math.exp on complex, real part

Percentage Accurate: 100.0% → 100.0%

Time: 3.3s

Alternatives: 13

Speedup: 1.0×

• 25.4% of points produce a very large (infinite) output. You may want to add a precondition.

Specification

Math

\[\begin{array}{l} \\ e^{re} \cdot \cos im \end{array} \]

FPCore

(FPCore (re im) :precision binary64 (* (exp re) (cos im)))

C

double code(double re, double im) {
	return exp(re) * cos(im);
}

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(re, im)
use fmin_fmax_functions
    real(8), intent (in) :: re
    real(8), intent (in) :: im
    code = exp(re) * cos(im)
end function

Java

public static double code(double re, double im) {
	return Math.exp(re) * Math.cos(im);
}

Python

def code(re, im):
	return math.exp(re) * math.cos(im)

Julia

function code(re, im)
	return Float64(exp(re) * cos(im))
end

MATLAB

function tmp = code(re, im)
	tmp = exp(re) * cos(im);
end

Wolfram

code[re_, im_] := N[(N[Exp[re], $MachinePrecision] * N[Cos[im], $MachinePrecision]), $MachinePrecision]

Initial Program: 100.0% accurate, 1.0× speedup

Math

\[\begin{array}{l} \\ e^{re} \cdot \cos im \end{array} \]

FPCore

(FPCore (re im) :precision binary64 (* (exp re) (cos im)))

C

double code(double re, double im) {
	return exp(re) * cos(im);
}

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(re, im)
use fmin_fmax_functions
    real(8), intent (in) :: re
    real(8), intent (in) :: im
    code = exp(re) * cos(im)
end function

Java

public static double code(double re, double im) {
	return Math.exp(re) * Math.cos(im);
}

Python

def code(re, im):
	return math.exp(re) * math.cos(im)

Julia

function code(re, im)
	return Float64(exp(re) * cos(im))
end

MATLAB

function tmp = code(re, im)
	tmp = exp(re) * cos(im);
end

Wolfram

code[re_, im_] := N[(N[Exp[re], $MachinePrecision] * N[Cos[im], $MachinePrecision]), $MachinePrecision]

Alternative 1: 100.0% accurate, 1.0× speedup

Math

\[\begin{array}{l} \\ e^{re} \cdot \cos im \end{array} \]

FPCore

(FPCore (re im) :precision binary64 (* (exp re) (cos im)))

C

double code(double re, double im) {
	return exp(re) * cos(im);
}

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(re, im)
use fmin_fmax_functions
    real(8), intent (in) :: re
    real(8), intent (in) :: im
    code = exp(re) * cos(im)
end function

Java

public static double code(double re, double im) {
	return Math.exp(re) * Math.cos(im);
}

Python

def code(re, im):
	return math.exp(re) * math.cos(im)

Julia

function code(re, im)
	return Float64(exp(re) * cos(im))
end

MATLAB

function tmp = code(re, im)
	tmp = exp(re) * cos(im);
end

Wolfram

code[re_, im_] := N[(N[Exp[re], $MachinePrecision] * N[Cos[im], $MachinePrecision]), $MachinePrecision]

Derivation

1. Initial program 100.0%

   \[e^{re} \cdot \cos im \]
2. Add Preprocessing

Alternative 2: 98.9% accurate, 0.2× speedup

Math

\[\begin{array}{l} \\ \begin{array}{l} t_0 := e^{re} \cdot \cos im\\ \mathbf{if}\;t\_0 \leq -\infty:\\ \;\;\;\;e^{re} \cdot \left(\left(im \cdot im\right) \cdot -0.5\right)\\ \mathbf{elif}\;t\_0 \leq -0.01:\\ \;\;\;\;\cos im\\ \mathbf{elif}\;t\_0 \leq 0:\\ \;\;\;\;e^{re}\\ \mathbf{elif}\;t\_0 \leq 0.9999:\\ \;\;\;\;\cos im \cdot \left(re - -1\right)\\ \mathbf{else}:\\ \;\;\;\;e^{re}\\ \end{array} \end{array} \]

FPCore

(FPCore (re im)
 :precision binary64
 (let* ((t_0 (* (exp re) (cos im))))
   (if (<= t_0 (- INFINITY))
     (* (exp re) (* (* im im) -0.5))
     (if (<= t_0 -0.01)
       (cos im)
       (if (<= t_0 0.0)
         (exp re)
         (if (<= t_0 0.9999) (* (cos im) (- re -1.0)) (exp re)))))))

C

double code(double re, double im) {
	double t_0 = exp(re) * cos(im);
	double tmp;
	if (t_0 <= -((double) INFINITY)) {
		tmp = exp(re) * ((im * im) * -0.5);
	} else if (t_0 <= -0.01) {
		tmp = cos(im);
	} else if (t_0 <= 0.0) {
		tmp = exp(re);
	} else if (t_0 <= 0.9999) {
		tmp = cos(im) * (re - -1.0);
	} else {
		tmp = exp(re);
	}
	return tmp;
}

Java

public static double code(double re, double im) {
	double t_0 = Math.exp(re) * Math.cos(im);
	double tmp;
	if (t_0 <= -Double.POSITIVE_INFINITY) {
		tmp = Math.exp(re) * ((im * im) * -0.5);
	} else if (t_0 <= -0.01) {
		tmp = Math.cos(im);
	} else if (t_0 <= 0.0) {
		tmp = Math.exp(re);
	} else if (t_0 <= 0.9999) {
		tmp = Math.cos(im) * (re - -1.0);
	} else {
		tmp = Math.exp(re);
	}
	return tmp;
}

Python

def code(re, im):
	t_0 = math.exp(re) * math.cos(im)
	tmp = 0
	if t_0 <= -math.inf:
		tmp = math.exp(re) * ((im * im) * -0.5)
	elif t_0 <= -0.01:
		tmp = math.cos(im)
	elif t_0 <= 0.0:
		tmp = math.exp(re)
	elif t_0 <= 0.9999:
		tmp = math.cos(im) * (re - -1.0)
	else:
		tmp = math.exp(re)
	return tmp

Julia

function code(re, im)
	t_0 = Float64(exp(re) * cos(im))
	tmp = 0.0
	if (t_0 <= Float64(-Inf))
		tmp = Float64(exp(re) * Float64(Float64(im * im) * -0.5));
	elseif (t_0 <= -0.01)
		tmp = cos(im);
	elseif (t_0 <= 0.0)
		tmp = exp(re);
	elseif (t_0 <= 0.9999)
		tmp = Float64(cos(im) * Float64(re - -1.0));
	else
		tmp = exp(re);
	end
	return tmp
end

MATLAB

function tmp_2 = code(re, im)
	t_0 = exp(re) * cos(im);
	tmp = 0.0;
	if (t_0 <= -Inf)
		tmp = exp(re) * ((im * im) * -0.5);
	elseif (t_0 <= -0.01)
		tmp = cos(im);
	elseif (t_0 <= 0.0)
		tmp = exp(re);
	elseif (t_0 <= 0.9999)
		tmp = cos(im) * (re - -1.0);
	else
		tmp = exp(re);
	end
	tmp_2 = tmp;
end

Wolfram

code[re_, im_] := Block[{t$95$0 = N[(N[Exp[re], $MachinePrecision] * N[Cos[im], $MachinePrecision]), $MachinePrecision]}, If[LessEqual[t$95$0, (-Infinity)], N[(N[Exp[re], $MachinePrecision] * N[(N[(im * im), $MachinePrecision] * -0.5), $MachinePrecision]), $MachinePrecision], If[LessEqual[t$95$0, -0.01], N[Cos[im], $MachinePrecision], If[LessEqual[t$95$0, 0.0], N[Exp[re], $MachinePrecision], If[LessEqual[t$95$0, 0.9999], N[(N[Cos[im], $MachinePrecision] * N[(re - -1.0), $MachinePrecision]), $MachinePrecision], N[Exp[re], $MachinePrecision]]]]]]

Derivation

1. Split input into 4 regimes

2. if (*.f64 (exp.f64 re) (cos.f64 im)) < -inf.0

   1. Initial program 100.0%

      \[e^{re} \cdot \cos im \]

   2. Taylor expanded in im around 0

      \[\leadsto e^{re} \cdot \color{blue}{\left(1 + \frac{-1}{2} \cdot {im}^{2}\right)} \]
   3. Step-by-step derivation

      1. +-commutative N/A

         \[\leadsto e^{re} \cdot \left(\frac{-1}{2} \cdot {im}^{2} + \color{blue}{1}\right) \]

      2. *-commutative N/A

         \[\leadsto e^{re} \cdot \left({im}^{2} \cdot \frac{-1}{2} + 1\right) \]

      3. lower-fma.f64 N/A

         \[\leadsto e^{re} \cdot \mathsf{fma}\left({im}^{2}, \color{blue}{\frac{-1}{2}}, 1\right) \]

      4. unpow2 N/A

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

      5. lower-*.f64 100.0

         \[\leadsto e^{re} \cdot \mathsf{fma}\left(im \cdot im, -0.5, 1\right) \]

   4. Applied rewrites 100.0%

      \[\leadsto e^{re} \cdot \color{blue}{\mathsf{fma}\left(im \cdot im, -0.5, 1\right)} \]

   5. Taylor expanded in im around inf

      \[\leadsto e^{re} \cdot \left(\frac{-1}{2} \cdot \color{blue}{{im}^{2}}\right) \]
   6. Step-by-step derivation

      1. *-commutative N/A

         \[\leadsto e^{re} \cdot \left({im}^{2} \cdot \frac{-1}{2}\right) \]

      2. lower-*.f64 N/A

         \[\leadsto e^{re} \cdot \left({im}^{2} \cdot \frac{-1}{2}\right) \]

      3. pow2 N/A

         \[\leadsto e^{re} \cdot \left(\left(im \cdot im\right) \cdot \frac{-1}{2}\right) \]

      4. lift-*.f64 100.0

         \[\leadsto e^{re} \cdot \left(\left(im \cdot im\right) \cdot -0.5\right) \]

   7. Applied rewrites 100.0%

      \[\leadsto e^{re} \cdot \left(\left(im \cdot im\right) \cdot \color{blue}{-0.5}\right) \]

   if -inf.0 < (*.f64 (exp.f64 re) (cos.f64 im)) < -0.0100000000000000002

   1. Initial program 100.0%

      \[e^{re} \cdot \cos im \]

   2. Taylor expanded in re around 0

      \[\leadsto \color{blue}{\cos im} \]
   3. Step-by-step derivation

      1. lift-cos.f64 97.6

         \[\leadsto \cos im \]

   4. Applied rewrites 97.6%

      \[\leadsto \color{blue}{\cos im} \]

   if -0.0100000000000000002 < (*.f64 (exp.f64 re) (cos.f64 im)) < 0.0 or 0.99990000000000001 < (*.f64 (exp.f64 re) (cos.f64 im))

   1. Initial program 100.0%

      \[e^{re} \cdot \cos im \]

   2. Taylor expanded in im around 0

      \[\leadsto \color{blue}{e^{re}} \]
   3. Step-by-step derivation

      1. lift-exp.f64 99.3

         \[\leadsto e^{re} \]

   4. Applied rewrites 99.3%

      \[\leadsto \color{blue}{e^{re}} \]

   if 0.0 < (*.f64 (exp.f64 re) (cos.f64 im)) < 0.99990000000000001

   1. Initial program 100.0%

      \[e^{re} \cdot \cos im \]

   2. Taylor expanded in re around 0

      \[\leadsto \color{blue}{\cos im + re \cdot \cos im} \]
   3. Step-by-step derivation

      1. distribute-rgt1-in N/A

         \[\leadsto \left(re + 1\right) \cdot \color{blue}{\cos im} \]

      2. +-commutative N/A

         \[\leadsto \left(1 + re\right) \cdot \cos \color{blue}{im} \]

      3. *-commutative N/A

         \[\leadsto \cos im \cdot \color{blue}{\left(1 + re\right)} \]

      4. lower-*.f64 N/A

         \[\leadsto \cos im \cdot \color{blue}{\left(1 + re\right)} \]

      5. lift-cos.f64 N/A

         \[\leadsto \cos im \cdot \left(\color{blue}{1} + re\right) \]

      6. +-commutative N/A

         \[\leadsto \cos im \cdot \left(re + \color{blue}{1}\right) \]

      7. add-flip N/A

         \[\leadsto \cos im \cdot \left(re - \color{blue}{\left(\mathsf{neg}\left(1\right)\right)}\right) \]

      8. lower--.f64 N/A

         \[\leadsto \cos im \cdot \left(re - \color{blue}{\left(\mathsf{neg}\left(1\right)\right)}\right) \]

      9. metadata-eval 97.3

         \[\leadsto \cos im \cdot \left(re - -1\right) \]

   4. Applied rewrites 97.3%

      \[\leadsto \color{blue}{\cos im \cdot \left(re - -1\right)} \]
3. Recombined 4 regimes into one program.
4. Add Preprocessing

Alternative 3: 98.7% accurate, 0.2× speedup

Math

\[\begin{array}{l} \\ \begin{array}{l} t_0 := e^{re} \cdot \cos im\\ \mathbf{if}\;t\_0 \leq -\infty:\\ \;\;\;\;e^{re} \cdot \left(\left(im \cdot im\right) \cdot -0.5\right)\\ \mathbf{elif}\;t\_0 \leq -0.01:\\ \;\;\;\;\cos im\\ \mathbf{elif}\;t\_0 \leq 0:\\ \;\;\;\;e^{re}\\ \mathbf{elif}\;t\_0 \leq 0.9999:\\ \;\;\;\;\cos im\\ \mathbf{else}:\\ \;\;\;\;e^{re}\\ \end{array} \end{array} \]

FPCore

(FPCore (re im)
 :precision binary64
 (let* ((t_0 (* (exp re) (cos im))))
   (if (<= t_0 (- INFINITY))
     (* (exp re) (* (* im im) -0.5))
     (if (<= t_0 -0.01)
       (cos im)
       (if (<= t_0 0.0) (exp re) (if (<= t_0 0.9999) (cos im) (exp re)))))))

C

double code(double re, double im) {
	double t_0 = exp(re) * cos(im);
	double tmp;
	if (t_0 <= -((double) INFINITY)) {
		tmp = exp(re) * ((im * im) * -0.5);
	} else if (t_0 <= -0.01) {
		tmp = cos(im);
	} else if (t_0 <= 0.0) {
		tmp = exp(re);
	} else if (t_0 <= 0.9999) {
		tmp = cos(im);
	} else {
		tmp = exp(re);
	}
	return tmp;
}

Java

public static double code(double re, double im) {
	double t_0 = Math.exp(re) * Math.cos(im);
	double tmp;
	if (t_0 <= -Double.POSITIVE_INFINITY) {
		tmp = Math.exp(re) * ((im * im) * -0.5);
	} else if (t_0 <= -0.01) {
		tmp = Math.cos(im);
	} else if (t_0 <= 0.0) {
		tmp = Math.exp(re);
	} else if (t_0 <= 0.9999) {
		tmp = Math.cos(im);
	} else {
		tmp = Math.exp(re);
	}
	return tmp;
}

Python

def code(re, im):
	t_0 = math.exp(re) * math.cos(im)
	tmp = 0
	if t_0 <= -math.inf:
		tmp = math.exp(re) * ((im * im) * -0.5)
	elif t_0 <= -0.01:
		tmp = math.cos(im)
	elif t_0 <= 0.0:
		tmp = math.exp(re)
	elif t_0 <= 0.9999:
		tmp = math.cos(im)
	else:
		tmp = math.exp(re)
	return tmp

Julia

function code(re, im)
	t_0 = Float64(exp(re) * cos(im))
	tmp = 0.0
	if (t_0 <= Float64(-Inf))
		tmp = Float64(exp(re) * Float64(Float64(im * im) * -0.5));
	elseif (t_0 <= -0.01)
		tmp = cos(im);
	elseif (t_0 <= 0.0)
		tmp = exp(re);
	elseif (t_0 <= 0.9999)
		tmp = cos(im);
	else
		tmp = exp(re);
	end
	return tmp
end

MATLAB

function tmp_2 = code(re, im)
	t_0 = exp(re) * cos(im);
	tmp = 0.0;
	if (t_0 <= -Inf)
		tmp = exp(re) * ((im * im) * -0.5);
	elseif (t_0 <= -0.01)
		tmp = cos(im);
	elseif (t_0 <= 0.0)
		tmp = exp(re);
	elseif (t_0 <= 0.9999)
		tmp = cos(im);
	else
		tmp = exp(re);
	end
	tmp_2 = tmp;
end

Wolfram

code[re_, im_] := Block[{t$95$0 = N[(N[Exp[re], $MachinePrecision] * N[Cos[im], $MachinePrecision]), $MachinePrecision]}, If[LessEqual[t$95$0, (-Infinity)], N[(N[Exp[re], $MachinePrecision] * N[(N[(im * im), $MachinePrecision] * -0.5), $MachinePrecision]), $MachinePrecision], If[LessEqual[t$95$0, -0.01], N[Cos[im], $MachinePrecision], If[LessEqual[t$95$0, 0.0], N[Exp[re], $MachinePrecision], If[LessEqual[t$95$0, 0.9999], N[Cos[im], $MachinePrecision], N[Exp[re], $MachinePrecision]]]]]]

Derivation

1. Split input into 3 regimes

2. if (*.f64 (exp.f64 re) (cos.f64 im)) < -inf.0

   1. Initial program 100.0%

      \[e^{re} \cdot \cos im \]

   2. Taylor expanded in im around 0

      \[\leadsto e^{re} \cdot \color{blue}{\left(1 + \frac{-1}{2} \cdot {im}^{2}\right)} \]
   3. Step-by-step derivation

      1. +-commutative N/A

         \[\leadsto e^{re} \cdot \left(\frac{-1}{2} \cdot {im}^{2} + \color{blue}{1}\right) \]

      2. *-commutative N/A

         \[\leadsto e^{re} \cdot \left({im}^{2} \cdot \frac{-1}{2} + 1\right) \]

      3. lower-fma.f64 N/A

         \[\leadsto e^{re} \cdot \mathsf{fma}\left({im}^{2}, \color{blue}{\frac{-1}{2}}, 1\right) \]

      4. unpow2 N/A

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

      5. lower-*.f64 100.0

         \[\leadsto e^{re} \cdot \mathsf{fma}\left(im \cdot im, -0.5, 1\right) \]

   4. Applied rewrites 100.0%

      \[\leadsto e^{re} \cdot \color{blue}{\mathsf{fma}\left(im \cdot im, -0.5, 1\right)} \]

   5. Taylor expanded in im around inf

      \[\leadsto e^{re} \cdot \left(\frac{-1}{2} \cdot \color{blue}{{im}^{2}}\right) \]
   6. Step-by-step derivation

      1. *-commutative N/A

         \[\leadsto e^{re} \cdot \left({im}^{2} \cdot \frac{-1}{2}\right) \]

      2. lower-*.f64 N/A

         \[\leadsto e^{re} \cdot \left({im}^{2} \cdot \frac{-1}{2}\right) \]

      3. pow2 N/A

         \[\leadsto e^{re} \cdot \left(\left(im \cdot im\right) \cdot \frac{-1}{2}\right) \]

      4. lift-*.f64 100.0

         \[\leadsto e^{re} \cdot \left(\left(im \cdot im\right) \cdot -0.5\right) \]

   7. Applied rewrites 100.0%

      \[\leadsto e^{re} \cdot \left(\left(im \cdot im\right) \cdot \color{blue}{-0.5}\right) \]

   if -inf.0 < (*.f64 (exp.f64 re) (cos.f64 im)) < -0.0100000000000000002 or 0.0 < (*.f64 (exp.f64 re) (cos.f64 im)) < 0.99990000000000001

   1. Initial program 100.0%

      \[e^{re} \cdot \cos im \]

   2. Taylor expanded in re around 0

      \[\leadsto \color{blue}{\cos im} \]
   3. Step-by-step derivation

      1. lift-cos.f64 96.7

         \[\leadsto \cos im \]

   4. Applied rewrites 96.7%

      \[\leadsto \color{blue}{\cos im} \]

   if -0.0100000000000000002 < (*.f64 (exp.f64 re) (cos.f64 im)) < 0.0 or 0.99990000000000001 < (*.f64 (exp.f64 re) (cos.f64 im))

   1. Initial program 100.0%

      \[e^{re} \cdot \cos im \]

   2. Taylor expanded in im around 0

      \[\leadsto \color{blue}{e^{re}} \]
   3. Step-by-step derivation

      1. lift-exp.f64 99.3

         \[\leadsto e^{re} \]

   4. Applied rewrites 99.3%

      \[\leadsto \color{blue}{e^{re}} \]
3. Recombined 3 regimes into one program.
4. Add Preprocessing

Alternative 4: 77.7% accurate, 0.7× speedup

Math

\[\begin{array}{l} \\ \begin{array}{l} \mathbf{if}\;e^{re} \cdot \cos im \leq -0.01:\\ \;\;\;\;e^{re} \cdot \left(\left(im \cdot im\right) \cdot -0.5\right)\\ \mathbf{else}:\\ \;\;\;\;e^{re}\\ \end{array} \end{array} \]

FPCore

(FPCore (re im)
 :precision binary64
 (if (<= (* (exp re) (cos im)) -0.01)
   (* (exp re) (* (* im im) -0.5))
   (exp re)))

C

double code(double re, double im) {
	double tmp;
	if ((exp(re) * cos(im)) <= -0.01) {
		tmp = exp(re) * ((im * im) * -0.5);
	} else {
		tmp = exp(re);
	}
	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(re, im)
use fmin_fmax_functions
    real(8), intent (in) :: re
    real(8), intent (in) :: im
    real(8) :: tmp
    if ((exp(re) * cos(im)) <= (-0.01d0)) then
        tmp = exp(re) * ((im * im) * (-0.5d0))
    else
        tmp = exp(re)
    end if
    code = tmp
end function

Java

public static double code(double re, double im) {
	double tmp;
	if ((Math.exp(re) * Math.cos(im)) <= -0.01) {
		tmp = Math.exp(re) * ((im * im) * -0.5);
	} else {
		tmp = Math.exp(re);
	}
	return tmp;
}

Python

def code(re, im):
	tmp = 0
	if (math.exp(re) * math.cos(im)) <= -0.01:
		tmp = math.exp(re) * ((im * im) * -0.5)
	else:
		tmp = math.exp(re)
	return tmp

Julia

function code(re, im)
	tmp = 0.0
	if (Float64(exp(re) * cos(im)) <= -0.01)
		tmp = Float64(exp(re) * Float64(Float64(im * im) * -0.5));
	else
		tmp = exp(re);
	end
	return tmp
end

MATLAB

function tmp_2 = code(re, im)
	tmp = 0.0;
	if ((exp(re) * cos(im)) <= -0.01)
		tmp = exp(re) * ((im * im) * -0.5);
	else
		tmp = exp(re);
	end
	tmp_2 = tmp;
end

Wolfram

code[re_, im_] := If[LessEqual[N[(N[Exp[re], $MachinePrecision] * N[Cos[im], $MachinePrecision]), $MachinePrecision], -0.01], N[(N[Exp[re], $MachinePrecision] * N[(N[(im * im), $MachinePrecision] * -0.5), $MachinePrecision]), $MachinePrecision], N[Exp[re], $MachinePrecision]]

Derivation

1. Split input into 2 regimes

2. if (*.f64 (exp.f64 re) (cos.f64 im)) < -0.0100000000000000002

   1. Initial program 100.0%

      \[e^{re} \cdot \cos im \]

   2. Taylor expanded in im around 0

      \[\leadsto e^{re} \cdot \color{blue}{\left(1 + \frac{-1}{2} \cdot {im}^{2}\right)} \]
   3. Step-by-step derivation

      1. +-commutative N/A

         \[\leadsto e^{re} \cdot \left(\frac{-1}{2} \cdot {im}^{2} + \color{blue}{1}\right) \]

      2. *-commutative N/A

         \[\leadsto e^{re} \cdot \left({im}^{2} \cdot \frac{-1}{2} + 1\right) \]

      3. lower-fma.f64 N/A

         \[\leadsto e^{re} \cdot \mathsf{fma}\left({im}^{2}, \color{blue}{\frac{-1}{2}}, 1\right) \]

      4. unpow2 N/A

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

      5. lower-*.f64 37.4

         \[\leadsto e^{re} \cdot \mathsf{fma}\left(im \cdot im, -0.5, 1\right) \]

   4. Applied rewrites 37.4%

      \[\leadsto e^{re} \cdot \color{blue}{\mathsf{fma}\left(im \cdot im, -0.5, 1\right)} \]

   5. Taylor expanded in im around inf

      \[\leadsto e^{re} \cdot \left(\frac{-1}{2} \cdot \color{blue}{{im}^{2}}\right) \]
   6. Step-by-step derivation

      1. *-commutative N/A

         \[\leadsto e^{re} \cdot \left({im}^{2} \cdot \frac{-1}{2}\right) \]

      2. lower-*.f64 N/A

         \[\leadsto e^{re} \cdot \left({im}^{2} \cdot \frac{-1}{2}\right) \]

      3. pow2 N/A

         \[\leadsto e^{re} \cdot \left(\left(im \cdot im\right) \cdot \frac{-1}{2}\right) \]

      4. lift-*.f64 37.4

         \[\leadsto e^{re} \cdot \left(\left(im \cdot im\right) \cdot -0.5\right) \]

   7. Applied rewrites 37.4%

      \[\leadsto e^{re} \cdot \left(\left(im \cdot im\right) \cdot \color{blue}{-0.5}\right) \]

   if -0.0100000000000000002 < (*.f64 (exp.f64 re) (cos.f64 im))

   1. Initial program 100.0%

      \[e^{re} \cdot \cos im \]

   2. Taylor expanded in im around 0

      \[\leadsto \color{blue}{e^{re}} \]
   3. Step-by-step derivation

      1. lift-exp.f64 87.2

         \[\leadsto e^{re} \]

   4. Applied rewrites 87.2%

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

Alternative 5: 76.1% accurate, 0.8× speedup

Math

\[\begin{array}{l} \\ \begin{array}{l} \mathbf{if}\;e^{re} \cdot \cos im \leq -0.01:\\ \;\;\;\;\left(re - -1\right) \cdot \left(\left(im \cdot im\right) \cdot -0.5\right)\\ \mathbf{else}:\\ \;\;\;\;e^{re}\\ \end{array} \end{array} \]

FPCore

(FPCore (re im)
 :precision binary64
 (if (<= (* (exp re) (cos im)) -0.01)
   (* (- re -1.0) (* (* im im) -0.5))
   (exp re)))

C

double code(double re, double im) {
	double tmp;
	if ((exp(re) * cos(im)) <= -0.01) {
		tmp = (re - -1.0) * ((im * im) * -0.5);
	} else {
		tmp = exp(re);
	}
	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(re, im)
use fmin_fmax_functions
    real(8), intent (in) :: re
    real(8), intent (in) :: im
    real(8) :: tmp
    if ((exp(re) * cos(im)) <= (-0.01d0)) then
        tmp = (re - (-1.0d0)) * ((im * im) * (-0.5d0))
    else
        tmp = exp(re)
    end if
    code = tmp
end function

Java

public static double code(double re, double im) {
	double tmp;
	if ((Math.exp(re) * Math.cos(im)) <= -0.01) {
		tmp = (re - -1.0) * ((im * im) * -0.5);
	} else {
		tmp = Math.exp(re);
	}
	return tmp;
}

Python

def code(re, im):
	tmp = 0
	if (math.exp(re) * math.cos(im)) <= -0.01:
		tmp = (re - -1.0) * ((im * im) * -0.5)
	else:
		tmp = math.exp(re)
	return tmp

Julia

function code(re, im)
	tmp = 0.0
	if (Float64(exp(re) * cos(im)) <= -0.01)
		tmp = Float64(Float64(re - -1.0) * Float64(Float64(im * im) * -0.5));
	else
		tmp = exp(re);
	end
	return tmp
end

MATLAB

function tmp_2 = code(re, im)
	tmp = 0.0;
	if ((exp(re) * cos(im)) <= -0.01)
		tmp = (re - -1.0) * ((im * im) * -0.5);
	else
		tmp = exp(re);
	end
	tmp_2 = tmp;
end

Wolfram

code[re_, im_] := If[LessEqual[N[(N[Exp[re], $MachinePrecision] * N[Cos[im], $MachinePrecision]), $MachinePrecision], -0.01], N[(N[(re - -1.0), $MachinePrecision] * N[(N[(im * im), $MachinePrecision] * -0.5), $MachinePrecision]), $MachinePrecision], N[Exp[re], $MachinePrecision]]

Derivation

1. Split input into 2 regimes

2. if (*.f64 (exp.f64 re) (cos.f64 im)) < -0.0100000000000000002

   1. Initial program 100.0%

      \[e^{re} \cdot \cos im \]

   2. Taylor expanded in im around 0

      \[\leadsto e^{re} \cdot \color{blue}{\left(1 + \frac{-1}{2} \cdot {im}^{2}\right)} \]
   3. Step-by-step derivation

      1. +-commutative N/A

         \[\leadsto e^{re} \cdot \left(\frac{-1}{2} \cdot {im}^{2} + \color{blue}{1}\right) \]

      2. *-commutative N/A

         \[\leadsto e^{re} \cdot \left({im}^{2} \cdot \frac{-1}{2} + 1\right) \]

      3. lower-fma.f64 N/A

         \[\leadsto e^{re} \cdot \mathsf{fma}\left({im}^{2}, \color{blue}{\frac{-1}{2}}, 1\right) \]

      4. unpow2 N/A

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

      5. lower-*.f64 37.4

         \[\leadsto e^{re} \cdot \mathsf{fma}\left(im \cdot im, -0.5, 1\right) \]

   4. Applied rewrites 37.4%

      \[\leadsto e^{re} \cdot \color{blue}{\mathsf{fma}\left(im \cdot im, -0.5, 1\right)} \]

   5. Taylor expanded in re around 0

      \[\leadsto \color{blue}{\left(1 + re\right)} \cdot \mathsf{fma}\left(im \cdot im, \frac{-1}{2}, 1\right) \]
   6. Step-by-step derivation

      1. +-commutative N/A

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

      2. metadata-eval N/A

         \[\leadsto \left(re + \left(\mathsf{neg}\left(-1\right)\right)\right) \cdot \mathsf{fma}\left(im \cdot im, \frac{-1}{2}, 1\right) \]

      3. sub-flip N/A

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

      4. lift--.f64 29.1

         \[\leadsto \left(re - \color{blue}{-1}\right) \cdot \mathsf{fma}\left(im \cdot im, -0.5, 1\right) \]

   7. Applied rewrites 29.1%

      \[\leadsto \color{blue}{\left(re - -1\right)} \cdot \mathsf{fma}\left(im \cdot im, -0.5, 1\right) \]
   8. Step-by-step derivation

      1. lift-*.f64 N/A

         \[\leadsto \left(re - -1\right) \cdot \mathsf{fma}\left(im \cdot im, \frac{-1}{2}, 1\right) \]

      2. pow2 N/A

         \[\leadsto \left(re - -1\right) \cdot \mathsf{fma}\left({im}^{2}, \frac{-1}{2}, 1\right) \]

      3. pow-to-exp N/A

         \[\leadsto \left(re - -1\right) \cdot \mathsf{fma}\left(e^{\log im \cdot 2}, \frac{-1}{2}, 1\right) \]

      4. lower-exp.f64 N/A

         \[\leadsto \left(re - -1\right) \cdot \mathsf{fma}\left(e^{\log im \cdot 2}, \frac{-1}{2}, 1\right) \]

      5. lower-*.f64 N/A

         \[\leadsto \left(re - -1\right) \cdot \mathsf{fma}\left(e^{\log im \cdot 2}, \frac{-1}{2}, 1\right) \]

      6. lower-log.f64 14.7

         \[\leadsto \left(re - -1\right) \cdot \mathsf{fma}\left(e^{\log im \cdot 2}, -0.5, 1\right) \]

   9. Applied rewrites 14.7%

      \[\leadsto \left(re - -1\right) \cdot \mathsf{fma}\left(e^{\log im \cdot 2}, -0.5, 1\right) \]

   10. Taylor expanded in im around inf

       \[\leadsto \left(re - -1\right) \cdot \left(\frac{-1}{2} \cdot \color{blue}{{im}^{2}}\right) \]
   11. Step-by-step derivation

       1. *-commutative N/A

          \[\leadsto \left(re - -1\right) \cdot \left({im}^{2} \cdot \frac{-1}{2}\right) \]

       2. lower-*.f64 N/A

          \[\leadsto \left(re - -1\right) \cdot \left({im}^{2} \cdot \frac{-1}{2}\right) \]

       3. pow2 N/A

          \[\leadsto \left(re - -1\right) \cdot \left(\left(im \cdot im\right) \cdot \frac{-1}{2}\right) \]

       4. lift-*.f64 29.1

          \[\leadsto \left(re - -1\right) \cdot \left(\left(im \cdot im\right) \cdot -0.5\right) \]

   12. Applied rewrites 29.1%

       \[\leadsto \left(re - -1\right) \cdot \left(\left(im \cdot im\right) \cdot \color{blue}{-0.5}\right) \]

   if -0.0100000000000000002 < (*.f64 (exp.f64 re) (cos.f64 im))

   1. Initial program 100.0%

      \[e^{re} \cdot \cos im \]

   2. Taylor expanded in im around 0

      \[\leadsto \color{blue}{e^{re}} \]
   3. Step-by-step derivation

      1. lift-exp.f64 87.2

         \[\leadsto e^{re} \]

   4. Applied rewrites 87.2%

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

Alternative 6: 74.6% accurate, 0.8× speedup

Math

\[\begin{array}{l} \\ \begin{array}{l} \mathbf{if}\;e^{re} \cdot \cos im \leq -0.01:\\ \;\;\;\;1 \cdot \mathsf{fma}\left(im \cdot im, -0.5, 1\right)\\ \mathbf{else}:\\ \;\;\;\;e^{re}\\ \end{array} \end{array} \]

FPCore

(FPCore (re im)
 :precision binary64
 (if (<= (* (exp re) (cos im)) -0.01)
   (* 1.0 (fma (* im im) -0.5 1.0))
   (exp re)))

C

double code(double re, double im) {
	double tmp;
	if ((exp(re) * cos(im)) <= -0.01) {
		tmp = 1.0 * fma((im * im), -0.5, 1.0);
	} else {
		tmp = exp(re);
	}
	return tmp;
}

Julia

function code(re, im)
	tmp = 0.0
	if (Float64(exp(re) * cos(im)) <= -0.01)
		tmp = Float64(1.0 * fma(Float64(im * im), -0.5, 1.0));
	else
		tmp = exp(re);
	end
	return tmp
end

Wolfram

code[re_, im_] := If[LessEqual[N[(N[Exp[re], $MachinePrecision] * N[Cos[im], $MachinePrecision]), $MachinePrecision], -0.01], N[(1.0 * N[(N[(im * im), $MachinePrecision] * -0.5 + 1.0), $MachinePrecision]), $MachinePrecision], N[Exp[re], $MachinePrecision]]

Derivation

1. Split input into 2 regimes

2. if (*.f64 (exp.f64 re) (cos.f64 im)) < -0.0100000000000000002

   1. Initial program 100.0%

      \[e^{re} \cdot \cos im \]

   2. Taylor expanded in im around 0

      \[\leadsto e^{re} \cdot \color{blue}{\left(1 + \frac{-1}{2} \cdot {im}^{2}\right)} \]
   3. Step-by-step derivation

      1. +-commutative N/A

         \[\leadsto e^{re} \cdot \left(\frac{-1}{2} \cdot {im}^{2} + \color{blue}{1}\right) \]

      2. *-commutative N/A

         \[\leadsto e^{re} \cdot \left({im}^{2} \cdot \frac{-1}{2} + 1\right) \]

      3. lower-fma.f64 N/A

         \[\leadsto e^{re} \cdot \mathsf{fma}\left({im}^{2}, \color{blue}{\frac{-1}{2}}, 1\right) \]

      4. unpow2 N/A

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

      5. lower-*.f64 37.4

         \[\leadsto e^{re} \cdot \mathsf{fma}\left(im \cdot im, -0.5, 1\right) \]

   4. Applied rewrites 37.4%

      \[\leadsto e^{re} \cdot \color{blue}{\mathsf{fma}\left(im \cdot im, -0.5, 1\right)} \]

   5. Taylor expanded in re around 0

      \[\leadsto \color{blue}{1} \cdot \mathsf{fma}\left(im \cdot im, \frac{-1}{2}, 1\right) \]
   6. Step-by-step derivation

   7. Applied rewrites 21.2%

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

   if -0.0100000000000000002 < (*.f64 (exp.f64 re) (cos.f64 im))

   1. Initial program 100.0%

      \[e^{re} \cdot \cos im \]

   2. Taylor expanded in im around 0

      \[\leadsto \color{blue}{e^{re}} \]
   3. Step-by-step derivation

      1. lift-exp.f64 87.2

         \[\leadsto e^{re} \]

   4. Applied rewrites 87.2%

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

Alternative 7: 70.8% accurate, 4.1× speedup

Math

\[\begin{array}{l} \\ e^{re} \end{array} \]

FPCore

(FPCore (re im) :precision binary64 (exp re))

C

double code(double re, double im) {
	return exp(re);
}

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(re, im)
use fmin_fmax_functions
    real(8), intent (in) :: re
    real(8), intent (in) :: im
    code = exp(re)
end function

Java

public static double code(double re, double im) {
	return Math.exp(re);
}

Python

def code(re, im):
	return math.exp(re)

Julia

function code(re, im)
	return exp(re)
end

MATLAB

function tmp = code(re, im)
	tmp = exp(re);
end

Wolfram

code[re_, im_] := N[Exp[re], $MachinePrecision]

Derivation

1. Initial program 100.0%

   \[e^{re} \cdot \cos im \]

2. Taylor expanded in im around 0

   \[\leadsto \color{blue}{e^{re}} \]
3. Step-by-step derivation

   1. lift-exp.f64 70.8

      \[\leadsto e^{re} \]

4. Applied rewrites 70.8%

   \[\leadsto \color{blue}{e^{re}} \]
5. Add Preprocessing

Alternative 8: 38.3% accurate, 0.8× speedup

Math

\[\begin{array}{l} \\ \begin{array}{l} \mathbf{if}\;e^{re} \cdot \cos im \leq 2:\\ \;\;\;\;1 \cdot \frac{-1}{re - 1}\\ \mathbf{else}:\\ \;\;\;\;\mathsf{fma}\left(re, 0.5 \cdot re, re\right)\\ \end{array} \end{array} \]

FPCore

(FPCore (re im)
 :precision binary64
 (if (<= (* (exp re) (cos im)) 2.0)
   (* 1.0 (/ -1.0 (- re 1.0)))
   (fma re (* 0.5 re) re)))

C

double code(double re, double im) {
	double tmp;
	if ((exp(re) * cos(im)) <= 2.0) {
		tmp = 1.0 * (-1.0 / (re - 1.0));
	} else {
		tmp = fma(re, (0.5 * re), re);
	}
	return tmp;
}

Julia

function code(re, im)
	tmp = 0.0
	if (Float64(exp(re) * cos(im)) <= 2.0)
		tmp = Float64(1.0 * Float64(-1.0 / Float64(re - 1.0)));
	else
		tmp = fma(re, Float64(0.5 * re), re);
	end
	return tmp
end

Wolfram

code[re_, im_] := If[LessEqual[N[(N[Exp[re], $MachinePrecision] * N[Cos[im], $MachinePrecision]), $MachinePrecision], 2.0], N[(1.0 * N[(-1.0 / N[(re - 1.0), $MachinePrecision]), $MachinePrecision]), $MachinePrecision], N[(re * N[(0.5 * re), $MachinePrecision] + re), $MachinePrecision]]

Derivation

1. Split input into 2 regimes

2. if (*.f64 (exp.f64 re) (cos.f64 im)) < 2

   1. Initial program 100.0%

      \[e^{re} \cdot \cos im \]

   2. Taylor expanded in re around 0

      \[\leadsto \color{blue}{\cos im + re \cdot \cos im} \]
   3. Step-by-step derivation

      1. distribute-rgt1-in N/A

         \[\leadsto \left(re + 1\right) \cdot \color{blue}{\cos im} \]

      2. +-commutative N/A

         \[\leadsto \left(1 + re\right) \cdot \cos \color{blue}{im} \]

      3. *-commutative N/A

         \[\leadsto \cos im \cdot \color{blue}{\left(1 + re\right)} \]

      4. lower-*.f64 N/A

         \[\leadsto \cos im \cdot \color{blue}{\left(1 + re\right)} \]

      5. lift-cos.f64 N/A

         \[\leadsto \cos im \cdot \left(\color{blue}{1} + re\right) \]

      6. +-commutative N/A

         \[\leadsto \cos im \cdot \left(re + \color{blue}{1}\right) \]

      7. add-flip N/A

         \[\leadsto \cos im \cdot \left(re - \color{blue}{\left(\mathsf{neg}\left(1\right)\right)}\right) \]

      8. lower--.f64 N/A

         \[\leadsto \cos im \cdot \left(re - \color{blue}{\left(\mathsf{neg}\left(1\right)\right)}\right) \]

      9. metadata-eval 61.7

         \[\leadsto \cos im \cdot \left(re - -1\right) \]

   4. Applied rewrites 61.7%

      \[\leadsto \color{blue}{\cos im \cdot \left(re - -1\right)} \]
   5. Step-by-step derivation

      1. lift--.f64 N/A

         \[\leadsto \cos im \cdot \left(re - \color{blue}{-1}\right) \]

      2. flip-- N/A

         \[\leadsto \cos im \cdot \frac{re \cdot re - -1 \cdot -1}{\color{blue}{re + -1}} \]

      3. lower-/.f64 N/A

         \[\leadsto \cos im \cdot \frac{re \cdot re - -1 \cdot -1}{\color{blue}{re + -1}} \]

      4. metadata-eval N/A

         \[\leadsto \cos im \cdot \frac{re \cdot re - 1}{re + -1} \]

      5. metadata-eval N/A

         \[\leadsto \cos im \cdot \frac{re \cdot re - \left(\mathsf{neg}\left(-1\right)\right)}{re + -1} \]

      6. add-flip-rev N/A

         \[\leadsto \cos im \cdot \frac{re \cdot re + -1}{\color{blue}{re} + -1} \]

      7. lower-fma.f64 N/A

         \[\leadsto \cos im \cdot \frac{\mathsf{fma}\left(re, re, -1\right)}{\color{blue}{re} + -1} \]

      8. add-flip N/A

         \[\leadsto \cos im \cdot \frac{\mathsf{fma}\left(re, re, -1\right)}{re - \color{blue}{\left(\mathsf{neg}\left(-1\right)\right)}} \]

      9. metadata-eval N/A

         \[\leadsto \cos im \cdot \frac{\mathsf{fma}\left(re, re, -1\right)}{re - 1} \]

      10. lower--.f64 65.4

          \[\leadsto \cos im \cdot \frac{\mathsf{fma}\left(re, re, -1\right)}{re - \color{blue}{1}} \]

   6. Applied rewrites 65.4%

      \[\leadsto \cos im \cdot \frac{\mathsf{fma}\left(re, re, -1\right)}{\color{blue}{re - 1}} \]

   7. Taylor expanded in re around 0

      \[\leadsto \cos im \cdot \frac{-1}{\color{blue}{re} - 1} \]
   8. Step-by-step derivation

   9. Applied rewrites 62.3%

      \[\leadsto \cos im \cdot \frac{-1}{\color{blue}{re} - 1} \]

   10. Taylor expanded in im around 0

       \[\leadsto 1 \cdot \frac{\color{blue}{-1}}{re - 1} \]
   11. Step-by-step derivation

   12. Applied rewrites 35.1%

       \[\leadsto 1 \cdot \frac{\color{blue}{-1}}{re - 1} \]

   if 2 < (*.f64 (exp.f64 re) (cos.f64 im))

   1. Initial program 100.0%

      \[e^{re} \cdot \cos im \]

   2. Taylor expanded in im around 0

      \[\leadsto \color{blue}{e^{re}} \]
   3. Step-by-step derivation

      1. lift-exp.f64 99.6

         \[\leadsto e^{re} \]

   4. Applied rewrites 99.6%

      \[\leadsto \color{blue}{e^{re}} \]

   5. Taylor expanded in re around 0

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

      1. +-commutative N/A

         \[\leadsto re \cdot \left(1 + \frac{1}{2} \cdot re\right) + 1 \]

      2. *-commutative N/A

         \[\leadsto \left(1 + \frac{1}{2} \cdot re\right) \cdot re + 1 \]

      3. lower-fma.f64 N/A

         \[\leadsto \mathsf{fma}\left(1 + \frac{1}{2} \cdot re, re, 1\right) \]

      4. +-commutative N/A

         \[\leadsto \mathsf{fma}\left(\frac{1}{2} \cdot re + 1, re, 1\right) \]

      5. lower-fma.f64 52.1

         \[\leadsto \mathsf{fma}\left(\mathsf{fma}\left(0.5, re, 1\right), re, 1\right) \]

   7. Applied rewrites 52.1%

      \[\leadsto \mathsf{fma}\left(\mathsf{fma}\left(0.5, re, 1\right), \color{blue}{re}, 1\right) \]

   8. Taylor expanded in re around inf

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

      1. distribute-lft-in N/A

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

      2. inv-pow N/A

         \[\leadsto {re}^{2} \cdot \frac{1}{2} + {re}^{2} \cdot {re}^{-1} \]

      3. pow-prod-up N/A

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

      4. metadata-eval N/A

         \[\leadsto {re}^{2} \cdot \frac{1}{2} + {re}^{1} \]

      5. unpow1 N/A

         \[\leadsto {re}^{2} \cdot \frac{1}{2} + re \]

      6. lower-fma.f64 N/A

         \[\leadsto \mathsf{fma}\left({re}^{2}, \frac{1}{2}, re\right) \]

      7. pow2 N/A

         \[\leadsto \mathsf{fma}\left(re \cdot re, \frac{1}{2}, re\right) \]

      8. lower-*.f64 52.1

         \[\leadsto \mathsf{fma}\left(re \cdot re, 0.5, re\right) \]

   10. Applied rewrites 52.1%

       \[\leadsto \mathsf{fma}\left(re \cdot re, 0.5, re\right) \]
   11. Step-by-step derivation

       1. lift-*.f64 N/A

          \[\leadsto \mathsf{fma}\left(re \cdot re, \frac{1}{2}, re\right) \]

       2. lift-fma.f64 N/A

          \[\leadsto \left(re \cdot re\right) \cdot \frac{1}{2} + re \]

       3. associate-*l* N/A

          \[\leadsto re \cdot \left(re \cdot \frac{1}{2}\right) + re \]

       4. *-commutative N/A

          \[\leadsto re \cdot \left(\frac{1}{2} \cdot re\right) + re \]

       5. lower-fma.f64 N/A

          \[\leadsto \mathsf{fma}\left(re, \frac{1}{2} \cdot re, re\right) \]

       6. lower-*.f64 52.1

          \[\leadsto \mathsf{fma}\left(re, 0.5 \cdot re, re\right) \]

   12. Applied rewrites 52.1%

       \[\leadsto \mathsf{fma}\left(re, 0.5 \cdot re, re\right) \]
3. Recombined 2 regimes into one program.
4. Add Preprocessing

Alternative 9: 37.4% accurate, 0.8× speedup

Math

\[\begin{array}{l} \\ \begin{array}{l} \mathbf{if}\;e^{re} \cdot \cos im \leq 2:\\ \;\;\;\;re - -1\\ \mathbf{else}:\\ \;\;\;\;\mathsf{fma}\left(re, 0.5 \cdot re, re\right)\\ \end{array} \end{array} \]

FPCore

(FPCore (re im)
 :precision binary64
 (if (<= (* (exp re) (cos im)) 2.0) (- re -1.0) (fma re (* 0.5 re) re)))

C

double code(double re, double im) {
	double tmp;
	if ((exp(re) * cos(im)) <= 2.0) {
		tmp = re - -1.0;
	} else {
		tmp = fma(re, (0.5 * re), re);
	}
	return tmp;
}

Julia

function code(re, im)
	tmp = 0.0
	if (Float64(exp(re) * cos(im)) <= 2.0)
		tmp = Float64(re - -1.0);
	else
		tmp = fma(re, Float64(0.5 * re), re);
	end
	return tmp
end

Wolfram

code[re_, im_] := If[LessEqual[N[(N[Exp[re], $MachinePrecision] * N[Cos[im], $MachinePrecision]), $MachinePrecision], 2.0], N[(re - -1.0), $MachinePrecision], N[(re * N[(0.5 * re), $MachinePrecision] + re), $MachinePrecision]]

Derivation

1. Split input into 2 regimes

2. if (*.f64 (exp.f64 re) (cos.f64 im)) < 2

   1. Initial program 100.0%

      \[e^{re} \cdot \cos im \]

   2. Taylor expanded in re around 0

      \[\leadsto \color{blue}{\cos im + re \cdot \cos im} \]
   3. Step-by-step derivation

      1. distribute-rgt1-in N/A

         \[\leadsto \left(re + 1\right) \cdot \color{blue}{\cos im} \]

      2. +-commutative N/A

         \[\leadsto \left(1 + re\right) \cdot \cos \color{blue}{im} \]

      3. *-commutative N/A

         \[\leadsto \cos im \cdot \color{blue}{\left(1 + re\right)} \]

      4. lower-*.f64 N/A

         \[\leadsto \cos im \cdot \color{blue}{\left(1 + re\right)} \]

      5. lift-cos.f64 N/A

         \[\leadsto \cos im \cdot \left(\color{blue}{1} + re\right) \]

      6. +-commutative N/A

         \[\leadsto \cos im \cdot \left(re + \color{blue}{1}\right) \]

      7. add-flip N/A

         \[\leadsto \cos im \cdot \left(re - \color{blue}{\left(\mathsf{neg}\left(1\right)\right)}\right) \]

      8. lower--.f64 N/A

         \[\leadsto \cos im \cdot \left(re - \color{blue}{\left(\mathsf{neg}\left(1\right)\right)}\right) \]

      9. metadata-eval 61.7

         \[\leadsto \cos im \cdot \left(re - -1\right) \]

   4. Applied rewrites 61.7%

      \[\leadsto \color{blue}{\cos im \cdot \left(re - -1\right)} \]

   5. Taylor expanded in im around 0

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

      1. +-commutative N/A

         \[\leadsto re + 1 \]

      2. metadata-eval N/A

         \[\leadsto re + \left(\mathsf{neg}\left(-1\right)\right) \]

      3. sub-flip N/A

         \[\leadsto re - -1 \]

      4. lift--.f64 33.9

         \[\leadsto re - -1 \]

   7. Applied rewrites 33.9%

      \[\leadsto re - \color{blue}{-1} \]

   if 2 < (*.f64 (exp.f64 re) (cos.f64 im))

   1. Initial program 100.0%

      \[e^{re} \cdot \cos im \]

   2. Taylor expanded in im around 0

      \[\leadsto \color{blue}{e^{re}} \]
   3. Step-by-step derivation

      1. lift-exp.f64 99.6

         \[\leadsto e^{re} \]

   4. Applied rewrites 99.6%

      \[\leadsto \color{blue}{e^{re}} \]

   5. Taylor expanded in re around 0

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

      1. +-commutative N/A

         \[\leadsto re \cdot \left(1 + \frac{1}{2} \cdot re\right) + 1 \]

      2. *-commutative N/A

         \[\leadsto \left(1 + \frac{1}{2} \cdot re\right) \cdot re + 1 \]

      3. lower-fma.f64 N/A

         \[\leadsto \mathsf{fma}\left(1 + \frac{1}{2} \cdot re, re, 1\right) \]

      4. +-commutative N/A

         \[\leadsto \mathsf{fma}\left(\frac{1}{2} \cdot re + 1, re, 1\right) \]

      5. lower-fma.f64 52.1

         \[\leadsto \mathsf{fma}\left(\mathsf{fma}\left(0.5, re, 1\right), re, 1\right) \]

   7. Applied rewrites 52.1%

      \[\leadsto \mathsf{fma}\left(\mathsf{fma}\left(0.5, re, 1\right), \color{blue}{re}, 1\right) \]

   8. Taylor expanded in re around inf

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

      1. distribute-lft-in N/A

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

      2. inv-pow N/A

         \[\leadsto {re}^{2} \cdot \frac{1}{2} + {re}^{2} \cdot {re}^{-1} \]

      3. pow-prod-up N/A

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

      4. metadata-eval N/A

         \[\leadsto {re}^{2} \cdot \frac{1}{2} + {re}^{1} \]

      5. unpow1 N/A

         \[\leadsto {re}^{2} \cdot \frac{1}{2} + re \]

      6. lower-fma.f64 N/A

         \[\leadsto \mathsf{fma}\left({re}^{2}, \frac{1}{2}, re\right) \]

      7. pow2 N/A

         \[\leadsto \mathsf{fma}\left(re \cdot re, \frac{1}{2}, re\right) \]

      8. lower-*.f64 52.1

         \[\leadsto \mathsf{fma}\left(re \cdot re, 0.5, re\right) \]

   10. Applied rewrites 52.1%

       \[\leadsto \mathsf{fma}\left(re \cdot re, 0.5, re\right) \]
   11. Step-by-step derivation

       1. lift-*.f64 N/A

          \[\leadsto \mathsf{fma}\left(re \cdot re, \frac{1}{2}, re\right) \]

       2. lift-fma.f64 N/A

          \[\leadsto \left(re \cdot re\right) \cdot \frac{1}{2} + re \]

       3. associate-*l* N/A

          \[\leadsto re \cdot \left(re \cdot \frac{1}{2}\right) + re \]

       4. *-commutative N/A

          \[\leadsto re \cdot \left(\frac{1}{2} \cdot re\right) + re \]

       5. lower-fma.f64 N/A

          \[\leadsto \mathsf{fma}\left(re, \frac{1}{2} \cdot re, re\right) \]

       6. lower-*.f64 52.1

          \[\leadsto \mathsf{fma}\left(re, 0.5 \cdot re, re\right) \]

   12. Applied rewrites 52.1%

       \[\leadsto \mathsf{fma}\left(re, 0.5 \cdot re, re\right) \]
3. Recombined 2 regimes into one program.
4. Add Preprocessing

Alternative 10: 37.4% accurate, 0.8× speedup

Math

\[\begin{array}{l} \\ \begin{array}{l} \mathbf{if}\;e^{re} \cdot \cos im \leq 2:\\ \;\;\;\;re - -1\\ \mathbf{else}:\\ \;\;\;\;\left(re \cdot re\right) \cdot 0.5\\ \end{array} \end{array} \]

FPCore

(FPCore (re im)
 :precision binary64
 (if (<= (* (exp re) (cos im)) 2.0) (- re -1.0) (* (* re re) 0.5)))

C

double code(double re, double im) {
	double tmp;
	if ((exp(re) * cos(im)) <= 2.0) {
		tmp = re - -1.0;
	} else {
		tmp = (re * re) * 0.5;
	}
	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(re, im)
use fmin_fmax_functions
    real(8), intent (in) :: re
    real(8), intent (in) :: im
    real(8) :: tmp
    if ((exp(re) * cos(im)) <= 2.0d0) then
        tmp = re - (-1.0d0)
    else
        tmp = (re * re) * 0.5d0
    end if
    code = tmp
end function

Java

public static double code(double re, double im) {
	double tmp;
	if ((Math.exp(re) * Math.cos(im)) <= 2.0) {
		tmp = re - -1.0;
	} else {
		tmp = (re * re) * 0.5;
	}
	return tmp;
}

Python

def code(re, im):
	tmp = 0
	if (math.exp(re) * math.cos(im)) <= 2.0:
		tmp = re - -1.0
	else:
		tmp = (re * re) * 0.5
	return tmp

Julia

function code(re, im)
	tmp = 0.0
	if (Float64(exp(re) * cos(im)) <= 2.0)
		tmp = Float64(re - -1.0);
	else
		tmp = Float64(Float64(re * re) * 0.5);
	end
	return tmp
end

MATLAB

function tmp_2 = code(re, im)
	tmp = 0.0;
	if ((exp(re) * cos(im)) <= 2.0)
		tmp = re - -1.0;
	else
		tmp = (re * re) * 0.5;
	end
	tmp_2 = tmp;
end

Wolfram

code[re_, im_] := If[LessEqual[N[(N[Exp[re], $MachinePrecision] * N[Cos[im], $MachinePrecision]), $MachinePrecision], 2.0], N[(re - -1.0), $MachinePrecision], N[(N[(re * re), $MachinePrecision] * 0.5), $MachinePrecision]]

Derivation

1. Split input into 2 regimes

2. if (*.f64 (exp.f64 re) (cos.f64 im)) < 2

   1. Initial program 100.0%

      \[e^{re} \cdot \cos im \]

   2. Taylor expanded in re around 0

      \[\leadsto \color{blue}{\cos im + re \cdot \cos im} \]
   3. Step-by-step derivation

      1. distribute-rgt1-in N/A

         \[\leadsto \left(re + 1\right) \cdot \color{blue}{\cos im} \]

      2. +-commutative N/A

         \[\leadsto \left(1 + re\right) \cdot \cos \color{blue}{im} \]

      3. *-commutative N/A

         \[\leadsto \cos im \cdot \color{blue}{\left(1 + re\right)} \]

      4. lower-*.f64 N/A

         \[\leadsto \cos im \cdot \color{blue}{\left(1 + re\right)} \]

      5. lift-cos.f64 N/A

         \[\leadsto \cos im \cdot \left(\color{blue}{1} + re\right) \]

      6. +-commutative N/A

         \[\leadsto \cos im \cdot \left(re + \color{blue}{1}\right) \]

      7. add-flip N/A

         \[\leadsto \cos im \cdot \left(re - \color{blue}{\left(\mathsf{neg}\left(1\right)\right)}\right) \]

      8. lower--.f64 N/A

         \[\leadsto \cos im \cdot \left(re - \color{blue}{\left(\mathsf{neg}\left(1\right)\right)}\right) \]

      9. metadata-eval 61.7

         \[\leadsto \cos im \cdot \left(re - -1\right) \]

   4. Applied rewrites 61.7%

      \[\leadsto \color{blue}{\cos im \cdot \left(re - -1\right)} \]

   5. Taylor expanded in im around 0

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

      1. +-commutative N/A

         \[\leadsto re + 1 \]

      2. metadata-eval N/A

         \[\leadsto re + \left(\mathsf{neg}\left(-1\right)\right) \]

      3. sub-flip N/A

         \[\leadsto re - -1 \]

      4. lift--.f64 33.9

         \[\leadsto re - -1 \]

   7. Applied rewrites 33.9%

      \[\leadsto re - \color{blue}{-1} \]

   if 2 < (*.f64 (exp.f64 re) (cos.f64 im))

   1. Initial program 100.0%

      \[e^{re} \cdot \cos im \]

   2. Taylor expanded in im around 0

      \[\leadsto \color{blue}{e^{re}} \]
   3. Step-by-step derivation

      1. lift-exp.f64 99.6

         \[\leadsto e^{re} \]

   4. Applied rewrites 99.6%

      \[\leadsto \color{blue}{e^{re}} \]

   5. Taylor expanded in re around 0

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

      1. +-commutative N/A

         \[\leadsto re \cdot \left(1 + \frac{1}{2} \cdot re\right) + 1 \]

      2. *-commutative N/A

         \[\leadsto \left(1 + \frac{1}{2} \cdot re\right) \cdot re + 1 \]

      3. lower-fma.f64 N/A

         \[\leadsto \mathsf{fma}\left(1 + \frac{1}{2} \cdot re, re, 1\right) \]

      4. +-commutative N/A

         \[\leadsto \mathsf{fma}\left(\frac{1}{2} \cdot re + 1, re, 1\right) \]

      5. lower-fma.f64 52.1

         \[\leadsto \mathsf{fma}\left(\mathsf{fma}\left(0.5, re, 1\right), re, 1\right) \]

   7. Applied rewrites 52.1%

      \[\leadsto \mathsf{fma}\left(\mathsf{fma}\left(0.5, re, 1\right), \color{blue}{re}, 1\right) \]

   8. Taylor expanded in re around inf

      \[\leadsto \frac{1}{2} \cdot {re}^{\color{blue}{2}} \]
   9. Step-by-step derivation

      1. *-commutative N/A

         \[\leadsto {re}^{2} \cdot \frac{1}{2} \]

      2. lower-*.f64 N/A

         \[\leadsto {re}^{2} \cdot \frac{1}{2} \]

      3. pow2 N/A

         \[\leadsto \left(re \cdot re\right) \cdot \frac{1}{2} \]

      4. lower-*.f64 52.1

         \[\leadsto \left(re \cdot re\right) \cdot 0.5 \]

   10. Applied rewrites 52.1%

       \[\leadsto \left(re \cdot re\right) \cdot 0.5 \]
3. Recombined 2 regimes into one program.
4. Add Preprocessing

Alternative 11: 37.4% accurate, 4.2× speedup

Math

\[\begin{array}{l} \\ \mathsf{fma}\left(\mathsf{fma}\left(0.5, re, 1\right), re, 1\right) \end{array} \]

FPCore

(FPCore (re im) :precision binary64 (fma (fma 0.5 re 1.0) re 1.0))

C

double code(double re, double im) {
	return fma(fma(0.5, re, 1.0), re, 1.0);
}

Julia

function code(re, im)
	return fma(fma(0.5, re, 1.0), re, 1.0)
end

Wolfram

code[re_, im_] := N[(N[(0.5 * re + 1.0), $MachinePrecision] * re + 1.0), $MachinePrecision]

Derivation

1. Initial program 100.0%

   \[e^{re} \cdot \cos im \]

2. Taylor expanded in im around 0

   \[\leadsto \color{blue}{e^{re}} \]
3. Step-by-step derivation

   1. lift-exp.f64 70.8

      \[\leadsto e^{re} \]

4. Applied rewrites 70.8%

   \[\leadsto \color{blue}{e^{re}} \]

5. Taylor expanded in re around 0

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

   1. +-commutative N/A

      \[\leadsto re \cdot \left(1 + \frac{1}{2} \cdot re\right) + 1 \]

   2. *-commutative N/A

      \[\leadsto \left(1 + \frac{1}{2} \cdot re\right) \cdot re + 1 \]

   3. lower-fma.f64 N/A

      \[\leadsto \mathsf{fma}\left(1 + \frac{1}{2} \cdot re, re, 1\right) \]

   4. +-commutative N/A

      \[\leadsto \mathsf{fma}\left(\frac{1}{2} \cdot re + 1, re, 1\right) \]

   5. lower-fma.f64 37.4

      \[\leadsto \mathsf{fma}\left(\mathsf{fma}\left(0.5, re, 1\right), re, 1\right) \]

7. Applied rewrites 37.4%

   \[\leadsto \mathsf{fma}\left(\mathsf{fma}\left(0.5, re, 1\right), \color{blue}{re}, 1\right) \]
8. Add Preprocessing

Alternative 12: 28.5% accurate, 12.7× speedup

Math

\[\begin{array}{l} \\ re - -1 \end{array} \]

FPCore

(FPCore (re im) :precision binary64 (- re -1.0))

C

double code(double re, double im) {
	return re - -1.0;
}

Fortran

module fmin_fmax_functions
    implicit none
    private
    public fmax
    public fmin

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

real(8) function code(re, im)
use fmin_fmax_functions
    real(8), intent (in) :: re
    real(8), intent (in) :: im
    code = re - (-1.0d0)
end function

Java

public static double code(double re, double im) {
	return re - -1.0;
}

Python

def code(re, im):
	return re - -1.0

Julia

function code(re, im)
	return Float64(re - -1.0)
end

MATLAB

function tmp = code(re, im)
	tmp = re - -1.0;
end

Wolfram

code[re_, im_] := N[(re - -1.0), $MachinePrecision]

Derivation

1. Initial program 100.0%

   \[e^{re} \cdot \cos im \]

2. Taylor expanded in re around 0

   \[\leadsto \color{blue}{\cos im + re \cdot \cos im} \]
3. Step-by-step derivation

   1. distribute-rgt1-in N/A

      \[\leadsto \left(re + 1\right) \cdot \color{blue}{\cos im} \]

   2. +-commutative N/A

      \[\leadsto \left(1 + re\right) \cdot \cos \color{blue}{im} \]

   3. *-commutative N/A

      \[\leadsto \cos im \cdot \color{blue}{\left(1 + re\right)} \]

   4. lower-*.f64 N/A

      \[\leadsto \cos im \cdot \color{blue}{\left(1 + re\right)} \]

   5. lift-cos.f64 N/A

      \[\leadsto \cos im \cdot \left(\color{blue}{1} + re\right) \]

   6. +-commutative N/A

      \[\leadsto \cos im \cdot \left(re + \color{blue}{1}\right) \]

   7. add-flip N/A

      \[\leadsto \cos im \cdot \left(re - \color{blue}{\left(\mathsf{neg}\left(1\right)\right)}\right) \]

   8. lower--.f64 N/A

      \[\leadsto \cos im \cdot \left(re - \color{blue}{\left(\mathsf{neg}\left(1\right)\right)}\right) \]

   9. metadata-eval 51.0

      \[\leadsto \cos im \cdot \left(re - -1\right) \]

4. Applied rewrites 51.0%

   \[\leadsto \color{blue}{\cos im \cdot \left(re - -1\right)} \]

5. Taylor expanded in im around 0

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

   1. +-commutative N/A

      \[\leadsto re + 1 \]

   2. metadata-eval N/A

      \[\leadsto re + \left(\mathsf{neg}\left(-1\right)\right) \]

   3. sub-flip N/A

      \[\leadsto re - -1 \]

   4. lift--.f64 28.5

      \[\leadsto re - -1 \]

7. Applied rewrites 28.5%

   \[\leadsto re - \color{blue}{-1} \]
8. Add Preprocessing

Alternative 13: 28.1% accurate, 46.0× speedup

Math

\[\begin{array}{l} \\ 1 \end{array} \]

FPCore

(FPCore (re im) :precision binary64 1.0)

C

double code(double re, double im) {
	return 1.0;
}

Fortran

module fmin_fmax_functions
    implicit none
    private
    public fmax
    public fmin

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

real(8) function code(re, im)
use fmin_fmax_functions
    real(8), intent (in) :: re
    real(8), intent (in) :: im
    code = 1.0d0
end function

Java

public static double code(double re, double im) {
	return 1.0;
}

Python

def code(re, im):
	return 1.0

Julia

function code(re, im)
	return 1.0
end

MATLAB

function tmp = code(re, im)
	tmp = 1.0;
end

Wolfram

code[re_, im_] := 1.0

Derivation

1. Initial program 100.0%

   \[e^{re} \cdot \cos im \]

2. Taylor expanded in im around 0

   \[\leadsto \color{blue}{e^{re}} \]
3. Step-by-step derivation

   1. lift-exp.f64 70.8

      \[\leadsto e^{re} \]

4. Applied rewrites 70.8%

   \[\leadsto \color{blue}{e^{re}} \]

5. Taylor expanded in re around 0

   \[\leadsto 1 \]
6. Step-by-step derivation

7. Applied rewrites 28.1%

   \[\leadsto 1 \]
8. Add Preprocessing

Reproduce

herbie shell --seed 2025128 
(FPCore (re im)
  :name "math.exp on complex, real part"
  :precision binary64
  (* (exp re) (cos im)))