
(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 8 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 (asin (- 1.0 x)))
(t_1 (* 0.25 (* (PI) (PI))))
(t_2 (fma (fma 0.5 (PI) t_0) t_0 t_1))
(t_3 (/ (pow t_0 3.0) t_2))
(t_4 (/ (* (pow (PI) 3.0) 0.125) t_2)))
(/
(fma
(pow (PI) 9.0)
(pow (/ 0.125 t_2) 3.0)
(* (- (pow t_0 9.0)) (pow (fma (fma (PI) 0.5 t_0) t_0 t_1) -3.0)))
(+ (+ (* t_3 t_4) (* t_3 t_3)) (pow t_4 2.0)))))\begin{array}{l}
\\
\begin{array}{l}
t_0 := \sin^{-1} \left(1 - x\right)\\
t_1 := 0.25 \cdot \left(\mathsf{PI}\left(\right) \cdot \mathsf{PI}\left(\right)\right)\\
t_2 := \mathsf{fma}\left(\mathsf{fma}\left(0.5, \mathsf{PI}\left(\right), t\_0\right), t\_0, t\_1\right)\\
t_3 := \frac{{t\_0}^{3}}{t\_2}\\
t_4 := \frac{{\mathsf{PI}\left(\right)}^{3} \cdot 0.125}{t\_2}\\
\frac{\mathsf{fma}\left({\mathsf{PI}\left(\right)}^{9}, {\left(\frac{0.125}{t\_2}\right)}^{3}, \left(-{t\_0}^{9}\right) \cdot {\left(\mathsf{fma}\left(\mathsf{fma}\left(\mathsf{PI}\left(\right), 0.5, t\_0\right), t\_0, t\_1\right)\right)}^{-3}\right)}{\left(t\_3 \cdot t\_4 + t\_3 \cdot t\_3\right) + {t\_4}^{2}}
\end{array}
\end{array}
Initial program 6.0%
lift-acos.f64N/A
acos-asinN/A
sub-negN/A
div-invN/A
add-cube-cbrtN/A
associate-*l*N/A
lower-fma.f64N/A
pow2N/A
lower-pow.f64N/A
lower-cbrt.f64N/A
lower-PI.f64N/A
lower-*.f64N/A
lower-cbrt.f64N/A
lower-PI.f64N/A
metadata-evalN/A
lower-neg.f64N/A
lower-asin.f644.2
Applied rewrites4.2%
Applied rewrites9.3%
lift-pow.f64N/A
lift-neg.f64N/A
cube-negN/A
metadata-evalN/A
pow-powN/A
Applied rewrites9.4%
Final simplification9.4%
(FPCore (x)
:precision binary64
(let* ((t_0 (* 0.25 (* (PI) (PI))))
(t_1 (asin (- 1.0 x)))
(t_2 (fma (fma (PI) 0.5 t_1) t_1 t_0))
(t_3 (fma (fma 0.5 (PI) t_1) t_1 t_0)))
(/
(fma
(pow (PI) 9.0)
(pow (/ 0.125 t_3) 3.0)
(* (- (pow t_1 9.0)) (pow t_2 -3.0)))
(+
(fma
(pow t_1 6.0)
(pow t_2 -2.0)
(/ (pow (* (* 0.5 (PI)) t_1) 3.0) (pow t_2 2.0)))
(pow (/ (* (pow (PI) 3.0) 0.125) t_3) 2.0)))))\begin{array}{l}
\\
\begin{array}{l}
t_0 := 0.25 \cdot \left(\mathsf{PI}\left(\right) \cdot \mathsf{PI}\left(\right)\right)\\
t_1 := \sin^{-1} \left(1 - x\right)\\
t_2 := \mathsf{fma}\left(\mathsf{fma}\left(\mathsf{PI}\left(\right), 0.5, t\_1\right), t\_1, t\_0\right)\\
t_3 := \mathsf{fma}\left(\mathsf{fma}\left(0.5, \mathsf{PI}\left(\right), t\_1\right), t\_1, t\_0\right)\\
\frac{\mathsf{fma}\left({\mathsf{PI}\left(\right)}^{9}, {\left(\frac{0.125}{t\_3}\right)}^{3}, \left(-{t\_1}^{9}\right) \cdot {t\_2}^{-3}\right)}{\mathsf{fma}\left({t\_1}^{6}, {t\_2}^{-2}, \frac{{\left(\left(0.5 \cdot \mathsf{PI}\left(\right)\right) \cdot t\_1\right)}^{3}}{{t\_2}^{2}}\right) + {\left(\frac{{\mathsf{PI}\left(\right)}^{3} \cdot 0.125}{t\_3}\right)}^{2}}
\end{array}
\end{array}
Initial program 6.0%
lift-acos.f64N/A
acos-asinN/A
sub-negN/A
div-invN/A
add-cube-cbrtN/A
associate-*l*N/A
lower-fma.f64N/A
pow2N/A
lower-pow.f64N/A
lower-cbrt.f64N/A
lower-PI.f64N/A
lower-*.f64N/A
lower-cbrt.f64N/A
lower-PI.f64N/A
metadata-evalN/A
lower-neg.f64N/A
lower-asin.f644.2
Applied rewrites4.2%
Applied rewrites9.3%
lift-pow.f64N/A
lift-neg.f64N/A
cube-negN/A
metadata-evalN/A
pow-powN/A
Applied rewrites9.4%
Applied rewrites9.4%
Final simplification9.4%
(FPCore (x) :precision binary64 (let* ((t_0 (asin (- 1.0 x))) (t_1 (fma (PI) 0.5 t_0))) (fma (/ 0.25 t_1) (* (PI) (PI)) (* (pow t_1 -1.0) (- (pow t_0 2.0))))))
\begin{array}{l}
\\
\begin{array}{l}
t_0 := \sin^{-1} \left(1 - x\right)\\
t_1 := \mathsf{fma}\left(\mathsf{PI}\left(\right), 0.5, t\_0\right)\\
\mathsf{fma}\left(\frac{0.25}{t\_1}, \mathsf{PI}\left(\right) \cdot \mathsf{PI}\left(\right), {t\_1}^{-1} \cdot \left(-{t\_0}^{2}\right)\right)
\end{array}
\end{array}
Initial program 6.0%
lift-acos.f64N/A
acos-asinN/A
sub-negN/A
div-invN/A
add-cube-cbrtN/A
associate-*l*N/A
lower-fma.f64N/A
pow2N/A
lower-pow.f64N/A
lower-cbrt.f64N/A
lower-PI.f64N/A
lower-*.f64N/A
lower-cbrt.f64N/A
lower-PI.f64N/A
metadata-evalN/A
lower-neg.f64N/A
lower-asin.f644.2
Applied rewrites4.2%
Applied rewrites9.3%
lift-fma.f64N/A
lift-*.f64N/A
associate-*l*N/A
*-commutativeN/A
lower-fma.f64N/A
Applied rewrites9.3%
Final simplification9.3%
(FPCore (x) :precision binary64 (fma (pow (PI) 0.75) (* (pow (PI) 0.25) 0.5) (- (asin (- 1.0 x)))))
\begin{array}{l}
\\
\mathsf{fma}\left({\mathsf{PI}\left(\right)}^{0.75}, {\mathsf{PI}\left(\right)}^{0.25} \cdot 0.5, -\sin^{-1} \left(1 - x\right)\right)
\end{array}
Initial program 6.0%
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.f646.0
Applied rewrites6.0%
lift-asin.f64N/A
asin-acosN/A
lift-PI.f64N/A
div-invN/A
metadata-evalN/A
*-commutativeN/A
sub-negN/A
*-commutativeN/A
acos-asinN/A
lift-PI.f64N/A
div-invN/A
metadata-evalN/A
lift-asin.f64N/A
*-commutativeN/A
lower-fma.f64N/A
lower-neg.f64N/A
*-commutativeN/A
metadata-evalN/A
div-invN/A
lift-PI.f64N/A
lift-asin.f64N/A
Applied rewrites6.0%
lift-fma.f64N/A
lift-neg.f64N/A
unsub-negN/A
lift-fma.f64N/A
lift-neg.f64N/A
unsub-negN/A
metadata-evalN/A
div-invN/A
lift-PI.f64N/A
lift-acos.f64N/A
asin-acosN/A
lift-asin.f64N/A
sub-negN/A
Applied rewrites9.3%
(FPCore (x) :precision binary64 (let* ((t_0 (sqrt (PI)))) (fma (PI) 0.5 (- (fma (* t_0 0.5) 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(t\_0 \cdot 0.5, t\_0, -\cos^{-1} \left(1 - x\right)\right)\right)
\end{array}
\end{array}
Initial program 6.0%
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.f646.0
Applied rewrites6.0%
lift-asin.f64N/A
asin-acosN/A
lift-PI.f64N/A
div-invN/A
metadata-evalN/A
*-commutativeN/A
sub-negN/A
rem-square-sqrtN/A
lift-sqrt.f64N/A
lift-sqrt.f64N/A
associate-*r*N/A
acos-asinN/A
lift-PI.f64N/A
div-invN/A
metadata-evalN/A
lift-asin.f64N/A
*-commutativeN/A
lower-fma.f64N/A
Applied rewrites9.3%
Final simplification9.3%
(FPCore (x) :precision binary64 (if (<= (- 1.0 x) 0.9999999999999999) (acos (- 1.0 x)) (acos (- x))))
double code(double x) {
double tmp;
if ((1.0 - x) <= 0.9999999999999999) {
tmp = acos((1.0 - x));
} else {
tmp = acos(-x);
}
return tmp;
}
real(8) function code(x)
real(8), intent (in) :: x
real(8) :: tmp
if ((1.0d0 - x) <= 0.9999999999999999d0) then
tmp = acos((1.0d0 - x))
else
tmp = acos(-x)
end if
code = tmp
end function
public static double code(double x) {
double tmp;
if ((1.0 - x) <= 0.9999999999999999) {
tmp = Math.acos((1.0 - x));
} else {
tmp = Math.acos(-x);
}
return tmp;
}
def code(x): tmp = 0 if (1.0 - x) <= 0.9999999999999999: tmp = math.acos((1.0 - x)) else: tmp = math.acos(-x) return tmp
function code(x) tmp = 0.0 if (Float64(1.0 - x) <= 0.9999999999999999) tmp = acos(Float64(1.0 - x)); else tmp = acos(Float64(-x)); end return tmp end
function tmp_2 = code(x) tmp = 0.0; if ((1.0 - x) <= 0.9999999999999999) tmp = acos((1.0 - x)); else tmp = acos(-x); end tmp_2 = tmp; end
code[x_] := If[LessEqual[N[(1.0 - x), $MachinePrecision], 0.9999999999999999], N[ArcCos[N[(1.0 - x), $MachinePrecision]], $MachinePrecision], N[ArcCos[(-x)], $MachinePrecision]]
\begin{array}{l}
\\
\begin{array}{l}
\mathbf{if}\;1 - x \leq 0.9999999999999999:\\
\;\;\;\;\cos^{-1} \left(1 - x\right)\\
\mathbf{else}:\\
\;\;\;\;\cos^{-1} \left(-x\right)\\
\end{array}
\end{array}
if (-.f64 #s(literal 1 binary64) x) < 0.999999999999999889Initial program 49.1%
if 0.999999999999999889 < (-.f64 #s(literal 1 binary64) x) Initial program 3.9%
Taylor expanded in x around inf
mul-1-negN/A
lower-neg.f646.4
Applied rewrites6.4%
(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 6.0%
Taylor expanded in x around inf
mul-1-negN/A
lower-neg.f646.7
Applied rewrites6.7%
(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 6.0%
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 2024241
(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)))