ab-angle->ABCF C

Percentage Accurate: 79.8% → 79.8%
Time: 9.4s
Alternatives: 6
Speedup: 1.9×

Specification

?
\[\begin{array}{l} \\ \begin{array}{l} t_0 := \mathsf{PI}\left(\right) \cdot \frac{angle}{180}\\ {\left(a \cdot \cos t\_0\right)}^{2} + {\left(b \cdot \sin t\_0\right)}^{2} \end{array} \end{array} \]
(FPCore (a b angle)
 :precision binary64
 (let* ((t_0 (* (PI) (/ angle 180.0))))
   (+ (pow (* a (cos t_0)) 2.0) (pow (* b (sin t_0)) 2.0))))
\begin{array}{l}

\\
\begin{array}{l}
t_0 := \mathsf{PI}\left(\right) \cdot \frac{angle}{180}\\
{\left(a \cdot \cos t\_0\right)}^{2} + {\left(b \cdot \sin t\_0\right)}^{2}
\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 6 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: 79.8% accurate, 1.0× speedup?

\[\begin{array}{l} \\ \begin{array}{l} t_0 := \mathsf{PI}\left(\right) \cdot \frac{angle}{180}\\ {\left(a \cdot \cos t\_0\right)}^{2} + {\left(b \cdot \sin t\_0\right)}^{2} \end{array} \end{array} \]
(FPCore (a b angle)
 :precision binary64
 (let* ((t_0 (* (PI) (/ angle 180.0))))
   (+ (pow (* a (cos t_0)) 2.0) (pow (* b (sin t_0)) 2.0))))
\begin{array}{l}

\\
\begin{array}{l}
t_0 := \mathsf{PI}\left(\right) \cdot \frac{angle}{180}\\
{\left(a \cdot \cos t\_0\right)}^{2} + {\left(b \cdot \sin t\_0\right)}^{2}
\end{array}
\end{array}

Alternative 1: 79.8% accurate, 1.9× speedup?

\[\begin{array}{l} \\ \mathsf{fma}\left(\left(1 \cdot a\right) \cdot 1, a, {\left(\sin \left(0.005555555555555556 \cdot \left(\mathsf{PI}\left(\right) \cdot angle\right)\right) \cdot b\right)}^{2}\right) \end{array} \]
(FPCore (a b angle)
 :precision binary64
 (fma
  (* (* 1.0 a) 1.0)
  a
  (pow (* (sin (* 0.005555555555555556 (* (PI) angle))) b) 2.0)))
\begin{array}{l}

\\
\mathsf{fma}\left(\left(1 \cdot a\right) \cdot 1, a, {\left(\sin \left(0.005555555555555556 \cdot \left(\mathsf{PI}\left(\right) \cdot angle\right)\right) \cdot b\right)}^{2}\right)
\end{array}
Derivation
  1. Initial program 81.4%

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

    \[\leadsto {\left(a \cdot \color{blue}{1}\right)}^{2} + {\left(b \cdot \sin \left(\mathsf{PI}\left(\right) \cdot \frac{angle}{180}\right)\right)}^{2} \]
  4. Step-by-step derivation
    1. Applied rewrites81.6%

      \[\leadsto {\left(a \cdot \color{blue}{1}\right)}^{2} + {\left(b \cdot \sin \left(\mathsf{PI}\left(\right) \cdot \frac{angle}{180}\right)\right)}^{2} \]
    2. Step-by-step derivation
      1. lift-*.f64N/A

        \[\leadsto {\left(a \cdot 1\right)}^{2} + {\left(b \cdot \sin \color{blue}{\left(\mathsf{PI}\left(\right) \cdot \frac{angle}{180}\right)}\right)}^{2} \]
      2. lift-/.f64N/A

        \[\leadsto {\left(a \cdot 1\right)}^{2} + {\left(b \cdot \sin \left(\mathsf{PI}\left(\right) \cdot \color{blue}{\frac{angle}{180}}\right)\right)}^{2} \]
      3. div-invN/A

        \[\leadsto {\left(a \cdot 1\right)}^{2} + {\left(b \cdot \sin \left(\mathsf{PI}\left(\right) \cdot \color{blue}{\left(angle \cdot \frac{1}{180}\right)}\right)\right)}^{2} \]
      4. metadata-evalN/A

        \[\leadsto {\left(a \cdot 1\right)}^{2} + {\left(b \cdot \sin \left(\mathsf{PI}\left(\right) \cdot \left(angle \cdot \color{blue}{\frac{1}{180}}\right)\right)\right)}^{2} \]
      5. associate-*r*N/A

        \[\leadsto {\left(a \cdot 1\right)}^{2} + {\left(b \cdot \sin \color{blue}{\left(\left(\mathsf{PI}\left(\right) \cdot angle\right) \cdot \frac{1}{180}\right)}\right)}^{2} \]
      6. *-commutativeN/A

        \[\leadsto {\left(a \cdot 1\right)}^{2} + {\left(b \cdot \sin \left(\color{blue}{\left(angle \cdot \mathsf{PI}\left(\right)\right)} \cdot \frac{1}{180}\right)\right)}^{2} \]
      7. lift-*.f64N/A

        \[\leadsto {\left(a \cdot 1\right)}^{2} + {\left(b \cdot \sin \left(\color{blue}{\left(angle \cdot \mathsf{PI}\left(\right)\right)} \cdot \frac{1}{180}\right)\right)}^{2} \]
      8. lower-*.f6481.7

        \[\leadsto {\left(a \cdot 1\right)}^{2} + {\left(b \cdot \sin \color{blue}{\left(\left(angle \cdot \mathsf{PI}\left(\right)\right) \cdot 0.005555555555555556\right)}\right)}^{2} \]
      9. lift-*.f64N/A

        \[\leadsto {\left(a \cdot 1\right)}^{2} + {\left(b \cdot \sin \left(\color{blue}{\left(angle \cdot \mathsf{PI}\left(\right)\right)} \cdot \frac{1}{180}\right)\right)}^{2} \]
      10. *-commutativeN/A

        \[\leadsto {\left(a \cdot 1\right)}^{2} + {\left(b \cdot \sin \left(\color{blue}{\left(\mathsf{PI}\left(\right) \cdot angle\right)} \cdot \frac{1}{180}\right)\right)}^{2} \]
      11. lower-*.f6481.7

        \[\leadsto {\left(a \cdot 1\right)}^{2} + {\left(b \cdot \sin \left(\color{blue}{\left(\mathsf{PI}\left(\right) \cdot angle\right)} \cdot 0.005555555555555556\right)\right)}^{2} \]
    3. Applied rewrites81.7%

      \[\leadsto {\left(a \cdot 1\right)}^{2} + {\left(b \cdot \sin \color{blue}{\left(\left(\mathsf{PI}\left(\right) \cdot angle\right) \cdot 0.005555555555555556\right)}\right)}^{2} \]
    4. Step-by-step derivation
      1. lift-+.f64N/A

        \[\leadsto \color{blue}{{\left(a \cdot 1\right)}^{2} + {\left(b \cdot \sin \left(\left(\mathsf{PI}\left(\right) \cdot angle\right) \cdot \frac{1}{180}\right)\right)}^{2}} \]
      2. lift-pow.f64N/A

        \[\leadsto \color{blue}{{\left(a \cdot 1\right)}^{2}} + {\left(b \cdot \sin \left(\left(\mathsf{PI}\left(\right) \cdot angle\right) \cdot \frac{1}{180}\right)\right)}^{2} \]
      3. lift-*.f64N/A

        \[\leadsto {\color{blue}{\left(a \cdot 1\right)}}^{2} + {\left(b \cdot \sin \left(\left(\mathsf{PI}\left(\right) \cdot angle\right) \cdot \frac{1}{180}\right)\right)}^{2} \]
      4. *-commutativeN/A

        \[\leadsto {\color{blue}{\left(1 \cdot a\right)}}^{2} + {\left(b \cdot \sin \left(\left(\mathsf{PI}\left(\right) \cdot angle\right) \cdot \frac{1}{180}\right)\right)}^{2} \]
      5. lift-*.f64N/A

        \[\leadsto {\color{blue}{\left(1 \cdot a\right)}}^{2} + {\left(b \cdot \sin \left(\left(\mathsf{PI}\left(\right) \cdot angle\right) \cdot \frac{1}{180}\right)\right)}^{2} \]
      6. unpow2N/A

        \[\leadsto \color{blue}{\left(1 \cdot a\right) \cdot \left(1 \cdot a\right)} + {\left(b \cdot \sin \left(\left(\mathsf{PI}\left(\right) \cdot angle\right) \cdot \frac{1}{180}\right)\right)}^{2} \]
      7. lift-*.f64N/A

        \[\leadsto \left(1 \cdot a\right) \cdot \color{blue}{\left(1 \cdot a\right)} + {\left(b \cdot \sin \left(\left(\mathsf{PI}\left(\right) \cdot angle\right) \cdot \frac{1}{180}\right)\right)}^{2} \]
      8. associate-*r*N/A

        \[\leadsto \color{blue}{\left(\left(1 \cdot a\right) \cdot 1\right) \cdot a} + {\left(b \cdot \sin \left(\left(\mathsf{PI}\left(\right) \cdot angle\right) \cdot \frac{1}{180}\right)\right)}^{2} \]
      9. lower-fma.f64N/A

        \[\leadsto \color{blue}{\mathsf{fma}\left(\left(1 \cdot a\right) \cdot 1, a, {\left(b \cdot \sin \left(\left(\mathsf{PI}\left(\right) \cdot angle\right) \cdot \frac{1}{180}\right)\right)}^{2}\right)} \]
    5. Applied rewrites81.7%

      \[\leadsto \color{blue}{\mathsf{fma}\left(\left(1 \cdot a\right) \cdot 1, a, {\left(\sin \left(0.005555555555555556 \cdot \left(\mathsf{PI}\left(\right) \cdot angle\right)\right) \cdot b\right)}^{2}\right)} \]
    6. Add Preprocessing

    Alternative 2: 67.3% accurate, 3.2× speedup?

    \[\begin{array}{l} \\ \begin{array}{l} \mathbf{if}\;b \leq 1.3 \cdot 10^{-143}:\\ \;\;\;\;a \cdot a\\ \mathbf{else}:\\ \;\;\;\;\mathsf{fma}\left(\left(1 \cdot a\right) \cdot 1, a, {\left(\left(\left(b \cdot \mathsf{PI}\left(\right)\right) \cdot angle\right) \cdot 0.005555555555555556\right)}^{2}\right)\\ \end{array} \end{array} \]
    (FPCore (a b angle)
     :precision binary64
     (if (<= b 1.3e-143)
       (* a a)
       (fma
        (* (* 1.0 a) 1.0)
        a
        (pow (* (* (* b (PI)) angle) 0.005555555555555556) 2.0))))
    \begin{array}{l}
    
    \\
    \begin{array}{l}
    \mathbf{if}\;b \leq 1.3 \cdot 10^{-143}:\\
    \;\;\;\;a \cdot a\\
    
    \mathbf{else}:\\
    \;\;\;\;\mathsf{fma}\left(\left(1 \cdot a\right) \cdot 1, a, {\left(\left(\left(b \cdot \mathsf{PI}\left(\right)\right) \cdot angle\right) \cdot 0.005555555555555556\right)}^{2}\right)\\
    
    
    \end{array}
    \end{array}
    
    Derivation
    1. Split input into 2 regimes
    2. if b < 1.29999999999999994e-143

      1. Initial program 84.5%

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

        \[\leadsto \color{blue}{{a}^{2}} \]
      4. Step-by-step derivation
        1. unpow2N/A

          \[\leadsto \color{blue}{a \cdot a} \]
        2. lower-*.f6460.8

          \[\leadsto \color{blue}{a \cdot a} \]
      5. Applied rewrites60.8%

        \[\leadsto \color{blue}{a \cdot a} \]

      if 1.29999999999999994e-143 < b

      1. Initial program 75.7%

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

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

          \[\leadsto {\left(a \cdot \color{blue}{1}\right)}^{2} + {\left(b \cdot \sin \left(\mathsf{PI}\left(\right) \cdot \frac{angle}{180}\right)\right)}^{2} \]
        2. Step-by-step derivation
          1. lift-*.f64N/A

            \[\leadsto {\left(a \cdot 1\right)}^{2} + {\left(b \cdot \sin \color{blue}{\left(\mathsf{PI}\left(\right) \cdot \frac{angle}{180}\right)}\right)}^{2} \]
          2. lift-/.f64N/A

            \[\leadsto {\left(a \cdot 1\right)}^{2} + {\left(b \cdot \sin \left(\mathsf{PI}\left(\right) \cdot \color{blue}{\frac{angle}{180}}\right)\right)}^{2} \]
          3. div-invN/A

            \[\leadsto {\left(a \cdot 1\right)}^{2} + {\left(b \cdot \sin \left(\mathsf{PI}\left(\right) \cdot \color{blue}{\left(angle \cdot \frac{1}{180}\right)}\right)\right)}^{2} \]
          4. metadata-evalN/A

            \[\leadsto {\left(a \cdot 1\right)}^{2} + {\left(b \cdot \sin \left(\mathsf{PI}\left(\right) \cdot \left(angle \cdot \color{blue}{\frac{1}{180}}\right)\right)\right)}^{2} \]
          5. associate-*r*N/A

            \[\leadsto {\left(a \cdot 1\right)}^{2} + {\left(b \cdot \sin \color{blue}{\left(\left(\mathsf{PI}\left(\right) \cdot angle\right) \cdot \frac{1}{180}\right)}\right)}^{2} \]
          6. *-commutativeN/A

            \[\leadsto {\left(a \cdot 1\right)}^{2} + {\left(b \cdot \sin \left(\color{blue}{\left(angle \cdot \mathsf{PI}\left(\right)\right)} \cdot \frac{1}{180}\right)\right)}^{2} \]
          7. lift-*.f64N/A

            \[\leadsto {\left(a \cdot 1\right)}^{2} + {\left(b \cdot \sin \left(\color{blue}{\left(angle \cdot \mathsf{PI}\left(\right)\right)} \cdot \frac{1}{180}\right)\right)}^{2} \]
          8. lower-*.f6476.0

            \[\leadsto {\left(a \cdot 1\right)}^{2} + {\left(b \cdot \sin \color{blue}{\left(\left(angle \cdot \mathsf{PI}\left(\right)\right) \cdot 0.005555555555555556\right)}\right)}^{2} \]
          9. lift-*.f64N/A

            \[\leadsto {\left(a \cdot 1\right)}^{2} + {\left(b \cdot \sin \left(\color{blue}{\left(angle \cdot \mathsf{PI}\left(\right)\right)} \cdot \frac{1}{180}\right)\right)}^{2} \]
          10. *-commutativeN/A

            \[\leadsto {\left(a \cdot 1\right)}^{2} + {\left(b \cdot \sin \left(\color{blue}{\left(\mathsf{PI}\left(\right) \cdot angle\right)} \cdot \frac{1}{180}\right)\right)}^{2} \]
          11. lower-*.f6476.0

            \[\leadsto {\left(a \cdot 1\right)}^{2} + {\left(b \cdot \sin \left(\color{blue}{\left(\mathsf{PI}\left(\right) \cdot angle\right)} \cdot 0.005555555555555556\right)\right)}^{2} \]
        3. Applied rewrites76.0%

          \[\leadsto {\left(a \cdot 1\right)}^{2} + {\left(b \cdot \sin \color{blue}{\left(\left(\mathsf{PI}\left(\right) \cdot angle\right) \cdot 0.005555555555555556\right)}\right)}^{2} \]
        4. Step-by-step derivation
          1. lift-+.f64N/A

            \[\leadsto \color{blue}{{\left(a \cdot 1\right)}^{2} + {\left(b \cdot \sin \left(\left(\mathsf{PI}\left(\right) \cdot angle\right) \cdot \frac{1}{180}\right)\right)}^{2}} \]
          2. lift-pow.f64N/A

            \[\leadsto \color{blue}{{\left(a \cdot 1\right)}^{2}} + {\left(b \cdot \sin \left(\left(\mathsf{PI}\left(\right) \cdot angle\right) \cdot \frac{1}{180}\right)\right)}^{2} \]
          3. lift-*.f64N/A

            \[\leadsto {\color{blue}{\left(a \cdot 1\right)}}^{2} + {\left(b \cdot \sin \left(\left(\mathsf{PI}\left(\right) \cdot angle\right) \cdot \frac{1}{180}\right)\right)}^{2} \]
          4. *-commutativeN/A

            \[\leadsto {\color{blue}{\left(1 \cdot a\right)}}^{2} + {\left(b \cdot \sin \left(\left(\mathsf{PI}\left(\right) \cdot angle\right) \cdot \frac{1}{180}\right)\right)}^{2} \]
          5. lift-*.f64N/A

            \[\leadsto {\color{blue}{\left(1 \cdot a\right)}}^{2} + {\left(b \cdot \sin \left(\left(\mathsf{PI}\left(\right) \cdot angle\right) \cdot \frac{1}{180}\right)\right)}^{2} \]
          6. unpow2N/A

            \[\leadsto \color{blue}{\left(1 \cdot a\right) \cdot \left(1 \cdot a\right)} + {\left(b \cdot \sin \left(\left(\mathsf{PI}\left(\right) \cdot angle\right) \cdot \frac{1}{180}\right)\right)}^{2} \]
          7. lift-*.f64N/A

            \[\leadsto \left(1 \cdot a\right) \cdot \color{blue}{\left(1 \cdot a\right)} + {\left(b \cdot \sin \left(\left(\mathsf{PI}\left(\right) \cdot angle\right) \cdot \frac{1}{180}\right)\right)}^{2} \]
          8. associate-*r*N/A

            \[\leadsto \color{blue}{\left(\left(1 \cdot a\right) \cdot 1\right) \cdot a} + {\left(b \cdot \sin \left(\left(\mathsf{PI}\left(\right) \cdot angle\right) \cdot \frac{1}{180}\right)\right)}^{2} \]
          9. lower-fma.f64N/A

            \[\leadsto \color{blue}{\mathsf{fma}\left(\left(1 \cdot a\right) \cdot 1, a, {\left(b \cdot \sin \left(\left(\mathsf{PI}\left(\right) \cdot angle\right) \cdot \frac{1}{180}\right)\right)}^{2}\right)} \]
        5. Applied rewrites76.0%

          \[\leadsto \color{blue}{\mathsf{fma}\left(\left(1 \cdot a\right) \cdot 1, a, {\left(\sin \left(0.005555555555555556 \cdot \left(\mathsf{PI}\left(\right) \cdot angle\right)\right) \cdot b\right)}^{2}\right)} \]
        6. Taylor expanded in angle around 0

          \[\leadsto \mathsf{fma}\left(\left(1 \cdot a\right) \cdot 1, a, {\color{blue}{\left(\frac{1}{180} \cdot \left(angle \cdot \left(b \cdot \mathsf{PI}\left(\right)\right)\right)\right)}}^{2}\right) \]
        7. Step-by-step derivation
          1. *-commutativeN/A

            \[\leadsto \mathsf{fma}\left(\left(1 \cdot a\right) \cdot 1, a, {\color{blue}{\left(\left(angle \cdot \left(b \cdot \mathsf{PI}\left(\right)\right)\right) \cdot \frac{1}{180}\right)}}^{2}\right) \]
          2. lower-*.f64N/A

            \[\leadsto \mathsf{fma}\left(\left(1 \cdot a\right) \cdot 1, a, {\color{blue}{\left(\left(angle \cdot \left(b \cdot \mathsf{PI}\left(\right)\right)\right) \cdot \frac{1}{180}\right)}}^{2}\right) \]
          3. *-commutativeN/A

            \[\leadsto \mathsf{fma}\left(\left(1 \cdot a\right) \cdot 1, a, {\left(\color{blue}{\left(\left(b \cdot \mathsf{PI}\left(\right)\right) \cdot angle\right)} \cdot \frac{1}{180}\right)}^{2}\right) \]
          4. lower-*.f64N/A

            \[\leadsto \mathsf{fma}\left(\left(1 \cdot a\right) \cdot 1, a, {\left(\color{blue}{\left(\left(b \cdot \mathsf{PI}\left(\right)\right) \cdot angle\right)} \cdot \frac{1}{180}\right)}^{2}\right) \]
          5. lower-*.f64N/A

            \[\leadsto \mathsf{fma}\left(\left(1 \cdot a\right) \cdot 1, a, {\left(\left(\color{blue}{\left(b \cdot \mathsf{PI}\left(\right)\right)} \cdot angle\right) \cdot \frac{1}{180}\right)}^{2}\right) \]
          6. lower-PI.f6472.0

            \[\leadsto \mathsf{fma}\left(\left(1 \cdot a\right) \cdot 1, a, {\left(\left(\left(b \cdot \color{blue}{\mathsf{PI}\left(\right)}\right) \cdot angle\right) \cdot 0.005555555555555556\right)}^{2}\right) \]
        8. Applied rewrites72.0%

          \[\leadsto \mathsf{fma}\left(\left(1 \cdot a\right) \cdot 1, a, {\color{blue}{\left(\left(\left(b \cdot \mathsf{PI}\left(\right)\right) \cdot angle\right) \cdot 0.005555555555555556\right)}}^{2}\right) \]
      5. Recombined 2 regimes into one program.
      6. Add Preprocessing

      Alternative 3: 56.7% accurate, 8.3× speedup?

      \[\begin{array}{l} \\ \begin{array}{l} \mathbf{if}\;a \leq 10^{+90}:\\ \;\;\;\;\mathsf{fma}\left(\mathsf{fma}\left(3.08641975308642 \cdot 10^{-5} \cdot b, b, \left(a \cdot a\right) \cdot -3.08641975308642 \cdot 10^{-5}\right) \cdot \left(\left(\mathsf{PI}\left(\right) \cdot \mathsf{PI}\left(\right)\right) \cdot angle\right), angle, a \cdot a\right)\\ \mathbf{else}:\\ \;\;\;\;a \cdot a\\ \end{array} \end{array} \]
      (FPCore (a b angle)
       :precision binary64
       (if (<= a 1e+90)
         (fma
          (*
           (fma (* 3.08641975308642e-5 b) b (* (* a a) -3.08641975308642e-5))
           (* (* (PI) (PI)) angle))
          angle
          (* a a))
         (* a a)))
      \begin{array}{l}
      
      \\
      \begin{array}{l}
      \mathbf{if}\;a \leq 10^{+90}:\\
      \;\;\;\;\mathsf{fma}\left(\mathsf{fma}\left(3.08641975308642 \cdot 10^{-5} \cdot b, b, \left(a \cdot a\right) \cdot -3.08641975308642 \cdot 10^{-5}\right) \cdot \left(\left(\mathsf{PI}\left(\right) \cdot \mathsf{PI}\left(\right)\right) \cdot angle\right), angle, a \cdot a\right)\\
      
      \mathbf{else}:\\
      \;\;\;\;a \cdot a\\
      
      
      \end{array}
      \end{array}
      
      Derivation
      1. Split input into 2 regimes
      2. if a < 9.99999999999999966e89

        1. Initial program 81.5%

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

          \[\leadsto \color{blue}{{angle}^{2} \cdot \left(\frac{-1}{32400} \cdot \left({a}^{2} \cdot {\mathsf{PI}\left(\right)}^{2}\right) + \frac{1}{32400} \cdot \left({b}^{2} \cdot {\mathsf{PI}\left(\right)}^{2}\right)\right) + {a}^{2}} \]
        4. Step-by-step derivation
          1. *-commutativeN/A

            \[\leadsto \color{blue}{\left(\frac{-1}{32400} \cdot \left({a}^{2} \cdot {\mathsf{PI}\left(\right)}^{2}\right) + \frac{1}{32400} \cdot \left({b}^{2} \cdot {\mathsf{PI}\left(\right)}^{2}\right)\right) \cdot {angle}^{2}} + {a}^{2} \]
          2. lower-fma.f64N/A

            \[\leadsto \color{blue}{\mathsf{fma}\left(\frac{-1}{32400} \cdot \left({a}^{2} \cdot {\mathsf{PI}\left(\right)}^{2}\right) + \frac{1}{32400} \cdot \left({b}^{2} \cdot {\mathsf{PI}\left(\right)}^{2}\right), {angle}^{2}, {a}^{2}\right)} \]
        5. Applied rewrites48.4%

          \[\leadsto \color{blue}{\mathsf{fma}\left(\left(\mathsf{PI}\left(\right) \cdot \mathsf{PI}\left(\right)\right) \cdot \mathsf{fma}\left(3.08641975308642 \cdot 10^{-5}, b \cdot b, -3.08641975308642 \cdot 10^{-5} \cdot \left(a \cdot a\right)\right), angle \cdot angle, a \cdot a\right)} \]
        6. Step-by-step derivation
          1. Applied rewrites53.7%

            \[\leadsto \mathsf{fma}\left(\mathsf{fma}\left(3.08641975308642 \cdot 10^{-5} \cdot b, b, \left(a \cdot a\right) \cdot -3.08641975308642 \cdot 10^{-5}\right) \cdot \left(\left(\mathsf{PI}\left(\right) \cdot \mathsf{PI}\left(\right)\right) \cdot angle\right), \color{blue}{angle}, a \cdot a\right) \]

          if 9.99999999999999966e89 < a

          1. Initial program 80.4%

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

            \[\leadsto \color{blue}{{a}^{2}} \]
          4. Step-by-step derivation
            1. unpow2N/A

              \[\leadsto \color{blue}{a \cdot a} \]
            2. lower-*.f6483.3

              \[\leadsto \color{blue}{a \cdot a} \]
          5. Applied rewrites83.3%

            \[\leadsto \color{blue}{a \cdot a} \]
        7. Recombined 2 regimes into one program.
        8. Add Preprocessing

        Alternative 4: 63.0% accurate, 9.1× speedup?

        \[\begin{array}{l} \\ \begin{array}{l} t_0 := \mathsf{PI}\left(\right) \cdot \mathsf{PI}\left(\right)\\ \mathbf{if}\;b \leq 1.3 \cdot 10^{-143}:\\ \;\;\;\;a \cdot a\\ \mathbf{elif}\;b \leq 5.6 \cdot 10^{+170}:\\ \;\;\;\;\mathsf{fma}\left(t\_0 \cdot \left(3.08641975308642 \cdot 10^{-5} \cdot \left(b \cdot b\right)\right), angle \cdot angle, a \cdot a\right)\\ \mathbf{else}:\\ \;\;\;\;\left(3.08641975308642 \cdot 10^{-5} \cdot \left(\left(\left(angle \cdot angle\right) \cdot b\right) \cdot b\right)\right) \cdot t\_0\\ \end{array} \end{array} \]
        (FPCore (a b angle)
         :precision binary64
         (let* ((t_0 (* (PI) (PI))))
           (if (<= b 1.3e-143)
             (* a a)
             (if (<= b 5.6e+170)
               (fma (* t_0 (* 3.08641975308642e-5 (* b b))) (* angle angle) (* a a))
               (* (* 3.08641975308642e-5 (* (* (* angle angle) b) b)) t_0)))))
        \begin{array}{l}
        
        \\
        \begin{array}{l}
        t_0 := \mathsf{PI}\left(\right) \cdot \mathsf{PI}\left(\right)\\
        \mathbf{if}\;b \leq 1.3 \cdot 10^{-143}:\\
        \;\;\;\;a \cdot a\\
        
        \mathbf{elif}\;b \leq 5.6 \cdot 10^{+170}:\\
        \;\;\;\;\mathsf{fma}\left(t\_0 \cdot \left(3.08641975308642 \cdot 10^{-5} \cdot \left(b \cdot b\right)\right), angle \cdot angle, a \cdot a\right)\\
        
        \mathbf{else}:\\
        \;\;\;\;\left(3.08641975308642 \cdot 10^{-5} \cdot \left(\left(\left(angle \cdot angle\right) \cdot b\right) \cdot b\right)\right) \cdot t\_0\\
        
        
        \end{array}
        \end{array}
        
        Derivation
        1. Split input into 3 regimes
        2. if b < 1.29999999999999994e-143

          1. Initial program 84.5%

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

            \[\leadsto \color{blue}{{a}^{2}} \]
          4. Step-by-step derivation
            1. unpow2N/A

              \[\leadsto \color{blue}{a \cdot a} \]
            2. lower-*.f6460.8

              \[\leadsto \color{blue}{a \cdot a} \]
          5. Applied rewrites60.8%

            \[\leadsto \color{blue}{a \cdot a} \]

          if 1.29999999999999994e-143 < b < 5.6000000000000003e170

          1. Initial program 68.5%

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

            \[\leadsto \color{blue}{{angle}^{2} \cdot \left(\frac{-1}{32400} \cdot \left({a}^{2} \cdot {\mathsf{PI}\left(\right)}^{2}\right) + \frac{1}{32400} \cdot \left({b}^{2} \cdot {\mathsf{PI}\left(\right)}^{2}\right)\right) + {a}^{2}} \]
          4. Step-by-step derivation
            1. *-commutativeN/A

              \[\leadsto \color{blue}{\left(\frac{-1}{32400} \cdot \left({a}^{2} \cdot {\mathsf{PI}\left(\right)}^{2}\right) + \frac{1}{32400} \cdot \left({b}^{2} \cdot {\mathsf{PI}\left(\right)}^{2}\right)\right) \cdot {angle}^{2}} + {a}^{2} \]
            2. lower-fma.f64N/A

              \[\leadsto \color{blue}{\mathsf{fma}\left(\frac{-1}{32400} \cdot \left({a}^{2} \cdot {\mathsf{PI}\left(\right)}^{2}\right) + \frac{1}{32400} \cdot \left({b}^{2} \cdot {\mathsf{PI}\left(\right)}^{2}\right), {angle}^{2}, {a}^{2}\right)} \]
          5. Applied rewrites39.4%

            \[\leadsto \color{blue}{\mathsf{fma}\left(\left(\mathsf{PI}\left(\right) \cdot \mathsf{PI}\left(\right)\right) \cdot \mathsf{fma}\left(3.08641975308642 \cdot 10^{-5}, b \cdot b, -3.08641975308642 \cdot 10^{-5} \cdot \left(a \cdot a\right)\right), angle \cdot angle, a \cdot a\right)} \]
          6. Taylor expanded in a around 0

            \[\leadsto \mathsf{fma}\left(\left(\mathsf{PI}\left(\right) \cdot \mathsf{PI}\left(\right)\right) \cdot \left(\frac{1}{32400} \cdot {b}^{2}\right), angle \cdot angle, a \cdot a\right) \]
          7. Step-by-step derivation
            1. Applied rewrites58.9%

              \[\leadsto \mathsf{fma}\left(\left(\mathsf{PI}\left(\right) \cdot \mathsf{PI}\left(\right)\right) \cdot \left(3.08641975308642 \cdot 10^{-5} \cdot \left(b \cdot b\right)\right), angle \cdot angle, a \cdot a\right) \]

            if 5.6000000000000003e170 < b

            1. Initial program 99.5%

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

              \[\leadsto \color{blue}{{angle}^{2} \cdot \left(\frac{-1}{32400} \cdot \left({a}^{2} \cdot {\mathsf{PI}\left(\right)}^{2}\right) + \frac{1}{32400} \cdot \left({b}^{2} \cdot {\mathsf{PI}\left(\right)}^{2}\right)\right) + {a}^{2}} \]
            4. Step-by-step derivation
              1. *-commutativeN/A

                \[\leadsto \color{blue}{\left(\frac{-1}{32400} \cdot \left({a}^{2} \cdot {\mathsf{PI}\left(\right)}^{2}\right) + \frac{1}{32400} \cdot \left({b}^{2} \cdot {\mathsf{PI}\left(\right)}^{2}\right)\right) \cdot {angle}^{2}} + {a}^{2} \]
              2. lower-fma.f64N/A

                \[\leadsto \color{blue}{\mathsf{fma}\left(\frac{-1}{32400} \cdot \left({a}^{2} \cdot {\mathsf{PI}\left(\right)}^{2}\right) + \frac{1}{32400} \cdot \left({b}^{2} \cdot {\mathsf{PI}\left(\right)}^{2}\right), {angle}^{2}, {a}^{2}\right)} \]
            5. Applied rewrites34.2%

              \[\leadsto \color{blue}{\mathsf{fma}\left(\left(\mathsf{PI}\left(\right) \cdot \mathsf{PI}\left(\right)\right) \cdot \mathsf{fma}\left(3.08641975308642 \cdot 10^{-5}, b \cdot b, -3.08641975308642 \cdot 10^{-5} \cdot \left(a \cdot a\right)\right), angle \cdot angle, a \cdot a\right)} \]
            6. Taylor expanded in a around 0

              \[\leadsto \frac{1}{32400} \cdot \color{blue}{\left({angle}^{2} \cdot \left({b}^{2} \cdot {\mathsf{PI}\left(\right)}^{2}\right)\right)} \]
            7. Step-by-step derivation
              1. Applied rewrites53.7%

                \[\leadsto \left(3.08641975308642 \cdot 10^{-5} \cdot \left(\left(\left(angle \cdot angle\right) \cdot b\right) \cdot b\right)\right) \cdot \color{blue}{\left(\mathsf{PI}\left(\right) \cdot \mathsf{PI}\left(\right)\right)} \]
            8. Recombined 3 regimes into one program.
            9. Add Preprocessing

            Alternative 5: 61.6% accurate, 12.1× speedup?

            \[\begin{array}{l} \\ \begin{array}{l} \mathbf{if}\;b \leq 1.7 \cdot 10^{+147}:\\ \;\;\;\;a \cdot a\\ \mathbf{else}:\\ \;\;\;\;\left(3.08641975308642 \cdot 10^{-5} \cdot \left(\left(\left(angle \cdot angle\right) \cdot b\right) \cdot b\right)\right) \cdot \left(\mathsf{PI}\left(\right) \cdot \mathsf{PI}\left(\right)\right)\\ \end{array} \end{array} \]
            (FPCore (a b angle)
             :precision binary64
             (if (<= b 1.7e+147)
               (* a a)
               (* (* 3.08641975308642e-5 (* (* (* angle angle) b) b)) (* (PI) (PI)))))
            \begin{array}{l}
            
            \\
            \begin{array}{l}
            \mathbf{if}\;b \leq 1.7 \cdot 10^{+147}:\\
            \;\;\;\;a \cdot a\\
            
            \mathbf{else}:\\
            \;\;\;\;\left(3.08641975308642 \cdot 10^{-5} \cdot \left(\left(\left(angle \cdot angle\right) \cdot b\right) \cdot b\right)\right) \cdot \left(\mathsf{PI}\left(\right) \cdot \mathsf{PI}\left(\right)\right)\\
            
            
            \end{array}
            \end{array}
            
            Derivation
            1. Split input into 2 regimes
            2. if b < 1.7e147

              1. Initial program 79.5%

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

                \[\leadsto \color{blue}{{a}^{2}} \]
              4. Step-by-step derivation
                1. unpow2N/A

                  \[\leadsto \color{blue}{a \cdot a} \]
                2. lower-*.f6456.8

                  \[\leadsto \color{blue}{a \cdot a} \]
              5. Applied rewrites56.8%

                \[\leadsto \color{blue}{a \cdot a} \]

              if 1.7e147 < b

              1. Initial program 96.5%

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

                \[\leadsto \color{blue}{{angle}^{2} \cdot \left(\frac{-1}{32400} \cdot \left({a}^{2} \cdot {\mathsf{PI}\left(\right)}^{2}\right) + \frac{1}{32400} \cdot \left({b}^{2} \cdot {\mathsf{PI}\left(\right)}^{2}\right)\right) + {a}^{2}} \]
              4. Step-by-step derivation
                1. *-commutativeN/A

                  \[\leadsto \color{blue}{\left(\frac{-1}{32400} \cdot \left({a}^{2} \cdot {\mathsf{PI}\left(\right)}^{2}\right) + \frac{1}{32400} \cdot \left({b}^{2} \cdot {\mathsf{PI}\left(\right)}^{2}\right)\right) \cdot {angle}^{2}} + {a}^{2} \]
                2. lower-fma.f64N/A

                  \[\leadsto \color{blue}{\mathsf{fma}\left(\frac{-1}{32400} \cdot \left({a}^{2} \cdot {\mathsf{PI}\left(\right)}^{2}\right) + \frac{1}{32400} \cdot \left({b}^{2} \cdot {\mathsf{PI}\left(\right)}^{2}\right), {angle}^{2}, {a}^{2}\right)} \]
              5. Applied rewrites37.0%

                \[\leadsto \color{blue}{\mathsf{fma}\left(\left(\mathsf{PI}\left(\right) \cdot \mathsf{PI}\left(\right)\right) \cdot \mathsf{fma}\left(3.08641975308642 \cdot 10^{-5}, b \cdot b, -3.08641975308642 \cdot 10^{-5} \cdot \left(a \cdot a\right)\right), angle \cdot angle, a \cdot a\right)} \]
              6. Taylor expanded in a around 0

                \[\leadsto \frac{1}{32400} \cdot \color{blue}{\left({angle}^{2} \cdot \left({b}^{2} \cdot {\mathsf{PI}\left(\right)}^{2}\right)\right)} \]
              7. Step-by-step derivation
                1. Applied rewrites51.5%

                  \[\leadsto \left(3.08641975308642 \cdot 10^{-5} \cdot \left(\left(\left(angle \cdot angle\right) \cdot b\right) \cdot b\right)\right) \cdot \color{blue}{\left(\mathsf{PI}\left(\right) \cdot \mathsf{PI}\left(\right)\right)} \]
              8. Recombined 2 regimes into one program.
              9. Add Preprocessing

              Alternative 6: 57.5% accurate, 74.7× speedup?

              \[\begin{array}{l} \\ a \cdot a \end{array} \]
              (FPCore (a b angle) :precision binary64 (* a a))
              double code(double a, double b, double angle) {
              	return a * a;
              }
              
              real(8) function code(a, b, angle)
                  real(8), intent (in) :: a
                  real(8), intent (in) :: b
                  real(8), intent (in) :: angle
                  code = a * a
              end function
              
              public static double code(double a, double b, double angle) {
              	return a * a;
              }
              
              def code(a, b, angle):
              	return a * a
              
              function code(a, b, angle)
              	return Float64(a * a)
              end
              
              function tmp = code(a, b, angle)
              	tmp = a * a;
              end
              
              code[a_, b_, angle_] := N[(a * a), $MachinePrecision]
              
              \begin{array}{l}
              
              \\
              a \cdot a
              \end{array}
              
              Derivation
              1. Initial program 81.4%

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

                \[\leadsto \color{blue}{{a}^{2}} \]
              4. Step-by-step derivation
                1. unpow2N/A

                  \[\leadsto \color{blue}{a \cdot a} \]
                2. lower-*.f6454.2

                  \[\leadsto \color{blue}{a \cdot a} \]
              5. Applied rewrites54.2%

                \[\leadsto \color{blue}{a \cdot a} \]
              6. Add Preprocessing

              Reproduce

              ?
              herbie shell --seed 2024342 
              (FPCore (a b angle)
                :name "ab-angle->ABCF C"
                :precision binary64
                (+ (pow (* a (cos (* (PI) (/ angle 180.0)))) 2.0) (pow (* b (sin (* (PI) (/ angle 180.0)))) 2.0)))