2atan (example 3.5)

Average Accuracy: 76.9% → 99.4%

Time: 7.0s

Precision: binary64

Cost: 7168

Math

\[\tan^{-1} \left(N + 1\right) - \tan^{-1} N \]

↓

\[\tan^{-1}_* \frac{1 + \left(N - N\right)}{1 + \left(N + N \cdot N\right)} \]

FPCore

(FPCore (N) :precision binary64 (- (atan (+ N 1.0)) (atan N)))

↓

(FPCore (N) :precision binary64 (atan2 (+ 1.0 (- N N)) (+ 1.0 (+ N (* N N)))))

C

double code(double N) {
	return atan((N + 1.0)) - atan(N);
}

↓

double code(double N) {
	return atan2((1.0 + (N - N)), (1.0 + (N + (N * N))));
}

Fortran

real(8) function code(n)
    real(8), intent (in) :: n
    code = atan((n + 1.0d0)) - atan(n)
end function

↓

real(8) function code(n)
    real(8), intent (in) :: n
    code = atan2((1.0d0 + (n - n)), (1.0d0 + (n + (n * n))))
end function

Java

public static double code(double N) {
	return Math.atan((N + 1.0)) - Math.atan(N);
}

↓

public static double code(double N) {
	return Math.atan2((1.0 + (N - N)), (1.0 + (N + (N * N))));
}

Python

def code(N):
	return math.atan((N + 1.0)) - math.atan(N)

↓

def code(N):
	return math.atan2((1.0 + (N - N)), (1.0 + (N + (N * N))))

Julia

function code(N)
	return Float64(atan(Float64(N + 1.0)) - atan(N))
end

↓

function code(N)
	return atan(Float64(1.0 + Float64(N - N)), Float64(1.0 + Float64(N + Float64(N * N))))
end

MATLAB

function tmp = code(N)
	tmp = atan((N + 1.0)) - atan(N);
end

↓

function tmp = code(N)
	tmp = atan2((1.0 + (N - N)), (1.0 + (N + (N * N))));
end

Wolfram

code[N_] := N[(N[ArcTan[N[(N + 1.0), $MachinePrecision]], $MachinePrecision] - N[ArcTan[N], $MachinePrecision]), $MachinePrecision]

↓

code[N_] := N[ArcTan[N[(1.0 + N[(N - N), $MachinePrecision]), $MachinePrecision] / N[(1.0 + N[(N + N[(N * N), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]], $MachinePrecision]

Target

Original	76.9%
Target	99.4%
Herbie	99.4%

\[\tan^{-1} \left(\frac{1}{1 + N \cdot \left(N + 1\right)}\right) \]

Derivation

1. Initial program 76.9%

   \[\tan^{-1} \left(N + 1\right) - \tan^{-1} N \]

2. Applied egg-rr 78.7%

   \[\leadsto \color{blue}{\tan^{-1}_* \frac{N + \left(1 - N\right)}{\left(N + 1\right) + N \cdot N}} \]

3. Simplified 99.4%

   \[\leadsto \color{blue}{\tan^{-1}_* \frac{1 + \left(N - N\right)}{\mathsf{fma}\left(N, N, 1 + N\right)}} \]
   Proof

   [Start] 78.7	\[ \tan^{-1}_* \frac{N + \left(1 - N\right)}{\left(N + 1\right) + N \cdot N} \]
   associate-+r- [=>] 78.7	\[ \tan^{-1}_* \frac{\color{blue}{\left(N + 1\right) - N}}{\left(N + 1\right) + N \cdot N} \]
   +-commutative [=>] 78.7	\[ \tan^{-1}_* \frac{\color{blue}{\left(1 + N\right)} - N}{\left(N + 1\right) + N \cdot N} \]
   associate--l+ [=>] 99.4	\[ \tan^{-1}_* \frac{\color{blue}{1 + \left(N - N\right)}}{\left(N + 1\right) + N \cdot N} \]
   +-commutative [=>] 99.4	\[ \tan^{-1}_* \frac{1 + \left(N - N\right)}{\color{blue}{N \cdot N + \left(N + 1\right)}} \]
   fma-def [=>] 99.4	\[ \tan^{-1}_* \frac{1 + \left(N - N\right)}{\color{blue}{\mathsf{fma}\left(N, N, N + 1\right)}} \]
   +-commutative [=>] 99.4	\[ \tan^{-1}_* \frac{1 + \left(N - N\right)}{\mathsf{fma}\left(N, N, \color{blue}{1 + N}\right)} \]

4. Applied egg-rr 99.4%

   \[\leadsto \tan^{-1}_* \frac{1 + \left(N - N\right)}{\color{blue}{\left(N \cdot N + N\right) + 1}} \]

5. Final simplification 99.4%

   \[\leadsto \tan^{-1}_* \frac{1 + \left(N - N\right)}{1 + \left(N + N \cdot N\right)} \]

Alternatives

Alternative 1
Accuracy	98.0%
Cost	6921

\[\begin{array}{l} \mathbf{if}\;N \leq -0.6 \lor \neg \left(N \leq 1.62\right):\\ \;\;\;\;\tan^{-1}_* \frac{1}{N \cdot N}\\ \mathbf{else}:\\ \;\;\;\;\tan^{-1}_* \frac{1}{1 + N}\\ \end{array} \]

Alternative 2
Accuracy	99.4%
Cost	6912

\[\tan^{-1}_* \frac{1}{N \cdot N + \left(1 + N\right)} \]

Alternative 3
Accuracy	57.2%
Cost	6656

\[\tan^{-1}_* \frac{1}{N \cdot N} \]

Reproduce

herbie shell --seed 2023141 
(FPCore (N)
  :name "2atan (example 3.5)"
  :precision binary64

  :herbie-target
  (atan (/ 1.0 (+ 1.0 (* N (+ N 1.0)))))

  (- (atan (+ N 1.0)) (atan N)))