
(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 10 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 (/ (PI) 2.0)) (t_1 (acos (- 1.0 x))))
(/
(fma
t_0
(* t_0 (PI))
(fma
(* t_1 (- t_1 (/ (PI) -2.0)))
(PI)
(* -2.0 (- (pow t_0 3.0) (pow t_1 3.0)))))
(*
2.0
(+ (pow (/ (cbrt (pow (PI) 3.0)) 2.0) 2.0) (* (fma 0.5 (PI) t_1) t_1))))))\begin{array}{l}
\\
\begin{array}{l}
t_0 := \frac{\mathsf{PI}\left(\right)}{2}\\
t_1 := \cos^{-1} \left(1 - x\right)\\
\frac{\mathsf{fma}\left(t\_0, t\_0 \cdot \mathsf{PI}\left(\right), \mathsf{fma}\left(t\_1 \cdot \left(t\_1 - \frac{\mathsf{PI}\left(\right)}{-2}\right), \mathsf{PI}\left(\right), -2 \cdot \left({t\_0}^{3} - {t\_1}^{3}\right)\right)\right)}{2 \cdot \left({\left(\frac{\sqrt[3]{{\mathsf{PI}\left(\right)}^{3}}}{2}\right)}^{2} + \mathsf{fma}\left(0.5, \mathsf{PI}\left(\right), t\_1\right) \cdot t\_1\right)}
\end{array}
\end{array}
Initial program 7.2%
lift-acos.f64N/A
acos-asinN/A
asin-acosN/A
lift-acos.f64N/A
flip3--N/A
frac-subN/A
lower-/.f64N/A
Applied rewrites7.2%
lift--.f64N/A
lift-*.f64N/A
fp-cancel-sub-sign-invN/A
Applied rewrites10.6%
Taylor expanded in x around 0
+-commutativeN/A
unpow2N/A
associate-*r*N/A
distribute-rgt-inN/A
metadata-evalN/A
fp-cancel-sub-sign-invN/A
*-commutativeN/A
lower-*.f64N/A
fp-cancel-sub-sign-invN/A
metadata-evalN/A
+-commutativeN/A
lower-fma.f64N/A
lower-PI.f64N/A
lower-acos.f64N/A
lower--.f64N/A
lower-acos.f64N/A
lower--.f6410.6
Applied rewrites10.6%
unpow1N/A
metadata-evalN/A
pow-powN/A
pow1/3N/A
lower-cbrt.f64N/A
lower-pow.f6410.6
Applied rewrites10.6%
(FPCore (x)
:precision binary64
(let* ((t_0 (/ (PI) 2.0)) (t_1 (acos (- 1.0 x))))
(/
(fma
t_0
(* t_0 (PI))
(fma
(* t_1 (- t_1 (/ (PI) -2.0)))
(PI)
(* -2.0 (- (pow t_0 3.0) (pow t_1 3.0)))))
(* 2.0 (+ (/ (* (PI) (PI)) 4.0) (* (fma 0.5 (PI) t_1) t_1))))))\begin{array}{l}
\\
\begin{array}{l}
t_0 := \frac{\mathsf{PI}\left(\right)}{2}\\
t_1 := \cos^{-1} \left(1 - x\right)\\
\frac{\mathsf{fma}\left(t\_0, t\_0 \cdot \mathsf{PI}\left(\right), \mathsf{fma}\left(t\_1 \cdot \left(t\_1 - \frac{\mathsf{PI}\left(\right)}{-2}\right), \mathsf{PI}\left(\right), -2 \cdot \left({t\_0}^{3} - {t\_1}^{3}\right)\right)\right)}{2 \cdot \left(\frac{\mathsf{PI}\left(\right) \cdot \mathsf{PI}\left(\right)}{4} + \mathsf{fma}\left(0.5, \mathsf{PI}\left(\right), t\_1\right) \cdot t\_1\right)}
\end{array}
\end{array}
Initial program 7.2%
lift-acos.f64N/A
acos-asinN/A
asin-acosN/A
lift-acos.f64N/A
flip3--N/A
frac-subN/A
lower-/.f64N/A
Applied rewrites7.2%
lift--.f64N/A
lift-*.f64N/A
fp-cancel-sub-sign-invN/A
Applied rewrites10.6%
Taylor expanded in x around 0
+-commutativeN/A
unpow2N/A
associate-*r*N/A
distribute-rgt-inN/A
metadata-evalN/A
fp-cancel-sub-sign-invN/A
*-commutativeN/A
lower-*.f64N/A
fp-cancel-sub-sign-invN/A
metadata-evalN/A
+-commutativeN/A
lower-fma.f64N/A
lower-PI.f64N/A
lower-acos.f64N/A
lower--.f64N/A
lower-acos.f64N/A
lower--.f6410.6
Applied rewrites10.6%
lift-pow.f64N/A
unpow2N/A
lift-/.f64N/A
lift-/.f64N/A
frac-timesN/A
lift-*.f64N/A
metadata-evalN/A
lower-/.f6410.6
Applied rewrites10.6%
(FPCore (x)
:precision binary64
(let* ((t_0 (acos (- 1.0 x)))
(t_1 (/ (PI) -2.0))
(t_2 (fma (- t_0 t_1) t_0 (pow t_1 2.0))))
(fma t_2 (/ (PI) (* t_2 2.0)) (- (asin (- 1.0 x))))))\begin{array}{l}
\\
\begin{array}{l}
t_0 := \cos^{-1} \left(1 - x\right)\\
t_1 := \frac{\mathsf{PI}\left(\right)}{-2}\\
t_2 := \mathsf{fma}\left(t\_0 - t\_1, t\_0, {t\_1}^{2}\right)\\
\mathsf{fma}\left(t\_2, \frac{\mathsf{PI}\left(\right)}{t\_2 \cdot 2}, -\sin^{-1} \left(1 - x\right)\right)
\end{array}
\end{array}
Initial program 7.2%
lift-acos.f64N/A
acos-asinN/A
asin-acosN/A
lift-acos.f64N/A
flip3--N/A
frac-subN/A
lower-/.f64N/A
Applied rewrites7.2%
lift--.f64N/A
lift-*.f64N/A
fp-cancel-sub-sign-invN/A
Applied rewrites10.6%
Applied rewrites10.6%
Final simplification10.6%
(FPCore (x)
:precision binary64
(let* ((t_0 (sqrt (PI))) (t_1 (asin (- 1.0 x))))
(/
(- (* (* (* t_0 t_0) (PI)) 0.25) (pow (sqrt t_1) 4.0))
(fma (PI) 0.5 t_1))))\begin{array}{l}
\\
\begin{array}{l}
t_0 := \sqrt{\mathsf{PI}\left(\right)}\\
t_1 := \sin^{-1} \left(1 - x\right)\\
\frac{\left(\left(t\_0 \cdot t\_0\right) \cdot \mathsf{PI}\left(\right)\right) \cdot 0.25 - {\left(\sqrt{t\_1}\right)}^{4}}{\mathsf{fma}\left(\mathsf{PI}\left(\right), 0.5, t\_1\right)}
\end{array}
\end{array}
Initial program 7.2%
lift-acos.f64N/A
acos-asinN/A
flip--N/A
lower-/.f64N/A
lower--.f64N/A
pow2N/A
lower-pow.f64N/A
lower-/.f64N/A
lower-PI.f64N/A
pow2N/A
lower-pow.f64N/A
lower-asin.f64N/A
+-commutativeN/A
lower-+.f64N/A
lower-asin.f64N/A
lower-/.f64N/A
lower-PI.f647.2
Applied rewrites7.2%
Taylor expanded in x around 0
Applied rewrites7.2%
Applied rewrites10.5%
Applied rewrites10.5%
(FPCore (x) :precision binary64 (let* ((t_0 (asin (- 1.0 x)))) (/ (- (* (* (PI) (PI)) 0.25) (pow (sqrt t_0) 4.0)) (fma (PI) 0.5 t_0))))
\begin{array}{l}
\\
\begin{array}{l}
t_0 := \sin^{-1} \left(1 - x\right)\\
\frac{\left(\mathsf{PI}\left(\right) \cdot \mathsf{PI}\left(\right)\right) \cdot 0.25 - {\left(\sqrt{t\_0}\right)}^{4}}{\mathsf{fma}\left(\mathsf{PI}\left(\right), 0.5, t\_0\right)}
\end{array}
\end{array}
Initial program 7.2%
lift-acos.f64N/A
acos-asinN/A
flip--N/A
lower-/.f64N/A
lower--.f64N/A
pow2N/A
lower-pow.f64N/A
lower-/.f64N/A
lower-PI.f64N/A
pow2N/A
lower-pow.f64N/A
lower-asin.f64N/A
+-commutativeN/A
lower-+.f64N/A
lower-asin.f64N/A
lower-/.f64N/A
lower-PI.f647.2
Applied rewrites7.2%
Taylor expanded in x around 0
Applied rewrites7.2%
Applied rewrites10.5%
(FPCore (x) :precision binary64 (- (* 0.5 (cbrt (pow (PI) 3.0))) (asin (- 1.0 x))))
\begin{array}{l}
\\
0.5 \cdot \sqrt[3]{{\mathsf{PI}\left(\right)}^{3}} - \sin^{-1} \left(1 - x\right)
\end{array}
Initial program 7.2%
lift-acos.f64N/A
acos-asinN/A
flip--N/A
lower-/.f64N/A
lower--.f64N/A
pow2N/A
lower-pow.f64N/A
lower-/.f64N/A
lower-PI.f64N/A
pow2N/A
lower-pow.f64N/A
lower-asin.f64N/A
+-commutativeN/A
lower-+.f64N/A
lower-asin.f64N/A
lower-/.f64N/A
lower-PI.f647.2
Applied rewrites7.2%
Taylor expanded in x around 0
Applied rewrites7.2%
Applied rewrites7.2%
Applied rewrites10.5%
(FPCore (x)
:precision binary64
(let* ((t_0 (sqrt (PI))))
(if (<= (acos (- 1.0 x)) 0.0)
(acos (- x))
(- (* (* 0.5 t_0) t_0) (asin (- 1.0 x))))))\begin{array}{l}
\\
\begin{array}{l}
t_0 := \sqrt{\mathsf{PI}\left(\right)}\\
\mathbf{if}\;\cos^{-1} \left(1 - x\right) \leq 0:\\
\;\;\;\;\cos^{-1} \left(-x\right)\\
\mathbf{else}:\\
\;\;\;\;\left(0.5 \cdot t\_0\right) \cdot t\_0 - \sin^{-1} \left(1 - x\right)\\
\end{array}
\end{array}
if (acos.f64 (-.f64 #s(literal 1 binary64) x)) < 0.0Initial program 3.9%
Taylor expanded in x around inf
mul-1-negN/A
lower-neg.f646.5
Applied rewrites6.5%
if 0.0 < (acos.f64 (-.f64 #s(literal 1 binary64) x)) Initial program 74.1%
lift-acos.f64N/A
acos-asinN/A
flip--N/A
lower-/.f64N/A
lower--.f64N/A
pow2N/A
lower-pow.f64N/A
lower-/.f64N/A
lower-PI.f64N/A
pow2N/A
lower-pow.f64N/A
lower-asin.f64N/A
+-commutativeN/A
lower-+.f64N/A
lower-asin.f64N/A
lower-/.f64N/A
lower-PI.f6474.0
Applied rewrites74.0%
Taylor expanded in x around 0
Applied rewrites74.0%
Applied rewrites74.1%
Applied rewrites74.3%
(FPCore (x) :precision binary64 (let* ((t_0 (acos (- 1.0 x)))) (if (<= t_0 0.0) (acos (- x)) t_0)))
double code(double x) {
double t_0 = acos((1.0 - x));
double tmp;
if (t_0 <= 0.0) {
tmp = acos(-x);
} else {
tmp = t_0;
}
return tmp;
}
real(8) function code(x)
real(8), intent (in) :: x
real(8) :: t_0
real(8) :: tmp
t_0 = acos((1.0d0 - x))
if (t_0 <= 0.0d0) then
tmp = acos(-x)
else
tmp = t_0
end if
code = tmp
end function
public static double code(double x) {
double t_0 = Math.acos((1.0 - x));
double tmp;
if (t_0 <= 0.0) {
tmp = Math.acos(-x);
} else {
tmp = t_0;
}
return tmp;
}
def code(x): t_0 = math.acos((1.0 - x)) tmp = 0 if t_0 <= 0.0: tmp = math.acos(-x) else: tmp = t_0 return tmp
function code(x) t_0 = acos(Float64(1.0 - x)) tmp = 0.0 if (t_0 <= 0.0) tmp = acos(Float64(-x)); else tmp = t_0; end return tmp end
function tmp_2 = code(x) t_0 = acos((1.0 - x)); tmp = 0.0; if (t_0 <= 0.0) tmp = acos(-x); else tmp = t_0; end tmp_2 = tmp; end
code[x_] := Block[{t$95$0 = N[ArcCos[N[(1.0 - x), $MachinePrecision]], $MachinePrecision]}, If[LessEqual[t$95$0, 0.0], N[ArcCos[(-x)], $MachinePrecision], t$95$0]]
\begin{array}{l}
\\
\begin{array}{l}
t_0 := \cos^{-1} \left(1 - x\right)\\
\mathbf{if}\;t\_0 \leq 0:\\
\;\;\;\;\cos^{-1} \left(-x\right)\\
\mathbf{else}:\\
\;\;\;\;t\_0\\
\end{array}
\end{array}
if (acos.f64 (-.f64 #s(literal 1 binary64) x)) < 0.0Initial program 3.9%
Taylor expanded in x around inf
mul-1-negN/A
lower-neg.f646.5
Applied rewrites6.5%
if 0.0 < (acos.f64 (-.f64 #s(literal 1 binary64) x)) Initial program 74.1%
(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 7.2%
Taylor expanded in x around inf
mul-1-negN/A
lower-neg.f646.9
Applied rewrites6.9%
(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 7.2%
Taylor expanded in x around 0
Applied rewrites3.8%
(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 2024343
(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)))