Example 2 from Robby

Percentage Accurate: 99.8% → 99.8%

Time: 8.6s

Alternatives: 9

Speedup: N/A×

Specification

Math

\[\begin{array}{l} \\ \begin{array}{l} t_1 := \tan^{-1} \left(\frac{\left(-eh\right) \cdot \tan t}{ew}\right)\\ \left|\left(ew \cdot \cos t\right) \cdot \cos t\_1 - \left(eh \cdot \sin t\right) \cdot \sin t\_1\right| \end{array} \end{array} \]

FPCore

(FPCore (eh ew t)
 :precision binary64
 (let* ((t_1 (atan (/ (* (- eh) (tan t)) ew))))
   (fabs (- (* (* ew (cos t)) (cos t_1)) (* (* eh (sin t)) (sin t_1))))))

C

double code(double eh, double ew, double t) {
	double t_1 = atan(((-eh * tan(t)) / ew));
	return fabs((((ew * cos(t)) * cos(t_1)) - ((eh * sin(t)) * sin(t_1))));
}

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(eh, ew, t)
use fmin_fmax_functions
    real(8), intent (in) :: eh
    real(8), intent (in) :: ew
    real(8), intent (in) :: t
    real(8) :: t_1
    t_1 = atan(((-eh * tan(t)) / ew))
    code = abs((((ew * cos(t)) * cos(t_1)) - ((eh * sin(t)) * sin(t_1))))
end function

Java

public static double code(double eh, double ew, double t) {
	double t_1 = Math.atan(((-eh * Math.tan(t)) / ew));
	return Math.abs((((ew * Math.cos(t)) * Math.cos(t_1)) - ((eh * Math.sin(t)) * Math.sin(t_1))));
}

Python

def code(eh, ew, t):
	t_1 = math.atan(((-eh * math.tan(t)) / ew))
	return math.fabs((((ew * math.cos(t)) * math.cos(t_1)) - ((eh * math.sin(t)) * math.sin(t_1))))

Julia

function code(eh, ew, t)
	t_1 = atan(Float64(Float64(Float64(-eh) * tan(t)) / ew))
	return abs(Float64(Float64(Float64(ew * cos(t)) * cos(t_1)) - Float64(Float64(eh * sin(t)) * sin(t_1))))
end

MATLAB

function tmp = code(eh, ew, t)
	t_1 = atan(((-eh * tan(t)) / ew));
	tmp = abs((((ew * cos(t)) * cos(t_1)) - ((eh * sin(t)) * sin(t_1))));
end

Wolfram

code[eh_, ew_, t_] := Block[{t$95$1 = N[ArcTan[N[(N[((-eh) * N[Tan[t], $MachinePrecision]), $MachinePrecision] / ew), $MachinePrecision]], $MachinePrecision]}, N[Abs[N[(N[(N[(ew * N[Cos[t], $MachinePrecision]), $MachinePrecision] * N[Cos[t$95$1], $MachinePrecision]), $MachinePrecision] - N[(N[(eh * N[Sin[t], $MachinePrecision]), $MachinePrecision] * N[Sin[t$95$1], $MachinePrecision]), $MachinePrecision]), $MachinePrecision]], $MachinePrecision]]

Initial Program: 99.8% accurate, 1.0× speedup

Math

\[\begin{array}{l} \\ \begin{array}{l} t_1 := \tan^{-1} \left(\frac{\left(-eh\right) \cdot \tan t}{ew}\right)\\ \left|\left(ew \cdot \cos t\right) \cdot \cos t\_1 - \left(eh \cdot \sin t\right) \cdot \sin t\_1\right| \end{array} \end{array} \]

FPCore

(FPCore (eh ew t)
 :precision binary64
 (let* ((t_1 (atan (/ (* (- eh) (tan t)) ew))))
   (fabs (- (* (* ew (cos t)) (cos t_1)) (* (* eh (sin t)) (sin t_1))))))

C

double code(double eh, double ew, double t) {
	double t_1 = atan(((-eh * tan(t)) / ew));
	return fabs((((ew * cos(t)) * cos(t_1)) - ((eh * sin(t)) * sin(t_1))));
}

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(eh, ew, t)
use fmin_fmax_functions
    real(8), intent (in) :: eh
    real(8), intent (in) :: ew
    real(8), intent (in) :: t
    real(8) :: t_1
    t_1 = atan(((-eh * tan(t)) / ew))
    code = abs((((ew * cos(t)) * cos(t_1)) - ((eh * sin(t)) * sin(t_1))))
end function

Java

public static double code(double eh, double ew, double t) {
	double t_1 = Math.atan(((-eh * Math.tan(t)) / ew));
	return Math.abs((((ew * Math.cos(t)) * Math.cos(t_1)) - ((eh * Math.sin(t)) * Math.sin(t_1))));
}

Python

def code(eh, ew, t):
	t_1 = math.atan(((-eh * math.tan(t)) / ew))
	return math.fabs((((ew * math.cos(t)) * math.cos(t_1)) - ((eh * math.sin(t)) * math.sin(t_1))))

Julia

function code(eh, ew, t)
	t_1 = atan(Float64(Float64(Float64(-eh) * tan(t)) / ew))
	return abs(Float64(Float64(Float64(ew * cos(t)) * cos(t_1)) - Float64(Float64(eh * sin(t)) * sin(t_1))))
end

MATLAB

function tmp = code(eh, ew, t)
	t_1 = atan(((-eh * tan(t)) / ew));
	tmp = abs((((ew * cos(t)) * cos(t_1)) - ((eh * sin(t)) * sin(t_1))));
end

Wolfram

code[eh_, ew_, t_] := Block[{t$95$1 = N[ArcTan[N[(N[((-eh) * N[Tan[t], $MachinePrecision]), $MachinePrecision] / ew), $MachinePrecision]], $MachinePrecision]}, N[Abs[N[(N[(N[(ew * N[Cos[t], $MachinePrecision]), $MachinePrecision] * N[Cos[t$95$1], $MachinePrecision]), $MachinePrecision] - N[(N[(eh * N[Sin[t], $MachinePrecision]), $MachinePrecision] * N[Sin[t$95$1], $MachinePrecision]), $MachinePrecision]), $MachinePrecision]], $MachinePrecision]]

Alternative 1: 99.8% accurate, 1.2× speedup

Math

\[\begin{array}{l} \\ \begin{array}{l} t_1 := \sinh^{-1} \left(\frac{\tan t}{ew} \cdot eh\right)\\ \left|\mathsf{fma}\left(\frac{ew}{\cosh t\_1}, \cos t, \tanh t\_1 \cdot \left(\sin t \cdot eh\right)\right)\right| \end{array} \end{array} \]

FPCore

(FPCore (eh ew t)
 :precision binary64
 (let* ((t_1 (asinh (* (/ (tan t) ew) eh))))
   (fabs (fma (/ ew (cosh t_1)) (cos t) (* (tanh t_1) (* (sin t) eh))))))

C

double code(double eh, double ew, double t) {
	double t_1 = asinh(((tan(t) / ew) * eh));
	return fabs(fma((ew / cosh(t_1)), cos(t), (tanh(t_1) * (sin(t) * eh))));
}

Julia

function code(eh, ew, t)
	t_1 = asinh(Float64(Float64(tan(t) / ew) * eh))
	return abs(fma(Float64(ew / cosh(t_1)), cos(t), Float64(tanh(t_1) * Float64(sin(t) * eh))))
end

Wolfram

code[eh_, ew_, t_] := Block[{t$95$1 = N[ArcSinh[N[(N[(N[Tan[t], $MachinePrecision] / ew), $MachinePrecision] * eh), $MachinePrecision]], $MachinePrecision]}, N[Abs[N[(N[(ew / N[Cosh[t$95$1], $MachinePrecision]), $MachinePrecision] * N[Cos[t], $MachinePrecision] + N[(N[Tanh[t$95$1], $MachinePrecision] * N[(N[Sin[t], $MachinePrecision] * eh), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]], $MachinePrecision]]

Derivation

1. Initial program 99.8%

   \[\left|\left(ew \cdot \cos t\right) \cdot \cos \tan^{-1} \left(\frac{\left(-eh\right) \cdot \tan t}{ew}\right) - \left(eh \cdot \sin t\right) \cdot \sin \tan^{-1} \left(\frac{\left(-eh\right) \cdot \tan t}{ew}\right)\right| \]

3. Applied rewrites 99.8%

   \[\leadsto \left|\color{blue}{\mathsf{fma}\left(\frac{ew}{\cosh \sinh^{-1} \left(\frac{\tan t}{ew} \cdot eh\right)}, \cos t, \tanh \sinh^{-1} \left(\frac{\tan t}{ew} \cdot eh\right) \cdot \left(\sin t \cdot eh\right)\right)}\right| \]
4. Add Preprocessing

Alternative 2: 98.5% accurate, 1.8× speedup

Math

\[\begin{array}{l} \\ \left|\mathsf{fma}\left(ew, \cos t, \tanh \sinh^{-1} \left(\frac{\tan t}{ew} \cdot eh\right) \cdot \left(\sin t \cdot eh\right)\right)\right| \end{array} \]

FPCore

(FPCore (eh ew t)
 :precision binary64
 (fabs
  (fma ew (cos t) (* (tanh (asinh (* (/ (tan t) ew) eh))) (* (sin t) eh)))))

C

double code(double eh, double ew, double t) {
	return fabs(fma(ew, cos(t), (tanh(asinh(((tan(t) / ew) * eh))) * (sin(t) * eh))));
}

Julia

function code(eh, ew, t)
	return abs(fma(ew, cos(t), Float64(tanh(asinh(Float64(Float64(tan(t) / ew) * eh))) * Float64(sin(t) * eh))))
end

Wolfram

code[eh_, ew_, t_] := N[Abs[N[(ew * N[Cos[t], $MachinePrecision] + N[(N[Tanh[N[ArcSinh[N[(N[(N[Tan[t], $MachinePrecision] / ew), $MachinePrecision] * eh), $MachinePrecision]], $MachinePrecision]], $MachinePrecision] * N[(N[Sin[t], $MachinePrecision] * eh), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]], $MachinePrecision]

Derivation

1. Initial program 99.8%

   \[\left|\left(ew \cdot \cos t\right) \cdot \cos \tan^{-1} \left(\frac{\left(-eh\right) \cdot \tan t}{ew}\right) - \left(eh \cdot \sin t\right) \cdot \sin \tan^{-1} \left(\frac{\left(-eh\right) \cdot \tan t}{ew}\right)\right| \]

3. Applied rewrites 99.8%

   \[\leadsto \left|\color{blue}{\mathsf{fma}\left(\frac{ew}{\cosh \sinh^{-1} \left(\frac{\tan t}{ew} \cdot eh\right)}, \cos t, \tanh \sinh^{-1} \left(\frac{\tan t}{ew} \cdot eh\right) \cdot \left(\sin t \cdot eh\right)\right)}\right| \]

4. Taylor expanded in eh around 0

   \[\leadsto \left|\mathsf{fma}\left(\color{blue}{ew}, \cos t, \tanh \sinh^{-1} \left(\frac{\tan t}{ew} \cdot eh\right) \cdot \left(\sin t \cdot eh\right)\right)\right| \]
5. Step-by-step derivation

6. Applied rewrites 98.5%

   \[\leadsto \left|\mathsf{fma}\left(\color{blue}{ew}, \cos t, \tanh \sinh^{-1} \left(\frac{\tan t}{ew} \cdot eh\right) \cdot \left(\sin t \cdot eh\right)\right)\right| \]
7. Add Preprocessing

Alternative 3: 90.4% accurate, 1.9× speedup

Math

\[\begin{array}{l} \\ \begin{array}{l} t_1 := \sinh^{-1} \left(\frac{t}{ew} \cdot eh\right)\\ \mathbf{if}\;ew \leq 4.2 \cdot 10^{+93}:\\ \;\;\;\;\left|\mathsf{fma}\left(\frac{ew}{\cosh t\_1}, \cos t, \tanh t\_1 \cdot \left(\sin t \cdot eh\right)\right)\right|\\ \mathbf{else}:\\ \;\;\;\;\left|ew \cdot \cos t\right|\\ \end{array} \end{array} \]

FPCore

(FPCore (eh ew t)
 :precision binary64
 (let* ((t_1 (asinh (* (/ t ew) eh))))
   (if (<= ew 4.2e+93)
     (fabs (fma (/ ew (cosh t_1)) (cos t) (* (tanh t_1) (* (sin t) eh))))
     (fabs (* ew (cos t))))))

C

double code(double eh, double ew, double t) {
	double t_1 = asinh(((t / ew) * eh));
	double tmp;
	if (ew <= 4.2e+93) {
		tmp = fabs(fma((ew / cosh(t_1)), cos(t), (tanh(t_1) * (sin(t) * eh))));
	} else {
		tmp = fabs((ew * cos(t)));
	}
	return tmp;
}

Julia

function code(eh, ew, t)
	t_1 = asinh(Float64(Float64(t / ew) * eh))
	tmp = 0.0
	if (ew <= 4.2e+93)
		tmp = abs(fma(Float64(ew / cosh(t_1)), cos(t), Float64(tanh(t_1) * Float64(sin(t) * eh))));
	else
		tmp = abs(Float64(ew * cos(t)));
	end
	return tmp
end

Wolfram

code[eh_, ew_, t_] := Block[{t$95$1 = N[ArcSinh[N[(N[(t / ew), $MachinePrecision] * eh), $MachinePrecision]], $MachinePrecision]}, If[LessEqual[ew, 4.2e+93], N[Abs[N[(N[(ew / N[Cosh[t$95$1], $MachinePrecision]), $MachinePrecision] * N[Cos[t], $MachinePrecision] + N[(N[Tanh[t$95$1], $MachinePrecision] * N[(N[Sin[t], $MachinePrecision] * eh), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]], $MachinePrecision], N[Abs[N[(ew * N[Cos[t], $MachinePrecision]), $MachinePrecision]], $MachinePrecision]]]

Derivation

1. Split input into 2 regimes

2. if ew < 4.1999999999999996e93

   1. Initial program 99.8%

      \[\left|\left(ew \cdot \cos t\right) \cdot \cos \tan^{-1} \left(\frac{\left(-eh\right) \cdot \tan t}{ew}\right) - \left(eh \cdot \sin t\right) \cdot \sin \tan^{-1} \left(\frac{\left(-eh\right) \cdot \tan t}{ew}\right)\right| \]

   3. Applied rewrites 99.8%

      \[\leadsto \left|\color{blue}{\mathsf{fma}\left(\frac{ew}{\cosh \sinh^{-1} \left(\frac{\tan t}{ew} \cdot eh\right)}, \cos t, \tanh \sinh^{-1} \left(\frac{\tan t}{ew} \cdot eh\right) \cdot \left(\sin t \cdot eh\right)\right)}\right| \]

   4. Taylor expanded in t around 0

      \[\leadsto \left|\mathsf{fma}\left(\frac{ew}{\cosh \sinh^{-1} \left(\frac{\color{blue}{t}}{ew} \cdot eh\right)}, \cos t, \tanh \sinh^{-1} \left(\frac{\tan t}{ew} \cdot eh\right) \cdot \left(\sin t \cdot eh\right)\right)\right| \]
   5. Step-by-step derivation

   6. Applied rewrites 90.1%

      \[\leadsto \left|\mathsf{fma}\left(\frac{ew}{\cosh \sinh^{-1} \left(\frac{\color{blue}{t}}{ew} \cdot eh\right)}, \cos t, \tanh \sinh^{-1} \left(\frac{\tan t}{ew} \cdot eh\right) \cdot \left(\sin t \cdot eh\right)\right)\right| \]

   7. Taylor expanded in t around 0

      \[\leadsto \left|\mathsf{fma}\left(\frac{ew}{\cosh \sinh^{-1} \left(\frac{t}{ew} \cdot eh\right)}, \cos t, \tanh \sinh^{-1} \left(\frac{\color{blue}{t}}{ew} \cdot eh\right) \cdot \left(\sin t \cdot eh\right)\right)\right| \]
   8. Step-by-step derivation

   9. Applied rewrites 90.2%

      \[\leadsto \left|\mathsf{fma}\left(\frac{ew}{\cosh \sinh^{-1} \left(\frac{t}{ew} \cdot eh\right)}, \cos t, \tanh \sinh^{-1} \left(\frac{\color{blue}{t}}{ew} \cdot eh\right) \cdot \left(\sin t \cdot eh\right)\right)\right| \]

   if 4.1999999999999996e93 < ew

   1. Initial program 99.8%

      \[\left|\left(ew \cdot \cos t\right) \cdot \cos \tan^{-1} \left(\frac{\left(-eh\right) \cdot \tan t}{ew}\right) - \left(eh \cdot \sin t\right) \cdot \sin \tan^{-1} \left(\frac{\left(-eh\right) \cdot \tan t}{ew}\right)\right| \]

   3. Applied rewrites 99.8%

      \[\leadsto \left|\color{blue}{\mathsf{fma}\left(\frac{ew}{\cosh \sinh^{-1} \left(\frac{\tan t}{ew} \cdot eh\right)}, \cos t, \tanh \sinh^{-1} \left(\frac{\tan t}{ew} \cdot eh\right) \cdot \left(\sin t \cdot eh\right)\right)}\right| \]

   4. Taylor expanded in eh around 0

      \[\leadsto \left|\color{blue}{ew \cdot \cos t}\right| \]
   5. Step-by-step derivation

      1. lower-*.f64 N/A

         \[\leadsto \left|ew \cdot \color{blue}{\cos t}\right| \]

      2. lower-cos.f64 61.7

         \[\leadsto \left|ew \cdot \cos t\right| \]

   6. Applied rewrites 61.7%

      \[\leadsto \left|\color{blue}{ew \cdot \cos t}\right| \]
3. Recombined 2 regimes into one program.
4. Add Preprocessing

Alternative 4: 89.9% accurate, 2.1× speedup

Math

\[\begin{array}{l} \\ \begin{array}{l} t_1 := \frac{eh \cdot t}{ew}\\ \mathbf{if}\;ew \leq 3.8 \cdot 10^{+76}:\\ \;\;\;\;\left|\mathsf{fma}\left(\frac{ew}{\cosh t\_1}, \cos t, \tanh t\_1 \cdot \left(\sin t \cdot eh\right)\right)\right|\\ \mathbf{else}:\\ \;\;\;\;\left|ew \cdot \cos t\right|\\ \end{array} \end{array} \]

FPCore

(FPCore (eh ew t)
 :precision binary64
 (let* ((t_1 (/ (* eh t) ew)))
   (if (<= ew 3.8e+76)
     (fabs (fma (/ ew (cosh t_1)) (cos t) (* (tanh t_1) (* (sin t) eh))))
     (fabs (* ew (cos t))))))

C

double code(double eh, double ew, double t) {
	double t_1 = (eh * t) / ew;
	double tmp;
	if (ew <= 3.8e+76) {
		tmp = fabs(fma((ew / cosh(t_1)), cos(t), (tanh(t_1) * (sin(t) * eh))));
	} else {
		tmp = fabs((ew * cos(t)));
	}
	return tmp;
}

Julia

function code(eh, ew, t)
	t_1 = Float64(Float64(eh * t) / ew)
	tmp = 0.0
	if (ew <= 3.8e+76)
		tmp = abs(fma(Float64(ew / cosh(t_1)), cos(t), Float64(tanh(t_1) * Float64(sin(t) * eh))));
	else
		tmp = abs(Float64(ew * cos(t)));
	end
	return tmp
end

Wolfram

code[eh_, ew_, t_] := Block[{t$95$1 = N[(N[(eh * t), $MachinePrecision] / ew), $MachinePrecision]}, If[LessEqual[ew, 3.8e+76], N[Abs[N[(N[(ew / N[Cosh[t$95$1], $MachinePrecision]), $MachinePrecision] * N[Cos[t], $MachinePrecision] + N[(N[Tanh[t$95$1], $MachinePrecision] * N[(N[Sin[t], $MachinePrecision] * eh), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]], $MachinePrecision], N[Abs[N[(ew * N[Cos[t], $MachinePrecision]), $MachinePrecision]], $MachinePrecision]]]

Derivation

1. Split input into 2 regimes

2. if ew < 3.80000000000000024e76

   1. Initial program 99.8%

      \[\left|\left(ew \cdot \cos t\right) \cdot \cos \tan^{-1} \left(\frac{\left(-eh\right) \cdot \tan t}{ew}\right) - \left(eh \cdot \sin t\right) \cdot \sin \tan^{-1} \left(\frac{\left(-eh\right) \cdot \tan t}{ew}\right)\right| \]

   3. Applied rewrites 99.8%

      \[\leadsto \left|\color{blue}{\mathsf{fma}\left(\frac{ew}{\cosh \sinh^{-1} \left(\frac{\tan t}{ew} \cdot eh\right)}, \cos t, \tanh \sinh^{-1} \left(\frac{\tan t}{ew} \cdot eh\right) \cdot \left(\sin t \cdot eh\right)\right)}\right| \]

   4. Taylor expanded in t around 0

      \[\leadsto \left|\mathsf{fma}\left(\frac{ew}{\cosh \color{blue}{\left(\frac{eh \cdot t}{ew}\right)}}, \cos t, \tanh \sinh^{-1} \left(\frac{\tan t}{ew} \cdot eh\right) \cdot \left(\sin t \cdot eh\right)\right)\right| \]
   5. Step-by-step derivation

      1. lower-/.f64 N/A

         \[\leadsto \left|\mathsf{fma}\left(\frac{ew}{\cosh \left(\frac{eh \cdot t}{\color{blue}{ew}}\right)}, \cos t, \tanh \sinh^{-1} \left(\frac{\tan t}{ew} \cdot eh\right) \cdot \left(\sin t \cdot eh\right)\right)\right| \]

      2. lower-*.f64 89.6

         \[\leadsto \left|\mathsf{fma}\left(\frac{ew}{\cosh \left(\frac{eh \cdot t}{ew}\right)}, \cos t, \tanh \sinh^{-1} \left(\frac{\tan t}{ew} \cdot eh\right) \cdot \left(\sin t \cdot eh\right)\right)\right| \]

   6. Applied rewrites 89.6%

      \[\leadsto \left|\mathsf{fma}\left(\frac{ew}{\cosh \color{blue}{\left(\frac{eh \cdot t}{ew}\right)}}, \cos t, \tanh \sinh^{-1} \left(\frac{\tan t}{ew} \cdot eh\right) \cdot \left(\sin t \cdot eh\right)\right)\right| \]

   7. Taylor expanded in t around 0

      \[\leadsto \left|\mathsf{fma}\left(\frac{ew}{\cosh \left(\frac{eh \cdot t}{ew}\right)}, \cos t, \tanh \color{blue}{\left(\frac{eh \cdot t}{ew}\right)} \cdot \left(\sin t \cdot eh\right)\right)\right| \]
   8. Step-by-step derivation

      1. lower-/.f64 N/A

         \[\leadsto \left|\mathsf{fma}\left(\frac{ew}{\cosh \left(\frac{eh \cdot t}{ew}\right)}, \cos t, \tanh \left(\frac{eh \cdot t}{\color{blue}{ew}}\right) \cdot \left(\sin t \cdot eh\right)\right)\right| \]

      2. lower-*.f64 89.8

         \[\leadsto \left|\mathsf{fma}\left(\frac{ew}{\cosh \left(\frac{eh \cdot t}{ew}\right)}, \cos t, \tanh \left(\frac{eh \cdot t}{ew}\right) \cdot \left(\sin t \cdot eh\right)\right)\right| \]

   9. Applied rewrites 89.8%

      \[\leadsto \left|\mathsf{fma}\left(\frac{ew}{\cosh \left(\frac{eh \cdot t}{ew}\right)}, \cos t, \tanh \color{blue}{\left(\frac{eh \cdot t}{ew}\right)} \cdot \left(\sin t \cdot eh\right)\right)\right| \]

   if 3.80000000000000024e76 < ew

   1. Initial program 99.8%

      \[\left|\left(ew \cdot \cos t\right) \cdot \cos \tan^{-1} \left(\frac{\left(-eh\right) \cdot \tan t}{ew}\right) - \left(eh \cdot \sin t\right) \cdot \sin \tan^{-1} \left(\frac{\left(-eh\right) \cdot \tan t}{ew}\right)\right| \]

   3. Applied rewrites 99.8%

      \[\leadsto \left|\color{blue}{\mathsf{fma}\left(\frac{ew}{\cosh \sinh^{-1} \left(\frac{\tan t}{ew} \cdot eh\right)}, \cos t, \tanh \sinh^{-1} \left(\frac{\tan t}{ew} \cdot eh\right) \cdot \left(\sin t \cdot eh\right)\right)}\right| \]

   4. Taylor expanded in eh around 0

      \[\leadsto \left|\color{blue}{ew \cdot \cos t}\right| \]
   5. Step-by-step derivation

      1. lower-*.f64 N/A

         \[\leadsto \left|ew \cdot \color{blue}{\cos t}\right| \]

      2. lower-cos.f64 61.7

         \[\leadsto \left|ew \cdot \cos t\right| \]

   6. Applied rewrites 61.7%

      \[\leadsto \left|\color{blue}{ew \cdot \cos t}\right| \]
3. Recombined 2 regimes into one program.
4. Add Preprocessing

Alternative 5: 68.5% accurate, 2.4× speedup

Math

\[\begin{array}{l} \\ \begin{array}{l} \mathbf{if}\;eh \leq 5.4 \cdot 10^{+133}:\\ \;\;\;\;\left|ew \cdot \cos t\right|\\ \mathbf{else}:\\ \;\;\;\;\left|\left(\sin t \cdot eh\right) \cdot \tanh \sinh^{-1} \left(eh \cdot \frac{\tan \left(\pi + t\right)}{ew}\right)\right|\\ \end{array} \end{array} \]

FPCore

(FPCore (eh ew t)
 :precision binary64
 (if (<= eh 5.4e+133)
   (fabs (* ew (cos t)))
   (fabs (* (* (sin t) eh) (tanh (asinh (* eh (/ (tan (+ PI t)) ew))))))))

C

double code(double eh, double ew, double t) {
	double tmp;
	if (eh <= 5.4e+133) {
		tmp = fabs((ew * cos(t)));
	} else {
		tmp = fabs(((sin(t) * eh) * tanh(asinh((eh * (tan((((double) M_PI) + t)) / ew))))));
	}
	return tmp;
}

Python

def code(eh, ew, t):
	tmp = 0
	if eh <= 5.4e+133:
		tmp = math.fabs((ew * math.cos(t)))
	else:
		tmp = math.fabs(((math.sin(t) * eh) * math.tanh(math.asinh((eh * (math.tan((math.pi + t)) / ew))))))
	return tmp

Julia

function code(eh, ew, t)
	tmp = 0.0
	if (eh <= 5.4e+133)
		tmp = abs(Float64(ew * cos(t)));
	else
		tmp = abs(Float64(Float64(sin(t) * eh) * tanh(asinh(Float64(eh * Float64(tan(Float64(pi + t)) / ew))))));
	end
	return tmp
end

MATLAB

function tmp_2 = code(eh, ew, t)
	tmp = 0.0;
	if (eh <= 5.4e+133)
		tmp = abs((ew * cos(t)));
	else
		tmp = abs(((sin(t) * eh) * tanh(asinh((eh * (tan((pi + t)) / ew))))));
	end
	tmp_2 = tmp;
end

Wolfram

code[eh_, ew_, t_] := If[LessEqual[eh, 5.4e+133], N[Abs[N[(ew * N[Cos[t], $MachinePrecision]), $MachinePrecision]], $MachinePrecision], N[Abs[N[(N[(N[Sin[t], $MachinePrecision] * eh), $MachinePrecision] * N[Tanh[N[ArcSinh[N[(eh * N[(N[Tan[N[(Pi + t), $MachinePrecision]], $MachinePrecision] / ew), $MachinePrecision]), $MachinePrecision]], $MachinePrecision]], $MachinePrecision]), $MachinePrecision]], $MachinePrecision]]

Derivation

1. Split input into 2 regimes

2. if eh < 5.4000000000000004e133

   1. Initial program 99.8%

      \[\left|\left(ew \cdot \cos t\right) \cdot \cos \tan^{-1} \left(\frac{\left(-eh\right) \cdot \tan t}{ew}\right) - \left(eh \cdot \sin t\right) \cdot \sin \tan^{-1} \left(\frac{\left(-eh\right) \cdot \tan t}{ew}\right)\right| \]

   3. Applied rewrites 99.8%

      \[\leadsto \left|\color{blue}{\mathsf{fma}\left(\frac{ew}{\cosh \sinh^{-1} \left(\frac{\tan t}{ew} \cdot eh\right)}, \cos t, \tanh \sinh^{-1} \left(\frac{\tan t}{ew} \cdot eh\right) \cdot \left(\sin t \cdot eh\right)\right)}\right| \]

   4. Taylor expanded in eh around 0

      \[\leadsto \left|\color{blue}{ew \cdot \cos t}\right| \]
   5. Step-by-step derivation

      1. lower-*.f64 N/A

         \[\leadsto \left|ew \cdot \color{blue}{\cos t}\right| \]

      2. lower-cos.f64 61.7

         \[\leadsto \left|ew \cdot \cos t\right| \]

   6. Applied rewrites 61.7%

      \[\leadsto \left|\color{blue}{ew \cdot \cos t}\right| \]

   if 5.4000000000000004e133 < eh

   1. Initial program 99.8%

      \[\left|\left(ew \cdot \cos t\right) \cdot \cos \tan^{-1} \left(\frac{\left(-eh\right) \cdot \tan t}{ew}\right) - \left(eh \cdot \sin t\right) \cdot \sin \tan^{-1} \left(\frac{\left(-eh\right) \cdot \tan t}{ew}\right)\right| \]

   2. Taylor expanded in eh around inf

      \[\leadsto \left|\color{blue}{-1 \cdot \left(eh \cdot \left(\sin t \cdot \sin \tan^{-1} \left(-1 \cdot \frac{eh \cdot \sin t}{ew \cdot \cos t}\right)\right)\right)}\right| \]
   3. Step-by-step derivation

      1. lower-*.f64 N/A

         \[\leadsto \left|-1 \cdot \color{blue}{\left(eh \cdot \left(\sin t \cdot \sin \tan^{-1} \left(-1 \cdot \frac{eh \cdot \sin t}{ew \cdot \cos t}\right)\right)\right)}\right| \]

      2. lower-*.f64 N/A

         \[\leadsto \left|-1 \cdot \left(eh \cdot \color{blue}{\left(\sin t \cdot \sin \tan^{-1} \left(-1 \cdot \frac{eh \cdot \sin t}{ew \cdot \cos t}\right)\right)}\right)\right| \]

      3. lower-*.f64 N/A

         \[\leadsto \left|-1 \cdot \left(eh \cdot \left(\sin t \cdot \color{blue}{\sin \tan^{-1} \left(-1 \cdot \frac{eh \cdot \sin t}{ew \cdot \cos t}\right)}\right)\right)\right| \]

      4. lower-sin.f64 N/A

         \[\leadsto \left|-1 \cdot \left(eh \cdot \left(\sin t \cdot \sin \color{blue}{\tan^{-1} \left(-1 \cdot \frac{eh \cdot \sin t}{ew \cdot \cos t}\right)}\right)\right)\right| \]

      5. lower-sin.f64 N/A

         \[\leadsto \left|-1 \cdot \left(eh \cdot \left(\sin t \cdot \sin \tan^{-1} \left(-1 \cdot \frac{eh \cdot \sin t}{ew \cdot \cos t}\right)\right)\right)\right| \]

      6. lower-atan.f64 N/A

         \[\leadsto \left|-1 \cdot \left(eh \cdot \left(\sin t \cdot \sin \tan^{-1} \left(-1 \cdot \frac{eh \cdot \sin t}{ew \cdot \cos t}\right)\right)\right)\right| \]

      7. lower-*.f64 N/A

         \[\leadsto \left|-1 \cdot \left(eh \cdot \left(\sin t \cdot \sin \tan^{-1} \left(-1 \cdot \frac{eh \cdot \sin t}{ew \cdot \cos t}\right)\right)\right)\right| \]

      8. lower-/.f64 N/A

         \[\leadsto \left|-1 \cdot \left(eh \cdot \left(\sin t \cdot \sin \tan^{-1} \left(-1 \cdot \frac{eh \cdot \sin t}{ew \cdot \cos t}\right)\right)\right)\right| \]

      9. lower-*.f64 N/A

         \[\leadsto \left|-1 \cdot \left(eh \cdot \left(\sin t \cdot \sin \tan^{-1} \left(-1 \cdot \frac{eh \cdot \sin t}{ew \cdot \cos t}\right)\right)\right)\right| \]

      10. lower-sin.f64 N/A

          \[\leadsto \left|-1 \cdot \left(eh \cdot \left(\sin t \cdot \sin \tan^{-1} \left(-1 \cdot \frac{eh \cdot \sin t}{ew \cdot \cos t}\right)\right)\right)\right| \]

      11. lower-*.f64 N/A

          \[\leadsto \left|-1 \cdot \left(eh \cdot \left(\sin t \cdot \sin \tan^{-1} \left(-1 \cdot \frac{eh \cdot \sin t}{ew \cdot \cos t}\right)\right)\right)\right| \]

   4. Applied rewrites 41.5%

      \[\leadsto \left|\color{blue}{-1 \cdot \left(eh \cdot \left(\sin t \cdot \sin \tan^{-1} \left(-1 \cdot \frac{eh \cdot \sin t}{ew \cdot \cos t}\right)\right)\right)}\right| \]

   6. Applied rewrites 41.5%

      \[\leadsto \left|\color{blue}{\left(\sin t \cdot eh\right) \cdot \tanh \sinh^{-1} \left(eh \cdot \frac{\tan t}{ew}\right)}\right| \]
   7. Step-by-step derivation

      1. lift-tan.f64 N/A

         \[\leadsto \left|\left(\sin t \cdot eh\right) \cdot \tanh \sinh^{-1} \left(eh \cdot \frac{\tan t}{ew}\right)\right| \]

      2. tan-+PI-rev N/A

         \[\leadsto \left|\left(\sin t \cdot eh\right) \cdot \tanh \sinh^{-1} \left(eh \cdot \frac{\tan \left(t + \mathsf{PI}\left(\right)\right)}{ew}\right)\right| \]

      3. lower-tan.f64 N/A

         \[\leadsto \left|\left(\sin t \cdot eh\right) \cdot \tanh \sinh^{-1} \left(eh \cdot \frac{\tan \left(t + \mathsf{PI}\left(\right)\right)}{ew}\right)\right| \]

      4. +-commutative N/A

         \[\leadsto \left|\left(\sin t \cdot eh\right) \cdot \tanh \sinh^{-1} \left(eh \cdot \frac{\tan \left(\mathsf{PI}\left(\right) + t\right)}{ew}\right)\right| \]

      5. lower-+.f64 N/A

         \[\leadsto \left|\left(\sin t \cdot eh\right) \cdot \tanh \sinh^{-1} \left(eh \cdot \frac{\tan \left(\mathsf{PI}\left(\right) + t\right)}{ew}\right)\right| \]

      6. lower-PI.f64 41.7

         \[\leadsto \left|\left(\sin t \cdot eh\right) \cdot \tanh \sinh^{-1} \left(eh \cdot \frac{\tan \left(\pi + t\right)}{ew}\right)\right| \]

   8. Applied rewrites 41.7%

      \[\leadsto \left|\left(\sin t \cdot eh\right) \cdot \tanh \sinh^{-1} \left(eh \cdot \frac{\tan \left(\pi + t\right)}{ew}\right)\right| \]
3. Recombined 2 regimes into one program.
4. Add Preprocessing

Alternative 6: 68.5% accurate, 3.7× speedup

Math

\[\begin{array}{l} \\ \begin{array}{l} \mathbf{if}\;eh \leq 5.4 \cdot 10^{+133}:\\ \;\;\;\;\left|ew \cdot \cos t\right|\\ \mathbf{else}:\\ \;\;\;\;\left|\left(\sin t \cdot eh\right) \cdot \tanh \sinh^{-1} \left(eh \cdot \frac{t}{ew}\right)\right|\\ \end{array} \end{array} \]

FPCore

(FPCore (eh ew t)
 :precision binary64
 (if (<= eh 5.4e+133)
   (fabs (* ew (cos t)))
   (fabs (* (* (sin t) eh) (tanh (asinh (* eh (/ t ew))))))))

C

double code(double eh, double ew, double t) {
	double tmp;
	if (eh <= 5.4e+133) {
		tmp = fabs((ew * cos(t)));
	} else {
		tmp = fabs(((sin(t) * eh) * tanh(asinh((eh * (t / ew))))));
	}
	return tmp;
}

Python

def code(eh, ew, t):
	tmp = 0
	if eh <= 5.4e+133:
		tmp = math.fabs((ew * math.cos(t)))
	else:
		tmp = math.fabs(((math.sin(t) * eh) * math.tanh(math.asinh((eh * (t / ew))))))
	return tmp

Julia

function code(eh, ew, t)
	tmp = 0.0
	if (eh <= 5.4e+133)
		tmp = abs(Float64(ew * cos(t)));
	else
		tmp = abs(Float64(Float64(sin(t) * eh) * tanh(asinh(Float64(eh * Float64(t / ew))))));
	end
	return tmp
end

MATLAB

function tmp_2 = code(eh, ew, t)
	tmp = 0.0;
	if (eh <= 5.4e+133)
		tmp = abs((ew * cos(t)));
	else
		tmp = abs(((sin(t) * eh) * tanh(asinh((eh * (t / ew))))));
	end
	tmp_2 = tmp;
end

Wolfram

code[eh_, ew_, t_] := If[LessEqual[eh, 5.4e+133], N[Abs[N[(ew * N[Cos[t], $MachinePrecision]), $MachinePrecision]], $MachinePrecision], N[Abs[N[(N[(N[Sin[t], $MachinePrecision] * eh), $MachinePrecision] * N[Tanh[N[ArcSinh[N[(eh * N[(t / ew), $MachinePrecision]), $MachinePrecision]], $MachinePrecision]], $MachinePrecision]), $MachinePrecision]], $MachinePrecision]]

Derivation

1. Split input into 2 regimes

2. if eh < 5.4000000000000004e133

   1. Initial program 99.8%

      \[\left|\left(ew \cdot \cos t\right) \cdot \cos \tan^{-1} \left(\frac{\left(-eh\right) \cdot \tan t}{ew}\right) - \left(eh \cdot \sin t\right) \cdot \sin \tan^{-1} \left(\frac{\left(-eh\right) \cdot \tan t}{ew}\right)\right| \]

   3. Applied rewrites 99.8%

      \[\leadsto \left|\color{blue}{\mathsf{fma}\left(\frac{ew}{\cosh \sinh^{-1} \left(\frac{\tan t}{ew} \cdot eh\right)}, \cos t, \tanh \sinh^{-1} \left(\frac{\tan t}{ew} \cdot eh\right) \cdot \left(\sin t \cdot eh\right)\right)}\right| \]

   4. Taylor expanded in eh around 0

      \[\leadsto \left|\color{blue}{ew \cdot \cos t}\right| \]
   5. Step-by-step derivation

      1. lower-*.f64 N/A

         \[\leadsto \left|ew \cdot \color{blue}{\cos t}\right| \]

      2. lower-cos.f64 61.7

         \[\leadsto \left|ew \cdot \cos t\right| \]

   6. Applied rewrites 61.7%

      \[\leadsto \left|\color{blue}{ew \cdot \cos t}\right| \]

   if 5.4000000000000004e133 < eh

   1. Initial program 99.8%

      \[\left|\left(ew \cdot \cos t\right) \cdot \cos \tan^{-1} \left(\frac{\left(-eh\right) \cdot \tan t}{ew}\right) - \left(eh \cdot \sin t\right) \cdot \sin \tan^{-1} \left(\frac{\left(-eh\right) \cdot \tan t}{ew}\right)\right| \]

   2. Taylor expanded in eh around inf

      \[\leadsto \left|\color{blue}{-1 \cdot \left(eh \cdot \left(\sin t \cdot \sin \tan^{-1} \left(-1 \cdot \frac{eh \cdot \sin t}{ew \cdot \cos t}\right)\right)\right)}\right| \]
   3. Step-by-step derivation

      1. lower-*.f64 N/A

         \[\leadsto \left|-1 \cdot \color{blue}{\left(eh \cdot \left(\sin t \cdot \sin \tan^{-1} \left(-1 \cdot \frac{eh \cdot \sin t}{ew \cdot \cos t}\right)\right)\right)}\right| \]

      2. lower-*.f64 N/A

         \[\leadsto \left|-1 \cdot \left(eh \cdot \color{blue}{\left(\sin t \cdot \sin \tan^{-1} \left(-1 \cdot \frac{eh \cdot \sin t}{ew \cdot \cos t}\right)\right)}\right)\right| \]

      3. lower-*.f64 N/A

         \[\leadsto \left|-1 \cdot \left(eh \cdot \left(\sin t \cdot \color{blue}{\sin \tan^{-1} \left(-1 \cdot \frac{eh \cdot \sin t}{ew \cdot \cos t}\right)}\right)\right)\right| \]

      4. lower-sin.f64 N/A

         \[\leadsto \left|-1 \cdot \left(eh \cdot \left(\sin t \cdot \sin \color{blue}{\tan^{-1} \left(-1 \cdot \frac{eh \cdot \sin t}{ew \cdot \cos t}\right)}\right)\right)\right| \]

      5. lower-sin.f64 N/A

         \[\leadsto \left|-1 \cdot \left(eh \cdot \left(\sin t \cdot \sin \tan^{-1} \left(-1 \cdot \frac{eh \cdot \sin t}{ew \cdot \cos t}\right)\right)\right)\right| \]

      6. lower-atan.f64 N/A

         \[\leadsto \left|-1 \cdot \left(eh \cdot \left(\sin t \cdot \sin \tan^{-1} \left(-1 \cdot \frac{eh \cdot \sin t}{ew \cdot \cos t}\right)\right)\right)\right| \]

      7. lower-*.f64 N/A

         \[\leadsto \left|-1 \cdot \left(eh \cdot \left(\sin t \cdot \sin \tan^{-1} \left(-1 \cdot \frac{eh \cdot \sin t}{ew \cdot \cos t}\right)\right)\right)\right| \]

      8. lower-/.f64 N/A

         \[\leadsto \left|-1 \cdot \left(eh \cdot \left(\sin t \cdot \sin \tan^{-1} \left(-1 \cdot \frac{eh \cdot \sin t}{ew \cdot \cos t}\right)\right)\right)\right| \]

      9. lower-*.f64 N/A

         \[\leadsto \left|-1 \cdot \left(eh \cdot \left(\sin t \cdot \sin \tan^{-1} \left(-1 \cdot \frac{eh \cdot \sin t}{ew \cdot \cos t}\right)\right)\right)\right| \]

      10. lower-sin.f64 N/A

          \[\leadsto \left|-1 \cdot \left(eh \cdot \left(\sin t \cdot \sin \tan^{-1} \left(-1 \cdot \frac{eh \cdot \sin t}{ew \cdot \cos t}\right)\right)\right)\right| \]

      11. lower-*.f64 N/A

          \[\leadsto \left|-1 \cdot \left(eh \cdot \left(\sin t \cdot \sin \tan^{-1} \left(-1 \cdot \frac{eh \cdot \sin t}{ew \cdot \cos t}\right)\right)\right)\right| \]

   4. Applied rewrites 41.5%

      \[\leadsto \left|\color{blue}{-1 \cdot \left(eh \cdot \left(\sin t \cdot \sin \tan^{-1} \left(-1 \cdot \frac{eh \cdot \sin t}{ew \cdot \cos t}\right)\right)\right)}\right| \]

   6. Applied rewrites 41.5%

      \[\leadsto \left|\color{blue}{\left(\sin t \cdot eh\right) \cdot \tanh \sinh^{-1} \left(eh \cdot \frac{\tan t}{ew}\right)}\right| \]

   7. Taylor expanded in t around 0

      \[\leadsto \left|\left(\sin t \cdot eh\right) \cdot \tanh \sinh^{-1} \left(eh \cdot \frac{t}{ew}\right)\right| \]
   8. Step-by-step derivation

      1. lower-/.f64 41.7

         \[\leadsto \left|\left(\sin t \cdot eh\right) \cdot \tanh \sinh^{-1} \left(eh \cdot \frac{t}{ew}\right)\right| \]

   9. Applied rewrites 41.7%

      \[\leadsto \left|\left(\sin t \cdot eh\right) \cdot \tanh \sinh^{-1} \left(eh \cdot \frac{t}{ew}\right)\right| \]
3. Recombined 2 regimes into one program.
4. Add Preprocessing

Alternative 7: 61.7% accurate, 6.7× speedup

Math

\[\begin{array}{l} \\ \left|ew \cdot \cos t\right| \end{array} \]

FPCore

(FPCore (eh ew t) :precision binary64 (fabs (* ew (cos t))))

C

double code(double eh, double ew, double t) {
	return fabs((ew * cos(t)));
}

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(eh, ew, t)
use fmin_fmax_functions
    real(8), intent (in) :: eh
    real(8), intent (in) :: ew
    real(8), intent (in) :: t
    code = abs((ew * cos(t)))
end function

Java

public static double code(double eh, double ew, double t) {
	return Math.abs((ew * Math.cos(t)));
}

Python

def code(eh, ew, t):
	return math.fabs((ew * math.cos(t)))

Julia

function code(eh, ew, t)
	return abs(Float64(ew * cos(t)))
end

MATLAB

function tmp = code(eh, ew, t)
	tmp = abs((ew * cos(t)));
end

Wolfram

code[eh_, ew_, t_] := N[Abs[N[(ew * N[Cos[t], $MachinePrecision]), $MachinePrecision]], $MachinePrecision]

Derivation

1. Initial program 99.8%

   \[\left|\left(ew \cdot \cos t\right) \cdot \cos \tan^{-1} \left(\frac{\left(-eh\right) \cdot \tan t}{ew}\right) - \left(eh \cdot \sin t\right) \cdot \sin \tan^{-1} \left(\frac{\left(-eh\right) \cdot \tan t}{ew}\right)\right| \]

3. Applied rewrites 99.8%

   \[\leadsto \left|\color{blue}{\mathsf{fma}\left(\frac{ew}{\cosh \sinh^{-1} \left(\frac{\tan t}{ew} \cdot eh\right)}, \cos t, \tanh \sinh^{-1} \left(\frac{\tan t}{ew} \cdot eh\right) \cdot \left(\sin t \cdot eh\right)\right)}\right| \]

4. Taylor expanded in eh around 0

   \[\leadsto \left|\color{blue}{ew \cdot \cos t}\right| \]
5. Step-by-step derivation

   1. lower-*.f64 N/A

      \[\leadsto \left|ew \cdot \color{blue}{\cos t}\right| \]

   2. lower-cos.f64 61.7

      \[\leadsto \left|ew \cdot \cos t\right| \]

6. Applied rewrites 61.7%

   \[\leadsto \left|\color{blue}{ew \cdot \cos t}\right| \]
7. Add Preprocessing

Alternative 8: 38.9% accurate, 11.0× speedup

Math

\[\begin{array}{l} \\ \begin{array}{l} \mathbf{if}\;t \leq 380000000:\\ \;\;\;\;\left|\mathsf{fma}\left(t \cdot t, -0.5, 1\right) \cdot ew\right|\\ \mathbf{else}:\\ \;\;\;\;\sqrt{{ew}^{2}}\\ \end{array} \end{array} \]

FPCore

(FPCore (eh ew t)
 :precision binary64
 (if (<= t 380000000.0)
   (fabs (* (fma (* t t) -0.5 1.0) ew))
   (sqrt (pow ew 2.0))))

C

double code(double eh, double ew, double t) {
	double tmp;
	if (t <= 380000000.0) {
		tmp = fabs((fma((t * t), -0.5, 1.0) * ew));
	} else {
		tmp = sqrt(pow(ew, 2.0));
	}
	return tmp;
}

Julia

function code(eh, ew, t)
	tmp = 0.0
	if (t <= 380000000.0)
		tmp = abs(Float64(fma(Float64(t * t), -0.5, 1.0) * ew));
	else
		tmp = sqrt((ew ^ 2.0));
	end
	return tmp
end

Wolfram

code[eh_, ew_, t_] := If[LessEqual[t, 380000000.0], N[Abs[N[(N[(N[(t * t), $MachinePrecision] * -0.5 + 1.0), $MachinePrecision] * ew), $MachinePrecision]], $MachinePrecision], N[Sqrt[N[Power[ew, 2.0], $MachinePrecision]], $MachinePrecision]]

Derivation

1. Split input into 2 regimes

2. if t < 3.8e8

   1. Initial program 99.8%

      \[\left|\left(ew \cdot \cos t\right) \cdot \cos \tan^{-1} \left(\frac{\left(-eh\right) \cdot \tan t}{ew}\right) - \left(eh \cdot \sin t\right) \cdot \sin \tan^{-1} \left(\frac{\left(-eh\right) \cdot \tan t}{ew}\right)\right| \]

   3. Applied rewrites 99.8%

      \[\leadsto \left|\color{blue}{\mathsf{fma}\left(\frac{ew}{\cosh \sinh^{-1} \left(\frac{\tan t}{ew} \cdot eh\right)}, \cos t, \tanh \sinh^{-1} \left(\frac{\tan t}{ew} \cdot eh\right) \cdot \left(\sin t \cdot eh\right)\right)}\right| \]

   4. Taylor expanded in eh around 0

      \[\leadsto \left|\color{blue}{ew \cdot \cos t}\right| \]
   5. Step-by-step derivation

      1. lower-*.f64 N/A

         \[\leadsto \left|ew \cdot \color{blue}{\cos t}\right| \]

      2. lower-cos.f64 61.7

         \[\leadsto \left|ew \cdot \cos t\right| \]

   6. Applied rewrites 61.7%

      \[\leadsto \left|\color{blue}{ew \cdot \cos t}\right| \]

   7. Taylor expanded in t around 0

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

      1. lower-+.f64 N/A

         \[\leadsto \left|ew \cdot \left(1 + \frac{-1}{2} \cdot \color{blue}{{t}^{2}}\right)\right| \]

      2. lower-*.f64 N/A

         \[\leadsto \left|ew \cdot \left(1 + \frac{-1}{2} \cdot {t}^{\color{blue}{2}}\right)\right| \]

      3. lower-pow.f64 38.1

         \[\leadsto \left|ew \cdot \left(1 + -0.5 \cdot {t}^{2}\right)\right| \]

   9. Applied rewrites 38.1%

      \[\leadsto \left|ew \cdot \left(1 + \color{blue}{-0.5 \cdot {t}^{2}}\right)\right| \]
   10. Step-by-step derivation

       1. lift-*.f64 N/A

          \[\leadsto \left|ew \cdot \color{blue}{\left(1 + \frac{-1}{2} \cdot {t}^{2}\right)}\right| \]

       2. *-commutative N/A

          \[\leadsto \left|\left(1 + \frac{-1}{2} \cdot {t}^{2}\right) \cdot \color{blue}{ew}\right| \]

       3. lower-*.f64 38.1

          \[\leadsto \left|\left(1 + -0.5 \cdot {t}^{2}\right) \cdot \color{blue}{ew}\right| \]

       4. lift-+.f64 N/A

          \[\leadsto \left|\left(1 + \frac{-1}{2} \cdot {t}^{2}\right) \cdot ew\right| \]

       5. +-commutative N/A

          \[\leadsto \left|\left(\frac{-1}{2} \cdot {t}^{2} + 1\right) \cdot ew\right| \]

       6. lift-*.f64 N/A

          \[\leadsto \left|\left(\frac{-1}{2} \cdot {t}^{2} + 1\right) \cdot ew\right| \]

       7. *-commutative N/A

          \[\leadsto \left|\left({t}^{2} \cdot \frac{-1}{2} + 1\right) \cdot ew\right| \]

       8. lower-fma.f64 38.1

          \[\leadsto \left|\mathsf{fma}\left({t}^{2}, -0.5, 1\right) \cdot ew\right| \]

       9. lift-pow.f64 N/A

          \[\leadsto \left|\mathsf{fma}\left({t}^{2}, \frac{-1}{2}, 1\right) \cdot ew\right| \]

       10. unpow2 N/A

           \[\leadsto \left|\mathsf{fma}\left(t \cdot t, \frac{-1}{2}, 1\right) \cdot ew\right| \]

       11. lower-*.f64 38.1

           \[\leadsto \left|\mathsf{fma}\left(t \cdot t, -0.5, 1\right) \cdot ew\right| \]

   11. Applied rewrites 38.1%

       \[\leadsto \left|\color{blue}{\mathsf{fma}\left(t \cdot t, -0.5, 1\right) \cdot ew}\right| \]

   if 3.8e8 < t

   1. Initial program 99.8%

      \[\left|\left(ew \cdot \cos t\right) \cdot \cos \tan^{-1} \left(\frac{\left(-eh\right) \cdot \tan t}{ew}\right) - \left(eh \cdot \sin t\right) \cdot \sin \tan^{-1} \left(\frac{\left(-eh\right) \cdot \tan t}{ew}\right)\right| \]

   3. Applied rewrites 99.8%

      \[\leadsto \left|\color{blue}{\mathsf{fma}\left(\frac{ew}{\cosh \sinh^{-1} \left(\frac{\tan t}{ew} \cdot eh\right)}, \cos t, \tanh \sinh^{-1} \left(\frac{\tan t}{ew} \cdot eh\right) \cdot \left(\sin t \cdot eh\right)\right)}\right| \]

   4. Taylor expanded in eh around 0

      \[\leadsto \left|\color{blue}{ew \cdot \cos t}\right| \]
   5. Step-by-step derivation

      1. lower-*.f64 N/A

         \[\leadsto \left|ew \cdot \color{blue}{\cos t}\right| \]

      2. lower-cos.f64 61.7

         \[\leadsto \left|ew \cdot \cos t\right| \]

   6. Applied rewrites 61.7%

      \[\leadsto \left|\color{blue}{ew \cdot \cos t}\right| \]
   7. Step-by-step derivation

      1. lift-fabs.f64 N/A

         \[\leadsto \color{blue}{\left|ew \cdot \cos t\right|} \]

      2. rem-sqrt-square-rev N/A

         \[\leadsto \color{blue}{\sqrt{\left(ew \cdot \cos t\right) \cdot \left(ew \cdot \cos t\right)}} \]

      3. lower-sqrt.f64 N/A

         \[\leadsto \color{blue}{\sqrt{\left(ew \cdot \cos t\right) \cdot \left(ew \cdot \cos t\right)}} \]

      4. pow2 N/A

         \[\leadsto \sqrt{\color{blue}{{\left(ew \cdot \cos t\right)}^{2}}} \]

   8. Applied rewrites 34.2%

      \[\leadsto \color{blue}{\sqrt{{\left(\cos t \cdot ew\right)}^{2}}} \]

   9. Taylor expanded in t around 0

      \[\leadsto \sqrt{\color{blue}{{ew}^{2}}} \]
   10. Step-by-step derivation

       1. lower-pow.f64 24.5

          \[\leadsto \sqrt{{ew}^{\color{blue}{2}}} \]

   11. Applied rewrites 24.5%

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

Alternative 9: 38.1% accurate, 18.9× speedup

Math

\[\begin{array}{l} \\ \left|\mathsf{fma}\left(t \cdot t, -0.5, 1\right) \cdot ew\right| \end{array} \]

FPCore

(FPCore (eh ew t) :precision binary64 (fabs (* (fma (* t t) -0.5 1.0) ew)))

C

double code(double eh, double ew, double t) {
	return fabs((fma((t * t), -0.5, 1.0) * ew));
}

Julia

function code(eh, ew, t)
	return abs(Float64(fma(Float64(t * t), -0.5, 1.0) * ew))
end

Wolfram

code[eh_, ew_, t_] := N[Abs[N[(N[(N[(t * t), $MachinePrecision] * -0.5 + 1.0), $MachinePrecision] * ew), $MachinePrecision]], $MachinePrecision]

Derivation

1. Initial program 99.8%

   \[\left|\left(ew \cdot \cos t\right) \cdot \cos \tan^{-1} \left(\frac{\left(-eh\right) \cdot \tan t}{ew}\right) - \left(eh \cdot \sin t\right) \cdot \sin \tan^{-1} \left(\frac{\left(-eh\right) \cdot \tan t}{ew}\right)\right| \]

3. Applied rewrites 99.8%

   \[\leadsto \left|\color{blue}{\mathsf{fma}\left(\frac{ew}{\cosh \sinh^{-1} \left(\frac{\tan t}{ew} \cdot eh\right)}, \cos t, \tanh \sinh^{-1} \left(\frac{\tan t}{ew} \cdot eh\right) \cdot \left(\sin t \cdot eh\right)\right)}\right| \]

4. Taylor expanded in eh around 0

   \[\leadsto \left|\color{blue}{ew \cdot \cos t}\right| \]
5. Step-by-step derivation

   1. lower-*.f64 N/A

      \[\leadsto \left|ew \cdot \color{blue}{\cos t}\right| \]

   2. lower-cos.f64 61.7

      \[\leadsto \left|ew \cdot \cos t\right| \]

6. Applied rewrites 61.7%

   \[\leadsto \left|\color{blue}{ew \cdot \cos t}\right| \]

7. Taylor expanded in t around 0

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

   1. lower-+.f64 N/A

      \[\leadsto \left|ew \cdot \left(1 + \frac{-1}{2} \cdot \color{blue}{{t}^{2}}\right)\right| \]

   2. lower-*.f64 N/A

      \[\leadsto \left|ew \cdot \left(1 + \frac{-1}{2} \cdot {t}^{\color{blue}{2}}\right)\right| \]

   3. lower-pow.f64 38.1

      \[\leadsto \left|ew \cdot \left(1 + -0.5 \cdot {t}^{2}\right)\right| \]

9. Applied rewrites 38.1%

   \[\leadsto \left|ew \cdot \left(1 + \color{blue}{-0.5 \cdot {t}^{2}}\right)\right| \]
10. Step-by-step derivation

    1. lift-*.f64 N/A

       \[\leadsto \left|ew \cdot \color{blue}{\left(1 + \frac{-1}{2} \cdot {t}^{2}\right)}\right| \]

    2. *-commutative N/A

       \[\leadsto \left|\left(1 + \frac{-1}{2} \cdot {t}^{2}\right) \cdot \color{blue}{ew}\right| \]

    3. lower-*.f64 38.1

       \[\leadsto \left|\left(1 + -0.5 \cdot {t}^{2}\right) \cdot \color{blue}{ew}\right| \]

    4. lift-+.f64 N/A

       \[\leadsto \left|\left(1 + \frac{-1}{2} \cdot {t}^{2}\right) \cdot ew\right| \]

    5. +-commutative N/A

       \[\leadsto \left|\left(\frac{-1}{2} \cdot {t}^{2} + 1\right) \cdot ew\right| \]

    6. lift-*.f64 N/A

       \[\leadsto \left|\left(\frac{-1}{2} \cdot {t}^{2} + 1\right) \cdot ew\right| \]

    7. *-commutative N/A

       \[\leadsto \left|\left({t}^{2} \cdot \frac{-1}{2} + 1\right) \cdot ew\right| \]

    8. lower-fma.f64 38.1

       \[\leadsto \left|\mathsf{fma}\left({t}^{2}, -0.5, 1\right) \cdot ew\right| \]

    9. lift-pow.f64 N/A

       \[\leadsto \left|\mathsf{fma}\left({t}^{2}, \frac{-1}{2}, 1\right) \cdot ew\right| \]

    10. unpow2 N/A

        \[\leadsto \left|\mathsf{fma}\left(t \cdot t, \frac{-1}{2}, 1\right) \cdot ew\right| \]

    11. lower-*.f64 38.1

        \[\leadsto \left|\mathsf{fma}\left(t \cdot t, -0.5, 1\right) \cdot ew\right| \]

11. Applied rewrites 38.1%

    \[\leadsto \left|\color{blue}{\mathsf{fma}\left(t \cdot t, -0.5, 1\right) \cdot ew}\right| \]
12. Add Preprocessing

Reproduce

herbie shell --seed 2025156 
(FPCore (eh ew t)
  :name "Example 2 from Robby"
  :precision binary64
  (fabs (- (* (* ew (cos t)) (cos (atan (/ (* (- eh) (tan t)) ew)))) (* (* eh (sin t)) (sin (atan (/ (* (- eh) (tan t)) ew)))))))