VandenBroeck and Keller, Equation (6)

Percentage Accurate: 76.5% → 99.1%
Time: 7.4s
Alternatives: 9
Speedup: 4.4×

Specification

?
\[\begin{array}{l} \\ \begin{array}{l} t_0 := \mathsf{PI}\left(\right) \cdot \ell\\ t\_0 - \frac{1}{F \cdot F} \cdot \tan t\_0 \end{array} \end{array} \]
(FPCore (F l)
 :precision binary64
 (let* ((t_0 (* (PI) l))) (- t_0 (* (/ 1.0 (* F F)) (tan t_0)))))
\begin{array}{l}

\\
\begin{array}{l}
t_0 := \mathsf{PI}\left(\right) \cdot \ell\\
t\_0 - \frac{1}{F \cdot F} \cdot \tan t\_0
\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 9 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: 76.5% accurate, 1.0× speedup?

\[\begin{array}{l} \\ \begin{array}{l} t_0 := \mathsf{PI}\left(\right) \cdot \ell\\ t\_0 - \frac{1}{F \cdot F} \cdot \tan t\_0 \end{array} \end{array} \]
(FPCore (F l)
 :precision binary64
 (let* ((t_0 (* (PI) l))) (- t_0 (* (/ 1.0 (* F F)) (tan t_0)))))
\begin{array}{l}

\\
\begin{array}{l}
t_0 := \mathsf{PI}\left(\right) \cdot \ell\\
t\_0 - \frac{1}{F \cdot F} \cdot \tan t\_0
\end{array}
\end{array}

Alternative 1: 99.1% accurate, 0.6× speedup?

\[\begin{array}{l} l\_m = \left|\ell\right| \\ l\_s = \mathsf{copysign}\left(1, \ell\right) \\ l\_s \cdot \begin{array}{l} \mathbf{if}\;l\_m \leq 9.5 \cdot 10^{+14}:\\ \;\;\;\;\mathsf{PI}\left(\right) \cdot l\_m - {F}^{-1} \cdot \frac{\tan \left(l\_m \cdot \mathsf{PI}\left(\right)\right)}{F}\\ \mathbf{else}:\\ \;\;\;\;{\mathsf{PI}\left(\right)}^{0.25} \cdot \left({\mathsf{PI}\left(\right)}^{0.75} \cdot l\_m\right)\\ \end{array} \end{array} \]
l\_m = (fabs.f64 l)
l\_s = (copysign.f64 #s(literal 1 binary64) l)
(FPCore (l_s F l_m)
 :precision binary64
 (*
  l_s
  (if (<= l_m 9.5e+14)
    (- (* (PI) l_m) (* (pow F -1.0) (/ (tan (* l_m (PI))) F)))
    (* (pow (PI) 0.25) (* (pow (PI) 0.75) l_m)))))
\begin{array}{l}
l\_m = \left|\ell\right|
\\
l\_s = \mathsf{copysign}\left(1, \ell\right)

\\
l\_s \cdot \begin{array}{l}
\mathbf{if}\;l\_m \leq 9.5 \cdot 10^{+14}:\\
\;\;\;\;\mathsf{PI}\left(\right) \cdot l\_m - {F}^{-1} \cdot \frac{\tan \left(l\_m \cdot \mathsf{PI}\left(\right)\right)}{F}\\

\mathbf{else}:\\
\;\;\;\;{\mathsf{PI}\left(\right)}^{0.25} \cdot \left({\mathsf{PI}\left(\right)}^{0.75} \cdot l\_m\right)\\


\end{array}
\end{array}
Derivation
  1. Split input into 2 regimes
  2. if l < 9.5e14

    1. Initial program 82.7%

      \[\mathsf{PI}\left(\right) \cdot \ell - \frac{1}{F \cdot F} \cdot \tan \left(\mathsf{PI}\left(\right) \cdot \ell\right) \]
    2. Add Preprocessing
    3. Step-by-step derivation
      1. lift-*.f64N/A

        \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \color{blue}{\frac{1}{F \cdot F} \cdot \tan \left(\mathsf{PI}\left(\right) \cdot \ell\right)} \]
      2. lift-/.f64N/A

        \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \color{blue}{\frac{1}{F \cdot F}} \cdot \tan \left(\mathsf{PI}\left(\right) \cdot \ell\right) \]
      3. associate-*l/N/A

        \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \color{blue}{\frac{1 \cdot \tan \left(\mathsf{PI}\left(\right) \cdot \ell\right)}{F \cdot F}} \]
      4. lift-*.f64N/A

        \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \frac{1 \cdot \tan \left(\mathsf{PI}\left(\right) \cdot \ell\right)}{\color{blue}{F \cdot F}} \]
      5. times-fracN/A

        \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \color{blue}{\frac{1}{F} \cdot \frac{\tan \left(\mathsf{PI}\left(\right) \cdot \ell\right)}{F}} \]
      6. lower-*.f64N/A

        \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \color{blue}{\frac{1}{F} \cdot \frac{\tan \left(\mathsf{PI}\left(\right) \cdot \ell\right)}{F}} \]
      7. inv-powN/A

        \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \color{blue}{{F}^{-1}} \cdot \frac{\tan \left(\mathsf{PI}\left(\right) \cdot \ell\right)}{F} \]
      8. lower-pow.f64N/A

        \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \color{blue}{{F}^{-1}} \cdot \frac{\tan \left(\mathsf{PI}\left(\right) \cdot \ell\right)}{F} \]
      9. lower-/.f6488.7

        \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - {F}^{-1} \cdot \color{blue}{\frac{\tan \left(\mathsf{PI}\left(\right) \cdot \ell\right)}{F}} \]
      10. lift-*.f64N/A

        \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - {F}^{-1} \cdot \frac{\tan \color{blue}{\left(\mathsf{PI}\left(\right) \cdot \ell\right)}}{F} \]
      11. *-commutativeN/A

        \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - {F}^{-1} \cdot \frac{\tan \color{blue}{\left(\ell \cdot \mathsf{PI}\left(\right)\right)}}{F} \]
      12. lower-*.f6488.7

        \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - {F}^{-1} \cdot \frac{\tan \color{blue}{\left(\ell \cdot \mathsf{PI}\left(\right)\right)}}{F} \]
    4. Applied rewrites88.7%

      \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \color{blue}{{F}^{-1} \cdot \frac{\tan \left(\ell \cdot \mathsf{PI}\left(\right)\right)}{F}} \]

    if 9.5e14 < l

    1. Initial program 54.3%

      \[\mathsf{PI}\left(\right) \cdot \ell - \frac{1}{F \cdot F} \cdot \tan \left(\mathsf{PI}\left(\right) \cdot \ell\right) \]
    2. Add Preprocessing
    3. Taylor expanded in F around inf

      \[\leadsto \color{blue}{\ell \cdot \mathsf{PI}\left(\right)} \]
    4. Step-by-step derivation
      1. Applied rewrites99.4%

        \[\leadsto \color{blue}{\mathsf{PI}\left(\right) \cdot \ell} \]
      2. Step-by-step derivation
        1. Applied rewrites99.6%

          \[\leadsto {\mathsf{PI}\left(\right)}^{0.25} \cdot \color{blue}{\left({\mathsf{PI}\left(\right)}^{0.75} \cdot \ell\right)} \]
      3. Recombined 2 regimes into one program.
      4. Add Preprocessing

      Alternative 2: 99.1% accurate, 0.6× speedup?

      \[\begin{array}{l} l\_m = \left|\ell\right| \\ l\_s = \mathsf{copysign}\left(1, \ell\right) \\ l\_s \cdot \begin{array}{l} \mathbf{if}\;l\_m \leq 9.5 \cdot 10^{+14}:\\ \;\;\;\;\mathsf{PI}\left(\right) \cdot l\_m - \frac{\frac{\tan \left(l\_m \cdot \mathsf{PI}\left(\right)\right)}{F}}{F}\\ \mathbf{else}:\\ \;\;\;\;{\mathsf{PI}\left(\right)}^{0.25} \cdot \left({\mathsf{PI}\left(\right)}^{0.75} \cdot l\_m\right)\\ \end{array} \end{array} \]
      l\_m = (fabs.f64 l)
      l\_s = (copysign.f64 #s(literal 1 binary64) l)
      (FPCore (l_s F l_m)
       :precision binary64
       (*
        l_s
        (if (<= l_m 9.5e+14)
          (- (* (PI) l_m) (/ (/ (tan (* l_m (PI))) F) F))
          (* (pow (PI) 0.25) (* (pow (PI) 0.75) l_m)))))
      \begin{array}{l}
      l\_m = \left|\ell\right|
      \\
      l\_s = \mathsf{copysign}\left(1, \ell\right)
      
      \\
      l\_s \cdot \begin{array}{l}
      \mathbf{if}\;l\_m \leq 9.5 \cdot 10^{+14}:\\
      \;\;\;\;\mathsf{PI}\left(\right) \cdot l\_m - \frac{\frac{\tan \left(l\_m \cdot \mathsf{PI}\left(\right)\right)}{F}}{F}\\
      
      \mathbf{else}:\\
      \;\;\;\;{\mathsf{PI}\left(\right)}^{0.25} \cdot \left({\mathsf{PI}\left(\right)}^{0.75} \cdot l\_m\right)\\
      
      
      \end{array}
      \end{array}
      
      Derivation
      1. Split input into 2 regimes
      2. if l < 9.5e14

        1. Initial program 82.7%

          \[\mathsf{PI}\left(\right) \cdot \ell - \frac{1}{F \cdot F} \cdot \tan \left(\mathsf{PI}\left(\right) \cdot \ell\right) \]
        2. Add Preprocessing
        3. Step-by-step derivation
          1. lift-*.f64N/A

            \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \color{blue}{\frac{1}{F \cdot F} \cdot \tan \left(\mathsf{PI}\left(\right) \cdot \ell\right)} \]
          2. lift-/.f64N/A

            \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \color{blue}{\frac{1}{F \cdot F}} \cdot \tan \left(\mathsf{PI}\left(\right) \cdot \ell\right) \]
          3. associate-*l/N/A

            \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \color{blue}{\frac{1 \cdot \tan \left(\mathsf{PI}\left(\right) \cdot \ell\right)}{F \cdot F}} \]
          4. lift-*.f64N/A

            \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \frac{1 \cdot \tan \left(\mathsf{PI}\left(\right) \cdot \ell\right)}{\color{blue}{F \cdot F}} \]
          5. *-lft-identityN/A

            \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \frac{\color{blue}{\tan \left(\mathsf{PI}\left(\right) \cdot \ell\right)}}{F \cdot F} \]
          6. associate-/r*N/A

            \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \color{blue}{\frac{\frac{\tan \left(\mathsf{PI}\left(\right) \cdot \ell\right)}{F}}{F}} \]
          7. lower-/.f64N/A

            \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \color{blue}{\frac{\frac{\tan \left(\mathsf{PI}\left(\right) \cdot \ell\right)}{F}}{F}} \]
          8. lower-/.f6488.7

            \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \frac{\color{blue}{\frac{\tan \left(\mathsf{PI}\left(\right) \cdot \ell\right)}{F}}}{F} \]
          9. lift-*.f64N/A

            \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \frac{\frac{\tan \color{blue}{\left(\mathsf{PI}\left(\right) \cdot \ell\right)}}{F}}{F} \]
          10. *-commutativeN/A

            \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \frac{\frac{\tan \color{blue}{\left(\ell \cdot \mathsf{PI}\left(\right)\right)}}{F}}{F} \]
          11. lower-*.f6488.7

            \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \frac{\frac{\tan \color{blue}{\left(\ell \cdot \mathsf{PI}\left(\right)\right)}}{F}}{F} \]
        4. Applied rewrites88.7%

          \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \color{blue}{\frac{\frac{\tan \left(\ell \cdot \mathsf{PI}\left(\right)\right)}{F}}{F}} \]

        if 9.5e14 < l

        1. Initial program 54.3%

          \[\mathsf{PI}\left(\right) \cdot \ell - \frac{1}{F \cdot F} \cdot \tan \left(\mathsf{PI}\left(\right) \cdot \ell\right) \]
        2. Add Preprocessing
        3. Taylor expanded in F around inf

          \[\leadsto \color{blue}{\ell \cdot \mathsf{PI}\left(\right)} \]
        4. Step-by-step derivation
          1. Applied rewrites99.4%

            \[\leadsto \color{blue}{\mathsf{PI}\left(\right) \cdot \ell} \]
          2. Step-by-step derivation
            1. Applied rewrites99.6%

              \[\leadsto {\mathsf{PI}\left(\right)}^{0.25} \cdot \color{blue}{\left({\mathsf{PI}\left(\right)}^{0.75} \cdot \ell\right)} \]
          3. Recombined 2 regimes into one program.
          4. Add Preprocessing

          Alternative 3: 99.1% accurate, 1.0× speedup?

          \[\begin{array}{l} l\_m = \left|\ell\right| \\ l\_s = \mathsf{copysign}\left(1, \ell\right) \\ \begin{array}{l} t_0 := \mathsf{PI}\left(\right) \cdot l\_m\\ l\_s \cdot \begin{array}{l} \mathbf{if}\;l\_m \leq 9.5 \cdot 10^{+14}:\\ \;\;\;\;t\_0 - \frac{\frac{\tan \left(l\_m \cdot \mathsf{PI}\left(\right)\right)}{F}}{F}\\ \mathbf{else}:\\ \;\;\;\;t\_0\\ \end{array} \end{array} \end{array} \]
          l\_m = (fabs.f64 l)
          l\_s = (copysign.f64 #s(literal 1 binary64) l)
          (FPCore (l_s F l_m)
           :precision binary64
           (let* ((t_0 (* (PI) l_m)))
             (* l_s (if (<= l_m 9.5e+14) (- t_0 (/ (/ (tan (* l_m (PI))) F) F)) t_0))))
          \begin{array}{l}
          l\_m = \left|\ell\right|
          \\
          l\_s = \mathsf{copysign}\left(1, \ell\right)
          
          \\
          \begin{array}{l}
          t_0 := \mathsf{PI}\left(\right) \cdot l\_m\\
          l\_s \cdot \begin{array}{l}
          \mathbf{if}\;l\_m \leq 9.5 \cdot 10^{+14}:\\
          \;\;\;\;t\_0 - \frac{\frac{\tan \left(l\_m \cdot \mathsf{PI}\left(\right)\right)}{F}}{F}\\
          
          \mathbf{else}:\\
          \;\;\;\;t\_0\\
          
          
          \end{array}
          \end{array}
          \end{array}
          
          Derivation
          1. Split input into 2 regimes
          2. if l < 9.5e14

            1. Initial program 82.7%

              \[\mathsf{PI}\left(\right) \cdot \ell - \frac{1}{F \cdot F} \cdot \tan \left(\mathsf{PI}\left(\right) \cdot \ell\right) \]
            2. Add Preprocessing
            3. Step-by-step derivation
              1. lift-*.f64N/A

                \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \color{blue}{\frac{1}{F \cdot F} \cdot \tan \left(\mathsf{PI}\left(\right) \cdot \ell\right)} \]
              2. lift-/.f64N/A

                \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \color{blue}{\frac{1}{F \cdot F}} \cdot \tan \left(\mathsf{PI}\left(\right) \cdot \ell\right) \]
              3. associate-*l/N/A

                \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \color{blue}{\frac{1 \cdot \tan \left(\mathsf{PI}\left(\right) \cdot \ell\right)}{F \cdot F}} \]
              4. lift-*.f64N/A

                \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \frac{1 \cdot \tan \left(\mathsf{PI}\left(\right) \cdot \ell\right)}{\color{blue}{F \cdot F}} \]
              5. *-lft-identityN/A

                \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \frac{\color{blue}{\tan \left(\mathsf{PI}\left(\right) \cdot \ell\right)}}{F \cdot F} \]
              6. associate-/r*N/A

                \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \color{blue}{\frac{\frac{\tan \left(\mathsf{PI}\left(\right) \cdot \ell\right)}{F}}{F}} \]
              7. lower-/.f64N/A

                \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \color{blue}{\frac{\frac{\tan \left(\mathsf{PI}\left(\right) \cdot \ell\right)}{F}}{F}} \]
              8. lower-/.f6488.7

                \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \frac{\color{blue}{\frac{\tan \left(\mathsf{PI}\left(\right) \cdot \ell\right)}{F}}}{F} \]
              9. lift-*.f64N/A

                \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \frac{\frac{\tan \color{blue}{\left(\mathsf{PI}\left(\right) \cdot \ell\right)}}{F}}{F} \]
              10. *-commutativeN/A

                \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \frac{\frac{\tan \color{blue}{\left(\ell \cdot \mathsf{PI}\left(\right)\right)}}{F}}{F} \]
              11. lower-*.f6488.7

                \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \frac{\frac{\tan \color{blue}{\left(\ell \cdot \mathsf{PI}\left(\right)\right)}}{F}}{F} \]
            4. Applied rewrites88.7%

              \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \color{blue}{\frac{\frac{\tan \left(\ell \cdot \mathsf{PI}\left(\right)\right)}{F}}{F}} \]

            if 9.5e14 < l

            1. Initial program 54.3%

              \[\mathsf{PI}\left(\right) \cdot \ell - \frac{1}{F \cdot F} \cdot \tan \left(\mathsf{PI}\left(\right) \cdot \ell\right) \]
            2. Add Preprocessing
            3. Taylor expanded in F around inf

              \[\leadsto \color{blue}{\ell \cdot \mathsf{PI}\left(\right)} \]
            4. Step-by-step derivation
              1. Applied rewrites99.4%

                \[\leadsto \color{blue}{\mathsf{PI}\left(\right) \cdot \ell} \]
            5. Recombined 2 regimes into one program.
            6. Add Preprocessing

            Alternative 4: 98.4% accurate, 3.0× speedup?

            \[\begin{array}{l} l\_m = \left|\ell\right| \\ l\_s = \mathsf{copysign}\left(1, \ell\right) \\ l\_s \cdot \begin{array}{l} \mathbf{if}\;l\_m \leq 8 \cdot 10^{+14}:\\ \;\;\;\;\mathsf{fma}\left(\frac{-1}{F}, \frac{\mathsf{PI}\left(\right)}{F} \cdot l\_m, l\_m \cdot \mathsf{PI}\left(\right)\right)\\ \mathbf{else}:\\ \;\;\;\;\mathsf{PI}\left(\right) \cdot l\_m\\ \end{array} \end{array} \]
            l\_m = (fabs.f64 l)
            l\_s = (copysign.f64 #s(literal 1 binary64) l)
            (FPCore (l_s F l_m)
             :precision binary64
             (*
              l_s
              (if (<= l_m 8e+14)
                (fma (/ -1.0 F) (* (/ (PI) F) l_m) (* l_m (PI)))
                (* (PI) l_m))))
            \begin{array}{l}
            l\_m = \left|\ell\right|
            \\
            l\_s = \mathsf{copysign}\left(1, \ell\right)
            
            \\
            l\_s \cdot \begin{array}{l}
            \mathbf{if}\;l\_m \leq 8 \cdot 10^{+14}:\\
            \;\;\;\;\mathsf{fma}\left(\frac{-1}{F}, \frac{\mathsf{PI}\left(\right)}{F} \cdot l\_m, l\_m \cdot \mathsf{PI}\left(\right)\right)\\
            
            \mathbf{else}:\\
            \;\;\;\;\mathsf{PI}\left(\right) \cdot l\_m\\
            
            
            \end{array}
            \end{array}
            
            Derivation
            1. Split input into 2 regimes
            2. if l < 8e14

              1. Initial program 82.7%

                \[\mathsf{PI}\left(\right) \cdot \ell - \frac{1}{F \cdot F} \cdot \tan \left(\mathsf{PI}\left(\right) \cdot \ell\right) \]
              2. Add Preprocessing
              3. Step-by-step derivation
                1. lift-*.f64N/A

                  \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \color{blue}{\frac{1}{F \cdot F} \cdot \tan \left(\mathsf{PI}\left(\right) \cdot \ell\right)} \]
                2. lift-/.f64N/A

                  \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \color{blue}{\frac{1}{F \cdot F}} \cdot \tan \left(\mathsf{PI}\left(\right) \cdot \ell\right) \]
                3. associate-*l/N/A

                  \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \color{blue}{\frac{1 \cdot \tan \left(\mathsf{PI}\left(\right) \cdot \ell\right)}{F \cdot F}} \]
                4. lift-*.f64N/A

                  \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \frac{1 \cdot \tan \left(\mathsf{PI}\left(\right) \cdot \ell\right)}{\color{blue}{F \cdot F}} \]
                5. times-fracN/A

                  \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \color{blue}{\frac{1}{F} \cdot \frac{\tan \left(\mathsf{PI}\left(\right) \cdot \ell\right)}{F}} \]
                6. lower-*.f64N/A

                  \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \color{blue}{\frac{1}{F} \cdot \frac{\tan \left(\mathsf{PI}\left(\right) \cdot \ell\right)}{F}} \]
                7. inv-powN/A

                  \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \color{blue}{{F}^{-1}} \cdot \frac{\tan \left(\mathsf{PI}\left(\right) \cdot \ell\right)}{F} \]
                8. lower-pow.f64N/A

                  \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \color{blue}{{F}^{-1}} \cdot \frac{\tan \left(\mathsf{PI}\left(\right) \cdot \ell\right)}{F} \]
                9. lower-/.f6488.7

                  \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - {F}^{-1} \cdot \color{blue}{\frac{\tan \left(\mathsf{PI}\left(\right) \cdot \ell\right)}{F}} \]
                10. lift-*.f64N/A

                  \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - {F}^{-1} \cdot \frac{\tan \color{blue}{\left(\mathsf{PI}\left(\right) \cdot \ell\right)}}{F} \]
                11. *-commutativeN/A

                  \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - {F}^{-1} \cdot \frac{\tan \color{blue}{\left(\ell \cdot \mathsf{PI}\left(\right)\right)}}{F} \]
                12. lower-*.f6488.7

                  \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - {F}^{-1} \cdot \frac{\tan \color{blue}{\left(\ell \cdot \mathsf{PI}\left(\right)\right)}}{F} \]
              4. Applied rewrites88.7%

                \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \color{blue}{{F}^{-1} \cdot \frac{\tan \left(\ell \cdot \mathsf{PI}\left(\right)\right)}{F}} \]
              5. Taylor expanded in l around 0

                \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - {F}^{-1} \cdot \color{blue}{\frac{\ell \cdot \mathsf{PI}\left(\right)}{F}} \]
              6. Step-by-step derivation
                1. Applied rewrites83.7%

                  \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - {F}^{-1} \cdot \color{blue}{\left(\frac{\mathsf{PI}\left(\right)}{F} \cdot \ell\right)} \]
                2. Step-by-step derivation
                  1. lift--.f64N/A

                    \[\leadsto \color{blue}{\mathsf{PI}\left(\right) \cdot \ell - {F}^{-1} \cdot \left(\frac{\mathsf{PI}\left(\right)}{F} \cdot \ell\right)} \]
                  2. lift-*.f64N/A

                    \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \color{blue}{{F}^{-1} \cdot \left(\frac{\mathsf{PI}\left(\right)}{F} \cdot \ell\right)} \]
                  3. fp-cancel-sub-sign-invN/A

                    \[\leadsto \color{blue}{\mathsf{PI}\left(\right) \cdot \ell + \left(\mathsf{neg}\left({F}^{-1}\right)\right) \cdot \left(\frac{\mathsf{PI}\left(\right)}{F} \cdot \ell\right)} \]
                  4. +-commutativeN/A

                    \[\leadsto \color{blue}{\left(\mathsf{neg}\left({F}^{-1}\right)\right) \cdot \left(\frac{\mathsf{PI}\left(\right)}{F} \cdot \ell\right) + \mathsf{PI}\left(\right) \cdot \ell} \]
                  5. lower-fma.f64N/A

                    \[\leadsto \color{blue}{\mathsf{fma}\left(\mathsf{neg}\left({F}^{-1}\right), \frac{\mathsf{PI}\left(\right)}{F} \cdot \ell, \mathsf{PI}\left(\right) \cdot \ell\right)} \]
                  6. lift-pow.f64N/A

                    \[\leadsto \mathsf{fma}\left(\mathsf{neg}\left(\color{blue}{{F}^{-1}}\right), \frac{\mathsf{PI}\left(\right)}{F} \cdot \ell, \mathsf{PI}\left(\right) \cdot \ell\right) \]
                  7. unpow-1N/A

                    \[\leadsto \mathsf{fma}\left(\mathsf{neg}\left(\color{blue}{\frac{1}{F}}\right), \frac{\mathsf{PI}\left(\right)}{F} \cdot \ell, \mathsf{PI}\left(\right) \cdot \ell\right) \]
                  8. distribute-neg-fracN/A

                    \[\leadsto \mathsf{fma}\left(\color{blue}{\frac{\mathsf{neg}\left(1\right)}{F}}, \frac{\mathsf{PI}\left(\right)}{F} \cdot \ell, \mathsf{PI}\left(\right) \cdot \ell\right) \]
                  9. metadata-evalN/A

                    \[\leadsto \mathsf{fma}\left(\frac{\color{blue}{-1}}{F}, \frac{\mathsf{PI}\left(\right)}{F} \cdot \ell, \mathsf{PI}\left(\right) \cdot \ell\right) \]
                  10. lower-/.f6483.7

                    \[\leadsto \mathsf{fma}\left(\color{blue}{\frac{-1}{F}}, \frac{\mathsf{PI}\left(\right)}{F} \cdot \ell, \mathsf{PI}\left(\right) \cdot \ell\right) \]
                  11. lift-*.f64N/A

                    \[\leadsto \mathsf{fma}\left(\frac{-1}{F}, \frac{\mathsf{PI}\left(\right)}{F} \cdot \ell, \color{blue}{\mathsf{PI}\left(\right) \cdot \ell}\right) \]
                  12. *-commutativeN/A

                    \[\leadsto \mathsf{fma}\left(\frac{-1}{F}, \frac{\mathsf{PI}\left(\right)}{F} \cdot \ell, \color{blue}{\ell \cdot \mathsf{PI}\left(\right)}\right) \]
                  13. lower-*.f6483.7

                    \[\leadsto \mathsf{fma}\left(\frac{-1}{F}, \frac{\mathsf{PI}\left(\right)}{F} \cdot \ell, \color{blue}{\ell \cdot \mathsf{PI}\left(\right)}\right) \]
                3. Applied rewrites83.7%

                  \[\leadsto \color{blue}{\mathsf{fma}\left(\frac{-1}{F}, \frac{\mathsf{PI}\left(\right)}{F} \cdot \ell, \ell \cdot \mathsf{PI}\left(\right)\right)} \]

                if 8e14 < l

                1. Initial program 54.3%

                  \[\mathsf{PI}\left(\right) \cdot \ell - \frac{1}{F \cdot F} \cdot \tan \left(\mathsf{PI}\left(\right) \cdot \ell\right) \]
                2. Add Preprocessing
                3. Taylor expanded in F around inf

                  \[\leadsto \color{blue}{\ell \cdot \mathsf{PI}\left(\right)} \]
                4. Step-by-step derivation
                  1. Applied rewrites99.4%

                    \[\leadsto \color{blue}{\mathsf{PI}\left(\right) \cdot \ell} \]
                5. Recombined 2 regimes into one program.
                6. Add Preprocessing

                Alternative 5: 98.4% accurate, 3.2× speedup?

                \[\begin{array}{l} l\_m = \left|\ell\right| \\ l\_s = \mathsf{copysign}\left(1, \ell\right) \\ \begin{array}{l} t_0 := \mathsf{PI}\left(\right) \cdot l\_m\\ l\_s \cdot \begin{array}{l} \mathbf{if}\;l\_m \leq 8 \cdot 10^{+14}:\\ \;\;\;\;t\_0 - \frac{\frac{\mathsf{PI}\left(\right)}{F} \cdot l\_m}{F}\\ \mathbf{else}:\\ \;\;\;\;t\_0\\ \end{array} \end{array} \end{array} \]
                l\_m = (fabs.f64 l)
                l\_s = (copysign.f64 #s(literal 1 binary64) l)
                (FPCore (l_s F l_m)
                 :precision binary64
                 (let* ((t_0 (* (PI) l_m)))
                   (* l_s (if (<= l_m 8e+14) (- t_0 (/ (* (/ (PI) F) l_m) F)) t_0))))
                \begin{array}{l}
                l\_m = \left|\ell\right|
                \\
                l\_s = \mathsf{copysign}\left(1, \ell\right)
                
                \\
                \begin{array}{l}
                t_0 := \mathsf{PI}\left(\right) \cdot l\_m\\
                l\_s \cdot \begin{array}{l}
                \mathbf{if}\;l\_m \leq 8 \cdot 10^{+14}:\\
                \;\;\;\;t\_0 - \frac{\frac{\mathsf{PI}\left(\right)}{F} \cdot l\_m}{F}\\
                
                \mathbf{else}:\\
                \;\;\;\;t\_0\\
                
                
                \end{array}
                \end{array}
                \end{array}
                
                Derivation
                1. Split input into 2 regimes
                2. if l < 8e14

                  1. Initial program 82.7%

                    \[\mathsf{PI}\left(\right) \cdot \ell - \frac{1}{F \cdot F} \cdot \tan \left(\mathsf{PI}\left(\right) \cdot \ell\right) \]
                  2. Add Preprocessing
                  3. Step-by-step derivation
                    1. lift-*.f64N/A

                      \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \color{blue}{\frac{1}{F \cdot F} \cdot \tan \left(\mathsf{PI}\left(\right) \cdot \ell\right)} \]
                    2. lift-/.f64N/A

                      \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \color{blue}{\frac{1}{F \cdot F}} \cdot \tan \left(\mathsf{PI}\left(\right) \cdot \ell\right) \]
                    3. associate-*l/N/A

                      \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \color{blue}{\frac{1 \cdot \tan \left(\mathsf{PI}\left(\right) \cdot \ell\right)}{F \cdot F}} \]
                    4. lift-*.f64N/A

                      \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \frac{1 \cdot \tan \left(\mathsf{PI}\left(\right) \cdot \ell\right)}{\color{blue}{F \cdot F}} \]
                    5. times-fracN/A

                      \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \color{blue}{\frac{1}{F} \cdot \frac{\tan \left(\mathsf{PI}\left(\right) \cdot \ell\right)}{F}} \]
                    6. lower-*.f64N/A

                      \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \color{blue}{\frac{1}{F} \cdot \frac{\tan \left(\mathsf{PI}\left(\right) \cdot \ell\right)}{F}} \]
                    7. inv-powN/A

                      \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \color{blue}{{F}^{-1}} \cdot \frac{\tan \left(\mathsf{PI}\left(\right) \cdot \ell\right)}{F} \]
                    8. lower-pow.f64N/A

                      \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \color{blue}{{F}^{-1}} \cdot \frac{\tan \left(\mathsf{PI}\left(\right) \cdot \ell\right)}{F} \]
                    9. lower-/.f6488.7

                      \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - {F}^{-1} \cdot \color{blue}{\frac{\tan \left(\mathsf{PI}\left(\right) \cdot \ell\right)}{F}} \]
                    10. lift-*.f64N/A

                      \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - {F}^{-1} \cdot \frac{\tan \color{blue}{\left(\mathsf{PI}\left(\right) \cdot \ell\right)}}{F} \]
                    11. *-commutativeN/A

                      \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - {F}^{-1} \cdot \frac{\tan \color{blue}{\left(\ell \cdot \mathsf{PI}\left(\right)\right)}}{F} \]
                    12. lower-*.f6488.7

                      \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - {F}^{-1} \cdot \frac{\tan \color{blue}{\left(\ell \cdot \mathsf{PI}\left(\right)\right)}}{F} \]
                  4. Applied rewrites88.7%

                    \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \color{blue}{{F}^{-1} \cdot \frac{\tan \left(\ell \cdot \mathsf{PI}\left(\right)\right)}{F}} \]
                  5. Step-by-step derivation
                    1. lift-*.f64N/A

                      \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \color{blue}{{F}^{-1} \cdot \frac{\tan \left(\ell \cdot \mathsf{PI}\left(\right)\right)}{F}} \]
                    2. *-commutativeN/A

                      \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \color{blue}{\frac{\tan \left(\ell \cdot \mathsf{PI}\left(\right)\right)}{F} \cdot {F}^{-1}} \]
                    3. lift-/.f64N/A

                      \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \color{blue}{\frac{\tan \left(\ell \cdot \mathsf{PI}\left(\right)\right)}{F}} \cdot {F}^{-1} \]
                    4. associate-*l/N/A

                      \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \color{blue}{\frac{\tan \left(\ell \cdot \mathsf{PI}\left(\right)\right) \cdot {F}^{-1}}{F}} \]
                    5. *-commutativeN/A

                      \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \frac{\color{blue}{{F}^{-1} \cdot \tan \left(\ell \cdot \mathsf{PI}\left(\right)\right)}}{F} \]
                    6. lift-pow.f64N/A

                      \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \frac{\color{blue}{{F}^{-1}} \cdot \tan \left(\ell \cdot \mathsf{PI}\left(\right)\right)}{F} \]
                    7. unpow-1N/A

                      \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \frac{\color{blue}{\frac{1}{F}} \cdot \tan \left(\ell \cdot \mathsf{PI}\left(\right)\right)}{F} \]
                    8. associate-*l/N/A

                      \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \frac{\color{blue}{\frac{1 \cdot \tan \left(\ell \cdot \mathsf{PI}\left(\right)\right)}{F}}}{F} \]
                    9. *-lft-identityN/A

                      \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \frac{\frac{\color{blue}{\tan \left(\ell \cdot \mathsf{PI}\left(\right)\right)}}{F}}{F} \]
                    10. lift-/.f64N/A

                      \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \frac{\color{blue}{\frac{\tan \left(\ell \cdot \mathsf{PI}\left(\right)\right)}{F}}}{F} \]
                    11. lower-/.f6488.7

                      \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \color{blue}{\frac{\frac{\tan \left(\ell \cdot \mathsf{PI}\left(\right)\right)}{F}}{F}} \]
                  6. Applied rewrites88.7%

                    \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \color{blue}{\frac{\frac{\tan \left(\mathsf{PI}\left(\right) \cdot \ell\right)}{F}}{F}} \]
                  7. Taylor expanded in l around 0

                    \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \frac{\color{blue}{\frac{\ell \cdot \mathsf{PI}\left(\right)}{F}}}{F} \]
                  8. Step-by-step derivation
                    1. Applied rewrites83.7%

                      \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \frac{\color{blue}{\frac{\mathsf{PI}\left(\right)}{F} \cdot \ell}}{F} \]

                    if 8e14 < l

                    1. Initial program 54.3%

                      \[\mathsf{PI}\left(\right) \cdot \ell - \frac{1}{F \cdot F} \cdot \tan \left(\mathsf{PI}\left(\right) \cdot \ell\right) \]
                    2. Add Preprocessing
                    3. Taylor expanded in F around inf

                      \[\leadsto \color{blue}{\ell \cdot \mathsf{PI}\left(\right)} \]
                    4. Step-by-step derivation
                      1. Applied rewrites99.4%

                        \[\leadsto \color{blue}{\mathsf{PI}\left(\right) \cdot \ell} \]
                    5. Recombined 2 regimes into one program.
                    6. Final simplification88.1%

                      \[\leadsto \begin{array}{l} \mathbf{if}\;\ell \leq 8 \cdot 10^{+14}:\\ \;\;\;\;\mathsf{PI}\left(\right) \cdot \ell - \frac{\frac{\mathsf{PI}\left(\right)}{F} \cdot \ell}{F}\\ \mathbf{else}:\\ \;\;\;\;\mathsf{PI}\left(\right) \cdot \ell\\ \end{array} \]
                    7. Add Preprocessing

                    Alternative 6: 92.6% accurate, 3.6× speedup?

                    \[\begin{array}{l} l\_m = \left|\ell\right| \\ l\_s = \mathsf{copysign}\left(1, \ell\right) \\ l\_s \cdot \begin{array}{l} \mathbf{if}\;l\_m \leq 8 \cdot 10^{+14}:\\ \;\;\;\;\left(\mathsf{PI}\left(\right) - \frac{\frac{\mathsf{PI}\left(\right)}{F}}{F}\right) \cdot l\_m\\ \mathbf{else}:\\ \;\;\;\;\mathsf{PI}\left(\right) \cdot l\_m\\ \end{array} \end{array} \]
                    l\_m = (fabs.f64 l)
                    l\_s = (copysign.f64 #s(literal 1 binary64) l)
                    (FPCore (l_s F l_m)
                     :precision binary64
                     (* l_s (if (<= l_m 8e+14) (* (- (PI) (/ (/ (PI) F) F)) l_m) (* (PI) l_m))))
                    \begin{array}{l}
                    l\_m = \left|\ell\right|
                    \\
                    l\_s = \mathsf{copysign}\left(1, \ell\right)
                    
                    \\
                    l\_s \cdot \begin{array}{l}
                    \mathbf{if}\;l\_m \leq 8 \cdot 10^{+14}:\\
                    \;\;\;\;\left(\mathsf{PI}\left(\right) - \frac{\frac{\mathsf{PI}\left(\right)}{F}}{F}\right) \cdot l\_m\\
                    
                    \mathbf{else}:\\
                    \;\;\;\;\mathsf{PI}\left(\right) \cdot l\_m\\
                    
                    
                    \end{array}
                    \end{array}
                    
                    Derivation
                    1. Split input into 2 regimes
                    2. if l < 8e14

                      1. Initial program 82.7%

                        \[\mathsf{PI}\left(\right) \cdot \ell - \frac{1}{F \cdot F} \cdot \tan \left(\mathsf{PI}\left(\right) \cdot \ell\right) \]
                      2. Add Preprocessing
                      3. Step-by-step derivation
                        1. lift-*.f64N/A

                          \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \color{blue}{\frac{1}{F \cdot F} \cdot \tan \left(\mathsf{PI}\left(\right) \cdot \ell\right)} \]
                        2. lift-/.f64N/A

                          \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \color{blue}{\frac{1}{F \cdot F}} \cdot \tan \left(\mathsf{PI}\left(\right) \cdot \ell\right) \]
                        3. associate-*l/N/A

                          \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \color{blue}{\frac{1 \cdot \tan \left(\mathsf{PI}\left(\right) \cdot \ell\right)}{F \cdot F}} \]
                        4. lift-*.f64N/A

                          \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \frac{1 \cdot \tan \left(\mathsf{PI}\left(\right) \cdot \ell\right)}{\color{blue}{F \cdot F}} \]
                        5. times-fracN/A

                          \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \color{blue}{\frac{1}{F} \cdot \frac{\tan \left(\mathsf{PI}\left(\right) \cdot \ell\right)}{F}} \]
                        6. lower-*.f64N/A

                          \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \color{blue}{\frac{1}{F} \cdot \frac{\tan \left(\mathsf{PI}\left(\right) \cdot \ell\right)}{F}} \]
                        7. inv-powN/A

                          \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \color{blue}{{F}^{-1}} \cdot \frac{\tan \left(\mathsf{PI}\left(\right) \cdot \ell\right)}{F} \]
                        8. lower-pow.f64N/A

                          \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \color{blue}{{F}^{-1}} \cdot \frac{\tan \left(\mathsf{PI}\left(\right) \cdot \ell\right)}{F} \]
                        9. lower-/.f6488.7

                          \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - {F}^{-1} \cdot \color{blue}{\frac{\tan \left(\mathsf{PI}\left(\right) \cdot \ell\right)}{F}} \]
                        10. lift-*.f64N/A

                          \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - {F}^{-1} \cdot \frac{\tan \color{blue}{\left(\mathsf{PI}\left(\right) \cdot \ell\right)}}{F} \]
                        11. *-commutativeN/A

                          \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - {F}^{-1} \cdot \frac{\tan \color{blue}{\left(\ell \cdot \mathsf{PI}\left(\right)\right)}}{F} \]
                        12. lower-*.f6488.7

                          \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - {F}^{-1} \cdot \frac{\tan \color{blue}{\left(\ell \cdot \mathsf{PI}\left(\right)\right)}}{F} \]
                      4. Applied rewrites88.7%

                        \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \color{blue}{{F}^{-1} \cdot \frac{\tan \left(\ell \cdot \mathsf{PI}\left(\right)\right)}{F}} \]
                      5. Step-by-step derivation
                        1. lift-*.f64N/A

                          \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \color{blue}{{F}^{-1} \cdot \frac{\tan \left(\ell \cdot \mathsf{PI}\left(\right)\right)}{F}} \]
                        2. *-commutativeN/A

                          \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \color{blue}{\frac{\tan \left(\ell \cdot \mathsf{PI}\left(\right)\right)}{F} \cdot {F}^{-1}} \]
                        3. lift-/.f64N/A

                          \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \color{blue}{\frac{\tan \left(\ell \cdot \mathsf{PI}\left(\right)\right)}{F}} \cdot {F}^{-1} \]
                        4. associate-*l/N/A

                          \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \color{blue}{\frac{\tan \left(\ell \cdot \mathsf{PI}\left(\right)\right) \cdot {F}^{-1}}{F}} \]
                        5. *-commutativeN/A

                          \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \frac{\color{blue}{{F}^{-1} \cdot \tan \left(\ell \cdot \mathsf{PI}\left(\right)\right)}}{F} \]
                        6. lift-pow.f64N/A

                          \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \frac{\color{blue}{{F}^{-1}} \cdot \tan \left(\ell \cdot \mathsf{PI}\left(\right)\right)}{F} \]
                        7. unpow-1N/A

                          \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \frac{\color{blue}{\frac{1}{F}} \cdot \tan \left(\ell \cdot \mathsf{PI}\left(\right)\right)}{F} \]
                        8. associate-*l/N/A

                          \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \frac{\color{blue}{\frac{1 \cdot \tan \left(\ell \cdot \mathsf{PI}\left(\right)\right)}{F}}}{F} \]
                        9. *-lft-identityN/A

                          \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \frac{\frac{\color{blue}{\tan \left(\ell \cdot \mathsf{PI}\left(\right)\right)}}{F}}{F} \]
                        10. lift-/.f64N/A

                          \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \frac{\color{blue}{\frac{\tan \left(\ell \cdot \mathsf{PI}\left(\right)\right)}{F}}}{F} \]
                        11. lower-/.f6488.7

                          \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \color{blue}{\frac{\frac{\tan \left(\ell \cdot \mathsf{PI}\left(\right)\right)}{F}}{F}} \]
                      6. Applied rewrites88.7%

                        \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \color{blue}{\frac{\frac{\tan \left(\mathsf{PI}\left(\right) \cdot \ell\right)}{F}}{F}} \]
                      7. Taylor expanded in l around 0

                        \[\leadsto \color{blue}{\ell \cdot \left(\mathsf{PI}\left(\right) - \frac{\mathsf{PI}\left(\right)}{{F}^{2}}\right)} \]
                      8. Step-by-step derivation
                        1. Applied rewrites77.7%

                          \[\leadsto \color{blue}{\left(\mathsf{PI}\left(\right) - \frac{\frac{\mathsf{PI}\left(\right)}{F}}{F}\right) \cdot \ell} \]

                        if 8e14 < l

                        1. Initial program 54.3%

                          \[\mathsf{PI}\left(\right) \cdot \ell - \frac{1}{F \cdot F} \cdot \tan \left(\mathsf{PI}\left(\right) \cdot \ell\right) \]
                        2. Add Preprocessing
                        3. Taylor expanded in F around inf

                          \[\leadsto \color{blue}{\ell \cdot \mathsf{PI}\left(\right)} \]
                        4. Step-by-step derivation
                          1. Applied rewrites99.4%

                            \[\leadsto \color{blue}{\mathsf{PI}\left(\right) \cdot \ell} \]
                        5. Recombined 2 regimes into one program.
                        6. Final simplification83.7%

                          \[\leadsto \begin{array}{l} \mathbf{if}\;\ell \leq 8 \cdot 10^{+14}:\\ \;\;\;\;\left(\mathsf{PI}\left(\right) - \frac{\frac{\mathsf{PI}\left(\right)}{F}}{F}\right) \cdot \ell\\ \mathbf{else}:\\ \;\;\;\;\mathsf{PI}\left(\right) \cdot \ell\\ \end{array} \]
                        7. Add Preprocessing

                        Alternative 7: 92.6% accurate, 4.4× speedup?

                        \[\begin{array}{l} l\_m = \left|\ell\right| \\ l\_s = \mathsf{copysign}\left(1, \ell\right) \\ l\_s \cdot \begin{array}{l} \mathbf{if}\;l\_m \leq 8 \cdot 10^{+14}:\\ \;\;\;\;\left(\mathsf{PI}\left(\right) - \frac{\mathsf{PI}\left(\right)}{F \cdot F}\right) \cdot l\_m\\ \mathbf{else}:\\ \;\;\;\;\mathsf{PI}\left(\right) \cdot l\_m\\ \end{array} \end{array} \]
                        l\_m = (fabs.f64 l)
                        l\_s = (copysign.f64 #s(literal 1 binary64) l)
                        (FPCore (l_s F l_m)
                         :precision binary64
                         (* l_s (if (<= l_m 8e+14) (* (- (PI) (/ (PI) (* F F))) l_m) (* (PI) l_m))))
                        \begin{array}{l}
                        l\_m = \left|\ell\right|
                        \\
                        l\_s = \mathsf{copysign}\left(1, \ell\right)
                        
                        \\
                        l\_s \cdot \begin{array}{l}
                        \mathbf{if}\;l\_m \leq 8 \cdot 10^{+14}:\\
                        \;\;\;\;\left(\mathsf{PI}\left(\right) - \frac{\mathsf{PI}\left(\right)}{F \cdot F}\right) \cdot l\_m\\
                        
                        \mathbf{else}:\\
                        \;\;\;\;\mathsf{PI}\left(\right) \cdot l\_m\\
                        
                        
                        \end{array}
                        \end{array}
                        
                        Derivation
                        1. Split input into 2 regimes
                        2. if l < 8e14

                          1. Initial program 82.7%

                            \[\mathsf{PI}\left(\right) \cdot \ell - \frac{1}{F \cdot F} \cdot \tan \left(\mathsf{PI}\left(\right) \cdot \ell\right) \]
                          2. Add Preprocessing
                          3. Taylor expanded in l around 0

                            \[\leadsto \color{blue}{\ell \cdot \left(\mathsf{PI}\left(\right) - \frac{\mathsf{PI}\left(\right)}{{F}^{2}}\right)} \]
                          4. Step-by-step derivation
                            1. Applied rewrites77.7%

                              \[\leadsto \color{blue}{\left(\mathsf{PI}\left(\right) - \frac{\mathsf{PI}\left(\right)}{F \cdot F}\right) \cdot \ell} \]

                            if 8e14 < l

                            1. Initial program 54.3%

                              \[\mathsf{PI}\left(\right) \cdot \ell - \frac{1}{F \cdot F} \cdot \tan \left(\mathsf{PI}\left(\right) \cdot \ell\right) \]
                            2. Add Preprocessing
                            3. Taylor expanded in F around inf

                              \[\leadsto \color{blue}{\ell \cdot \mathsf{PI}\left(\right)} \]
                            4. Step-by-step derivation
                              1. Applied rewrites99.4%

                                \[\leadsto \color{blue}{\mathsf{PI}\left(\right) \cdot \ell} \]
                            5. Recombined 2 regimes into one program.
                            6. Add Preprocessing

                            Alternative 8: 73.1% accurate, 22.5× speedup?

                            \[\begin{array}{l} l\_m = \left|\ell\right| \\ l\_s = \mathsf{copysign}\left(1, \ell\right) \\ l\_s \cdot \left(\mathsf{PI}\left(\right) \cdot l\_m\right) \end{array} \]
                            l\_m = (fabs.f64 l)
                            l\_s = (copysign.f64 #s(literal 1 binary64) l)
                            (FPCore (l_s F l_m) :precision binary64 (* l_s (* (PI) l_m)))
                            \begin{array}{l}
                            l\_m = \left|\ell\right|
                            \\
                            l\_s = \mathsf{copysign}\left(1, \ell\right)
                            
                            \\
                            l\_s \cdot \left(\mathsf{PI}\left(\right) \cdot l\_m\right)
                            \end{array}
                            
                            Derivation
                            1. Initial program 74.8%

                              \[\mathsf{PI}\left(\right) \cdot \ell - \frac{1}{F \cdot F} \cdot \tan \left(\mathsf{PI}\left(\right) \cdot \ell\right) \]
                            2. Add Preprocessing
                            3. Taylor expanded in F around inf

                              \[\leadsto \color{blue}{\ell \cdot \mathsf{PI}\left(\right)} \]
                            4. Step-by-step derivation
                              1. Applied rewrites76.7%

                                \[\leadsto \color{blue}{\mathsf{PI}\left(\right) \cdot \ell} \]
                              2. Add Preprocessing

                              Alternative 9: 3.1% accurate, 135.0× speedup?

                              \[\begin{array}{l} l\_m = \left|\ell\right| \\ l\_s = \mathsf{copysign}\left(1, \ell\right) \\ l\_s \cdot 0 \end{array} \]
                              l\_m = (fabs.f64 l)
                              l\_s = (copysign.f64 #s(literal 1 binary64) l)
                              (FPCore (l_s F l_m) :precision binary64 (* l_s 0.0))
                              l\_m = fabs(l);
                              l\_s = copysign(1.0, l);
                              double code(double l_s, double F, double l_m) {
                              	return l_s * 0.0;
                              }
                              
                              l\_m =     private
                              l\_s =     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(l_s, f, l_m)
                              use fmin_fmax_functions
                                  real(8), intent (in) :: l_s
                                  real(8), intent (in) :: f
                                  real(8), intent (in) :: l_m
                                  code = l_s * 0.0d0
                              end function
                              
                              l\_m = Math.abs(l);
                              l\_s = Math.copySign(1.0, l);
                              public static double code(double l_s, double F, double l_m) {
                              	return l_s * 0.0;
                              }
                              
                              l\_m = math.fabs(l)
                              l\_s = math.copysign(1.0, l)
                              def code(l_s, F, l_m):
                              	return l_s * 0.0
                              
                              l\_m = abs(l)
                              l\_s = copysign(1.0, l)
                              function code(l_s, F, l_m)
                              	return Float64(l_s * 0.0)
                              end
                              
                              l\_m = abs(l);
                              l\_s = sign(l) * abs(1.0);
                              function tmp = code(l_s, F, l_m)
                              	tmp = l_s * 0.0;
                              end
                              
                              l\_m = N[Abs[l], $MachinePrecision]
                              l\_s = N[With[{TMP1 = Abs[1.0], TMP2 = Sign[l]}, TMP1 * If[TMP2 == 0, 1, TMP2]], $MachinePrecision]
                              code[l$95$s_, F_, l$95$m_] := N[(l$95$s * 0.0), $MachinePrecision]
                              
                              \begin{array}{l}
                              l\_m = \left|\ell\right|
                              \\
                              l\_s = \mathsf{copysign}\left(1, \ell\right)
                              
                              \\
                              l\_s \cdot 0
                              \end{array}
                              
                              Derivation
                              1. Initial program 74.8%

                                \[\mathsf{PI}\left(\right) \cdot \ell - \frac{1}{F \cdot F} \cdot \tan \left(\mathsf{PI}\left(\right) \cdot \ell\right) \]
                              2. Add Preprocessing
                              3. Step-by-step derivation
                                1. lift-tan.f64N/A

                                  \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \frac{1}{F \cdot F} \cdot \color{blue}{\tan \left(\mathsf{PI}\left(\right) \cdot \ell\right)} \]
                                2. tan-+PI-revN/A

                                  \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \frac{1}{F \cdot F} \cdot \color{blue}{\tan \left(\mathsf{PI}\left(\right) \cdot \ell + \mathsf{PI}\left(\right)\right)} \]
                                3. lower-tan.f64N/A

                                  \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \frac{1}{F \cdot F} \cdot \color{blue}{\tan \left(\mathsf{PI}\left(\right) \cdot \ell + \mathsf{PI}\left(\right)\right)} \]
                                4. lift-*.f64N/A

                                  \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \frac{1}{F \cdot F} \cdot \tan \left(\color{blue}{\mathsf{PI}\left(\right) \cdot \ell} + \mathsf{PI}\left(\right)\right) \]
                                5. *-commutativeN/A

                                  \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \frac{1}{F \cdot F} \cdot \tan \left(\color{blue}{\ell \cdot \mathsf{PI}\left(\right)} + \mathsf{PI}\left(\right)\right) \]
                                6. lift-PI.f64N/A

                                  \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \frac{1}{F \cdot F} \cdot \tan \left(\ell \cdot \mathsf{PI}\left(\right) + \color{blue}{\mathsf{PI}\left(\right)}\right) \]
                                7. lower-fma.f6458.0

                                  \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \frac{1}{F \cdot F} \cdot \tan \color{blue}{\left(\mathsf{fma}\left(\ell, \mathsf{PI}\left(\right), \mathsf{PI}\left(\right)\right)\right)} \]
                              4. Applied rewrites58.0%

                                \[\leadsto \mathsf{PI}\left(\right) \cdot \ell - \frac{1}{F \cdot F} \cdot \color{blue}{\tan \left(\mathsf{fma}\left(\ell, \mathsf{PI}\left(\right), \mathsf{PI}\left(\right)\right)\right)} \]
                              5. Taylor expanded in l around 0

                                \[\leadsto \color{blue}{-1 \cdot \frac{\sin \mathsf{PI}\left(\right)}{{F}^{2} \cdot \cos \mathsf{PI}\left(\right)}} \]
                              6. Step-by-step derivation
                                1. Applied rewrites3.1%

                                  \[\leadsto \color{blue}{0} \]
                                2. Final simplification3.1%

                                  \[\leadsto 0 \]
                                3. Add Preprocessing

                                Reproduce

                                ?
                                herbie shell --seed 2025018 
                                (FPCore (F l)
                                  :name "VandenBroeck and Keller, Equation (6)"
                                  :precision binary64
                                  (- (* (PI) l) (* (/ 1.0 (* F F)) (tan (* (PI) l)))))