exp-w (used to crash)

Percentage Accurate: 99.4% → 99.4%

Time: 13.1s

Alternatives: 11

Speedup: N/A×

• could not determine a ground truth

  1. w = 9.865302980425481e-70
  2. l = -9.163678204766524e-8

Specification

Math

\[e^{-w} \cdot {\ell}^{\left(e^{w}\right)} \]

FPCore

(FPCore (w l) :precision binary64 (* (exp (- w)) (pow l (exp w))))

C

double code(double w, double l) {
	return exp(-w) * pow(l, exp(w));
}

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(w, l)
use fmin_fmax_functions
    real(8), intent (in) :: w
    real(8), intent (in) :: l
    code = exp(-w) * (l ** exp(w))
end function

Java

public static double code(double w, double l) {
	return Math.exp(-w) * Math.pow(l, Math.exp(w));
}

Python

def code(w, l):
	return math.exp(-w) * math.pow(l, math.exp(w))

Julia

function code(w, l)
	return Float64(exp(Float64(-w)) * (l ^ exp(w)))
end

MATLAB

function tmp = code(w, l)
	tmp = exp(-w) * (l ^ exp(w));
end

Wolfram

code[w_, l_] := N[(N[Exp[(-w)], $MachinePrecision] * N[Power[l, N[Exp[w], $MachinePrecision]], $MachinePrecision]), $MachinePrecision]

Initial Program: 99.4% accurate, 1.0× speedup

Math

\[e^{-w} \cdot {\ell}^{\left(e^{w}\right)} \]

FPCore

(FPCore (w l) :precision binary64 (* (exp (- w)) (pow l (exp w))))

C

double code(double w, double l) {
	return exp(-w) * pow(l, exp(w));
}

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(w, l)
use fmin_fmax_functions
    real(8), intent (in) :: w
    real(8), intent (in) :: l
    code = exp(-w) * (l ** exp(w))
end function

Java

public static double code(double w, double l) {
	return Math.exp(-w) * Math.pow(l, Math.exp(w));
}

Python

def code(w, l):
	return math.exp(-w) * math.pow(l, math.exp(w))

Julia

function code(w, l)
	return Float64(exp(Float64(-w)) * (l ^ exp(w)))
end

MATLAB

function tmp = code(w, l)
	tmp = exp(-w) * (l ^ exp(w));
end

Wolfram

code[w_, l_] := N[(N[Exp[(-w)], $MachinePrecision] * N[Power[l, N[Exp[w], $MachinePrecision]], $MachinePrecision]), $MachinePrecision]

Alternative 1: 99.4% accurate, 1.0× speedup

Math

\[\frac{{\ell}^{\left(e^{w}\right)}}{e^{w}} \]

FPCore

(FPCore (w l) :precision binary64 (/ (pow l (exp w)) (exp w)))

C

double code(double w, double l) {
	return pow(l, exp(w)) / exp(w);
}

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(w, l)
use fmin_fmax_functions
    real(8), intent (in) :: w
    real(8), intent (in) :: l
    code = (l ** exp(w)) / exp(w)
end function

Java

public static double code(double w, double l) {
	return Math.pow(l, Math.exp(w)) / Math.exp(w);
}

Python

def code(w, l):
	return math.pow(l, math.exp(w)) / math.exp(w)

Julia

function code(w, l)
	return Float64((l ^ exp(w)) / exp(w))
end

MATLAB

function tmp = code(w, l)
	tmp = (l ^ exp(w)) / exp(w);
end

Wolfram

code[w_, l_] := N[(N[Power[l, N[Exp[w], $MachinePrecision]], $MachinePrecision] / N[Exp[w], $MachinePrecision]), $MachinePrecision]

Derivation

1. Initial program 99.4%

   \[e^{-w} \cdot {\ell}^{\left(e^{w}\right)} \]
2. Step-by-step derivation

   1. lift-*.f64 N/A

      \[\leadsto \color{blue}{e^{-w} \cdot {\ell}^{\left(e^{w}\right)}} \]

   2. *-commutative N/A

      \[\leadsto \color{blue}{{\ell}^{\left(e^{w}\right)} \cdot e^{-w}} \]

   3. lift-exp.64 N/A

      \[\leadsto {\ell}^{\left(e^{w}\right)} \cdot \color{blue}{e^{-w}} \]

   4. lift-neg.f64 N/A

      \[\leadsto {\ell}^{\left(e^{w}\right)} \cdot e^{\color{blue}{\mathsf{neg}\left(w\right)}} \]

   5. exp-neg N/A

      \[\leadsto {\ell}^{\left(e^{w}\right)} \cdot \color{blue}{\frac{1}{e^{w}}} \]

   6. lift-exp.64 N/A

      \[\leadsto {\ell}^{\left(e^{w}\right)} \cdot \frac{1}{\color{blue}{e^{w}}} \]

   7. mult-flip-rev N/A

      \[\leadsto \color{blue}{\frac{{\ell}^{\left(e^{w}\right)}}{e^{w}}} \]

   8. lower-/.f64 99.4%

      \[\leadsto \color{blue}{\frac{{\ell}^{\left(e^{w}\right)}}{e^{w}}} \]

3. Applied rewrites 99.4%

   \[\leadsto \color{blue}{\frac{{\ell}^{\left(e^{w}\right)}}{e^{w}}} \]
4. Add Preprocessing

Alternative 2: 99.0% accurate, 1.1× speedup

Math

\[\begin{array}{l} \mathbf{if}\;w \leq -2.1 \cdot 10^{-14}:\\ \;\;\;\;e^{\mathsf{fma}\left(\log \ell, e^{w}, w \cdot -1\right)}\\ \mathbf{else}:\\ \;\;\;\;{\ell}^{\left(e^{w}\right)} \cdot \left(1 - w\right)\\ \end{array} \]

FPCore

(FPCore (w l)
 :precision binary64
 (if (<= w -2.1e-14)
   (exp (fma (log l) (exp w) (* w -1.0)))
   (* (pow l (exp w)) (- 1.0 w))))

C

double code(double w, double l) {
	double tmp;
	if (w <= -2.1e-14) {
		tmp = exp(fma(log(l), exp(w), (w * -1.0)));
	} else {
		tmp = pow(l, exp(w)) * (1.0 - w);
	}
	return tmp;
}

Julia

function code(w, l)
	tmp = 0.0
	if (w <= -2.1e-14)
		tmp = exp(fma(log(l), exp(w), Float64(w * -1.0)));
	else
		tmp = Float64((l ^ exp(w)) * Float64(1.0 - w));
	end
	return tmp
end

Wolfram

code[w_, l_] := If[LessEqual[w, -2.1e-14], N[Exp[N[(N[Log[l], $MachinePrecision] * N[Exp[w], $MachinePrecision] + N[(w * -1.0), $MachinePrecision]), $MachinePrecision]], $MachinePrecision], N[(N[Power[l, N[Exp[w], $MachinePrecision]], $MachinePrecision] * N[(1.0 - w), $MachinePrecision]), $MachinePrecision]]

Derivation

1. Split input into 2 regimes

2. if w < -2.0999999999999999e-14

   1. Initial program 99.4%

      \[e^{-w} \cdot {\ell}^{\left(e^{w}\right)} \]
   2. Step-by-step derivation

      1. lift-*.f64 N/A

         \[\leadsto \color{blue}{e^{-w} \cdot {\ell}^{\left(e^{w}\right)}} \]

      2. *-commutative N/A

         \[\leadsto \color{blue}{{\ell}^{\left(e^{w}\right)} \cdot e^{-w}} \]

      3. lift-exp.64 N/A

         \[\leadsto {\ell}^{\left(e^{w}\right)} \cdot \color{blue}{e^{-w}} \]

      4. lift-neg.f64 N/A

         \[\leadsto {\ell}^{\left(e^{w}\right)} \cdot e^{\color{blue}{\mathsf{neg}\left(w\right)}} \]

      5. exp-neg N/A

         \[\leadsto {\ell}^{\left(e^{w}\right)} \cdot \color{blue}{\frac{1}{e^{w}}} \]

      6. lift-exp.64 N/A

         \[\leadsto {\ell}^{\left(e^{w}\right)} \cdot \frac{1}{\color{blue}{e^{w}}} \]

      7. mult-flip-rev N/A

         \[\leadsto \color{blue}{\frac{{\ell}^{\left(e^{w}\right)}}{e^{w}}} \]

      8. div-flip N/A

         \[\leadsto \color{blue}{\frac{1}{\frac{e^{w}}{{\ell}^{\left(e^{w}\right)}}}} \]

      9. lower-unsound-/.f64 N/A

         \[\leadsto \color{blue}{\frac{1}{\frac{e^{w}}{{\ell}^{\left(e^{w}\right)}}}} \]

      10. lower-unsound-/.f64 99.3%

          \[\leadsto \frac{1}{\color{blue}{\frac{e^{w}}{{\ell}^{\left(e^{w}\right)}}}} \]

   3. Applied rewrites 99.3%

      \[\leadsto \color{blue}{\frac{1}{\frac{e^{w}}{{\ell}^{\left(e^{w}\right)}}}} \]
   4. Step-by-step derivation

      1. lift-/.f64 N/A

         \[\leadsto \color{blue}{\frac{1}{\frac{e^{w}}{{\ell}^{\left(e^{w}\right)}}}} \]

      2. lift-/.f64 N/A

         \[\leadsto \frac{1}{\color{blue}{\frac{e^{w}}{{\ell}^{\left(e^{w}\right)}}}} \]

      3. div-flip-rev N/A

         \[\leadsto \color{blue}{\frac{{\ell}^{\left(e^{w}\right)}}{e^{w}}} \]

      4. lower-/.f64 99.4%

         \[\leadsto \color{blue}{\frac{{\ell}^{\left(e^{w}\right)}}{e^{w}}} \]

      5. lift-pow.64 N/A

         \[\leadsto \frac{\color{blue}{{\ell}^{\left(e^{w}\right)}}}{e^{w}} \]

      6. lift-exp.64 N/A

         \[\leadsto \frac{{\ell}^{\color{blue}{\left(e^{w}\right)}}}{e^{w}} \]

      7. sinh-+-cosh-rev N/A

         \[\leadsto \frac{{\ell}^{\color{blue}{\left(\cosh w + \sinh w\right)}}}{e^{w}} \]

      8. lift-cosh.64 N/A

         \[\leadsto \frac{{\ell}^{\left(\color{blue}{\cosh w} + \sinh w\right)}}{e^{w}} \]

      9. lift-sinh.64 N/A

         \[\leadsto \frac{{\ell}^{\left(\cosh w + \color{blue}{\sinh w}\right)}}{e^{w}} \]

      10. pow-prod-up N/A

          \[\leadsto \frac{\color{blue}{{\ell}^{\cosh w} \cdot {\ell}^{\sinh w}}}{e^{w}} \]

      11. lift-pow.64 N/A

          \[\leadsto \frac{\color{blue}{{\ell}^{\cosh w}} \cdot {\ell}^{\sinh w}}{e^{w}} \]

      12. lift-pow.64 N/A

          \[\leadsto \frac{{\ell}^{\cosh w} \cdot \color{blue}{{\ell}^{\sinh w}}}{e^{w}} \]

      13. *-commutative N/A

          \[\leadsto \frac{\color{blue}{{\ell}^{\sinh w} \cdot {\ell}^{\cosh w}}}{e^{w}} \]

      14. lift-*.f64 70.7%

          \[\leadsto \frac{\color{blue}{{\ell}^{\sinh w} \cdot {\ell}^{\cosh w}}}{e^{w}} \]

      15. lift-/.f64 N/A

          \[\leadsto \color{blue}{\frac{{\ell}^{\sinh w} \cdot {\ell}^{\cosh w}}{e^{w}}} \]

      16. mult-flip N/A

          \[\leadsto \color{blue}{\left({\ell}^{\sinh w} \cdot {\ell}^{\cosh w}\right) \cdot \frac{1}{e^{w}}} \]

   5. Applied rewrites 94.8%

      \[\leadsto \color{blue}{e^{\mathsf{fma}\left(\log \ell, e^{w}, w \cdot -1\right)}} \]

   if -2.0999999999999999e-14 < w

   1. Initial program 99.4%

      \[e^{-w} \cdot {\ell}^{\left(e^{w}\right)} \]

   2. Taylor expanded in w around 0

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

      1. lower-+.f64 N/A

         \[\leadsto \left(1 + \color{blue}{-1 \cdot w}\right) \cdot {\ell}^{\left(e^{w}\right)} \]

      2. lower-*.f64 71.5%

         \[\leadsto \left(1 + -1 \cdot \color{blue}{w}\right) \cdot {\ell}^{\left(e^{w}\right)} \]

   4. Applied rewrites 71.5%

      \[\leadsto \color{blue}{\left(1 + -1 \cdot w\right)} \cdot {\ell}^{\left(e^{w}\right)} \]
   5. Step-by-step derivation

      1. lift-*.f64 N/A

         \[\leadsto \color{blue}{\left(1 + -1 \cdot w\right) \cdot {\ell}^{\left(e^{w}\right)}} \]

      2. *-commutative N/A

         \[\leadsto \color{blue}{{\ell}^{\left(e^{w}\right)} \cdot \left(1 + -1 \cdot w\right)} \]

      3. lower-*.f64 71.5%

         \[\leadsto \color{blue}{{\ell}^{\left(e^{w}\right)} \cdot \left(1 + -1 \cdot w\right)} \]

      4. lift-+.f64 N/A

         \[\leadsto {\ell}^{\left(e^{w}\right)} \cdot \left(1 + \color{blue}{-1 \cdot w}\right) \]

      5. lift-*.f64 N/A

         \[\leadsto {\ell}^{\left(e^{w}\right)} \cdot \left(1 + -1 \cdot \color{blue}{w}\right) \]

      6. mul-1-neg N/A

         \[\leadsto {\ell}^{\left(e^{w}\right)} \cdot \left(1 + \left(\mathsf{neg}\left(w\right)\right)\right) \]

      7. sub-flip-reverse N/A

         \[\leadsto {\ell}^{\left(e^{w}\right)} \cdot \left(1 - \color{blue}{w}\right) \]

      8. lower--.f64 71.5%

         \[\leadsto {\ell}^{\left(e^{w}\right)} \cdot \left(1 - \color{blue}{w}\right) \]

   6. Applied rewrites 71.5%

      \[\leadsto \color{blue}{{\ell}^{\left(e^{w}\right)} \cdot \left(1 - w\right)} \]
3. Recombined 2 regimes into one program.
4. Add Preprocessing

Alternative 3: 99.0% accurate, 1.1× speedup

Math

\[\begin{array}{l} \mathbf{if}\;w \leq -5.2 \cdot 10^{-11}:\\ \;\;\;\;e^{\log \ell \cdot e^{w} - w}\\ \mathbf{else}:\\ \;\;\;\;{\ell}^{\left(e^{w}\right)} \cdot \left(1 - w\right)\\ \end{array} \]

FPCore

(FPCore (w l)
 :precision binary64
 (if (<= w -5.2e-11)
   (exp (- (* (log l) (exp w)) w))
   (* (pow l (exp w)) (- 1.0 w))))

C

double code(double w, double l) {
	double tmp;
	if (w <= -5.2e-11) {
		tmp = exp(((log(l) * exp(w)) - w));
	} else {
		tmp = pow(l, exp(w)) * (1.0 - w);
	}
	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(w, l)
use fmin_fmax_functions
    real(8), intent (in) :: w
    real(8), intent (in) :: l
    real(8) :: tmp
    if (w <= (-5.2d-11)) then
        tmp = exp(((log(l) * exp(w)) - w))
    else
        tmp = (l ** exp(w)) * (1.0d0 - w)
    end if
    code = tmp
end function

Java

public static double code(double w, double l) {
	double tmp;
	if (w <= -5.2e-11) {
		tmp = Math.exp(((Math.log(l) * Math.exp(w)) - w));
	} else {
		tmp = Math.pow(l, Math.exp(w)) * (1.0 - w);
	}
	return tmp;
}

Python

def code(w, l):
	tmp = 0
	if w <= -5.2e-11:
		tmp = math.exp(((math.log(l) * math.exp(w)) - w))
	else:
		tmp = math.pow(l, math.exp(w)) * (1.0 - w)
	return tmp

Julia

function code(w, l)
	tmp = 0.0
	if (w <= -5.2e-11)
		tmp = exp(Float64(Float64(log(l) * exp(w)) - w));
	else
		tmp = Float64((l ^ exp(w)) * Float64(1.0 - w));
	end
	return tmp
end

MATLAB

function tmp_2 = code(w, l)
	tmp = 0.0;
	if (w <= -5.2e-11)
		tmp = exp(((log(l) * exp(w)) - w));
	else
		tmp = (l ^ exp(w)) * (1.0 - w);
	end
	tmp_2 = tmp;
end

Wolfram

code[w_, l_] := If[LessEqual[w, -5.2e-11], N[Exp[N[(N[(N[Log[l], $MachinePrecision] * N[Exp[w], $MachinePrecision]), $MachinePrecision] - w), $MachinePrecision]], $MachinePrecision], N[(N[Power[l, N[Exp[w], $MachinePrecision]], $MachinePrecision] * N[(1.0 - w), $MachinePrecision]), $MachinePrecision]]

Derivation

1. Split input into 2 regimes

2. if w < -5.2000000000000001e-11

   1. Initial program 99.4%

      \[e^{-w} \cdot {\ell}^{\left(e^{w}\right)} \]
   2. Step-by-step derivation

      1. lift-*.f64 N/A

         \[\leadsto \color{blue}{e^{-w} \cdot {\ell}^{\left(e^{w}\right)}} \]

      2. *-commutative N/A

         \[\leadsto \color{blue}{{\ell}^{\left(e^{w}\right)} \cdot e^{-w}} \]

      3. lift-exp.64 N/A

         \[\leadsto {\ell}^{\left(e^{w}\right)} \cdot \color{blue}{e^{-w}} \]

      4. lift-neg.f64 N/A

         \[\leadsto {\ell}^{\left(e^{w}\right)} \cdot e^{\color{blue}{\mathsf{neg}\left(w\right)}} \]

      5. exp-neg N/A

         \[\leadsto {\ell}^{\left(e^{w}\right)} \cdot \color{blue}{\frac{1}{e^{w}}} \]

      6. lift-exp.64 N/A

         \[\leadsto {\ell}^{\left(e^{w}\right)} \cdot \frac{1}{\color{blue}{e^{w}}} \]

      7. mult-flip-rev N/A

         \[\leadsto \color{blue}{\frac{{\ell}^{\left(e^{w}\right)}}{e^{w}}} \]

      8. lower-/.f64 99.4%

         \[\leadsto \color{blue}{\frac{{\ell}^{\left(e^{w}\right)}}{e^{w}}} \]

   3. Applied rewrites 99.4%

      \[\leadsto \color{blue}{\frac{{\ell}^{\left(e^{w}\right)}}{e^{w}}} \]
   4. Step-by-step derivation

      1. lift-/.f64 N/A

         \[\leadsto \color{blue}{\frac{{\ell}^{\left(e^{w}\right)}}{e^{w}}} \]

      2. lift-pow.64 N/A

         \[\leadsto \frac{\color{blue}{{\ell}^{\left(e^{w}\right)}}}{e^{w}} \]

      3. exp-to-pow N/A

         \[\leadsto \frac{\color{blue}{e^{\log \ell \cdot e^{w}}}}{e^{w}} \]

      4. lift-log.64 N/A

         \[\leadsto \frac{e^{\color{blue}{\log \ell} \cdot e^{w}}}{e^{w}} \]

      5. lift-*.f64 N/A

         \[\leadsto \frac{e^{\color{blue}{\log \ell \cdot e^{w}}}}{e^{w}} \]

      6. lift-exp.64 N/A

         \[\leadsto \frac{e^{\log \ell \cdot e^{w}}}{\color{blue}{e^{w}}} \]

      7. div-exp N/A

         \[\leadsto \color{blue}{e^{\log \ell \cdot e^{w} - w}} \]

      8. sub-flip-reverse N/A

         \[\leadsto e^{\color{blue}{\log \ell \cdot e^{w} + \left(\mathsf{neg}\left(w\right)\right)}} \]

      9. lift-neg.f64 N/A

         \[\leadsto e^{\log \ell \cdot e^{w} + \color{blue}{\left(-w\right)}} \]

      10. +-commutative N/A

          \[\leadsto e^{\color{blue}{\left(-w\right) + \log \ell \cdot e^{w}}} \]

      11. lower-exp.64 N/A

          \[\leadsto \color{blue}{e^{\left(-w\right) + \log \ell \cdot e^{w}}} \]

      12. +-commutative N/A

          \[\leadsto e^{\color{blue}{\log \ell \cdot e^{w} + \left(-w\right)}} \]

      13. lift-neg.f64 N/A

          \[\leadsto e^{\log \ell \cdot e^{w} + \color{blue}{\left(\mathsf{neg}\left(w\right)\right)}} \]

      14. sub-flip-reverse N/A

          \[\leadsto e^{\color{blue}{\log \ell \cdot e^{w} - w}} \]

      15. lower--.f64 94.8%

          \[\leadsto e^{\color{blue}{\log \ell \cdot e^{w} - w}} \]

   5. Applied rewrites 94.8%

      \[\leadsto \color{blue}{e^{\log \ell \cdot e^{w} - w}} \]

   if -5.2000000000000001e-11 < w

   1. Initial program 99.4%

      \[e^{-w} \cdot {\ell}^{\left(e^{w}\right)} \]

   2. Taylor expanded in w around 0

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

      1. lower-+.f64 N/A

         \[\leadsto \left(1 + \color{blue}{-1 \cdot w}\right) \cdot {\ell}^{\left(e^{w}\right)} \]

      2. lower-*.f64 71.5%

         \[\leadsto \left(1 + -1 \cdot \color{blue}{w}\right) \cdot {\ell}^{\left(e^{w}\right)} \]

   4. Applied rewrites 71.5%

      \[\leadsto \color{blue}{\left(1 + -1 \cdot w\right)} \cdot {\ell}^{\left(e^{w}\right)} \]
   5. Step-by-step derivation

      1. lift-*.f64 N/A

         \[\leadsto \color{blue}{\left(1 + -1 \cdot w\right) \cdot {\ell}^{\left(e^{w}\right)}} \]

      2. *-commutative N/A

         \[\leadsto \color{blue}{{\ell}^{\left(e^{w}\right)} \cdot \left(1 + -1 \cdot w\right)} \]

      3. lower-*.f64 71.5%

         \[\leadsto \color{blue}{{\ell}^{\left(e^{w}\right)} \cdot \left(1 + -1 \cdot w\right)} \]

      4. lift-+.f64 N/A

         \[\leadsto {\ell}^{\left(e^{w}\right)} \cdot \left(1 + \color{blue}{-1 \cdot w}\right) \]

      5. lift-*.f64 N/A

         \[\leadsto {\ell}^{\left(e^{w}\right)} \cdot \left(1 + -1 \cdot \color{blue}{w}\right) \]

      6. mul-1-neg N/A

         \[\leadsto {\ell}^{\left(e^{w}\right)} \cdot \left(1 + \left(\mathsf{neg}\left(w\right)\right)\right) \]

      7. sub-flip-reverse N/A

         \[\leadsto {\ell}^{\left(e^{w}\right)} \cdot \left(1 - \color{blue}{w}\right) \]

      8. lower--.f64 71.5%

         \[\leadsto {\ell}^{\left(e^{w}\right)} \cdot \left(1 - \color{blue}{w}\right) \]

   6. Applied rewrites 71.5%

      \[\leadsto \color{blue}{{\ell}^{\left(e^{w}\right)} \cdot \left(1 - w\right)} \]
3. Recombined 2 regimes into one program.
4. Add Preprocessing

Alternative 4: 98.6% accurate, 1.1× speedup

Math

\[\begin{array}{l} \mathbf{if}\;w \leq -550:\\ \;\;\;\;e^{-w} \cdot e^{\log \ell}\\ \mathbf{else}:\\ \;\;\;\;{\ell}^{\left(e^{w}\right)} \cdot \left(1 - w\right)\\ \end{array} \]

FPCore

(FPCore (w l)
 :precision binary64
 (if (<= w -550.0)
   (* (exp (- w)) (exp (log l)))
   (* (pow l (exp w)) (- 1.0 w))))

C

double code(double w, double l) {
	double tmp;
	if (w <= -550.0) {
		tmp = exp(-w) * exp(log(l));
	} else {
		tmp = pow(l, exp(w)) * (1.0 - w);
	}
	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(w, l)
use fmin_fmax_functions
    real(8), intent (in) :: w
    real(8), intent (in) :: l
    real(8) :: tmp
    if (w <= (-550.0d0)) then
        tmp = exp(-w) * exp(log(l))
    else
        tmp = (l ** exp(w)) * (1.0d0 - w)
    end if
    code = tmp
end function

Java

public static double code(double w, double l) {
	double tmp;
	if (w <= -550.0) {
		tmp = Math.exp(-w) * Math.exp(Math.log(l));
	} else {
		tmp = Math.pow(l, Math.exp(w)) * (1.0 - w);
	}
	return tmp;
}

Python

def code(w, l):
	tmp = 0
	if w <= -550.0:
		tmp = math.exp(-w) * math.exp(math.log(l))
	else:
		tmp = math.pow(l, math.exp(w)) * (1.0 - w)
	return tmp

Julia

function code(w, l)
	tmp = 0.0
	if (w <= -550.0)
		tmp = Float64(exp(Float64(-w)) * exp(log(l)));
	else
		tmp = Float64((l ^ exp(w)) * Float64(1.0 - w));
	end
	return tmp
end

MATLAB

function tmp_2 = code(w, l)
	tmp = 0.0;
	if (w <= -550.0)
		tmp = exp(-w) * exp(log(l));
	else
		tmp = (l ^ exp(w)) * (1.0 - w);
	end
	tmp_2 = tmp;
end

Wolfram

code[w_, l_] := If[LessEqual[w, -550.0], N[(N[Exp[(-w)], $MachinePrecision] * N[Exp[N[Log[l], $MachinePrecision]], $MachinePrecision]), $MachinePrecision], N[(N[Power[l, N[Exp[w], $MachinePrecision]], $MachinePrecision] * N[(1.0 - w), $MachinePrecision]), $MachinePrecision]]

Derivation

1. Split input into 2 regimes

2. if w < -550

   1. Initial program 99.4%

      \[e^{-w} \cdot {\ell}^{\left(e^{w}\right)} \]
   2. Step-by-step derivation

      1. lift-pow.64 N/A

         \[\leadsto e^{-w} \cdot \color{blue}{{\ell}^{\left(e^{w}\right)}} \]

      2. pow-to-exp N/A

         \[\leadsto e^{-w} \cdot \color{blue}{e^{\log \ell \cdot e^{w}}} \]

      3. lower-unsound-exp.64 N/A

         \[\leadsto e^{-w} \cdot \color{blue}{e^{\log \ell \cdot e^{w}}} \]

      4. lower-unsound-*.f64 N/A

         \[\leadsto e^{-w} \cdot e^{\color{blue}{\log \ell \cdot e^{w}}} \]

      5. lower-unsound-log.64 94.5%

         \[\leadsto e^{-w} \cdot e^{\color{blue}{\log \ell} \cdot e^{w}} \]

   3. Applied rewrites 94.5%

      \[\leadsto e^{-w} \cdot \color{blue}{e^{\log \ell \cdot e^{w}}} \]

   4. Taylor expanded in w around 0

      \[\leadsto e^{-w} \cdot e^{\color{blue}{\log \ell}} \]
   5. Step-by-step derivation

      1. lower-log.64 92.7%

         \[\leadsto e^{-w} \cdot e^{\log \ell} \]

   6. Applied rewrites 92.7%

      \[\leadsto e^{-w} \cdot e^{\color{blue}{\log \ell}} \]

   if -550 < w

   1. Initial program 99.4%

      \[e^{-w} \cdot {\ell}^{\left(e^{w}\right)} \]

   2. Taylor expanded in w around 0

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

      1. lower-+.f64 N/A

         \[\leadsto \left(1 + \color{blue}{-1 \cdot w}\right) \cdot {\ell}^{\left(e^{w}\right)} \]

      2. lower-*.f64 71.5%

         \[\leadsto \left(1 + -1 \cdot \color{blue}{w}\right) \cdot {\ell}^{\left(e^{w}\right)} \]

   4. Applied rewrites 71.5%

      \[\leadsto \color{blue}{\left(1 + -1 \cdot w\right)} \cdot {\ell}^{\left(e^{w}\right)} \]
   5. Step-by-step derivation

      1. lift-*.f64 N/A

         \[\leadsto \color{blue}{\left(1 + -1 \cdot w\right) \cdot {\ell}^{\left(e^{w}\right)}} \]

      2. *-commutative N/A

         \[\leadsto \color{blue}{{\ell}^{\left(e^{w}\right)} \cdot \left(1 + -1 \cdot w\right)} \]

      3. lower-*.f64 71.5%

         \[\leadsto \color{blue}{{\ell}^{\left(e^{w}\right)} \cdot \left(1 + -1 \cdot w\right)} \]

      4. lift-+.f64 N/A

         \[\leadsto {\ell}^{\left(e^{w}\right)} \cdot \left(1 + \color{blue}{-1 \cdot w}\right) \]

      5. lift-*.f64 N/A

         \[\leadsto {\ell}^{\left(e^{w}\right)} \cdot \left(1 + -1 \cdot \color{blue}{w}\right) \]

      6. mul-1-neg N/A

         \[\leadsto {\ell}^{\left(e^{w}\right)} \cdot \left(1 + \left(\mathsf{neg}\left(w\right)\right)\right) \]

      7. sub-flip-reverse N/A

         \[\leadsto {\ell}^{\left(e^{w}\right)} \cdot \left(1 - \color{blue}{w}\right) \]

      8. lower--.f64 71.5%

         \[\leadsto {\ell}^{\left(e^{w}\right)} \cdot \left(1 - \color{blue}{w}\right) \]

   6. Applied rewrites 71.5%

      \[\leadsto \color{blue}{{\ell}^{\left(e^{w}\right)} \cdot \left(1 - w\right)} \]
3. Recombined 2 regimes into one program.
4. Add Preprocessing

Alternative 5: 98.4% accurate, 1.2× speedup

Math

\[\begin{array}{l} \mathbf{if}\;w \leq -0.019:\\ \;\;\;\;e^{-w} \cdot e^{\log \ell}\\ \mathbf{else}:\\ \;\;\;\;\left(1 + -1 \cdot w\right) \cdot {\ell}^{\left(1 + w\right)}\\ \end{array} \]

FPCore

(FPCore (w l)
 :precision binary64
 (if (<= w -0.019)
   (* (exp (- w)) (exp (log l)))
   (* (+ 1.0 (* -1.0 w)) (pow l (+ 1.0 w)))))

C

double code(double w, double l) {
	double tmp;
	if (w <= -0.019) {
		tmp = exp(-w) * exp(log(l));
	} else {
		tmp = (1.0 + (-1.0 * w)) * pow(l, (1.0 + w));
	}
	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(w, l)
use fmin_fmax_functions
    real(8), intent (in) :: w
    real(8), intent (in) :: l
    real(8) :: tmp
    if (w <= (-0.019d0)) then
        tmp = exp(-w) * exp(log(l))
    else
        tmp = (1.0d0 + ((-1.0d0) * w)) * (l ** (1.0d0 + w))
    end if
    code = tmp
end function

Java

public static double code(double w, double l) {
	double tmp;
	if (w <= -0.019) {
		tmp = Math.exp(-w) * Math.exp(Math.log(l));
	} else {
		tmp = (1.0 + (-1.0 * w)) * Math.pow(l, (1.0 + w));
	}
	return tmp;
}

Python

def code(w, l):
	tmp = 0
	if w <= -0.019:
		tmp = math.exp(-w) * math.exp(math.log(l))
	else:
		tmp = (1.0 + (-1.0 * w)) * math.pow(l, (1.0 + w))
	return tmp

Julia

function code(w, l)
	tmp = 0.0
	if (w <= -0.019)
		tmp = Float64(exp(Float64(-w)) * exp(log(l)));
	else
		tmp = Float64(Float64(1.0 + Float64(-1.0 * w)) * (l ^ Float64(1.0 + w)));
	end
	return tmp
end

MATLAB

function tmp_2 = code(w, l)
	tmp = 0.0;
	if (w <= -0.019)
		tmp = exp(-w) * exp(log(l));
	else
		tmp = (1.0 + (-1.0 * w)) * (l ^ (1.0 + w));
	end
	tmp_2 = tmp;
end

Wolfram

code[w_, l_] := If[LessEqual[w, -0.019], N[(N[Exp[(-w)], $MachinePrecision] * N[Exp[N[Log[l], $MachinePrecision]], $MachinePrecision]), $MachinePrecision], N[(N[(1.0 + N[(-1.0 * w), $MachinePrecision]), $MachinePrecision] * N[Power[l, N[(1.0 + w), $MachinePrecision]], $MachinePrecision]), $MachinePrecision]]

Derivation

1. Split input into 2 regimes

2. if w < -0.0189999999999999995

   1. Initial program 99.4%

      \[e^{-w} \cdot {\ell}^{\left(e^{w}\right)} \]
   2. Step-by-step derivation

      1. lift-pow.64 N/A

         \[\leadsto e^{-w} \cdot \color{blue}{{\ell}^{\left(e^{w}\right)}} \]

      2. pow-to-exp N/A

         \[\leadsto e^{-w} \cdot \color{blue}{e^{\log \ell \cdot e^{w}}} \]

      3. lower-unsound-exp.64 N/A

         \[\leadsto e^{-w} \cdot \color{blue}{e^{\log \ell \cdot e^{w}}} \]

      4. lower-unsound-*.f64 N/A

         \[\leadsto e^{-w} \cdot e^{\color{blue}{\log \ell \cdot e^{w}}} \]

      5. lower-unsound-log.64 94.5%

         \[\leadsto e^{-w} \cdot e^{\color{blue}{\log \ell} \cdot e^{w}} \]

   3. Applied rewrites 94.5%

      \[\leadsto e^{-w} \cdot \color{blue}{e^{\log \ell \cdot e^{w}}} \]

   4. Taylor expanded in w around 0

      \[\leadsto e^{-w} \cdot e^{\color{blue}{\log \ell}} \]
   5. Step-by-step derivation

      1. lower-log.64 92.7%

         \[\leadsto e^{-w} \cdot e^{\log \ell} \]

   6. Applied rewrites 92.7%

      \[\leadsto e^{-w} \cdot e^{\color{blue}{\log \ell}} \]

   if -0.0189999999999999995 < w

   1. Initial program 99.4%

      \[e^{-w} \cdot {\ell}^{\left(e^{w}\right)} \]

   2. Taylor expanded in w around 0

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

      1. lower-+.f64 84.5%

         \[\leadsto e^{-w} \cdot {\ell}^{\left(1 + \color{blue}{w}\right)} \]

   4. Applied rewrites 84.5%

      \[\leadsto e^{-w} \cdot {\ell}^{\color{blue}{\left(1 + w\right)}} \]

   5. Taylor expanded in w around 0

      \[\leadsto \color{blue}{\left(1 + -1 \cdot w\right)} \cdot {\ell}^{\left(1 + w\right)} \]
   6. Step-by-step derivation

      1. lower-+.f64 N/A

         \[\leadsto \left(1 + \color{blue}{-1 \cdot w}\right) \cdot {\ell}^{\left(1 + w\right)} \]

      2. lower-*.f64 84.5%

         \[\leadsto \left(1 + -1 \cdot \color{blue}{w}\right) \cdot {\ell}^{\left(1 + w\right)} \]

   7. Applied rewrites 84.5%

      \[\leadsto \color{blue}{\left(1 + -1 \cdot w\right)} \cdot {\ell}^{\left(1 + w\right)} \]
3. Recombined 2 regimes into one program.
4. Add Preprocessing

Alternative 6: 90.8% accurate, 1.3× speedup

Math

\[\begin{array}{l} \mathbf{if}\;\ell \leq 3.5 \cdot 10^{+49}:\\ \;\;\;\;\left(1 + -1 \cdot w\right) \cdot {\ell}^{\left(1 + w\right)}\\ \mathbf{else}:\\ \;\;\;\;\ell + \mathsf{fma}\left(\log \ell \cdot \ell, w, \left(-\ell\right) \cdot w\right)\\ \end{array} \]

FPCore

(FPCore (w l)
 :precision binary64
 (if (<= l 3.5e+49)
   (* (+ 1.0 (* -1.0 w)) (pow l (+ 1.0 w)))
   (+ l (fma (* (log l) l) w (* (- l) w)))))

C

double code(double w, double l) {
	double tmp;
	if (l <= 3.5e+49) {
		tmp = (1.0 + (-1.0 * w)) * pow(l, (1.0 + w));
	} else {
		tmp = l + fma((log(l) * l), w, (-l * w));
	}
	return tmp;
}

Julia

function code(w, l)
	tmp = 0.0
	if (l <= 3.5e+49)
		tmp = Float64(Float64(1.0 + Float64(-1.0 * w)) * (l ^ Float64(1.0 + w)));
	else
		tmp = Float64(l + fma(Float64(log(l) * l), w, Float64(Float64(-l) * w)));
	end
	return tmp
end

Wolfram

code[w_, l_] := If[LessEqual[l, 3.5e+49], N[(N[(1.0 + N[(-1.0 * w), $MachinePrecision]), $MachinePrecision] * N[Power[l, N[(1.0 + w), $MachinePrecision]], $MachinePrecision]), $MachinePrecision], N[(l + N[(N[(N[Log[l], $MachinePrecision] * l), $MachinePrecision] * w + N[((-l) * w), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]]

Derivation

1. Split input into 2 regimes

2. if l < 3.49999999999999975e49

   1. Initial program 99.4%

      \[e^{-w} \cdot {\ell}^{\left(e^{w}\right)} \]

   2. Taylor expanded in w around 0

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

      1. lower-+.f64 84.5%

         \[\leadsto e^{-w} \cdot {\ell}^{\left(1 + \color{blue}{w}\right)} \]

   4. Applied rewrites 84.5%

      \[\leadsto e^{-w} \cdot {\ell}^{\color{blue}{\left(1 + w\right)}} \]

   5. Taylor expanded in w around 0

      \[\leadsto \color{blue}{\left(1 + -1 \cdot w\right)} \cdot {\ell}^{\left(1 + w\right)} \]
   6. Step-by-step derivation

      1. lower-+.f64 N/A

         \[\leadsto \left(1 + \color{blue}{-1 \cdot w}\right) \cdot {\ell}^{\left(1 + w\right)} \]

      2. lower-*.f64 84.5%

         \[\leadsto \left(1 + -1 \cdot \color{blue}{w}\right) \cdot {\ell}^{\left(1 + w\right)} \]

   7. Applied rewrites 84.5%

      \[\leadsto \color{blue}{\left(1 + -1 \cdot w\right)} \cdot {\ell}^{\left(1 + w\right)} \]

   if 3.49999999999999975e49 < l

   1. Initial program 99.4%

      \[e^{-w} \cdot {\ell}^{\left(e^{w}\right)} \]
   2. Step-by-step derivation

      1. lift-*.f64 N/A

         \[\leadsto \color{blue}{e^{-w} \cdot {\ell}^{\left(e^{w}\right)}} \]

      2. *-commutative N/A

         \[\leadsto \color{blue}{{\ell}^{\left(e^{w}\right)} \cdot e^{-w}} \]

      3. lift-exp.64 N/A

         \[\leadsto {\ell}^{\left(e^{w}\right)} \cdot \color{blue}{e^{-w}} \]

      4. lift-neg.f64 N/A

         \[\leadsto {\ell}^{\left(e^{w}\right)} \cdot e^{\color{blue}{\mathsf{neg}\left(w\right)}} \]

      5. exp-neg N/A

         \[\leadsto {\ell}^{\left(e^{w}\right)} \cdot \color{blue}{\frac{1}{e^{w}}} \]

      6. lift-exp.64 N/A

         \[\leadsto {\ell}^{\left(e^{w}\right)} \cdot \frac{1}{\color{blue}{e^{w}}} \]

      7. mult-flip-rev N/A

         \[\leadsto \color{blue}{\frac{{\ell}^{\left(e^{w}\right)}}{e^{w}}} \]

      8. lower-/.f64 99.4%

         \[\leadsto \color{blue}{\frac{{\ell}^{\left(e^{w}\right)}}{e^{w}}} \]

   3. Applied rewrites 99.4%

      \[\leadsto \color{blue}{\frac{{\ell}^{\left(e^{w}\right)}}{e^{w}}} \]

   4. Taylor expanded in w around 0

      \[\leadsto \color{blue}{\ell + w \cdot \left(\ell \cdot \log \ell - \ell\right)} \]
   5. Step-by-step derivation

      1. lower-+.f64 N/A

         \[\leadsto \ell + \color{blue}{w \cdot \left(\ell \cdot \log \ell - \ell\right)} \]

      2. lower-*.f64 N/A

         \[\leadsto \ell + w \cdot \color{blue}{\left(\ell \cdot \log \ell - \ell\right)} \]

      3. lower--.f64 N/A

         \[\leadsto \ell + w \cdot \left(\ell \cdot \log \ell - \color{blue}{\ell}\right) \]

      4. lower-*.f64 N/A

         \[\leadsto \ell + w \cdot \left(\ell \cdot \log \ell - \ell\right) \]

      5. lower-log.64 56.6%

         \[\leadsto \ell + w \cdot \left(\ell \cdot \log \ell - \ell\right) \]

   6. Applied rewrites 56.6%

      \[\leadsto \color{blue}{\ell + w \cdot \left(\ell \cdot \log \ell - \ell\right)} \]
   7. Step-by-step derivation

      1. lift-*.f64 N/A

         \[\leadsto \ell + w \cdot \color{blue}{\left(\ell \cdot \log \ell - \ell\right)} \]

      2. lift--.f64 N/A

         \[\leadsto \ell + w \cdot \left(\ell \cdot \log \ell - \color{blue}{\ell}\right) \]

      3. sub-flip N/A

         \[\leadsto \ell + w \cdot \left(\ell \cdot \log \ell + \color{blue}{\left(\mathsf{neg}\left(\ell\right)\right)}\right) \]

      4. distribute-rgt-in N/A

         \[\leadsto \ell + \left(\left(\ell \cdot \log \ell\right) \cdot w + \color{blue}{\left(\mathsf{neg}\left(\ell\right)\right) \cdot w}\right) \]

      5. lower-fma.f64 N/A

         \[\leadsto \ell + \mathsf{fma}\left(\ell \cdot \log \ell, \color{blue}{w}, \left(\mathsf{neg}\left(\ell\right)\right) \cdot w\right) \]

      6. lift-*.f64 N/A

         \[\leadsto \ell + \mathsf{fma}\left(\ell \cdot \log \ell, w, \left(\mathsf{neg}\left(\ell\right)\right) \cdot w\right) \]

      7. *-commutative N/A

         \[\leadsto \ell + \mathsf{fma}\left(\log \ell \cdot \ell, w, \left(\mathsf{neg}\left(\ell\right)\right) \cdot w\right) \]

      8. lower-*.f64 N/A

         \[\leadsto \ell + \mathsf{fma}\left(\log \ell \cdot \ell, w, \left(\mathsf{neg}\left(\ell\right)\right) \cdot w\right) \]

      9. lower-*.f64 N/A

         \[\leadsto \ell + \mathsf{fma}\left(\log \ell \cdot \ell, w, \left(\mathsf{neg}\left(\ell\right)\right) \cdot w\right) \]

      10. lower-neg.f64 63.6%

          \[\leadsto \ell + \mathsf{fma}\left(\log \ell \cdot \ell, w, \left(-\ell\right) \cdot w\right) \]

   8. Applied rewrites 63.6%

      \[\leadsto \ell + \mathsf{fma}\left(\log \ell \cdot \ell, \color{blue}{w}, \left(-\ell\right) \cdot w\right) \]
3. Recombined 2 regimes into one program.
4. Add Preprocessing

Alternative 7: 77.7% accurate, 1.4× speedup

Math

\[\begin{array}{l} \mathbf{if}\;w \leq -260:\\ \;\;\;\;\ell + \mathsf{fma}\left(\log \ell \cdot \ell, w, \left(-\ell\right) \cdot w\right)\\ \mathbf{else}:\\ \;\;\;\;\frac{{\ell}^{\left(w - -1\right)}}{w - -1}\\ \end{array} \]

FPCore

(FPCore (w l)
 :precision binary64
 (if (<= w -260.0)
   (+ l (fma (* (log l) l) w (* (- l) w)))
   (/ (pow l (- w -1.0)) (- w -1.0))))

C

double code(double w, double l) {
	double tmp;
	if (w <= -260.0) {
		tmp = l + fma((log(l) * l), w, (-l * w));
	} else {
		tmp = pow(l, (w - -1.0)) / (w - -1.0);
	}
	return tmp;
}

Julia

function code(w, l)
	tmp = 0.0
	if (w <= -260.0)
		tmp = Float64(l + fma(Float64(log(l) * l), w, Float64(Float64(-l) * w)));
	else
		tmp = Float64((l ^ Float64(w - -1.0)) / Float64(w - -1.0));
	end
	return tmp
end

Wolfram

code[w_, l_] := If[LessEqual[w, -260.0], N[(l + N[(N[(N[Log[l], $MachinePrecision] * l), $MachinePrecision] * w + N[((-l) * w), $MachinePrecision]), $MachinePrecision]), $MachinePrecision], N[(N[Power[l, N[(w - -1.0), $MachinePrecision]], $MachinePrecision] / N[(w - -1.0), $MachinePrecision]), $MachinePrecision]]

Derivation

1. Split input into 2 regimes

2. if w < -260

   1. Initial program 99.4%

      \[e^{-w} \cdot {\ell}^{\left(e^{w}\right)} \]
   2. Step-by-step derivation

      1. lift-*.f64 N/A

         \[\leadsto \color{blue}{e^{-w} \cdot {\ell}^{\left(e^{w}\right)}} \]

      2. *-commutative N/A

         \[\leadsto \color{blue}{{\ell}^{\left(e^{w}\right)} \cdot e^{-w}} \]

      3. lift-exp.64 N/A

         \[\leadsto {\ell}^{\left(e^{w}\right)} \cdot \color{blue}{e^{-w}} \]

      4. lift-neg.f64 N/A

         \[\leadsto {\ell}^{\left(e^{w}\right)} \cdot e^{\color{blue}{\mathsf{neg}\left(w\right)}} \]

      5. exp-neg N/A

         \[\leadsto {\ell}^{\left(e^{w}\right)} \cdot \color{blue}{\frac{1}{e^{w}}} \]

      6. lift-exp.64 N/A

         \[\leadsto {\ell}^{\left(e^{w}\right)} \cdot \frac{1}{\color{blue}{e^{w}}} \]

      7. mult-flip-rev N/A

         \[\leadsto \color{blue}{\frac{{\ell}^{\left(e^{w}\right)}}{e^{w}}} \]

      8. lower-/.f64 99.4%

         \[\leadsto \color{blue}{\frac{{\ell}^{\left(e^{w}\right)}}{e^{w}}} \]

   3. Applied rewrites 99.4%

      \[\leadsto \color{blue}{\frac{{\ell}^{\left(e^{w}\right)}}{e^{w}}} \]

   4. Taylor expanded in w around 0

      \[\leadsto \color{blue}{\ell + w \cdot \left(\ell \cdot \log \ell - \ell\right)} \]
   5. Step-by-step derivation

      1. lower-+.f64 N/A

         \[\leadsto \ell + \color{blue}{w \cdot \left(\ell \cdot \log \ell - \ell\right)} \]

      2. lower-*.f64 N/A

         \[\leadsto \ell + w \cdot \color{blue}{\left(\ell \cdot \log \ell - \ell\right)} \]

      3. lower--.f64 N/A

         \[\leadsto \ell + w \cdot \left(\ell \cdot \log \ell - \color{blue}{\ell}\right) \]

      4. lower-*.f64 N/A

         \[\leadsto \ell + w \cdot \left(\ell \cdot \log \ell - \ell\right) \]

      5. lower-log.64 56.6%

         \[\leadsto \ell + w \cdot \left(\ell \cdot \log \ell - \ell\right) \]

   6. Applied rewrites 56.6%

      \[\leadsto \color{blue}{\ell + w \cdot \left(\ell \cdot \log \ell - \ell\right)} \]
   7. Step-by-step derivation

      1. lift-*.f64 N/A

         \[\leadsto \ell + w \cdot \color{blue}{\left(\ell \cdot \log \ell - \ell\right)} \]

      2. lift--.f64 N/A

         \[\leadsto \ell + w \cdot \left(\ell \cdot \log \ell - \color{blue}{\ell}\right) \]

      3. sub-flip N/A

         \[\leadsto \ell + w \cdot \left(\ell \cdot \log \ell + \color{blue}{\left(\mathsf{neg}\left(\ell\right)\right)}\right) \]

      4. distribute-rgt-in N/A

         \[\leadsto \ell + \left(\left(\ell \cdot \log \ell\right) \cdot w + \color{blue}{\left(\mathsf{neg}\left(\ell\right)\right) \cdot w}\right) \]

      5. lower-fma.f64 N/A

         \[\leadsto \ell + \mathsf{fma}\left(\ell \cdot \log \ell, \color{blue}{w}, \left(\mathsf{neg}\left(\ell\right)\right) \cdot w\right) \]

      6. lift-*.f64 N/A

         \[\leadsto \ell + \mathsf{fma}\left(\ell \cdot \log \ell, w, \left(\mathsf{neg}\left(\ell\right)\right) \cdot w\right) \]

      7. *-commutative N/A

         \[\leadsto \ell + \mathsf{fma}\left(\log \ell \cdot \ell, w, \left(\mathsf{neg}\left(\ell\right)\right) \cdot w\right) \]

      8. lower-*.f64 N/A

         \[\leadsto \ell + \mathsf{fma}\left(\log \ell \cdot \ell, w, \left(\mathsf{neg}\left(\ell\right)\right) \cdot w\right) \]

      9. lower-*.f64 N/A

         \[\leadsto \ell + \mathsf{fma}\left(\log \ell \cdot \ell, w, \left(\mathsf{neg}\left(\ell\right)\right) \cdot w\right) \]

      10. lower-neg.f64 63.6%

          \[\leadsto \ell + \mathsf{fma}\left(\log \ell \cdot \ell, w, \left(-\ell\right) \cdot w\right) \]

   8. Applied rewrites 63.6%

      \[\leadsto \ell + \mathsf{fma}\left(\log \ell \cdot \ell, \color{blue}{w}, \left(-\ell\right) \cdot w\right) \]

   if -260 < w

   1. Initial program 99.4%

      \[e^{-w} \cdot {\ell}^{\left(e^{w}\right)} \]
   2. Step-by-step derivation

      1. lift-exp.64 N/A

         \[\leadsto \color{blue}{e^{-w}} \cdot {\ell}^{\left(e^{w}\right)} \]

      2. lift-neg.f64 N/A

         \[\leadsto e^{\color{blue}{\mathsf{neg}\left(w\right)}} \cdot {\ell}^{\left(e^{w}\right)} \]

      3. exp-neg N/A

         \[\leadsto \color{blue}{\frac{1}{e^{w}}} \cdot {\ell}^{\left(e^{w}\right)} \]

      4. lift-exp.64 N/A

         \[\leadsto \frac{1}{\color{blue}{e^{w}}} \cdot {\ell}^{\left(e^{w}\right)} \]

      5. lower-/.f64 99.4%

         \[\leadsto \color{blue}{\frac{1}{e^{w}}} \cdot {\ell}^{\left(e^{w}\right)} \]

   3. Applied rewrites 99.4%

      \[\leadsto \color{blue}{\frac{1}{e^{w}}} \cdot {\ell}^{\left(e^{w}\right)} \]

   4. Taylor expanded in w around 0

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

      1. lower-+.f64 70.8%

         \[\leadsto \frac{1}{1 + \color{blue}{w}} \cdot {\ell}^{\left(e^{w}\right)} \]

   6. Applied rewrites 70.8%

      \[\leadsto \frac{1}{\color{blue}{1 + w}} \cdot {\ell}^{\left(e^{w}\right)} \]

   7. Taylor expanded in w around 0

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

      1. lower-+.f64 70.5%

         \[\leadsto \frac{1}{1 + w} \cdot {\ell}^{\left(1 + \color{blue}{w}\right)} \]

   9. Applied rewrites 70.5%

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

       1. lift-*.f64 N/A

          \[\leadsto \color{blue}{\frac{1}{1 + w} \cdot {\ell}^{\left(1 + w\right)}} \]

       2. *-commutative N/A

          \[\leadsto \color{blue}{{\ell}^{\left(1 + w\right)} \cdot \frac{1}{1 + w}} \]

       3. lift-/.f64 N/A

          \[\leadsto {\ell}^{\left(1 + w\right)} \cdot \color{blue}{\frac{1}{1 + w}} \]

       4. mult-flip-rev N/A

          \[\leadsto \color{blue}{\frac{{\ell}^{\left(1 + w\right)}}{1 + w}} \]

       5. lower-/.f64 70.5%

          \[\leadsto \color{blue}{\frac{{\ell}^{\left(1 + w\right)}}{1 + w}} \]

       6. lift-+.f64 N/A

          \[\leadsto \frac{{\ell}^{\left(1 + \color{blue}{w}\right)}}{1 + w} \]

       7. +-commutative N/A

          \[\leadsto \frac{{\ell}^{\left(w + \color{blue}{1}\right)}}{1 + w} \]

       8. add-flip N/A

          \[\leadsto \frac{{\ell}^{\left(w - \color{blue}{\left(\mathsf{neg}\left(1\right)\right)}\right)}}{1 + w} \]

       9. metadata-eval N/A

          \[\leadsto \frac{{\ell}^{\left(w - -1\right)}}{1 + w} \]

       10. lower--.f64 70.5%

           \[\leadsto \frac{{\ell}^{\left(w - \color{blue}{-1}\right)}}{1 + w} \]

       11. lower--.f64 N/A

           \[\leadsto \frac{{\ell}^{\left(w - \color{blue}{-1}\right)}}{\mathsf{Rewrite=>}\left(lift-+.f64, \left(1 + w\right)\right)} \]

       12. lower--.f64 N/A

           \[\leadsto \frac{{\ell}^{\left(w - \color{blue}{-1}\right)}}{\mathsf{Rewrite=>}\left(+-commutative, \left(w + 1\right)\right)} \]

       13. lower--.f64 N/A

           \[\leadsto \frac{{\ell}^{\left(w - \color{blue}{-1}\right)}}{\mathsf{Rewrite=>}\left(add-flip, \left(w - \left(\mathsf{neg}\left(1\right)\right)\right)\right)} \]

       14. lower--.f64 N/A

           \[\leadsto \frac{{\ell}^{\left(w - \color{blue}{-1}\right)}}{w - \mathsf{Rewrite=>}\left(metadata-eval, -1\right)} \]

       15. lower--.f64 N/A

           \[\leadsto \frac{{\ell}^{\left(w - \color{blue}{-1}\right)}}{\mathsf{Rewrite=>}\left(lower--.f64, \left(w - -1\right)\right)} \]

   11. Applied rewrites 70.5%

       \[\leadsto \color{blue}{\frac{{\ell}^{\left(w - -1\right)}}{w - -1}} \]
3. Recombined 2 regimes into one program.
4. Add Preprocessing

Alternative 8: 66.9% accurate, 1.5× speedup

Math

\[\begin{array}{l} \mathbf{if}\;\ell \leq 0.00018:\\ \;\;\;\;\left(1 + -1 \cdot w\right) \cdot \left(\ell + \ell \cdot \left(w \cdot \log \ell\right)\right)\\ \mathbf{else}:\\ \;\;\;\;\ell + \mathsf{fma}\left(\log \ell \cdot \ell, w, \left(-\ell\right) \cdot w\right)\\ \end{array} \]

FPCore

(FPCore (w l)
 :precision binary64
 (if (<= l 0.00018)
   (* (+ 1.0 (* -1.0 w)) (+ l (* l (* w (log l)))))
   (+ l (fma (* (log l) l) w (* (- l) w)))))

C

double code(double w, double l) {
	double tmp;
	if (l <= 0.00018) {
		tmp = (1.0 + (-1.0 * w)) * (l + (l * (w * log(l))));
	} else {
		tmp = l + fma((log(l) * l), w, (-l * w));
	}
	return tmp;
}

Julia

function code(w, l)
	tmp = 0.0
	if (l <= 0.00018)
		tmp = Float64(Float64(1.0 + Float64(-1.0 * w)) * Float64(l + Float64(l * Float64(w * log(l)))));
	else
		tmp = Float64(l + fma(Float64(log(l) * l), w, Float64(Float64(-l) * w)));
	end
	return tmp
end

Wolfram

code[w_, l_] := If[LessEqual[l, 0.00018], N[(N[(1.0 + N[(-1.0 * w), $MachinePrecision]), $MachinePrecision] * N[(l + N[(l * N[(w * N[Log[l], $MachinePrecision]), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]), $MachinePrecision], N[(l + N[(N[(N[Log[l], $MachinePrecision] * l), $MachinePrecision] * w + N[((-l) * w), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]]

Derivation

1. Split input into 2 regimes

2. if l < 1.80000000000000011e-4

   1. Initial program 99.4%

      \[e^{-w} \cdot {\ell}^{\left(e^{w}\right)} \]

   2. Taylor expanded in w around 0

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

      1. lower-+.f64 N/A

         \[\leadsto \left(1 + \color{blue}{-1 \cdot w}\right) \cdot {\ell}^{\left(e^{w}\right)} \]

      2. lower-*.f64 71.5%

         \[\leadsto \left(1 + -1 \cdot \color{blue}{w}\right) \cdot {\ell}^{\left(e^{w}\right)} \]

   4. Applied rewrites 71.5%

      \[\leadsto \color{blue}{\left(1 + -1 \cdot w\right)} \cdot {\ell}^{\left(e^{w}\right)} \]

   5. Taylor expanded in w around 0

      \[\leadsto \left(1 + -1 \cdot w\right) \cdot \color{blue}{\left(\ell + \ell \cdot \left(w \cdot \log \ell\right)\right)} \]
   6. Step-by-step derivation

      1. lower-+.f64 N/A

         \[\leadsto \left(1 + -1 \cdot w\right) \cdot \left(\ell + \color{blue}{\ell \cdot \left(w \cdot \log \ell\right)}\right) \]

      2. lower-*.f64 N/A

         \[\leadsto \left(1 + -1 \cdot w\right) \cdot \left(\ell + \ell \cdot \color{blue}{\left(w \cdot \log \ell\right)}\right) \]

      3. lower-*.f64 N/A

         \[\leadsto \left(1 + -1 \cdot w\right) \cdot \left(\ell + \ell \cdot \left(w \cdot \color{blue}{\log \ell}\right)\right) \]

      4. lower-log.64 60.3%

         \[\leadsto \left(1 + -1 \cdot w\right) \cdot \left(\ell + \ell \cdot \left(w \cdot \log \ell\right)\right) \]

   7. Applied rewrites 60.3%

      \[\leadsto \left(1 + -1 \cdot w\right) \cdot \color{blue}{\left(\ell + \ell \cdot \left(w \cdot \log \ell\right)\right)} \]

   if 1.80000000000000011e-4 < l

   1. Initial program 99.4%

      \[e^{-w} \cdot {\ell}^{\left(e^{w}\right)} \]
   2. Step-by-step derivation

      1. lift-*.f64 N/A

         \[\leadsto \color{blue}{e^{-w} \cdot {\ell}^{\left(e^{w}\right)}} \]

      2. *-commutative N/A

         \[\leadsto \color{blue}{{\ell}^{\left(e^{w}\right)} \cdot e^{-w}} \]

      3. lift-exp.64 N/A

         \[\leadsto {\ell}^{\left(e^{w}\right)} \cdot \color{blue}{e^{-w}} \]

      4. lift-neg.f64 N/A

         \[\leadsto {\ell}^{\left(e^{w}\right)} \cdot e^{\color{blue}{\mathsf{neg}\left(w\right)}} \]

      5. exp-neg N/A

         \[\leadsto {\ell}^{\left(e^{w}\right)} \cdot \color{blue}{\frac{1}{e^{w}}} \]

      6. lift-exp.64 N/A

         \[\leadsto {\ell}^{\left(e^{w}\right)} \cdot \frac{1}{\color{blue}{e^{w}}} \]

      7. mult-flip-rev N/A

         \[\leadsto \color{blue}{\frac{{\ell}^{\left(e^{w}\right)}}{e^{w}}} \]

      8. lower-/.f64 99.4%

         \[\leadsto \color{blue}{\frac{{\ell}^{\left(e^{w}\right)}}{e^{w}}} \]

   3. Applied rewrites 99.4%

      \[\leadsto \color{blue}{\frac{{\ell}^{\left(e^{w}\right)}}{e^{w}}} \]

   4. Taylor expanded in w around 0

      \[\leadsto \color{blue}{\ell + w \cdot \left(\ell \cdot \log \ell - \ell\right)} \]
   5. Step-by-step derivation

      1. lower-+.f64 N/A

         \[\leadsto \ell + \color{blue}{w \cdot \left(\ell \cdot \log \ell - \ell\right)} \]

      2. lower-*.f64 N/A

         \[\leadsto \ell + w \cdot \color{blue}{\left(\ell \cdot \log \ell - \ell\right)} \]

      3. lower--.f64 N/A

         \[\leadsto \ell + w \cdot \left(\ell \cdot \log \ell - \color{blue}{\ell}\right) \]

      4. lower-*.f64 N/A

         \[\leadsto \ell + w \cdot \left(\ell \cdot \log \ell - \ell\right) \]

      5. lower-log.64 56.6%

         \[\leadsto \ell + w \cdot \left(\ell \cdot \log \ell - \ell\right) \]

   6. Applied rewrites 56.6%

      \[\leadsto \color{blue}{\ell + w \cdot \left(\ell \cdot \log \ell - \ell\right)} \]
   7. Step-by-step derivation

      1. lift-*.f64 N/A

         \[\leadsto \ell + w \cdot \color{blue}{\left(\ell \cdot \log \ell - \ell\right)} \]

      2. lift--.f64 N/A

         \[\leadsto \ell + w \cdot \left(\ell \cdot \log \ell - \color{blue}{\ell}\right) \]

      3. sub-flip N/A

         \[\leadsto \ell + w \cdot \left(\ell \cdot \log \ell + \color{blue}{\left(\mathsf{neg}\left(\ell\right)\right)}\right) \]

      4. distribute-rgt-in N/A

         \[\leadsto \ell + \left(\left(\ell \cdot \log \ell\right) \cdot w + \color{blue}{\left(\mathsf{neg}\left(\ell\right)\right) \cdot w}\right) \]

      5. lower-fma.f64 N/A

         \[\leadsto \ell + \mathsf{fma}\left(\ell \cdot \log \ell, \color{blue}{w}, \left(\mathsf{neg}\left(\ell\right)\right) \cdot w\right) \]

      6. lift-*.f64 N/A

         \[\leadsto \ell + \mathsf{fma}\left(\ell \cdot \log \ell, w, \left(\mathsf{neg}\left(\ell\right)\right) \cdot w\right) \]

      7. *-commutative N/A

         \[\leadsto \ell + \mathsf{fma}\left(\log \ell \cdot \ell, w, \left(\mathsf{neg}\left(\ell\right)\right) \cdot w\right) \]

      8. lower-*.f64 N/A

         \[\leadsto \ell + \mathsf{fma}\left(\log \ell \cdot \ell, w, \left(\mathsf{neg}\left(\ell\right)\right) \cdot w\right) \]

      9. lower-*.f64 N/A

         \[\leadsto \ell + \mathsf{fma}\left(\log \ell \cdot \ell, w, \left(\mathsf{neg}\left(\ell\right)\right) \cdot w\right) \]

      10. lower-neg.f64 63.6%

          \[\leadsto \ell + \mathsf{fma}\left(\log \ell \cdot \ell, w, \left(-\ell\right) \cdot w\right) \]

   8. Applied rewrites 63.6%

      \[\leadsto \ell + \mathsf{fma}\left(\log \ell \cdot \ell, \color{blue}{w}, \left(-\ell\right) \cdot w\right) \]
3. Recombined 2 regimes into one program.
4. Add Preprocessing

Alternative 9: 63.6% accurate, 1.9× speedup

Math

\[\ell + \mathsf{fma}\left(\log \ell \cdot \ell, w, \left(-\ell\right) \cdot w\right) \]

FPCore

(FPCore (w l) :precision binary64 (+ l (fma (* (log l) l) w (* (- l) w))))

C

double code(double w, double l) {
	return l + fma((log(l) * l), w, (-l * w));
}

Julia

function code(w, l)
	return Float64(l + fma(Float64(log(l) * l), w, Float64(Float64(-l) * w)))
end

Wolfram

code[w_, l_] := N[(l + N[(N[(N[Log[l], $MachinePrecision] * l), $MachinePrecision] * w + N[((-l) * w), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]

Derivation

1. Initial program 99.4%

   \[e^{-w} \cdot {\ell}^{\left(e^{w}\right)} \]
2. Step-by-step derivation

   1. lift-*.f64 N/A

      \[\leadsto \color{blue}{e^{-w} \cdot {\ell}^{\left(e^{w}\right)}} \]

   2. *-commutative N/A

      \[\leadsto \color{blue}{{\ell}^{\left(e^{w}\right)} \cdot e^{-w}} \]

   3. lift-exp.64 N/A

      \[\leadsto {\ell}^{\left(e^{w}\right)} \cdot \color{blue}{e^{-w}} \]

   4. lift-neg.f64 N/A

      \[\leadsto {\ell}^{\left(e^{w}\right)} \cdot e^{\color{blue}{\mathsf{neg}\left(w\right)}} \]

   5. exp-neg N/A

      \[\leadsto {\ell}^{\left(e^{w}\right)} \cdot \color{blue}{\frac{1}{e^{w}}} \]

   6. lift-exp.64 N/A

      \[\leadsto {\ell}^{\left(e^{w}\right)} \cdot \frac{1}{\color{blue}{e^{w}}} \]

   7. mult-flip-rev N/A

      \[\leadsto \color{blue}{\frac{{\ell}^{\left(e^{w}\right)}}{e^{w}}} \]

   8. lower-/.f64 99.4%

      \[\leadsto \color{blue}{\frac{{\ell}^{\left(e^{w}\right)}}{e^{w}}} \]

3. Applied rewrites 99.4%

   \[\leadsto \color{blue}{\frac{{\ell}^{\left(e^{w}\right)}}{e^{w}}} \]

4. Taylor expanded in w around 0

   \[\leadsto \color{blue}{\ell + w \cdot \left(\ell \cdot \log \ell - \ell\right)} \]
5. Step-by-step derivation

   1. lower-+.f64 N/A

      \[\leadsto \ell + \color{blue}{w \cdot \left(\ell \cdot \log \ell - \ell\right)} \]

   2. lower-*.f64 N/A

      \[\leadsto \ell + w \cdot \color{blue}{\left(\ell \cdot \log \ell - \ell\right)} \]

   3. lower--.f64 N/A

      \[\leadsto \ell + w \cdot \left(\ell \cdot \log \ell - \color{blue}{\ell}\right) \]

   4. lower-*.f64 N/A

      \[\leadsto \ell + w \cdot \left(\ell \cdot \log \ell - \ell\right) \]

   5. lower-log.64 56.6%

      \[\leadsto \ell + w \cdot \left(\ell \cdot \log \ell - \ell\right) \]

6. Applied rewrites 56.6%

   \[\leadsto \color{blue}{\ell + w \cdot \left(\ell \cdot \log \ell - \ell\right)} \]
7. Step-by-step derivation

   1. lift-*.f64 N/A

      \[\leadsto \ell + w \cdot \color{blue}{\left(\ell \cdot \log \ell - \ell\right)} \]

   2. lift--.f64 N/A

      \[\leadsto \ell + w \cdot \left(\ell \cdot \log \ell - \color{blue}{\ell}\right) \]

   3. sub-flip N/A

      \[\leadsto \ell + w \cdot \left(\ell \cdot \log \ell + \color{blue}{\left(\mathsf{neg}\left(\ell\right)\right)}\right) \]

   4. distribute-rgt-in N/A

      \[\leadsto \ell + \left(\left(\ell \cdot \log \ell\right) \cdot w + \color{blue}{\left(\mathsf{neg}\left(\ell\right)\right) \cdot w}\right) \]

   5. lower-fma.f64 N/A

      \[\leadsto \ell + \mathsf{fma}\left(\ell \cdot \log \ell, \color{blue}{w}, \left(\mathsf{neg}\left(\ell\right)\right) \cdot w\right) \]

   6. lift-*.f64 N/A

      \[\leadsto \ell + \mathsf{fma}\left(\ell \cdot \log \ell, w, \left(\mathsf{neg}\left(\ell\right)\right) \cdot w\right) \]

   7. *-commutative N/A

      \[\leadsto \ell + \mathsf{fma}\left(\log \ell \cdot \ell, w, \left(\mathsf{neg}\left(\ell\right)\right) \cdot w\right) \]

   8. lower-*.f64 N/A

      \[\leadsto \ell + \mathsf{fma}\left(\log \ell \cdot \ell, w, \left(\mathsf{neg}\left(\ell\right)\right) \cdot w\right) \]

   9. lower-*.f64 N/A

      \[\leadsto \ell + \mathsf{fma}\left(\log \ell \cdot \ell, w, \left(\mathsf{neg}\left(\ell\right)\right) \cdot w\right) \]

   10. lower-neg.f64 63.6%

       \[\leadsto \ell + \mathsf{fma}\left(\log \ell \cdot \ell, w, \left(-\ell\right) \cdot w\right) \]

8. Applied rewrites 63.6%

   \[\leadsto \ell + \mathsf{fma}\left(\log \ell \cdot \ell, \color{blue}{w}, \left(-\ell\right) \cdot w\right) \]
9. Add Preprocessing

Alternative 10: 57.2% accurate, 2.3× speedup

Math

\[\ell + \ell \cdot \left(w \cdot \left(\log \ell - 1\right)\right) \]

FPCore

(FPCore (w l) :precision binary64 (+ l (* l (* w (- (log l) 1.0)))))

C

double code(double w, double l) {
	return l + (l * (w * (log(l) - 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(w, l)
use fmin_fmax_functions
    real(8), intent (in) :: w
    real(8), intent (in) :: l
    code = l + (l * (w * (log(l) - 1.0d0)))
end function

Java

public static double code(double w, double l) {
	return l + (l * (w * (Math.log(l) - 1.0)));
}

Python

def code(w, l):
	return l + (l * (w * (math.log(l) - 1.0)))

Julia

function code(w, l)
	return Float64(l + Float64(l * Float64(w * Float64(log(l) - 1.0))))
end

MATLAB

function tmp = code(w, l)
	tmp = l + (l * (w * (log(l) - 1.0)));
end

Wolfram

code[w_, l_] := N[(l + N[(l * N[(w * N[(N[Log[l], $MachinePrecision] - 1.0), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]

Derivation

1. Initial program 99.4%

   \[e^{-w} \cdot {\ell}^{\left(e^{w}\right)} \]
2. Step-by-step derivation

   1. lift-*.f64 N/A

      \[\leadsto \color{blue}{e^{-w} \cdot {\ell}^{\left(e^{w}\right)}} \]

   2. *-commutative N/A

      \[\leadsto \color{blue}{{\ell}^{\left(e^{w}\right)} \cdot e^{-w}} \]

   3. lift-exp.64 N/A

      \[\leadsto {\ell}^{\left(e^{w}\right)} \cdot \color{blue}{e^{-w}} \]

   4. lift-neg.f64 N/A

      \[\leadsto {\ell}^{\left(e^{w}\right)} \cdot e^{\color{blue}{\mathsf{neg}\left(w\right)}} \]

   5. exp-neg N/A

      \[\leadsto {\ell}^{\left(e^{w}\right)} \cdot \color{blue}{\frac{1}{e^{w}}} \]

   6. lift-exp.64 N/A

      \[\leadsto {\ell}^{\left(e^{w}\right)} \cdot \frac{1}{\color{blue}{e^{w}}} \]

   7. mult-flip-rev N/A

      \[\leadsto \color{blue}{\frac{{\ell}^{\left(e^{w}\right)}}{e^{w}}} \]

   8. lower-/.f64 99.4%

      \[\leadsto \color{blue}{\frac{{\ell}^{\left(e^{w}\right)}}{e^{w}}} \]

3. Applied rewrites 99.4%

   \[\leadsto \color{blue}{\frac{{\ell}^{\left(e^{w}\right)}}{e^{w}}} \]

4. Taylor expanded in w around 0

   \[\leadsto \color{blue}{\ell + w \cdot \left(\ell \cdot \log \ell - \ell\right)} \]
5. Step-by-step derivation

   1. lower-+.f64 N/A

      \[\leadsto \ell + \color{blue}{w \cdot \left(\ell \cdot \log \ell - \ell\right)} \]

   2. lower-*.f64 N/A

      \[\leadsto \ell + w \cdot \color{blue}{\left(\ell \cdot \log \ell - \ell\right)} \]

   3. lower--.f64 N/A

      \[\leadsto \ell + w \cdot \left(\ell \cdot \log \ell - \color{blue}{\ell}\right) \]

   4. lower-*.f64 N/A

      \[\leadsto \ell + w \cdot \left(\ell \cdot \log \ell - \ell\right) \]

   5. lower-log.64 56.6%

      \[\leadsto \ell + w \cdot \left(\ell \cdot \log \ell - \ell\right) \]

6. Applied rewrites 56.6%

   \[\leadsto \color{blue}{\ell + w \cdot \left(\ell \cdot \log \ell - \ell\right)} \]

7. Taylor expanded in l around 0

   \[\leadsto \ell + \ell \cdot \color{blue}{\left(w \cdot \left(\log \ell - 1\right)\right)} \]
8. Step-by-step derivation

   1. lower-*.f64 N/A

      \[\leadsto \ell + \ell \cdot \left(w \cdot \color{blue}{\left(\log \ell - 1\right)}\right) \]

   2. lower-*.f64 N/A

      \[\leadsto \ell + \ell \cdot \left(w \cdot \left(\log \ell - \color{blue}{1}\right)\right) \]

   3. lower--.f64 N/A

      \[\leadsto \ell + \ell \cdot \left(w \cdot \left(\log \ell - 1\right)\right) \]

   4. lower-log.64 57.1%

      \[\leadsto \ell + \ell \cdot \left(w \cdot \left(\log \ell - 1\right)\right) \]

9. Applied rewrites 57.1%

   \[\leadsto \ell + \ell \cdot \color{blue}{\left(w \cdot \left(\log \ell - 1\right)\right)} \]
10. Add Preprocessing

Alternative 11: 57.1% accurate, 41.0× speedup

Math

\[\ell \]

FPCore

(FPCore (w l) :precision binary64 l)

C

double code(double w, double l) {
	return l;
}

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(w, l)
use fmin_fmax_functions
    real(8), intent (in) :: w
    real(8), intent (in) :: l
    code = l
end function

Java

public static double code(double w, double l) {
	return l;
}

Python

def code(w, l):
	return l

Julia

function code(w, l)
	return l
end

MATLAB

function tmp = code(w, l)
	tmp = l;
end

Wolfram

code[w_, l_] := l

Derivation

1. Initial program 99.4%

   \[e^{-w} \cdot {\ell}^{\left(e^{w}\right)} \]

2. Taylor expanded in w around 0

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

4. Applied rewrites 57.2%

   \[\leadsto \color{blue}{\ell} \]
5. Add Preprocessing

Reproduce

herbie shell --seed 2025183 
(FPCore (w l)
  :name "exp-w (used to crash)"
  :precision binary64
  (* (exp (- w)) (pow l (exp w))))