math

\[\begin{array}{l} \\ \mathsf{fma}\left(x + x, -wj, x\right) \end{array} \]

FPCore

(FPCore (wj x) :precision binary64 (fma (+ x x) (- wj) x))

C

double code(double wj, double x) {
	return fma((x + x), -wj, x);
}

Julia

function code(wj, x)
	return fma(Float64(x + x), Float64(-wj), x)
end

Wolfram

code[wj_, x_] := N[(N[(x + x), $MachinePrecision] * (-wj) + x), $MachinePrecision]

Derivation

1. Initial program 77.4%

   \[wj - \frac{wj \cdot e^{wj} - x}{e^{wj} + wj \cdot e^{wj}} \]
2. Add Preprocessing

3. Taylor expanded in wj around 0

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

5. Applied rewrites 85.5%

   \[\leadsto \color{blue}{\mathsf{fma}\left(x + x, -wj, x\right)} \]
6. Add Preprocessing

Alternative 12: 84.4% accurate, 331.0× speedup

Math

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

FPCore

(FPCore (wj x) :precision binary64 x)

C

double code(double wj, double x) {
	return x;
}

Fortran

module fmin_fmax_functions
    implicit none
    private
    public fmax
    public fmin

    interface fmax
        module procedure fmax88
        module procedure fmax44
        module procedure fmax84
        module procedure fmax48
    end interface
    interface fmin
        module procedure fmin88
        module procedure fmin44
        module procedure fmin84
        module procedure fmin48
    end interface
contains
    real(8) function fmax88(x, y) result (res)
        real(8), intent (in) :: x
        real(8), intent (in) :: y
        res = merge(y, merge(x, max(x, y), y /= y), x /= x)
    end function
    real(4) function fmax44(x, y) result (res)
        real(4), intent (in) :: x
        real(4), intent (in) :: y
        res = merge(y, merge(x, max(x, y), y /= y), x /= x)
    end function
    real(8) function fmax84(x, y) result(res)
        real(8), intent (in) :: x
        real(4), intent (in) :: y
        res = merge(dble(y), merge(x, max(x, dble(y)), y /= y), x /= x)
    end function
    real(8) function fmax48(x, y) result(res)
        real(4), intent (in) :: x
        real(8), intent (in) :: y
        res = merge(y, merge(dble(x), max(dble(x), y), y /= y), x /= x)
    end function
    real(8) function fmin88(x, y) result (res)
        real(8), intent (in) :: x
        real(8), intent (in) :: y
        res = merge(y, merge(x, min(x, y), y /= y), x /= x)
    end function
    real(4) function fmin44(x, y) result (res)
        real(4), intent (in) :: x
        real(4), intent (in) :: y
        res = merge(y, merge(x, min(x, y), y /= y), x /= x)
    end function
    real(8) function fmin84(x, y) result(res)
        real(8), intent (in) :: x
        real(4), intent (in) :: y
        res = merge(dble(y), merge(x, min(x, dble(y)), y /= y), x /= x)
    end function
    real(8) function fmin48(x, y) result(res)
        real(4), intent (in) :: x
        real(8), intent (in) :: y
        res = merge(y, merge(dble(x), min(dble(x), y), y /= y), x /= x)
    end function
end module

real(8) function code(wj, x)
use fmin_fmax_functions
    real(8), intent (in) :: wj
    real(8), intent (in) :: x
    code = x
end function

Java

public static double code(double wj, double x) {
	return x;
}

Python

def code(wj, x):
	return x

Julia

function code(wj, x)
	return x
end

MATLAB

function tmp = code(wj, x)
	tmp = x;
end

Wolfram

code[wj_, x_] := x

Derivation

1. Initial program 77.4%

   \[wj - \frac{wj \cdot e^{wj} - x}{e^{wj} + wj \cdot e^{wj}} \]
2. Add Preprocessing

3. Taylor expanded in wj around 0

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

5. Applied rewrites 84.7%

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

Alternative 13: 4.4% accurate, 331.0× speedup

Math

\[\begin{array}{l} \\ wj \end{array} \]

FPCore

(FPCore (wj x) :precision binary64 wj)

C

double code(double wj, double x) {
	return wj;
}

Fortran

module fmin_fmax_functions
    implicit none
    private
    public fmax
    public fmin

    interface fmax
        module procedure fmax88
        module procedure fmax44
        module procedure fmax84
        module procedure fmax48
    end interface
    interface fmin
        module procedure fmin88
        module procedure fmin44
        module procedure fmin84
        module procedure fmin48
    end interface
contains
    real(8) function fmax88(x, y) result (res)
        real(8), intent (in) :: x
        real(8), intent (in) :: y
        res = merge(y, merge(x, max(x, y), y /= y), x /= x)
    end function
    real(4) function fmax44(x, y) result (res)
        real(4), intent (in) :: x
        real(4), intent (in) :: y
        res = merge(y, merge(x, max(x, y), y /= y), x /= x)
    end function
    real(8) function fmax84(x, y) result(res)
        real(8), intent (in) :: x
        real(4), intent (in) :: y
        res = merge(dble(y), merge(x, max(x, dble(y)), y /= y), x /= x)
    end function
    real(8) function fmax48(x, y) result(res)
        real(4), intent (in) :: x
        real(8), intent (in) :: y
        res = merge(y, merge(dble(x), max(dble(x), y), y /= y), x /= x)
    end function
    real(8) function fmin88(x, y) result (res)
        real(8), intent (in) :: x
        real(8), intent (in) :: y
        res = merge(y, merge(x, min(x, y), y /= y), x /= x)
    end function
    real(4) function fmin44(x, y) result (res)
        real(4), intent (in) :: x
        real(4), intent (in) :: y
        res = merge(y, merge(x, min(x, y), y /= y), x /= x)
    end function
    real(8) function fmin84(x, y) result(res)
        real(8), intent (in) :: x
        real(4), intent (in) :: y
        res = merge(dble(y), merge(x, min(x, dble(y)), y /= y), x /= x)
    end function
    real(8) function fmin48(x, y) result(res)
        real(4), intent (in) :: x
        real(8), intent (in) :: y
        res = merge(y, merge(dble(x), min(dble(x), y), y /= y), x /= x)
    end function
end module

real(8) function code(wj, x)
use fmin_fmax_functions
    real(8), intent (in) :: wj
    real(8), intent (in) :: x
    code = wj
end function

Java

public static double code(double wj, double x) {
	return wj;
}

Python

def code(wj, x):
	return wj

Julia

function code(wj, x)
	return wj
end

MATLAB

function tmp = code(wj, x)
	tmp = wj;
end

Wolfram

code[wj_, x_] := wj

Derivation

1. Initial program 77.4%

   \[wj - \frac{wj \cdot e^{wj} - x}{e^{wj} + wj \cdot e^{wj}} \]
2. Add Preprocessing

3. Taylor expanded in wj around inf

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

5. Applied rewrites 4.6%

   \[\leadsto \color{blue}{wj} \]
6. Add Preprocessing

Developer Target 1: 79.4% accurate, 1.4× speedup

Math

\[\begin{array}{l} \\ wj - \left(\frac{wj}{wj + 1} - \frac{x}{e^{wj} + wj \cdot e^{wj}}\right) \end{array} \]

FPCore

(FPCore (wj x)
 :precision binary64
 (- wj (- (/ wj (+ wj 1.0)) (/ x (+ (exp wj) (* wj (exp wj)))))))

C

double code(double wj, double x) {
	return wj - ((wj / (wj + 1.0)) - (x / (exp(wj) + (wj * exp(wj)))));
}

Fortran

module fmin_fmax_functions
    implicit none
    private
    public fmax
    public fmin

    interface fmax
        module procedure fmax88
        module procedure fmax44
        module procedure fmax84
        module procedure fmax48
    end interface
    interface fmin
        module procedure fmin88
        module procedure fmin44
        module procedure fmin84
        module procedure fmin48
    end interface
contains
    real(8) function fmax88(x, y) result (res)
        real(8), intent (in) :: x
        real(8), intent (in) :: y
        res = merge(y, merge(x, max(x, y), y /= y), x /= x)
    end function
    real(4) function fmax44(x, y) result (res)
        real(4), intent (in) :: x
        real(4), intent (in) :: y
        res = merge(y, merge(x, max(x, y), y /= y), x /= x)
    end function
    real(8) function fmax84(x, y) result(res)
        real(8), intent (in) :: x
        real(4), intent (in) :: y
        res = merge(dble(y), merge(x, max(x, dble(y)), y /= y), x /= x)
    end function
    real(8) function fmax48(x, y) result(res)
        real(4), intent (in) :: x
        real(8), intent (in) :: y
        res = merge(y, merge(dble(x), max(dble(x), y), y /= y), x /= x)
    end function
    real(8) function fmin88(x, y) result (res)
        real(8), intent (in) :: x
        real(8), intent (in) :: y
        res = merge(y, merge(x, min(x, y), y /= y), x /= x)
    end function
    real(4) function fmin44(x, y) result (res)
        real(4), intent (in) :: x
        real(4), intent (in) :: y
        res = merge(y, merge(x, min(x, y), y /= y), x /= x)
    end function
    real(8) function fmin84(x, y) result(res)
        real(8), intent (in) :: x
        real(4), intent (in) :: y
        res = merge(dble(y), merge(x, min(x, dble(y)), y /= y), x /= x)
    end function
    real(8) function fmin48(x, y) result(res)
        real(4), intent (in) :: x
        real(8), intent (in) :: y
        res = merge(y, merge(dble(x), min(dble(x), y), y /= y), x /= x)
    end function
end module

real(8) function code(wj, x)
use fmin_fmax_functions
    real(8), intent (in) :: wj
    real(8), intent (in) :: x
    code = wj - ((wj / (wj + 1.0d0)) - (x / (exp(wj) + (wj * exp(wj)))))
end function

Java

public static double code(double wj, double x) {
	return wj - ((wj / (wj + 1.0)) - (x / (Math.exp(wj) + (wj * Math.exp(wj)))));
}

Python

def code(wj, x):
	return wj - ((wj / (wj + 1.0)) - (x / (math.exp(wj) + (wj * math.exp(wj)))))

Julia

function code(wj, x)
	return Float64(wj - Float64(Float64(wj / Float64(wj + 1.0)) - Float64(x / Float64(exp(wj) + Float64(wj * exp(wj))))))
end

MATLAB

function tmp = code(wj, x)
	tmp = wj - ((wj / (wj + 1.0)) - (x / (exp(wj) + (wj * exp(wj)))));
end

Wolfram

code[wj_, x_] := N[(wj - N[(N[(wj / N[(wj + 1.0), $MachinePrecision]), $MachinePrecision] - N[(x / N[(N[Exp[wj], $MachinePrecision] + N[(wj * N[Exp[wj], $MachinePrecision]), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]

Reproduce

herbie shell --seed 2025021 
(FPCore (wj x)
  :name "Jmat.Real.lambertw, newton loop step"
  :precision binary64

  :alt
  (! :herbie-platform default (let ((ew (exp wj))) (- wj (- (/ wj (+ wj 1)) (/ x (+ ew (* wj ew)))))))

  (- wj (/ (- (* wj (exp wj)) x) (+ (exp wj) (* wj (exp wj))))))