
(FPCore (x) :precision binary64 (let* ((t_0 (exp (- x)))) (/ (- (exp x) t_0) (+ (exp x) t_0))))
double code(double x) {
double t_0 = exp(-x);
return (exp(x) - t_0) / (exp(x) + t_0);
}
real(8) function code(x)
real(8), intent (in) :: x
real(8) :: t_0
t_0 = exp(-x)
code = (exp(x) - t_0) / (exp(x) + t_0)
end function
public static double code(double x) {
double t_0 = Math.exp(-x);
return (Math.exp(x) - t_0) / (Math.exp(x) + t_0);
}
def code(x): t_0 = math.exp(-x) return (math.exp(x) - t_0) / (math.exp(x) + t_0)
function code(x) t_0 = exp(Float64(-x)) return Float64(Float64(exp(x) - t_0) / Float64(exp(x) + t_0)) end
function tmp = code(x) t_0 = exp(-x); tmp = (exp(x) - t_0) / (exp(x) + t_0); end
code[x_] := Block[{t$95$0 = N[Exp[(-x)], $MachinePrecision]}, N[(N[(N[Exp[x], $MachinePrecision] - t$95$0), $MachinePrecision] / N[(N[Exp[x], $MachinePrecision] + t$95$0), $MachinePrecision]), $MachinePrecision]]
\begin{array}{l}
\\
\begin{array}{l}
t_0 := e^{-x}\\
\frac{e^{x} - t_0}{e^{x} + t_0}
\end{array}
\end{array}
Sampling outcomes in binary64 precision:
Herbie found 9 alternatives:
| Alternative | Accuracy | Speedup |
|---|
(FPCore (x) :precision binary64 (let* ((t_0 (exp (- x)))) (/ (- (exp x) t_0) (+ (exp x) t_0))))
double code(double x) {
double t_0 = exp(-x);
return (exp(x) - t_0) / (exp(x) + t_0);
}
real(8) function code(x)
real(8), intent (in) :: x
real(8) :: t_0
t_0 = exp(-x)
code = (exp(x) - t_0) / (exp(x) + t_0)
end function
public static double code(double x) {
double t_0 = Math.exp(-x);
return (Math.exp(x) - t_0) / (Math.exp(x) + t_0);
}
def code(x): t_0 = math.exp(-x) return (math.exp(x) - t_0) / (math.exp(x) + t_0)
function code(x) t_0 = exp(Float64(-x)) return Float64(Float64(exp(x) - t_0) / Float64(exp(x) + t_0)) end
function tmp = code(x) t_0 = exp(-x); tmp = (exp(x) - t_0) / (exp(x) + t_0); end
code[x_] := Block[{t$95$0 = N[Exp[(-x)], $MachinePrecision]}, N[(N[(N[Exp[x], $MachinePrecision] - t$95$0), $MachinePrecision] / N[(N[Exp[x], $MachinePrecision] + t$95$0), $MachinePrecision]), $MachinePrecision]]
\begin{array}{l}
\\
\begin{array}{l}
t_0 := e^{-x}\\
\frac{e^{x} - t_0}{e^{x} + t_0}
\end{array}
\end{array}
(FPCore (x)
:precision binary64
(/
(fma
2.0
x
(fma
0.3333333333333333
(pow x 3.0)
(fma
0.016666666666666666
(pow x 5.0)
(* 0.0003968253968253968 (pow x 7.0)))))
(+
2.0
(+
(* 0.002777777777777778 (pow x 6.0))
(+ (* 0.08333333333333333 (pow x 4.0)) (pow x 2.0))))))
double code(double x) {
return fma(2.0, x, fma(0.3333333333333333, pow(x, 3.0), fma(0.016666666666666666, pow(x, 5.0), (0.0003968253968253968 * pow(x, 7.0))))) / (2.0 + ((0.002777777777777778 * pow(x, 6.0)) + ((0.08333333333333333 * pow(x, 4.0)) + pow(x, 2.0))));
}
function code(x) return Float64(fma(2.0, x, fma(0.3333333333333333, (x ^ 3.0), fma(0.016666666666666666, (x ^ 5.0), Float64(0.0003968253968253968 * (x ^ 7.0))))) / Float64(2.0 + Float64(Float64(0.002777777777777778 * (x ^ 6.0)) + Float64(Float64(0.08333333333333333 * (x ^ 4.0)) + (x ^ 2.0))))) end
code[x_] := N[(N[(2.0 * x + N[(0.3333333333333333 * N[Power[x, 3.0], $MachinePrecision] + N[(0.016666666666666666 * N[Power[x, 5.0], $MachinePrecision] + N[(0.0003968253968253968 * N[Power[x, 7.0], $MachinePrecision]), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]), $MachinePrecision] / N[(2.0 + N[(N[(0.002777777777777778 * N[Power[x, 6.0], $MachinePrecision]), $MachinePrecision] + N[(N[(0.08333333333333333 * N[Power[x, 4.0], $MachinePrecision]), $MachinePrecision] + N[Power[x, 2.0], $MachinePrecision]), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]
\begin{array}{l}
\\
\frac{\mathsf{fma}\left(2, x, \mathsf{fma}\left(0.3333333333333333, {x}^{3}, \mathsf{fma}\left(0.016666666666666666, {x}^{5}, 0.0003968253968253968 \cdot {x}^{7}\right)\right)\right)}{2 + \left(0.002777777777777778 \cdot {x}^{6} + \left(0.08333333333333333 \cdot {x}^{4} + {x}^{2}\right)\right)}
\end{array}
Initial program 9.6%
Taylor expanded in x around 0 96.3%
associate-+r+96.3%
associate-+r+96.3%
+-commutative96.3%
fma-def96.3%
associate-+r+96.3%
+-commutative96.3%
fma-def96.3%
+-commutative96.3%
fma-def96.3%
Simplified96.3%
Taylor expanded in x around 0 96.6%
Final simplification96.6%
(FPCore (x)
:precision binary64
(*
0.5
(/
(fma
(pow x 7.0)
0.0003968253968253968
(fma
x
(fma 0.3333333333333333 (* x x) 2.0)
(* 0.016666666666666666 (pow x 5.0))))
(cosh x))))
double code(double x) {
return 0.5 * (fma(pow(x, 7.0), 0.0003968253968253968, fma(x, fma(0.3333333333333333, (x * x), 2.0), (0.016666666666666666 * pow(x, 5.0)))) / cosh(x));
}
function code(x) return Float64(0.5 * Float64(fma((x ^ 7.0), 0.0003968253968253968, fma(x, fma(0.3333333333333333, Float64(x * x), 2.0), Float64(0.016666666666666666 * (x ^ 5.0)))) / cosh(x))) end
code[x_] := N[(0.5 * N[(N[(N[Power[x, 7.0], $MachinePrecision] * 0.0003968253968253968 + N[(x * N[(0.3333333333333333 * N[(x * x), $MachinePrecision] + 2.0), $MachinePrecision] + N[(0.016666666666666666 * N[Power[x, 5.0], $MachinePrecision]), $MachinePrecision]), $MachinePrecision]), $MachinePrecision] / N[Cosh[x], $MachinePrecision]), $MachinePrecision]), $MachinePrecision]
\begin{array}{l}
\\
0.5 \cdot \frac{\mathsf{fma}\left({x}^{7}, 0.0003968253968253968, \mathsf{fma}\left(x, \mathsf{fma}\left(0.3333333333333333, x \cdot x, 2\right), 0.016666666666666666 \cdot {x}^{5}\right)\right)}{\cosh x}
\end{array}
Initial program 9.6%
Taylor expanded in x around 0 96.3%
div-inv96.3%
fma-def96.3%
fma-def96.3%
fma-def96.3%
*-commutative96.3%
cosh-undef96.3%
Applied egg-rr96.3%
associate-*r/96.3%
*-commutative96.3%
times-frac96.3%
metadata-eval96.3%
Simplified96.3%
Final simplification96.3%
(FPCore (x)
:precision binary64
(/
(+
(* 0.0003968253968253968 (pow x 7.0))
(+
(* 0.016666666666666666 (pow x 5.0))
(+ (* 0.3333333333333333 (pow x 3.0)) (* 2.0 x))))
(+ (exp x) (exp (- x)))))
double code(double x) {
return ((0.0003968253968253968 * pow(x, 7.0)) + ((0.016666666666666666 * pow(x, 5.0)) + ((0.3333333333333333 * pow(x, 3.0)) + (2.0 * x)))) / (exp(x) + exp(-x));
}
real(8) function code(x)
real(8), intent (in) :: x
code = ((0.0003968253968253968d0 * (x ** 7.0d0)) + ((0.016666666666666666d0 * (x ** 5.0d0)) + ((0.3333333333333333d0 * (x ** 3.0d0)) + (2.0d0 * x)))) / (exp(x) + exp(-x))
end function
public static double code(double x) {
return ((0.0003968253968253968 * Math.pow(x, 7.0)) + ((0.016666666666666666 * Math.pow(x, 5.0)) + ((0.3333333333333333 * Math.pow(x, 3.0)) + (2.0 * x)))) / (Math.exp(x) + Math.exp(-x));
}
def code(x): return ((0.0003968253968253968 * math.pow(x, 7.0)) + ((0.016666666666666666 * math.pow(x, 5.0)) + ((0.3333333333333333 * math.pow(x, 3.0)) + (2.0 * x)))) / (math.exp(x) + math.exp(-x))
function code(x) return Float64(Float64(Float64(0.0003968253968253968 * (x ^ 7.0)) + Float64(Float64(0.016666666666666666 * (x ^ 5.0)) + Float64(Float64(0.3333333333333333 * (x ^ 3.0)) + Float64(2.0 * x)))) / Float64(exp(x) + exp(Float64(-x)))) end
function tmp = code(x) tmp = ((0.0003968253968253968 * (x ^ 7.0)) + ((0.016666666666666666 * (x ^ 5.0)) + ((0.3333333333333333 * (x ^ 3.0)) + (2.0 * x)))) / (exp(x) + exp(-x)); end
code[x_] := N[(N[(N[(0.0003968253968253968 * N[Power[x, 7.0], $MachinePrecision]), $MachinePrecision] + N[(N[(0.016666666666666666 * N[Power[x, 5.0], $MachinePrecision]), $MachinePrecision] + N[(N[(0.3333333333333333 * N[Power[x, 3.0], $MachinePrecision]), $MachinePrecision] + N[(2.0 * x), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]), $MachinePrecision] / N[(N[Exp[x], $MachinePrecision] + N[Exp[(-x)], $MachinePrecision]), $MachinePrecision]), $MachinePrecision]
\begin{array}{l}
\\
\frac{0.0003968253968253968 \cdot {x}^{7} + \left(0.016666666666666666 \cdot {x}^{5} + \left(0.3333333333333333 \cdot {x}^{3} + 2 \cdot x\right)\right)}{e^{x} + e^{-x}}
\end{array}
Initial program 9.6%
Taylor expanded in x around 0 96.3%
Final simplification96.3%
(FPCore (x) :precision binary64 (/ (* x (fma x (* x 0.3333333333333333) 2.0)) (+ 2.0 (+ (* 0.08333333333333333 (pow x 4.0)) (* x x)))))
double code(double x) {
return (x * fma(x, (x * 0.3333333333333333), 2.0)) / (2.0 + ((0.08333333333333333 * pow(x, 4.0)) + (x * x)));
}
function code(x) return Float64(Float64(x * fma(x, Float64(x * 0.3333333333333333), 2.0)) / Float64(2.0 + Float64(Float64(0.08333333333333333 * (x ^ 4.0)) + Float64(x * x)))) end
code[x_] := N[(N[(x * N[(x * N[(x * 0.3333333333333333), $MachinePrecision] + 2.0), $MachinePrecision]), $MachinePrecision] / N[(2.0 + N[(N[(0.08333333333333333 * N[Power[x, 4.0], $MachinePrecision]), $MachinePrecision] + N[(x * x), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]
\begin{array}{l}
\\
\frac{x \cdot \mathsf{fma}\left(x, x \cdot 0.3333333333333333, 2\right)}{2 + \left(0.08333333333333333 \cdot {x}^{4} + x \cdot x\right)}
\end{array}
Initial program 9.6%
Taylor expanded in x around 0 95.9%
unpow396.2%
unpow296.2%
associate-*r*96.2%
distribute-rgt-out96.2%
*-commutative96.2%
unpow296.2%
associate-*l*96.2%
fma-def96.2%
Simplified95.9%
Taylor expanded in x around 0 96.2%
fma-def96.2%
unpow296.2%
Simplified96.2%
fma-udef96.2%
Applied egg-rr96.2%
Final simplification96.2%
(FPCore (x) :precision binary64 (/ (+ (* 0.3333333333333333 (pow x 3.0)) (* 2.0 x)) (+ 2.0 (* x x))))
double code(double x) {
return ((0.3333333333333333 * pow(x, 3.0)) + (2.0 * x)) / (2.0 + (x * x));
}
real(8) function code(x)
real(8), intent (in) :: x
code = ((0.3333333333333333d0 * (x ** 3.0d0)) + (2.0d0 * x)) / (2.0d0 + (x * x))
end function
public static double code(double x) {
return ((0.3333333333333333 * Math.pow(x, 3.0)) + (2.0 * x)) / (2.0 + (x * x));
}
def code(x): return ((0.3333333333333333 * math.pow(x, 3.0)) + (2.0 * x)) / (2.0 + (x * x))
function code(x) return Float64(Float64(Float64(0.3333333333333333 * (x ^ 3.0)) + Float64(2.0 * x)) / Float64(2.0 + Float64(x * x))) end
function tmp = code(x) tmp = ((0.3333333333333333 * (x ^ 3.0)) + (2.0 * x)) / (2.0 + (x * x)); end
code[x_] := N[(N[(N[(0.3333333333333333 * N[Power[x, 3.0], $MachinePrecision]), $MachinePrecision] + N[(2.0 * x), $MachinePrecision]), $MachinePrecision] / N[(2.0 + N[(x * x), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]
\begin{array}{l}
\\
\frac{0.3333333333333333 \cdot {x}^{3} + 2 \cdot x}{2 + x \cdot x}
\end{array}
Initial program 9.6%
Taylor expanded in x around 0 8.8%
unpow28.8%
Simplified8.8%
Taylor expanded in x around 0 96.2%
Final simplification96.2%
(FPCore (x) :precision binary64 (/ (* x (+ 2.0 (* x (* x 0.3333333333333333)))) (+ 2.0 (* x x))))
double code(double x) {
return (x * (2.0 + (x * (x * 0.3333333333333333)))) / (2.0 + (x * x));
}
real(8) function code(x)
real(8), intent (in) :: x
code = (x * (2.0d0 + (x * (x * 0.3333333333333333d0)))) / (2.0d0 + (x * x))
end function
public static double code(double x) {
return (x * (2.0 + (x * (x * 0.3333333333333333)))) / (2.0 + (x * x));
}
def code(x): return (x * (2.0 + (x * (x * 0.3333333333333333)))) / (2.0 + (x * x))
function code(x) return Float64(Float64(x * Float64(2.0 + Float64(x * Float64(x * 0.3333333333333333)))) / Float64(2.0 + Float64(x * x))) end
function tmp = code(x) tmp = (x * (2.0 + (x * (x * 0.3333333333333333)))) / (2.0 + (x * x)); end
code[x_] := N[(N[(x * N[(2.0 + N[(x * N[(x * 0.3333333333333333), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]), $MachinePrecision] / N[(2.0 + N[(x * x), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]
\begin{array}{l}
\\
\frac{x \cdot \left(2 + x \cdot \left(x \cdot 0.3333333333333333\right)\right)}{2 + x \cdot x}
\end{array}
Initial program 9.6%
Taylor expanded in x around 0 8.8%
unpow28.8%
Simplified8.8%
Taylor expanded in x around 0 96.2%
unpow396.2%
unpow296.2%
associate-*r*96.2%
distribute-rgt-out96.2%
*-commutative96.2%
unpow296.2%
associate-*l*96.2%
fma-def96.2%
Simplified96.2%
fma-udef96.2%
Applied egg-rr96.2%
Final simplification96.2%
(FPCore (x) :precision binary64 (/ (* 2.0 x) (+ 2.0 (* x x))))
double code(double x) {
return (2.0 * x) / (2.0 + (x * x));
}
real(8) function code(x)
real(8), intent (in) :: x
code = (2.0d0 * x) / (2.0d0 + (x * x))
end function
public static double code(double x) {
return (2.0 * x) / (2.0 + (x * x));
}
def code(x): return (2.0 * x) / (2.0 + (x * x))
function code(x) return Float64(Float64(2.0 * x) / Float64(2.0 + Float64(x * x))) end
function tmp = code(x) tmp = (2.0 * x) / (2.0 + (x * x)); end
code[x_] := N[(N[(2.0 * x), $MachinePrecision] / N[(2.0 + N[(x * x), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]
\begin{array}{l}
\\
\frac{2 \cdot x}{2 + x \cdot x}
\end{array}
Initial program 9.6%
Taylor expanded in x around 0 8.8%
unpow28.8%
Simplified8.8%
Taylor expanded in x around 0 95.7%
Final simplification95.7%
(FPCore (x) :precision binary64 0.75)
double code(double x) {
return 0.75;
}
real(8) function code(x)
real(8), intent (in) :: x
code = 0.75d0
end function
public static double code(double x) {
return 0.75;
}
def code(x): return 0.75
function code(x) return 0.75 end
function tmp = code(x) tmp = 0.75; end
code[x_] := 0.75
\begin{array}{l}
\\
0.75
\end{array}
Initial program 9.6%
Taylor expanded in x around 0 8.8%
unpow28.8%
Simplified8.8%
Applied egg-rr4.0%
Taylor expanded in x around 0 4.2%
Final simplification4.2%
(FPCore (x) :precision binary64 x)
double code(double x) {
return x;
}
real(8) function code(x)
real(8), intent (in) :: x
code = x
end function
public static double code(double x) {
return x;
}
def code(x): return x
function code(x) return x end
function tmp = code(x) tmp = x; end
code[x_] := x
\begin{array}{l}
\\
x
\end{array}
Initial program 9.6%
Taylor expanded in x around 0 95.6%
Final simplification95.6%
herbie shell --seed 2023292
(FPCore (x)
:name "Hyperbolic tangent"
:precision binary64
(/ (- (exp x) (exp (- x))) (+ (exp x) (exp (- x)))))