cos2 (problem 3.4.1)

Average Error: 31.9 → 0.5

Time: 9.9s

Precision: binary64

Cost: 13768

Math

\[\frac{1 - \cos x}{x \cdot x} \]

↓

\[\begin{array}{l} t_0 := 1 - \cos x\\ \mathbf{if}\;x \leq -0.032:\\ \;\;\;\;\frac{t_0}{x \cdot x}\\ \mathbf{elif}\;x \leq 0.03:\\ \;\;\;\;0.5 + \left(-0.041666666666666664 \cdot {x}^{2} + 0.001388888888888889 \cdot {x}^{4}\right)\\ \mathbf{else}:\\ \;\;\;\;\frac{t_0}{{x}^{2}}\\ \end{array} \]

FPCore

(FPCore (x) :precision binary64 (/ (- 1.0 (cos x)) (* x x)))

↓

(FPCore (x)
 :precision binary64
 (let* ((t_0 (- 1.0 (cos x))))
   (if (<= x -0.032)
     (/ t_0 (* x x))
     (if (<= x 0.03)
       (+
        0.5
        (+
         (* -0.041666666666666664 (pow x 2.0))
         (* 0.001388888888888889 (pow x 4.0))))
       (/ t_0 (pow x 2.0))))))

C

double code(double x) {
	return (1.0 - cos(x)) / (x * x);
}

↓

double code(double x) {
	double t_0 = 1.0 - cos(x);
	double tmp;
	if (x <= -0.032) {
		tmp = t_0 / (x * x);
	} else if (x <= 0.03) {
		tmp = 0.5 + ((-0.041666666666666664 * pow(x, 2.0)) + (0.001388888888888889 * pow(x, 4.0)));
	} else {
		tmp = t_0 / pow(x, 2.0);
	}
	return tmp;
}

Fortran

real(8) function code(x)
    real(8), intent (in) :: x
    code = (1.0d0 - cos(x)) / (x * x)
end function

↓

real(8) function code(x)
    real(8), intent (in) :: x
    real(8) :: t_0
    real(8) :: tmp
    t_0 = 1.0d0 - cos(x)
    if (x <= (-0.032d0)) then
        tmp = t_0 / (x * x)
    else if (x <= 0.03d0) then
        tmp = 0.5d0 + (((-0.041666666666666664d0) * (x ** 2.0d0)) + (0.001388888888888889d0 * (x ** 4.0d0)))
    else
        tmp = t_0 / (x ** 2.0d0)
    end if
    code = tmp
end function

Java

public static double code(double x) {
	return (1.0 - Math.cos(x)) / (x * x);
}

↓

public static double code(double x) {
	double t_0 = 1.0 - Math.cos(x);
	double tmp;
	if (x <= -0.032) {
		tmp = t_0 / (x * x);
	} else if (x <= 0.03) {
		tmp = 0.5 + ((-0.041666666666666664 * Math.pow(x, 2.0)) + (0.001388888888888889 * Math.pow(x, 4.0)));
	} else {
		tmp = t_0 / Math.pow(x, 2.0);
	}
	return tmp;
}

Python

def code(x):
	return (1.0 - math.cos(x)) / (x * x)

↓

def code(x):
	t_0 = 1.0 - math.cos(x)
	tmp = 0
	if x <= -0.032:
		tmp = t_0 / (x * x)
	elif x <= 0.03:
		tmp = 0.5 + ((-0.041666666666666664 * math.pow(x, 2.0)) + (0.001388888888888889 * math.pow(x, 4.0)))
	else:
		tmp = t_0 / math.pow(x, 2.0)
	return tmp

Julia

function code(x)
	return Float64(Float64(1.0 - cos(x)) / Float64(x * x))
end

↓

function code(x)
	t_0 = Float64(1.0 - cos(x))
	tmp = 0.0
	if (x <= -0.032)
		tmp = Float64(t_0 / Float64(x * x));
	elseif (x <= 0.03)
		tmp = Float64(0.5 + Float64(Float64(-0.041666666666666664 * (x ^ 2.0)) + Float64(0.001388888888888889 * (x ^ 4.0))));
	else
		tmp = Float64(t_0 / (x ^ 2.0));
	end
	return tmp
end

MATLAB

function tmp = code(x)
	tmp = (1.0 - cos(x)) / (x * x);
end

↓

function tmp_2 = code(x)
	t_0 = 1.0 - cos(x);
	tmp = 0.0;
	if (x <= -0.032)
		tmp = t_0 / (x * x);
	elseif (x <= 0.03)
		tmp = 0.5 + ((-0.041666666666666664 * (x ^ 2.0)) + (0.001388888888888889 * (x ^ 4.0)));
	else
		tmp = t_0 / (x ^ 2.0);
	end
	tmp_2 = tmp;
end

Wolfram

code[x_] := N[(N[(1.0 - N[Cos[x], $MachinePrecision]), $MachinePrecision] / N[(x * x), $MachinePrecision]), $MachinePrecision]

↓

code[x_] := Block[{t$95$0 = N[(1.0 - N[Cos[x], $MachinePrecision]), $MachinePrecision]}, If[LessEqual[x, -0.032], N[(t$95$0 / N[(x * x), $MachinePrecision]), $MachinePrecision], If[LessEqual[x, 0.03], N[(0.5 + N[(N[(-0.041666666666666664 * N[Power[x, 2.0], $MachinePrecision]), $MachinePrecision] + N[(0.001388888888888889 * N[Power[x, 4.0], $MachinePrecision]), $MachinePrecision]), $MachinePrecision]), $MachinePrecision], N[(t$95$0 / N[Power[x, 2.0], $MachinePrecision]), $MachinePrecision]]]]

Derivation

1. Split input into 3 regimes

2. if x < -0.032000000000000001

   1. Initial program 1.0

      \[\frac{1 - \cos x}{x \cdot x} \]

   if -0.032000000000000001 < x < 0.029999999999999999

   1. Initial program 62.4

      \[\frac{1 - \cos x}{x \cdot x} \]

   2. Taylor expanded in x around 0 0.0

      \[\leadsto \color{blue}{0.5 + \left(-0.041666666666666664 \cdot {x}^{2} + 0.001388888888888889 \cdot {x}^{4}\right)} \]

   if 0.029999999999999999 < x

   1. Initial program 1.1

      \[\frac{1 - \cos x}{x \cdot x} \]

   2. Taylor expanded in x around inf 1.1

      \[\leadsto \color{blue}{\frac{1 - \cos x}{{x}^{2}}} \]
3. Recombined 3 regimes into one program.

4. Final simplification 0.5

   \[\leadsto \begin{array}{l} \mathbf{if}\;x \leq -0.032:\\ \;\;\;\;\frac{1 - \cos x}{x \cdot x}\\ \mathbf{elif}\;x \leq 0.03:\\ \;\;\;\;0.5 + \left(-0.041666666666666664 \cdot {x}^{2} + 0.001388888888888889 \cdot {x}^{4}\right)\\ \mathbf{else}:\\ \;\;\;\;\frac{1 - \cos x}{{x}^{2}}\\ \end{array} \]

Alternatives

Alternative 1
Error	0.6
Cost	13448

\[\begin{array}{l} t_0 := 1 - \cos x\\ \mathbf{if}\;x \leq -0.0055:\\ \;\;\;\;\frac{t_0}{x \cdot x}\\ \mathbf{elif}\;x \leq 0.0048:\\ \;\;\;\;0.5 + -0.041666666666666664 \cdot {x}^{2}\\ \mathbf{else}:\\ \;\;\;\;\frac{t_0}{{x}^{2}}\\ \end{array} \]

Alternative 2
Error	0.6
Cost	7112

\[\begin{array}{l} t_0 := \frac{1 - \cos x}{x \cdot x}\\ \mathbf{if}\;x \leq -0.0055:\\ \;\;\;\;t_0\\ \mathbf{elif}\;x \leq 0.0048:\\ \;\;\;\;0.5 + -0.041666666666666664 \cdot {x}^{2}\\ \mathbf{else}:\\ \;\;\;\;t_0\\ \end{array} \]

Alternative 3
Error	30.6
Cost	64

\[0.5 \]

Reproduce

herbie shell --seed 2023090 
(FPCore (x)
  :name "cos2 (problem 3.4.1)"
  :precision binary64
  (/ (- 1.0 (cos x)) (* x x)))