a from scale-rotated-ellipse

Percentage Accurate: 2.6% → 51.5%
Time: 17.9s
Alternatives: 8
Speedup: 484.7×

Specification

?
\[\begin{array}{l} \\ \begin{array}{l} t_0 := \frac{angle}{180} \cdot \mathsf{PI}\left(\right)\\ t_1 := \sin t\_0\\ t_2 := \cos t\_0\\ t_3 := \frac{\frac{{\left(a \cdot t\_2\right)}^{2} + {\left(b \cdot t\_1\right)}^{2}}{y-scale}}{y-scale}\\ t_4 := \frac{\frac{{\left(a \cdot t\_1\right)}^{2} + {\left(b \cdot t\_2\right)}^{2}}{x-scale}}{x-scale}\\ t_5 := \left(b \cdot a\right) \cdot \left(b \cdot \left(-a\right)\right)\\ t_6 := \frac{4 \cdot t\_5}{{\left(x-scale \cdot y-scale\right)}^{2}}\\ \frac{-\sqrt{\left(\left(2 \cdot t\_6\right) \cdot t\_5\right) \cdot \left(\left(t\_4 + t\_3\right) + \sqrt{{\left(t\_4 - t\_3\right)}^{2} + {\left(\frac{\frac{\left(\left(2 \cdot \left({b}^{2} - {a}^{2}\right)\right) \cdot t\_1\right) \cdot t\_2}{x-scale}}{y-scale}\right)}^{2}}\right)}}{t\_6} \end{array} \end{array} \]
(FPCore (a b angle x-scale y-scale)
 :precision binary64
 (let* ((t_0 (* (/ angle 180.0) (PI)))
        (t_1 (sin t_0))
        (t_2 (cos t_0))
        (t_3
         (/ (/ (+ (pow (* a t_2) 2.0) (pow (* b t_1) 2.0)) y-scale) y-scale))
        (t_4
         (/ (/ (+ (pow (* a t_1) 2.0) (pow (* b t_2) 2.0)) x-scale) x-scale))
        (t_5 (* (* b a) (* b (- a))))
        (t_6 (/ (* 4.0 t_5) (pow (* x-scale y-scale) 2.0))))
   (/
    (-
     (sqrt
      (*
       (* (* 2.0 t_6) t_5)
       (+
        (+ t_4 t_3)
        (sqrt
         (+
          (pow (- t_4 t_3) 2.0)
          (pow
           (/
            (/ (* (* (* 2.0 (- (pow b 2.0) (pow a 2.0))) t_1) t_2) x-scale)
            y-scale)
           2.0)))))))
    t_6)))
\begin{array}{l}

\\
\begin{array}{l}
t_0 := \frac{angle}{180} \cdot \mathsf{PI}\left(\right)\\
t_1 := \sin t\_0\\
t_2 := \cos t\_0\\
t_3 := \frac{\frac{{\left(a \cdot t\_2\right)}^{2} + {\left(b \cdot t\_1\right)}^{2}}{y-scale}}{y-scale}\\
t_4 := \frac{\frac{{\left(a \cdot t\_1\right)}^{2} + {\left(b \cdot t\_2\right)}^{2}}{x-scale}}{x-scale}\\
t_5 := \left(b \cdot a\right) \cdot \left(b \cdot \left(-a\right)\right)\\
t_6 := \frac{4 \cdot t\_5}{{\left(x-scale \cdot y-scale\right)}^{2}}\\
\frac{-\sqrt{\left(\left(2 \cdot t\_6\right) \cdot t\_5\right) \cdot \left(\left(t\_4 + t\_3\right) + \sqrt{{\left(t\_4 - t\_3\right)}^{2} + {\left(\frac{\frac{\left(\left(2 \cdot \left({b}^{2} - {a}^{2}\right)\right) \cdot t\_1\right) \cdot t\_2}{x-scale}}{y-scale}\right)}^{2}}\right)}}{t\_6}
\end{array}
\end{array}

Sampling outcomes in binary64 precision:

Local Percentage Accuracy vs ?

The average percentage accuracy by input value. Horizontal axis shows value of an input variable; the variable is choosen in the title. Vertical axis is accuracy; higher is better. Red represent the original program, while blue represents Herbie's suggestion. These can be toggled with buttons below the plot. The line is an average while dots represent individual samples.

Accuracy vs Speed?

Herbie found 8 alternatives:

AlternativeAccuracySpeedup
The accuracy (vertical axis) and speed (horizontal axis) of each alternatives. Up and to the right is better. The red square shows the initial program, and each blue circle shows an alternative.The line shows the best available speed-accuracy tradeoffs.

Initial Program: 2.6% accurate, 1.0× speedup?

\[\begin{array}{l} \\ \begin{array}{l} t_0 := \frac{angle}{180} \cdot \mathsf{PI}\left(\right)\\ t_1 := \sin t\_0\\ t_2 := \cos t\_0\\ t_3 := \frac{\frac{{\left(a \cdot t\_2\right)}^{2} + {\left(b \cdot t\_1\right)}^{2}}{y-scale}}{y-scale}\\ t_4 := \frac{\frac{{\left(a \cdot t\_1\right)}^{2} + {\left(b \cdot t\_2\right)}^{2}}{x-scale}}{x-scale}\\ t_5 := \left(b \cdot a\right) \cdot \left(b \cdot \left(-a\right)\right)\\ t_6 := \frac{4 \cdot t\_5}{{\left(x-scale \cdot y-scale\right)}^{2}}\\ \frac{-\sqrt{\left(\left(2 \cdot t\_6\right) \cdot t\_5\right) \cdot \left(\left(t\_4 + t\_3\right) + \sqrt{{\left(t\_4 - t\_3\right)}^{2} + {\left(\frac{\frac{\left(\left(2 \cdot \left({b}^{2} - {a}^{2}\right)\right) \cdot t\_1\right) \cdot t\_2}{x-scale}}{y-scale}\right)}^{2}}\right)}}{t\_6} \end{array} \end{array} \]
(FPCore (a b angle x-scale y-scale)
 :precision binary64
 (let* ((t_0 (* (/ angle 180.0) (PI)))
        (t_1 (sin t_0))
        (t_2 (cos t_0))
        (t_3
         (/ (/ (+ (pow (* a t_2) 2.0) (pow (* b t_1) 2.0)) y-scale) y-scale))
        (t_4
         (/ (/ (+ (pow (* a t_1) 2.0) (pow (* b t_2) 2.0)) x-scale) x-scale))
        (t_5 (* (* b a) (* b (- a))))
        (t_6 (/ (* 4.0 t_5) (pow (* x-scale y-scale) 2.0))))
   (/
    (-
     (sqrt
      (*
       (* (* 2.0 t_6) t_5)
       (+
        (+ t_4 t_3)
        (sqrt
         (+
          (pow (- t_4 t_3) 2.0)
          (pow
           (/
            (/ (* (* (* 2.0 (- (pow b 2.0) (pow a 2.0))) t_1) t_2) x-scale)
            y-scale)
           2.0)))))))
    t_6)))
\begin{array}{l}

\\
\begin{array}{l}
t_0 := \frac{angle}{180} \cdot \mathsf{PI}\left(\right)\\
t_1 := \sin t\_0\\
t_2 := \cos t\_0\\
t_3 := \frac{\frac{{\left(a \cdot t\_2\right)}^{2} + {\left(b \cdot t\_1\right)}^{2}}{y-scale}}{y-scale}\\
t_4 := \frac{\frac{{\left(a \cdot t\_1\right)}^{2} + {\left(b \cdot t\_2\right)}^{2}}{x-scale}}{x-scale}\\
t_5 := \left(b \cdot a\right) \cdot \left(b \cdot \left(-a\right)\right)\\
t_6 := \frac{4 \cdot t\_5}{{\left(x-scale \cdot y-scale\right)}^{2}}\\
\frac{-\sqrt{\left(\left(2 \cdot t\_6\right) \cdot t\_5\right) \cdot \left(\left(t\_4 + t\_3\right) + \sqrt{{\left(t\_4 - t\_3\right)}^{2} + {\left(\frac{\frac{\left(\left(2 \cdot \left({b}^{2} - {a}^{2}\right)\right) \cdot t\_1\right) \cdot t\_2}{x-scale}}{y-scale}\right)}^{2}}\right)}}{t\_6}
\end{array}
\end{array}

Alternative 1: 51.5% accurate, 5.9× speedup?

\[\begin{array}{l} x-scale_m = \left|x-scale\right| \\ y-scale_m = \left|y-scale\right| \\ \begin{array}{l} t_0 := angle \cdot \mathsf{PI}\left(\right)\\ t_1 := \sin \left(0.005555555555555556 \cdot t\_0\right)\\ \mathbf{if}\;x-scale\_m \leq 5.8 \cdot 10^{+50}:\\ \;\;\;\;0.25 \cdot \left(\left(y-scale\_m \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, {\left(a \cdot t\_1\right)}^{2}, 2 \cdot {\left(b \cdot \sin \left(\mathsf{fma}\left(0.005555555555555556, t\_0, \frac{\mathsf{PI}\left(\right)}{2}\right)\right)\right)}^{2}\right)}\right)\\ \mathbf{else}:\\ \;\;\;\;0.25 \cdot \left(\left(x-scale\_m \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, a \cdot a, 2 \cdot {\left(b \cdot t\_1\right)}^{2}\right)}\right)\\ \end{array} \end{array} \]
x-scale_m = (fabs.f64 x-scale)
y-scale_m = (fabs.f64 y-scale)
(FPCore (a b angle x-scale_m y-scale_m)
 :precision binary64
 (let* ((t_0 (* angle (PI))) (t_1 (sin (* 0.005555555555555556 t_0))))
   (if (<= x-scale_m 5.8e+50)
     (*
      0.25
      (*
       (* y-scale_m (sqrt 8.0))
       (sqrt
        (fma
         2.0
         (pow (* a t_1) 2.0)
         (*
          2.0
          (pow
           (* b (sin (fma 0.005555555555555556 t_0 (/ (PI) 2.0))))
           2.0))))))
     (*
      0.25
      (*
       (* x-scale_m (sqrt 8.0))
       (sqrt (fma 2.0 (* a a) (* 2.0 (pow (* b t_1) 2.0)))))))))
\begin{array}{l}
x-scale_m = \left|x-scale\right|
\\
y-scale_m = \left|y-scale\right|

\\
\begin{array}{l}
t_0 := angle \cdot \mathsf{PI}\left(\right)\\
t_1 := \sin \left(0.005555555555555556 \cdot t\_0\right)\\
\mathbf{if}\;x-scale\_m \leq 5.8 \cdot 10^{+50}:\\
\;\;\;\;0.25 \cdot \left(\left(y-scale\_m \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, {\left(a \cdot t\_1\right)}^{2}, 2 \cdot {\left(b \cdot \sin \left(\mathsf{fma}\left(0.005555555555555556, t\_0, \frac{\mathsf{PI}\left(\right)}{2}\right)\right)\right)}^{2}\right)}\right)\\

\mathbf{else}:\\
\;\;\;\;0.25 \cdot \left(\left(x-scale\_m \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, a \cdot a, 2 \cdot {\left(b \cdot t\_1\right)}^{2}\right)}\right)\\


\end{array}
\end{array}
Derivation
  1. Split input into 2 regimes
  2. if x-scale < 5.8e50

    1. Initial program 4.1%

      \[\frac{-\sqrt{\left(\left(2 \cdot \frac{4 \cdot \left(\left(b \cdot a\right) \cdot \left(b \cdot \left(-a\right)\right)\right)}{{\left(x-scale \cdot y-scale\right)}^{2}}\right) \cdot \left(\left(b \cdot a\right) \cdot \left(b \cdot \left(-a\right)\right)\right)\right) \cdot \left(\left(\frac{\frac{{\left(a \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2} + {\left(b \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}}{x-scale}}{x-scale} + \frac{\frac{{\left(a \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2} + {\left(b \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}}{y-scale}}{y-scale}\right) + \sqrt{{\left(\frac{\frac{{\left(a \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2} + {\left(b \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}}{x-scale}}{x-scale} - \frac{\frac{{\left(a \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2} + {\left(b \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}}{y-scale}}{y-scale}\right)}^{2} + {\left(\frac{\frac{\left(\left(2 \cdot \left({b}^{2} - {a}^{2}\right)\right) \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right) \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)}{x-scale}}{y-scale}\right)}^{2}}\right)}}{\frac{4 \cdot \left(\left(b \cdot a\right) \cdot \left(b \cdot \left(-a\right)\right)\right)}{{\left(x-scale \cdot y-scale\right)}^{2}}} \]
    2. Add Preprocessing
    3. Taylor expanded in x-scale around 0

      \[\leadsto \color{blue}{\frac{1}{4} \cdot \left(\left(y-scale \cdot \sqrt{8}\right) \cdot \sqrt{2 \cdot \left({a}^{2} \cdot {\sin \left(\frac{1}{180} \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}\right) + 2 \cdot \left({b}^{2} \cdot {\cos \left(\frac{1}{180} \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}\right)}\right)} \]
    4. Step-by-step derivation
      1. Applied rewrites26.5%

        \[\leadsto \color{blue}{0.25 \cdot \left(\left(y-scale \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, {\left(a \cdot \sin \left(0.005555555555555556 \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}, 2 \cdot {\left(b \cdot \cos \left(0.005555555555555556 \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}\right)}\right)} \]
      2. Step-by-step derivation
        1. lift-cos.f64N/A

          \[\leadsto \frac{1}{4} \cdot \left(\left(y-scale \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, {\left(a \cdot \sin \left(\frac{1}{180} \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}, 2 \cdot {\left(b \cdot \cos \left(\frac{1}{180} \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}\right)}\right) \]
        2. lift-*.f64N/A

          \[\leadsto \frac{1}{4} \cdot \left(\left(y-scale \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, {\left(a \cdot \sin \left(\frac{1}{180} \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}, 2 \cdot {\left(b \cdot \cos \left(\frac{1}{180} \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}\right)}\right) \]
        3. lift-PI.f64N/A

          \[\leadsto \frac{1}{4} \cdot \left(\left(y-scale \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, {\left(a \cdot \sin \left(\frac{1}{180} \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}, 2 \cdot {\left(b \cdot \cos \left(\frac{1}{180} \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}\right)}\right) \]
        4. lift-*.f64N/A

          \[\leadsto \frac{1}{4} \cdot \left(\left(y-scale \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, {\left(a \cdot \sin \left(\frac{1}{180} \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}, 2 \cdot {\left(b \cdot \cos \left(\frac{1}{180} \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}\right)}\right) \]
        5. sin-+PI/2-revN/A

          \[\leadsto \frac{1}{4} \cdot \left(\left(y-scale \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, {\left(a \cdot \sin \left(\frac{1}{180} \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}, 2 \cdot {\left(b \cdot \sin \left(\frac{1}{180} \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right) + \frac{\mathsf{PI}\left(\right)}{2}\right)\right)}^{2}\right)}\right) \]
        6. lower-sin.f64N/A

          \[\leadsto \frac{1}{4} \cdot \left(\left(y-scale \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, {\left(a \cdot \sin \left(\frac{1}{180} \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}, 2 \cdot {\left(b \cdot \sin \left(\frac{1}{180} \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right) + \frac{\mathsf{PI}\left(\right)}{2}\right)\right)}^{2}\right)}\right) \]
        7. lower-fma.f64N/A

          \[\leadsto \frac{1}{4} \cdot \left(\left(y-scale \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, {\left(a \cdot \sin \left(\frac{1}{180} \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}, 2 \cdot {\left(b \cdot \sin \left(\mathsf{fma}\left(\frac{1}{180}, angle \cdot \mathsf{PI}\left(\right), \frac{\mathsf{PI}\left(\right)}{2}\right)\right)\right)}^{2}\right)}\right) \]
        8. lift-*.f64N/A

          \[\leadsto \frac{1}{4} \cdot \left(\left(y-scale \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, {\left(a \cdot \sin \left(\frac{1}{180} \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}, 2 \cdot {\left(b \cdot \sin \left(\mathsf{fma}\left(\frac{1}{180}, angle \cdot \mathsf{PI}\left(\right), \frac{\mathsf{PI}\left(\right)}{2}\right)\right)\right)}^{2}\right)}\right) \]
        9. lift-PI.f64N/A

          \[\leadsto \frac{1}{4} \cdot \left(\left(y-scale \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, {\left(a \cdot \sin \left(\frac{1}{180} \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}, 2 \cdot {\left(b \cdot \sin \left(\mathsf{fma}\left(\frac{1}{180}, angle \cdot \mathsf{PI}\left(\right), \frac{\mathsf{PI}\left(\right)}{2}\right)\right)\right)}^{2}\right)}\right) \]
        10. lower-/.f64N/A

          \[\leadsto \frac{1}{4} \cdot \left(\left(y-scale \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, {\left(a \cdot \sin \left(\frac{1}{180} \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}, 2 \cdot {\left(b \cdot \sin \left(\mathsf{fma}\left(\frac{1}{180}, angle \cdot \mathsf{PI}\left(\right), \frac{\mathsf{PI}\left(\right)}{2}\right)\right)\right)}^{2}\right)}\right) \]
        11. lift-PI.f6426.5

          \[\leadsto 0.25 \cdot \left(\left(y-scale \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, {\left(a \cdot \sin \left(0.005555555555555556 \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}, 2 \cdot {\left(b \cdot \sin \left(\mathsf{fma}\left(0.005555555555555556, angle \cdot \mathsf{PI}\left(\right), \frac{\mathsf{PI}\left(\right)}{2}\right)\right)\right)}^{2}\right)}\right) \]
      3. Applied rewrites26.5%

        \[\leadsto 0.25 \cdot \left(\left(y-scale \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, {\left(a \cdot \sin \left(0.005555555555555556 \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}, 2 \cdot {\left(b \cdot \sin \left(\mathsf{fma}\left(0.005555555555555556, angle \cdot \mathsf{PI}\left(\right), \frac{\mathsf{PI}\left(\right)}{2}\right)\right)\right)}^{2}\right)}\right) \]

      if 5.8e50 < x-scale

      1. Initial program 2.6%

        \[\frac{-\sqrt{\left(\left(2 \cdot \frac{4 \cdot \left(\left(b \cdot a\right) \cdot \left(b \cdot \left(-a\right)\right)\right)}{{\left(x-scale \cdot y-scale\right)}^{2}}\right) \cdot \left(\left(b \cdot a\right) \cdot \left(b \cdot \left(-a\right)\right)\right)\right) \cdot \left(\left(\frac{\frac{{\left(a \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2} + {\left(b \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}}{x-scale}}{x-scale} + \frac{\frac{{\left(a \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2} + {\left(b \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}}{y-scale}}{y-scale}\right) + \sqrt{{\left(\frac{\frac{{\left(a \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2} + {\left(b \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}}{x-scale}}{x-scale} - \frac{\frac{{\left(a \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2} + {\left(b \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}}{y-scale}}{y-scale}\right)}^{2} + {\left(\frac{\frac{\left(\left(2 \cdot \left({b}^{2} - {a}^{2}\right)\right) \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right) \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)}{x-scale}}{y-scale}\right)}^{2}}\right)}}{\frac{4 \cdot \left(\left(b \cdot a\right) \cdot \left(b \cdot \left(-a\right)\right)\right)}{{\left(x-scale \cdot y-scale\right)}^{2}}} \]
      2. Add Preprocessing
      3. Taylor expanded in y-scale around 0

        \[\leadsto \color{blue}{\frac{1}{4} \cdot \left(\left(x-scale \cdot \sqrt{8}\right) \cdot \sqrt{2 \cdot \left({a}^{2} \cdot {\cos \left(\frac{1}{180} \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}\right) + 2 \cdot \left({b}^{2} \cdot {\sin \left(\frac{1}{180} \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}\right)}\right)} \]
      4. Step-by-step derivation
        1. Applied rewrites75.7%

          \[\leadsto \color{blue}{0.25 \cdot \left(\left(x-scale \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, {\left(a \cdot \cos \left(0.005555555555555556 \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}, 2 \cdot {\left(b \cdot \sin \left(0.005555555555555556 \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}\right)}\right)} \]
        2. Taylor expanded in angle around 0

          \[\leadsto \frac{1}{4} \cdot \left(\left(x-scale \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, {a}^{2}, 2 \cdot {\left(b \cdot \sin \left(\frac{1}{180} \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}\right)}\right) \]
        3. Step-by-step derivation
          1. unpow2N/A

            \[\leadsto \frac{1}{4} \cdot \left(\left(x-scale \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, a \cdot a, 2 \cdot {\left(b \cdot \sin \left(\frac{1}{180} \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}\right)}\right) \]
          2. lower-*.f6477.6

            \[\leadsto 0.25 \cdot \left(\left(x-scale \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, a \cdot a, 2 \cdot {\left(b \cdot \sin \left(0.005555555555555556 \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}\right)}\right) \]
        4. Applied rewrites77.6%

          \[\leadsto 0.25 \cdot \left(\left(x-scale \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, a \cdot a, 2 \cdot {\left(b \cdot \sin \left(0.005555555555555556 \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}\right)}\right) \]
      5. Recombined 2 regimes into one program.
      6. Add Preprocessing

      Alternative 2: 51.5% accurate, 6.0× speedup?

      \[\begin{array}{l} x-scale_m = \left|x-scale\right| \\ y-scale_m = \left|y-scale\right| \\ \begin{array}{l} t_0 := 0.005555555555555556 \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\\ t_1 := \sin t\_0\\ \mathbf{if}\;x-scale\_m \leq 5.8 \cdot 10^{+50}:\\ \;\;\;\;0.25 \cdot \left(\left(y-scale\_m \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, {\left(a \cdot t\_1\right)}^{2}, 2 \cdot {\left(b \cdot \cos t\_0\right)}^{2}\right)}\right)\\ \mathbf{else}:\\ \;\;\;\;0.25 \cdot \left(\left(x-scale\_m \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, a \cdot a, 2 \cdot {\left(b \cdot t\_1\right)}^{2}\right)}\right)\\ \end{array} \end{array} \]
      x-scale_m = (fabs.f64 x-scale)
      y-scale_m = (fabs.f64 y-scale)
      (FPCore (a b angle x-scale_m y-scale_m)
       :precision binary64
       (let* ((t_0 (* 0.005555555555555556 (* angle (PI)))) (t_1 (sin t_0)))
         (if (<= x-scale_m 5.8e+50)
           (*
            0.25
            (*
             (* y-scale_m (sqrt 8.0))
             (sqrt (fma 2.0 (pow (* a t_1) 2.0) (* 2.0 (pow (* b (cos t_0)) 2.0))))))
           (*
            0.25
            (*
             (* x-scale_m (sqrt 8.0))
             (sqrt (fma 2.0 (* a a) (* 2.0 (pow (* b t_1) 2.0)))))))))
      \begin{array}{l}
      x-scale_m = \left|x-scale\right|
      \\
      y-scale_m = \left|y-scale\right|
      
      \\
      \begin{array}{l}
      t_0 := 0.005555555555555556 \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\\
      t_1 := \sin t\_0\\
      \mathbf{if}\;x-scale\_m \leq 5.8 \cdot 10^{+50}:\\
      \;\;\;\;0.25 \cdot \left(\left(y-scale\_m \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, {\left(a \cdot t\_1\right)}^{2}, 2 \cdot {\left(b \cdot \cos t\_0\right)}^{2}\right)}\right)\\
      
      \mathbf{else}:\\
      \;\;\;\;0.25 \cdot \left(\left(x-scale\_m \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, a \cdot a, 2 \cdot {\left(b \cdot t\_1\right)}^{2}\right)}\right)\\
      
      
      \end{array}
      \end{array}
      
      Derivation
      1. Split input into 2 regimes
      2. if x-scale < 5.8e50

        1. Initial program 4.1%

          \[\frac{-\sqrt{\left(\left(2 \cdot \frac{4 \cdot \left(\left(b \cdot a\right) \cdot \left(b \cdot \left(-a\right)\right)\right)}{{\left(x-scale \cdot y-scale\right)}^{2}}\right) \cdot \left(\left(b \cdot a\right) \cdot \left(b \cdot \left(-a\right)\right)\right)\right) \cdot \left(\left(\frac{\frac{{\left(a \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2} + {\left(b \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}}{x-scale}}{x-scale} + \frac{\frac{{\left(a \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2} + {\left(b \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}}{y-scale}}{y-scale}\right) + \sqrt{{\left(\frac{\frac{{\left(a \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2} + {\left(b \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}}{x-scale}}{x-scale} - \frac{\frac{{\left(a \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2} + {\left(b \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}}{y-scale}}{y-scale}\right)}^{2} + {\left(\frac{\frac{\left(\left(2 \cdot \left({b}^{2} - {a}^{2}\right)\right) \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right) \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)}{x-scale}}{y-scale}\right)}^{2}}\right)}}{\frac{4 \cdot \left(\left(b \cdot a\right) \cdot \left(b \cdot \left(-a\right)\right)\right)}{{\left(x-scale \cdot y-scale\right)}^{2}}} \]
        2. Add Preprocessing
        3. Taylor expanded in x-scale around 0

          \[\leadsto \color{blue}{\frac{1}{4} \cdot \left(\left(y-scale \cdot \sqrt{8}\right) \cdot \sqrt{2 \cdot \left({a}^{2} \cdot {\sin \left(\frac{1}{180} \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}\right) + 2 \cdot \left({b}^{2} \cdot {\cos \left(\frac{1}{180} \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}\right)}\right)} \]
        4. Step-by-step derivation
          1. Applied rewrites26.5%

            \[\leadsto \color{blue}{0.25 \cdot \left(\left(y-scale \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, {\left(a \cdot \sin \left(0.005555555555555556 \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}, 2 \cdot {\left(b \cdot \cos \left(0.005555555555555556 \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}\right)}\right)} \]

          if 5.8e50 < x-scale

          1. Initial program 2.6%

            \[\frac{-\sqrt{\left(\left(2 \cdot \frac{4 \cdot \left(\left(b \cdot a\right) \cdot \left(b \cdot \left(-a\right)\right)\right)}{{\left(x-scale \cdot y-scale\right)}^{2}}\right) \cdot \left(\left(b \cdot a\right) \cdot \left(b \cdot \left(-a\right)\right)\right)\right) \cdot \left(\left(\frac{\frac{{\left(a \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2} + {\left(b \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}}{x-scale}}{x-scale} + \frac{\frac{{\left(a \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2} + {\left(b \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}}{y-scale}}{y-scale}\right) + \sqrt{{\left(\frac{\frac{{\left(a \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2} + {\left(b \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}}{x-scale}}{x-scale} - \frac{\frac{{\left(a \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2} + {\left(b \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}}{y-scale}}{y-scale}\right)}^{2} + {\left(\frac{\frac{\left(\left(2 \cdot \left({b}^{2} - {a}^{2}\right)\right) \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right) \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)}{x-scale}}{y-scale}\right)}^{2}}\right)}}{\frac{4 \cdot \left(\left(b \cdot a\right) \cdot \left(b \cdot \left(-a\right)\right)\right)}{{\left(x-scale \cdot y-scale\right)}^{2}}} \]
          2. Add Preprocessing
          3. Taylor expanded in y-scale around 0

            \[\leadsto \color{blue}{\frac{1}{4} \cdot \left(\left(x-scale \cdot \sqrt{8}\right) \cdot \sqrt{2 \cdot \left({a}^{2} \cdot {\cos \left(\frac{1}{180} \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}\right) + 2 \cdot \left({b}^{2} \cdot {\sin \left(\frac{1}{180} \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}\right)}\right)} \]
          4. Step-by-step derivation
            1. Applied rewrites75.7%

              \[\leadsto \color{blue}{0.25 \cdot \left(\left(x-scale \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, {\left(a \cdot \cos \left(0.005555555555555556 \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}, 2 \cdot {\left(b \cdot \sin \left(0.005555555555555556 \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}\right)}\right)} \]
            2. Taylor expanded in angle around 0

              \[\leadsto \frac{1}{4} \cdot \left(\left(x-scale \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, {a}^{2}, 2 \cdot {\left(b \cdot \sin \left(\frac{1}{180} \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}\right)}\right) \]
            3. Step-by-step derivation
              1. unpow2N/A

                \[\leadsto \frac{1}{4} \cdot \left(\left(x-scale \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, a \cdot a, 2 \cdot {\left(b \cdot \sin \left(\frac{1}{180} \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}\right)}\right) \]
              2. lower-*.f6477.6

                \[\leadsto 0.25 \cdot \left(\left(x-scale \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, a \cdot a, 2 \cdot {\left(b \cdot \sin \left(0.005555555555555556 \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}\right)}\right) \]
            4. Applied rewrites77.6%

              \[\leadsto 0.25 \cdot \left(\left(x-scale \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, a \cdot a, 2 \cdot {\left(b \cdot \sin \left(0.005555555555555556 \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}\right)}\right) \]
          5. Recombined 2 regimes into one program.
          6. Add Preprocessing

          Alternative 3: 51.5% accurate, 10.6× speedup?

          \[\begin{array}{l} x-scale_m = \left|x-scale\right| \\ y-scale_m = \left|y-scale\right| \\ \begin{array}{l} t_0 := \sin \left(0.005555555555555556 \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\\ \mathbf{if}\;x-scale\_m \leq 5.8 \cdot 10^{+50}:\\ \;\;\;\;0.25 \cdot \left(\left(y-scale\_m \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, {\left(a \cdot t\_0\right)}^{2}, 2 \cdot \left(b \cdot b\right)\right)}\right)\\ \mathbf{else}:\\ \;\;\;\;0.25 \cdot \left(\left(x-scale\_m \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, a \cdot a, 2 \cdot {\left(b \cdot t\_0\right)}^{2}\right)}\right)\\ \end{array} \end{array} \]
          x-scale_m = (fabs.f64 x-scale)
          y-scale_m = (fabs.f64 y-scale)
          (FPCore (a b angle x-scale_m y-scale_m)
           :precision binary64
           (let* ((t_0 (sin (* 0.005555555555555556 (* angle (PI))))))
             (if (<= x-scale_m 5.8e+50)
               (*
                0.25
                (*
                 (* y-scale_m (sqrt 8.0))
                 (sqrt (fma 2.0 (pow (* a t_0) 2.0) (* 2.0 (* b b))))))
               (*
                0.25
                (*
                 (* x-scale_m (sqrt 8.0))
                 (sqrt (fma 2.0 (* a a) (* 2.0 (pow (* b t_0) 2.0)))))))))
          \begin{array}{l}
          x-scale_m = \left|x-scale\right|
          \\
          y-scale_m = \left|y-scale\right|
          
          \\
          \begin{array}{l}
          t_0 := \sin \left(0.005555555555555556 \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\\
          \mathbf{if}\;x-scale\_m \leq 5.8 \cdot 10^{+50}:\\
          \;\;\;\;0.25 \cdot \left(\left(y-scale\_m \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, {\left(a \cdot t\_0\right)}^{2}, 2 \cdot \left(b \cdot b\right)\right)}\right)\\
          
          \mathbf{else}:\\
          \;\;\;\;0.25 \cdot \left(\left(x-scale\_m \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, a \cdot a, 2 \cdot {\left(b \cdot t\_0\right)}^{2}\right)}\right)\\
          
          
          \end{array}
          \end{array}
          
          Derivation
          1. Split input into 2 regimes
          2. if x-scale < 5.8e50

            1. Initial program 4.1%

              \[\frac{-\sqrt{\left(\left(2 \cdot \frac{4 \cdot \left(\left(b \cdot a\right) \cdot \left(b \cdot \left(-a\right)\right)\right)}{{\left(x-scale \cdot y-scale\right)}^{2}}\right) \cdot \left(\left(b \cdot a\right) \cdot \left(b \cdot \left(-a\right)\right)\right)\right) \cdot \left(\left(\frac{\frac{{\left(a \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2} + {\left(b \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}}{x-scale}}{x-scale} + \frac{\frac{{\left(a \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2} + {\left(b \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}}{y-scale}}{y-scale}\right) + \sqrt{{\left(\frac{\frac{{\left(a \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2} + {\left(b \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}}{x-scale}}{x-scale} - \frac{\frac{{\left(a \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2} + {\left(b \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}}{y-scale}}{y-scale}\right)}^{2} + {\left(\frac{\frac{\left(\left(2 \cdot \left({b}^{2} - {a}^{2}\right)\right) \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right) \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)}{x-scale}}{y-scale}\right)}^{2}}\right)}}{\frac{4 \cdot \left(\left(b \cdot a\right) \cdot \left(b \cdot \left(-a\right)\right)\right)}{{\left(x-scale \cdot y-scale\right)}^{2}}} \]
            2. Add Preprocessing
            3. Taylor expanded in x-scale around 0

              \[\leadsto \color{blue}{\frac{1}{4} \cdot \left(\left(y-scale \cdot \sqrt{8}\right) \cdot \sqrt{2 \cdot \left({a}^{2} \cdot {\sin \left(\frac{1}{180} \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}\right) + 2 \cdot \left({b}^{2} \cdot {\cos \left(\frac{1}{180} \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}\right)}\right)} \]
            4. Step-by-step derivation
              1. Applied rewrites26.5%

                \[\leadsto \color{blue}{0.25 \cdot \left(\left(y-scale \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, {\left(a \cdot \sin \left(0.005555555555555556 \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}, 2 \cdot {\left(b \cdot \cos \left(0.005555555555555556 \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}\right)}\right)} \]
              2. Taylor expanded in angle around 0

                \[\leadsto \frac{1}{4} \cdot \left(\left(y-scale \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, {\left(a \cdot \sin \left(\frac{1}{180} \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}, 2 \cdot {b}^{2}\right)}\right) \]
              3. Step-by-step derivation
                1. pow2N/A

                  \[\leadsto \frac{1}{4} \cdot \left(\left(y-scale \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, {\left(a \cdot \sin \left(\frac{1}{180} \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}, 2 \cdot \left(b \cdot b\right)\right)}\right) \]
                2. lift-*.f6426.2

                  \[\leadsto 0.25 \cdot \left(\left(y-scale \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, {\left(a \cdot \sin \left(0.005555555555555556 \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}, 2 \cdot \left(b \cdot b\right)\right)}\right) \]
              4. Applied rewrites26.2%

                \[\leadsto 0.25 \cdot \left(\left(y-scale \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, {\left(a \cdot \sin \left(0.005555555555555556 \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}, 2 \cdot \left(b \cdot b\right)\right)}\right) \]

              if 5.8e50 < x-scale

              1. Initial program 2.6%

                \[\frac{-\sqrt{\left(\left(2 \cdot \frac{4 \cdot \left(\left(b \cdot a\right) \cdot \left(b \cdot \left(-a\right)\right)\right)}{{\left(x-scale \cdot y-scale\right)}^{2}}\right) \cdot \left(\left(b \cdot a\right) \cdot \left(b \cdot \left(-a\right)\right)\right)\right) \cdot \left(\left(\frac{\frac{{\left(a \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2} + {\left(b \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}}{x-scale}}{x-scale} + \frac{\frac{{\left(a \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2} + {\left(b \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}}{y-scale}}{y-scale}\right) + \sqrt{{\left(\frac{\frac{{\left(a \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2} + {\left(b \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}}{x-scale}}{x-scale} - \frac{\frac{{\left(a \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2} + {\left(b \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}}{y-scale}}{y-scale}\right)}^{2} + {\left(\frac{\frac{\left(\left(2 \cdot \left({b}^{2} - {a}^{2}\right)\right) \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right) \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)}{x-scale}}{y-scale}\right)}^{2}}\right)}}{\frac{4 \cdot \left(\left(b \cdot a\right) \cdot \left(b \cdot \left(-a\right)\right)\right)}{{\left(x-scale \cdot y-scale\right)}^{2}}} \]
              2. Add Preprocessing
              3. Taylor expanded in y-scale around 0

                \[\leadsto \color{blue}{\frac{1}{4} \cdot \left(\left(x-scale \cdot \sqrt{8}\right) \cdot \sqrt{2 \cdot \left({a}^{2} \cdot {\cos \left(\frac{1}{180} \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}\right) + 2 \cdot \left({b}^{2} \cdot {\sin \left(\frac{1}{180} \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}\right)}\right)} \]
              4. Step-by-step derivation
                1. Applied rewrites75.7%

                  \[\leadsto \color{blue}{0.25 \cdot \left(\left(x-scale \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, {\left(a \cdot \cos \left(0.005555555555555556 \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}, 2 \cdot {\left(b \cdot \sin \left(0.005555555555555556 \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}\right)}\right)} \]
                2. Taylor expanded in angle around 0

                  \[\leadsto \frac{1}{4} \cdot \left(\left(x-scale \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, {a}^{2}, 2 \cdot {\left(b \cdot \sin \left(\frac{1}{180} \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}\right)}\right) \]
                3. Step-by-step derivation
                  1. unpow2N/A

                    \[\leadsto \frac{1}{4} \cdot \left(\left(x-scale \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, a \cdot a, 2 \cdot {\left(b \cdot \sin \left(\frac{1}{180} \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}\right)}\right) \]
                  2. lower-*.f6477.6

                    \[\leadsto 0.25 \cdot \left(\left(x-scale \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, a \cdot a, 2 \cdot {\left(b \cdot \sin \left(0.005555555555555556 \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}\right)}\right) \]
                4. Applied rewrites77.6%

                  \[\leadsto 0.25 \cdot \left(\left(x-scale \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, a \cdot a, 2 \cdot {\left(b \cdot \sin \left(0.005555555555555556 \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}\right)}\right) \]
              5. Recombined 2 regimes into one program.
              6. Add Preprocessing

              Alternative 4: 51.6% accurate, 10.6× speedup?

              \[\begin{array}{l} x-scale_m = \left|x-scale\right| \\ y-scale_m = \left|y-scale\right| \\ \begin{array}{l} t_0 := 0.005555555555555556 \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\\ \mathbf{if}\;y-scale\_m \leq 2.9 \cdot 10^{+42}:\\ \;\;\;\;0.25 \cdot \left(\left(x-scale\_m \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, a \cdot a, 2 \cdot {\left(b \cdot \sin t\_0\right)}^{2}\right)}\right)\\ \mathbf{else}:\\ \;\;\;\;0.25 \cdot \left(\left(y-scale\_m \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, {\left(a \cdot t\_0\right)}^{2}, 2 \cdot \left(b \cdot b\right)\right)}\right)\\ \end{array} \end{array} \]
              x-scale_m = (fabs.f64 x-scale)
              y-scale_m = (fabs.f64 y-scale)
              (FPCore (a b angle x-scale_m y-scale_m)
               :precision binary64
               (let* ((t_0 (* 0.005555555555555556 (* angle (PI)))))
                 (if (<= y-scale_m 2.9e+42)
                   (*
                    0.25
                    (*
                     (* x-scale_m (sqrt 8.0))
                     (sqrt (fma 2.0 (* a a) (* 2.0 (pow (* b (sin t_0)) 2.0))))))
                   (*
                    0.25
                    (*
                     (* y-scale_m (sqrt 8.0))
                     (sqrt (fma 2.0 (pow (* a t_0) 2.0) (* 2.0 (* b b)))))))))
              \begin{array}{l}
              x-scale_m = \left|x-scale\right|
              \\
              y-scale_m = \left|y-scale\right|
              
              \\
              \begin{array}{l}
              t_0 := 0.005555555555555556 \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\\
              \mathbf{if}\;y-scale\_m \leq 2.9 \cdot 10^{+42}:\\
              \;\;\;\;0.25 \cdot \left(\left(x-scale\_m \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, a \cdot a, 2 \cdot {\left(b \cdot \sin t\_0\right)}^{2}\right)}\right)\\
              
              \mathbf{else}:\\
              \;\;\;\;0.25 \cdot \left(\left(y-scale\_m \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, {\left(a \cdot t\_0\right)}^{2}, 2 \cdot \left(b \cdot b\right)\right)}\right)\\
              
              
              \end{array}
              \end{array}
              
              Derivation
              1. Split input into 2 regimes
              2. if y-scale < 2.89999999999999981e42

                1. Initial program 2.9%

                  \[\frac{-\sqrt{\left(\left(2 \cdot \frac{4 \cdot \left(\left(b \cdot a\right) \cdot \left(b \cdot \left(-a\right)\right)\right)}{{\left(x-scale \cdot y-scale\right)}^{2}}\right) \cdot \left(\left(b \cdot a\right) \cdot \left(b \cdot \left(-a\right)\right)\right)\right) \cdot \left(\left(\frac{\frac{{\left(a \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2} + {\left(b \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}}{x-scale}}{x-scale} + \frac{\frac{{\left(a \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2} + {\left(b \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}}{y-scale}}{y-scale}\right) + \sqrt{{\left(\frac{\frac{{\left(a \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2} + {\left(b \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}}{x-scale}}{x-scale} - \frac{\frac{{\left(a \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2} + {\left(b \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}}{y-scale}}{y-scale}\right)}^{2} + {\left(\frac{\frac{\left(\left(2 \cdot \left({b}^{2} - {a}^{2}\right)\right) \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right) \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)}{x-scale}}{y-scale}\right)}^{2}}\right)}}{\frac{4 \cdot \left(\left(b \cdot a\right) \cdot \left(b \cdot \left(-a\right)\right)\right)}{{\left(x-scale \cdot y-scale\right)}^{2}}} \]
                2. Add Preprocessing
                3. Taylor expanded in y-scale around 0

                  \[\leadsto \color{blue}{\frac{1}{4} \cdot \left(\left(x-scale \cdot \sqrt{8}\right) \cdot \sqrt{2 \cdot \left({a}^{2} \cdot {\cos \left(\frac{1}{180} \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}\right) + 2 \cdot \left({b}^{2} \cdot {\sin \left(\frac{1}{180} \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}\right)}\right)} \]
                4. Step-by-step derivation
                  1. Applied rewrites23.5%

                    \[\leadsto \color{blue}{0.25 \cdot \left(\left(x-scale \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, {\left(a \cdot \cos \left(0.005555555555555556 \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}, 2 \cdot {\left(b \cdot \sin \left(0.005555555555555556 \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}\right)}\right)} \]
                  2. Taylor expanded in angle around 0

                    \[\leadsto \frac{1}{4} \cdot \left(\left(x-scale \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, {a}^{2}, 2 \cdot {\left(b \cdot \sin \left(\frac{1}{180} \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}\right)}\right) \]
                  3. Step-by-step derivation
                    1. unpow2N/A

                      \[\leadsto \frac{1}{4} \cdot \left(\left(x-scale \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, a \cdot a, 2 \cdot {\left(b \cdot \sin \left(\frac{1}{180} \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}\right)}\right) \]
                    2. lower-*.f6424.1

                      \[\leadsto 0.25 \cdot \left(\left(x-scale \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, a \cdot a, 2 \cdot {\left(b \cdot \sin \left(0.005555555555555556 \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}\right)}\right) \]
                  4. Applied rewrites24.1%

                    \[\leadsto 0.25 \cdot \left(\left(x-scale \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, a \cdot a, 2 \cdot {\left(b \cdot \sin \left(0.005555555555555556 \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}\right)}\right) \]

                  if 2.89999999999999981e42 < y-scale

                  1. Initial program 7.0%

                    \[\frac{-\sqrt{\left(\left(2 \cdot \frac{4 \cdot \left(\left(b \cdot a\right) \cdot \left(b \cdot \left(-a\right)\right)\right)}{{\left(x-scale \cdot y-scale\right)}^{2}}\right) \cdot \left(\left(b \cdot a\right) \cdot \left(b \cdot \left(-a\right)\right)\right)\right) \cdot \left(\left(\frac{\frac{{\left(a \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2} + {\left(b \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}}{x-scale}}{x-scale} + \frac{\frac{{\left(a \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2} + {\left(b \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}}{y-scale}}{y-scale}\right) + \sqrt{{\left(\frac{\frac{{\left(a \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2} + {\left(b \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}}{x-scale}}{x-scale} - \frac{\frac{{\left(a \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2} + {\left(b \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}}{y-scale}}{y-scale}\right)}^{2} + {\left(\frac{\frac{\left(\left(2 \cdot \left({b}^{2} - {a}^{2}\right)\right) \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right) \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)}{x-scale}}{y-scale}\right)}^{2}}\right)}}{\frac{4 \cdot \left(\left(b \cdot a\right) \cdot \left(b \cdot \left(-a\right)\right)\right)}{{\left(x-scale \cdot y-scale\right)}^{2}}} \]
                  2. Add Preprocessing
                  3. Taylor expanded in x-scale around 0

                    \[\leadsto \color{blue}{\frac{1}{4} \cdot \left(\left(y-scale \cdot \sqrt{8}\right) \cdot \sqrt{2 \cdot \left({a}^{2} \cdot {\sin \left(\frac{1}{180} \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}\right) + 2 \cdot \left({b}^{2} \cdot {\cos \left(\frac{1}{180} \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}\right)}\right)} \]
                  4. Step-by-step derivation
                    1. Applied rewrites76.6%

                      \[\leadsto \color{blue}{0.25 \cdot \left(\left(y-scale \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, {\left(a \cdot \sin \left(0.005555555555555556 \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}, 2 \cdot {\left(b \cdot \cos \left(0.005555555555555556 \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}\right)}\right)} \]
                    2. Taylor expanded in angle around 0

                      \[\leadsto \frac{1}{4} \cdot \left(\left(y-scale \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, {\left(a \cdot \sin \left(\frac{1}{180} \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}, 2 \cdot {b}^{2}\right)}\right) \]
                    3. Step-by-step derivation
                      1. pow2N/A

                        \[\leadsto \frac{1}{4} \cdot \left(\left(y-scale \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, {\left(a \cdot \sin \left(\frac{1}{180} \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}, 2 \cdot \left(b \cdot b\right)\right)}\right) \]
                      2. lift-*.f6476.6

                        \[\leadsto 0.25 \cdot \left(\left(y-scale \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, {\left(a \cdot \sin \left(0.005555555555555556 \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}, 2 \cdot \left(b \cdot b\right)\right)}\right) \]
                    4. Applied rewrites76.6%

                      \[\leadsto 0.25 \cdot \left(\left(y-scale \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, {\left(a \cdot \sin \left(0.005555555555555556 \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}, 2 \cdot \left(b \cdot b\right)\right)}\right) \]
                    5. Taylor expanded in angle around 0

                      \[\leadsto \frac{1}{4} \cdot \left(\left(y-scale \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, {\left(a \cdot \left(\frac{1}{180} \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}, 2 \cdot \left(b \cdot b\right)\right)}\right) \]
                    6. Step-by-step derivation
                      1. lift-*.f64N/A

                        \[\leadsto \frac{1}{4} \cdot \left(\left(y-scale \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, {\left(a \cdot \left(\frac{1}{180} \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}, 2 \cdot \left(b \cdot b\right)\right)}\right) \]
                      2. lift-PI.f64N/A

                        \[\leadsto \frac{1}{4} \cdot \left(\left(y-scale \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, {\left(a \cdot \left(\frac{1}{180} \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}, 2 \cdot \left(b \cdot b\right)\right)}\right) \]
                      3. lift-*.f6474.4

                        \[\leadsto 0.25 \cdot \left(\left(y-scale \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, {\left(a \cdot \left(0.005555555555555556 \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}, 2 \cdot \left(b \cdot b\right)\right)}\right) \]
                    7. Applied rewrites74.4%

                      \[\leadsto 0.25 \cdot \left(\left(y-scale \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, {\left(a \cdot \left(0.005555555555555556 \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}, 2 \cdot \left(b \cdot b\right)\right)}\right) \]
                  5. Recombined 2 regimes into one program.
                  6. Add Preprocessing

                  Alternative 5: 42.0% accurate, 16.7× speedup?

                  \[\begin{array}{l} x-scale_m = \left|x-scale\right| \\ y-scale_m = \left|y-scale\right| \\ \begin{array}{l} \mathbf{if}\;y-scale\_m \leq 2.1 \cdot 10^{-40}:\\ \;\;\;\;0.25 \cdot \left(a \cdot \left(x-scale\_m \cdot 4\right)\right)\\ \mathbf{else}:\\ \;\;\;\;0.25 \cdot \left(\left(y-scale\_m \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, {\left(a \cdot \left(0.005555555555555556 \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}, 2 \cdot \left(b \cdot b\right)\right)}\right)\\ \end{array} \end{array} \]
                  x-scale_m = (fabs.f64 x-scale)
                  y-scale_m = (fabs.f64 y-scale)
                  (FPCore (a b angle x-scale_m y-scale_m)
                   :precision binary64
                   (if (<= y-scale_m 2.1e-40)
                     (* 0.25 (* a (* x-scale_m 4.0)))
                     (*
                      0.25
                      (*
                       (* y-scale_m (sqrt 8.0))
                       (sqrt
                        (fma
                         2.0
                         (pow (* a (* 0.005555555555555556 (* angle (PI)))) 2.0)
                         (* 2.0 (* b b))))))))
                  \begin{array}{l}
                  x-scale_m = \left|x-scale\right|
                  \\
                  y-scale_m = \left|y-scale\right|
                  
                  \\
                  \begin{array}{l}
                  \mathbf{if}\;y-scale\_m \leq 2.1 \cdot 10^{-40}:\\
                  \;\;\;\;0.25 \cdot \left(a \cdot \left(x-scale\_m \cdot 4\right)\right)\\
                  
                  \mathbf{else}:\\
                  \;\;\;\;0.25 \cdot \left(\left(y-scale\_m \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, {\left(a \cdot \left(0.005555555555555556 \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}, 2 \cdot \left(b \cdot b\right)\right)}\right)\\
                  
                  
                  \end{array}
                  \end{array}
                  
                  Derivation
                  1. Split input into 2 regimes
                  2. if y-scale < 2.10000000000000018e-40

                    1. Initial program 3.2%

                      \[\frac{-\sqrt{\left(\left(2 \cdot \frac{4 \cdot \left(\left(b \cdot a\right) \cdot \left(b \cdot \left(-a\right)\right)\right)}{{\left(x-scale \cdot y-scale\right)}^{2}}\right) \cdot \left(\left(b \cdot a\right) \cdot \left(b \cdot \left(-a\right)\right)\right)\right) \cdot \left(\left(\frac{\frac{{\left(a \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2} + {\left(b \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}}{x-scale}}{x-scale} + \frac{\frac{{\left(a \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2} + {\left(b \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}}{y-scale}}{y-scale}\right) + \sqrt{{\left(\frac{\frac{{\left(a \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2} + {\left(b \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}}{x-scale}}{x-scale} - \frac{\frac{{\left(a \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2} + {\left(b \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}}{y-scale}}{y-scale}\right)}^{2} + {\left(\frac{\frac{\left(\left(2 \cdot \left({b}^{2} - {a}^{2}\right)\right) \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right) \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)}{x-scale}}{y-scale}\right)}^{2}}\right)}}{\frac{4 \cdot \left(\left(b \cdot a\right) \cdot \left(b \cdot \left(-a\right)\right)\right)}{{\left(x-scale \cdot y-scale\right)}^{2}}} \]
                    2. Add Preprocessing
                    3. Taylor expanded in x-scale around inf

                      \[\leadsto \color{blue}{\frac{1}{4} \cdot \left(\left(x-scale \cdot \left(y-scale \cdot \sqrt{8}\right)\right) \cdot \sqrt{2 \cdot \frac{{a}^{2} \cdot {\cos \left(\frac{1}{180} \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}}{{y-scale}^{2}} + 2 \cdot \frac{{b}^{2} \cdot {\sin \left(\frac{1}{180} \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}}{{y-scale}^{2}}}\right)} \]
                    4. Applied rewrites12.7%

                      \[\leadsto \color{blue}{0.25 \cdot \left(\left(x-scale \cdot \left(y-scale \cdot \sqrt{8}\right)\right) \cdot \sqrt{\mathsf{fma}\left(2, \frac{{\left(a \cdot \cos \left(0.005555555555555556 \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}}{y-scale \cdot y-scale}, 2 \cdot \frac{{\left(b \cdot \sin \left(0.005555555555555556 \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}}{y-scale \cdot y-scale}\right)}\right)} \]
                    5. Taylor expanded in angle around 0

                      \[\leadsto \frac{1}{4} \cdot \left(a \cdot \color{blue}{\left(x-scale \cdot \left(\sqrt{2} \cdot \sqrt{8}\right)\right)}\right) \]
                    6. Step-by-step derivation
                      1. lower-*.f64N/A

                        \[\leadsto \frac{1}{4} \cdot \left(a \cdot \left(x-scale \cdot \color{blue}{\left(\sqrt{2} \cdot \sqrt{8}\right)}\right)\right) \]
                      2. sqrt-unprodN/A

                        \[\leadsto \frac{1}{4} \cdot \left(a \cdot \left(x-scale \cdot \sqrt{2 \cdot 8}\right)\right) \]
                      3. metadata-evalN/A

                        \[\leadsto \frac{1}{4} \cdot \left(a \cdot \left(x-scale \cdot \sqrt{16}\right)\right) \]
                      4. metadata-evalN/A

                        \[\leadsto \frac{1}{4} \cdot \left(a \cdot \left(x-scale \cdot 4\right)\right) \]
                      5. lower-*.f6415.9

                        \[\leadsto 0.25 \cdot \left(a \cdot \left(x-scale \cdot 4\right)\right) \]
                    7. Applied rewrites15.9%

                      \[\leadsto 0.25 \cdot \left(a \cdot \color{blue}{\left(x-scale \cdot 4\right)}\right) \]

                    if 2.10000000000000018e-40 < y-scale

                    1. Initial program 5.3%

                      \[\frac{-\sqrt{\left(\left(2 \cdot \frac{4 \cdot \left(\left(b \cdot a\right) \cdot \left(b \cdot \left(-a\right)\right)\right)}{{\left(x-scale \cdot y-scale\right)}^{2}}\right) \cdot \left(\left(b \cdot a\right) \cdot \left(b \cdot \left(-a\right)\right)\right)\right) \cdot \left(\left(\frac{\frac{{\left(a \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2} + {\left(b \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}}{x-scale}}{x-scale} + \frac{\frac{{\left(a \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2} + {\left(b \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}}{y-scale}}{y-scale}\right) + \sqrt{{\left(\frac{\frac{{\left(a \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2} + {\left(b \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}}{x-scale}}{x-scale} - \frac{\frac{{\left(a \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2} + {\left(b \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}}{y-scale}}{y-scale}\right)}^{2} + {\left(\frac{\frac{\left(\left(2 \cdot \left({b}^{2} - {a}^{2}\right)\right) \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right) \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)}{x-scale}}{y-scale}\right)}^{2}}\right)}}{\frac{4 \cdot \left(\left(b \cdot a\right) \cdot \left(b \cdot \left(-a\right)\right)\right)}{{\left(x-scale \cdot y-scale\right)}^{2}}} \]
                    2. Add Preprocessing
                    3. Taylor expanded in x-scale around 0

                      \[\leadsto \color{blue}{\frac{1}{4} \cdot \left(\left(y-scale \cdot \sqrt{8}\right) \cdot \sqrt{2 \cdot \left({a}^{2} \cdot {\sin \left(\frac{1}{180} \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}\right) + 2 \cdot \left({b}^{2} \cdot {\cos \left(\frac{1}{180} \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}\right)}\right)} \]
                    4. Step-by-step derivation
                      1. Applied rewrites63.6%

                        \[\leadsto \color{blue}{0.25 \cdot \left(\left(y-scale \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, {\left(a \cdot \sin \left(0.005555555555555556 \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}, 2 \cdot {\left(b \cdot \cos \left(0.005555555555555556 \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}\right)}\right)} \]
                      2. Taylor expanded in angle around 0

                        \[\leadsto \frac{1}{4} \cdot \left(\left(y-scale \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, {\left(a \cdot \sin \left(\frac{1}{180} \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}, 2 \cdot {b}^{2}\right)}\right) \]
                      3. Step-by-step derivation
                        1. pow2N/A

                          \[\leadsto \frac{1}{4} \cdot \left(\left(y-scale \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, {\left(a \cdot \sin \left(\frac{1}{180} \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}, 2 \cdot \left(b \cdot b\right)\right)}\right) \]
                        2. lift-*.f6463.6

                          \[\leadsto 0.25 \cdot \left(\left(y-scale \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, {\left(a \cdot \sin \left(0.005555555555555556 \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}, 2 \cdot \left(b \cdot b\right)\right)}\right) \]
                      4. Applied rewrites63.6%

                        \[\leadsto 0.25 \cdot \left(\left(y-scale \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, {\left(a \cdot \sin \left(0.005555555555555556 \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}, 2 \cdot \left(b \cdot b\right)\right)}\right) \]
                      5. Taylor expanded in angle around 0

                        \[\leadsto \frac{1}{4} \cdot \left(\left(y-scale \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, {\left(a \cdot \left(\frac{1}{180} \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}, 2 \cdot \left(b \cdot b\right)\right)}\right) \]
                      6. Step-by-step derivation
                        1. lift-*.f64N/A

                          \[\leadsto \frac{1}{4} \cdot \left(\left(y-scale \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, {\left(a \cdot \left(\frac{1}{180} \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}, 2 \cdot \left(b \cdot b\right)\right)}\right) \]
                        2. lift-PI.f64N/A

                          \[\leadsto \frac{1}{4} \cdot \left(\left(y-scale \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, {\left(a \cdot \left(\frac{1}{180} \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}, 2 \cdot \left(b \cdot b\right)\right)}\right) \]
                        3. lift-*.f6461.8

                          \[\leadsto 0.25 \cdot \left(\left(y-scale \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, {\left(a \cdot \left(0.005555555555555556 \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}, 2 \cdot \left(b \cdot b\right)\right)}\right) \]
                      7. Applied rewrites61.8%

                        \[\leadsto 0.25 \cdot \left(\left(y-scale \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, {\left(a \cdot \left(0.005555555555555556 \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}, 2 \cdot \left(b \cdot b\right)\right)}\right) \]
                    5. Recombined 2 regimes into one program.
                    6. Add Preprocessing

                    Alternative 6: 26.8% accurate, 55.9× speedup?

                    \[\begin{array}{l} x-scale_m = \left|x-scale\right| \\ y-scale_m = \left|y-scale\right| \\ \begin{array}{l} \mathbf{if}\;b \leq 2.3 \cdot 10^{-41}:\\ \;\;\;\;0.25 \cdot \left(a \cdot \left(x-scale\_m \cdot 4\right)\right)\\ \mathbf{else}:\\ \;\;\;\;0.25 \cdot \left(\left(y-scale\_m \cdot \sqrt{8}\right) \cdot \sqrt{2 \cdot \left(b \cdot b\right)}\right)\\ \end{array} \end{array} \]
                    x-scale_m = (fabs.f64 x-scale)
                    y-scale_m = (fabs.f64 y-scale)
                    (FPCore (a b angle x-scale_m y-scale_m)
                     :precision binary64
                     (if (<= b 2.3e-41)
                       (* 0.25 (* a (* x-scale_m 4.0)))
                       (* 0.25 (* (* y-scale_m (sqrt 8.0)) (sqrt (* 2.0 (* b b)))))))
                    x-scale_m = fabs(x_45_scale);
                    y-scale_m = fabs(y_45_scale);
                    double code(double a, double b, double angle, double x_45_scale_m, double y_45_scale_m) {
                    	double tmp;
                    	if (b <= 2.3e-41) {
                    		tmp = 0.25 * (a * (x_45_scale_m * 4.0));
                    	} else {
                    		tmp = 0.25 * ((y_45_scale_m * sqrt(8.0)) * sqrt((2.0 * (b * b))));
                    	}
                    	return tmp;
                    }
                    
                    x-scale_m =     private
                    y-scale_m =     private
                    module fmin_fmax_functions
                        implicit none
                        private
                        public fmax
                        public fmin
                    
                        interface fmax
                            module procedure fmax88
                            module procedure fmax44
                            module procedure fmax84
                            module procedure fmax48
                        end interface
                        interface fmin
                            module procedure fmin88
                            module procedure fmin44
                            module procedure fmin84
                            module procedure fmin48
                        end interface
                    contains
                        real(8) function fmax88(x, y) result (res)
                            real(8), intent (in) :: x
                            real(8), intent (in) :: y
                            res = merge(y, merge(x, max(x, y), y /= y), x /= x)
                        end function
                        real(4) function fmax44(x, y) result (res)
                            real(4), intent (in) :: x
                            real(4), intent (in) :: y
                            res = merge(y, merge(x, max(x, y), y /= y), x /= x)
                        end function
                        real(8) function fmax84(x, y) result(res)
                            real(8), intent (in) :: x
                            real(4), intent (in) :: y
                            res = merge(dble(y), merge(x, max(x, dble(y)), y /= y), x /= x)
                        end function
                        real(8) function fmax48(x, y) result(res)
                            real(4), intent (in) :: x
                            real(8), intent (in) :: y
                            res = merge(y, merge(dble(x), max(dble(x), y), y /= y), x /= x)
                        end function
                        real(8) function fmin88(x, y) result (res)
                            real(8), intent (in) :: x
                            real(8), intent (in) :: y
                            res = merge(y, merge(x, min(x, y), y /= y), x /= x)
                        end function
                        real(4) function fmin44(x, y) result (res)
                            real(4), intent (in) :: x
                            real(4), intent (in) :: y
                            res = merge(y, merge(x, min(x, y), y /= y), x /= x)
                        end function
                        real(8) function fmin84(x, y) result(res)
                            real(8), intent (in) :: x
                            real(4), intent (in) :: y
                            res = merge(dble(y), merge(x, min(x, dble(y)), y /= y), x /= x)
                        end function
                        real(8) function fmin48(x, y) result(res)
                            real(4), intent (in) :: x
                            real(8), intent (in) :: y
                            res = merge(y, merge(dble(x), min(dble(x), y), y /= y), x /= x)
                        end function
                    end module
                    
                    real(8) function code(a, b, angle, x_45scale_m, y_45scale_m)
                    use fmin_fmax_functions
                        real(8), intent (in) :: a
                        real(8), intent (in) :: b
                        real(8), intent (in) :: angle
                        real(8), intent (in) :: x_45scale_m
                        real(8), intent (in) :: y_45scale_m
                        real(8) :: tmp
                        if (b <= 2.3d-41) then
                            tmp = 0.25d0 * (a * (x_45scale_m * 4.0d0))
                        else
                            tmp = 0.25d0 * ((y_45scale_m * sqrt(8.0d0)) * sqrt((2.0d0 * (b * b))))
                        end if
                        code = tmp
                    end function
                    
                    x-scale_m = Math.abs(x_45_scale);
                    y-scale_m = Math.abs(y_45_scale);
                    public static double code(double a, double b, double angle, double x_45_scale_m, double y_45_scale_m) {
                    	double tmp;
                    	if (b <= 2.3e-41) {
                    		tmp = 0.25 * (a * (x_45_scale_m * 4.0));
                    	} else {
                    		tmp = 0.25 * ((y_45_scale_m * Math.sqrt(8.0)) * Math.sqrt((2.0 * (b * b))));
                    	}
                    	return tmp;
                    }
                    
                    x-scale_m = math.fabs(x_45_scale)
                    y-scale_m = math.fabs(y_45_scale)
                    def code(a, b, angle, x_45_scale_m, y_45_scale_m):
                    	tmp = 0
                    	if b <= 2.3e-41:
                    		tmp = 0.25 * (a * (x_45_scale_m * 4.0))
                    	else:
                    		tmp = 0.25 * ((y_45_scale_m * math.sqrt(8.0)) * math.sqrt((2.0 * (b * b))))
                    	return tmp
                    
                    x-scale_m = abs(x_45_scale)
                    y-scale_m = abs(y_45_scale)
                    function code(a, b, angle, x_45_scale_m, y_45_scale_m)
                    	tmp = 0.0
                    	if (b <= 2.3e-41)
                    		tmp = Float64(0.25 * Float64(a * Float64(x_45_scale_m * 4.0)));
                    	else
                    		tmp = Float64(0.25 * Float64(Float64(y_45_scale_m * sqrt(8.0)) * sqrt(Float64(2.0 * Float64(b * b)))));
                    	end
                    	return tmp
                    end
                    
                    x-scale_m = abs(x_45_scale);
                    y-scale_m = abs(y_45_scale);
                    function tmp_2 = code(a, b, angle, x_45_scale_m, y_45_scale_m)
                    	tmp = 0.0;
                    	if (b <= 2.3e-41)
                    		tmp = 0.25 * (a * (x_45_scale_m * 4.0));
                    	else
                    		tmp = 0.25 * ((y_45_scale_m * sqrt(8.0)) * sqrt((2.0 * (b * b))));
                    	end
                    	tmp_2 = tmp;
                    end
                    
                    x-scale_m = N[Abs[x$45$scale], $MachinePrecision]
                    y-scale_m = N[Abs[y$45$scale], $MachinePrecision]
                    code[a_, b_, angle_, x$45$scale$95$m_, y$45$scale$95$m_] := If[LessEqual[b, 2.3e-41], N[(0.25 * N[(a * N[(x$45$scale$95$m * 4.0), $MachinePrecision]), $MachinePrecision]), $MachinePrecision], N[(0.25 * N[(N[(y$45$scale$95$m * N[Sqrt[8.0], $MachinePrecision]), $MachinePrecision] * N[Sqrt[N[(2.0 * N[(b * b), $MachinePrecision]), $MachinePrecision]], $MachinePrecision]), $MachinePrecision]), $MachinePrecision]]
                    
                    \begin{array}{l}
                    x-scale_m = \left|x-scale\right|
                    \\
                    y-scale_m = \left|y-scale\right|
                    
                    \\
                    \begin{array}{l}
                    \mathbf{if}\;b \leq 2.3 \cdot 10^{-41}:\\
                    \;\;\;\;0.25 \cdot \left(a \cdot \left(x-scale\_m \cdot 4\right)\right)\\
                    
                    \mathbf{else}:\\
                    \;\;\;\;0.25 \cdot \left(\left(y-scale\_m \cdot \sqrt{8}\right) \cdot \sqrt{2 \cdot \left(b \cdot b\right)}\right)\\
                    
                    
                    \end{array}
                    \end{array}
                    
                    Derivation
                    1. Split input into 2 regimes
                    2. if b < 2.3000000000000001e-41

                      1. Initial program 2.5%

                        \[\frac{-\sqrt{\left(\left(2 \cdot \frac{4 \cdot \left(\left(b \cdot a\right) \cdot \left(b \cdot \left(-a\right)\right)\right)}{{\left(x-scale \cdot y-scale\right)}^{2}}\right) \cdot \left(\left(b \cdot a\right) \cdot \left(b \cdot \left(-a\right)\right)\right)\right) \cdot \left(\left(\frac{\frac{{\left(a \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2} + {\left(b \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}}{x-scale}}{x-scale} + \frac{\frac{{\left(a \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2} + {\left(b \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}}{y-scale}}{y-scale}\right) + \sqrt{{\left(\frac{\frac{{\left(a \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2} + {\left(b \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}}{x-scale}}{x-scale} - \frac{\frac{{\left(a \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2} + {\left(b \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}}{y-scale}}{y-scale}\right)}^{2} + {\left(\frac{\frac{\left(\left(2 \cdot \left({b}^{2} - {a}^{2}\right)\right) \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right) \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)}{x-scale}}{y-scale}\right)}^{2}}\right)}}{\frac{4 \cdot \left(\left(b \cdot a\right) \cdot \left(b \cdot \left(-a\right)\right)\right)}{{\left(x-scale \cdot y-scale\right)}^{2}}} \]
                      2. Add Preprocessing
                      3. Taylor expanded in x-scale around inf

                        \[\leadsto \color{blue}{\frac{1}{4} \cdot \left(\left(x-scale \cdot \left(y-scale \cdot \sqrt{8}\right)\right) \cdot \sqrt{2 \cdot \frac{{a}^{2} \cdot {\cos \left(\frac{1}{180} \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}}{{y-scale}^{2}} + 2 \cdot \frac{{b}^{2} \cdot {\sin \left(\frac{1}{180} \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}}{{y-scale}^{2}}}\right)} \]
                      4. Applied rewrites9.7%

                        \[\leadsto \color{blue}{0.25 \cdot \left(\left(x-scale \cdot \left(y-scale \cdot \sqrt{8}\right)\right) \cdot \sqrt{\mathsf{fma}\left(2, \frac{{\left(a \cdot \cos \left(0.005555555555555556 \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}}{y-scale \cdot y-scale}, 2 \cdot \frac{{\left(b \cdot \sin \left(0.005555555555555556 \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}}{y-scale \cdot y-scale}\right)}\right)} \]
                      5. Taylor expanded in angle around 0

                        \[\leadsto \frac{1}{4} \cdot \left(a \cdot \color{blue}{\left(x-scale \cdot \left(\sqrt{2} \cdot \sqrt{8}\right)\right)}\right) \]
                      6. Step-by-step derivation
                        1. lower-*.f64N/A

                          \[\leadsto \frac{1}{4} \cdot \left(a \cdot \left(x-scale \cdot \color{blue}{\left(\sqrt{2} \cdot \sqrt{8}\right)}\right)\right) \]
                        2. sqrt-unprodN/A

                          \[\leadsto \frac{1}{4} \cdot \left(a \cdot \left(x-scale \cdot \sqrt{2 \cdot 8}\right)\right) \]
                        3. metadata-evalN/A

                          \[\leadsto \frac{1}{4} \cdot \left(a \cdot \left(x-scale \cdot \sqrt{16}\right)\right) \]
                        4. metadata-evalN/A

                          \[\leadsto \frac{1}{4} \cdot \left(a \cdot \left(x-scale \cdot 4\right)\right) \]
                        5. lower-*.f6420.1

                          \[\leadsto 0.25 \cdot \left(a \cdot \left(x-scale \cdot 4\right)\right) \]
                      7. Applied rewrites20.1%

                        \[\leadsto 0.25 \cdot \left(a \cdot \color{blue}{\left(x-scale \cdot 4\right)}\right) \]

                      if 2.3000000000000001e-41 < b

                      1. Initial program 6.8%

                        \[\frac{-\sqrt{\left(\left(2 \cdot \frac{4 \cdot \left(\left(b \cdot a\right) \cdot \left(b \cdot \left(-a\right)\right)\right)}{{\left(x-scale \cdot y-scale\right)}^{2}}\right) \cdot \left(\left(b \cdot a\right) \cdot \left(b \cdot \left(-a\right)\right)\right)\right) \cdot \left(\left(\frac{\frac{{\left(a \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2} + {\left(b \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}}{x-scale}}{x-scale} + \frac{\frac{{\left(a \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2} + {\left(b \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}}{y-scale}}{y-scale}\right) + \sqrt{{\left(\frac{\frac{{\left(a \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2} + {\left(b \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}}{x-scale}}{x-scale} - \frac{\frac{{\left(a \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2} + {\left(b \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}}{y-scale}}{y-scale}\right)}^{2} + {\left(\frac{\frac{\left(\left(2 \cdot \left({b}^{2} - {a}^{2}\right)\right) \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right) \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)}{x-scale}}{y-scale}\right)}^{2}}\right)}}{\frac{4 \cdot \left(\left(b \cdot a\right) \cdot \left(b \cdot \left(-a\right)\right)\right)}{{\left(x-scale \cdot y-scale\right)}^{2}}} \]
                      2. Add Preprocessing
                      3. Taylor expanded in x-scale around 0

                        \[\leadsto \color{blue}{\frac{1}{4} \cdot \left(\left(y-scale \cdot \sqrt{8}\right) \cdot \sqrt{2 \cdot \left({a}^{2} \cdot {\sin \left(\frac{1}{180} \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}\right) + 2 \cdot \left({b}^{2} \cdot {\cos \left(\frac{1}{180} \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}\right)}\right)} \]
                      4. Step-by-step derivation
                        1. Applied rewrites31.2%

                          \[\leadsto \color{blue}{0.25 \cdot \left(\left(y-scale \cdot \sqrt{8}\right) \cdot \sqrt{\mathsf{fma}\left(2, {\left(a \cdot \sin \left(0.005555555555555556 \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}, 2 \cdot {\left(b \cdot \cos \left(0.005555555555555556 \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}\right)}\right)} \]
                        2. Taylor expanded in angle around 0

                          \[\leadsto \frac{1}{4} \cdot \left(\left(y-scale \cdot \sqrt{8}\right) \cdot \sqrt{2 \cdot {b}^{2}}\right) \]
                        3. Step-by-step derivation
                          1. lower-*.f64N/A

                            \[\leadsto \frac{1}{4} \cdot \left(\left(y-scale \cdot \sqrt{8}\right) \cdot \sqrt{2 \cdot {b}^{2}}\right) \]
                          2. pow2N/A

                            \[\leadsto \frac{1}{4} \cdot \left(\left(y-scale \cdot \sqrt{8}\right) \cdot \sqrt{2 \cdot \left(b \cdot b\right)}\right) \]
                          3. lift-*.f6428.6

                            \[\leadsto 0.25 \cdot \left(\left(y-scale \cdot \sqrt{8}\right) \cdot \sqrt{2 \cdot \left(b \cdot b\right)}\right) \]
                        4. Applied rewrites28.6%

                          \[\leadsto 0.25 \cdot \left(\left(y-scale \cdot \sqrt{8}\right) \cdot \sqrt{2 \cdot \left(b \cdot b\right)}\right) \]
                      5. Recombined 2 regimes into one program.
                      6. Add Preprocessing

                      Alternative 7: 26.7% accurate, 132.1× speedup?

                      \[\begin{array}{l} x-scale_m = \left|x-scale\right| \\ y-scale_m = \left|y-scale\right| \\ \begin{array}{l} \mathbf{if}\;b \leq 2.3 \cdot 10^{-41}:\\ \;\;\;\;0.25 \cdot \left(a \cdot \left(x-scale\_m \cdot 4\right)\right)\\ \mathbf{else}:\\ \;\;\;\;b \cdot y-scale\_m\\ \end{array} \end{array} \]
                      x-scale_m = (fabs.f64 x-scale)
                      y-scale_m = (fabs.f64 y-scale)
                      (FPCore (a b angle x-scale_m y-scale_m)
                       :precision binary64
                       (if (<= b 2.3e-41) (* 0.25 (* a (* x-scale_m 4.0))) (* b y-scale_m)))
                      x-scale_m = fabs(x_45_scale);
                      y-scale_m = fabs(y_45_scale);
                      double code(double a, double b, double angle, double x_45_scale_m, double y_45_scale_m) {
                      	double tmp;
                      	if (b <= 2.3e-41) {
                      		tmp = 0.25 * (a * (x_45_scale_m * 4.0));
                      	} else {
                      		tmp = b * y_45_scale_m;
                      	}
                      	return tmp;
                      }
                      
                      x-scale_m =     private
                      y-scale_m =     private
                      module fmin_fmax_functions
                          implicit none
                          private
                          public fmax
                          public fmin
                      
                          interface fmax
                              module procedure fmax88
                              module procedure fmax44
                              module procedure fmax84
                              module procedure fmax48
                          end interface
                          interface fmin
                              module procedure fmin88
                              module procedure fmin44
                              module procedure fmin84
                              module procedure fmin48
                          end interface
                      contains
                          real(8) function fmax88(x, y) result (res)
                              real(8), intent (in) :: x
                              real(8), intent (in) :: y
                              res = merge(y, merge(x, max(x, y), y /= y), x /= x)
                          end function
                          real(4) function fmax44(x, y) result (res)
                              real(4), intent (in) :: x
                              real(4), intent (in) :: y
                              res = merge(y, merge(x, max(x, y), y /= y), x /= x)
                          end function
                          real(8) function fmax84(x, y) result(res)
                              real(8), intent (in) :: x
                              real(4), intent (in) :: y
                              res = merge(dble(y), merge(x, max(x, dble(y)), y /= y), x /= x)
                          end function
                          real(8) function fmax48(x, y) result(res)
                              real(4), intent (in) :: x
                              real(8), intent (in) :: y
                              res = merge(y, merge(dble(x), max(dble(x), y), y /= y), x /= x)
                          end function
                          real(8) function fmin88(x, y) result (res)
                              real(8), intent (in) :: x
                              real(8), intent (in) :: y
                              res = merge(y, merge(x, min(x, y), y /= y), x /= x)
                          end function
                          real(4) function fmin44(x, y) result (res)
                              real(4), intent (in) :: x
                              real(4), intent (in) :: y
                              res = merge(y, merge(x, min(x, y), y /= y), x /= x)
                          end function
                          real(8) function fmin84(x, y) result(res)
                              real(8), intent (in) :: x
                              real(4), intent (in) :: y
                              res = merge(dble(y), merge(x, min(x, dble(y)), y /= y), x /= x)
                          end function
                          real(8) function fmin48(x, y) result(res)
                              real(4), intent (in) :: x
                              real(8), intent (in) :: y
                              res = merge(y, merge(dble(x), min(dble(x), y), y /= y), x /= x)
                          end function
                      end module
                      
                      real(8) function code(a, b, angle, x_45scale_m, y_45scale_m)
                      use fmin_fmax_functions
                          real(8), intent (in) :: a
                          real(8), intent (in) :: b
                          real(8), intent (in) :: angle
                          real(8), intent (in) :: x_45scale_m
                          real(8), intent (in) :: y_45scale_m
                          real(8) :: tmp
                          if (b <= 2.3d-41) then
                              tmp = 0.25d0 * (a * (x_45scale_m * 4.0d0))
                          else
                              tmp = b * y_45scale_m
                          end if
                          code = tmp
                      end function
                      
                      x-scale_m = Math.abs(x_45_scale);
                      y-scale_m = Math.abs(y_45_scale);
                      public static double code(double a, double b, double angle, double x_45_scale_m, double y_45_scale_m) {
                      	double tmp;
                      	if (b <= 2.3e-41) {
                      		tmp = 0.25 * (a * (x_45_scale_m * 4.0));
                      	} else {
                      		tmp = b * y_45_scale_m;
                      	}
                      	return tmp;
                      }
                      
                      x-scale_m = math.fabs(x_45_scale)
                      y-scale_m = math.fabs(y_45_scale)
                      def code(a, b, angle, x_45_scale_m, y_45_scale_m):
                      	tmp = 0
                      	if b <= 2.3e-41:
                      		tmp = 0.25 * (a * (x_45_scale_m * 4.0))
                      	else:
                      		tmp = b * y_45_scale_m
                      	return tmp
                      
                      x-scale_m = abs(x_45_scale)
                      y-scale_m = abs(y_45_scale)
                      function code(a, b, angle, x_45_scale_m, y_45_scale_m)
                      	tmp = 0.0
                      	if (b <= 2.3e-41)
                      		tmp = Float64(0.25 * Float64(a * Float64(x_45_scale_m * 4.0)));
                      	else
                      		tmp = Float64(b * y_45_scale_m);
                      	end
                      	return tmp
                      end
                      
                      x-scale_m = abs(x_45_scale);
                      y-scale_m = abs(y_45_scale);
                      function tmp_2 = code(a, b, angle, x_45_scale_m, y_45_scale_m)
                      	tmp = 0.0;
                      	if (b <= 2.3e-41)
                      		tmp = 0.25 * (a * (x_45_scale_m * 4.0));
                      	else
                      		tmp = b * y_45_scale_m;
                      	end
                      	tmp_2 = tmp;
                      end
                      
                      x-scale_m = N[Abs[x$45$scale], $MachinePrecision]
                      y-scale_m = N[Abs[y$45$scale], $MachinePrecision]
                      code[a_, b_, angle_, x$45$scale$95$m_, y$45$scale$95$m_] := If[LessEqual[b, 2.3e-41], N[(0.25 * N[(a * N[(x$45$scale$95$m * 4.0), $MachinePrecision]), $MachinePrecision]), $MachinePrecision], N[(b * y$45$scale$95$m), $MachinePrecision]]
                      
                      \begin{array}{l}
                      x-scale_m = \left|x-scale\right|
                      \\
                      y-scale_m = \left|y-scale\right|
                      
                      \\
                      \begin{array}{l}
                      \mathbf{if}\;b \leq 2.3 \cdot 10^{-41}:\\
                      \;\;\;\;0.25 \cdot \left(a \cdot \left(x-scale\_m \cdot 4\right)\right)\\
                      
                      \mathbf{else}:\\
                      \;\;\;\;b \cdot y-scale\_m\\
                      
                      
                      \end{array}
                      \end{array}
                      
                      Derivation
                      1. Split input into 2 regimes
                      2. if b < 2.3000000000000001e-41

                        1. Initial program 2.5%

                          \[\frac{-\sqrt{\left(\left(2 \cdot \frac{4 \cdot \left(\left(b \cdot a\right) \cdot \left(b \cdot \left(-a\right)\right)\right)}{{\left(x-scale \cdot y-scale\right)}^{2}}\right) \cdot \left(\left(b \cdot a\right) \cdot \left(b \cdot \left(-a\right)\right)\right)\right) \cdot \left(\left(\frac{\frac{{\left(a \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2} + {\left(b \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}}{x-scale}}{x-scale} + \frac{\frac{{\left(a \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2} + {\left(b \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}}{y-scale}}{y-scale}\right) + \sqrt{{\left(\frac{\frac{{\left(a \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2} + {\left(b \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}}{x-scale}}{x-scale} - \frac{\frac{{\left(a \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2} + {\left(b \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}}{y-scale}}{y-scale}\right)}^{2} + {\left(\frac{\frac{\left(\left(2 \cdot \left({b}^{2} - {a}^{2}\right)\right) \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right) \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)}{x-scale}}{y-scale}\right)}^{2}}\right)}}{\frac{4 \cdot \left(\left(b \cdot a\right) \cdot \left(b \cdot \left(-a\right)\right)\right)}{{\left(x-scale \cdot y-scale\right)}^{2}}} \]
                        2. Add Preprocessing
                        3. Taylor expanded in x-scale around inf

                          \[\leadsto \color{blue}{\frac{1}{4} \cdot \left(\left(x-scale \cdot \left(y-scale \cdot \sqrt{8}\right)\right) \cdot \sqrt{2 \cdot \frac{{a}^{2} \cdot {\cos \left(\frac{1}{180} \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}}{{y-scale}^{2}} + 2 \cdot \frac{{b}^{2} \cdot {\sin \left(\frac{1}{180} \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}}{{y-scale}^{2}}}\right)} \]
                        4. Applied rewrites9.7%

                          \[\leadsto \color{blue}{0.25 \cdot \left(\left(x-scale \cdot \left(y-scale \cdot \sqrt{8}\right)\right) \cdot \sqrt{\mathsf{fma}\left(2, \frac{{\left(a \cdot \cos \left(0.005555555555555556 \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}}{y-scale \cdot y-scale}, 2 \cdot \frac{{\left(b \cdot \sin \left(0.005555555555555556 \cdot \left(angle \cdot \mathsf{PI}\left(\right)\right)\right)\right)}^{2}}{y-scale \cdot y-scale}\right)}\right)} \]
                        5. Taylor expanded in angle around 0

                          \[\leadsto \frac{1}{4} \cdot \left(a \cdot \color{blue}{\left(x-scale \cdot \left(\sqrt{2} \cdot \sqrt{8}\right)\right)}\right) \]
                        6. Step-by-step derivation
                          1. lower-*.f64N/A

                            \[\leadsto \frac{1}{4} \cdot \left(a \cdot \left(x-scale \cdot \color{blue}{\left(\sqrt{2} \cdot \sqrt{8}\right)}\right)\right) \]
                          2. sqrt-unprodN/A

                            \[\leadsto \frac{1}{4} \cdot \left(a \cdot \left(x-scale \cdot \sqrt{2 \cdot 8}\right)\right) \]
                          3. metadata-evalN/A

                            \[\leadsto \frac{1}{4} \cdot \left(a \cdot \left(x-scale \cdot \sqrt{16}\right)\right) \]
                          4. metadata-evalN/A

                            \[\leadsto \frac{1}{4} \cdot \left(a \cdot \left(x-scale \cdot 4\right)\right) \]
                          5. lower-*.f6420.1

                            \[\leadsto 0.25 \cdot \left(a \cdot \left(x-scale \cdot 4\right)\right) \]
                        7. Applied rewrites20.1%

                          \[\leadsto 0.25 \cdot \left(a \cdot \color{blue}{\left(x-scale \cdot 4\right)}\right) \]

                        if 2.3000000000000001e-41 < b

                        1. Initial program 6.8%

                          \[\frac{-\sqrt{\left(\left(2 \cdot \frac{4 \cdot \left(\left(b \cdot a\right) \cdot \left(b \cdot \left(-a\right)\right)\right)}{{\left(x-scale \cdot y-scale\right)}^{2}}\right) \cdot \left(\left(b \cdot a\right) \cdot \left(b \cdot \left(-a\right)\right)\right)\right) \cdot \left(\left(\frac{\frac{{\left(a \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2} + {\left(b \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}}{x-scale}}{x-scale} + \frac{\frac{{\left(a \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2} + {\left(b \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}}{y-scale}}{y-scale}\right) + \sqrt{{\left(\frac{\frac{{\left(a \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2} + {\left(b \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}}{x-scale}}{x-scale} - \frac{\frac{{\left(a \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2} + {\left(b \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}}{y-scale}}{y-scale}\right)}^{2} + {\left(\frac{\frac{\left(\left(2 \cdot \left({b}^{2} - {a}^{2}\right)\right) \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right) \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)}{x-scale}}{y-scale}\right)}^{2}}\right)}}{\frac{4 \cdot \left(\left(b \cdot a\right) \cdot \left(b \cdot \left(-a\right)\right)\right)}{{\left(x-scale \cdot y-scale\right)}^{2}}} \]
                        2. Add Preprocessing
                        3. Taylor expanded in angle around 0

                          \[\leadsto \color{blue}{\frac{1}{4} \cdot \left(b \cdot \left(y-scale \cdot \left(\sqrt{2} \cdot \sqrt{8}\right)\right)\right)} \]
                        4. Step-by-step derivation
                          1. Applied rewrites26.7%

                            \[\leadsto \color{blue}{0.25 \cdot \left(b \cdot \left(y-scale \cdot 4\right)\right)} \]
                          2. Taylor expanded in b around 0

                            \[\leadsto b \cdot \color{blue}{y-scale} \]
                          3. Step-by-step derivation
                            1. lower-*.f6426.7

                              \[\leadsto b \cdot y-scale \]
                          4. Applied rewrites26.7%

                            \[\leadsto b \cdot \color{blue}{y-scale} \]
                        5. Recombined 2 regimes into one program.
                        6. Add Preprocessing

                        Alternative 8: 17.6% accurate, 484.7× speedup?

                        \[\begin{array}{l} x-scale_m = \left|x-scale\right| \\ y-scale_m = \left|y-scale\right| \\ b \cdot y-scale\_m \end{array} \]
                        x-scale_m = (fabs.f64 x-scale)
                        y-scale_m = (fabs.f64 y-scale)
                        (FPCore (a b angle x-scale_m y-scale_m) :precision binary64 (* b y-scale_m))
                        x-scale_m = fabs(x_45_scale);
                        y-scale_m = fabs(y_45_scale);
                        double code(double a, double b, double angle, double x_45_scale_m, double y_45_scale_m) {
                        	return b * y_45_scale_m;
                        }
                        
                        x-scale_m =     private
                        y-scale_m =     private
                        module fmin_fmax_functions
                            implicit none
                            private
                            public fmax
                            public fmin
                        
                            interface fmax
                                module procedure fmax88
                                module procedure fmax44
                                module procedure fmax84
                                module procedure fmax48
                            end interface
                            interface fmin
                                module procedure fmin88
                                module procedure fmin44
                                module procedure fmin84
                                module procedure fmin48
                            end interface
                        contains
                            real(8) function fmax88(x, y) result (res)
                                real(8), intent (in) :: x
                                real(8), intent (in) :: y
                                res = merge(y, merge(x, max(x, y), y /= y), x /= x)
                            end function
                            real(4) function fmax44(x, y) result (res)
                                real(4), intent (in) :: x
                                real(4), intent (in) :: y
                                res = merge(y, merge(x, max(x, y), y /= y), x /= x)
                            end function
                            real(8) function fmax84(x, y) result(res)
                                real(8), intent (in) :: x
                                real(4), intent (in) :: y
                                res = merge(dble(y), merge(x, max(x, dble(y)), y /= y), x /= x)
                            end function
                            real(8) function fmax48(x, y) result(res)
                                real(4), intent (in) :: x
                                real(8), intent (in) :: y
                                res = merge(y, merge(dble(x), max(dble(x), y), y /= y), x /= x)
                            end function
                            real(8) function fmin88(x, y) result (res)
                                real(8), intent (in) :: x
                                real(8), intent (in) :: y
                                res = merge(y, merge(x, min(x, y), y /= y), x /= x)
                            end function
                            real(4) function fmin44(x, y) result (res)
                                real(4), intent (in) :: x
                                real(4), intent (in) :: y
                                res = merge(y, merge(x, min(x, y), y /= y), x /= x)
                            end function
                            real(8) function fmin84(x, y) result(res)
                                real(8), intent (in) :: x
                                real(4), intent (in) :: y
                                res = merge(dble(y), merge(x, min(x, dble(y)), y /= y), x /= x)
                            end function
                            real(8) function fmin48(x, y) result(res)
                                real(4), intent (in) :: x
                                real(8), intent (in) :: y
                                res = merge(y, merge(dble(x), min(dble(x), y), y /= y), x /= x)
                            end function
                        end module
                        
                        real(8) function code(a, b, angle, x_45scale_m, y_45scale_m)
                        use fmin_fmax_functions
                            real(8), intent (in) :: a
                            real(8), intent (in) :: b
                            real(8), intent (in) :: angle
                            real(8), intent (in) :: x_45scale_m
                            real(8), intent (in) :: y_45scale_m
                            code = b * y_45scale_m
                        end function
                        
                        x-scale_m = Math.abs(x_45_scale);
                        y-scale_m = Math.abs(y_45_scale);
                        public static double code(double a, double b, double angle, double x_45_scale_m, double y_45_scale_m) {
                        	return b * y_45_scale_m;
                        }
                        
                        x-scale_m = math.fabs(x_45_scale)
                        y-scale_m = math.fabs(y_45_scale)
                        def code(a, b, angle, x_45_scale_m, y_45_scale_m):
                        	return b * y_45_scale_m
                        
                        x-scale_m = abs(x_45_scale)
                        y-scale_m = abs(y_45_scale)
                        function code(a, b, angle, x_45_scale_m, y_45_scale_m)
                        	return Float64(b * y_45_scale_m)
                        end
                        
                        x-scale_m = abs(x_45_scale);
                        y-scale_m = abs(y_45_scale);
                        function tmp = code(a, b, angle, x_45_scale_m, y_45_scale_m)
                        	tmp = b * y_45_scale_m;
                        end
                        
                        x-scale_m = N[Abs[x$45$scale], $MachinePrecision]
                        y-scale_m = N[Abs[y$45$scale], $MachinePrecision]
                        code[a_, b_, angle_, x$45$scale$95$m_, y$45$scale$95$m_] := N[(b * y$45$scale$95$m), $MachinePrecision]
                        
                        \begin{array}{l}
                        x-scale_m = \left|x-scale\right|
                        \\
                        y-scale_m = \left|y-scale\right|
                        
                        \\
                        b \cdot y-scale\_m
                        \end{array}
                        
                        Derivation
                        1. Initial program 3.8%

                          \[\frac{-\sqrt{\left(\left(2 \cdot \frac{4 \cdot \left(\left(b \cdot a\right) \cdot \left(b \cdot \left(-a\right)\right)\right)}{{\left(x-scale \cdot y-scale\right)}^{2}}\right) \cdot \left(\left(b \cdot a\right) \cdot \left(b \cdot \left(-a\right)\right)\right)\right) \cdot \left(\left(\frac{\frac{{\left(a \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2} + {\left(b \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}}{x-scale}}{x-scale} + \frac{\frac{{\left(a \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2} + {\left(b \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}}{y-scale}}{y-scale}\right) + \sqrt{{\left(\frac{\frac{{\left(a \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2} + {\left(b \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}}{x-scale}}{x-scale} - \frac{\frac{{\left(a \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2} + {\left(b \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right)}^{2}}{y-scale}}{y-scale}\right)}^{2} + {\left(\frac{\frac{\left(\left(2 \cdot \left({b}^{2} - {a}^{2}\right)\right) \cdot \sin \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)\right) \cdot \cos \left(\frac{angle}{180} \cdot \mathsf{PI}\left(\right)\right)}{x-scale}}{y-scale}\right)}^{2}}\right)}}{\frac{4 \cdot \left(\left(b \cdot a\right) \cdot \left(b \cdot \left(-a\right)\right)\right)}{{\left(x-scale \cdot y-scale\right)}^{2}}} \]
                        2. Add Preprocessing
                        3. Taylor expanded in angle around 0

                          \[\leadsto \color{blue}{\frac{1}{4} \cdot \left(b \cdot \left(y-scale \cdot \left(\sqrt{2} \cdot \sqrt{8}\right)\right)\right)} \]
                        4. Step-by-step derivation
                          1. Applied rewrites19.3%

                            \[\leadsto \color{blue}{0.25 \cdot \left(b \cdot \left(y-scale \cdot 4\right)\right)} \]
                          2. Taylor expanded in b around 0

                            \[\leadsto b \cdot \color{blue}{y-scale} \]
                          3. Step-by-step derivation
                            1. lower-*.f6419.3

                              \[\leadsto b \cdot y-scale \]
                          4. Applied rewrites19.3%

                            \[\leadsto b \cdot \color{blue}{y-scale} \]
                          5. Add Preprocessing

                          Reproduce

                          ?
                          herbie shell --seed 2025056 
                          (FPCore (a b angle x-scale y-scale)
                            :name "a from scale-rotated-ellipse"
                            :precision binary64
                            (/ (- (sqrt (* (* (* 2.0 (/ (* 4.0 (* (* b a) (* b (- a)))) (pow (* x-scale y-scale) 2.0))) (* (* b a) (* b (- a)))) (+ (+ (/ (/ (+ (pow (* a (sin (* (/ angle 180.0) (PI)))) 2.0) (pow (* b (cos (* (/ angle 180.0) (PI)))) 2.0)) x-scale) x-scale) (/ (/ (+ (pow (* a (cos (* (/ angle 180.0) (PI)))) 2.0) (pow (* b (sin (* (/ angle 180.0) (PI)))) 2.0)) y-scale) y-scale)) (sqrt (+ (pow (- (/ (/ (+ (pow (* a (sin (* (/ angle 180.0) (PI)))) 2.0) (pow (* b (cos (* (/ angle 180.0) (PI)))) 2.0)) x-scale) x-scale) (/ (/ (+ (pow (* a (cos (* (/ angle 180.0) (PI)))) 2.0) (pow (* b (sin (* (/ angle 180.0) (PI)))) 2.0)) y-scale) y-scale)) 2.0) (pow (/ (/ (* (* (* 2.0 (- (pow b 2.0) (pow a 2.0))) (sin (* (/ angle 180.0) (PI)))) (cos (* (/ angle 180.0) (PI)))) x-scale) y-scale) 2.0))))))) (/ (* 4.0 (* (* b a) (* b (- a)))) (pow (* x-scale y-scale) 2.0))))