Numeric.SpecFunctions:invErfc from math-functions-0.1.5.2, A

Percentage Accurate: 95.6% → 99.8%

Time: 3.5s

Alternatives: 13

Speedup: 1.1×

Specification

Math

\[\begin{array}{l} \\ x + \frac{y}{1.1283791670955126 \cdot e^{z} - x \cdot y} \end{array} \]

FPCore

(FPCore (x y z)
 :precision binary64
 (+ x (/ y (- (* 1.1283791670955126 (exp z)) (* x y)))))

C

double code(double x, double y, double z) {
	return x + (y / ((1.1283791670955126 * exp(z)) - (x * y)));
}

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(x, y, z)
use fmin_fmax_functions
    real(8), intent (in) :: x
    real(8), intent (in) :: y
    real(8), intent (in) :: z
    code = x + (y / ((1.1283791670955126d0 * exp(z)) - (x * y)))
end function

Java

public static double code(double x, double y, double z) {
	return x + (y / ((1.1283791670955126 * Math.exp(z)) - (x * y)));
}

Python

def code(x, y, z):
	return x + (y / ((1.1283791670955126 * math.exp(z)) - (x * y)))

Julia

function code(x, y, z)
	return Float64(x + Float64(y / Float64(Float64(1.1283791670955126 * exp(z)) - Float64(x * y))))
end

MATLAB

function tmp = code(x, y, z)
	tmp = x + (y / ((1.1283791670955126 * exp(z)) - (x * y)));
end

Wolfram

code[x_, y_, z_] := N[(x + N[(y / N[(N[(1.1283791670955126 * N[Exp[z], $MachinePrecision]), $MachinePrecision] - N[(x * y), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]

Initial Program: 95.6% accurate, 1.0× speedup

Math

\[\begin{array}{l} \\ x + \frac{y}{1.1283791670955126 \cdot e^{z} - x \cdot y} \end{array} \]

FPCore

(FPCore (x y z)
 :precision binary64
 (+ x (/ y (- (* 1.1283791670955126 (exp z)) (* x y)))))

C

double code(double x, double y, double z) {
	return x + (y / ((1.1283791670955126 * exp(z)) - (x * y)));
}

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(x, y, z)
use fmin_fmax_functions
    real(8), intent (in) :: x
    real(8), intent (in) :: y
    real(8), intent (in) :: z
    code = x + (y / ((1.1283791670955126d0 * exp(z)) - (x * y)))
end function

Java

public static double code(double x, double y, double z) {
	return x + (y / ((1.1283791670955126 * Math.exp(z)) - (x * y)));
}

Python

def code(x, y, z):
	return x + (y / ((1.1283791670955126 * math.exp(z)) - (x * y)))

Julia

function code(x, y, z)
	return Float64(x + Float64(y / Float64(Float64(1.1283791670955126 * exp(z)) - Float64(x * y))))
end

MATLAB

function tmp = code(x, y, z)
	tmp = x + (y / ((1.1283791670955126 * exp(z)) - (x * y)));
end

Wolfram

code[x_, y_, z_] := N[(x + N[(y / N[(N[(1.1283791670955126 * N[Exp[z], $MachinePrecision]), $MachinePrecision] - N[(x * y), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]

Alternative 1: 99.8% accurate, 0.8× speedup

Math

\[\begin{array}{l} \\ \begin{array}{l} \mathbf{if}\;z \leq -6.5 \cdot 10^{+15}:\\ \;\;\;\;x - \frac{1}{x}\\ \mathbf{else}:\\ \;\;\;\;x + \frac{y}{\left(\frac{e^{z}}{x} \cdot 1.1283791670955126 - y\right) \cdot x}\\ \end{array} \end{array} \]

FPCore

(FPCore (x y z)
 :precision binary64
 (if (<= z -6.5e+15)
   (- x (/ 1.0 x))
   (+ x (/ y (* (- (* (/ (exp z) x) 1.1283791670955126) y) x)))))

C

double code(double x, double y, double z) {
	double tmp;
	if (z <= -6.5e+15) {
		tmp = x - (1.0 / x);
	} else {
		tmp = x + (y / ((((exp(z) / x) * 1.1283791670955126) - y) * x));
	}
	return tmp;
}

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(x, y, z)
use fmin_fmax_functions
    real(8), intent (in) :: x
    real(8), intent (in) :: y
    real(8), intent (in) :: z
    real(8) :: tmp
    if (z <= (-6.5d+15)) then
        tmp = x - (1.0d0 / x)
    else
        tmp = x + (y / ((((exp(z) / x) * 1.1283791670955126d0) - y) * x))
    end if
    code = tmp
end function

Java

public static double code(double x, double y, double z) {
	double tmp;
	if (z <= -6.5e+15) {
		tmp = x - (1.0 / x);
	} else {
		tmp = x + (y / ((((Math.exp(z) / x) * 1.1283791670955126) - y) * x));
	}
	return tmp;
}

Python

def code(x, y, z):
	tmp = 0
	if z <= -6.5e+15:
		tmp = x - (1.0 / x)
	else:
		tmp = x + (y / ((((math.exp(z) / x) * 1.1283791670955126) - y) * x))
	return tmp

Julia

function code(x, y, z)
	tmp = 0.0
	if (z <= -6.5e+15)
		tmp = Float64(x - Float64(1.0 / x));
	else
		tmp = Float64(x + Float64(y / Float64(Float64(Float64(Float64(exp(z) / x) * 1.1283791670955126) - y) * x)));
	end
	return tmp
end

MATLAB

function tmp_2 = code(x, y, z)
	tmp = 0.0;
	if (z <= -6.5e+15)
		tmp = x - (1.0 / x);
	else
		tmp = x + (y / ((((exp(z) / x) * 1.1283791670955126) - y) * x));
	end
	tmp_2 = tmp;
end

Wolfram

code[x_, y_, z_] := If[LessEqual[z, -6.5e+15], N[(x - N[(1.0 / x), $MachinePrecision]), $MachinePrecision], N[(x + N[(y / N[(N[(N[(N[(N[Exp[z], $MachinePrecision] / x), $MachinePrecision] * 1.1283791670955126), $MachinePrecision] - y), $MachinePrecision] * x), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]]

Derivation

1. Split input into 2 regimes

2. if z < -6.5e15

   1. Initial program 95.6%

      \[x + \frac{y}{1.1283791670955126 \cdot e^{z} - x \cdot y} \]

   2. Taylor expanded in y around inf

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

      1. lower--.f64 N/A

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

      2. lower-/.f64 69.4

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

   4. Applied rewrites 69.4%

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

   if -6.5e15 < z

   1. Initial program 95.6%

      \[x + \frac{y}{1.1283791670955126 \cdot e^{z} - x \cdot y} \]

   2. Taylor expanded in x around inf

      \[\leadsto x + \frac{y}{\color{blue}{x \cdot \left(\frac{5641895835477563}{5000000000000000} \cdot \frac{e^{z}}{x} - y\right)}} \]
   3. Step-by-step derivation

      1. *-commutative N/A

         \[\leadsto x + \frac{y}{\left(\frac{5641895835477563}{5000000000000000} \cdot \frac{e^{z}}{x} - y\right) \cdot \color{blue}{x}} \]

      2. lower-*.f64 N/A

         \[\leadsto x + \frac{y}{\left(\frac{5641895835477563}{5000000000000000} \cdot \frac{e^{z}}{x} - y\right) \cdot \color{blue}{x}} \]

      3. lower--.f64 N/A

         \[\leadsto x + \frac{y}{\left(\frac{5641895835477563}{5000000000000000} \cdot \frac{e^{z}}{x} - y\right) \cdot x} \]

      4. *-commutative N/A

         \[\leadsto x + \frac{y}{\left(\frac{e^{z}}{x} \cdot \frac{5641895835477563}{5000000000000000} - y\right) \cdot x} \]

      5. lower-*.f64 N/A

         \[\leadsto x + \frac{y}{\left(\frac{e^{z}}{x} \cdot \frac{5641895835477563}{5000000000000000} - y\right) \cdot x} \]

      6. lower-/.f64 N/A

         \[\leadsto x + \frac{y}{\left(\frac{e^{z}}{x} \cdot \frac{5641895835477563}{5000000000000000} - y\right) \cdot x} \]

      7. lift-exp.f64 97.2

         \[\leadsto x + \frac{y}{\left(\frac{e^{z}}{x} \cdot 1.1283791670955126 - y\right) \cdot x} \]

   4. Applied rewrites 97.2%

      \[\leadsto x + \frac{y}{\color{blue}{\left(\frac{e^{z}}{x} \cdot 1.1283791670955126 - y\right) \cdot x}} \]
3. Recombined 2 regimes into one program.
4. Add Preprocessing

Alternative 2: 99.5% accurate, 0.9× speedup

Math

\[\begin{array}{l} \\ \begin{array}{l} \mathbf{if}\;z \leq -6.5 \cdot 10^{+15}:\\ \;\;\;\;x - \frac{1}{x}\\ \mathbf{else}:\\ \;\;\;\;x + \frac{y}{1.1283791670955126 \cdot e^{z} - x \cdot y}\\ \end{array} \end{array} \]

FPCore

(FPCore (x y z)
 :precision binary64
 (if (<= z -6.5e+15)
   (- x (/ 1.0 x))
   (+ x (/ y (- (* 1.1283791670955126 (exp z)) (* x y))))))

C

double code(double x, double y, double z) {
	double tmp;
	if (z <= -6.5e+15) {
		tmp = x - (1.0 / x);
	} else {
		tmp = x + (y / ((1.1283791670955126 * exp(z)) - (x * y)));
	}
	return tmp;
}

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(x, y, z)
use fmin_fmax_functions
    real(8), intent (in) :: x
    real(8), intent (in) :: y
    real(8), intent (in) :: z
    real(8) :: tmp
    if (z <= (-6.5d+15)) then
        tmp = x - (1.0d0 / x)
    else
        tmp = x + (y / ((1.1283791670955126d0 * exp(z)) - (x * y)))
    end if
    code = tmp
end function

Java

public static double code(double x, double y, double z) {
	double tmp;
	if (z <= -6.5e+15) {
		tmp = x - (1.0 / x);
	} else {
		tmp = x + (y / ((1.1283791670955126 * Math.exp(z)) - (x * y)));
	}
	return tmp;
}

Python

def code(x, y, z):
	tmp = 0
	if z <= -6.5e+15:
		tmp = x - (1.0 / x)
	else:
		tmp = x + (y / ((1.1283791670955126 * math.exp(z)) - (x * y)))
	return tmp

Julia

function code(x, y, z)
	tmp = 0.0
	if (z <= -6.5e+15)
		tmp = Float64(x - Float64(1.0 / x));
	else
		tmp = Float64(x + Float64(y / Float64(Float64(1.1283791670955126 * exp(z)) - Float64(x * y))));
	end
	return tmp
end

MATLAB

function tmp_2 = code(x, y, z)
	tmp = 0.0;
	if (z <= -6.5e+15)
		tmp = x - (1.0 / x);
	else
		tmp = x + (y / ((1.1283791670955126 * exp(z)) - (x * y)));
	end
	tmp_2 = tmp;
end

Wolfram

code[x_, y_, z_] := If[LessEqual[z, -6.5e+15], N[(x - N[(1.0 / x), $MachinePrecision]), $MachinePrecision], N[(x + N[(y / N[(N[(1.1283791670955126 * N[Exp[z], $MachinePrecision]), $MachinePrecision] - N[(x * y), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]]

Derivation

1. Split input into 2 regimes

2. if z < -6.5e15

   1. Initial program 95.6%

      \[x + \frac{y}{1.1283791670955126 \cdot e^{z} - x \cdot y} \]

   2. Taylor expanded in y around inf

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

      1. lower--.f64 N/A

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

      2. lower-/.f64 69.4

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

   4. Applied rewrites 69.4%

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

   if -6.5e15 < z

   1. Initial program 95.6%

      \[x + \frac{y}{1.1283791670955126 \cdot e^{z} - x \cdot y} \]
3. Recombined 2 regimes into one program.
4. Add Preprocessing

Alternative 3: 98.3% accurate, 1.0× speedup

Math

\[\begin{array}{l} \\ \begin{array}{l} \mathbf{if}\;z \leq -5:\\ \;\;\;\;x - \frac{1}{x}\\ \mathbf{elif}\;z \leq 4.2 \cdot 10^{-14}:\\ \;\;\;\;x + \frac{y}{\mathsf{fma}\left(z, 1.1283791670955126, 1.1283791670955126\right) - x \cdot y}\\ \mathbf{else}:\\ \;\;\;\;\mathsf{fma}\left(\frac{y}{e^{z}}, 0.8862269254527579, x\right)\\ \end{array} \end{array} \]

FPCore

(FPCore (x y z)
 :precision binary64
 (if (<= z -5.0)
   (- x (/ 1.0 x))
   (if (<= z 4.2e-14)
     (+ x (/ y (- (fma z 1.1283791670955126 1.1283791670955126) (* x y))))
     (fma (/ y (exp z)) 0.8862269254527579 x))))

C

double code(double x, double y, double z) {
	double tmp;
	if (z <= -5.0) {
		tmp = x - (1.0 / x);
	} else if (z <= 4.2e-14) {
		tmp = x + (y / (fma(z, 1.1283791670955126, 1.1283791670955126) - (x * y)));
	} else {
		tmp = fma((y / exp(z)), 0.8862269254527579, x);
	}
	return tmp;
}

Julia

function code(x, y, z)
	tmp = 0.0
	if (z <= -5.0)
		tmp = Float64(x - Float64(1.0 / x));
	elseif (z <= 4.2e-14)
		tmp = Float64(x + Float64(y / Float64(fma(z, 1.1283791670955126, 1.1283791670955126) - Float64(x * y))));
	else
		tmp = fma(Float64(y / exp(z)), 0.8862269254527579, x);
	end
	return tmp
end

Wolfram

code[x_, y_, z_] := If[LessEqual[z, -5.0], N[(x - N[(1.0 / x), $MachinePrecision]), $MachinePrecision], If[LessEqual[z, 4.2e-14], N[(x + N[(y / N[(N[(z * 1.1283791670955126 + 1.1283791670955126), $MachinePrecision] - N[(x * y), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]), $MachinePrecision], N[(N[(y / N[Exp[z], $MachinePrecision]), $MachinePrecision] * 0.8862269254527579 + x), $MachinePrecision]]]

Derivation

1. Split input into 3 regimes

2. if z < -5

   1. Initial program 95.6%

      \[x + \frac{y}{1.1283791670955126 \cdot e^{z} - x \cdot y} \]

   2. Taylor expanded in y around inf

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

      1. lower--.f64 N/A

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

      2. lower-/.f64 69.4

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

   4. Applied rewrites 69.4%

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

   if -5 < z < 4.1999999999999998e-14

   1. Initial program 95.6%

      \[x + \frac{y}{1.1283791670955126 \cdot e^{z} - x \cdot y} \]

   2. Taylor expanded in z around 0

      \[\leadsto x + \frac{y}{\color{blue}{\left(\frac{5641895835477563}{5000000000000000} + \frac{5641895835477563}{5000000000000000} \cdot z\right)} - x \cdot y} \]
   3. Step-by-step derivation

      1. +-commutative N/A

         \[\leadsto x + \frac{y}{\left(\frac{5641895835477563}{5000000000000000} \cdot z + \color{blue}{\frac{5641895835477563}{5000000000000000}}\right) - x \cdot y} \]

      2. *-commutative N/A

         \[\leadsto x + \frac{y}{\left(z \cdot \frac{5641895835477563}{5000000000000000} + \frac{5641895835477563}{5000000000000000}\right) - x \cdot y} \]

      3. lower-fma.f64 81.3

         \[\leadsto x + \frac{y}{\mathsf{fma}\left(z, \color{blue}{1.1283791670955126}, 1.1283791670955126\right) - x \cdot y} \]

   4. Applied rewrites 81.3%

      \[\leadsto x + \frac{y}{\color{blue}{\mathsf{fma}\left(z, 1.1283791670955126, 1.1283791670955126\right)} - x \cdot y} \]

   if 4.1999999999999998e-14 < z

   1. Initial program 95.6%

      \[x + \frac{y}{1.1283791670955126 \cdot e^{z} - x \cdot y} \]

   2. Taylor expanded in y around 0

      \[\leadsto \color{blue}{x + \frac{5000000000000000}{5641895835477563} \cdot \frac{y}{e^{z}}} \]
   3. Step-by-step derivation

      1. +-commutative N/A

         \[\leadsto \frac{5000000000000000}{5641895835477563} \cdot \frac{y}{e^{z}} + \color{blue}{x} \]

      2. *-commutative N/A

         \[\leadsto \frac{y}{e^{z}} \cdot \frac{5000000000000000}{5641895835477563} + x \]

      3. lower-fma.f64 N/A

         \[\leadsto \mathsf{fma}\left(\frac{y}{e^{z}}, \color{blue}{\frac{5000000000000000}{5641895835477563}}, x\right) \]

      4. lower-/.f64 N/A

         \[\leadsto \mathsf{fma}\left(\frac{y}{e^{z}}, \frac{5000000000000000}{5641895835477563}, x\right) \]

      5. lift-exp.f64 62.6

         \[\leadsto \mathsf{fma}\left(\frac{y}{e^{z}}, 0.8862269254527579, x\right) \]

   4. Applied rewrites 62.6%

      \[\leadsto \color{blue}{\mathsf{fma}\left(\frac{y}{e^{z}}, 0.8862269254527579, x\right)} \]
3. Recombined 3 regimes into one program.
4. Add Preprocessing

Alternative 4: 96.2% accurate, 1.0× speedup

Math

\[\begin{array}{l} \\ \begin{array}{l} \mathbf{if}\;z \leq -5:\\ \;\;\;\;x - \frac{1}{x}\\ \mathbf{elif}\;z \leq 1.2 \cdot 10^{+73}:\\ \;\;\;\;x + \frac{y}{\mathsf{fma}\left(z, 1.1283791670955126, 1.1283791670955126\right) - x \cdot y}\\ \mathbf{else}:\\ \;\;\;\;\mathsf{fma}\left(\frac{y}{\left(z \cdot z\right) \cdot 0.5}, 0.8862269254527579, x\right)\\ \end{array} \end{array} \]

FPCore

(FPCore (x y z)
 :precision binary64
 (if (<= z -5.0)
   (- x (/ 1.0 x))
   (if (<= z 1.2e+73)
     (+ x (/ y (- (fma z 1.1283791670955126 1.1283791670955126) (* x y))))
     (fma (/ y (* (* z z) 0.5)) 0.8862269254527579 x))))

C

double code(double x, double y, double z) {
	double tmp;
	if (z <= -5.0) {
		tmp = x - (1.0 / x);
	} else if (z <= 1.2e+73) {
		tmp = x + (y / (fma(z, 1.1283791670955126, 1.1283791670955126) - (x * y)));
	} else {
		tmp = fma((y / ((z * z) * 0.5)), 0.8862269254527579, x);
	}
	return tmp;
}

Julia

function code(x, y, z)
	tmp = 0.0
	if (z <= -5.0)
		tmp = Float64(x - Float64(1.0 / x));
	elseif (z <= 1.2e+73)
		tmp = Float64(x + Float64(y / Float64(fma(z, 1.1283791670955126, 1.1283791670955126) - Float64(x * y))));
	else
		tmp = fma(Float64(y / Float64(Float64(z * z) * 0.5)), 0.8862269254527579, x);
	end
	return tmp
end

Wolfram

code[x_, y_, z_] := If[LessEqual[z, -5.0], N[(x - N[(1.0 / x), $MachinePrecision]), $MachinePrecision], If[LessEqual[z, 1.2e+73], N[(x + N[(y / N[(N[(z * 1.1283791670955126 + 1.1283791670955126), $MachinePrecision] - N[(x * y), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]), $MachinePrecision], N[(N[(y / N[(N[(z * z), $MachinePrecision] * 0.5), $MachinePrecision]), $MachinePrecision] * 0.8862269254527579 + x), $MachinePrecision]]]

Derivation

1. Split input into 3 regimes

2. if z < -5

   1. Initial program 95.6%

      \[x + \frac{y}{1.1283791670955126 \cdot e^{z} - x \cdot y} \]

   2. Taylor expanded in y around inf

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

      1. lower--.f64 N/A

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

      2. lower-/.f64 69.4

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

   4. Applied rewrites 69.4%

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

   if -5 < z < 1.20000000000000001e73

   1. Initial program 95.6%

      \[x + \frac{y}{1.1283791670955126 \cdot e^{z} - x \cdot y} \]

   2. Taylor expanded in z around 0

      \[\leadsto x + \frac{y}{\color{blue}{\left(\frac{5641895835477563}{5000000000000000} + \frac{5641895835477563}{5000000000000000} \cdot z\right)} - x \cdot y} \]
   3. Step-by-step derivation

      1. +-commutative N/A

         \[\leadsto x + \frac{y}{\left(\frac{5641895835477563}{5000000000000000} \cdot z + \color{blue}{\frac{5641895835477563}{5000000000000000}}\right) - x \cdot y} \]

      2. *-commutative N/A

         \[\leadsto x + \frac{y}{\left(z \cdot \frac{5641895835477563}{5000000000000000} + \frac{5641895835477563}{5000000000000000}\right) - x \cdot y} \]

      3. lower-fma.f64 81.3

         \[\leadsto x + \frac{y}{\mathsf{fma}\left(z, \color{blue}{1.1283791670955126}, 1.1283791670955126\right) - x \cdot y} \]

   4. Applied rewrites 81.3%

      \[\leadsto x + \frac{y}{\color{blue}{\mathsf{fma}\left(z, 1.1283791670955126, 1.1283791670955126\right)} - x \cdot y} \]

   if 1.20000000000000001e73 < z

   1. Initial program 95.6%

      \[x + \frac{y}{1.1283791670955126 \cdot e^{z} - x \cdot y} \]

   2. Taylor expanded in y around 0

      \[\leadsto \color{blue}{x + \frac{5000000000000000}{5641895835477563} \cdot \frac{y}{e^{z}}} \]
   3. Step-by-step derivation

      1. +-commutative N/A

         \[\leadsto \frac{5000000000000000}{5641895835477563} \cdot \frac{y}{e^{z}} + \color{blue}{x} \]

      2. *-commutative N/A

         \[\leadsto \frac{y}{e^{z}} \cdot \frac{5000000000000000}{5641895835477563} + x \]

      3. lower-fma.f64 N/A

         \[\leadsto \mathsf{fma}\left(\frac{y}{e^{z}}, \color{blue}{\frac{5000000000000000}{5641895835477563}}, x\right) \]

      4. lower-/.f64 N/A

         \[\leadsto \mathsf{fma}\left(\frac{y}{e^{z}}, \frac{5000000000000000}{5641895835477563}, x\right) \]

      5. lift-exp.f64 62.6

         \[\leadsto \mathsf{fma}\left(\frac{y}{e^{z}}, 0.8862269254527579, x\right) \]

   4. Applied rewrites 62.6%

      \[\leadsto \color{blue}{\mathsf{fma}\left(\frac{y}{e^{z}}, 0.8862269254527579, x\right)} \]

   5. Taylor expanded in z around 0

      \[\leadsto \mathsf{fma}\left(\frac{y}{1 + z \cdot \left(1 + \frac{1}{2} \cdot z\right)}, \frac{5000000000000000}{5641895835477563}, x\right) \]
   6. Step-by-step derivation

      1. +-commutative N/A

         \[\leadsto \mathsf{fma}\left(\frac{y}{z \cdot \left(1 + \frac{1}{2} \cdot z\right) + 1}, \frac{5000000000000000}{5641895835477563}, x\right) \]

      2. *-commutative N/A

         \[\leadsto \mathsf{fma}\left(\frac{y}{\left(1 + \frac{1}{2} \cdot z\right) \cdot z + 1}, \frac{5000000000000000}{5641895835477563}, x\right) \]

      3. lower-fma.f64 N/A

         \[\leadsto \mathsf{fma}\left(\frac{y}{\mathsf{fma}\left(1 + \frac{1}{2} \cdot z, z, 1\right)}, \frac{5000000000000000}{5641895835477563}, x\right) \]

      4. +-commutative N/A

         \[\leadsto \mathsf{fma}\left(\frac{y}{\mathsf{fma}\left(\frac{1}{2} \cdot z + 1, z, 1\right)}, \frac{5000000000000000}{5641895835477563}, x\right) \]

      5. lower-fma.f64 69.8

         \[\leadsto \mathsf{fma}\left(\frac{y}{\mathsf{fma}\left(\mathsf{fma}\left(0.5, z, 1\right), z, 1\right)}, 0.8862269254527579, x\right) \]

   7. Applied rewrites 69.8%

      \[\leadsto \mathsf{fma}\left(\frac{y}{\mathsf{fma}\left(\mathsf{fma}\left(0.5, z, 1\right), z, 1\right)}, 0.8862269254527579, x\right) \]

   8. Taylor expanded in z around inf

      \[\leadsto \mathsf{fma}\left(\frac{y}{\frac{1}{2} \cdot {z}^{2}}, \frac{5000000000000000}{5641895835477563}, x\right) \]
   9. Step-by-step derivation

      1. *-commutative N/A

         \[\leadsto \mathsf{fma}\left(\frac{y}{{z}^{2} \cdot \frac{1}{2}}, \frac{5000000000000000}{5641895835477563}, x\right) \]

      2. lower-*.f64 N/A

         \[\leadsto \mathsf{fma}\left(\frac{y}{{z}^{2} \cdot \frac{1}{2}}, \frac{5000000000000000}{5641895835477563}, x\right) \]

      3. unpow2 N/A

         \[\leadsto \mathsf{fma}\left(\frac{y}{\left(z \cdot z\right) \cdot \frac{1}{2}}, \frac{5000000000000000}{5641895835477563}, x\right) \]

      4. lower-*.f64 41.6

         \[\leadsto \mathsf{fma}\left(\frac{y}{\left(z \cdot z\right) \cdot 0.5}, 0.8862269254527579, x\right) \]

   10. Applied rewrites 41.6%

       \[\leadsto \mathsf{fma}\left(\frac{y}{\left(z \cdot z\right) \cdot 0.5}, 0.8862269254527579, x\right) \]
3. Recombined 3 regimes into one program.
4. Add Preprocessing

Alternative 5: 96.2% accurate, 1.0× speedup

Math

\[\begin{array}{l} \\ \begin{array}{l} \mathbf{if}\;z \leq -5:\\ \;\;\;\;x - \frac{1}{x}\\ \mathbf{elif}\;z \leq 1.2 \cdot 10^{+73}:\\ \;\;\;\;x + \frac{y}{\mathsf{fma}\left(z, 1.1283791670955126, 1.1283791670955126\right) - x \cdot y}\\ \mathbf{else}:\\ \;\;\;\;\mathsf{fma}\left(\frac{y}{\mathsf{fma}\left(\mathsf{fma}\left(0.5, z, 1\right), z, 1\right)}, 0.8862269254527579, x\right)\\ \end{array} \end{array} \]

FPCore

(FPCore (x y z)
 :precision binary64
 (if (<= z -5.0)
   (- x (/ 1.0 x))
   (if (<= z 1.2e+73)
     (+ x (/ y (- (fma z 1.1283791670955126 1.1283791670955126) (* x y))))
     (fma (/ y (fma (fma 0.5 z 1.0) z 1.0)) 0.8862269254527579 x))))

C

double code(double x, double y, double z) {
	double tmp;
	if (z <= -5.0) {
		tmp = x - (1.0 / x);
	} else if (z <= 1.2e+73) {
		tmp = x + (y / (fma(z, 1.1283791670955126, 1.1283791670955126) - (x * y)));
	} else {
		tmp = fma((y / fma(fma(0.5, z, 1.0), z, 1.0)), 0.8862269254527579, x);
	}
	return tmp;
}

Julia

function code(x, y, z)
	tmp = 0.0
	if (z <= -5.0)
		tmp = Float64(x - Float64(1.0 / x));
	elseif (z <= 1.2e+73)
		tmp = Float64(x + Float64(y / Float64(fma(z, 1.1283791670955126, 1.1283791670955126) - Float64(x * y))));
	else
		tmp = fma(Float64(y / fma(fma(0.5, z, 1.0), z, 1.0)), 0.8862269254527579, x);
	end
	return tmp
end

Wolfram

code[x_, y_, z_] := If[LessEqual[z, -5.0], N[(x - N[(1.0 / x), $MachinePrecision]), $MachinePrecision], If[LessEqual[z, 1.2e+73], N[(x + N[(y / N[(N[(z * 1.1283791670955126 + 1.1283791670955126), $MachinePrecision] - N[(x * y), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]), $MachinePrecision], N[(N[(y / N[(N[(0.5 * z + 1.0), $MachinePrecision] * z + 1.0), $MachinePrecision]), $MachinePrecision] * 0.8862269254527579 + x), $MachinePrecision]]]

Derivation

1. Split input into 3 regimes

2. if z < -5

   1. Initial program 95.6%

      \[x + \frac{y}{1.1283791670955126 \cdot e^{z} - x \cdot y} \]

   2. Taylor expanded in y around inf

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

      1. lower--.f64 N/A

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

      2. lower-/.f64 69.4

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

   4. Applied rewrites 69.4%

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

   if -5 < z < 1.20000000000000001e73

   1. Initial program 95.6%

      \[x + \frac{y}{1.1283791670955126 \cdot e^{z} - x \cdot y} \]

   2. Taylor expanded in z around 0

      \[\leadsto x + \frac{y}{\color{blue}{\left(\frac{5641895835477563}{5000000000000000} + \frac{5641895835477563}{5000000000000000} \cdot z\right)} - x \cdot y} \]
   3. Step-by-step derivation

      1. +-commutative N/A

         \[\leadsto x + \frac{y}{\left(\frac{5641895835477563}{5000000000000000} \cdot z + \color{blue}{\frac{5641895835477563}{5000000000000000}}\right) - x \cdot y} \]

      2. *-commutative N/A

         \[\leadsto x + \frac{y}{\left(z \cdot \frac{5641895835477563}{5000000000000000} + \frac{5641895835477563}{5000000000000000}\right) - x \cdot y} \]

      3. lower-fma.f64 81.3

         \[\leadsto x + \frac{y}{\mathsf{fma}\left(z, \color{blue}{1.1283791670955126}, 1.1283791670955126\right) - x \cdot y} \]

   4. Applied rewrites 81.3%

      \[\leadsto x + \frac{y}{\color{blue}{\mathsf{fma}\left(z, 1.1283791670955126, 1.1283791670955126\right)} - x \cdot y} \]

   if 1.20000000000000001e73 < z

   1. Initial program 95.6%

      \[x + \frac{y}{1.1283791670955126 \cdot e^{z} - x \cdot y} \]

   2. Taylor expanded in y around 0

      \[\leadsto \color{blue}{x + \frac{5000000000000000}{5641895835477563} \cdot \frac{y}{e^{z}}} \]
   3. Step-by-step derivation

      1. +-commutative N/A

         \[\leadsto \frac{5000000000000000}{5641895835477563} \cdot \frac{y}{e^{z}} + \color{blue}{x} \]

      2. *-commutative N/A

         \[\leadsto \frac{y}{e^{z}} \cdot \frac{5000000000000000}{5641895835477563} + x \]

      3. lower-fma.f64 N/A

         \[\leadsto \mathsf{fma}\left(\frac{y}{e^{z}}, \color{blue}{\frac{5000000000000000}{5641895835477563}}, x\right) \]

      4. lower-/.f64 N/A

         \[\leadsto \mathsf{fma}\left(\frac{y}{e^{z}}, \frac{5000000000000000}{5641895835477563}, x\right) \]

      5. lift-exp.f64 62.6

         \[\leadsto \mathsf{fma}\left(\frac{y}{e^{z}}, 0.8862269254527579, x\right) \]

   4. Applied rewrites 62.6%

      \[\leadsto \color{blue}{\mathsf{fma}\left(\frac{y}{e^{z}}, 0.8862269254527579, x\right)} \]

   5. Taylor expanded in z around 0

      \[\leadsto \mathsf{fma}\left(\frac{y}{1 + z \cdot \left(1 + \frac{1}{2} \cdot z\right)}, \frac{5000000000000000}{5641895835477563}, x\right) \]
   6. Step-by-step derivation

      1. +-commutative N/A

         \[\leadsto \mathsf{fma}\left(\frac{y}{z \cdot \left(1 + \frac{1}{2} \cdot z\right) + 1}, \frac{5000000000000000}{5641895835477563}, x\right) \]

      2. *-commutative N/A

         \[\leadsto \mathsf{fma}\left(\frac{y}{\left(1 + \frac{1}{2} \cdot z\right) \cdot z + 1}, \frac{5000000000000000}{5641895835477563}, x\right) \]

      3. lower-fma.f64 N/A

         \[\leadsto \mathsf{fma}\left(\frac{y}{\mathsf{fma}\left(1 + \frac{1}{2} \cdot z, z, 1\right)}, \frac{5000000000000000}{5641895835477563}, x\right) \]

      4. +-commutative N/A

         \[\leadsto \mathsf{fma}\left(\frac{y}{\mathsf{fma}\left(\frac{1}{2} \cdot z + 1, z, 1\right)}, \frac{5000000000000000}{5641895835477563}, x\right) \]

      5. lower-fma.f64 69.8

         \[\leadsto \mathsf{fma}\left(\frac{y}{\mathsf{fma}\left(\mathsf{fma}\left(0.5, z, 1\right), z, 1\right)}, 0.8862269254527579, x\right) \]

   7. Applied rewrites 69.8%

      \[\leadsto \mathsf{fma}\left(\frac{y}{\mathsf{fma}\left(\mathsf{fma}\left(0.5, z, 1\right), z, 1\right)}, 0.8862269254527579, x\right) \]
3. Recombined 3 regimes into one program.
4. Add Preprocessing

Alternative 6: 96.0% accurate, 1.1× speedup

Math

\[\begin{array}{l} \\ \begin{array}{l} \mathbf{if}\;z \leq -63:\\ \;\;\;\;x - \frac{1}{x}\\ \mathbf{elif}\;z \leq 5.4 \cdot 10^{+57}:\\ \;\;\;\;x + \frac{y}{\left(\frac{1.1283791670955126}{x} - y\right) \cdot x}\\ \mathbf{else}:\\ \;\;\;\;\mathsf{fma}\left(\frac{y}{\left(z \cdot z\right) \cdot 0.5}, 0.8862269254527579, x\right)\\ \end{array} \end{array} \]

FPCore

(FPCore (x y z)
 :precision binary64
 (if (<= z -63.0)
   (- x (/ 1.0 x))
   (if (<= z 5.4e+57)
     (+ x (/ y (* (- (/ 1.1283791670955126 x) y) x)))
     (fma (/ y (* (* z z) 0.5)) 0.8862269254527579 x))))

C

double code(double x, double y, double z) {
	double tmp;
	if (z <= -63.0) {
		tmp = x - (1.0 / x);
	} else if (z <= 5.4e+57) {
		tmp = x + (y / (((1.1283791670955126 / x) - y) * x));
	} else {
		tmp = fma((y / ((z * z) * 0.5)), 0.8862269254527579, x);
	}
	return tmp;
}

Julia

function code(x, y, z)
	tmp = 0.0
	if (z <= -63.0)
		tmp = Float64(x - Float64(1.0 / x));
	elseif (z <= 5.4e+57)
		tmp = Float64(x + Float64(y / Float64(Float64(Float64(1.1283791670955126 / x) - y) * x)));
	else
		tmp = fma(Float64(y / Float64(Float64(z * z) * 0.5)), 0.8862269254527579, x);
	end
	return tmp
end

Wolfram

code[x_, y_, z_] := If[LessEqual[z, -63.0], N[(x - N[(1.0 / x), $MachinePrecision]), $MachinePrecision], If[LessEqual[z, 5.4e+57], N[(x + N[(y / N[(N[(N[(1.1283791670955126 / x), $MachinePrecision] - y), $MachinePrecision] * x), $MachinePrecision]), $MachinePrecision]), $MachinePrecision], N[(N[(y / N[(N[(z * z), $MachinePrecision] * 0.5), $MachinePrecision]), $MachinePrecision] * 0.8862269254527579 + x), $MachinePrecision]]]

Derivation

1. Split input into 3 regimes

2. if z < -63

   1. Initial program 95.6%

      \[x + \frac{y}{1.1283791670955126 \cdot e^{z} - x \cdot y} \]

   2. Taylor expanded in y around inf

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

      1. lower--.f64 N/A

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

      2. lower-/.f64 69.4

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

   4. Applied rewrites 69.4%

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

   if -63 < z < 5.3999999999999997e57

   1. Initial program 95.6%

      \[x + \frac{y}{1.1283791670955126 \cdot e^{z} - x \cdot y} \]

   2. Taylor expanded in x around inf

      \[\leadsto x + \frac{y}{\color{blue}{x \cdot \left(\frac{5641895835477563}{5000000000000000} \cdot \frac{e^{z}}{x} - y\right)}} \]
   3. Step-by-step derivation

      1. *-commutative N/A

         \[\leadsto x + \frac{y}{\left(\frac{5641895835477563}{5000000000000000} \cdot \frac{e^{z}}{x} - y\right) \cdot \color{blue}{x}} \]

      2. lower-*.f64 N/A

         \[\leadsto x + \frac{y}{\left(\frac{5641895835477563}{5000000000000000} \cdot \frac{e^{z}}{x} - y\right) \cdot \color{blue}{x}} \]

      3. lower--.f64 N/A

         \[\leadsto x + \frac{y}{\left(\frac{5641895835477563}{5000000000000000} \cdot \frac{e^{z}}{x} - y\right) \cdot x} \]

      4. *-commutative N/A

         \[\leadsto x + \frac{y}{\left(\frac{e^{z}}{x} \cdot \frac{5641895835477563}{5000000000000000} - y\right) \cdot x} \]

      5. lower-*.f64 N/A

         \[\leadsto x + \frac{y}{\left(\frac{e^{z}}{x} \cdot \frac{5641895835477563}{5000000000000000} - y\right) \cdot x} \]

      6. lower-/.f64 N/A

         \[\leadsto x + \frac{y}{\left(\frac{e^{z}}{x} \cdot \frac{5641895835477563}{5000000000000000} - y\right) \cdot x} \]

      7. lift-exp.f64 97.2

         \[\leadsto x + \frac{y}{\left(\frac{e^{z}}{x} \cdot 1.1283791670955126 - y\right) \cdot x} \]

   4. Applied rewrites 97.2%

      \[\leadsto x + \frac{y}{\color{blue}{\left(\frac{e^{z}}{x} \cdot 1.1283791670955126 - y\right) \cdot x}} \]

   5. Taylor expanded in z around 0

      \[\leadsto x + \frac{y}{\left(\frac{\frac{5641895835477563}{5000000000000000}}{x} - y\right) \cdot x} \]
   6. Step-by-step derivation

      1. lower-/.f64 80.4

         \[\leadsto x + \frac{y}{\left(\frac{1.1283791670955126}{x} - y\right) \cdot x} \]

   7. Applied rewrites 80.4%

      \[\leadsto x + \frac{y}{\left(\frac{1.1283791670955126}{x} - y\right) \cdot x} \]

   if 5.3999999999999997e57 < z

   1. Initial program 95.6%

      \[x + \frac{y}{1.1283791670955126 \cdot e^{z} - x \cdot y} \]

   2. Taylor expanded in y around 0

      \[\leadsto \color{blue}{x + \frac{5000000000000000}{5641895835477563} \cdot \frac{y}{e^{z}}} \]
   3. Step-by-step derivation

      1. +-commutative N/A

         \[\leadsto \frac{5000000000000000}{5641895835477563} \cdot \frac{y}{e^{z}} + \color{blue}{x} \]

      2. *-commutative N/A

         \[\leadsto \frac{y}{e^{z}} \cdot \frac{5000000000000000}{5641895835477563} + x \]

      3. lower-fma.f64 N/A

         \[\leadsto \mathsf{fma}\left(\frac{y}{e^{z}}, \color{blue}{\frac{5000000000000000}{5641895835477563}}, x\right) \]

      4. lower-/.f64 N/A

         \[\leadsto \mathsf{fma}\left(\frac{y}{e^{z}}, \frac{5000000000000000}{5641895835477563}, x\right) \]

      5. lift-exp.f64 62.6

         \[\leadsto \mathsf{fma}\left(\frac{y}{e^{z}}, 0.8862269254527579, x\right) \]

   4. Applied rewrites 62.6%

      \[\leadsto \color{blue}{\mathsf{fma}\left(\frac{y}{e^{z}}, 0.8862269254527579, x\right)} \]

   5. Taylor expanded in z around 0

      \[\leadsto \mathsf{fma}\left(\frac{y}{1 + z \cdot \left(1 + \frac{1}{2} \cdot z\right)}, \frac{5000000000000000}{5641895835477563}, x\right) \]
   6. Step-by-step derivation

      1. +-commutative N/A

         \[\leadsto \mathsf{fma}\left(\frac{y}{z \cdot \left(1 + \frac{1}{2} \cdot z\right) + 1}, \frac{5000000000000000}{5641895835477563}, x\right) \]

      2. *-commutative N/A

         \[\leadsto \mathsf{fma}\left(\frac{y}{\left(1 + \frac{1}{2} \cdot z\right) \cdot z + 1}, \frac{5000000000000000}{5641895835477563}, x\right) \]

      3. lower-fma.f64 N/A

         \[\leadsto \mathsf{fma}\left(\frac{y}{\mathsf{fma}\left(1 + \frac{1}{2} \cdot z, z, 1\right)}, \frac{5000000000000000}{5641895835477563}, x\right) \]

      4. +-commutative N/A

         \[\leadsto \mathsf{fma}\left(\frac{y}{\mathsf{fma}\left(\frac{1}{2} \cdot z + 1, z, 1\right)}, \frac{5000000000000000}{5641895835477563}, x\right) \]

      5. lower-fma.f64 69.8

         \[\leadsto \mathsf{fma}\left(\frac{y}{\mathsf{fma}\left(\mathsf{fma}\left(0.5, z, 1\right), z, 1\right)}, 0.8862269254527579, x\right) \]

   7. Applied rewrites 69.8%

      \[\leadsto \mathsf{fma}\left(\frac{y}{\mathsf{fma}\left(\mathsf{fma}\left(0.5, z, 1\right), z, 1\right)}, 0.8862269254527579, x\right) \]

   8. Taylor expanded in z around inf

      \[\leadsto \mathsf{fma}\left(\frac{y}{\frac{1}{2} \cdot {z}^{2}}, \frac{5000000000000000}{5641895835477563}, x\right) \]
   9. Step-by-step derivation

      1. *-commutative N/A

         \[\leadsto \mathsf{fma}\left(\frac{y}{{z}^{2} \cdot \frac{1}{2}}, \frac{5000000000000000}{5641895835477563}, x\right) \]

      2. lower-*.f64 N/A

         \[\leadsto \mathsf{fma}\left(\frac{y}{{z}^{2} \cdot \frac{1}{2}}, \frac{5000000000000000}{5641895835477563}, x\right) \]

      3. unpow2 N/A

         \[\leadsto \mathsf{fma}\left(\frac{y}{\left(z \cdot z\right) \cdot \frac{1}{2}}, \frac{5000000000000000}{5641895835477563}, x\right) \]

      4. lower-*.f64 41.6

         \[\leadsto \mathsf{fma}\left(\frac{y}{\left(z \cdot z\right) \cdot 0.5}, 0.8862269254527579, x\right) \]

   10. Applied rewrites 41.6%

       \[\leadsto \mathsf{fma}\left(\frac{y}{\left(z \cdot z\right) \cdot 0.5}, 0.8862269254527579, x\right) \]
3. Recombined 3 regimes into one program.
4. Add Preprocessing

Alternative 7: 95.9% accurate, 1.1× speedup

Math

\[\begin{array}{l} \\ \begin{array}{l} \mathbf{if}\;z \leq -63:\\ \;\;\;\;x - \frac{1}{x}\\ \mathbf{elif}\;z \leq 5.4 \cdot 10^{+57}:\\ \;\;\;\;x + \frac{y}{1.1283791670955126 - x \cdot y}\\ \mathbf{else}:\\ \;\;\;\;\mathsf{fma}\left(\frac{y}{\left(z \cdot z\right) \cdot 0.5}, 0.8862269254527579, x\right)\\ \end{array} \end{array} \]

FPCore

(FPCore (x y z)
 :precision binary64
 (if (<= z -63.0)
   (- x (/ 1.0 x))
   (if (<= z 5.4e+57)
     (+ x (/ y (- 1.1283791670955126 (* x y))))
     (fma (/ y (* (* z z) 0.5)) 0.8862269254527579 x))))

C

double code(double x, double y, double z) {
	double tmp;
	if (z <= -63.0) {
		tmp = x - (1.0 / x);
	} else if (z <= 5.4e+57) {
		tmp = x + (y / (1.1283791670955126 - (x * y)));
	} else {
		tmp = fma((y / ((z * z) * 0.5)), 0.8862269254527579, x);
	}
	return tmp;
}

Julia

function code(x, y, z)
	tmp = 0.0
	if (z <= -63.0)
		tmp = Float64(x - Float64(1.0 / x));
	elseif (z <= 5.4e+57)
		tmp = Float64(x + Float64(y / Float64(1.1283791670955126 - Float64(x * y))));
	else
		tmp = fma(Float64(y / Float64(Float64(z * z) * 0.5)), 0.8862269254527579, x);
	end
	return tmp
end

Wolfram

code[x_, y_, z_] := If[LessEqual[z, -63.0], N[(x - N[(1.0 / x), $MachinePrecision]), $MachinePrecision], If[LessEqual[z, 5.4e+57], N[(x + N[(y / N[(1.1283791670955126 - N[(x * y), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]), $MachinePrecision], N[(N[(y / N[(N[(z * z), $MachinePrecision] * 0.5), $MachinePrecision]), $MachinePrecision] * 0.8862269254527579 + x), $MachinePrecision]]]

Derivation

1. Split input into 3 regimes

2. if z < -63

   1. Initial program 95.6%

      \[x + \frac{y}{1.1283791670955126 \cdot e^{z} - x \cdot y} \]

   2. Taylor expanded in y around inf

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

      1. lower--.f64 N/A

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

      2. lower-/.f64 69.4

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

   4. Applied rewrites 69.4%

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

   if -63 < z < 5.3999999999999997e57

   1. Initial program 95.6%

      \[x + \frac{y}{1.1283791670955126 \cdot e^{z} - x \cdot y} \]

   2. Taylor expanded in z around 0

      \[\leadsto x + \frac{y}{\color{blue}{\frac{5641895835477563}{5000000000000000}} - x \cdot y} \]
   3. Step-by-step derivation

   4. Applied rewrites 80.4%

      \[\leadsto x + \frac{y}{\color{blue}{1.1283791670955126} - x \cdot y} \]

   if 5.3999999999999997e57 < z

   1. Initial program 95.6%

      \[x + \frac{y}{1.1283791670955126 \cdot e^{z} - x \cdot y} \]

   2. Taylor expanded in y around 0

      \[\leadsto \color{blue}{x + \frac{5000000000000000}{5641895835477563} \cdot \frac{y}{e^{z}}} \]
   3. Step-by-step derivation

      1. +-commutative N/A

         \[\leadsto \frac{5000000000000000}{5641895835477563} \cdot \frac{y}{e^{z}} + \color{blue}{x} \]

      2. *-commutative N/A

         \[\leadsto \frac{y}{e^{z}} \cdot \frac{5000000000000000}{5641895835477563} + x \]

      3. lower-fma.f64 N/A

         \[\leadsto \mathsf{fma}\left(\frac{y}{e^{z}}, \color{blue}{\frac{5000000000000000}{5641895835477563}}, x\right) \]

      4. lower-/.f64 N/A

         \[\leadsto \mathsf{fma}\left(\frac{y}{e^{z}}, \frac{5000000000000000}{5641895835477563}, x\right) \]

      5. lift-exp.f64 62.6

         \[\leadsto \mathsf{fma}\left(\frac{y}{e^{z}}, 0.8862269254527579, x\right) \]

   4. Applied rewrites 62.6%

      \[\leadsto \color{blue}{\mathsf{fma}\left(\frac{y}{e^{z}}, 0.8862269254527579, x\right)} \]

   5. Taylor expanded in z around 0

      \[\leadsto \mathsf{fma}\left(\frac{y}{1 + z \cdot \left(1 + \frac{1}{2} \cdot z\right)}, \frac{5000000000000000}{5641895835477563}, x\right) \]
   6. Step-by-step derivation

      1. +-commutative N/A

         \[\leadsto \mathsf{fma}\left(\frac{y}{z \cdot \left(1 + \frac{1}{2} \cdot z\right) + 1}, \frac{5000000000000000}{5641895835477563}, x\right) \]

      2. *-commutative N/A

         \[\leadsto \mathsf{fma}\left(\frac{y}{\left(1 + \frac{1}{2} \cdot z\right) \cdot z + 1}, \frac{5000000000000000}{5641895835477563}, x\right) \]

      3. lower-fma.f64 N/A

         \[\leadsto \mathsf{fma}\left(\frac{y}{\mathsf{fma}\left(1 + \frac{1}{2} \cdot z, z, 1\right)}, \frac{5000000000000000}{5641895835477563}, x\right) \]

      4. +-commutative N/A

         \[\leadsto \mathsf{fma}\left(\frac{y}{\mathsf{fma}\left(\frac{1}{2} \cdot z + 1, z, 1\right)}, \frac{5000000000000000}{5641895835477563}, x\right) \]

      5. lower-fma.f64 69.8

         \[\leadsto \mathsf{fma}\left(\frac{y}{\mathsf{fma}\left(\mathsf{fma}\left(0.5, z, 1\right), z, 1\right)}, 0.8862269254527579, x\right) \]

   7. Applied rewrites 69.8%

      \[\leadsto \mathsf{fma}\left(\frac{y}{\mathsf{fma}\left(\mathsf{fma}\left(0.5, z, 1\right), z, 1\right)}, 0.8862269254527579, x\right) \]

   8. Taylor expanded in z around inf

      \[\leadsto \mathsf{fma}\left(\frac{y}{\frac{1}{2} \cdot {z}^{2}}, \frac{5000000000000000}{5641895835477563}, x\right) \]
   9. Step-by-step derivation

      1. *-commutative N/A

         \[\leadsto \mathsf{fma}\left(\frac{y}{{z}^{2} \cdot \frac{1}{2}}, \frac{5000000000000000}{5641895835477563}, x\right) \]

      2. lower-*.f64 N/A

         \[\leadsto \mathsf{fma}\left(\frac{y}{{z}^{2} \cdot \frac{1}{2}}, \frac{5000000000000000}{5641895835477563}, x\right) \]

      3. unpow2 N/A

         \[\leadsto \mathsf{fma}\left(\frac{y}{\left(z \cdot z\right) \cdot \frac{1}{2}}, \frac{5000000000000000}{5641895835477563}, x\right) \]

      4. lower-*.f64 41.6

         \[\leadsto \mathsf{fma}\left(\frac{y}{\left(z \cdot z\right) \cdot 0.5}, 0.8862269254527579, x\right) \]

   10. Applied rewrites 41.6%

       \[\leadsto \mathsf{fma}\left(\frac{y}{\left(z \cdot z\right) \cdot 0.5}, 0.8862269254527579, x\right) \]
3. Recombined 3 regimes into one program.
4. Add Preprocessing

Alternative 8: 92.5% accurate, 1.3× speedup

Math

\[\begin{array}{l} \\ \begin{array}{l} \mathbf{if}\;z \leq -63:\\ \;\;\;\;x - \frac{1}{x}\\ \mathbf{elif}\;z \leq 3.6 \cdot 10^{+74}:\\ \;\;\;\;x + \frac{y}{1.1283791670955126 - x \cdot y}\\ \mathbf{else}:\\ \;\;\;\;\mathsf{fma}\left(\frac{y}{z - -1}, 0.8862269254527579, x\right)\\ \end{array} \end{array} \]

FPCore

(FPCore (x y z)
 :precision binary64
 (if (<= z -63.0)
   (- x (/ 1.0 x))
   (if (<= z 3.6e+74)
     (+ x (/ y (- 1.1283791670955126 (* x y))))
     (fma (/ y (- z -1.0)) 0.8862269254527579 x))))

C

double code(double x, double y, double z) {
	double tmp;
	if (z <= -63.0) {
		tmp = x - (1.0 / x);
	} else if (z <= 3.6e+74) {
		tmp = x + (y / (1.1283791670955126 - (x * y)));
	} else {
		tmp = fma((y / (z - -1.0)), 0.8862269254527579, x);
	}
	return tmp;
}

Julia

function code(x, y, z)
	tmp = 0.0
	if (z <= -63.0)
		tmp = Float64(x - Float64(1.0 / x));
	elseif (z <= 3.6e+74)
		tmp = Float64(x + Float64(y / Float64(1.1283791670955126 - Float64(x * y))));
	else
		tmp = fma(Float64(y / Float64(z - -1.0)), 0.8862269254527579, x);
	end
	return tmp
end

Wolfram

code[x_, y_, z_] := If[LessEqual[z, -63.0], N[(x - N[(1.0 / x), $MachinePrecision]), $MachinePrecision], If[LessEqual[z, 3.6e+74], N[(x + N[(y / N[(1.1283791670955126 - N[(x * y), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]), $MachinePrecision], N[(N[(y / N[(z - -1.0), $MachinePrecision]), $MachinePrecision] * 0.8862269254527579 + x), $MachinePrecision]]]

Derivation

1. Split input into 3 regimes

2. if z < -63

   1. Initial program 95.6%

      \[x + \frac{y}{1.1283791670955126 \cdot e^{z} - x \cdot y} \]

   2. Taylor expanded in y around inf

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

      1. lower--.f64 N/A

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

      2. lower-/.f64 69.4

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

   4. Applied rewrites 69.4%

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

   if -63 < z < 3.59999999999999988e74

   1. Initial program 95.6%

      \[x + \frac{y}{1.1283791670955126 \cdot e^{z} - x \cdot y} \]

   2. Taylor expanded in z around 0

      \[\leadsto x + \frac{y}{\color{blue}{\frac{5641895835477563}{5000000000000000}} - x \cdot y} \]
   3. Step-by-step derivation

   4. Applied rewrites 80.4%

      \[\leadsto x + \frac{y}{\color{blue}{1.1283791670955126} - x \cdot y} \]

   if 3.59999999999999988e74 < z

   1. Initial program 95.6%

      \[x + \frac{y}{1.1283791670955126 \cdot e^{z} - x \cdot y} \]

   2. Taylor expanded in y around 0

      \[\leadsto \color{blue}{x + \frac{5000000000000000}{5641895835477563} \cdot \frac{y}{e^{z}}} \]
   3. Step-by-step derivation

      1. +-commutative N/A

         \[\leadsto \frac{5000000000000000}{5641895835477563} \cdot \frac{y}{e^{z}} + \color{blue}{x} \]

      2. *-commutative N/A

         \[\leadsto \frac{y}{e^{z}} \cdot \frac{5000000000000000}{5641895835477563} + x \]

      3. lower-fma.f64 N/A

         \[\leadsto \mathsf{fma}\left(\frac{y}{e^{z}}, \color{blue}{\frac{5000000000000000}{5641895835477563}}, x\right) \]

      4. lower-/.f64 N/A

         \[\leadsto \mathsf{fma}\left(\frac{y}{e^{z}}, \frac{5000000000000000}{5641895835477563}, x\right) \]

      5. lift-exp.f64 62.6

         \[\leadsto \mathsf{fma}\left(\frac{y}{e^{z}}, 0.8862269254527579, x\right) \]

   4. Applied rewrites 62.6%

      \[\leadsto \color{blue}{\mathsf{fma}\left(\frac{y}{e^{z}}, 0.8862269254527579, x\right)} \]

   5. Taylor expanded in z around 0

      \[\leadsto \mathsf{fma}\left(\frac{y}{1 + z}, \frac{5000000000000000}{5641895835477563}, x\right) \]
   6. Step-by-step derivation

      1. lower-+.f64 65.6

         \[\leadsto \mathsf{fma}\left(\frac{y}{1 + z}, 0.8862269254527579, x\right) \]

   7. Applied rewrites 65.6%

      \[\leadsto \mathsf{fma}\left(\frac{y}{1 + z}, 0.8862269254527579, x\right) \]
   8. Step-by-step derivation

      1. lift-+.f64 N/A

         \[\leadsto \mathsf{fma}\left(\frac{y}{1 + z}, \frac{5000000000000000}{5641895835477563}, x\right) \]

      2. +-commutative N/A

         \[\leadsto \mathsf{fma}\left(\frac{y}{z + 1}, \frac{5000000000000000}{5641895835477563}, x\right) \]

      3. add-flip N/A

         \[\leadsto \mathsf{fma}\left(\frac{y}{z - \left(\mathsf{neg}\left(1\right)\right)}, \frac{5000000000000000}{5641895835477563}, x\right) \]

      4. metadata-eval N/A

         \[\leadsto \mathsf{fma}\left(\frac{y}{z - -1}, \frac{5000000000000000}{5641895835477563}, x\right) \]

      5. lower--.f64 65.6

         \[\leadsto \mathsf{fma}\left(\frac{y}{z - -1}, 0.8862269254527579, x\right) \]

   9. Applied rewrites 65.6%

      \[\leadsto \mathsf{fma}\left(\frac{y}{z - -1}, 0.8862269254527579, x\right) \]
3. Recombined 3 regimes into one program.
4. Add Preprocessing

Alternative 9: 87.1% accurate, 0.4× speedup

Math

\[\begin{array}{l} \\ \begin{array}{l} t_0 := x - \frac{1}{x}\\ t_1 := x + \frac{y}{1.1283791670955126 \cdot e^{z} - x \cdot y}\\ \mathbf{if}\;t\_1 \leq -1:\\ \;\;\;\;t\_0\\ \mathbf{elif}\;t\_1 \leq 0.0004:\\ \;\;\;\;\mathsf{fma}\left(\frac{y}{z - -1}, 0.8862269254527579, x\right)\\ \mathbf{else}:\\ \;\;\;\;t\_0\\ \end{array} \end{array} \]

FPCore

(FPCore (x y z)
 :precision binary64
 (let* ((t_0 (- x (/ 1.0 x)))
        (t_1 (+ x (/ y (- (* 1.1283791670955126 (exp z)) (* x y))))))
   (if (<= t_1 -1.0)
     t_0
     (if (<= t_1 0.0004) (fma (/ y (- z -1.0)) 0.8862269254527579 x) t_0))))

C

double code(double x, double y, double z) {
	double t_0 = x - (1.0 / x);
	double t_1 = x + (y / ((1.1283791670955126 * exp(z)) - (x * y)));
	double tmp;
	if (t_1 <= -1.0) {
		tmp = t_0;
	} else if (t_1 <= 0.0004) {
		tmp = fma((y / (z - -1.0)), 0.8862269254527579, x);
	} else {
		tmp = t_0;
	}
	return tmp;
}

Julia

function code(x, y, z)
	t_0 = Float64(x - Float64(1.0 / x))
	t_1 = Float64(x + Float64(y / Float64(Float64(1.1283791670955126 * exp(z)) - Float64(x * y))))
	tmp = 0.0
	if (t_1 <= -1.0)
		tmp = t_0;
	elseif (t_1 <= 0.0004)
		tmp = fma(Float64(y / Float64(z - -1.0)), 0.8862269254527579, x);
	else
		tmp = t_0;
	end
	return tmp
end

Wolfram

code[x_, y_, z_] := Block[{t$95$0 = N[(x - N[(1.0 / x), $MachinePrecision]), $MachinePrecision]}, Block[{t$95$1 = N[(x + N[(y / N[(N[(1.1283791670955126 * N[Exp[z], $MachinePrecision]), $MachinePrecision] - N[(x * y), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]}, If[LessEqual[t$95$1, -1.0], t$95$0, If[LessEqual[t$95$1, 0.0004], N[(N[(y / N[(z - -1.0), $MachinePrecision]), $MachinePrecision] * 0.8862269254527579 + x), $MachinePrecision], t$95$0]]]]

Derivation

1. Split input into 2 regimes

2. if (+.f64 x (/.f64 y (-.f64 (*.f64 #s(literal 5641895835477563/5000000000000000 binary64) (exp.f64 z)) (*.f64 x y)))) < -1 or 4.00000000000000019e-4 < (+.f64 x (/.f64 y (-.f64 (*.f64 #s(literal 5641895835477563/5000000000000000 binary64) (exp.f64 z)) (*.f64 x y))))

   1. Initial program 95.6%

      \[x + \frac{y}{1.1283791670955126 \cdot e^{z} - x \cdot y} \]

   2. Taylor expanded in y around inf

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

      1. lower--.f64 N/A

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

      2. lower-/.f64 69.4

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

   4. Applied rewrites 69.4%

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

   if -1 < (+.f64 x (/.f64 y (-.f64 (*.f64 #s(literal 5641895835477563/5000000000000000 binary64) (exp.f64 z)) (*.f64 x y)))) < 4.00000000000000019e-4

   1. Initial program 95.6%

      \[x + \frac{y}{1.1283791670955126 \cdot e^{z} - x \cdot y} \]

   2. Taylor expanded in y around 0

      \[\leadsto \color{blue}{x + \frac{5000000000000000}{5641895835477563} \cdot \frac{y}{e^{z}}} \]
   3. Step-by-step derivation

      1. +-commutative N/A

         \[\leadsto \frac{5000000000000000}{5641895835477563} \cdot \frac{y}{e^{z}} + \color{blue}{x} \]

      2. *-commutative N/A

         \[\leadsto \frac{y}{e^{z}} \cdot \frac{5000000000000000}{5641895835477563} + x \]

      3. lower-fma.f64 N/A

         \[\leadsto \mathsf{fma}\left(\frac{y}{e^{z}}, \color{blue}{\frac{5000000000000000}{5641895835477563}}, x\right) \]

      4. lower-/.f64 N/A

         \[\leadsto \mathsf{fma}\left(\frac{y}{e^{z}}, \frac{5000000000000000}{5641895835477563}, x\right) \]

      5. lift-exp.f64 62.6

         \[\leadsto \mathsf{fma}\left(\frac{y}{e^{z}}, 0.8862269254527579, x\right) \]

   4. Applied rewrites 62.6%

      \[\leadsto \color{blue}{\mathsf{fma}\left(\frac{y}{e^{z}}, 0.8862269254527579, x\right)} \]

   5. Taylor expanded in z around 0

      \[\leadsto \mathsf{fma}\left(\frac{y}{1 + z}, \frac{5000000000000000}{5641895835477563}, x\right) \]
   6. Step-by-step derivation

      1. lower-+.f64 65.6

         \[\leadsto \mathsf{fma}\left(\frac{y}{1 + z}, 0.8862269254527579, x\right) \]

   7. Applied rewrites 65.6%

      \[\leadsto \mathsf{fma}\left(\frac{y}{1 + z}, 0.8862269254527579, x\right) \]
   8. Step-by-step derivation

      1. lift-+.f64 N/A

         \[\leadsto \mathsf{fma}\left(\frac{y}{1 + z}, \frac{5000000000000000}{5641895835477563}, x\right) \]

      2. +-commutative N/A

         \[\leadsto \mathsf{fma}\left(\frac{y}{z + 1}, \frac{5000000000000000}{5641895835477563}, x\right) \]

      3. add-flip N/A

         \[\leadsto \mathsf{fma}\left(\frac{y}{z - \left(\mathsf{neg}\left(1\right)\right)}, \frac{5000000000000000}{5641895835477563}, x\right) \]

      4. metadata-eval N/A

         \[\leadsto \mathsf{fma}\left(\frac{y}{z - -1}, \frac{5000000000000000}{5641895835477563}, x\right) \]

      5. lower--.f64 65.6

         \[\leadsto \mathsf{fma}\left(\frac{y}{z - -1}, 0.8862269254527579, x\right) \]

   9. Applied rewrites 65.6%

      \[\leadsto \mathsf{fma}\left(\frac{y}{z - -1}, 0.8862269254527579, x\right) \]
3. Recombined 2 regimes into one program.
4. Add Preprocessing

Alternative 10: 84.1% accurate, 0.4× speedup

Math

\[\begin{array}{l} \\ \begin{array}{l} t_0 := x - \frac{1}{x}\\ t_1 := x + \frac{y}{1.1283791670955126 \cdot e^{z} - x \cdot y}\\ \mathbf{if}\;t\_1 \leq -1:\\ \;\;\;\;t\_0\\ \mathbf{elif}\;t\_1 \leq 0.0004:\\ \;\;\;\;\mathsf{fma}\left(0.8862269254527579, y, x\right)\\ \mathbf{else}:\\ \;\;\;\;t\_0\\ \end{array} \end{array} \]

FPCore

(FPCore (x y z)
 :precision binary64
 (let* ((t_0 (- x (/ 1.0 x)))
        (t_1 (+ x (/ y (- (* 1.1283791670955126 (exp z)) (* x y))))))
   (if (<= t_1 -1.0)
     t_0
     (if (<= t_1 0.0004) (fma 0.8862269254527579 y x) t_0))))

C

double code(double x, double y, double z) {
	double t_0 = x - (1.0 / x);
	double t_1 = x + (y / ((1.1283791670955126 * exp(z)) - (x * y)));
	double tmp;
	if (t_1 <= -1.0) {
		tmp = t_0;
	} else if (t_1 <= 0.0004) {
		tmp = fma(0.8862269254527579, y, x);
	} else {
		tmp = t_0;
	}
	return tmp;
}

Julia

function code(x, y, z)
	t_0 = Float64(x - Float64(1.0 / x))
	t_1 = Float64(x + Float64(y / Float64(Float64(1.1283791670955126 * exp(z)) - Float64(x * y))))
	tmp = 0.0
	if (t_1 <= -1.0)
		tmp = t_0;
	elseif (t_1 <= 0.0004)
		tmp = fma(0.8862269254527579, y, x);
	else
		tmp = t_0;
	end
	return tmp
end

Wolfram

code[x_, y_, z_] := Block[{t$95$0 = N[(x - N[(1.0 / x), $MachinePrecision]), $MachinePrecision]}, Block[{t$95$1 = N[(x + N[(y / N[(N[(1.1283791670955126 * N[Exp[z], $MachinePrecision]), $MachinePrecision] - N[(x * y), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]}, If[LessEqual[t$95$1, -1.0], t$95$0, If[LessEqual[t$95$1, 0.0004], N[(0.8862269254527579 * y + x), $MachinePrecision], t$95$0]]]]

Derivation

1. Split input into 2 regimes

2. if (+.f64 x (/.f64 y (-.f64 (*.f64 #s(literal 5641895835477563/5000000000000000 binary64) (exp.f64 z)) (*.f64 x y)))) < -1 or 4.00000000000000019e-4 < (+.f64 x (/.f64 y (-.f64 (*.f64 #s(literal 5641895835477563/5000000000000000 binary64) (exp.f64 z)) (*.f64 x y))))

   1. Initial program 95.6%

      \[x + \frac{y}{1.1283791670955126 \cdot e^{z} - x \cdot y} \]

   2. Taylor expanded in y around inf

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

      1. lower--.f64 N/A

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

      2. lower-/.f64 69.4

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

   4. Applied rewrites 69.4%

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

   if -1 < (+.f64 x (/.f64 y (-.f64 (*.f64 #s(literal 5641895835477563/5000000000000000 binary64) (exp.f64 z)) (*.f64 x y)))) < 4.00000000000000019e-4

   1. Initial program 95.6%

      \[x + \frac{y}{1.1283791670955126 \cdot e^{z} - x \cdot y} \]

   2. Taylor expanded in y around 0

      \[\leadsto \color{blue}{x + \frac{5000000000000000}{5641895835477563} \cdot \frac{y}{e^{z}}} \]
   3. Step-by-step derivation

      1. +-commutative N/A

         \[\leadsto \frac{5000000000000000}{5641895835477563} \cdot \frac{y}{e^{z}} + \color{blue}{x} \]

      2. *-commutative N/A

         \[\leadsto \frac{y}{e^{z}} \cdot \frac{5000000000000000}{5641895835477563} + x \]

      3. lower-fma.f64 N/A

         \[\leadsto \mathsf{fma}\left(\frac{y}{e^{z}}, \color{blue}{\frac{5000000000000000}{5641895835477563}}, x\right) \]

      4. lower-/.f64 N/A

         \[\leadsto \mathsf{fma}\left(\frac{y}{e^{z}}, \frac{5000000000000000}{5641895835477563}, x\right) \]

      5. lift-exp.f64 62.6

         \[\leadsto \mathsf{fma}\left(\frac{y}{e^{z}}, 0.8862269254527579, x\right) \]

   4. Applied rewrites 62.6%

      \[\leadsto \color{blue}{\mathsf{fma}\left(\frac{y}{e^{z}}, 0.8862269254527579, x\right)} \]

   5. Taylor expanded in z around 0

      \[\leadsto x + \color{blue}{\frac{5000000000000000}{5641895835477563} \cdot y} \]
   6. Step-by-step derivation

      1. +-commutative N/A

         \[\leadsto \frac{5000000000000000}{5641895835477563} \cdot y + x \]

      2. lower-fma.f64 58.5

         \[\leadsto \mathsf{fma}\left(0.8862269254527579, y, x\right) \]

   7. Applied rewrites 58.5%

      \[\leadsto \mathsf{fma}\left(0.8862269254527579, \color{blue}{y}, x\right) \]
3. Recombined 2 regimes into one program.
4. Add Preprocessing

Alternative 11: 62.0% accurate, 2.6× speedup

Math

\[\begin{array}{l} \\ \begin{array}{l} \mathbf{if}\;z \leq -1100:\\ \;\;\;\;\frac{-1}{x}\\ \mathbf{else}:\\ \;\;\;\;\mathsf{fma}\left(0.8862269254527579, y, x\right)\\ \end{array} \end{array} \]

FPCore

(FPCore (x y z)
 :precision binary64
 (if (<= z -1100.0) (/ -1.0 x) (fma 0.8862269254527579 y x)))

C

double code(double x, double y, double z) {
	double tmp;
	if (z <= -1100.0) {
		tmp = -1.0 / x;
	} else {
		tmp = fma(0.8862269254527579, y, x);
	}
	return tmp;
}

Julia

function code(x, y, z)
	tmp = 0.0
	if (z <= -1100.0)
		tmp = Float64(-1.0 / x);
	else
		tmp = fma(0.8862269254527579, y, x);
	end
	return tmp
end

Wolfram

code[x_, y_, z_] := If[LessEqual[z, -1100.0], N[(-1.0 / x), $MachinePrecision], N[(0.8862269254527579 * y + x), $MachinePrecision]]

Derivation

1. Split input into 2 regimes

2. if z < -1100

   1. Initial program 95.6%

      \[x + \frac{y}{1.1283791670955126 \cdot e^{z} - x \cdot y} \]

   2. Taylor expanded in y around inf

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

      1. lower--.f64 N/A

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

      2. lower-/.f64 69.4

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

   4. Applied rewrites 69.4%

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

   5. Taylor expanded in x around 0

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

      1. lower-/.f64 20.1

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

   7. Applied rewrites 20.1%

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

   if -1100 < z

   1. Initial program 95.6%

      \[x + \frac{y}{1.1283791670955126 \cdot e^{z} - x \cdot y} \]

   2. Taylor expanded in y around 0

      \[\leadsto \color{blue}{x + \frac{5000000000000000}{5641895835477563} \cdot \frac{y}{e^{z}}} \]
   3. Step-by-step derivation

      1. +-commutative N/A

         \[\leadsto \frac{5000000000000000}{5641895835477563} \cdot \frac{y}{e^{z}} + \color{blue}{x} \]

      2. *-commutative N/A

         \[\leadsto \frac{y}{e^{z}} \cdot \frac{5000000000000000}{5641895835477563} + x \]

      3. lower-fma.f64 N/A

         \[\leadsto \mathsf{fma}\left(\frac{y}{e^{z}}, \color{blue}{\frac{5000000000000000}{5641895835477563}}, x\right) \]

      4. lower-/.f64 N/A

         \[\leadsto \mathsf{fma}\left(\frac{y}{e^{z}}, \frac{5000000000000000}{5641895835477563}, x\right) \]

      5. lift-exp.f64 62.6

         \[\leadsto \mathsf{fma}\left(\frac{y}{e^{z}}, 0.8862269254527579, x\right) \]

   4. Applied rewrites 62.6%

      \[\leadsto \color{blue}{\mathsf{fma}\left(\frac{y}{e^{z}}, 0.8862269254527579, x\right)} \]

   5. Taylor expanded in z around 0

      \[\leadsto x + \color{blue}{\frac{5000000000000000}{5641895835477563} \cdot y} \]
   6. Step-by-step derivation

      1. +-commutative N/A

         \[\leadsto \frac{5000000000000000}{5641895835477563} \cdot y + x \]

      2. lower-fma.f64 58.5

         \[\leadsto \mathsf{fma}\left(0.8862269254527579, y, x\right) \]

   7. Applied rewrites 58.5%

      \[\leadsto \mathsf{fma}\left(0.8862269254527579, \color{blue}{y}, x\right) \]
3. Recombined 2 regimes into one program.
4. Add Preprocessing

Alternative 12: 26.1% accurate, 1.2× speedup

Math

\[\begin{array}{l} \\ \begin{array}{l} \mathbf{if}\;1.1283791670955126 \cdot e^{z} \leq 0.00019:\\ \;\;\;\;\frac{-1}{x}\\ \mathbf{else}:\\ \;\;\;\;0.8862269254527579 \cdot y\\ \end{array} \end{array} \]

FPCore

(FPCore (x y z)
 :precision binary64
 (if (<= (* 1.1283791670955126 (exp z)) 0.00019)
   (/ -1.0 x)
   (* 0.8862269254527579 y)))

C

double code(double x, double y, double z) {
	double tmp;
	if ((1.1283791670955126 * exp(z)) <= 0.00019) {
		tmp = -1.0 / x;
	} else {
		tmp = 0.8862269254527579 * y;
	}
	return tmp;
}

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(x, y, z)
use fmin_fmax_functions
    real(8), intent (in) :: x
    real(8), intent (in) :: y
    real(8), intent (in) :: z
    real(8) :: tmp
    if ((1.1283791670955126d0 * exp(z)) <= 0.00019d0) then
        tmp = (-1.0d0) / x
    else
        tmp = 0.8862269254527579d0 * y
    end if
    code = tmp
end function

Java

public static double code(double x, double y, double z) {
	double tmp;
	if ((1.1283791670955126 * Math.exp(z)) <= 0.00019) {
		tmp = -1.0 / x;
	} else {
		tmp = 0.8862269254527579 * y;
	}
	return tmp;
}

Python

def code(x, y, z):
	tmp = 0
	if (1.1283791670955126 * math.exp(z)) <= 0.00019:
		tmp = -1.0 / x
	else:
		tmp = 0.8862269254527579 * y
	return tmp

Julia

function code(x, y, z)
	tmp = 0.0
	if (Float64(1.1283791670955126 * exp(z)) <= 0.00019)
		tmp = Float64(-1.0 / x);
	else
		tmp = Float64(0.8862269254527579 * y);
	end
	return tmp
end

MATLAB

function tmp_2 = code(x, y, z)
	tmp = 0.0;
	if ((1.1283791670955126 * exp(z)) <= 0.00019)
		tmp = -1.0 / x;
	else
		tmp = 0.8862269254527579 * y;
	end
	tmp_2 = tmp;
end

Wolfram

code[x_, y_, z_] := If[LessEqual[N[(1.1283791670955126 * N[Exp[z], $MachinePrecision]), $MachinePrecision], 0.00019], N[(-1.0 / x), $MachinePrecision], N[(0.8862269254527579 * y), $MachinePrecision]]

Derivation

1. Split input into 2 regimes

2. if (*.f64 #s(literal 5641895835477563/5000000000000000 binary64) (exp.f64 z)) < 1.9000000000000001e-4

   1. Initial program 95.6%

      \[x + \frac{y}{1.1283791670955126 \cdot e^{z} - x \cdot y} \]

   2. Taylor expanded in y around inf

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

      1. lower--.f64 N/A

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

      2. lower-/.f64 69.4

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

   4. Applied rewrites 69.4%

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

   5. Taylor expanded in x around 0

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

      1. lower-/.f64 20.1

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

   7. Applied rewrites 20.1%

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

   if 1.9000000000000001e-4 < (*.f64 #s(literal 5641895835477563/5000000000000000 binary64) (exp.f64 z))

   1. Initial program 95.6%

      \[x + \frac{y}{1.1283791670955126 \cdot e^{z} - x \cdot y} \]

   2. Taylor expanded in x around 0

      \[\leadsto \color{blue}{\frac{5000000000000000}{5641895835477563} \cdot \frac{y}{e^{z}}} \]
   3. Step-by-step derivation

      1. *-commutative N/A

         \[\leadsto \frac{y}{e^{z}} \cdot \color{blue}{\frac{5000000000000000}{5641895835477563}} \]

      2. lower-*.f64 N/A

         \[\leadsto \frac{y}{e^{z}} \cdot \color{blue}{\frac{5000000000000000}{5641895835477563}} \]

      3. lower-/.f64 N/A

         \[\leadsto \frac{y}{e^{z}} \cdot \frac{5000000000000000}{5641895835477563} \]

      4. lift-exp.f64 14.2

         \[\leadsto \frac{y}{e^{z}} \cdot 0.8862269254527579 \]

   4. Applied rewrites 14.2%

      \[\leadsto \color{blue}{\frac{y}{e^{z}} \cdot 0.8862269254527579} \]

   5. Taylor expanded in z around 0

      \[\leadsto \frac{5000000000000000}{5641895835477563} \cdot \color{blue}{y} \]
   6. Step-by-step derivation

      1. lower-*.f64 13.8

         \[\leadsto 0.8862269254527579 \cdot y \]

   7. Applied rewrites 13.8%

      \[\leadsto 0.8862269254527579 \cdot \color{blue}{y} \]
3. Recombined 2 regimes into one program.
4. Add Preprocessing

Alternative 13: 13.8% accurate, 6.5× speedup

Math

\[\begin{array}{l} \\ 0.8862269254527579 \cdot y \end{array} \]

FPCore

(FPCore (x y z) :precision binary64 (* 0.8862269254527579 y))

C

double code(double x, double y, double z) {
	return 0.8862269254527579 * y;
}

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(x, y, z)
use fmin_fmax_functions
    real(8), intent (in) :: x
    real(8), intent (in) :: y
    real(8), intent (in) :: z
    code = 0.8862269254527579d0 * y
end function

Java

public static double code(double x, double y, double z) {
	return 0.8862269254527579 * y;
}

Python

def code(x, y, z):
	return 0.8862269254527579 * y

Julia

function code(x, y, z)
	return Float64(0.8862269254527579 * y)
end

MATLAB

function tmp = code(x, y, z)
	tmp = 0.8862269254527579 * y;
end

Wolfram

code[x_, y_, z_] := N[(0.8862269254527579 * y), $MachinePrecision]

Derivation

1. Initial program 95.6%

   \[x + \frac{y}{1.1283791670955126 \cdot e^{z} - x \cdot y} \]

2. Taylor expanded in x around 0

   \[\leadsto \color{blue}{\frac{5000000000000000}{5641895835477563} \cdot \frac{y}{e^{z}}} \]
3. Step-by-step derivation

   1. *-commutative N/A

      \[\leadsto \frac{y}{e^{z}} \cdot \color{blue}{\frac{5000000000000000}{5641895835477563}} \]

   2. lower-*.f64 N/A

      \[\leadsto \frac{y}{e^{z}} \cdot \color{blue}{\frac{5000000000000000}{5641895835477563}} \]

   3. lower-/.f64 N/A

      \[\leadsto \frac{y}{e^{z}} \cdot \frac{5000000000000000}{5641895835477563} \]

   4. lift-exp.f64 14.2

      \[\leadsto \frac{y}{e^{z}} \cdot 0.8862269254527579 \]

4. Applied rewrites 14.2%

   \[\leadsto \color{blue}{\frac{y}{e^{z}} \cdot 0.8862269254527579} \]

5. Taylor expanded in z around 0

   \[\leadsto \frac{5000000000000000}{5641895835477563} \cdot \color{blue}{y} \]
6. Step-by-step derivation

   1. lower-*.f64 13.8

      \[\leadsto 0.8862269254527579 \cdot y \]

7. Applied rewrites 13.8%

   \[\leadsto 0.8862269254527579 \cdot \color{blue}{y} \]
8. Add Preprocessing

Reproduce

herbie shell --seed 2025136 
(FPCore (x y z)
  :name "Numeric.SpecFunctions:invErfc from math-functions-0.1.5.2, A"
  :precision binary64
  (+ x (/ y (- (* 1.1283791670955126 (exp z)) (* x y)))))