exp-w (used to crash)

Average Accuracy: 99.6% → 99.6%

Time: 18.2s

Precision: binary64

Cost: 19456

Math

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

↓

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

FPCore

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

↓

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

C

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

↓

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

Fortran

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

↓

real(8) function code(w, l)
    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.exp(-w) * Math.pow(l, Math.exp(w));
}

↓

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.exp(-w) * math.pow(l, math.exp(w))

↓

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

Julia

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

↓

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

MATLAB

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

↓

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

Wolfram

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

↓

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

Derivation

1. Initial program 99.6%

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

2. Simplified 99.6%

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

   [Start] 99.6	\[ e^{-w} \cdot {\ell}^{\left(e^{w}\right)} \]
   exp-neg [=>] 99.6	\[ \color{blue}{\frac{1}{e^{w}}} \cdot {\ell}^{\left(e^{w}\right)} \]
   associate-*l/ [=>] 99.6	\[ \color{blue}{\frac{1 \cdot {\ell}^{\left(e^{w}\right)}}{e^{w}}} \]
   *-lft-identity [=>] 99.6	\[ \frac{\color{blue}{{\ell}^{\left(e^{w}\right)}}}{e^{w}} \]

3. Final simplification 99.6%

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

Alternatives

Alternative 1
Accuracy	98.0%
Cost	13376

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

Alternative 2
Accuracy	98.0%
Cost	13376

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

Alternative 3
Accuracy	98.1%
Cost	13376

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

Alternative 4
Accuracy	97.4%
Cost	6660

\[\begin{array}{l} \mathbf{if}\;w \leq 430:\\ \;\;\;\;\ell + \left(w \cdot w\right) \cdot \left(\ell \cdot 0.5 + w \cdot \left(\ell \cdot -0.16666666666666666\right)\right)\\ \mathbf{else}:\\ \;\;\;\;e^{-w}\\ \end{array} \]

Alternative 5
Accuracy	97.4%
Cost	6656

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

Alternative 6
Accuracy	97.4%
Cost	6592

\[\frac{\ell}{e^{w}} \]

Alternative 7
Accuracy	86.9%
Cost	1092

\[\begin{array}{l} \mathbf{if}\;w \leq 0.048:\\ \;\;\;\;\ell + \left(w \cdot w\right) \cdot \left(\ell \cdot 0.5 + w \cdot \left(\ell \cdot -0.16666666666666666\right)\right)\\ \mathbf{else}:\\ \;\;\;\;\left(1 + \ell \cdot \left(1 - w\right)\right) + -1\\ \end{array} \]

Alternative 8
Accuracy	86.9%
Cost	708

\[\begin{array}{l} \mathbf{if}\;w \leq 0.048:\\ \;\;\;\;\ell\\ \mathbf{else}:\\ \;\;\;\;\left(1 + \ell \cdot \left(1 - w\right)\right) + -1\\ \end{array} \]

Alternative 9
Accuracy	78.4%
Cost	64

\[\ell \]

Reproduce

herbie shell --seed 2023126 
(FPCore (w l)
  :name "exp-w (used to crash)"
  :precision binary64
  (* (exp (- w)) (pow l (exp w))))