Example from Robby

Percentage Accurate: 99.8% → 99.8%

Time: 11.0s

Alternatives: 12

Speedup: N/A×

Specification

Math

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

FPCore

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

C

double code(double eh, double ew, double t) {
	double t_1 = atan(((eh / ew) / tan(t)));
	return fabs((((ew * sin(t)) * cos(t_1)) + ((eh * cos(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 / ew) / tan(t)))
    code = abs((((ew * sin(t)) * cos(t_1)) + ((eh * cos(t)) * sin(t_1))))
end function

Java

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

Python

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

Julia

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

MATLAB

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

Wolfram

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

Initial Program: 99.8% accurate, 1.0× speedup

Math

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

FPCore

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

C

double code(double eh, double ew, double t) {
	double t_1 = atan(((eh / ew) / tan(t)));
	return fabs((((ew * sin(t)) * cos(t_1)) + ((eh * cos(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 / ew) / tan(t)))
    code = abs((((ew * sin(t)) * cos(t_1)) + ((eh * cos(t)) * sin(t_1))))
end function

Java

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

Python

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

Julia

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

MATLAB

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

Wolfram

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

Alternative 1: 99.8% accurate, 0.9× speedup

Math

\[\begin{array}{l} \\ \left|\left(\frac{1}{\sqrt{1 + {\left(\frac{\cos t \cdot eh}{\sin t \cdot ew}\right)}^{2}}} \cdot ew\right) \cdot \sin t + \left(eh \cdot \cos t\right) \cdot \sin \tan^{-1} \left(\frac{\frac{eh}{ew}}{\tan t}\right)\right| \end{array} \]

FPCore

(FPCore (eh ew t)
 :precision binary64
 (fabs
  (+
   (*
    (* (/ 1.0 (sqrt (+ 1.0 (pow (/ (* (cos t) eh) (* (sin t) ew)) 2.0)))) ew)
    (sin t))
   (* (* eh (cos t)) (sin (atan (/ (/ eh ew) (tan t))))))))

C

double code(double eh, double ew, double t) {
	return fabs(((((1.0 / sqrt((1.0 + pow(((cos(t) * eh) / (sin(t) * ew)), 2.0)))) * ew) * sin(t)) + ((eh * cos(t)) * sin(atan(((eh / ew) / tan(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(((((1.0d0 / sqrt((1.0d0 + (((cos(t) * eh) / (sin(t) * ew)) ** 2.0d0)))) * ew) * sin(t)) + ((eh * cos(t)) * sin(atan(((eh / ew) / tan(t)))))))
end function

Java

public static double code(double eh, double ew, double t) {
	return Math.abs(((((1.0 / Math.sqrt((1.0 + Math.pow(((Math.cos(t) * eh) / (Math.sin(t) * ew)), 2.0)))) * ew) * Math.sin(t)) + ((eh * Math.cos(t)) * Math.sin(Math.atan(((eh / ew) / Math.tan(t)))))));
}

Python

def code(eh, ew, t):
	return math.fabs(((((1.0 / math.sqrt((1.0 + math.pow(((math.cos(t) * eh) / (math.sin(t) * ew)), 2.0)))) * ew) * math.sin(t)) + ((eh * math.cos(t)) * math.sin(math.atan(((eh / ew) / math.tan(t)))))))

Julia

function code(eh, ew, t)
	return abs(Float64(Float64(Float64(Float64(1.0 / sqrt(Float64(1.0 + (Float64(Float64(cos(t) * eh) / Float64(sin(t) * ew)) ^ 2.0)))) * ew) * sin(t)) + Float64(Float64(eh * cos(t)) * sin(atan(Float64(Float64(eh / ew) / tan(t)))))))
end

MATLAB

function tmp = code(eh, ew, t)
	tmp = abs(((((1.0 / sqrt((1.0 + (((cos(t) * eh) / (sin(t) * ew)) ^ 2.0)))) * ew) * sin(t)) + ((eh * cos(t)) * sin(atan(((eh / ew) / tan(t)))))));
end

Wolfram

code[eh_, ew_, t_] := N[Abs[N[(N[(N[(N[(1.0 / N[Sqrt[N[(1.0 + N[Power[N[(N[(N[Cos[t], $MachinePrecision] * eh), $MachinePrecision] / N[(N[Sin[t], $MachinePrecision] * ew), $MachinePrecision]), $MachinePrecision], 2.0], $MachinePrecision]), $MachinePrecision]], $MachinePrecision]), $MachinePrecision] * ew), $MachinePrecision] * N[Sin[t], $MachinePrecision]), $MachinePrecision] + N[(N[(eh * N[Cos[t], $MachinePrecision]), $MachinePrecision] * N[Sin[N[ArcTan[N[(N[(eh / ew), $MachinePrecision] / N[Tan[t], $MachinePrecision]), $MachinePrecision]], $MachinePrecision]], $MachinePrecision]), $MachinePrecision]), $MachinePrecision]], $MachinePrecision]

Derivation

1. Initial program 99.8%

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

2. Taylor expanded in eh around 0

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

   1. associate-*r* N/A

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

   2. lower-*.f64 N/A

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

4. Applied rewrites 99.8%

   \[\leadsto \left|\color{blue}{\left(\frac{1}{\sqrt{1 + {\left(\frac{\cos t \cdot eh}{\sin t \cdot ew}\right)}^{2}}} \cdot ew\right) \cdot \sin t} + \left(eh \cdot \cos t\right) \cdot \sin \tan^{-1} \left(\frac{\frac{eh}{ew}}{\tan t}\right)\right| \]
5. Add Preprocessing

Alternative 2: 99.8% accurate, 1.2× speedup

Math

\[\begin{array}{l} \\ \begin{array}{l} t_1 := \frac{eh}{ew \cdot \tan t}\\ \left|\mathsf{fma}\left(\sin t \cdot ew, \frac{1}{\sqrt{1 + {t\_1}^{2}}}, \left(\cos t \cdot eh\right) \cdot \tanh \sinh^{-1} t\_1\right)\right| \end{array} \end{array} \]

FPCore

(FPCore (eh ew t)
 :precision binary64
 (let* ((t_1 (/ eh (* ew (tan t)))))
   (fabs
    (fma
     (* (sin t) ew)
     (/ 1.0 (sqrt (+ 1.0 (pow t_1 2.0))))
     (* (* (cos t) eh) (tanh (asinh t_1)))))))

C

double code(double eh, double ew, double t) {
	double t_1 = eh / (ew * tan(t));
	return fabs(fma((sin(t) * ew), (1.0 / sqrt((1.0 + pow(t_1, 2.0)))), ((cos(t) * eh) * tanh(asinh(t_1)))));
}

Julia

function code(eh, ew, t)
	t_1 = Float64(eh / Float64(ew * tan(t)))
	return abs(fma(Float64(sin(t) * ew), Float64(1.0 / sqrt(Float64(1.0 + (t_1 ^ 2.0)))), Float64(Float64(cos(t) * eh) * tanh(asinh(t_1)))))
end

Wolfram

code[eh_, ew_, t_] := Block[{t$95$1 = N[(eh / N[(ew * N[Tan[t], $MachinePrecision]), $MachinePrecision]), $MachinePrecision]}, N[Abs[N[(N[(N[Sin[t], $MachinePrecision] * ew), $MachinePrecision] * N[(1.0 / N[Sqrt[N[(1.0 + N[Power[t$95$1, 2.0], $MachinePrecision]), $MachinePrecision]], $MachinePrecision]), $MachinePrecision] + N[(N[(N[Cos[t], $MachinePrecision] * eh), $MachinePrecision] * N[Tanh[N[ArcSinh[t$95$1], $MachinePrecision]], $MachinePrecision]), $MachinePrecision]), $MachinePrecision]], $MachinePrecision]]

Derivation

1. Initial program 99.8%

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

3. Applied rewrites 99.8%

   \[\leadsto \left|\color{blue}{\mathsf{fma}\left(\sin t \cdot ew, \frac{1}{\sqrt{1 + {\left(\frac{eh}{ew \cdot \tan t}\right)}^{2}}}, \left(\cos t \cdot eh\right) \cdot \tanh \sinh^{-1} \left(\frac{eh}{ew \cdot \tan t}\right)\right)}\right| \]
4. Add Preprocessing

Alternative 3: 98.5% accurate, 1.8× speedup

Math

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

FPCore

(FPCore (eh ew t)
 :precision binary64
 (fabs
  (fma
   (* (sin t) ew)
   1.0
   (* (* (cos t) eh) (tanh (asinh (/ eh (* ew (tan t)))))))))

C

double code(double eh, double ew, double t) {
	return fabs(fma((sin(t) * ew), 1.0, ((cos(t) * eh) * tanh(asinh((eh / (ew * tan(t))))))));
}

Julia

function code(eh, ew, t)
	return abs(fma(Float64(sin(t) * ew), 1.0, Float64(Float64(cos(t) * eh) * tanh(asinh(Float64(eh / Float64(ew * tan(t))))))))
end

Wolfram

code[eh_, ew_, t_] := N[Abs[N[(N[(N[Sin[t], $MachinePrecision] * ew), $MachinePrecision] * 1.0 + N[(N[(N[Cos[t], $MachinePrecision] * eh), $MachinePrecision] * N[Tanh[N[ArcSinh[N[(eh / N[(ew * N[Tan[t], $MachinePrecision]), $MachinePrecision]), $MachinePrecision]], $MachinePrecision]], $MachinePrecision]), $MachinePrecision]), $MachinePrecision]], $MachinePrecision]

Derivation

1. Initial program 99.8%

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

3. Applied rewrites 99.8%

   \[\leadsto \left|\color{blue}{\mathsf{fma}\left(\sin t \cdot ew, \frac{1}{\sqrt{1 + {\left(\frac{eh}{ew \cdot \tan t}\right)}^{2}}}, \left(\cos t \cdot eh\right) \cdot \tanh \sinh^{-1} \left(\frac{eh}{ew \cdot \tan t}\right)\right)}\right| \]

4. Taylor expanded in eh around 0

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

6. Applied rewrites 98.5%

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

Alternative 4: 90.1% accurate, 1.9× speedup

Math

\[\begin{array}{l} \\ \begin{array}{l} t_1 := \frac{eh}{ew \cdot t}\\ \left|\mathsf{fma}\left(\sin t \cdot ew, \frac{1}{\sqrt{1 + {t\_1}^{2}}}, \left(\cos t \cdot eh\right) \cdot \tanh \sinh^{-1} t\_1\right)\right| \end{array} \end{array} \]

FPCore

(FPCore (eh ew t)
 :precision binary64
 (let* ((t_1 (/ eh (* ew t))))
   (fabs
    (fma
     (* (sin t) ew)
     (/ 1.0 (sqrt (+ 1.0 (pow t_1 2.0))))
     (* (* (cos t) eh) (tanh (asinh t_1)))))))

C

double code(double eh, double ew, double t) {
	double t_1 = eh / (ew * t);
	return fabs(fma((sin(t) * ew), (1.0 / sqrt((1.0 + pow(t_1, 2.0)))), ((cos(t) * eh) * tanh(asinh(t_1)))));
}

Julia

function code(eh, ew, t)
	t_1 = Float64(eh / Float64(ew * t))
	return abs(fma(Float64(sin(t) * ew), Float64(1.0 / sqrt(Float64(1.0 + (t_1 ^ 2.0)))), Float64(Float64(cos(t) * eh) * tanh(asinh(t_1)))))
end

Wolfram

code[eh_, ew_, t_] := Block[{t$95$1 = N[(eh / N[(ew * t), $MachinePrecision]), $MachinePrecision]}, N[Abs[N[(N[(N[Sin[t], $MachinePrecision] * ew), $MachinePrecision] * N[(1.0 / N[Sqrt[N[(1.0 + N[Power[t$95$1, 2.0], $MachinePrecision]), $MachinePrecision]], $MachinePrecision]), $MachinePrecision] + N[(N[(N[Cos[t], $MachinePrecision] * eh), $MachinePrecision] * N[Tanh[N[ArcSinh[t$95$1], $MachinePrecision]], $MachinePrecision]), $MachinePrecision]), $MachinePrecision]], $MachinePrecision]]

Derivation

1. Initial program 99.8%

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

3. Applied rewrites 99.8%

   \[\leadsto \left|\color{blue}{\mathsf{fma}\left(\sin t \cdot ew, \frac{1}{\sqrt{1 + {\left(\frac{eh}{ew \cdot \tan t}\right)}^{2}}}, \left(\cos t \cdot eh\right) \cdot \tanh \sinh^{-1} \left(\frac{eh}{ew \cdot \tan t}\right)\right)}\right| \]

4. Taylor expanded in t around 0

   \[\leadsto \left|\mathsf{fma}\left(\sin t \cdot ew, \frac{1}{\sqrt{1 + {\left(\frac{eh}{ew \cdot \color{blue}{t}}\right)}^{2}}}, \left(\cos t \cdot eh\right) \cdot \tanh \sinh^{-1} \left(\frac{eh}{ew \cdot \tan t}\right)\right)\right| \]
5. Step-by-step derivation

6. Applied rewrites 99.1%

   \[\leadsto \left|\mathsf{fma}\left(\sin t \cdot ew, \frac{1}{\sqrt{1 + {\left(\frac{eh}{ew \cdot \color{blue}{t}}\right)}^{2}}}, \left(\cos t \cdot eh\right) \cdot \tanh \sinh^{-1} \left(\frac{eh}{ew \cdot \tan t}\right)\right)\right| \]

7. Taylor expanded in t around 0

   \[\leadsto \left|\mathsf{fma}\left(\sin t \cdot ew, \frac{1}{\sqrt{1 + {\left(\frac{eh}{ew \cdot t}\right)}^{2}}}, \left(\cos t \cdot eh\right) \cdot \tanh \sinh^{-1} \left(\frac{eh}{ew \cdot \color{blue}{t}}\right)\right)\right| \]
8. Step-by-step derivation

9. Applied rewrites 90.1%

   \[\leadsto \left|\mathsf{fma}\left(\sin t \cdot ew, \frac{1}{\sqrt{1 + {\left(\frac{eh}{ew \cdot t}\right)}^{2}}}, \left(\cos t \cdot eh\right) \cdot \tanh \sinh^{-1} \left(\frac{eh}{ew \cdot \color{blue}{t}}\right)\right)\right| \]
10. Add Preprocessing

Alternative 5: 89.5% accurate, 2.4× speedup

Math

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

FPCore

(FPCore (eh ew t)
 :precision binary64
 (fabs
  (fma (* (sin t) ew) 1.0 (* (* (cos t) eh) (tanh (asinh (/ eh (* ew t))))))))

C

double code(double eh, double ew, double t) {
	return fabs(fma((sin(t) * ew), 1.0, ((cos(t) * eh) * tanh(asinh((eh / (ew * t)))))));
}

Julia

function code(eh, ew, t)
	return abs(fma(Float64(sin(t) * ew), 1.0, Float64(Float64(cos(t) * eh) * tanh(asinh(Float64(eh / Float64(ew * t)))))))
end

Wolfram

code[eh_, ew_, t_] := N[Abs[N[(N[(N[Sin[t], $MachinePrecision] * ew), $MachinePrecision] * 1.0 + N[(N[(N[Cos[t], $MachinePrecision] * eh), $MachinePrecision] * N[Tanh[N[ArcSinh[N[(eh / N[(ew * t), $MachinePrecision]), $MachinePrecision]], $MachinePrecision]], $MachinePrecision]), $MachinePrecision]), $MachinePrecision]], $MachinePrecision]

Derivation

1. Initial program 99.8%

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

3. Applied rewrites 99.8%

   \[\leadsto \left|\color{blue}{\mathsf{fma}\left(\sin t \cdot ew, \frac{1}{\sqrt{1 + {\left(\frac{eh}{ew \cdot \tan t}\right)}^{2}}}, \left(\cos t \cdot eh\right) \cdot \tanh \sinh^{-1} \left(\frac{eh}{ew \cdot \tan t}\right)\right)}\right| \]

4. Taylor expanded in t around 0

   \[\leadsto \left|\mathsf{fma}\left(\sin t \cdot ew, \frac{1}{\sqrt{1 + {\left(\frac{eh}{ew \cdot \color{blue}{t}}\right)}^{2}}}, \left(\cos t \cdot eh\right) \cdot \tanh \sinh^{-1} \left(\frac{eh}{ew \cdot \tan t}\right)\right)\right| \]
5. Step-by-step derivation

6. Applied rewrites 99.1%

   \[\leadsto \left|\mathsf{fma}\left(\sin t \cdot ew, \frac{1}{\sqrt{1 + {\left(\frac{eh}{ew \cdot \color{blue}{t}}\right)}^{2}}}, \left(\cos t \cdot eh\right) \cdot \tanh \sinh^{-1} \left(\frac{eh}{ew \cdot \tan t}\right)\right)\right| \]

7. Taylor expanded in t around 0

   \[\leadsto \left|\mathsf{fma}\left(\sin t \cdot ew, \frac{1}{\sqrt{1 + {\left(\frac{eh}{ew \cdot t}\right)}^{2}}}, \left(\cos t \cdot eh\right) \cdot \tanh \sinh^{-1} \left(\frac{eh}{ew \cdot \color{blue}{t}}\right)\right)\right| \]
8. Step-by-step derivation

9. Applied rewrites 90.1%

   \[\leadsto \left|\mathsf{fma}\left(\sin t \cdot ew, \frac{1}{\sqrt{1 + {\left(\frac{eh}{ew \cdot t}\right)}^{2}}}, \left(\cos t \cdot eh\right) \cdot \tanh \sinh^{-1} \left(\frac{eh}{ew \cdot \color{blue}{t}}\right)\right)\right| \]

10. Taylor expanded in eh around 0

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

12. Applied rewrites 89.5%

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

Alternative 6: 82.6% accurate, 2.3× speedup

Math

\[\begin{array}{l} \\ \begin{array}{l} t_1 := \frac{eh}{ew \cdot t}\\ t_2 := \tanh \sinh^{-1} t\_1\\ t_3 := \frac{1}{\sqrt{1 + {t\_1}^{2}}}\\ t_4 := \left|\mathsf{fma}\left(\sin t \cdot ew, t\_3, eh \cdot t\_2\right)\right|\\ \mathbf{if}\;ew \leq -1100000000:\\ \;\;\;\;t\_4\\ \mathbf{elif}\;ew \leq 6 \cdot 10^{-118}:\\ \;\;\;\;\left|\mathsf{fma}\left(ew \cdot t, t\_3, \left(\cos t \cdot eh\right) \cdot t\_2\right)\right|\\ \mathbf{else}:\\ \;\;\;\;t\_4\\ \end{array} \end{array} \]

FPCore

(FPCore (eh ew t)
 :precision binary64
 (let* ((t_1 (/ eh (* ew t)))
        (t_2 (tanh (asinh t_1)))
        (t_3 (/ 1.0 (sqrt (+ 1.0 (pow t_1 2.0)))))
        (t_4 (fabs (fma (* (sin t) ew) t_3 (* eh t_2)))))
   (if (<= ew -1100000000.0)
     t_4
     (if (<= ew 6e-118)
       (fabs (fma (* ew t) t_3 (* (* (cos t) eh) t_2)))
       t_4))))

C

double code(double eh, double ew, double t) {
	double t_1 = eh / (ew * t);
	double t_2 = tanh(asinh(t_1));
	double t_3 = 1.0 / sqrt((1.0 + pow(t_1, 2.0)));
	double t_4 = fabs(fma((sin(t) * ew), t_3, (eh * t_2)));
	double tmp;
	if (ew <= -1100000000.0) {
		tmp = t_4;
	} else if (ew <= 6e-118) {
		tmp = fabs(fma((ew * t), t_3, ((cos(t) * eh) * t_2)));
	} else {
		tmp = t_4;
	}
	return tmp;
}

Julia

function code(eh, ew, t)
	t_1 = Float64(eh / Float64(ew * t))
	t_2 = tanh(asinh(t_1))
	t_3 = Float64(1.0 / sqrt(Float64(1.0 + (t_1 ^ 2.0))))
	t_4 = abs(fma(Float64(sin(t) * ew), t_3, Float64(eh * t_2)))
	tmp = 0.0
	if (ew <= -1100000000.0)
		tmp = t_4;
	elseif (ew <= 6e-118)
		tmp = abs(fma(Float64(ew * t), t_3, Float64(Float64(cos(t) * eh) * t_2)));
	else
		tmp = t_4;
	end
	return tmp
end

Wolfram

code[eh_, ew_, t_] := Block[{t$95$1 = N[(eh / N[(ew * t), $MachinePrecision]), $MachinePrecision]}, Block[{t$95$2 = N[Tanh[N[ArcSinh[t$95$1], $MachinePrecision]], $MachinePrecision]}, Block[{t$95$3 = N[(1.0 / N[Sqrt[N[(1.0 + N[Power[t$95$1, 2.0], $MachinePrecision]), $MachinePrecision]], $MachinePrecision]), $MachinePrecision]}, Block[{t$95$4 = N[Abs[N[(N[(N[Sin[t], $MachinePrecision] * ew), $MachinePrecision] * t$95$3 + N[(eh * t$95$2), $MachinePrecision]), $MachinePrecision]], $MachinePrecision]}, If[LessEqual[ew, -1100000000.0], t$95$4, If[LessEqual[ew, 6e-118], N[Abs[N[(N[(ew * t), $MachinePrecision] * t$95$3 + N[(N[(N[Cos[t], $MachinePrecision] * eh), $MachinePrecision] * t$95$2), $MachinePrecision]), $MachinePrecision]], $MachinePrecision], t$95$4]]]]]]

Derivation

1. Split input into 2 regimes

2. if ew < -1.1e9 or 6.00000000000000035e-118 < ew

   1. Initial program 99.8%

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

   3. Applied rewrites 99.8%

      \[\leadsto \left|\color{blue}{\mathsf{fma}\left(\sin t \cdot ew, \frac{1}{\sqrt{1 + {\left(\frac{eh}{ew \cdot \tan t}\right)}^{2}}}, \left(\cos t \cdot eh\right) \cdot \tanh \sinh^{-1} \left(\frac{eh}{ew \cdot \tan t}\right)\right)}\right| \]

   4. Taylor expanded in t around 0

      \[\leadsto \left|\mathsf{fma}\left(\sin t \cdot ew, \frac{1}{\sqrt{1 + {\left(\frac{eh}{ew \cdot \color{blue}{t}}\right)}^{2}}}, \left(\cos t \cdot eh\right) \cdot \tanh \sinh^{-1} \left(\frac{eh}{ew \cdot \tan t}\right)\right)\right| \]
   5. Step-by-step derivation

   6. Applied rewrites 99.1%

      \[\leadsto \left|\mathsf{fma}\left(\sin t \cdot ew, \frac{1}{\sqrt{1 + {\left(\frac{eh}{ew \cdot \color{blue}{t}}\right)}^{2}}}, \left(\cos t \cdot eh\right) \cdot \tanh \sinh^{-1} \left(\frac{eh}{ew \cdot \tan t}\right)\right)\right| \]

   7. Taylor expanded in t around 0

      \[\leadsto \left|\mathsf{fma}\left(\sin t \cdot ew, \frac{1}{\sqrt{1 + {\left(\frac{eh}{ew \cdot t}\right)}^{2}}}, \left(\cos t \cdot eh\right) \cdot \tanh \sinh^{-1} \left(\frac{eh}{ew \cdot \color{blue}{t}}\right)\right)\right| \]
   8. Step-by-step derivation

   9. Applied rewrites 91.4%

      \[\leadsto \left|\mathsf{fma}\left(\sin t \cdot ew, \frac{1}{\sqrt{1 + {\left(\frac{eh}{ew \cdot t}\right)}^{2}}}, \left(\cos t \cdot eh\right) \cdot \tanh \sinh^{-1} \left(\frac{eh}{ew \cdot \color{blue}{t}}\right)\right)\right| \]

   10. Taylor expanded in t around 0

       \[\leadsto \left|\mathsf{fma}\left(\sin t \cdot ew, \frac{1}{\sqrt{1 + {\left(\frac{eh}{ew \cdot t}\right)}^{2}}}, \color{blue}{eh} \cdot \tanh \sinh^{-1} \left(\frac{eh}{ew \cdot t}\right)\right)\right| \]
   11. Step-by-step derivation

   12. Applied rewrites 86.1%

       \[\leadsto \left|\mathsf{fma}\left(\sin t \cdot ew, \frac{1}{\sqrt{1 + {\left(\frac{eh}{ew \cdot t}\right)}^{2}}}, \color{blue}{eh} \cdot \tanh \sinh^{-1} \left(\frac{eh}{ew \cdot t}\right)\right)\right| \]

   if -1.1e9 < ew < 6.00000000000000035e-118

   1. Initial program 99.8%

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

   3. Applied rewrites 99.8%

      \[\leadsto \left|\color{blue}{\mathsf{fma}\left(\sin t \cdot ew, \frac{1}{\sqrt{1 + {\left(\frac{eh}{ew \cdot \tan t}\right)}^{2}}}, \left(\cos t \cdot eh\right) \cdot \tanh \sinh^{-1} \left(\frac{eh}{ew \cdot \tan t}\right)\right)}\right| \]

   4. Taylor expanded in t around 0

      \[\leadsto \left|\mathsf{fma}\left(\sin t \cdot ew, \frac{1}{\sqrt{1 + {\left(\frac{eh}{ew \cdot \color{blue}{t}}\right)}^{2}}}, \left(\cos t \cdot eh\right) \cdot \tanh \sinh^{-1} \left(\frac{eh}{ew \cdot \tan t}\right)\right)\right| \]
   5. Step-by-step derivation

   6. Applied rewrites 99.1%

      \[\leadsto \left|\mathsf{fma}\left(\sin t \cdot ew, \frac{1}{\sqrt{1 + {\left(\frac{eh}{ew \cdot \color{blue}{t}}\right)}^{2}}}, \left(\cos t \cdot eh\right) \cdot \tanh \sinh^{-1} \left(\frac{eh}{ew \cdot \tan t}\right)\right)\right| \]

   7. Taylor expanded in t around 0

      \[\leadsto \left|\mathsf{fma}\left(\sin t \cdot ew, \frac{1}{\sqrt{1 + {\left(\frac{eh}{ew \cdot t}\right)}^{2}}}, \left(\cos t \cdot eh\right) \cdot \tanh \sinh^{-1} \left(\frac{eh}{ew \cdot \color{blue}{t}}\right)\right)\right| \]
   8. Step-by-step derivation

   9. Applied rewrites 88.0%

      \[\leadsto \left|\mathsf{fma}\left(\sin t \cdot ew, \frac{1}{\sqrt{1 + {\left(\frac{eh}{ew \cdot t}\right)}^{2}}}, \left(\cos t \cdot eh\right) \cdot \tanh \sinh^{-1} \left(\frac{eh}{ew \cdot \color{blue}{t}}\right)\right)\right| \]

   10. Taylor expanded in t around 0

       \[\leadsto \left|\mathsf{fma}\left(\color{blue}{ew \cdot t}, \frac{1}{\sqrt{1 + {\left(\frac{eh}{ew \cdot t}\right)}^{2}}}, \left(\cos t \cdot eh\right) \cdot \tanh \sinh^{-1} \left(\frac{eh}{ew \cdot t}\right)\right)\right| \]
   11. Step-by-step derivation

       1. lift-*.f64 77.5

          \[\leadsto \left|\mathsf{fma}\left(ew \cdot \color{blue}{t}, \frac{1}{\sqrt{1 + {\left(\frac{eh}{ew \cdot t}\right)}^{2}}}, \left(\cos t \cdot eh\right) \cdot \tanh \sinh^{-1} \left(\frac{eh}{ew \cdot t}\right)\right)\right| \]

   12. Applied rewrites 77.5%

       \[\leadsto \left|\mathsf{fma}\left(\color{blue}{ew \cdot t}, \frac{1}{\sqrt{1 + {\left(\frac{eh}{ew \cdot t}\right)}^{2}}}, \left(\cos t \cdot eh\right) \cdot \tanh \sinh^{-1} \left(\frac{eh}{ew \cdot t}\right)\right)\right| \]
3. Recombined 2 regimes into one program.
4. Add Preprocessing

Alternative 7: 78.3% accurate, 2.5× speedup

Math

\[\begin{array}{l} \\ \begin{array}{l} t_1 := \frac{eh}{ew \cdot t}\\ \left|\mathsf{fma}\left(\sin t \cdot ew, \frac{1}{\sqrt{1 + {t\_1}^{2}}}, eh \cdot \tanh \sinh^{-1} t\_1\right)\right| \end{array} \end{array} \]

FPCore

(FPCore (eh ew t)
 :precision binary64
 (let* ((t_1 (/ eh (* ew t))))
   (fabs
    (fma
     (* (sin t) ew)
     (/ 1.0 (sqrt (+ 1.0 (pow t_1 2.0))))
     (* eh (tanh (asinh t_1)))))))

C

double code(double eh, double ew, double t) {
	double t_1 = eh / (ew * t);
	return fabs(fma((sin(t) * ew), (1.0 / sqrt((1.0 + pow(t_1, 2.0)))), (eh * tanh(asinh(t_1)))));
}

Julia

function code(eh, ew, t)
	t_1 = Float64(eh / Float64(ew * t))
	return abs(fma(Float64(sin(t) * ew), Float64(1.0 / sqrt(Float64(1.0 + (t_1 ^ 2.0)))), Float64(eh * tanh(asinh(t_1)))))
end

Wolfram

code[eh_, ew_, t_] := Block[{t$95$1 = N[(eh / N[(ew * t), $MachinePrecision]), $MachinePrecision]}, N[Abs[N[(N[(N[Sin[t], $MachinePrecision] * ew), $MachinePrecision] * N[(1.0 / N[Sqrt[N[(1.0 + N[Power[t$95$1, 2.0], $MachinePrecision]), $MachinePrecision]], $MachinePrecision]), $MachinePrecision] + N[(eh * N[Tanh[N[ArcSinh[t$95$1], $MachinePrecision]], $MachinePrecision]), $MachinePrecision]), $MachinePrecision]], $MachinePrecision]]

Derivation

1. Initial program 99.8%

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

3. Applied rewrites 99.8%

   \[\leadsto \left|\color{blue}{\mathsf{fma}\left(\sin t \cdot ew, \frac{1}{\sqrt{1 + {\left(\frac{eh}{ew \cdot \tan t}\right)}^{2}}}, \left(\cos t \cdot eh\right) \cdot \tanh \sinh^{-1} \left(\frac{eh}{ew \cdot \tan t}\right)\right)}\right| \]

4. Taylor expanded in t around 0

   \[\leadsto \left|\mathsf{fma}\left(\sin t \cdot ew, \frac{1}{\sqrt{1 + {\left(\frac{eh}{ew \cdot \color{blue}{t}}\right)}^{2}}}, \left(\cos t \cdot eh\right) \cdot \tanh \sinh^{-1} \left(\frac{eh}{ew \cdot \tan t}\right)\right)\right| \]
5. Step-by-step derivation

6. Applied rewrites 99.1%

   \[\leadsto \left|\mathsf{fma}\left(\sin t \cdot ew, \frac{1}{\sqrt{1 + {\left(\frac{eh}{ew \cdot \color{blue}{t}}\right)}^{2}}}, \left(\cos t \cdot eh\right) \cdot \tanh \sinh^{-1} \left(\frac{eh}{ew \cdot \tan t}\right)\right)\right| \]

7. Taylor expanded in t around 0

   \[\leadsto \left|\mathsf{fma}\left(\sin t \cdot ew, \frac{1}{\sqrt{1 + {\left(\frac{eh}{ew \cdot t}\right)}^{2}}}, \left(\cos t \cdot eh\right) \cdot \tanh \sinh^{-1} \left(\frac{eh}{ew \cdot \color{blue}{t}}\right)\right)\right| \]
8. Step-by-step derivation

9. Applied rewrites 90.1%

   \[\leadsto \left|\mathsf{fma}\left(\sin t \cdot ew, \frac{1}{\sqrt{1 + {\left(\frac{eh}{ew \cdot t}\right)}^{2}}}, \left(\cos t \cdot eh\right) \cdot \tanh \sinh^{-1} \left(\frac{eh}{ew \cdot \color{blue}{t}}\right)\right)\right| \]

10. Taylor expanded in t around 0

    \[\leadsto \left|\mathsf{fma}\left(\sin t \cdot ew, \frac{1}{\sqrt{1 + {\left(\frac{eh}{ew \cdot t}\right)}^{2}}}, \color{blue}{eh} \cdot \tanh \sinh^{-1} \left(\frac{eh}{ew \cdot t}\right)\right)\right| \]
11. Step-by-step derivation

12. Applied rewrites 78.3%

    \[\leadsto \left|\mathsf{fma}\left(\sin t \cdot ew, \frac{1}{\sqrt{1 + {\left(\frac{eh}{ew \cdot t}\right)}^{2}}}, \color{blue}{eh} \cdot \tanh \sinh^{-1} \left(\frac{eh}{ew \cdot t}\right)\right)\right| \]
13. Add Preprocessing

Alternative 8: 61.4% accurate, 2.6× speedup

Math

\[\begin{array}{l} \\ \begin{array}{l} t_1 := ew \cdot \sin t\\ t_2 := \left|\tanh \left(\frac{eh \cdot \cos t}{t\_1}\right) \cdot eh\right|\\ \mathbf{if}\;eh \leq -7.5 \cdot 10^{-70}:\\ \;\;\;\;t\_2\\ \mathbf{elif}\;eh \leq 1.55 \cdot 10^{-30}:\\ \;\;\;\;\left|t\_1\right|\\ \mathbf{else}:\\ \;\;\;\;t\_2\\ \end{array} \end{array} \]

FPCore

(FPCore (eh ew t)
 :precision binary64
 (let* ((t_1 (* ew (sin t))) (t_2 (fabs (* (tanh (/ (* eh (cos t)) t_1)) eh))))
   (if (<= eh -7.5e-70) t_2 (if (<= eh 1.55e-30) (fabs t_1) t_2))))

C

double code(double eh, double ew, double t) {
	double t_1 = ew * sin(t);
	double t_2 = fabs((tanh(((eh * cos(t)) / t_1)) * eh));
	double tmp;
	if (eh <= -7.5e-70) {
		tmp = t_2;
	} else if (eh <= 1.55e-30) {
		tmp = fabs(t_1);
	} else {
		tmp = t_2;
	}
	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(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
    real(8) :: t_2
    real(8) :: tmp
    t_1 = ew * sin(t)
    t_2 = abs((tanh(((eh * cos(t)) / t_1)) * eh))
    if (eh <= (-7.5d-70)) then
        tmp = t_2
    else if (eh <= 1.55d-30) then
        tmp = abs(t_1)
    else
        tmp = t_2
    end if
    code = tmp
end function

Java

public static double code(double eh, double ew, double t) {
	double t_1 = ew * Math.sin(t);
	double t_2 = Math.abs((Math.tanh(((eh * Math.cos(t)) / t_1)) * eh));
	double tmp;
	if (eh <= -7.5e-70) {
		tmp = t_2;
	} else if (eh <= 1.55e-30) {
		tmp = Math.abs(t_1);
	} else {
		tmp = t_2;
	}
	return tmp;
}

Python

def code(eh, ew, t):
	t_1 = ew * math.sin(t)
	t_2 = math.fabs((math.tanh(((eh * math.cos(t)) / t_1)) * eh))
	tmp = 0
	if eh <= -7.5e-70:
		tmp = t_2
	elif eh <= 1.55e-30:
		tmp = math.fabs(t_1)
	else:
		tmp = t_2
	return tmp

Julia

function code(eh, ew, t)
	t_1 = Float64(ew * sin(t))
	t_2 = abs(Float64(tanh(Float64(Float64(eh * cos(t)) / t_1)) * eh))
	tmp = 0.0
	if (eh <= -7.5e-70)
		tmp = t_2;
	elseif (eh <= 1.55e-30)
		tmp = abs(t_1);
	else
		tmp = t_2;
	end
	return tmp
end

MATLAB

function tmp_2 = code(eh, ew, t)
	t_1 = ew * sin(t);
	t_2 = abs((tanh(((eh * cos(t)) / t_1)) * eh));
	tmp = 0.0;
	if (eh <= -7.5e-70)
		tmp = t_2;
	elseif (eh <= 1.55e-30)
		tmp = abs(t_1);
	else
		tmp = t_2;
	end
	tmp_2 = tmp;
end

Wolfram

code[eh_, ew_, t_] := Block[{t$95$1 = N[(ew * N[Sin[t], $MachinePrecision]), $MachinePrecision]}, Block[{t$95$2 = N[Abs[N[(N[Tanh[N[(N[(eh * N[Cos[t], $MachinePrecision]), $MachinePrecision] / t$95$1), $MachinePrecision]], $MachinePrecision] * eh), $MachinePrecision]], $MachinePrecision]}, If[LessEqual[eh, -7.5e-70], t$95$2, If[LessEqual[eh, 1.55e-30], N[Abs[t$95$1], $MachinePrecision], t$95$2]]]]

Derivation

1. Split input into 2 regimes

2. if eh < -7.49999999999999973e-70 or 1.54999999999999995e-30 < eh

   1. Initial program 99.8%

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

   2. Taylor expanded in t around 0

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

      1. *-commutative N/A

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

      2. lower-*.f64 N/A

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

   4. Applied rewrites 52.1%

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

   5. Taylor expanded in eh around 0

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

      1. lower-/.f64 N/A

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

      2. lift-cos.f64 N/A

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

      3. lift-*.f64 N/A

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

      4. lower-*.f64 N/A

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

      5. lift-sin.f64 52.1

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

   7. Applied rewrites 52.1%

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

   if -7.49999999999999973e-70 < eh < 1.54999999999999995e-30

   1. Initial program 99.8%

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

   3. Applied rewrites 99.8%

      \[\leadsto \left|\color{blue}{\mathsf{fma}\left(\sin t \cdot ew, \frac{1}{\sqrt{1 + {\left(\frac{eh}{ew \cdot \tan t}\right)}^{2}}}, \left(\cos t \cdot eh\right) \cdot \tanh \sinh^{-1} \left(\frac{eh}{ew \cdot \tan t}\right)\right)}\right| \]

   4. Taylor expanded in eh around 0

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

      1. lower-*.f64 N/A

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

      2. lift-sin.f64 68.1

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

   6. Applied rewrites 68.1%

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

Alternative 9: 58.9% accurate, 5.5× speedup

Math

\[\begin{array}{l} \\ \begin{array}{l} t_1 := \left|ew \cdot \sin t\right|\\ \mathbf{if}\;t \leq -3.3 \cdot 10^{-7}:\\ \;\;\;\;t\_1\\ \mathbf{elif}\;t \leq 8.5 \cdot 10^{-36}:\\ \;\;\;\;\left|\tanh \left(\frac{eh}{ew \cdot t}\right) \cdot eh\right|\\ \mathbf{else}:\\ \;\;\;\;t\_1\\ \end{array} \end{array} \]

FPCore

(FPCore (eh ew t)
 :precision binary64
 (let* ((t_1 (fabs (* ew (sin t)))))
   (if (<= t -3.3e-7)
     t_1
     (if (<= t 8.5e-36) (fabs (* (tanh (/ eh (* ew t))) eh)) t_1))))

C

double code(double eh, double ew, double t) {
	double t_1 = fabs((ew * sin(t)));
	double tmp;
	if (t <= -3.3e-7) {
		tmp = t_1;
	} else if (t <= 8.5e-36) {
		tmp = fabs((tanh((eh / (ew * t))) * eh));
	} else {
		tmp = t_1;
	}
	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(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
    real(8) :: tmp
    t_1 = abs((ew * sin(t)))
    if (t <= (-3.3d-7)) then
        tmp = t_1
    else if (t <= 8.5d-36) then
        tmp = abs((tanh((eh / (ew * t))) * eh))
    else
        tmp = t_1
    end if
    code = tmp
end function

Java

public static double code(double eh, double ew, double t) {
	double t_1 = Math.abs((ew * Math.sin(t)));
	double tmp;
	if (t <= -3.3e-7) {
		tmp = t_1;
	} else if (t <= 8.5e-36) {
		tmp = Math.abs((Math.tanh((eh / (ew * t))) * eh));
	} else {
		tmp = t_1;
	}
	return tmp;
}

Python

def code(eh, ew, t):
	t_1 = math.fabs((ew * math.sin(t)))
	tmp = 0
	if t <= -3.3e-7:
		tmp = t_1
	elif t <= 8.5e-36:
		tmp = math.fabs((math.tanh((eh / (ew * t))) * eh))
	else:
		tmp = t_1
	return tmp

Julia

function code(eh, ew, t)
	t_1 = abs(Float64(ew * sin(t)))
	tmp = 0.0
	if (t <= -3.3e-7)
		tmp = t_1;
	elseif (t <= 8.5e-36)
		tmp = abs(Float64(tanh(Float64(eh / Float64(ew * t))) * eh));
	else
		tmp = t_1;
	end
	return tmp
end

MATLAB

function tmp_2 = code(eh, ew, t)
	t_1 = abs((ew * sin(t)));
	tmp = 0.0;
	if (t <= -3.3e-7)
		tmp = t_1;
	elseif (t <= 8.5e-36)
		tmp = abs((tanh((eh / (ew * t))) * eh));
	else
		tmp = t_1;
	end
	tmp_2 = tmp;
end

Wolfram

code[eh_, ew_, t_] := Block[{t$95$1 = N[Abs[N[(ew * N[Sin[t], $MachinePrecision]), $MachinePrecision]], $MachinePrecision]}, If[LessEqual[t, -3.3e-7], t$95$1, If[LessEqual[t, 8.5e-36], N[Abs[N[(N[Tanh[N[(eh / N[(ew * t), $MachinePrecision]), $MachinePrecision]], $MachinePrecision] * eh), $MachinePrecision]], $MachinePrecision], t$95$1]]]

Derivation

1. Split input into 2 regimes

2. if t < -3.3000000000000002e-7 or 8.5000000000000007e-36 < t

   1. Initial program 99.6%

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

   3. Applied rewrites 99.6%

      \[\leadsto \left|\color{blue}{\mathsf{fma}\left(\sin t \cdot ew, \frac{1}{\sqrt{1 + {\left(\frac{eh}{ew \cdot \tan t}\right)}^{2}}}, \left(\cos t \cdot eh\right) \cdot \tanh \sinh^{-1} \left(\frac{eh}{ew \cdot \tan t}\right)\right)}\right| \]

   4. Taylor expanded in eh around 0

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

      1. lower-*.f64 N/A

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

      2. lift-sin.f64 51.7

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

   6. Applied rewrites 51.7%

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

   if -3.3000000000000002e-7 < t < 8.5000000000000007e-36

   1. Initial program 100.0%

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

   2. Taylor expanded in t around 0

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

      1. *-commutative N/A

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

      2. lower-*.f64 N/A

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

   4. Applied rewrites 72.6%

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

   5. Taylor expanded in eh around 0

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

      1. lower-/.f64 N/A

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

      2. lift-cos.f64 N/A

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

      3. lift-*.f64 N/A

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

      4. lower-*.f64 N/A

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

      5. lift-sin.f64 72.6

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

   7. Applied rewrites 72.6%

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

   8. Taylor expanded in t around 0

      \[\leadsto \left|\tanh \left(\frac{eh}{ew \cdot t}\right) \cdot eh\right| \]
   9. Step-by-step derivation

      1. lower-/.f64 N/A

         \[\leadsto \left|\tanh \left(\frac{eh}{ew \cdot t}\right) \cdot eh\right| \]

      2. lift-*.f64 72.6

         \[\leadsto \left|\tanh \left(\frac{eh}{ew \cdot t}\right) \cdot eh\right| \]

   10. Applied rewrites 72.6%

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

Alternative 10: 43.8% accurate, 5.2× speedup

Math

\[\begin{array}{l} \\ \begin{array}{l} t_1 := \frac{eh}{ew \cdot t}\\ \mathbf{if}\;eh \leq -3 \cdot 10^{-88}:\\ \;\;\;\;\left|\tanh t\_1 \cdot eh\right|\\ \mathbf{elif}\;eh \leq -5.2 \cdot 10^{-251}:\\ \;\;\;\;\left|ew \cdot t\right|\\ \mathbf{elif}\;eh \leq 9.5 \cdot 10^{-103}:\\ \;\;\;\;ew \cdot \sin t\\ \mathbf{else}:\\ \;\;\;\;\left|\tanh \sinh^{-1} t\_1 \cdot eh\right|\\ \end{array} \end{array} \]

FPCore

(FPCore (eh ew t)
 :precision binary64
 (let* ((t_1 (/ eh (* ew t))))
   (if (<= eh -3e-88)
     (fabs (* (tanh t_1) eh))
     (if (<= eh -5.2e-251)
       (fabs (* ew t))
       (if (<= eh 9.5e-103)
         (* ew (sin t))
         (fabs (* (tanh (asinh t_1)) eh)))))))

C

double code(double eh, double ew, double t) {
	double t_1 = eh / (ew * t);
	double tmp;
	if (eh <= -3e-88) {
		tmp = fabs((tanh(t_1) * eh));
	} else if (eh <= -5.2e-251) {
		tmp = fabs((ew * t));
	} else if (eh <= 9.5e-103) {
		tmp = ew * sin(t);
	} else {
		tmp = fabs((tanh(asinh(t_1)) * eh));
	}
	return tmp;
}

Python

def code(eh, ew, t):
	t_1 = eh / (ew * t)
	tmp = 0
	if eh <= -3e-88:
		tmp = math.fabs((math.tanh(t_1) * eh))
	elif eh <= -5.2e-251:
		tmp = math.fabs((ew * t))
	elif eh <= 9.5e-103:
		tmp = ew * math.sin(t)
	else:
		tmp = math.fabs((math.tanh(math.asinh(t_1)) * eh))
	return tmp

Julia

function code(eh, ew, t)
	t_1 = Float64(eh / Float64(ew * t))
	tmp = 0.0
	if (eh <= -3e-88)
		tmp = abs(Float64(tanh(t_1) * eh));
	elseif (eh <= -5.2e-251)
		tmp = abs(Float64(ew * t));
	elseif (eh <= 9.5e-103)
		tmp = Float64(ew * sin(t));
	else
		tmp = abs(Float64(tanh(asinh(t_1)) * eh));
	end
	return tmp
end

MATLAB

function tmp_2 = code(eh, ew, t)
	t_1 = eh / (ew * t);
	tmp = 0.0;
	if (eh <= -3e-88)
		tmp = abs((tanh(t_1) * eh));
	elseif (eh <= -5.2e-251)
		tmp = abs((ew * t));
	elseif (eh <= 9.5e-103)
		tmp = ew * sin(t);
	else
		tmp = abs((tanh(asinh(t_1)) * eh));
	end
	tmp_2 = tmp;
end

Wolfram

code[eh_, ew_, t_] := Block[{t$95$1 = N[(eh / N[(ew * t), $MachinePrecision]), $MachinePrecision]}, If[LessEqual[eh, -3e-88], N[Abs[N[(N[Tanh[t$95$1], $MachinePrecision] * eh), $MachinePrecision]], $MachinePrecision], If[LessEqual[eh, -5.2e-251], N[Abs[N[(ew * t), $MachinePrecision]], $MachinePrecision], If[LessEqual[eh, 9.5e-103], N[(ew * N[Sin[t], $MachinePrecision]), $MachinePrecision], N[Abs[N[(N[Tanh[N[ArcSinh[t$95$1], $MachinePrecision]], $MachinePrecision] * eh), $MachinePrecision]], $MachinePrecision]]]]]

Derivation

1. Split input into 4 regimes

2. if eh < -2.9999999999999999e-88

   1. Initial program 99.8%

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

   2. Taylor expanded in t around 0

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

      1. *-commutative N/A

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

      2. lower-*.f64 N/A

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

   4. Applied rewrites 50.9%

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

   5. Taylor expanded in eh around 0

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

      1. lower-/.f64 N/A

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

      2. lift-cos.f64 N/A

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

      3. lift-*.f64 N/A

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

      4. lower-*.f64 N/A

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

      5. lift-sin.f64 50.9

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

   7. Applied rewrites 50.9%

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

   8. Taylor expanded in t around 0

      \[\leadsto \left|\tanh \left(\frac{eh}{ew \cdot t}\right) \cdot eh\right| \]
   9. Step-by-step derivation

      1. lower-/.f64 N/A

         \[\leadsto \left|\tanh \left(\frac{eh}{ew \cdot t}\right) \cdot eh\right| \]

      2. lift-*.f64 48.5

         \[\leadsto \left|\tanh \left(\frac{eh}{ew \cdot t}\right) \cdot eh\right| \]

   10. Applied rewrites 48.5%

       \[\leadsto \left|\tanh \left(\frac{eh}{ew \cdot t}\right) \cdot eh\right| \]

   if -2.9999999999999999e-88 < eh < -5.1999999999999998e-251

   1. Initial program 99.8%

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

   2. Taylor expanded in eh around 0

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

      1. associate-*r* N/A

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

      2. lower-*.f64 N/A

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

   4. Applied rewrites 67.1%

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

   5. Taylor expanded in t around 0

      \[\leadsto \left|\left(\frac{ew \cdot t}{eh} \cdot ew\right) \cdot \sin t\right| \]
   6. Step-by-step derivation

      1. lower-/.f64 N/A

         \[\leadsto \left|\left(\frac{ew \cdot t}{eh} \cdot ew\right) \cdot \sin t\right| \]

      2. lower-*.f64 5.3

         \[\leadsto \left|\left(\frac{ew \cdot t}{eh} \cdot ew\right) \cdot \sin t\right| \]

   7. Applied rewrites 5.3%

      \[\leadsto \left|\left(\frac{ew \cdot t}{eh} \cdot ew\right) \cdot \sin t\right| \]

   8. Taylor expanded in t around 0

      \[\leadsto \left|\left(\frac{ew \cdot t}{eh} \cdot ew\right) \cdot t\right| \]
   9. Step-by-step derivation

   10. Applied rewrites 5.1%

       \[\leadsto \left|\left(\frac{ew \cdot t}{eh} \cdot ew\right) \cdot t\right| \]

   11. Taylor expanded in eh around 0

       \[\leadsto \left|ew \cdot t\right| \]
   12. Step-by-step derivation

   13. Applied rewrites 32.8%

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

   if -5.1999999999999998e-251 < eh < 9.50000000000000065e-103

   1. Initial program 99.8%

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

   3. Applied rewrites 49.5%

      \[\leadsto \color{blue}{\sqrt{\mathsf{fma}\left(\sin t \cdot ew, \frac{1}{\sqrt{1 + {\left(\frac{eh}{ew \cdot \tan t}\right)}^{2}}}, \left(\cos t \cdot eh\right) \cdot \tanh \sinh^{-1} \left(\frac{eh}{ew \cdot \tan t}\right)\right)} \cdot \sqrt{\mathsf{fma}\left(\sin t \cdot ew, \frac{1}{\sqrt{1 + {\left(\frac{eh}{ew \cdot \tan t}\right)}^{2}}}, \left(\cos t \cdot eh\right) \cdot \tanh \sinh^{-1} \left(\frac{eh}{ew \cdot \tan t}\right)\right)}} \]

   4. Taylor expanded in eh around 0

      \[\leadsto \sqrt{\color{blue}{ew \cdot \sin t}} \cdot \sqrt{\mathsf{fma}\left(\sin t \cdot ew, \frac{1}{\sqrt{1 + {\left(\frac{eh}{ew \cdot \tan t}\right)}^{2}}}, \left(\cos t \cdot eh\right) \cdot \tanh \sinh^{-1} \left(\frac{eh}{ew \cdot \tan t}\right)\right)} \]
   5. Step-by-step derivation

      1. lower-*.f64 N/A

         \[\leadsto \sqrt{ew \cdot \color{blue}{\sin t}} \cdot \sqrt{\mathsf{fma}\left(\sin t \cdot ew, \frac{1}{\sqrt{1 + {\left(\frac{eh}{ew \cdot \tan t}\right)}^{2}}}, \left(\cos t \cdot eh\right) \cdot \tanh \sinh^{-1} \left(\frac{eh}{ew \cdot \tan t}\right)\right)} \]

      2. lift-sin.f64 36.7

         \[\leadsto \sqrt{ew \cdot \sin t} \cdot \sqrt{\mathsf{fma}\left(\sin t \cdot ew, \frac{1}{\sqrt{1 + {\left(\frac{eh}{ew \cdot \tan t}\right)}^{2}}}, \left(\cos t \cdot eh\right) \cdot \tanh \sinh^{-1} \left(\frac{eh}{ew \cdot \tan t}\right)\right)} \]

   6. Applied rewrites 36.7%

      \[\leadsto \sqrt{\color{blue}{ew \cdot \sin t}} \cdot \sqrt{\mathsf{fma}\left(\sin t \cdot ew, \frac{1}{\sqrt{1 + {\left(\frac{eh}{ew \cdot \tan t}\right)}^{2}}}, \left(\cos t \cdot eh\right) \cdot \tanh \sinh^{-1} \left(\frac{eh}{ew \cdot \tan t}\right)\right)} \]

   7. Taylor expanded in eh around 0

      \[\leadsto \sqrt{ew \cdot \sin t} \cdot \sqrt{\color{blue}{ew \cdot \sin t}} \]
   8. Step-by-step derivation

      1. lower-*.f64 N/A

         \[\leadsto \sqrt{ew \cdot \sin t} \cdot \sqrt{ew \cdot \color{blue}{\sin t}} \]

      2. lift-sin.f64 36.9

         \[\leadsto \sqrt{ew \cdot \sin t} \cdot \sqrt{ew \cdot \sin t} \]

   9. Applied rewrites 36.9%

      \[\leadsto \sqrt{ew \cdot \sin t} \cdot \sqrt{\color{blue}{ew \cdot \sin t}} \]
   10. Step-by-step derivation

       1. lift-*.f64 N/A

          \[\leadsto \color{blue}{\sqrt{ew \cdot \sin t} \cdot \sqrt{ew \cdot \sin t}} \]

       2. lift-sqrt.f64 N/A

          \[\leadsto \color{blue}{\sqrt{ew \cdot \sin t}} \cdot \sqrt{ew \cdot \sin t} \]

       3. lift-sqrt.f64 N/A

          \[\leadsto \sqrt{ew \cdot \sin t} \cdot \color{blue}{\sqrt{ew \cdot \sin t}} \]

   11. Applied rewrites 38.2%

       \[\leadsto \color{blue}{ew \cdot \sin t} \]

   if 9.50000000000000065e-103 < eh

   1. Initial program 99.8%

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

   2. Taylor expanded in t around 0

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

      1. *-commutative N/A

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

      2. lower-*.f64 N/A

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

   4. Applied rewrites 49.6%

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

   5. Taylor expanded in t around 0

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

      1. lower-/.f64 N/A

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

      2. lower-*.f64 47.5

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

   7. Applied rewrites 47.5%

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

Alternative 11: 42.8% accurate, 9.1× speedup

Math

\[\begin{array}{l} \\ \begin{array}{l} t_1 := \left|\tanh \left(\frac{eh}{ew \cdot t}\right) \cdot eh\right|\\ \mathbf{if}\;eh \leq -3 \cdot 10^{-88}:\\ \;\;\;\;t\_1\\ \mathbf{elif}\;eh \leq 1.55 \cdot 10^{-116}:\\ \;\;\;\;\left|ew \cdot t\right|\\ \mathbf{else}:\\ \;\;\;\;t\_1\\ \end{array} \end{array} \]

FPCore

(FPCore (eh ew t)
 :precision binary64
 (let* ((t_1 (fabs (* (tanh (/ eh (* ew t))) eh))))
   (if (<= eh -3e-88) t_1 (if (<= eh 1.55e-116) (fabs (* ew t)) t_1))))

C

double code(double eh, double ew, double t) {
	double t_1 = fabs((tanh((eh / (ew * t))) * eh));
	double tmp;
	if (eh <= -3e-88) {
		tmp = t_1;
	} else if (eh <= 1.55e-116) {
		tmp = fabs((ew * t));
	} else {
		tmp = t_1;
	}
	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(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
    real(8) :: tmp
    t_1 = abs((tanh((eh / (ew * t))) * eh))
    if (eh <= (-3d-88)) then
        tmp = t_1
    else if (eh <= 1.55d-116) then
        tmp = abs((ew * t))
    else
        tmp = t_1
    end if
    code = tmp
end function

Java

public static double code(double eh, double ew, double t) {
	double t_1 = Math.abs((Math.tanh((eh / (ew * t))) * eh));
	double tmp;
	if (eh <= -3e-88) {
		tmp = t_1;
	} else if (eh <= 1.55e-116) {
		tmp = Math.abs((ew * t));
	} else {
		tmp = t_1;
	}
	return tmp;
}

Python

def code(eh, ew, t):
	t_1 = math.fabs((math.tanh((eh / (ew * t))) * eh))
	tmp = 0
	if eh <= -3e-88:
		tmp = t_1
	elif eh <= 1.55e-116:
		tmp = math.fabs((ew * t))
	else:
		tmp = t_1
	return tmp

Julia

function code(eh, ew, t)
	t_1 = abs(Float64(tanh(Float64(eh / Float64(ew * t))) * eh))
	tmp = 0.0
	if (eh <= -3e-88)
		tmp = t_1;
	elseif (eh <= 1.55e-116)
		tmp = abs(Float64(ew * t));
	else
		tmp = t_1;
	end
	return tmp
end

MATLAB

function tmp_2 = code(eh, ew, t)
	t_1 = abs((tanh((eh / (ew * t))) * eh));
	tmp = 0.0;
	if (eh <= -3e-88)
		tmp = t_1;
	elseif (eh <= 1.55e-116)
		tmp = abs((ew * t));
	else
		tmp = t_1;
	end
	tmp_2 = tmp;
end

Wolfram

code[eh_, ew_, t_] := Block[{t$95$1 = N[Abs[N[(N[Tanh[N[(eh / N[(ew * t), $MachinePrecision]), $MachinePrecision]], $MachinePrecision] * eh), $MachinePrecision]], $MachinePrecision]}, If[LessEqual[eh, -3e-88], t$95$1, If[LessEqual[eh, 1.55e-116], N[Abs[N[(ew * t), $MachinePrecision]], $MachinePrecision], t$95$1]]]

Derivation

1. Split input into 2 regimes

2. if eh < -2.9999999999999999e-88 or 1.55000000000000009e-116 < eh

   1. Initial program 99.8%

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

   2. Taylor expanded in t around 0

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

      1. *-commutative N/A

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

      2. lower-*.f64 N/A

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

   4. Applied rewrites 49.8%

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

   5. Taylor expanded in eh around 0

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

      1. lower-/.f64 N/A

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

      2. lift-cos.f64 N/A

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

      3. lift-*.f64 N/A

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

      4. lower-*.f64 N/A

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

      5. lift-sin.f64 49.8

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

   7. Applied rewrites 49.8%

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

   8. Taylor expanded in t around 0

      \[\leadsto \left|\tanh \left(\frac{eh}{ew \cdot t}\right) \cdot eh\right| \]
   9. Step-by-step derivation

      1. lower-/.f64 N/A

         \[\leadsto \left|\tanh \left(\frac{eh}{ew \cdot t}\right) \cdot eh\right| \]

      2. lift-*.f64 47.6

         \[\leadsto \left|\tanh \left(\frac{eh}{ew \cdot t}\right) \cdot eh\right| \]

   10. Applied rewrites 47.6%

       \[\leadsto \left|\tanh \left(\frac{eh}{ew \cdot t}\right) \cdot eh\right| \]

   if -2.9999999999999999e-88 < eh < 1.55000000000000009e-116

   1. Initial program 99.8%

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

   2. Taylor expanded in eh around 0

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

      1. associate-*r* N/A

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

      2. lower-*.f64 N/A

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

   4. Applied rewrites 71.8%

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

   5. Taylor expanded in t around 0

      \[\leadsto \left|\left(\frac{ew \cdot t}{eh} \cdot ew\right) \cdot \sin t\right| \]
   6. Step-by-step derivation

      1. lower-/.f64 N/A

         \[\leadsto \left|\left(\frac{ew \cdot t}{eh} \cdot ew\right) \cdot \sin t\right| \]

      2. lower-*.f64 5.3

         \[\leadsto \left|\left(\frac{ew \cdot t}{eh} \cdot ew\right) \cdot \sin t\right| \]

   7. Applied rewrites 5.3%

      \[\leadsto \left|\left(\frac{ew \cdot t}{eh} \cdot ew\right) \cdot \sin t\right| \]

   8. Taylor expanded in t around 0

      \[\leadsto \left|\left(\frac{ew \cdot t}{eh} \cdot ew\right) \cdot t\right| \]
   9. Step-by-step derivation

   10. Applied rewrites 5.1%

       \[\leadsto \left|\left(\frac{ew \cdot t}{eh} \cdot ew\right) \cdot t\right| \]

   11. Taylor expanded in eh around 0

       \[\leadsto \left|ew \cdot t\right| \]
   12. Step-by-step derivation

   13. Applied rewrites 33.2%

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

Alternative 12: 19.1% accurate, 47.8× speedup

Math

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

FPCore

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

C

double code(double eh, double ew, double t) {
	return fabs((ew * 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 * t))
end function

Java

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

Python

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

Julia

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

MATLAB

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

Wolfram

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

Derivation

1. Initial program 99.8%

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

2. Taylor expanded in eh around 0

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

   1. associate-*r* N/A

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

   2. lower-*.f64 N/A

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

4. Applied rewrites 41.5%

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

5. Taylor expanded in t around 0

   \[\leadsto \left|\left(\frac{ew \cdot t}{eh} \cdot ew\right) \cdot \sin t\right| \]
6. Step-by-step derivation

   1. lower-/.f64 N/A

      \[\leadsto \left|\left(\frac{ew \cdot t}{eh} \cdot ew\right) \cdot \sin t\right| \]

   2. lower-*.f64 4.6

      \[\leadsto \left|\left(\frac{ew \cdot t}{eh} \cdot ew\right) \cdot \sin t\right| \]

7. Applied rewrites 4.6%

   \[\leadsto \left|\left(\frac{ew \cdot t}{eh} \cdot ew\right) \cdot \sin t\right| \]

8. Taylor expanded in t around 0

   \[\leadsto \left|\left(\frac{ew \cdot t}{eh} \cdot ew\right) \cdot t\right| \]
9. Step-by-step derivation

10. Applied rewrites 4.6%

    \[\leadsto \left|\left(\frac{ew \cdot t}{eh} \cdot ew\right) \cdot t\right| \]

11. Taylor expanded in eh around 0

    \[\leadsto \left|ew \cdot t\right| \]
12. Step-by-step derivation

13. Applied rewrites 19.1%

    \[\leadsto \left|ew \cdot t\right| \]
14. Add Preprocessing

Reproduce

herbie shell --seed 2025116 
(FPCore (eh ew t)
  :name "Example from Robby"
  :precision binary64
  (fabs (+ (* (* ew (sin t)) (cos (atan (/ (/ eh ew) (tan t))))) (* (* eh (cos t)) (sin (atan (/ (/ eh ew) (tan t))))))))