
(FPCore (x) :precision binary64 (acos (- 1.0 x)))
double code(double x) {
return acos((1.0 - x));
}
real(8) function code(x)
real(8), intent (in) :: x
code = acos((1.0d0 - x))
end function
public static double code(double x) {
return Math.acos((1.0 - x));
}
def code(x): return math.acos((1.0 - x))
function code(x) return acos(Float64(1.0 - x)) end
function tmp = code(x) tmp = acos((1.0 - x)); end
code[x_] := N[ArcCos[N[(1.0 - x), $MachinePrecision]], $MachinePrecision]
\begin{array}{l}
\\
\cos^{-1} \left(1 - x\right)
\end{array}
Sampling outcomes in binary64 precision:
Herbie found 9 alternatives:
| Alternative | Accuracy | Speedup |
|---|
(FPCore (x) :precision binary64 (acos (- 1.0 x)))
double code(double x) {
return acos((1.0 - x));
}
real(8) function code(x)
real(8), intent (in) :: x
code = acos((1.0d0 - x))
end function
public static double code(double x) {
return Math.acos((1.0 - x));
}
def code(x): return math.acos((1.0 - x))
function code(x) return acos(Float64(1.0 - x)) end
function tmp = code(x) tmp = acos((1.0 - x)); end
code[x_] := N[ArcCos[N[(1.0 - x), $MachinePrecision]], $MachinePrecision]
\begin{array}{l}
\\
\cos^{-1} \left(1 - x\right)
\end{array}
(FPCore (x) :precision binary64 (let* ((t_0 (sqrt (PI))) (t_1 (* 0.5 t_0))) (fma (* (sqrt (* 2.0 t_0)) (sqrt t_1)) t_1 (- (asin (- 1.0 x))))))
\begin{array}{l}
\\
\begin{array}{l}
t_0 := \sqrt{\mathsf{PI}\left(\right)}\\
t_1 := 0.5 \cdot t\_0\\
\mathsf{fma}\left(\sqrt{2 \cdot t\_0} \cdot \sqrt{t\_1}, t\_1, -\sin^{-1} \left(1 - x\right)\right)
\end{array}
\end{array}
Initial program 7.1%
lift-acos.f64N/A
acos-asinN/A
sub-negN/A
div-invN/A
add-sqr-sqrtN/A
associate-*l*N/A
lower-fma.f64N/A
lower-sqrt.f64N/A
lower-PI.f64N/A
lower-*.f64N/A
lower-sqrt.f64N/A
lower-PI.f64N/A
metadata-evalN/A
lower-neg.f64N/A
lower-asin.f645.2
Applied rewrites5.2%
lift-sqrt.f64N/A
*-lft-identityN/A
*-commutativeN/A
metadata-evalN/A
associate-*l*N/A
rem-square-sqrtN/A
lift-sqrt.f64N/A
lift-sqrt.f64N/A
associate-*r*N/A
lift-*.f64N/A
*-commutativeN/A
associate-*l*N/A
sqrt-prodN/A
lower-*.f64N/A
lower-sqrt.f64N/A
lower-sqrt.f64N/A
lower-*.f6410.7
Applied rewrites10.7%
Final simplification10.7%
(FPCore (x) :precision binary64 (let* ((t_0 (sqrt (PI))) (t_1 (* 0.5 t_0))) (fma t_0 t_1 (fma t_1 (- t_0) (acos (- 1.0 x))))))
\begin{array}{l}
\\
\begin{array}{l}
t_0 := \sqrt{\mathsf{PI}\left(\right)}\\
t_1 := 0.5 \cdot t\_0\\
\mathsf{fma}\left(t\_0, t\_1, \mathsf{fma}\left(t\_1, -t\_0, \cos^{-1} \left(1 - x\right)\right)\right)
\end{array}
\end{array}
Initial program 7.1%
lift-acos.f64N/A
acos-asinN/A
sub-negN/A
div-invN/A
add-sqr-sqrtN/A
associate-*l*N/A
lower-fma.f64N/A
lower-sqrt.f64N/A
lower-PI.f64N/A
lower-*.f64N/A
lower-sqrt.f64N/A
lower-PI.f64N/A
metadata-evalN/A
lower-neg.f64N/A
lower-asin.f645.2
Applied rewrites5.2%
lift-asin.f64N/A
asin-acosN/A
lift-PI.f64N/A
div-invN/A
metadata-evalN/A
*-commutativeN/A
lift-acos.f64N/A
sub-negN/A
*-commutativeN/A
rem-square-sqrtN/A
lift-sqrt.f64N/A
lift-sqrt.f64N/A
associate-*r*N/A
lift-*.f64N/A
*-commutativeN/A
lower-fma.f64N/A
lower-neg.f6410.5
Applied rewrites10.5%
lift-neg.f64N/A
lift-fma.f64N/A
distribute-neg-inN/A
distribute-rgt-neg-inN/A
lift-neg.f64N/A
remove-double-negN/A
lower-fma.f64N/A
lift-*.f64N/A
*-commutativeN/A
lift-*.f64N/A
lower-neg.f6410.5
Applied rewrites10.5%
Final simplification10.5%
(FPCore (x) :precision binary64 (let* ((t_0 (sqrt (PI)))) (fma (* 0.5 t_0) (- t_0) (+ (* 0.5 (PI)) (acos (- 1.0 x))))))
\begin{array}{l}
\\
\begin{array}{l}
t_0 := \sqrt{\mathsf{PI}\left(\right)}\\
\mathsf{fma}\left(0.5 \cdot t\_0, -t\_0, 0.5 \cdot \mathsf{PI}\left(\right) + \cos^{-1} \left(1 - x\right)\right)
\end{array}
\end{array}
Initial program 7.1%
lift-acos.f64N/A
acos-asinN/A
sub-negN/A
div-invN/A
add-sqr-sqrtN/A
associate-*l*N/A
lower-fma.f64N/A
lower-sqrt.f64N/A
lower-PI.f64N/A
lower-*.f64N/A
lower-sqrt.f64N/A
lower-PI.f64N/A
metadata-evalN/A
lower-neg.f64N/A
lower-asin.f645.2
Applied rewrites5.2%
lift-asin.f64N/A
asin-acosN/A
lift-PI.f64N/A
div-invN/A
metadata-evalN/A
*-commutativeN/A
lift-acos.f64N/A
sub-negN/A
*-commutativeN/A
rem-square-sqrtN/A
lift-sqrt.f64N/A
lift-sqrt.f64N/A
associate-*r*N/A
lift-*.f64N/A
*-commutativeN/A
lower-fma.f64N/A
lower-neg.f6410.5
Applied rewrites10.5%
Applied rewrites10.5%
Final simplification10.5%
(FPCore (x) :precision binary64 (let* ((t_0 (sqrt (PI)))) (fma (PI) 0.5 (fma (* 0.5 t_0) (- t_0) (acos (- 1.0 x))))))
\begin{array}{l}
\\
\begin{array}{l}
t_0 := \sqrt{\mathsf{PI}\left(\right)}\\
\mathsf{fma}\left(\mathsf{PI}\left(\right), 0.5, \mathsf{fma}\left(0.5 \cdot t\_0, -t\_0, \cos^{-1} \left(1 - x\right)\right)\right)
\end{array}
\end{array}
Initial program 7.1%
lift-acos.f64N/A
acos-asinN/A
sub-negN/A
div-invN/A
lower-fma.f64N/A
lower-PI.f64N/A
metadata-evalN/A
lower-neg.f64N/A
lower-asin.f647.1
Applied rewrites7.1%
lift-neg.f64N/A
neg-sub0N/A
lift-asin.f64N/A
asin-acosN/A
lift-acos.f64N/A
lift-PI.f64N/A
div-invN/A
metadata-evalN/A
rem-cube-cbrtN/A
lift-cbrt.f64N/A
pow3N/A
unpow2N/A
lift-pow.f64N/A
associate-*r*N/A
lift-*.f64N/A
associate--r-N/A
Applied rewrites10.5%
(FPCore (x) :precision binary64 (fma (* (sqrt 2.0) (* (sqrt 0.5) 0.5)) (PI) (- (asin (- 1.0 x)))))
\begin{array}{l}
\\
\mathsf{fma}\left(\sqrt{2} \cdot \left(\sqrt{0.5} \cdot 0.5\right), \mathsf{PI}\left(\right), -\sin^{-1} \left(1 - x\right)\right)
\end{array}
Initial program 7.1%
lift-acos.f64N/A
acos-asinN/A
sub-negN/A
div-invN/A
add-sqr-sqrtN/A
associate-*l*N/A
lower-fma.f64N/A
lower-sqrt.f64N/A
lower-PI.f64N/A
lower-*.f64N/A
lower-sqrt.f64N/A
lower-PI.f64N/A
metadata-evalN/A
lower-neg.f64N/A
lower-asin.f645.2
Applied rewrites5.2%
lift-sqrt.f64N/A
*-lft-identityN/A
*-commutativeN/A
metadata-evalN/A
associate-*l*N/A
rem-square-sqrtN/A
lift-sqrt.f64N/A
lift-sqrt.f64N/A
associate-*r*N/A
lift-*.f64N/A
*-commutativeN/A
associate-*l*N/A
sqrt-prodN/A
lower-*.f64N/A
lower-sqrt.f64N/A
lower-sqrt.f64N/A
lower-*.f6410.7
Applied rewrites10.7%
Taylor expanded in x around 0
sub-negN/A
associate-*r*N/A
*-commutativeN/A
sub-negN/A
mul-1-negN/A
associate-*r*N/A
lower-fma.f64N/A
associate-*l*N/A
*-commutativeN/A
associate-*r*N/A
lower-*.f64N/A
lower-*.f64N/A
lower-sqrt.f64N/A
lower-sqrt.f64N/A
lower-PI.f64N/A
lower-neg.f64N/A
lower-asin.f64N/A
mul-1-negN/A
sub-negN/A
Applied rewrites10.4%
Final simplification10.4%
(FPCore (x) :precision binary64 (if (<= x 5.5e-17) (acos (- x)) (fma (PI) 0.5 (fma (PI) -0.5 (acos (- 1.0 x))))))
\begin{array}{l}
\\
\begin{array}{l}
\mathbf{if}\;x \leq 5.5 \cdot 10^{-17}:\\
\;\;\;\;\cos^{-1} \left(-x\right)\\
\mathbf{else}:\\
\;\;\;\;\mathsf{fma}\left(\mathsf{PI}\left(\right), 0.5, \mathsf{fma}\left(\mathsf{PI}\left(\right), -0.5, \cos^{-1} \left(1 - x\right)\right)\right)\\
\end{array}
\end{array}
if x < 5.50000000000000001e-17Initial program 3.9%
Taylor expanded in x around inf
mul-1-negN/A
lower-neg.f646.5
Applied rewrites6.5%
if 5.50000000000000001e-17 < x Initial program 72.2%
lift-acos.f64N/A
acos-asinN/A
sub-negN/A
div-invN/A
lower-fma.f64N/A
lower-PI.f64N/A
metadata-evalN/A
lower-neg.f64N/A
lower-asin.f6472.1
Applied rewrites72.1%
lift-neg.f64N/A
neg-sub0N/A
lift-asin.f64N/A
asin-acosN/A
lift-acos.f64N/A
lift-PI.f64N/A
div-invN/A
metadata-evalN/A
rem-cube-cbrtN/A
lift-cbrt.f64N/A
pow3N/A
unpow2N/A
lift-pow.f64N/A
associate-*r*N/A
lift-*.f64N/A
associate--r-N/A
Applied rewrites72.4%
(FPCore (x) :precision binary64 (if (<= x 5.5e-17) (acos (- x)) (acos (- 1.0 x))))
double code(double x) {
double tmp;
if (x <= 5.5e-17) {
tmp = acos(-x);
} else {
tmp = acos((1.0 - x));
}
return tmp;
}
real(8) function code(x)
real(8), intent (in) :: x
real(8) :: tmp
if (x <= 5.5d-17) then
tmp = acos(-x)
else
tmp = acos((1.0d0 - x))
end if
code = tmp
end function
public static double code(double x) {
double tmp;
if (x <= 5.5e-17) {
tmp = Math.acos(-x);
} else {
tmp = Math.acos((1.0 - x));
}
return tmp;
}
def code(x): tmp = 0 if x <= 5.5e-17: tmp = math.acos(-x) else: tmp = math.acos((1.0 - x)) return tmp
function code(x) tmp = 0.0 if (x <= 5.5e-17) tmp = acos(Float64(-x)); else tmp = acos(Float64(1.0 - x)); end return tmp end
function tmp_2 = code(x) tmp = 0.0; if (x <= 5.5e-17) tmp = acos(-x); else tmp = acos((1.0 - x)); end tmp_2 = tmp; end
code[x_] := If[LessEqual[x, 5.5e-17], N[ArcCos[(-x)], $MachinePrecision], N[ArcCos[N[(1.0 - x), $MachinePrecision]], $MachinePrecision]]
\begin{array}{l}
\\
\begin{array}{l}
\mathbf{if}\;x \leq 5.5 \cdot 10^{-17}:\\
\;\;\;\;\cos^{-1} \left(-x\right)\\
\mathbf{else}:\\
\;\;\;\;\cos^{-1} \left(1 - x\right)\\
\end{array}
\end{array}
if x < 5.50000000000000001e-17Initial program 3.9%
Taylor expanded in x around inf
mul-1-negN/A
lower-neg.f646.5
Applied rewrites6.5%
if 5.50000000000000001e-17 < x Initial program 72.2%
(FPCore (x) :precision binary64 (acos (- x)))
double code(double x) {
return acos(-x);
}
real(8) function code(x)
real(8), intent (in) :: x
code = acos(-x)
end function
public static double code(double x) {
return Math.acos(-x);
}
def code(x): return math.acos(-x)
function code(x) return acos(Float64(-x)) end
function tmp = code(x) tmp = acos(-x); end
code[x_] := N[ArcCos[(-x)], $MachinePrecision]
\begin{array}{l}
\\
\cos^{-1} \left(-x\right)
\end{array}
Initial program 7.1%
Taylor expanded in x around inf
mul-1-negN/A
lower-neg.f646.8
Applied rewrites6.8%
(FPCore (x) :precision binary64 (acos 1.0))
double code(double x) {
return acos(1.0);
}
real(8) function code(x)
real(8), intent (in) :: x
code = acos(1.0d0)
end function
public static double code(double x) {
return Math.acos(1.0);
}
def code(x): return math.acos(1.0)
function code(x) return acos(1.0) end
function tmp = code(x) tmp = acos(1.0); end
code[x_] := N[ArcCos[1.0], $MachinePrecision]
\begin{array}{l}
\\
\cos^{-1} 1
\end{array}
Initial program 7.1%
Taylor expanded in x around 0
Applied rewrites3.9%
(FPCore (x) :precision binary64 (* 2.0 (asin (sqrt (/ x 2.0)))))
double code(double x) {
return 2.0 * asin(sqrt((x / 2.0)));
}
real(8) function code(x)
real(8), intent (in) :: x
code = 2.0d0 * asin(sqrt((x / 2.0d0)))
end function
public static double code(double x) {
return 2.0 * Math.asin(Math.sqrt((x / 2.0)));
}
def code(x): return 2.0 * math.asin(math.sqrt((x / 2.0)))
function code(x) return Float64(2.0 * asin(sqrt(Float64(x / 2.0)))) end
function tmp = code(x) tmp = 2.0 * asin(sqrt((x / 2.0))); end
code[x_] := N[(2.0 * N[ArcSin[N[Sqrt[N[(x / 2.0), $MachinePrecision]], $MachinePrecision]], $MachinePrecision]), $MachinePrecision]
\begin{array}{l}
\\
2 \cdot \sin^{-1} \left(\sqrt{\frac{x}{2}}\right)
\end{array}
herbie shell --seed 2024249
(FPCore (x)
:name "bug323 (missed optimization)"
:precision binary64
:pre (and (<= 0.0 x) (<= x 0.5))
:alt
(! :herbie-platform default (* 2 (asin (sqrt (/ x 2)))))
(acos (- 1.0 x)))