2sin (example 3.3)

Average Accuracy: 42.4% → 99.4%

Time: 14.7s

Precision: binary64

Cost: 45248

Math

\[\sin \left(x + \varepsilon\right) - \sin x \]

↓

\[\mathsf{fma}\left(\sin \varepsilon, \cos x, \sin x \cdot \log \left(e^{\cos \varepsilon + -1}\right)\right) \]

FPCore

(FPCore (x eps) :precision binary64 (- (sin (+ x eps)) (sin x)))

↓

(FPCore (x eps)
 :precision binary64
 (fma (sin eps) (cos x) (* (sin x) (log (exp (+ (cos eps) -1.0))))))

C

double code(double x, double eps) {
	return sin((x + eps)) - sin(x);
}

↓

double code(double x, double eps) {
	return fma(sin(eps), cos(x), (sin(x) * log(exp((cos(eps) + -1.0)))));
}

Julia

function code(x, eps)
	return Float64(sin(Float64(x + eps)) - sin(x))
end

↓

function code(x, eps)
	return fma(sin(eps), cos(x), Float64(sin(x) * log(exp(Float64(cos(eps) + -1.0)))))
end

Wolfram

code[x_, eps_] := N[(N[Sin[N[(x + eps), $MachinePrecision]], $MachinePrecision] - N[Sin[x], $MachinePrecision]), $MachinePrecision]

↓

code[x_, eps_] := N[(N[Sin[eps], $MachinePrecision] * N[Cos[x], $MachinePrecision] + N[(N[Sin[x], $MachinePrecision] * N[Log[N[Exp[N[(N[Cos[eps], $MachinePrecision] + -1.0), $MachinePrecision]], $MachinePrecision]], $MachinePrecision]), $MachinePrecision]), $MachinePrecision]

Target

Original	42.4%
Target	76.7%
Herbie	99.4%

\[2 \cdot \left(\cos \left(x + \frac{\varepsilon}{2}\right) \cdot \sin \left(\frac{\varepsilon}{2}\right)\right) \]

Derivation

1. Initial program 42.4%

   \[\sin \left(x + \varepsilon\right) - \sin x \]

2. Applied egg-rr 65.8%

   \[\leadsto \color{blue}{\sin x \cdot \cos \varepsilon + \left(\left(-\sin x\right) + \cos x \cdot \sin \varepsilon\right)} \]

3. Simplified 99.4%

   \[\leadsto \color{blue}{\mathsf{fma}\left(\sin \varepsilon, \cos x, \sin x \cdot \left(\cos \varepsilon + -1\right)\right)} \]
   Proof

   [Start] 65.8	\[ \sin x \cdot \cos \varepsilon + \left(\left(-\sin x\right) + \cos x \cdot \sin \varepsilon\right) \]
   associate-+r+ [=>] 99.4	\[ \color{blue}{\left(\sin x \cdot \cos \varepsilon + \left(-\sin x\right)\right) + \cos x \cdot \sin \varepsilon} \]
   +-commutative [=>] 99.4	\[ \color{blue}{\left(\left(-\sin x\right) + \sin x \cdot \cos \varepsilon\right)} + \cos x \cdot \sin \varepsilon \]
   +-commutative [=>] 99.4	\[ \color{blue}{\cos x \cdot \sin \varepsilon + \left(\left(-\sin x\right) + \sin x \cdot \cos \varepsilon\right)} \]
   *-commutative [=>] 99.4	\[ \color{blue}{\sin \varepsilon \cdot \cos x} + \left(\left(-\sin x\right) + \sin x \cdot \cos \varepsilon\right) \]
   fma-def [=>] 99.4	\[ \color{blue}{\mathsf{fma}\left(\sin \varepsilon, \cos x, \left(-\sin x\right) + \sin x \cdot \cos \varepsilon\right)} \]
   neg-mul-1 [=>] 99.4	\[ \mathsf{fma}\left(\sin \varepsilon, \cos x, \color{blue}{-1 \cdot \sin x} + \sin x \cdot \cos \varepsilon\right) \]
   *-commutative [=>] 99.4	\[ \mathsf{fma}\left(\sin \varepsilon, \cos x, -1 \cdot \sin x + \color{blue}{\cos \varepsilon \cdot \sin x}\right) \]
   distribute-rgt-out [=>] 99.4	\[ \mathsf{fma}\left(\sin \varepsilon, \cos x, \color{blue}{\sin x \cdot \left(-1 + \cos \varepsilon\right)}\right) \]
   +-commutative [<=] 99.4	\[ \mathsf{fma}\left(\sin \varepsilon, \cos x, \sin x \cdot \color{blue}{\left(\cos \varepsilon + -1\right)}\right) \]

4. Applied egg-rr 99.4%

   \[\leadsto \mathsf{fma}\left(\sin \varepsilon, \cos x, \sin x \cdot \color{blue}{\log \left(e^{\cos \varepsilon + -1}\right)}\right) \]

5. Final simplification 99.4%

   \[\leadsto \mathsf{fma}\left(\sin \varepsilon, \cos x, \sin x \cdot \log \left(e^{\cos \varepsilon + -1}\right)\right) \]

Alternatives

Alternative 1
Accuracy	99.4%
Cost	32448

\[\mathsf{fma}\left(\sin \varepsilon, \cos x, \sin x \cdot \left(\cos \varepsilon + -1\right)\right) \]

Alternative 2
Accuracy	99.4%
Cost	26176

\[\sin x \cdot \left(\cos \varepsilon + -1\right) + \sin \varepsilon \cdot \cos x \]

Alternative 3
Accuracy	77.9%
Cost	26048

\[\sin \varepsilon \cdot \cos x + \left(\sin x - \sin x\right) \]

Alternative 4
Accuracy	77.6%
Cost	13641

\[\begin{array}{l} \mathbf{if}\;\varepsilon \leq -0.00046 \lor \neg \left(\varepsilon \leq 0.0126\right):\\ \;\;\;\;\sin \varepsilon - \sin x\\ \mathbf{else}:\\ \;\;\;\;\varepsilon \cdot \left(\cos x + \varepsilon \cdot \left(\sin x \cdot -0.5\right)\right)\\ \end{array} \]

Alternative 5
Accuracy	76.7%
Cost	13632

\[2 \cdot \left(\sin \left(\varepsilon \cdot 0.5\right) \cdot \cos \left(0.5 \cdot \left(\varepsilon + \left(x + x\right)\right)\right)\right) \]

Alternative 6
Accuracy	77.4%
Cost	13257

\[\begin{array}{l} \mathbf{if}\;\varepsilon \leq -0.000196 \lor \neg \left(\varepsilon \leq 0.00125\right):\\ \;\;\;\;\sin \varepsilon - \sin x\\ \mathbf{else}:\\ \;\;\;\;\varepsilon \cdot \cos x\\ \end{array} \]

Alternative 7
Accuracy	76.7%
Cost	6856

\[\begin{array}{l} \mathbf{if}\;\varepsilon \leq -3.7 \cdot 10^{-5}:\\ \;\;\;\;\sin \varepsilon\\ \mathbf{elif}\;\varepsilon \leq 0.00092:\\ \;\;\;\;\varepsilon \cdot \cos x\\ \mathbf{else}:\\ \;\;\;\;\sin \varepsilon\\ \end{array} \]

Alternative 8
Accuracy	55.0%
Cost	6464

\[\sin \varepsilon \]

Alternative 9
Accuracy	4.3%
Cost	64

\[0 \]

Alternative 10
Accuracy	28.9%
Cost	64

\[\varepsilon \]

Reproduce

herbie shell --seed 2023122 
(FPCore (x eps)
  :name "2sin (example 3.3)"
  :precision binary64

  :herbie-target
  (* 2.0 (* (cos (+ x (/ eps 2.0))) (sin (/ eps 2.0))))

  (- (sin (+ x eps)) (sin x)))