
(FPCore (a b) :precision binary64 (- (+ (pow (+ (* a a) (* b b)) 2.0) (* 4.0 (+ (* (* a a) (+ 1.0 a)) (* (* b b) (- 1.0 (* 3.0 a)))))) 1.0))
double code(double a, double b) {
return (pow(((a * a) + (b * b)), 2.0) + (4.0 * (((a * a) * (1.0 + a)) + ((b * b) * (1.0 - (3.0 * a)))))) - 1.0;
}
real(8) function code(a, b)
real(8), intent (in) :: a
real(8), intent (in) :: b
code = ((((a * a) + (b * b)) ** 2.0d0) + (4.0d0 * (((a * a) * (1.0d0 + a)) + ((b * b) * (1.0d0 - (3.0d0 * a)))))) - 1.0d0
end function
public static double code(double a, double b) {
return (Math.pow(((a * a) + (b * b)), 2.0) + (4.0 * (((a * a) * (1.0 + a)) + ((b * b) * (1.0 - (3.0 * a)))))) - 1.0;
}
def code(a, b): return (math.pow(((a * a) + (b * b)), 2.0) + (4.0 * (((a * a) * (1.0 + a)) + ((b * b) * (1.0 - (3.0 * a)))))) - 1.0
function code(a, b) return Float64(Float64((Float64(Float64(a * a) + Float64(b * b)) ^ 2.0) + Float64(4.0 * Float64(Float64(Float64(a * a) * Float64(1.0 + a)) + Float64(Float64(b * b) * Float64(1.0 - Float64(3.0 * a)))))) - 1.0) end
function tmp = code(a, b) tmp = ((((a * a) + (b * b)) ^ 2.0) + (4.0 * (((a * a) * (1.0 + a)) + ((b * b) * (1.0 - (3.0 * a)))))) - 1.0; end
code[a_, b_] := N[(N[(N[Power[N[(N[(a * a), $MachinePrecision] + N[(b * b), $MachinePrecision]), $MachinePrecision], 2.0], $MachinePrecision] + N[(4.0 * N[(N[(N[(a * a), $MachinePrecision] * N[(1.0 + a), $MachinePrecision]), $MachinePrecision] + N[(N[(b * b), $MachinePrecision] * N[(1.0 - N[(3.0 * a), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]), $MachinePrecision] - 1.0), $MachinePrecision]
\begin{array}{l}
\\
\left({\left(a \cdot a + b \cdot b\right)}^{2} + 4 \cdot \left(\left(a \cdot a\right) \cdot \left(1 + a\right) + \left(b \cdot b\right) \cdot \left(1 - 3 \cdot a\right)\right)\right) - 1
\end{array}
Sampling outcomes in binary64 precision:
Herbie found 15 alternatives:
| Alternative | Accuracy | Speedup |
|---|
(FPCore (a b) :precision binary64 (- (+ (pow (+ (* a a) (* b b)) 2.0) (* 4.0 (+ (* (* a a) (+ 1.0 a)) (* (* b b) (- 1.0 (* 3.0 a)))))) 1.0))
double code(double a, double b) {
return (pow(((a * a) + (b * b)), 2.0) + (4.0 * (((a * a) * (1.0 + a)) + ((b * b) * (1.0 - (3.0 * a)))))) - 1.0;
}
real(8) function code(a, b)
real(8), intent (in) :: a
real(8), intent (in) :: b
code = ((((a * a) + (b * b)) ** 2.0d0) + (4.0d0 * (((a * a) * (1.0d0 + a)) + ((b * b) * (1.0d0 - (3.0d0 * a)))))) - 1.0d0
end function
public static double code(double a, double b) {
return (Math.pow(((a * a) + (b * b)), 2.0) + (4.0 * (((a * a) * (1.0 + a)) + ((b * b) * (1.0 - (3.0 * a)))))) - 1.0;
}
def code(a, b): return (math.pow(((a * a) + (b * b)), 2.0) + (4.0 * (((a * a) * (1.0 + a)) + ((b * b) * (1.0 - (3.0 * a)))))) - 1.0
function code(a, b) return Float64(Float64((Float64(Float64(a * a) + Float64(b * b)) ^ 2.0) + Float64(4.0 * Float64(Float64(Float64(a * a) * Float64(1.0 + a)) + Float64(Float64(b * b) * Float64(1.0 - Float64(3.0 * a)))))) - 1.0) end
function tmp = code(a, b) tmp = ((((a * a) + (b * b)) ^ 2.0) + (4.0 * (((a * a) * (1.0 + a)) + ((b * b) * (1.0 - (3.0 * a)))))) - 1.0; end
code[a_, b_] := N[(N[(N[Power[N[(N[(a * a), $MachinePrecision] + N[(b * b), $MachinePrecision]), $MachinePrecision], 2.0], $MachinePrecision] + N[(4.0 * N[(N[(N[(a * a), $MachinePrecision] * N[(1.0 + a), $MachinePrecision]), $MachinePrecision] + N[(N[(b * b), $MachinePrecision] * N[(1.0 - N[(3.0 * a), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]), $MachinePrecision] - 1.0), $MachinePrecision]
\begin{array}{l}
\\
\left({\left(a \cdot a + b \cdot b\right)}^{2} + 4 \cdot \left(\left(a \cdot a\right) \cdot \left(1 + a\right) + \left(b \cdot b\right) \cdot \left(1 - 3 \cdot a\right)\right)\right) - 1
\end{array}
(FPCore (a b) :precision binary64 (+ (+ (/ (fma a a (* b b)) (/ 1.0 (fma 1.0 (/ (* b (* a b)) a) (* a a)))) (* 4.0 (* b b))) -1.0))
double code(double a, double b) {
return ((fma(a, a, (b * b)) / (1.0 / fma(1.0, ((b * (a * b)) / a), (a * a)))) + (4.0 * (b * b))) + -1.0;
}
function code(a, b) return Float64(Float64(Float64(fma(a, a, Float64(b * b)) / Float64(1.0 / fma(1.0, Float64(Float64(b * Float64(a * b)) / a), Float64(a * a)))) + Float64(4.0 * Float64(b * b))) + -1.0) end
code[a_, b_] := N[(N[(N[(N[(a * a + N[(b * b), $MachinePrecision]), $MachinePrecision] / N[(1.0 / N[(1.0 * N[(N[(b * N[(a * b), $MachinePrecision]), $MachinePrecision] / a), $MachinePrecision] + N[(a * a), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]), $MachinePrecision] + N[(4.0 * N[(b * b), $MachinePrecision]), $MachinePrecision]), $MachinePrecision] + -1.0), $MachinePrecision]
\begin{array}{l}
\\
\left(\frac{\mathsf{fma}\left(a, a, b \cdot b\right)}{\frac{1}{\mathsf{fma}\left(1, \frac{b \cdot \left(a \cdot b\right)}{a}, a \cdot a\right)}} + 4 \cdot \left(b \cdot b\right)\right) + -1
\end{array}
Initial program 74.5%
lift-pow.f64N/A
unpow2N/A
lift-+.f64N/A
flip-+N/A
clear-numN/A
un-div-invN/A
lower-/.f64N/A
lift-+.f64N/A
lift-*.f64N/A
lower-fma.f64N/A
clear-numN/A
flip-+N/A
lift-+.f64N/A
lower-/.f6474.5
lift-+.f64N/A
lift-*.f64N/A
lower-fma.f6474.5
Applied rewrites74.5%
Taylor expanded in a around 0
unpow2N/A
lower-*.f6499.3
Applied rewrites99.3%
Taylor expanded in a around inf
+-commutativeN/A
distribute-lft-inN/A
associate-*r/N/A
unpow2N/A
associate-*l*N/A
unpow2N/A
times-fracN/A
*-rgt-identityN/A
associate-*r/N/A
rgt-mult-inverseN/A
*-rgt-identityN/A
lower-fma.f64N/A
lower-/.f64N/A
lower-*.f64N/A
unpow2N/A
lower-*.f64N/A
unpow2N/A
lower-*.f6499.4
Applied rewrites99.4%
Applied rewrites99.4%
Final simplification99.4%
(FPCore (a b) :precision binary64 (+ (+ (* 4.0 (* b b)) (/ (fma a a (* b b)) (/ 1.0 (fma 1.0 (/ (* a (* b b)) a) (* a a))))) -1.0))
double code(double a, double b) {
return ((4.0 * (b * b)) + (fma(a, a, (b * b)) / (1.0 / fma(1.0, ((a * (b * b)) / a), (a * a))))) + -1.0;
}
function code(a, b) return Float64(Float64(Float64(4.0 * Float64(b * b)) + Float64(fma(a, a, Float64(b * b)) / Float64(1.0 / fma(1.0, Float64(Float64(a * Float64(b * b)) / a), Float64(a * a))))) + -1.0) end
code[a_, b_] := N[(N[(N[(4.0 * N[(b * b), $MachinePrecision]), $MachinePrecision] + N[(N[(a * a + N[(b * b), $MachinePrecision]), $MachinePrecision] / N[(1.0 / N[(1.0 * N[(N[(a * N[(b * b), $MachinePrecision]), $MachinePrecision] / a), $MachinePrecision] + N[(a * a), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]), $MachinePrecision] + -1.0), $MachinePrecision]
\begin{array}{l}
\\
\left(4 \cdot \left(b \cdot b\right) + \frac{\mathsf{fma}\left(a, a, b \cdot b\right)}{\frac{1}{\mathsf{fma}\left(1, \frac{a \cdot \left(b \cdot b\right)}{a}, a \cdot a\right)}}\right) + -1
\end{array}
Initial program 73.5%
lift-pow.f64N/A
unpow2N/A
lift-+.f64N/A
flip-+N/A
clear-numN/A
un-div-invN/A
lower-/.f64N/A
lift-+.f64N/A
lift-*.f64N/A
lower-fma.f64N/A
clear-numN/A
flip-+N/A
lift-+.f64N/A
lower-/.f6473.5
lift-+.f64N/A
lift-*.f64N/A
lower-fma.f6473.5
Applied rewrites73.5%
Taylor expanded in a around 0
unpow2N/A
lower-*.f6499.1
Applied rewrites99.1%
Taylor expanded in a around inf
+-commutativeN/A
distribute-lft-inN/A
associate-*r/N/A
unpow2N/A
associate-*l*N/A
unpow2N/A
times-fracN/A
*-rgt-identityN/A
associate-*r/N/A
rgt-mult-inverseN/A
*-rgt-identityN/A
lower-fma.f64N/A
lower-/.f64N/A
lower-*.f64N/A
unpow2N/A
lower-*.f64N/A
unpow2N/A
lower-*.f6499.1
Applied rewrites99.1%
Final simplification99.1%
herbie shell --seed 2024226
(FPCore (a b)
:name "Bouland and Aaronson, Equation (25)"
:precision binary64
(- (+ (pow (+ (* a a) (* b b)) 2.0) (* 4.0 (+ (* (* a a) (+ 1.0 a)) (* (* b b) (- 1.0 (* 3.0 a)))))) 1.0))