
(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 (cbrt (PI))) (t_1 (acos (- 1.0 x))) (t_2 (cbrt (- (sqrt (PI))))))
(/
(-
(fma 0.5 (PI) t_1)
(*
(/ 2.0 (PI))
(- (* 0.25 (* (* (pow t_0 2.0) t_0) (PI))) (pow t_1 2.0))))
(/ (/ (* 2.0 (fma (PI) 0.5 t_1)) t_0) (pow (* t_2 t_2) 2.0)))))\begin{array}{l}
\\
\begin{array}{l}
t_0 := \sqrt[3]{\mathsf{PI}\left(\right)}\\
t_1 := \cos^{-1} \left(1 - x\right)\\
t_2 := \sqrt[3]{-\sqrt{\mathsf{PI}\left(\right)}}\\
\frac{\mathsf{fma}\left(0.5, \mathsf{PI}\left(\right), t\_1\right) - \frac{2}{\mathsf{PI}\left(\right)} \cdot \left(0.25 \cdot \left(\left({t\_0}^{2} \cdot t\_0\right) \cdot \mathsf{PI}\left(\right)\right) - {t\_1}^{2}\right)}{\frac{\frac{2 \cdot \mathsf{fma}\left(\mathsf{PI}\left(\right), 0.5, t\_1\right)}{t\_0}}{{\left(t\_2 \cdot t\_2\right)}^{2}}}
\end{array}
\end{array}
Initial program 5.8%
lift-acos.f64N/A
acos-asinN/A
clear-numN/A
asin-acosN/A
lift-acos.f64N/A
flip--N/A
frac-subN/A
lower-/.f64N/A
Applied rewrites5.8%
lift-PI.f64N/A
add-cube-cbrtN/A
lower-*.f64N/A
pow2N/A
lower-pow.f64N/A
lift-PI.f64N/A
lower-cbrt.f64N/A
lift-PI.f64N/A
lower-cbrt.f649.7
Applied rewrites9.7%
lift-*.f64N/A
*-lft-identityN/A
lift-*.f64N/A
lift-/.f64N/A
associate-*l/N/A
rem-cube-cbrtN/A
lift-cbrt.f64N/A
cube-unmultN/A
unpow2N/A
lift-pow.f64N/A
associate-/r*N/A
lower-/.f64N/A
Applied rewrites9.7%
lift-cbrt.f64N/A
unpow1N/A
metadata-evalN/A
pow-powN/A
metadata-evalN/A
pow-powN/A
lift-pow.f64N/A
pow1/3N/A
lift-cbrt.f64N/A
pow2N/A
sqr-neg-revN/A
cbrt-prodN/A
lower-*.f64N/A
Applied rewrites9.7%
Final simplification9.7%
(FPCore (x)
:precision binary64
(let* ((t_0 (acos (- 1.0 x)))
(t_1 (fma 0.5 (PI) t_0))
(t_2 (sqrt (sqrt (PI)))))
(/
(-
t_1
(*
(/ 2.0 (* (* t_2 t_2) (cbrt (pow (PI) 1.5))))
(- (* 0.25 (* (PI) (PI))) (pow t_0 2.0))))
(* (/ 2.0 (PI)) t_1))))\begin{array}{l}
\\
\begin{array}{l}
t_0 := \cos^{-1} \left(1 - x\right)\\
t_1 := \mathsf{fma}\left(0.5, \mathsf{PI}\left(\right), t\_0\right)\\
t_2 := \sqrt{\sqrt{\mathsf{PI}\left(\right)}}\\
\frac{t\_1 - \frac{2}{\left(t\_2 \cdot t\_2\right) \cdot \sqrt[3]{{\mathsf{PI}\left(\right)}^{1.5}}} \cdot \left(0.25 \cdot \left(\mathsf{PI}\left(\right) \cdot \mathsf{PI}\left(\right)\right) - {t\_0}^{2}\right)}{\frac{2}{\mathsf{PI}\left(\right)} \cdot t\_1}
\end{array}
\end{array}
Initial program 5.8%
lift-acos.f64N/A
acos-asinN/A
clear-numN/A
asin-acosN/A
lift-acos.f64N/A
flip--N/A
frac-subN/A
lower-/.f64N/A
Applied rewrites5.8%
rem-cbrt-cubeN/A
lift-PI.f64N/A
add-sqr-sqrtN/A
unpow-prod-downN/A
cbrt-prodN/A
lower-*.f64N/A
lower-cbrt.f64N/A
cube-unmultN/A
lift-PI.f64N/A
pow1/2N/A
add-sqr-sqrtN/A
lift-PI.f64N/A
pow-plusN/A
metadata-evalN/A
metadata-evalN/A
lower-pow.f64N/A
metadata-evalN/A
lower-cbrt.f64N/A
cube-unmultN/A
lift-PI.f64N/A
pow1/2N/A
add-sqr-sqrtN/A
Applied rewrites9.7%
lift-cbrt.f64N/A
pow1/3N/A
lift-pow.f64N/A
pow-powN/A
metadata-evalN/A
pow1/2N/A
unpow1N/A
metadata-evalN/A
pow-powN/A
metadata-evalN/A
pow-powN/A
lift-pow.f64N/A
pow1/3N/A
lift-cbrt.f64N/A
pow2N/A
sqrt-prodN/A
lower-*.f64N/A
Applied rewrites9.7%
Final simplification9.7%
(FPCore (x)
:precision binary64
(let* ((t_0 (/ 2.0 (PI)))
(t_1 (sqrt (PI)))
(t_2 (fma 0.5 (PI) (asin (- 1.0 x)))))
(/
(-
(* (/ 2.0 (* t_1 t_1)) t_2)
(* t_0 (- t_2 (* t_0 (* t_2 (acos (- 1.0 x)))))))
(* t_0 (* t_0 t_2)))))\begin{array}{l}
\\
\begin{array}{l}
t_0 := \frac{2}{\mathsf{PI}\left(\right)}\\
t_1 := \sqrt{\mathsf{PI}\left(\right)}\\
t_2 := \mathsf{fma}\left(0.5, \mathsf{PI}\left(\right), \sin^{-1} \left(1 - x\right)\right)\\
\frac{\frac{2}{t\_1 \cdot t\_1} \cdot t\_2 - t\_0 \cdot \left(t\_2 - t\_0 \cdot \left(t\_2 \cdot \cos^{-1} \left(1 - x\right)\right)\right)}{t\_0 \cdot \left(t\_0 \cdot t\_2\right)}
\end{array}
\end{array}
Initial program 5.8%
lift-acos.f64N/A
acos-asinN/A
clear-numN/A
asin-acosN/A
clear-numN/A
acos-asinN/A
flip--N/A
frac-subN/A
frac-subN/A
lower-/.f64N/A
Applied rewrites5.8%
lift-PI.f64N/A
add-sqr-sqrtN/A
lower-*.f64N/A
lift-PI.f64N/A
lower-sqrt.f64N/A
lift-PI.f64N/A
lower-sqrt.f649.7
Applied rewrites9.7%
Final simplification9.7%
(FPCore (x)
:precision binary64
(let* ((t_0 (sqrt (PI)))
(t_1 (acos (- 1.0 x)))
(t_2 (fma 0.5 (PI) t_1))
(t_3 (/ 2.0 (PI))))
(/
(- t_2 (* t_3 (- (* 0.25 (* (* t_0 t_0) (PI))) (pow t_1 2.0))))
(* t_3 t_2))))\begin{array}{l}
\\
\begin{array}{l}
t_0 := \sqrt{\mathsf{PI}\left(\right)}\\
t_1 := \cos^{-1} \left(1 - x\right)\\
t_2 := \mathsf{fma}\left(0.5, \mathsf{PI}\left(\right), t\_1\right)\\
t_3 := \frac{2}{\mathsf{PI}\left(\right)}\\
\frac{t\_2 - t\_3 \cdot \left(0.25 \cdot \left(\left(t\_0 \cdot t\_0\right) \cdot \mathsf{PI}\left(\right)\right) - {t\_1}^{2}\right)}{t\_3 \cdot t\_2}
\end{array}
\end{array}
Initial program 5.8%
lift-acos.f64N/A
acos-asinN/A
clear-numN/A
asin-acosN/A
lift-acos.f64N/A
flip--N/A
frac-subN/A
lower-/.f64N/A
Applied rewrites5.8%
lift-PI.f64N/A
add-sqr-sqrtN/A
lower-*.f64N/A
lift-PI.f64N/A
lower-sqrt.f64N/A
lift-PI.f64N/A
lower-sqrt.f649.6
Applied rewrites9.6%
Final simplification9.6%
(FPCore (x) :precision binary64 (fma (/ 2.0 (PI)) (* (* (PI) (PI)) 0.25) (asin (+ -1.0 x))))
\begin{array}{l}
\\
\mathsf{fma}\left(\frac{2}{\mathsf{PI}\left(\right)}, \left(\mathsf{PI}\left(\right) \cdot \mathsf{PI}\left(\right)\right) \cdot 0.25, \sin^{-1} \left(-1 + x\right)\right)
\end{array}
Initial program 5.8%
lift-acos.f64N/A
acos-asinN/A
flip--N/A
div-invN/A
lower-*.f64N/A
difference-of-squaresN/A
acos-asinN/A
lift-acos.f64N/A
lower-*.f64N/A
clear-numN/A
associate-/r/N/A
lower-fma.f64N/A
metadata-evalN/A
lower-PI.f64N/A
lower-asin.f64N/A
inv-powN/A
Applied rewrites5.8%
Applied rewrites5.8%
lift-fma.f64N/A
metadata-evalN/A
div-invN/A
clear-numN/A
clear-numN/A
div-invN/A
metadata-evalN/A
lift-*.f64N/A
inv-powN/A
pow-flipN/A
metadata-evalN/A
metadata-evalN/A
pow-prod-upN/A
inv-powN/A
lift-*.f64N/A
metadata-evalN/A
div-invN/A
clear-numN/A
lift-/.f64N/A
pow2N/A
lower-fma.f64N/A
Applied rewrites9.6%
(FPCore (x) :precision binary64 (if (<= x 5.5e-17) (acos (- x)) (fma (PI) 0.5 (asin (+ -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, \sin^{-1} \left(-1 + x\right)\right)\\
\end{array}
\end{array}
if x < 5.50000000000000001e-17Initial program 3.8%
Taylor expanded in x around inf
mul-1-negN/A
lower-neg.f646.7
Applied rewrites6.7%
if 5.50000000000000001e-17 < x Initial program 54.5%
lift-acos.f64N/A
acos-asinN/A
flip--N/A
div-invN/A
lower-*.f64N/A
difference-of-squaresN/A
acos-asinN/A
lift-acos.f64N/A
lower-*.f64N/A
clear-numN/A
associate-/r/N/A
lower-fma.f64N/A
metadata-evalN/A
lower-PI.f64N/A
lower-asin.f64N/A
inv-powN/A
Applied rewrites54.6%
Applied rewrites54.6%
lift-neg.f64N/A
neg-sub0N/A
lift--.f64N/A
associate--r-N/A
metadata-evalN/A
lower-+.f6454.6
Applied rewrites54.6%
(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 3.8%
Taylor expanded in x around inf
mul-1-negN/A
lower-neg.f646.7
Applied rewrites6.7%
if 5.50000000000000001e-17 < x Initial program 54.5%
(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 5.8%
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 5.8%
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 2024305
(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)))