UniformSampleCone, y

Percentage Accurate: 57.6% → 98.3%
Time: 10.7s
Alternatives: 21
Speedup: 4.7×

Specification

?
\[\left(\left(2.328306437 \cdot 10^{-10} \leq ux \land ux \leq 1\right) \land \left(2.328306437 \cdot 10^{-10} \leq uy \land uy \leq 1\right)\right) \land \left(0 \leq maxCos \land maxCos \leq 1\right)\]
\[\begin{array}{l} \\ \begin{array}{l} t_0 := \left(1 - ux\right) + ux \cdot maxCos\\ \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{1 - t\_0 \cdot t\_0} \end{array} \end{array} \]
(FPCore (ux uy maxCos)
 :precision binary32
 (let* ((t_0 (+ (- 1.0 ux) (* ux maxCos))))
   (* (sin (* (* uy 2.0) PI)) (sqrt (- 1.0 (* t_0 t_0))))))
float code(float ux, float uy, float maxCos) {
	float t_0 = (1.0f - ux) + (ux * maxCos);
	return sinf(((uy * 2.0f) * ((float) M_PI))) * sqrtf((1.0f - (t_0 * t_0)));
}
function code(ux, uy, maxCos)
	t_0 = Float32(Float32(Float32(1.0) - ux) + Float32(ux * maxCos))
	return Float32(sin(Float32(Float32(uy * Float32(2.0)) * Float32(pi))) * sqrt(Float32(Float32(1.0) - Float32(t_0 * t_0))))
end
function tmp = code(ux, uy, maxCos)
	t_0 = (single(1.0) - ux) + (ux * maxCos);
	tmp = sin(((uy * single(2.0)) * single(pi))) * sqrt((single(1.0) - (t_0 * t_0)));
end
\begin{array}{l}

\\
\begin{array}{l}
t_0 := \left(1 - ux\right) + ux \cdot maxCos\\
\sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{1 - t\_0 \cdot t\_0}
\end{array}
\end{array}

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 21 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: 57.6% accurate, 1.0× speedup?

\[\begin{array}{l} \\ \begin{array}{l} t_0 := \left(1 - ux\right) + ux \cdot maxCos\\ \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{1 - t\_0 \cdot t\_0} \end{array} \end{array} \]
(FPCore (ux uy maxCos)
 :precision binary32
 (let* ((t_0 (+ (- 1.0 ux) (* ux maxCos))))
   (* (sin (* (* uy 2.0) PI)) (sqrt (- 1.0 (* t_0 t_0))))))
float code(float ux, float uy, float maxCos) {
	float t_0 = (1.0f - ux) + (ux * maxCos);
	return sinf(((uy * 2.0f) * ((float) M_PI))) * sqrtf((1.0f - (t_0 * t_0)));
}
function code(ux, uy, maxCos)
	t_0 = Float32(Float32(Float32(1.0) - ux) + Float32(ux * maxCos))
	return Float32(sin(Float32(Float32(uy * Float32(2.0)) * Float32(pi))) * sqrt(Float32(Float32(1.0) - Float32(t_0 * t_0))))
end
function tmp = code(ux, uy, maxCos)
	t_0 = (single(1.0) - ux) + (ux * maxCos);
	tmp = sin(((uy * single(2.0)) * single(pi))) * sqrt((single(1.0) - (t_0 * t_0)));
end
\begin{array}{l}

\\
\begin{array}{l}
t_0 := \left(1 - ux\right) + ux \cdot maxCos\\
\sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{1 - t\_0 \cdot t\_0}
\end{array}
\end{array}

Alternative 1: 98.3% accurate, 0.6× speedup?

\[\begin{array}{l} \\ \left(2 \cdot \left(\sin \left(\pi \cdot uy\right) \cdot \cos \left(\pi \cdot uy\right)\right)\right) \cdot \sqrt{\left(2 - \mathsf{fma}\left(\left(maxCos - 1\right) \cdot ux, maxCos - 1, maxCos + maxCos\right)\right) \cdot ux} \end{array} \]
(FPCore (ux uy maxCos)
 :precision binary32
 (*
  (* 2.0 (* (sin (* PI uy)) (cos (* PI uy))))
  (sqrt
   (*
    (- 2.0 (fma (* (- maxCos 1.0) ux) (- maxCos 1.0) (+ maxCos maxCos)))
    ux))))
float code(float ux, float uy, float maxCos) {
	return (2.0f * (sinf((((float) M_PI) * uy)) * cosf((((float) M_PI) * uy)))) * sqrtf(((2.0f - fmaf(((maxCos - 1.0f) * ux), (maxCos - 1.0f), (maxCos + maxCos))) * ux));
}
function code(ux, uy, maxCos)
	return Float32(Float32(Float32(2.0) * Float32(sin(Float32(Float32(pi) * uy)) * cos(Float32(Float32(pi) * uy)))) * sqrt(Float32(Float32(Float32(2.0) - fma(Float32(Float32(maxCos - Float32(1.0)) * ux), Float32(maxCos - Float32(1.0)), Float32(maxCos + maxCos))) * ux)))
end
\begin{array}{l}

\\
\left(2 \cdot \left(\sin \left(\pi \cdot uy\right) \cdot \cos \left(\pi \cdot uy\right)\right)\right) \cdot \sqrt{\left(2 - \mathsf{fma}\left(\left(maxCos - 1\right) \cdot ux, maxCos - 1, maxCos + maxCos\right)\right) \cdot ux}
\end{array}
Derivation
  1. Initial program 57.6%

    \[\sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{1 - \left(\left(1 - ux\right) + ux \cdot maxCos\right) \cdot \left(\left(1 - ux\right) + ux \cdot maxCos\right)} \]
  2. Taylor expanded in ux around 0

    \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\color{blue}{ux \cdot \left(\left(2 + -1 \cdot \left(ux \cdot {\left(maxCos - 1\right)}^{2}\right)\right) - 2 \cdot maxCos\right)}} \]
  3. Step-by-step derivation
    1. *-commutativeN/A

      \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(\left(2 + -1 \cdot \left(ux \cdot {\left(maxCos - 1\right)}^{2}\right)\right) - 2 \cdot maxCos\right) \cdot \color{blue}{ux}} \]
    2. lower-*.f32N/A

      \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(\left(2 + -1 \cdot \left(ux \cdot {\left(maxCos - 1\right)}^{2}\right)\right) - 2 \cdot maxCos\right) \cdot \color{blue}{ux}} \]
  4. Applied rewrites98.3%

    \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\color{blue}{\left(2 - \mathsf{fma}\left(\left(maxCos - 1\right) \cdot ux, maxCos - 1, maxCos + maxCos\right)\right) \cdot ux}} \]
  5. Step-by-step derivation
    1. lift-sin.f32N/A

      \[\leadsto \color{blue}{\sin \left(\left(uy \cdot 2\right) \cdot \pi\right)} \cdot \sqrt{\left(2 - \mathsf{fma}\left(\left(maxCos - 1\right) \cdot ux, maxCos - 1, maxCos + maxCos\right)\right) \cdot ux} \]
    2. lift-PI.f32N/A

      \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \color{blue}{\mathsf{PI}\left(\right)}\right) \cdot \sqrt{\left(2 - \mathsf{fma}\left(\left(maxCos - 1\right) \cdot ux, maxCos - 1, maxCos + maxCos\right)\right) \cdot ux} \]
    3. lift-*.f32N/A

      \[\leadsto \sin \color{blue}{\left(\left(uy \cdot 2\right) \cdot \mathsf{PI}\left(\right)\right)} \cdot \sqrt{\left(2 - \mathsf{fma}\left(\left(maxCos - 1\right) \cdot ux, maxCos - 1, maxCos + maxCos\right)\right) \cdot ux} \]
    4. lift-*.f32N/A

      \[\leadsto \sin \left(\color{blue}{\left(uy \cdot 2\right)} \cdot \mathsf{PI}\left(\right)\right) \cdot \sqrt{\left(2 - \mathsf{fma}\left(\left(maxCos - 1\right) \cdot ux, maxCos - 1, maxCos + maxCos\right)\right) \cdot ux} \]
    5. *-commutativeN/A

      \[\leadsto \sin \left(\color{blue}{\left(2 \cdot uy\right)} \cdot \mathsf{PI}\left(\right)\right) \cdot \sqrt{\left(2 - \mathsf{fma}\left(\left(maxCos - 1\right) \cdot ux, maxCos - 1, maxCos + maxCos\right)\right) \cdot ux} \]
    6. associate-*r*N/A

      \[\leadsto \sin \color{blue}{\left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right)} \cdot \sqrt{\left(2 - \mathsf{fma}\left(\left(maxCos - 1\right) \cdot ux, maxCos - 1, maxCos + maxCos\right)\right) \cdot ux} \]
    7. sin-2N/A

      \[\leadsto \color{blue}{\left(2 \cdot \left(\sin \left(uy \cdot \mathsf{PI}\left(\right)\right) \cdot \cos \left(uy \cdot \mathsf{PI}\left(\right)\right)\right)\right)} \cdot \sqrt{\left(2 - \mathsf{fma}\left(\left(maxCos - 1\right) \cdot ux, maxCos - 1, maxCos + maxCos\right)\right) \cdot ux} \]
    8. lower-*.f32N/A

      \[\leadsto \color{blue}{\left(2 \cdot \left(\sin \left(uy \cdot \mathsf{PI}\left(\right)\right) \cdot \cos \left(uy \cdot \mathsf{PI}\left(\right)\right)\right)\right)} \cdot \sqrt{\left(2 - \mathsf{fma}\left(\left(maxCos - 1\right) \cdot ux, maxCos - 1, maxCos + maxCos\right)\right) \cdot ux} \]
    9. lower-*.f32N/A

      \[\leadsto \left(2 \cdot \color{blue}{\left(\sin \left(uy \cdot \mathsf{PI}\left(\right)\right) \cdot \cos \left(uy \cdot \mathsf{PI}\left(\right)\right)\right)}\right) \cdot \sqrt{\left(2 - \mathsf{fma}\left(\left(maxCos - 1\right) \cdot ux, maxCos - 1, maxCos + maxCos\right)\right) \cdot ux} \]
    10. lower-sin.f32N/A

      \[\leadsto \left(2 \cdot \left(\color{blue}{\sin \left(uy \cdot \mathsf{PI}\left(\right)\right)} \cdot \cos \left(uy \cdot \mathsf{PI}\left(\right)\right)\right)\right) \cdot \sqrt{\left(2 - \mathsf{fma}\left(\left(maxCos - 1\right) \cdot ux, maxCos - 1, maxCos + maxCos\right)\right) \cdot ux} \]
    11. *-commutativeN/A

      \[\leadsto \left(2 \cdot \left(\sin \color{blue}{\left(\mathsf{PI}\left(\right) \cdot uy\right)} \cdot \cos \left(uy \cdot \mathsf{PI}\left(\right)\right)\right)\right) \cdot \sqrt{\left(2 - \mathsf{fma}\left(\left(maxCos - 1\right) \cdot ux, maxCos - 1, maxCos + maxCos\right)\right) \cdot ux} \]
    12. lower-*.f32N/A

      \[\leadsto \left(2 \cdot \left(\sin \color{blue}{\left(\mathsf{PI}\left(\right) \cdot uy\right)} \cdot \cos \left(uy \cdot \mathsf{PI}\left(\right)\right)\right)\right) \cdot \sqrt{\left(2 - \mathsf{fma}\left(\left(maxCos - 1\right) \cdot ux, maxCos - 1, maxCos + maxCos\right)\right) \cdot ux} \]
    13. lift-PI.f32N/A

      \[\leadsto \left(2 \cdot \left(\sin \left(\color{blue}{\pi} \cdot uy\right) \cdot \cos \left(uy \cdot \mathsf{PI}\left(\right)\right)\right)\right) \cdot \sqrt{\left(2 - \mathsf{fma}\left(\left(maxCos - 1\right) \cdot ux, maxCos - 1, maxCos + maxCos\right)\right) \cdot ux} \]
    14. lower-cos.f32N/A

      \[\leadsto \left(2 \cdot \left(\sin \left(\pi \cdot uy\right) \cdot \color{blue}{\cos \left(uy \cdot \mathsf{PI}\left(\right)\right)}\right)\right) \cdot \sqrt{\left(2 - \mathsf{fma}\left(\left(maxCos - 1\right) \cdot ux, maxCos - 1, maxCos + maxCos\right)\right) \cdot ux} \]
    15. *-commutativeN/A

      \[\leadsto \left(2 \cdot \left(\sin \left(\pi \cdot uy\right) \cdot \cos \color{blue}{\left(\mathsf{PI}\left(\right) \cdot uy\right)}\right)\right) \cdot \sqrt{\left(2 - \mathsf{fma}\left(\left(maxCos - 1\right) \cdot ux, maxCos - 1, maxCos + maxCos\right)\right) \cdot ux} \]
    16. lower-*.f32N/A

      \[\leadsto \left(2 \cdot \left(\sin \left(\pi \cdot uy\right) \cdot \cos \color{blue}{\left(\mathsf{PI}\left(\right) \cdot uy\right)}\right)\right) \cdot \sqrt{\left(2 - \mathsf{fma}\left(\left(maxCos - 1\right) \cdot ux, maxCos - 1, maxCos + maxCos\right)\right) \cdot ux} \]
    17. lift-PI.f3298.2

      \[\leadsto \left(2 \cdot \left(\sin \left(\pi \cdot uy\right) \cdot \cos \left(\color{blue}{\pi} \cdot uy\right)\right)\right) \cdot \sqrt{\left(2 - \mathsf{fma}\left(\left(maxCos - 1\right) \cdot ux, maxCos - 1, maxCos + maxCos\right)\right) \cdot ux} \]
  6. Applied rewrites98.2%

    \[\leadsto \color{blue}{\left(2 \cdot \left(\sin \left(\pi \cdot uy\right) \cdot \cos \left(\pi \cdot uy\right)\right)\right)} \cdot \sqrt{\left(2 - \mathsf{fma}\left(\left(maxCos - 1\right) \cdot ux, maxCos - 1, maxCos + maxCos\right)\right) \cdot ux} \]
  7. Add Preprocessing

Alternative 2: 98.2% accurate, 0.8× speedup?

\[\begin{array}{l} \\ \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\mathsf{fma}\left(-\mathsf{fma}\left(ux \cdot ux, maxCos, \left(\left(2 - ux\right) - ux\right) \cdot ux\right), maxCos, \left(\frac{2}{ux} - 1\right) \cdot \left(ux \cdot ux\right)\right)} \end{array} \]
(FPCore (ux uy maxCos)
 :precision binary32
 (*
  (sin (* (* uy 2.0) PI))
  (sqrt
   (fma
    (- (fma (* ux ux) maxCos (* (- (- 2.0 ux) ux) ux)))
    maxCos
    (* (- (/ 2.0 ux) 1.0) (* ux ux))))))
float code(float ux, float uy, float maxCos) {
	return sinf(((uy * 2.0f) * ((float) M_PI))) * sqrtf(fmaf(-fmaf((ux * ux), maxCos, (((2.0f - ux) - ux) * ux)), maxCos, (((2.0f / ux) - 1.0f) * (ux * ux))));
}
function code(ux, uy, maxCos)
	return Float32(sin(Float32(Float32(uy * Float32(2.0)) * Float32(pi))) * sqrt(fma(Float32(-fma(Float32(ux * ux), maxCos, Float32(Float32(Float32(Float32(2.0) - ux) - ux) * ux))), maxCos, Float32(Float32(Float32(Float32(2.0) / ux) - Float32(1.0)) * Float32(ux * ux)))))
end
\begin{array}{l}

\\
\sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\mathsf{fma}\left(-\mathsf{fma}\left(ux \cdot ux, maxCos, \left(\left(2 - ux\right) - ux\right) \cdot ux\right), maxCos, \left(\frac{2}{ux} - 1\right) \cdot \left(ux \cdot ux\right)\right)}
\end{array}
Derivation
  1. Initial program 57.6%

    \[\sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{1 - \left(\left(1 - ux\right) + ux \cdot maxCos\right) \cdot \left(\left(1 - ux\right) + ux \cdot maxCos\right)} \]
  2. Taylor expanded in ux around 0

    \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\color{blue}{ux \cdot \left(\left(2 + -1 \cdot \left(ux \cdot {\left(maxCos - 1\right)}^{2}\right)\right) - 2 \cdot maxCos\right)}} \]
  3. Step-by-step derivation
    1. *-commutativeN/A

      \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(\left(2 + -1 \cdot \left(ux \cdot {\left(maxCos - 1\right)}^{2}\right)\right) - 2 \cdot maxCos\right) \cdot \color{blue}{ux}} \]
    2. lower-*.f32N/A

      \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(\left(2 + -1 \cdot \left(ux \cdot {\left(maxCos - 1\right)}^{2}\right)\right) - 2 \cdot maxCos\right) \cdot \color{blue}{ux}} \]
  4. Applied rewrites98.3%

    \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\color{blue}{\left(2 - \mathsf{fma}\left(\left(maxCos - 1\right) \cdot ux, maxCos - 1, maxCos + maxCos\right)\right) \cdot ux}} \]
  5. Taylor expanded in maxCos around 0

    \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{maxCos \cdot \left(-1 \cdot \left(maxCos \cdot {ux}^{2}\right) + -1 \cdot \left(ux \cdot \left(2 + -2 \cdot ux\right)\right)\right) + \color{blue}{ux \cdot \left(2 - ux\right)}} \]
  6. Step-by-step derivation
    1. *-commutativeN/A

      \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(-1 \cdot \left(maxCos \cdot {ux}^{2}\right) + -1 \cdot \left(ux \cdot \left(2 + -2 \cdot ux\right)\right)\right) \cdot maxCos + ux \cdot \left(\color{blue}{2} - ux\right)} \]
    2. lower-fma.f32N/A

      \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\mathsf{fma}\left(-1 \cdot \left(maxCos \cdot {ux}^{2}\right) + -1 \cdot \left(ux \cdot \left(2 + -2 \cdot ux\right)\right), maxCos, ux \cdot \left(2 - ux\right)\right)} \]
  7. Applied rewrites98.3%

    \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\mathsf{fma}\left(-\mathsf{fma}\left(ux \cdot ux, maxCos, \left(\left(2 - ux\right) - ux\right) \cdot ux\right), \color{blue}{maxCos}, \left(2 - ux\right) \cdot ux\right)} \]
  8. Taylor expanded in ux around inf

    \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\mathsf{fma}\left(-\mathsf{fma}\left(ux \cdot ux, maxCos, \left(\left(2 - ux\right) - ux\right) \cdot ux\right), maxCos, {ux}^{2} \cdot \left(2 \cdot \frac{1}{ux} - 1\right)\right)} \]
  9. Step-by-step derivation
    1. *-commutativeN/A

      \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\mathsf{fma}\left(-\mathsf{fma}\left(ux \cdot ux, maxCos, \left(\left(2 - ux\right) - ux\right) \cdot ux\right), maxCos, \left(2 \cdot \frac{1}{ux} - 1\right) \cdot {ux}^{2}\right)} \]
    2. lower-*.f32N/A

      \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\mathsf{fma}\left(-\mathsf{fma}\left(ux \cdot ux, maxCos, \left(\left(2 - ux\right) - ux\right) \cdot ux\right), maxCos, \left(2 \cdot \frac{1}{ux} - 1\right) \cdot {ux}^{2}\right)} \]
    3. lower--.f32N/A

      \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\mathsf{fma}\left(-\mathsf{fma}\left(ux \cdot ux, maxCos, \left(\left(2 - ux\right) - ux\right) \cdot ux\right), maxCos, \left(2 \cdot \frac{1}{ux} - 1\right) \cdot {ux}^{2}\right)} \]
    4. associate-*r/N/A

      \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\mathsf{fma}\left(-\mathsf{fma}\left(ux \cdot ux, maxCos, \left(\left(2 - ux\right) - ux\right) \cdot ux\right), maxCos, \left(\frac{2 \cdot 1}{ux} - 1\right) \cdot {ux}^{2}\right)} \]
    5. metadata-evalN/A

      \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\mathsf{fma}\left(-\mathsf{fma}\left(ux \cdot ux, maxCos, \left(\left(2 - ux\right) - ux\right) \cdot ux\right), maxCos, \left(\frac{2}{ux} - 1\right) \cdot {ux}^{2}\right)} \]
    6. lower-/.f32N/A

      \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\mathsf{fma}\left(-\mathsf{fma}\left(ux \cdot ux, maxCos, \left(\left(2 - ux\right) - ux\right) \cdot ux\right), maxCos, \left(\frac{2}{ux} - 1\right) \cdot {ux}^{2}\right)} \]
    7. pow2N/A

      \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\mathsf{fma}\left(-\mathsf{fma}\left(ux \cdot ux, maxCos, \left(\left(2 - ux\right) - ux\right) \cdot ux\right), maxCos, \left(\frac{2}{ux} - 1\right) \cdot \left(ux \cdot ux\right)\right)} \]
    8. lift-*.f3298.2

      \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\mathsf{fma}\left(-\mathsf{fma}\left(ux \cdot ux, maxCos, \left(\left(2 - ux\right) - ux\right) \cdot ux\right), maxCos, \left(\frac{2}{ux} - 1\right) \cdot \left(ux \cdot ux\right)\right)} \]
  10. Applied rewrites98.2%

    \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\mathsf{fma}\left(-\mathsf{fma}\left(ux \cdot ux, maxCos, \left(\left(2 - ux\right) - ux\right) \cdot ux\right), maxCos, \left(\frac{2}{ux} - 1\right) \cdot \left(ux \cdot ux\right)\right)} \]
  11. Add Preprocessing

Alternative 3: 98.2% accurate, 1.1× speedup?

\[\begin{array}{l} \\ \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(2 - \mathsf{fma}\left(\mathsf{fma}\left(ux, maxCos + -2, 2\right), maxCos, ux\right)\right) \cdot ux} \end{array} \]
(FPCore (ux uy maxCos)
 :precision binary32
 (*
  (sin (* (* uy 2.0) PI))
  (sqrt (* (- 2.0 (fma (fma ux (+ maxCos -2.0) 2.0) maxCos ux)) ux))))
float code(float ux, float uy, float maxCos) {
	return sinf(((uy * 2.0f) * ((float) M_PI))) * sqrtf(((2.0f - fmaf(fmaf(ux, (maxCos + -2.0f), 2.0f), maxCos, ux)) * ux));
}
function code(ux, uy, maxCos)
	return Float32(sin(Float32(Float32(uy * Float32(2.0)) * Float32(pi))) * sqrt(Float32(Float32(Float32(2.0) - fma(fma(ux, Float32(maxCos + Float32(-2.0)), Float32(2.0)), maxCos, ux)) * ux)))
end
\begin{array}{l}

\\
\sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(2 - \mathsf{fma}\left(\mathsf{fma}\left(ux, maxCos + -2, 2\right), maxCos, ux\right)\right) \cdot ux}
\end{array}
Derivation
  1. Initial program 57.6%

    \[\sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{1 - \left(\left(1 - ux\right) + ux \cdot maxCos\right) \cdot \left(\left(1 - ux\right) + ux \cdot maxCos\right)} \]
  2. Taylor expanded in ux around 0

    \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\color{blue}{ux \cdot \left(\left(2 + -1 \cdot \left(ux \cdot {\left(maxCos - 1\right)}^{2}\right)\right) - 2 \cdot maxCos\right)}} \]
  3. Step-by-step derivation
    1. *-commutativeN/A

      \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(\left(2 + -1 \cdot \left(ux \cdot {\left(maxCos - 1\right)}^{2}\right)\right) - 2 \cdot maxCos\right) \cdot \color{blue}{ux}} \]
    2. lower-*.f32N/A

      \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(\left(2 + -1 \cdot \left(ux \cdot {\left(maxCos - 1\right)}^{2}\right)\right) - 2 \cdot maxCos\right) \cdot \color{blue}{ux}} \]
  4. Applied rewrites98.3%

    \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\color{blue}{\left(2 - \mathsf{fma}\left(\left(maxCos - 1\right) \cdot ux, maxCos - 1, maxCos + maxCos\right)\right) \cdot ux}} \]
  5. Taylor expanded in maxCos around 0

    \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(2 - \left(ux + maxCos \cdot \left(2 + \left(-2 \cdot ux + maxCos \cdot ux\right)\right)\right)\right) \cdot ux} \]
  6. Step-by-step derivation
    1. +-commutativeN/A

      \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(2 - \left(maxCos \cdot \left(2 + \left(-2 \cdot ux + maxCos \cdot ux\right)\right) + ux\right)\right) \cdot ux} \]
    2. *-commutativeN/A

      \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(2 - \left(\left(2 + \left(-2 \cdot ux + maxCos \cdot ux\right)\right) \cdot maxCos + ux\right)\right) \cdot ux} \]
    3. lower-fma.f32N/A

      \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(2 - \mathsf{fma}\left(2 + \left(-2 \cdot ux + maxCos \cdot ux\right), maxCos, ux\right)\right) \cdot ux} \]
    4. +-commutativeN/A

      \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(2 - \mathsf{fma}\left(\left(-2 \cdot ux + maxCos \cdot ux\right) + 2, maxCos, ux\right)\right) \cdot ux} \]
    5. +-commutativeN/A

      \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(2 - \mathsf{fma}\left(\left(maxCos \cdot ux + -2 \cdot ux\right) + 2, maxCos, ux\right)\right) \cdot ux} \]
    6. distribute-rgt-outN/A

      \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(2 - \mathsf{fma}\left(ux \cdot \left(maxCos + -2\right) + 2, maxCos, ux\right)\right) \cdot ux} \]
    7. lower-fma.f32N/A

      \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(2 - \mathsf{fma}\left(\mathsf{fma}\left(ux, maxCos + -2, 2\right), maxCos, ux\right)\right) \cdot ux} \]
    8. lower-+.f3298.3

      \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(2 - \mathsf{fma}\left(\mathsf{fma}\left(ux, maxCos + -2, 2\right), maxCos, ux\right)\right) \cdot ux} \]
  7. Applied rewrites98.3%

    \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(2 - \mathsf{fma}\left(\mathsf{fma}\left(ux, maxCos + -2, 2\right), maxCos, ux\right)\right) \cdot ux} \]
  8. Add Preprocessing

Alternative 4: 97.6% accurate, 1.1× speedup?

\[\begin{array}{l} \\ \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(2 - \mathsf{fma}\left(\left(2 - ux\right) - ux, maxCos, ux\right)\right) \cdot ux} \end{array} \]
(FPCore (ux uy maxCos)
 :precision binary32
 (*
  (sin (* (* uy 2.0) PI))
  (sqrt (* (- 2.0 (fma (- (- 2.0 ux) ux) maxCos ux)) ux))))
float code(float ux, float uy, float maxCos) {
	return sinf(((uy * 2.0f) * ((float) M_PI))) * sqrtf(((2.0f - fmaf(((2.0f - ux) - ux), maxCos, ux)) * ux));
}
function code(ux, uy, maxCos)
	return Float32(sin(Float32(Float32(uy * Float32(2.0)) * Float32(pi))) * sqrt(Float32(Float32(Float32(2.0) - fma(Float32(Float32(Float32(2.0) - ux) - ux), maxCos, ux)) * ux)))
end
\begin{array}{l}

\\
\sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(2 - \mathsf{fma}\left(\left(2 - ux\right) - ux, maxCos, ux\right)\right) \cdot ux}
\end{array}
Derivation
  1. Initial program 57.6%

    \[\sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{1 - \left(\left(1 - ux\right) + ux \cdot maxCos\right) \cdot \left(\left(1 - ux\right) + ux \cdot maxCos\right)} \]
  2. Taylor expanded in ux around 0

    \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\color{blue}{ux \cdot \left(\left(2 + -1 \cdot \left(ux \cdot {\left(maxCos - 1\right)}^{2}\right)\right) - 2 \cdot maxCos\right)}} \]
  3. Step-by-step derivation
    1. *-commutativeN/A

      \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(\left(2 + -1 \cdot \left(ux \cdot {\left(maxCos - 1\right)}^{2}\right)\right) - 2 \cdot maxCos\right) \cdot \color{blue}{ux}} \]
    2. lower-*.f32N/A

      \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(\left(2 + -1 \cdot \left(ux \cdot {\left(maxCos - 1\right)}^{2}\right)\right) - 2 \cdot maxCos\right) \cdot \color{blue}{ux}} \]
  4. Applied rewrites98.3%

    \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\color{blue}{\left(2 - \mathsf{fma}\left(\left(maxCos - 1\right) \cdot ux, maxCos - 1, maxCos + maxCos\right)\right) \cdot ux}} \]
  5. Taylor expanded in maxCos around 0

    \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(2 - \left(ux + maxCos \cdot \left(2 + -2 \cdot ux\right)\right)\right) \cdot ux} \]
  6. Step-by-step derivation
    1. +-commutativeN/A

      \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(2 - \left(maxCos \cdot \left(2 + -2 \cdot ux\right) + ux\right)\right) \cdot ux} \]
    2. *-commutativeN/A

      \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(2 - \left(\left(2 + -2 \cdot ux\right) \cdot maxCos + ux\right)\right) \cdot ux} \]
    3. lower-fma.f32N/A

      \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(2 - \mathsf{fma}\left(2 + -2 \cdot ux, maxCos, ux\right)\right) \cdot ux} \]
    4. metadata-evalN/A

      \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(2 - \mathsf{fma}\left(2 + \left(\mathsf{neg}\left(2\right)\right) \cdot ux, maxCos, ux\right)\right) \cdot ux} \]
    5. fp-cancel-sub-sign-invN/A

      \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(2 - \mathsf{fma}\left(2 - 2 \cdot ux, maxCos, ux\right)\right) \cdot ux} \]
    6. count-2-revN/A

      \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(2 - \mathsf{fma}\left(2 - \left(ux + ux\right), maxCos, ux\right)\right) \cdot ux} \]
    7. associate--r+N/A

      \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(2 - \mathsf{fma}\left(\left(2 - ux\right) - ux, maxCos, ux\right)\right) \cdot ux} \]
    8. lower--.f32N/A

      \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(2 - \mathsf{fma}\left(\left(2 - ux\right) - ux, maxCos, ux\right)\right) \cdot ux} \]
    9. lower--.f3297.6

      \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(2 - \mathsf{fma}\left(\left(2 - ux\right) - ux, maxCos, ux\right)\right) \cdot ux} \]
  7. Applied rewrites97.6%

    \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(2 - \mathsf{fma}\left(\left(2 - ux\right) - ux, maxCos, ux\right)\right) \cdot ux} \]
  8. Add Preprocessing

Alternative 5: 96.8% accurate, 1.2× speedup?

\[\begin{array}{l} \\ \begin{array}{l} \mathbf{if}\;uy \leq 0.006899999920278788:\\ \;\;\;\;\left(\mathsf{fma}\left(\left(uy \cdot uy\right) \cdot \left(\left(\pi \cdot \pi\right) \cdot \pi\right), -1.3333333333333333, \pi + \pi\right) \cdot uy\right) \cdot \sqrt{\left(2 - \mathsf{fma}\left(\left(maxCos - 1\right) \cdot ux, maxCos - 1, maxCos + maxCos\right)\right) \cdot ux}\\ \mathbf{else}:\\ \;\;\;\;\sqrt{\left(2 - ux\right) \cdot ux} \cdot \sin \left(\pi \cdot \left(uy + uy\right)\right)\\ \end{array} \end{array} \]
(FPCore (ux uy maxCos)
 :precision binary32
 (if (<= uy 0.006899999920278788)
   (*
    (* (fma (* (* uy uy) (* (* PI PI) PI)) -1.3333333333333333 (+ PI PI)) uy)
    (sqrt
     (*
      (- 2.0 (fma (* (- maxCos 1.0) ux) (- maxCos 1.0) (+ maxCos maxCos)))
      ux)))
   (* (sqrt (* (- 2.0 ux) ux)) (sin (* PI (+ uy uy))))))
float code(float ux, float uy, float maxCos) {
	float tmp;
	if (uy <= 0.006899999920278788f) {
		tmp = (fmaf(((uy * uy) * ((((float) M_PI) * ((float) M_PI)) * ((float) M_PI))), -1.3333333333333333f, (((float) M_PI) + ((float) M_PI))) * uy) * sqrtf(((2.0f - fmaf(((maxCos - 1.0f) * ux), (maxCos - 1.0f), (maxCos + maxCos))) * ux));
	} else {
		tmp = sqrtf(((2.0f - ux) * ux)) * sinf((((float) M_PI) * (uy + uy)));
	}
	return tmp;
}
function code(ux, uy, maxCos)
	tmp = Float32(0.0)
	if (uy <= Float32(0.006899999920278788))
		tmp = Float32(Float32(fma(Float32(Float32(uy * uy) * Float32(Float32(Float32(pi) * Float32(pi)) * Float32(pi))), Float32(-1.3333333333333333), Float32(Float32(pi) + Float32(pi))) * uy) * sqrt(Float32(Float32(Float32(2.0) - fma(Float32(Float32(maxCos - Float32(1.0)) * ux), Float32(maxCos - Float32(1.0)), Float32(maxCos + maxCos))) * ux)));
	else
		tmp = Float32(sqrt(Float32(Float32(Float32(2.0) - ux) * ux)) * sin(Float32(Float32(pi) * Float32(uy + uy))));
	end
	return tmp
end
\begin{array}{l}

\\
\begin{array}{l}
\mathbf{if}\;uy \leq 0.006899999920278788:\\
\;\;\;\;\left(\mathsf{fma}\left(\left(uy \cdot uy\right) \cdot \left(\left(\pi \cdot \pi\right) \cdot \pi\right), -1.3333333333333333, \pi + \pi\right) \cdot uy\right) \cdot \sqrt{\left(2 - \mathsf{fma}\left(\left(maxCos - 1\right) \cdot ux, maxCos - 1, maxCos + maxCos\right)\right) \cdot ux}\\

\mathbf{else}:\\
\;\;\;\;\sqrt{\left(2 - ux\right) \cdot ux} \cdot \sin \left(\pi \cdot \left(uy + uy\right)\right)\\


\end{array}
\end{array}
Derivation
  1. Split input into 2 regimes
  2. if uy < 0.00689999992

    1. Initial program 57.7%

      \[\sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{1 - \left(\left(1 - ux\right) + ux \cdot maxCos\right) \cdot \left(\left(1 - ux\right) + ux \cdot maxCos\right)} \]
    2. Taylor expanded in ux around 0

      \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\color{blue}{ux \cdot \left(\left(2 + -1 \cdot \left(ux \cdot {\left(maxCos - 1\right)}^{2}\right)\right) - 2 \cdot maxCos\right)}} \]
    3. Step-by-step derivation
      1. *-commutativeN/A

        \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(\left(2 + -1 \cdot \left(ux \cdot {\left(maxCos - 1\right)}^{2}\right)\right) - 2 \cdot maxCos\right) \cdot \color{blue}{ux}} \]
      2. lower-*.f32N/A

        \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(\left(2 + -1 \cdot \left(ux \cdot {\left(maxCos - 1\right)}^{2}\right)\right) - 2 \cdot maxCos\right) \cdot \color{blue}{ux}} \]
    4. Applied rewrites98.5%

      \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\color{blue}{\left(2 - \mathsf{fma}\left(\left(maxCos - 1\right) \cdot ux, maxCos - 1, maxCos + maxCos\right)\right) \cdot ux}} \]
    5. Taylor expanded in uy around 0

      \[\leadsto \color{blue}{\left(uy \cdot \left(\frac{-4}{3} \cdot \left({uy}^{2} \cdot {\mathsf{PI}\left(\right)}^{3}\right) + 2 \cdot \mathsf{PI}\left(\right)\right)\right)} \cdot \sqrt{\left(2 - \mathsf{fma}\left(\left(maxCos - 1\right) \cdot ux, maxCos - 1, maxCos + maxCos\right)\right) \cdot ux} \]
    6. Step-by-step derivation
      1. *-commutativeN/A

        \[\leadsto \left(uy \cdot \left(\frac{-4}{3} \cdot \left({uy}^{2} \cdot {\mathsf{PI}\left(\right)}^{3}\right) + 2 \cdot \mathsf{PI}\left(\right)\right)\right) \cdot \sqrt{\left(2 - \mathsf{fma}\left(\left(maxCos - 1\right) \cdot ux, maxCos - 1, maxCos + maxCos\right)\right) \cdot ux} \]
      2. associate-*r*N/A

        \[\leadsto \left(uy \cdot \left(\frac{-4}{3} \cdot \left({uy}^{2} \cdot {\mathsf{PI}\left(\right)}^{3}\right) + 2 \cdot \mathsf{PI}\left(\right)\right)\right) \cdot \sqrt{\left(2 - \mathsf{fma}\left(\left(maxCos - 1\right) \cdot ux, maxCos - 1, maxCos + maxCos\right)\right) \cdot ux} \]
      3. count-2-revN/A

        \[\leadsto \left(uy \cdot \left(\frac{-4}{3} \cdot \left({uy}^{2} \cdot {\mathsf{PI}\left(\right)}^{3}\right) + 2 \cdot \mathsf{PI}\left(\right)\right)\right) \cdot \sqrt{\left(2 - \mathsf{fma}\left(\left(maxCos - 1\right) \cdot ux, maxCos - 1, maxCos + maxCos\right)\right) \cdot ux} \]
      4. distribute-rgt-inN/A

        \[\leadsto \left(uy \cdot \left(\frac{-4}{3} \cdot \left({uy}^{2} \cdot {\mathsf{PI}\left(\right)}^{3}\right) + 2 \cdot \mathsf{PI}\left(\right)\right)\right) \cdot \sqrt{\left(2 - \mathsf{fma}\left(\left(maxCos - 1\right) \cdot ux, maxCos - 1, maxCos + maxCos\right)\right) \cdot ux} \]
      5. *-commutativeN/A

        \[\leadsto \left(\left(\frac{-4}{3} \cdot \left({uy}^{2} \cdot {\mathsf{PI}\left(\right)}^{3}\right) + 2 \cdot \mathsf{PI}\left(\right)\right) \cdot \color{blue}{uy}\right) \cdot \sqrt{\left(2 - \mathsf{fma}\left(\left(maxCos - 1\right) \cdot ux, maxCos - 1, maxCos + maxCos\right)\right) \cdot ux} \]
      6. lower-*.f32N/A

        \[\leadsto \left(\left(\frac{-4}{3} \cdot \left({uy}^{2} \cdot {\mathsf{PI}\left(\right)}^{3}\right) + 2 \cdot \mathsf{PI}\left(\right)\right) \cdot \color{blue}{uy}\right) \cdot \sqrt{\left(2 - \mathsf{fma}\left(\left(maxCos - 1\right) \cdot ux, maxCos - 1, maxCos + maxCos\right)\right) \cdot ux} \]
    7. Applied rewrites98.5%

      \[\leadsto \color{blue}{\left(\mathsf{fma}\left(\left(uy \cdot uy\right) \cdot \left(\left(\pi \cdot \pi\right) \cdot \pi\right), -1.3333333333333333, \pi + \pi\right) \cdot uy\right)} \cdot \sqrt{\left(2 - \mathsf{fma}\left(\left(maxCos - 1\right) \cdot ux, maxCos - 1, maxCos + maxCos\right)\right) \cdot ux} \]

    if 0.00689999992 < uy

    1. Initial program 57.3%

      \[\sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{1 - \left(\left(1 - ux\right) + ux \cdot maxCos\right) \cdot \left(\left(1 - ux\right) + ux \cdot maxCos\right)} \]
    2. Taylor expanded in maxCos around 0

      \[\leadsto \color{blue}{\sin \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \cdot \sqrt{1 - {\left(1 - ux\right)}^{2}}} \]
    3. Step-by-step derivation
      1. *-commutativeN/A

        \[\leadsto \sqrt{1 - {\left(1 - ux\right)}^{2}} \cdot \color{blue}{\sin \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right)} \]
      2. lower-*.f32N/A

        \[\leadsto \sqrt{1 - {\left(1 - ux\right)}^{2}} \cdot \color{blue}{\sin \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right)} \]
      3. lower-sqrt.f32N/A

        \[\leadsto \sqrt{1 - {\left(1 - ux\right)}^{2}} \cdot \sin \color{blue}{\left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right)} \]
      4. lower--.f32N/A

        \[\leadsto \sqrt{1 - {\left(1 - ux\right)}^{2}} \cdot \sin \left(\color{blue}{2} \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \]
      5. unpow2N/A

        \[\leadsto \sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \cdot \sin \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \]
      6. lower-*.f32N/A

        \[\leadsto \sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \cdot \sin \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \]
      7. lift--.f32N/A

        \[\leadsto \sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \cdot \sin \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \]
      8. lift--.f32N/A

        \[\leadsto \sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \cdot \sin \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \]
      9. associate-*r*N/A

        \[\leadsto \sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \cdot \sin \left(\left(2 \cdot uy\right) \cdot \mathsf{PI}\left(\right)\right) \]
      10. *-commutativeN/A

        \[\leadsto \sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \cdot \sin \left(\left(uy \cdot 2\right) \cdot \mathsf{PI}\left(\right)\right) \]
      11. lower-sin.f32N/A

        \[\leadsto \sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \cdot \sin \left(\left(uy \cdot 2\right) \cdot \mathsf{PI}\left(\right)\right) \]
      12. *-commutativeN/A

        \[\leadsto \sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \cdot \sin \left(\mathsf{PI}\left(\right) \cdot \left(uy \cdot 2\right)\right) \]
      13. lower-*.f32N/A

        \[\leadsto \sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \cdot \sin \left(\mathsf{PI}\left(\right) \cdot \left(uy \cdot 2\right)\right) \]
      14. lift-PI.f32N/A

        \[\leadsto \sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \cdot \sin \left(\pi \cdot \left(uy \cdot 2\right)\right) \]
      15. *-commutativeN/A

        \[\leadsto \sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \cdot \sin \left(\pi \cdot \left(2 \cdot uy\right)\right) \]
      16. count-2-revN/A

        \[\leadsto \sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \cdot \sin \left(\pi \cdot \left(uy + uy\right)\right) \]
      17. lower-+.f3254.8

        \[\leadsto \sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \cdot \sin \left(\pi \cdot \left(uy + uy\right)\right) \]
    4. Applied rewrites54.8%

      \[\leadsto \color{blue}{\sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \cdot \sin \left(\pi \cdot \left(uy + uy\right)\right)} \]
    5. Taylor expanded in ux around 0

      \[\leadsto \sqrt{ux \cdot \left(2 + -1 \cdot ux\right)} \cdot \sin \left(\color{blue}{\pi} \cdot \left(uy + uy\right)\right) \]
    6. Step-by-step derivation
      1. *-commutativeN/A

        \[\leadsto \sqrt{\left(2 + -1 \cdot ux\right) \cdot ux} \cdot \sin \left(\pi \cdot \left(uy + uy\right)\right) \]
      2. fp-cancel-sign-sub-invN/A

        \[\leadsto \sqrt{\left(2 - \left(\mathsf{neg}\left(-1\right)\right) \cdot ux\right) \cdot ux} \cdot \sin \left(\pi \cdot \left(uy + uy\right)\right) \]
      3. metadata-evalN/A

        \[\leadsto \sqrt{\left(2 - 1 \cdot ux\right) \cdot ux} \cdot \sin \left(\pi \cdot \left(uy + uy\right)\right) \]
      4. *-lft-identityN/A

        \[\leadsto \sqrt{\left(2 - ux\right) \cdot ux} \cdot \sin \left(\pi \cdot \left(uy + uy\right)\right) \]
      5. lower-*.f32N/A

        \[\leadsto \sqrt{\left(2 - ux\right) \cdot ux} \cdot \sin \left(\pi \cdot \left(uy + uy\right)\right) \]
      6. lower--.f3291.2

        \[\leadsto \sqrt{\left(2 - ux\right) \cdot ux} \cdot \sin \left(\pi \cdot \left(uy + uy\right)\right) \]
    7. Applied rewrites91.2%

      \[\leadsto \sqrt{\left(2 - ux\right) \cdot ux} \cdot \sin \left(\color{blue}{\pi} \cdot \left(uy + uy\right)\right) \]
  3. Recombined 2 regimes into one program.
  4. Add Preprocessing

Alternative 6: 96.7% accurate, 1.1× speedup?

\[\begin{array}{l} \\ \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\mathsf{fma}\left(-2 \cdot ux, maxCos, \left(2 - ux\right) \cdot ux\right)} \end{array} \]
(FPCore (ux uy maxCos)
 :precision binary32
 (* (sin (* (* uy 2.0) PI)) (sqrt (fma (* -2.0 ux) maxCos (* (- 2.0 ux) ux)))))
float code(float ux, float uy, float maxCos) {
	return sinf(((uy * 2.0f) * ((float) M_PI))) * sqrtf(fmaf((-2.0f * ux), maxCos, ((2.0f - ux) * ux)));
}
function code(ux, uy, maxCos)
	return Float32(sin(Float32(Float32(uy * Float32(2.0)) * Float32(pi))) * sqrt(fma(Float32(Float32(-2.0) * ux), maxCos, Float32(Float32(Float32(2.0) - ux) * ux))))
end
\begin{array}{l}

\\
\sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\mathsf{fma}\left(-2 \cdot ux, maxCos, \left(2 - ux\right) \cdot ux\right)}
\end{array}
Derivation
  1. Initial program 57.6%

    \[\sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{1 - \left(\left(1 - ux\right) + ux \cdot maxCos\right) \cdot \left(\left(1 - ux\right) + ux \cdot maxCos\right)} \]
  2. Taylor expanded in ux around 0

    \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\color{blue}{ux \cdot \left(\left(2 + -1 \cdot \left(ux \cdot {\left(maxCos - 1\right)}^{2}\right)\right) - 2 \cdot maxCos\right)}} \]
  3. Step-by-step derivation
    1. *-commutativeN/A

      \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(\left(2 + -1 \cdot \left(ux \cdot {\left(maxCos - 1\right)}^{2}\right)\right) - 2 \cdot maxCos\right) \cdot \color{blue}{ux}} \]
    2. lower-*.f32N/A

      \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(\left(2 + -1 \cdot \left(ux \cdot {\left(maxCos - 1\right)}^{2}\right)\right) - 2 \cdot maxCos\right) \cdot \color{blue}{ux}} \]
  4. Applied rewrites98.3%

    \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\color{blue}{\left(2 - \mathsf{fma}\left(\left(maxCos - 1\right) \cdot ux, maxCos - 1, maxCos + maxCos\right)\right) \cdot ux}} \]
  5. Taylor expanded in maxCos around 0

    \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{maxCos \cdot \left(-1 \cdot \left(maxCos \cdot {ux}^{2}\right) + -1 \cdot \left(ux \cdot \left(2 + -2 \cdot ux\right)\right)\right) + \color{blue}{ux \cdot \left(2 - ux\right)}} \]
  6. Step-by-step derivation
    1. *-commutativeN/A

      \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(-1 \cdot \left(maxCos \cdot {ux}^{2}\right) + -1 \cdot \left(ux \cdot \left(2 + -2 \cdot ux\right)\right)\right) \cdot maxCos + ux \cdot \left(\color{blue}{2} - ux\right)} \]
    2. lower-fma.f32N/A

      \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\mathsf{fma}\left(-1 \cdot \left(maxCos \cdot {ux}^{2}\right) + -1 \cdot \left(ux \cdot \left(2 + -2 \cdot ux\right)\right), maxCos, ux \cdot \left(2 - ux\right)\right)} \]
  7. Applied rewrites98.3%

    \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\mathsf{fma}\left(-\mathsf{fma}\left(ux \cdot ux, maxCos, \left(\left(2 - ux\right) - ux\right) \cdot ux\right), \color{blue}{maxCos}, \left(2 - ux\right) \cdot ux\right)} \]
  8. Taylor expanded in ux around 0

    \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\mathsf{fma}\left(-2 \cdot ux, maxCos, \left(2 - ux\right) \cdot ux\right)} \]
  9. Step-by-step derivation
    1. lower-*.f3296.7

      \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\mathsf{fma}\left(-2 \cdot ux, maxCos, \left(2 - ux\right) \cdot ux\right)} \]
  10. Applied rewrites96.7%

    \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\mathsf{fma}\left(-2 \cdot ux, maxCos, \left(2 - ux\right) \cdot ux\right)} \]
  11. Add Preprocessing

Alternative 7: 93.6% accurate, 1.2× speedup?

\[\begin{array}{l} \\ \begin{array}{l} \mathbf{if}\;uy \leq 0.07999999821186066:\\ \;\;\;\;\left(\mathsf{fma}\left(\left(uy \cdot uy\right) \cdot \left(\left(\pi \cdot \pi\right) \cdot \pi\right), -1.3333333333333333, \pi + \pi\right) \cdot uy\right) \cdot \sqrt{\left(2 - \mathsf{fma}\left(\left(maxCos - 1\right) \cdot ux, maxCos - 1, maxCos + maxCos\right)\right) \cdot ux}\\ \mathbf{else}:\\ \;\;\;\;\sqrt{ux + ux} \cdot \sin \left(\pi \cdot \left(uy + uy\right)\right)\\ \end{array} \end{array} \]
(FPCore (ux uy maxCos)
 :precision binary32
 (if (<= uy 0.07999999821186066)
   (*
    (* (fma (* (* uy uy) (* (* PI PI) PI)) -1.3333333333333333 (+ PI PI)) uy)
    (sqrt
     (*
      (- 2.0 (fma (* (- maxCos 1.0) ux) (- maxCos 1.0) (+ maxCos maxCos)))
      ux)))
   (* (sqrt (+ ux ux)) (sin (* PI (+ uy uy))))))
float code(float ux, float uy, float maxCos) {
	float tmp;
	if (uy <= 0.07999999821186066f) {
		tmp = (fmaf(((uy * uy) * ((((float) M_PI) * ((float) M_PI)) * ((float) M_PI))), -1.3333333333333333f, (((float) M_PI) + ((float) M_PI))) * uy) * sqrtf(((2.0f - fmaf(((maxCos - 1.0f) * ux), (maxCos - 1.0f), (maxCos + maxCos))) * ux));
	} else {
		tmp = sqrtf((ux + ux)) * sinf((((float) M_PI) * (uy + uy)));
	}
	return tmp;
}
function code(ux, uy, maxCos)
	tmp = Float32(0.0)
	if (uy <= Float32(0.07999999821186066))
		tmp = Float32(Float32(fma(Float32(Float32(uy * uy) * Float32(Float32(Float32(pi) * Float32(pi)) * Float32(pi))), Float32(-1.3333333333333333), Float32(Float32(pi) + Float32(pi))) * uy) * sqrt(Float32(Float32(Float32(2.0) - fma(Float32(Float32(maxCos - Float32(1.0)) * ux), Float32(maxCos - Float32(1.0)), Float32(maxCos + maxCos))) * ux)));
	else
		tmp = Float32(sqrt(Float32(ux + ux)) * sin(Float32(Float32(pi) * Float32(uy + uy))));
	end
	return tmp
end
\begin{array}{l}

\\
\begin{array}{l}
\mathbf{if}\;uy \leq 0.07999999821186066:\\
\;\;\;\;\left(\mathsf{fma}\left(\left(uy \cdot uy\right) \cdot \left(\left(\pi \cdot \pi\right) \cdot \pi\right), -1.3333333333333333, \pi + \pi\right) \cdot uy\right) \cdot \sqrt{\left(2 - \mathsf{fma}\left(\left(maxCos - 1\right) \cdot ux, maxCos - 1, maxCos + maxCos\right)\right) \cdot ux}\\

\mathbf{else}:\\
\;\;\;\;\sqrt{ux + ux} \cdot \sin \left(\pi \cdot \left(uy + uy\right)\right)\\


\end{array}
\end{array}
Derivation
  1. Split input into 2 regimes
  2. if uy < 0.0799999982

    1. Initial program 57.7%

      \[\sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{1 - \left(\left(1 - ux\right) + ux \cdot maxCos\right) \cdot \left(\left(1 - ux\right) + ux \cdot maxCos\right)} \]
    2. Taylor expanded in ux around 0

      \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\color{blue}{ux \cdot \left(\left(2 + -1 \cdot \left(ux \cdot {\left(maxCos - 1\right)}^{2}\right)\right) - 2 \cdot maxCos\right)}} \]
    3. Step-by-step derivation
      1. *-commutativeN/A

        \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(\left(2 + -1 \cdot \left(ux \cdot {\left(maxCos - 1\right)}^{2}\right)\right) - 2 \cdot maxCos\right) \cdot \color{blue}{ux}} \]
      2. lower-*.f32N/A

        \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(\left(2 + -1 \cdot \left(ux \cdot {\left(maxCos - 1\right)}^{2}\right)\right) - 2 \cdot maxCos\right) \cdot \color{blue}{ux}} \]
    4. Applied rewrites98.5%

      \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\color{blue}{\left(2 - \mathsf{fma}\left(\left(maxCos - 1\right) \cdot ux, maxCos - 1, maxCos + maxCos\right)\right) \cdot ux}} \]
    5. Taylor expanded in uy around 0

      \[\leadsto \color{blue}{\left(uy \cdot \left(\frac{-4}{3} \cdot \left({uy}^{2} \cdot {\mathsf{PI}\left(\right)}^{3}\right) + 2 \cdot \mathsf{PI}\left(\right)\right)\right)} \cdot \sqrt{\left(2 - \mathsf{fma}\left(\left(maxCos - 1\right) \cdot ux, maxCos - 1, maxCos + maxCos\right)\right) \cdot ux} \]
    6. Step-by-step derivation
      1. *-commutativeN/A

        \[\leadsto \left(uy \cdot \left(\frac{-4}{3} \cdot \left({uy}^{2} \cdot {\mathsf{PI}\left(\right)}^{3}\right) + 2 \cdot \mathsf{PI}\left(\right)\right)\right) \cdot \sqrt{\left(2 - \mathsf{fma}\left(\left(maxCos - 1\right) \cdot ux, maxCos - 1, maxCos + maxCos\right)\right) \cdot ux} \]
      2. associate-*r*N/A

        \[\leadsto \left(uy \cdot \left(\frac{-4}{3} \cdot \left({uy}^{2} \cdot {\mathsf{PI}\left(\right)}^{3}\right) + 2 \cdot \mathsf{PI}\left(\right)\right)\right) \cdot \sqrt{\left(2 - \mathsf{fma}\left(\left(maxCos - 1\right) \cdot ux, maxCos - 1, maxCos + maxCos\right)\right) \cdot ux} \]
      3. count-2-revN/A

        \[\leadsto \left(uy \cdot \left(\frac{-4}{3} \cdot \left({uy}^{2} \cdot {\mathsf{PI}\left(\right)}^{3}\right) + 2 \cdot \mathsf{PI}\left(\right)\right)\right) \cdot \sqrt{\left(2 - \mathsf{fma}\left(\left(maxCos - 1\right) \cdot ux, maxCos - 1, maxCos + maxCos\right)\right) \cdot ux} \]
      4. distribute-rgt-inN/A

        \[\leadsto \left(uy \cdot \left(\frac{-4}{3} \cdot \left({uy}^{2} \cdot {\mathsf{PI}\left(\right)}^{3}\right) + 2 \cdot \mathsf{PI}\left(\right)\right)\right) \cdot \sqrt{\left(2 - \mathsf{fma}\left(\left(maxCos - 1\right) \cdot ux, maxCos - 1, maxCos + maxCos\right)\right) \cdot ux} \]
      5. *-commutativeN/A

        \[\leadsto \left(\left(\frac{-4}{3} \cdot \left({uy}^{2} \cdot {\mathsf{PI}\left(\right)}^{3}\right) + 2 \cdot \mathsf{PI}\left(\right)\right) \cdot \color{blue}{uy}\right) \cdot \sqrt{\left(2 - \mathsf{fma}\left(\left(maxCos - 1\right) \cdot ux, maxCos - 1, maxCos + maxCos\right)\right) \cdot ux} \]
      6. lower-*.f32N/A

        \[\leadsto \left(\left(\frac{-4}{3} \cdot \left({uy}^{2} \cdot {\mathsf{PI}\left(\right)}^{3}\right) + 2 \cdot \mathsf{PI}\left(\right)\right) \cdot \color{blue}{uy}\right) \cdot \sqrt{\left(2 - \mathsf{fma}\left(\left(maxCos - 1\right) \cdot ux, maxCos - 1, maxCos + maxCos\right)\right) \cdot ux} \]
    7. Applied rewrites96.3%

      \[\leadsto \color{blue}{\left(\mathsf{fma}\left(\left(uy \cdot uy\right) \cdot \left(\left(\pi \cdot \pi\right) \cdot \pi\right), -1.3333333333333333, \pi + \pi\right) \cdot uy\right)} \cdot \sqrt{\left(2 - \mathsf{fma}\left(\left(maxCos - 1\right) \cdot ux, maxCos - 1, maxCos + maxCos\right)\right) \cdot ux} \]

    if 0.0799999982 < uy

    1. Initial program 56.8%

      \[\sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{1 - \left(\left(1 - ux\right) + ux \cdot maxCos\right) \cdot \left(\left(1 - ux\right) + ux \cdot maxCos\right)} \]
    2. Taylor expanded in maxCos around 0

      \[\leadsto \color{blue}{\sin \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \cdot \sqrt{1 - {\left(1 - ux\right)}^{2}}} \]
    3. Step-by-step derivation
      1. *-commutativeN/A

        \[\leadsto \sqrt{1 - {\left(1 - ux\right)}^{2}} \cdot \color{blue}{\sin \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right)} \]
      2. lower-*.f32N/A

        \[\leadsto \sqrt{1 - {\left(1 - ux\right)}^{2}} \cdot \color{blue}{\sin \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right)} \]
      3. lower-sqrt.f32N/A

        \[\leadsto \sqrt{1 - {\left(1 - ux\right)}^{2}} \cdot \sin \color{blue}{\left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right)} \]
      4. lower--.f32N/A

        \[\leadsto \sqrt{1 - {\left(1 - ux\right)}^{2}} \cdot \sin \left(\color{blue}{2} \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \]
      5. unpow2N/A

        \[\leadsto \sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \cdot \sin \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \]
      6. lower-*.f32N/A

        \[\leadsto \sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \cdot \sin \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \]
      7. lift--.f32N/A

        \[\leadsto \sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \cdot \sin \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \]
      8. lift--.f32N/A

        \[\leadsto \sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \cdot \sin \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \]
      9. associate-*r*N/A

        \[\leadsto \sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \cdot \sin \left(\left(2 \cdot uy\right) \cdot \mathsf{PI}\left(\right)\right) \]
      10. *-commutativeN/A

        \[\leadsto \sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \cdot \sin \left(\left(uy \cdot 2\right) \cdot \mathsf{PI}\left(\right)\right) \]
      11. lower-sin.f32N/A

        \[\leadsto \sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \cdot \sin \left(\left(uy \cdot 2\right) \cdot \mathsf{PI}\left(\right)\right) \]
      12. *-commutativeN/A

        \[\leadsto \sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \cdot \sin \left(\mathsf{PI}\left(\right) \cdot \left(uy \cdot 2\right)\right) \]
      13. lower-*.f32N/A

        \[\leadsto \sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \cdot \sin \left(\mathsf{PI}\left(\right) \cdot \left(uy \cdot 2\right)\right) \]
      14. lift-PI.f32N/A

        \[\leadsto \sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \cdot \sin \left(\pi \cdot \left(uy \cdot 2\right)\right) \]
      15. *-commutativeN/A

        \[\leadsto \sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \cdot \sin \left(\pi \cdot \left(2 \cdot uy\right)\right) \]
      16. count-2-revN/A

        \[\leadsto \sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \cdot \sin \left(\pi \cdot \left(uy + uy\right)\right) \]
      17. lower-+.f3254.2

        \[\leadsto \sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \cdot \sin \left(\pi \cdot \left(uy + uy\right)\right) \]
    4. Applied rewrites54.2%

      \[\leadsto \color{blue}{\sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \cdot \sin \left(\pi \cdot \left(uy + uy\right)\right)} \]
    5. Taylor expanded in ux around 0

      \[\leadsto \sqrt{2 \cdot ux} \cdot \sin \left(\color{blue}{\pi} \cdot \left(uy + uy\right)\right) \]
    6. Step-by-step derivation
      1. count-2-revN/A

        \[\leadsto \sqrt{ux + ux} \cdot \sin \left(\pi \cdot \left(uy + uy\right)\right) \]
      2. lower-+.f3272.1

        \[\leadsto \sqrt{ux + ux} \cdot \sin \left(\pi \cdot \left(uy + uy\right)\right) \]
    7. Applied rewrites72.1%

      \[\leadsto \sqrt{ux + ux} \cdot \sin \left(\color{blue}{\pi} \cdot \left(uy + uy\right)\right) \]
  3. Recombined 2 regimes into one program.
  4. Add Preprocessing

Alternative 8: 89.7% accurate, 1.2× speedup?

\[\begin{array}{l} \\ \begin{array}{l} \mathbf{if}\;uy \leq 0.003000000026077032:\\ \;\;\;\;\left(\sqrt{ux \cdot \left(\mathsf{fma}\left(-ux, \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - \left(maxCos + maxCos\right)\right)} \cdot \pi\right) \cdot \left(uy + uy\right)\\ \mathbf{else}:\\ \;\;\;\;\left(\sqrt{2} \cdot \sin \left(\left(uy + uy\right) \cdot \pi\right)\right) \cdot \sqrt{ux}\\ \end{array} \end{array} \]
(FPCore (ux uy maxCos)
 :precision binary32
 (if (<= uy 0.003000000026077032)
   (*
    (*
     (sqrt
      (*
       ux
       (-
        (fma (- ux) (* (- maxCos 1.0) (- maxCos 1.0)) 2.0)
        (+ maxCos maxCos))))
     PI)
    (+ uy uy))
   (* (* (sqrt 2.0) (sin (* (+ uy uy) PI))) (sqrt ux))))
float code(float ux, float uy, float maxCos) {
	float tmp;
	if (uy <= 0.003000000026077032f) {
		tmp = (sqrtf((ux * (fmaf(-ux, ((maxCos - 1.0f) * (maxCos - 1.0f)), 2.0f) - (maxCos + maxCos)))) * ((float) M_PI)) * (uy + uy);
	} else {
		tmp = (sqrtf(2.0f) * sinf(((uy + uy) * ((float) M_PI)))) * sqrtf(ux);
	}
	return tmp;
}
function code(ux, uy, maxCos)
	tmp = Float32(0.0)
	if (uy <= Float32(0.003000000026077032))
		tmp = Float32(Float32(sqrt(Float32(ux * Float32(fma(Float32(-ux), Float32(Float32(maxCos - Float32(1.0)) * Float32(maxCos - Float32(1.0))), Float32(2.0)) - Float32(maxCos + maxCos)))) * Float32(pi)) * Float32(uy + uy));
	else
		tmp = Float32(Float32(sqrt(Float32(2.0)) * sin(Float32(Float32(uy + uy) * Float32(pi)))) * sqrt(ux));
	end
	return tmp
end
\begin{array}{l}

\\
\begin{array}{l}
\mathbf{if}\;uy \leq 0.003000000026077032:\\
\;\;\;\;\left(\sqrt{ux \cdot \left(\mathsf{fma}\left(-ux, \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - \left(maxCos + maxCos\right)\right)} \cdot \pi\right) \cdot \left(uy + uy\right)\\

\mathbf{else}:\\
\;\;\;\;\left(\sqrt{2} \cdot \sin \left(\left(uy + uy\right) \cdot \pi\right)\right) \cdot \sqrt{ux}\\


\end{array}
\end{array}
Derivation
  1. Split input into 2 regimes
  2. if uy < 0.00300000003

    1. Initial program 57.8%

      \[\sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{1 - \left(\left(1 - ux\right) + ux \cdot maxCos\right) \cdot \left(\left(1 - ux\right) + ux \cdot maxCos\right)} \]
    2. Taylor expanded in uy around 0

      \[\leadsto \color{blue}{2 \cdot \left(\left(uy \cdot \mathsf{PI}\left(\right)\right) \cdot \sqrt{1 - {\left(\left(1 + maxCos \cdot ux\right) - ux\right)}^{2}}\right)} \]
    3. Applied rewrites57.1%

      \[\leadsto \color{blue}{\left(\sqrt{1 - \mathsf{fma}\left(maxCos, ux, 1 - ux\right) \cdot \mathsf{fma}\left(maxCos, ux, 1 - ux\right)} \cdot \pi\right) \cdot \left(uy + uy\right)} \]
    4. Taylor expanded in ux around 0

      \[\leadsto \left(\sqrt{1 - 1} \cdot \pi\right) \cdot \left(uy + uy\right) \]
    5. Step-by-step derivation
      1. Applied rewrites7.3%

        \[\leadsto \left(\sqrt{1 - 1} \cdot \pi\right) \cdot \left(uy + uy\right) \]
      2. Taylor expanded in ux around 0

        \[\leadsto \left(\sqrt{ux \cdot \left(\left(2 + -1 \cdot \left(ux \cdot {\left(maxCos - 1\right)}^{2}\right)\right) - 2 \cdot maxCos\right)} \cdot \pi\right) \cdot \left(uy + uy\right) \]
      3. Step-by-step derivation
        1. lower-*.f32N/A

          \[\leadsto \left(\sqrt{ux \cdot \left(\left(2 + -1 \cdot \left(ux \cdot {\left(maxCos - 1\right)}^{2}\right)\right) - 2 \cdot maxCos\right)} \cdot \pi\right) \cdot \left(uy + uy\right) \]
        2. lower--.f32N/A

          \[\leadsto \left(\sqrt{ux \cdot \left(\left(2 + -1 \cdot \left(ux \cdot {\left(maxCos - 1\right)}^{2}\right)\right) - 2 \cdot maxCos\right)} \cdot \pi\right) \cdot \left(uy + uy\right) \]
        3. +-commutativeN/A

          \[\leadsto \left(\sqrt{ux \cdot \left(\left(-1 \cdot \left(ux \cdot {\left(maxCos - 1\right)}^{2}\right) + 2\right) - 2 \cdot maxCos\right)} \cdot \pi\right) \cdot \left(uy + uy\right) \]
        4. pow2N/A

          \[\leadsto \left(\sqrt{ux \cdot \left(\left(-1 \cdot \left(ux \cdot \left(\left(maxCos - 1\right) \cdot \left(maxCos - 1\right)\right)\right) + 2\right) - 2 \cdot maxCos\right)} \cdot \pi\right) \cdot \left(uy + uy\right) \]
        5. associate-*r*N/A

          \[\leadsto \left(\sqrt{ux \cdot \left(\left(\left(-1 \cdot ux\right) \cdot \left(\left(maxCos - 1\right) \cdot \left(maxCos - 1\right)\right) + 2\right) - 2 \cdot maxCos\right)} \cdot \pi\right) \cdot \left(uy + uy\right) \]
        6. mul-1-negN/A

          \[\leadsto \left(\sqrt{ux \cdot \left(\left(\left(\mathsf{neg}\left(ux\right)\right) \cdot \left(\left(maxCos - 1\right) \cdot \left(maxCos - 1\right)\right) + 2\right) - 2 \cdot maxCos\right)} \cdot \pi\right) \cdot \left(uy + uy\right) \]
        7. lower-fma.f32N/A

          \[\leadsto \left(\sqrt{ux \cdot \left(\mathsf{fma}\left(\mathsf{neg}\left(ux\right), \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - 2 \cdot maxCos\right)} \cdot \pi\right) \cdot \left(uy + uy\right) \]
        8. lower-neg.f32N/A

          \[\leadsto \left(\sqrt{ux \cdot \left(\mathsf{fma}\left(-ux, \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - 2 \cdot maxCos\right)} \cdot \pi\right) \cdot \left(uy + uy\right) \]
        9. lift--.f32N/A

          \[\leadsto \left(\sqrt{ux \cdot \left(\mathsf{fma}\left(-ux, \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - 2 \cdot maxCos\right)} \cdot \pi\right) \cdot \left(uy + uy\right) \]
        10. lift--.f32N/A

          \[\leadsto \left(\sqrt{ux \cdot \left(\mathsf{fma}\left(-ux, \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - 2 \cdot maxCos\right)} \cdot \pi\right) \cdot \left(uy + uy\right) \]
        11. lift-*.f32N/A

          \[\leadsto \left(\sqrt{ux \cdot \left(\mathsf{fma}\left(-ux, \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - 2 \cdot maxCos\right)} \cdot \pi\right) \cdot \left(uy + uy\right) \]
        12. count-2-revN/A

          \[\leadsto \left(\sqrt{ux \cdot \left(\mathsf{fma}\left(-ux, \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - \left(maxCos + maxCos\right)\right)} \cdot \pi\right) \cdot \left(uy + uy\right) \]
        13. lift-+.f3295.6

          \[\leadsto \left(\sqrt{ux \cdot \left(\mathsf{fma}\left(-ux, \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - \left(maxCos + maxCos\right)\right)} \cdot \pi\right) \cdot \left(uy + uy\right) \]
      4. Applied rewrites95.6%

        \[\leadsto \left(\sqrt{ux \cdot \left(\mathsf{fma}\left(-ux, \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - \left(maxCos + maxCos\right)\right)} \cdot \pi\right) \cdot \left(uy + uy\right) \]

      if 0.00300000003 < uy

      1. Initial program 56.9%

        \[\sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{1 - \left(\left(1 - ux\right) + ux \cdot maxCos\right) \cdot \left(\left(1 - ux\right) + ux \cdot maxCos\right)} \]
      2. Taylor expanded in maxCos around 0

        \[\leadsto \color{blue}{\sin \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \cdot \sqrt{1 - {\left(1 - ux\right)}^{2}}} \]
      3. Step-by-step derivation
        1. *-commutativeN/A

          \[\leadsto \sqrt{1 - {\left(1 - ux\right)}^{2}} \cdot \color{blue}{\sin \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right)} \]
        2. lower-*.f32N/A

          \[\leadsto \sqrt{1 - {\left(1 - ux\right)}^{2}} \cdot \color{blue}{\sin \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right)} \]
        3. lower-sqrt.f32N/A

          \[\leadsto \sqrt{1 - {\left(1 - ux\right)}^{2}} \cdot \sin \color{blue}{\left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right)} \]
        4. lower--.f32N/A

          \[\leadsto \sqrt{1 - {\left(1 - ux\right)}^{2}} \cdot \sin \left(\color{blue}{2} \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \]
        5. unpow2N/A

          \[\leadsto \sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \cdot \sin \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \]
        6. lower-*.f32N/A

          \[\leadsto \sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \cdot \sin \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \]
        7. lift--.f32N/A

          \[\leadsto \sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \cdot \sin \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \]
        8. lift--.f32N/A

          \[\leadsto \sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \cdot \sin \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \]
        9. associate-*r*N/A

          \[\leadsto \sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \cdot \sin \left(\left(2 \cdot uy\right) \cdot \mathsf{PI}\left(\right)\right) \]
        10. *-commutativeN/A

          \[\leadsto \sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \cdot \sin \left(\left(uy \cdot 2\right) \cdot \mathsf{PI}\left(\right)\right) \]
        11. lower-sin.f32N/A

          \[\leadsto \sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \cdot \sin \left(\left(uy \cdot 2\right) \cdot \mathsf{PI}\left(\right)\right) \]
        12. *-commutativeN/A

          \[\leadsto \sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \cdot \sin \left(\mathsf{PI}\left(\right) \cdot \left(uy \cdot 2\right)\right) \]
        13. lower-*.f32N/A

          \[\leadsto \sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \cdot \sin \left(\mathsf{PI}\left(\right) \cdot \left(uy \cdot 2\right)\right) \]
        14. lift-PI.f32N/A

          \[\leadsto \sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \cdot \sin \left(\pi \cdot \left(uy \cdot 2\right)\right) \]
        15. *-commutativeN/A

          \[\leadsto \sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \cdot \sin \left(\pi \cdot \left(2 \cdot uy\right)\right) \]
        16. count-2-revN/A

          \[\leadsto \sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \cdot \sin \left(\pi \cdot \left(uy + uy\right)\right) \]
        17. lower-+.f3254.6

          \[\leadsto \sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \cdot \sin \left(\pi \cdot \left(uy + uy\right)\right) \]
      4. Applied rewrites54.6%

        \[\leadsto \color{blue}{\sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \cdot \sin \left(\pi \cdot \left(uy + uy\right)\right)} \]
      5. Taylor expanded in ux around 0

        \[\leadsto \sqrt{ux} \cdot \color{blue}{\left(\sin \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \cdot \sqrt{2}\right)} \]
      6. Step-by-step derivation
        1. *-commutativeN/A

          \[\leadsto \left(\sin \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \cdot \sqrt{2}\right) \cdot \sqrt{ux} \]
        2. lower-*.f32N/A

          \[\leadsto \left(\sin \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \cdot \sqrt{2}\right) \cdot \sqrt{ux} \]
        3. *-commutativeN/A

          \[\leadsto \left(\sqrt{2} \cdot \sin \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right)\right) \cdot \sqrt{ux} \]
        4. lower-*.f32N/A

          \[\leadsto \left(\sqrt{2} \cdot \sin \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right)\right) \cdot \sqrt{ux} \]
        5. lower-sqrt.f32N/A

          \[\leadsto \left(\sqrt{2} \cdot \sin \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right)\right) \cdot \sqrt{ux} \]
        6. lower-sin.f32N/A

          \[\leadsto \left(\sqrt{2} \cdot \sin \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right)\right) \cdot \sqrt{ux} \]
        7. associate-*r*N/A

          \[\leadsto \left(\sqrt{2} \cdot \sin \left(\left(2 \cdot uy\right) \cdot \mathsf{PI}\left(\right)\right)\right) \cdot \sqrt{ux} \]
        8. *-commutativeN/A

          \[\leadsto \left(\sqrt{2} \cdot \sin \left(\left(uy \cdot 2\right) \cdot \mathsf{PI}\left(\right)\right)\right) \cdot \sqrt{ux} \]
        9. lift-*.f32N/A

          \[\leadsto \left(\sqrt{2} \cdot \sin \left(\left(uy \cdot 2\right) \cdot \mathsf{PI}\left(\right)\right)\right) \cdot \sqrt{ux} \]
        10. lift-*.f32N/A

          \[\leadsto \left(\sqrt{2} \cdot \sin \left(\left(uy \cdot 2\right) \cdot \mathsf{PI}\left(\right)\right)\right) \cdot \sqrt{ux} \]
        11. lift-PI.f32N/A

          \[\leadsto \left(\sqrt{2} \cdot \sin \left(\left(uy \cdot 2\right) \cdot \pi\right)\right) \cdot \sqrt{ux} \]
        12. lift-*.f32N/A

          \[\leadsto \left(\sqrt{2} \cdot \sin \left(\left(uy \cdot 2\right) \cdot \pi\right)\right) \cdot \sqrt{ux} \]
        13. *-commutativeN/A

          \[\leadsto \left(\sqrt{2} \cdot \sin \left(\left(2 \cdot uy\right) \cdot \pi\right)\right) \cdot \sqrt{ux} \]
        14. count-2-revN/A

          \[\leadsto \left(\sqrt{2} \cdot \sin \left(\left(uy + uy\right) \cdot \pi\right)\right) \cdot \sqrt{ux} \]
        15. lift-+.f32N/A

          \[\leadsto \left(\sqrt{2} \cdot \sin \left(\left(uy + uy\right) \cdot \pi\right)\right) \cdot \sqrt{ux} \]
        16. lower-sqrt.f3273.1

          \[\leadsto \left(\sqrt{2} \cdot \sin \left(\left(uy + uy\right) \cdot \pi\right)\right) \cdot \sqrt{ux} \]
      7. Applied rewrites73.1%

        \[\leadsto \left(\sqrt{2} \cdot \sin \left(\left(uy + uy\right) \cdot \pi\right)\right) \cdot \color{blue}{\sqrt{ux}} \]
    6. Recombined 2 regimes into one program.
    7. Add Preprocessing

    Alternative 9: 89.7% accurate, 1.3× speedup?

    \[\begin{array}{l} \\ \begin{array}{l} \mathbf{if}\;uy \leq 0.003000000026077032:\\ \;\;\;\;\left(\sqrt{ux \cdot \left(\mathsf{fma}\left(-ux, \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - \left(maxCos + maxCos\right)\right)} \cdot \pi\right) \cdot \left(uy + uy\right)\\ \mathbf{else}:\\ \;\;\;\;\sqrt{ux + ux} \cdot \sin \left(\pi \cdot \left(uy + uy\right)\right)\\ \end{array} \end{array} \]
    (FPCore (ux uy maxCos)
     :precision binary32
     (if (<= uy 0.003000000026077032)
       (*
        (*
         (sqrt
          (*
           ux
           (-
            (fma (- ux) (* (- maxCos 1.0) (- maxCos 1.0)) 2.0)
            (+ maxCos maxCos))))
         PI)
        (+ uy uy))
       (* (sqrt (+ ux ux)) (sin (* PI (+ uy uy))))))
    float code(float ux, float uy, float maxCos) {
    	float tmp;
    	if (uy <= 0.003000000026077032f) {
    		tmp = (sqrtf((ux * (fmaf(-ux, ((maxCos - 1.0f) * (maxCos - 1.0f)), 2.0f) - (maxCos + maxCos)))) * ((float) M_PI)) * (uy + uy);
    	} else {
    		tmp = sqrtf((ux + ux)) * sinf((((float) M_PI) * (uy + uy)));
    	}
    	return tmp;
    }
    
    function code(ux, uy, maxCos)
    	tmp = Float32(0.0)
    	if (uy <= Float32(0.003000000026077032))
    		tmp = Float32(Float32(sqrt(Float32(ux * Float32(fma(Float32(-ux), Float32(Float32(maxCos - Float32(1.0)) * Float32(maxCos - Float32(1.0))), Float32(2.0)) - Float32(maxCos + maxCos)))) * Float32(pi)) * Float32(uy + uy));
    	else
    		tmp = Float32(sqrt(Float32(ux + ux)) * sin(Float32(Float32(pi) * Float32(uy + uy))));
    	end
    	return tmp
    end
    
    \begin{array}{l}
    
    \\
    \begin{array}{l}
    \mathbf{if}\;uy \leq 0.003000000026077032:\\
    \;\;\;\;\left(\sqrt{ux \cdot \left(\mathsf{fma}\left(-ux, \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - \left(maxCos + maxCos\right)\right)} \cdot \pi\right) \cdot \left(uy + uy\right)\\
    
    \mathbf{else}:\\
    \;\;\;\;\sqrt{ux + ux} \cdot \sin \left(\pi \cdot \left(uy + uy\right)\right)\\
    
    
    \end{array}
    \end{array}
    
    Derivation
    1. Split input into 2 regimes
    2. if uy < 0.00300000003

      1. Initial program 57.8%

        \[\sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{1 - \left(\left(1 - ux\right) + ux \cdot maxCos\right) \cdot \left(\left(1 - ux\right) + ux \cdot maxCos\right)} \]
      2. Taylor expanded in uy around 0

        \[\leadsto \color{blue}{2 \cdot \left(\left(uy \cdot \mathsf{PI}\left(\right)\right) \cdot \sqrt{1 - {\left(\left(1 + maxCos \cdot ux\right) - ux\right)}^{2}}\right)} \]
      3. Applied rewrites57.1%

        \[\leadsto \color{blue}{\left(\sqrt{1 - \mathsf{fma}\left(maxCos, ux, 1 - ux\right) \cdot \mathsf{fma}\left(maxCos, ux, 1 - ux\right)} \cdot \pi\right) \cdot \left(uy + uy\right)} \]
      4. Taylor expanded in ux around 0

        \[\leadsto \left(\sqrt{1 - 1} \cdot \pi\right) \cdot \left(uy + uy\right) \]
      5. Step-by-step derivation
        1. Applied rewrites7.3%

          \[\leadsto \left(\sqrt{1 - 1} \cdot \pi\right) \cdot \left(uy + uy\right) \]
        2. Taylor expanded in ux around 0

          \[\leadsto \left(\sqrt{ux \cdot \left(\left(2 + -1 \cdot \left(ux \cdot {\left(maxCos - 1\right)}^{2}\right)\right) - 2 \cdot maxCos\right)} \cdot \pi\right) \cdot \left(uy + uy\right) \]
        3. Step-by-step derivation
          1. lower-*.f32N/A

            \[\leadsto \left(\sqrt{ux \cdot \left(\left(2 + -1 \cdot \left(ux \cdot {\left(maxCos - 1\right)}^{2}\right)\right) - 2 \cdot maxCos\right)} \cdot \pi\right) \cdot \left(uy + uy\right) \]
          2. lower--.f32N/A

            \[\leadsto \left(\sqrt{ux \cdot \left(\left(2 + -1 \cdot \left(ux \cdot {\left(maxCos - 1\right)}^{2}\right)\right) - 2 \cdot maxCos\right)} \cdot \pi\right) \cdot \left(uy + uy\right) \]
          3. +-commutativeN/A

            \[\leadsto \left(\sqrt{ux \cdot \left(\left(-1 \cdot \left(ux \cdot {\left(maxCos - 1\right)}^{2}\right) + 2\right) - 2 \cdot maxCos\right)} \cdot \pi\right) \cdot \left(uy + uy\right) \]
          4. pow2N/A

            \[\leadsto \left(\sqrt{ux \cdot \left(\left(-1 \cdot \left(ux \cdot \left(\left(maxCos - 1\right) \cdot \left(maxCos - 1\right)\right)\right) + 2\right) - 2 \cdot maxCos\right)} \cdot \pi\right) \cdot \left(uy + uy\right) \]
          5. associate-*r*N/A

            \[\leadsto \left(\sqrt{ux \cdot \left(\left(\left(-1 \cdot ux\right) \cdot \left(\left(maxCos - 1\right) \cdot \left(maxCos - 1\right)\right) + 2\right) - 2 \cdot maxCos\right)} \cdot \pi\right) \cdot \left(uy + uy\right) \]
          6. mul-1-negN/A

            \[\leadsto \left(\sqrt{ux \cdot \left(\left(\left(\mathsf{neg}\left(ux\right)\right) \cdot \left(\left(maxCos - 1\right) \cdot \left(maxCos - 1\right)\right) + 2\right) - 2 \cdot maxCos\right)} \cdot \pi\right) \cdot \left(uy + uy\right) \]
          7. lower-fma.f32N/A

            \[\leadsto \left(\sqrt{ux \cdot \left(\mathsf{fma}\left(\mathsf{neg}\left(ux\right), \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - 2 \cdot maxCos\right)} \cdot \pi\right) \cdot \left(uy + uy\right) \]
          8. lower-neg.f32N/A

            \[\leadsto \left(\sqrt{ux \cdot \left(\mathsf{fma}\left(-ux, \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - 2 \cdot maxCos\right)} \cdot \pi\right) \cdot \left(uy + uy\right) \]
          9. lift--.f32N/A

            \[\leadsto \left(\sqrt{ux \cdot \left(\mathsf{fma}\left(-ux, \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - 2 \cdot maxCos\right)} \cdot \pi\right) \cdot \left(uy + uy\right) \]
          10. lift--.f32N/A

            \[\leadsto \left(\sqrt{ux \cdot \left(\mathsf{fma}\left(-ux, \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - 2 \cdot maxCos\right)} \cdot \pi\right) \cdot \left(uy + uy\right) \]
          11. lift-*.f32N/A

            \[\leadsto \left(\sqrt{ux \cdot \left(\mathsf{fma}\left(-ux, \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - 2 \cdot maxCos\right)} \cdot \pi\right) \cdot \left(uy + uy\right) \]
          12. count-2-revN/A

            \[\leadsto \left(\sqrt{ux \cdot \left(\mathsf{fma}\left(-ux, \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - \left(maxCos + maxCos\right)\right)} \cdot \pi\right) \cdot \left(uy + uy\right) \]
          13. lift-+.f3295.6

            \[\leadsto \left(\sqrt{ux \cdot \left(\mathsf{fma}\left(-ux, \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - \left(maxCos + maxCos\right)\right)} \cdot \pi\right) \cdot \left(uy + uy\right) \]
        4. Applied rewrites95.6%

          \[\leadsto \left(\sqrt{ux \cdot \left(\mathsf{fma}\left(-ux, \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - \left(maxCos + maxCos\right)\right)} \cdot \pi\right) \cdot \left(uy + uy\right) \]

        if 0.00300000003 < uy

        1. Initial program 56.9%

          \[\sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{1 - \left(\left(1 - ux\right) + ux \cdot maxCos\right) \cdot \left(\left(1 - ux\right) + ux \cdot maxCos\right)} \]
        2. Taylor expanded in maxCos around 0

          \[\leadsto \color{blue}{\sin \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \cdot \sqrt{1 - {\left(1 - ux\right)}^{2}}} \]
        3. Step-by-step derivation
          1. *-commutativeN/A

            \[\leadsto \sqrt{1 - {\left(1 - ux\right)}^{2}} \cdot \color{blue}{\sin \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right)} \]
          2. lower-*.f32N/A

            \[\leadsto \sqrt{1 - {\left(1 - ux\right)}^{2}} \cdot \color{blue}{\sin \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right)} \]
          3. lower-sqrt.f32N/A

            \[\leadsto \sqrt{1 - {\left(1 - ux\right)}^{2}} \cdot \sin \color{blue}{\left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right)} \]
          4. lower--.f32N/A

            \[\leadsto \sqrt{1 - {\left(1 - ux\right)}^{2}} \cdot \sin \left(\color{blue}{2} \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \]
          5. unpow2N/A

            \[\leadsto \sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \cdot \sin \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \]
          6. lower-*.f32N/A

            \[\leadsto \sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \cdot \sin \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \]
          7. lift--.f32N/A

            \[\leadsto \sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \cdot \sin \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \]
          8. lift--.f32N/A

            \[\leadsto \sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \cdot \sin \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \]
          9. associate-*r*N/A

            \[\leadsto \sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \cdot \sin \left(\left(2 \cdot uy\right) \cdot \mathsf{PI}\left(\right)\right) \]
          10. *-commutativeN/A

            \[\leadsto \sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \cdot \sin \left(\left(uy \cdot 2\right) \cdot \mathsf{PI}\left(\right)\right) \]
          11. lower-sin.f32N/A

            \[\leadsto \sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \cdot \sin \left(\left(uy \cdot 2\right) \cdot \mathsf{PI}\left(\right)\right) \]
          12. *-commutativeN/A

            \[\leadsto \sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \cdot \sin \left(\mathsf{PI}\left(\right) \cdot \left(uy \cdot 2\right)\right) \]
          13. lower-*.f32N/A

            \[\leadsto \sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \cdot \sin \left(\mathsf{PI}\left(\right) \cdot \left(uy \cdot 2\right)\right) \]
          14. lift-PI.f32N/A

            \[\leadsto \sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \cdot \sin \left(\pi \cdot \left(uy \cdot 2\right)\right) \]
          15. *-commutativeN/A

            \[\leadsto \sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \cdot \sin \left(\pi \cdot \left(2 \cdot uy\right)\right) \]
          16. count-2-revN/A

            \[\leadsto \sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \cdot \sin \left(\pi \cdot \left(uy + uy\right)\right) \]
          17. lower-+.f3254.6

            \[\leadsto \sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \cdot \sin \left(\pi \cdot \left(uy + uy\right)\right) \]
        4. Applied rewrites54.6%

          \[\leadsto \color{blue}{\sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \cdot \sin \left(\pi \cdot \left(uy + uy\right)\right)} \]
        5. Taylor expanded in ux around 0

          \[\leadsto \sqrt{2 \cdot ux} \cdot \sin \left(\color{blue}{\pi} \cdot \left(uy + uy\right)\right) \]
        6. Step-by-step derivation
          1. count-2-revN/A

            \[\leadsto \sqrt{ux + ux} \cdot \sin \left(\pi \cdot \left(uy + uy\right)\right) \]
          2. lower-+.f3273.1

            \[\leadsto \sqrt{ux + ux} \cdot \sin \left(\pi \cdot \left(uy + uy\right)\right) \]
        7. Applied rewrites73.1%

          \[\leadsto \sqrt{ux + ux} \cdot \sin \left(\color{blue}{\pi} \cdot \left(uy + uy\right)\right) \]
      6. Recombined 2 regimes into one program.
      7. Add Preprocessing

      Alternative 10: 81.4% accurate, 2.0× speedup?

      \[\begin{array}{l} \\ \left(\sqrt{ux \cdot \left(\mathsf{fma}\left(-ux, \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - \left(maxCos + maxCos\right)\right)} \cdot \pi\right) \cdot \left(uy + uy\right) \end{array} \]
      (FPCore (ux uy maxCos)
       :precision binary32
       (*
        (*
         (sqrt
          (*
           ux
           (- (fma (- ux) (* (- maxCos 1.0) (- maxCos 1.0)) 2.0) (+ maxCos maxCos))))
         PI)
        (+ uy uy)))
      float code(float ux, float uy, float maxCos) {
      	return (sqrtf((ux * (fmaf(-ux, ((maxCos - 1.0f) * (maxCos - 1.0f)), 2.0f) - (maxCos + maxCos)))) * ((float) M_PI)) * (uy + uy);
      }
      
      function code(ux, uy, maxCos)
      	return Float32(Float32(sqrt(Float32(ux * Float32(fma(Float32(-ux), Float32(Float32(maxCos - Float32(1.0)) * Float32(maxCos - Float32(1.0))), Float32(2.0)) - Float32(maxCos + maxCos)))) * Float32(pi)) * Float32(uy + uy))
      end
      
      \begin{array}{l}
      
      \\
      \left(\sqrt{ux \cdot \left(\mathsf{fma}\left(-ux, \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - \left(maxCos + maxCos\right)\right)} \cdot \pi\right) \cdot \left(uy + uy\right)
      \end{array}
      
      Derivation
      1. Initial program 57.6%

        \[\sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{1 - \left(\left(1 - ux\right) + ux \cdot maxCos\right) \cdot \left(\left(1 - ux\right) + ux \cdot maxCos\right)} \]
      2. Taylor expanded in uy around 0

        \[\leadsto \color{blue}{2 \cdot \left(\left(uy \cdot \mathsf{PI}\left(\right)\right) \cdot \sqrt{1 - {\left(\left(1 + maxCos \cdot ux\right) - ux\right)}^{2}}\right)} \]
      3. Applied rewrites50.7%

        \[\leadsto \color{blue}{\left(\sqrt{1 - \mathsf{fma}\left(maxCos, ux, 1 - ux\right) \cdot \mathsf{fma}\left(maxCos, ux, 1 - ux\right)} \cdot \pi\right) \cdot \left(uy + uy\right)} \]
      4. Taylor expanded in ux around 0

        \[\leadsto \left(\sqrt{1 - 1} \cdot \pi\right) \cdot \left(uy + uy\right) \]
      5. Step-by-step derivation
        1. Applied rewrites7.1%

          \[\leadsto \left(\sqrt{1 - 1} \cdot \pi\right) \cdot \left(uy + uy\right) \]
        2. Taylor expanded in ux around 0

          \[\leadsto \left(\sqrt{ux \cdot \left(\left(2 + -1 \cdot \left(ux \cdot {\left(maxCos - 1\right)}^{2}\right)\right) - 2 \cdot maxCos\right)} \cdot \pi\right) \cdot \left(uy + uy\right) \]
        3. Step-by-step derivation
          1. lower-*.f32N/A

            \[\leadsto \left(\sqrt{ux \cdot \left(\left(2 + -1 \cdot \left(ux \cdot {\left(maxCos - 1\right)}^{2}\right)\right) - 2 \cdot maxCos\right)} \cdot \pi\right) \cdot \left(uy + uy\right) \]
          2. lower--.f32N/A

            \[\leadsto \left(\sqrt{ux \cdot \left(\left(2 + -1 \cdot \left(ux \cdot {\left(maxCos - 1\right)}^{2}\right)\right) - 2 \cdot maxCos\right)} \cdot \pi\right) \cdot \left(uy + uy\right) \]
          3. +-commutativeN/A

            \[\leadsto \left(\sqrt{ux \cdot \left(\left(-1 \cdot \left(ux \cdot {\left(maxCos - 1\right)}^{2}\right) + 2\right) - 2 \cdot maxCos\right)} \cdot \pi\right) \cdot \left(uy + uy\right) \]
          4. pow2N/A

            \[\leadsto \left(\sqrt{ux \cdot \left(\left(-1 \cdot \left(ux \cdot \left(\left(maxCos - 1\right) \cdot \left(maxCos - 1\right)\right)\right) + 2\right) - 2 \cdot maxCos\right)} \cdot \pi\right) \cdot \left(uy + uy\right) \]
          5. associate-*r*N/A

            \[\leadsto \left(\sqrt{ux \cdot \left(\left(\left(-1 \cdot ux\right) \cdot \left(\left(maxCos - 1\right) \cdot \left(maxCos - 1\right)\right) + 2\right) - 2 \cdot maxCos\right)} \cdot \pi\right) \cdot \left(uy + uy\right) \]
          6. mul-1-negN/A

            \[\leadsto \left(\sqrt{ux \cdot \left(\left(\left(\mathsf{neg}\left(ux\right)\right) \cdot \left(\left(maxCos - 1\right) \cdot \left(maxCos - 1\right)\right) + 2\right) - 2 \cdot maxCos\right)} \cdot \pi\right) \cdot \left(uy + uy\right) \]
          7. lower-fma.f32N/A

            \[\leadsto \left(\sqrt{ux \cdot \left(\mathsf{fma}\left(\mathsf{neg}\left(ux\right), \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - 2 \cdot maxCos\right)} \cdot \pi\right) \cdot \left(uy + uy\right) \]
          8. lower-neg.f32N/A

            \[\leadsto \left(\sqrt{ux \cdot \left(\mathsf{fma}\left(-ux, \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - 2 \cdot maxCos\right)} \cdot \pi\right) \cdot \left(uy + uy\right) \]
          9. lift--.f32N/A

            \[\leadsto \left(\sqrt{ux \cdot \left(\mathsf{fma}\left(-ux, \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - 2 \cdot maxCos\right)} \cdot \pi\right) \cdot \left(uy + uy\right) \]
          10. lift--.f32N/A

            \[\leadsto \left(\sqrt{ux \cdot \left(\mathsf{fma}\left(-ux, \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - 2 \cdot maxCos\right)} \cdot \pi\right) \cdot \left(uy + uy\right) \]
          11. lift-*.f32N/A

            \[\leadsto \left(\sqrt{ux \cdot \left(\mathsf{fma}\left(-ux, \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - 2 \cdot maxCos\right)} \cdot \pi\right) \cdot \left(uy + uy\right) \]
          12. count-2-revN/A

            \[\leadsto \left(\sqrt{ux \cdot \left(\mathsf{fma}\left(-ux, \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - \left(maxCos + maxCos\right)\right)} \cdot \pi\right) \cdot \left(uy + uy\right) \]
          13. lift-+.f3281.4

            \[\leadsto \left(\sqrt{ux \cdot \left(\mathsf{fma}\left(-ux, \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - \left(maxCos + maxCos\right)\right)} \cdot \pi\right) \cdot \left(uy + uy\right) \]
        4. Applied rewrites81.4%

          \[\leadsto \left(\sqrt{ux \cdot \left(\mathsf{fma}\left(-ux, \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - \left(maxCos + maxCos\right)\right)} \cdot \pi\right) \cdot \left(uy + uy\right) \]
        5. Add Preprocessing

        Alternative 11: 81.4% accurate, 2.0× speedup?

        \[\begin{array}{l} \\ \left(\left(uy + uy\right) \cdot \pi\right) \cdot \sqrt{\left(2 - \mathsf{fma}\left(\left(maxCos - 1\right) \cdot ux, maxCos - 1, maxCos + maxCos\right)\right) \cdot ux} \end{array} \]
        (FPCore (ux uy maxCos)
         :precision binary32
         (*
          (* (+ uy uy) PI)
          (sqrt
           (*
            (- 2.0 (fma (* (- maxCos 1.0) ux) (- maxCos 1.0) (+ maxCos maxCos)))
            ux))))
        float code(float ux, float uy, float maxCos) {
        	return ((uy + uy) * ((float) M_PI)) * sqrtf(((2.0f - fmaf(((maxCos - 1.0f) * ux), (maxCos - 1.0f), (maxCos + maxCos))) * ux));
        }
        
        function code(ux, uy, maxCos)
        	return Float32(Float32(Float32(uy + uy) * Float32(pi)) * sqrt(Float32(Float32(Float32(2.0) - fma(Float32(Float32(maxCos - Float32(1.0)) * ux), Float32(maxCos - Float32(1.0)), Float32(maxCos + maxCos))) * ux)))
        end
        
        \begin{array}{l}
        
        \\
        \left(\left(uy + uy\right) \cdot \pi\right) \cdot \sqrt{\left(2 - \mathsf{fma}\left(\left(maxCos - 1\right) \cdot ux, maxCos - 1, maxCos + maxCos\right)\right) \cdot ux}
        \end{array}
        
        Derivation
        1. Initial program 57.6%

          \[\sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{1 - \left(\left(1 - ux\right) + ux \cdot maxCos\right) \cdot \left(\left(1 - ux\right) + ux \cdot maxCos\right)} \]
        2. Taylor expanded in ux around 0

          \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\color{blue}{ux \cdot \left(\left(2 + -1 \cdot \left(ux \cdot {\left(maxCos - 1\right)}^{2}\right)\right) - 2 \cdot maxCos\right)}} \]
        3. Step-by-step derivation
          1. *-commutativeN/A

            \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(\left(2 + -1 \cdot \left(ux \cdot {\left(maxCos - 1\right)}^{2}\right)\right) - 2 \cdot maxCos\right) \cdot \color{blue}{ux}} \]
          2. lower-*.f32N/A

            \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(\left(2 + -1 \cdot \left(ux \cdot {\left(maxCos - 1\right)}^{2}\right)\right) - 2 \cdot maxCos\right) \cdot \color{blue}{ux}} \]
        4. Applied rewrites98.3%

          \[\leadsto \sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\color{blue}{\left(2 - \mathsf{fma}\left(\left(maxCos - 1\right) \cdot ux, maxCos - 1, maxCos + maxCos\right)\right) \cdot ux}} \]
        5. Taylor expanded in uy around 0

          \[\leadsto \color{blue}{\left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right)} \cdot \sqrt{\left(2 - \mathsf{fma}\left(\left(maxCos - 1\right) \cdot ux, maxCos - 1, maxCos + maxCos\right)\right) \cdot ux} \]
        6. Step-by-step derivation
          1. *-commutativeN/A

            \[\leadsto \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \cdot \sqrt{\left(2 - \mathsf{fma}\left(\left(maxCos - 1\right) \cdot ux, maxCos - 1, maxCos + maxCos\right)\right) \cdot ux} \]
          2. associate-*r*N/A

            \[\leadsto \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \cdot \sqrt{\left(2 - \mathsf{fma}\left(\left(maxCos - 1\right) \cdot ux, maxCos - 1, maxCos + maxCos\right)\right) \cdot ux} \]
          3. count-2-revN/A

            \[\leadsto \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \cdot \sqrt{\left(2 - \mathsf{fma}\left(\left(maxCos - 1\right) \cdot ux, maxCos - 1, maxCos + maxCos\right)\right) \cdot ux} \]
          4. distribute-rgt-inN/A

            \[\leadsto \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \cdot \sqrt{\left(2 - \mathsf{fma}\left(\left(maxCos - 1\right) \cdot ux, maxCos - 1, maxCos + maxCos\right)\right) \cdot ux} \]
          5. associate-*r*N/A

            \[\leadsto \left(\left(2 \cdot uy\right) \cdot \color{blue}{\mathsf{PI}\left(\right)}\right) \cdot \sqrt{\left(2 - \mathsf{fma}\left(\left(maxCos - 1\right) \cdot ux, maxCos - 1, maxCos + maxCos\right)\right) \cdot ux} \]
          6. count-2-revN/A

            \[\leadsto \left(\left(uy + uy\right) \cdot \mathsf{PI}\left(\right)\right) \cdot \sqrt{\left(2 - \mathsf{fma}\left(\left(maxCos - 1\right) \cdot ux, maxCos - 1, maxCos + maxCos\right)\right) \cdot ux} \]
          7. lift-+.f32N/A

            \[\leadsto \left(\left(uy + uy\right) \cdot \mathsf{PI}\left(\right)\right) \cdot \sqrt{\left(2 - \mathsf{fma}\left(\left(maxCos - 1\right) \cdot ux, maxCos - 1, maxCos + maxCos\right)\right) \cdot ux} \]
          8. lower-*.f32N/A

            \[\leadsto \left(\left(uy + uy\right) \cdot \color{blue}{\mathsf{PI}\left(\right)}\right) \cdot \sqrt{\left(2 - \mathsf{fma}\left(\left(maxCos - 1\right) \cdot ux, maxCos - 1, maxCos + maxCos\right)\right) \cdot ux} \]
          9. lift-PI.f3281.4

            \[\leadsto \left(\left(uy + uy\right) \cdot \pi\right) \cdot \sqrt{\left(2 - \mathsf{fma}\left(\left(maxCos - 1\right) \cdot ux, maxCos - 1, maxCos + maxCos\right)\right) \cdot ux} \]
        7. Applied rewrites81.4%

          \[\leadsto \color{blue}{\left(\left(uy + uy\right) \cdot \pi\right)} \cdot \sqrt{\left(2 - \mathsf{fma}\left(\left(maxCos - 1\right) \cdot ux, maxCos - 1, maxCos + maxCos\right)\right) \cdot ux} \]
        8. Add Preprocessing

        Alternative 12: 76.7% accurate, 1.2× speedup?

        \[\begin{array}{l} \\ \begin{array}{l} t_0 := \mathsf{fma}\left(maxCos - 1, ux, 1\right)\\ t_1 := \left(1 - ux\right) + ux \cdot maxCos\\ \mathbf{if}\;t\_1 \cdot t\_1 \leq 0.9998000264167786:\\ \;\;\;\;\left(\sqrt{1 - t\_0 \cdot t\_0} \cdot \pi\right) \cdot \left(uy + uy\right)\\ \mathbf{else}:\\ \;\;\;\;\sqrt{\mathsf{fma}\left(maxCos \cdot ux, -2, ux + ux\right)} \cdot \left(\left(uy + uy\right) \cdot \pi\right)\\ \end{array} \end{array} \]
        (FPCore (ux uy maxCos)
         :precision binary32
         (let* ((t_0 (fma (- maxCos 1.0) ux 1.0)) (t_1 (+ (- 1.0 ux) (* ux maxCos))))
           (if (<= (* t_1 t_1) 0.9998000264167786)
             (* (* (sqrt (- 1.0 (* t_0 t_0))) PI) (+ uy uy))
             (* (sqrt (fma (* maxCos ux) -2.0 (+ ux ux))) (* (+ uy uy) PI)))))
        float code(float ux, float uy, float maxCos) {
        	float t_0 = fmaf((maxCos - 1.0f), ux, 1.0f);
        	float t_1 = (1.0f - ux) + (ux * maxCos);
        	float tmp;
        	if ((t_1 * t_1) <= 0.9998000264167786f) {
        		tmp = (sqrtf((1.0f - (t_0 * t_0))) * ((float) M_PI)) * (uy + uy);
        	} else {
        		tmp = sqrtf(fmaf((maxCos * ux), -2.0f, (ux + ux))) * ((uy + uy) * ((float) M_PI));
        	}
        	return tmp;
        }
        
        function code(ux, uy, maxCos)
        	t_0 = fma(Float32(maxCos - Float32(1.0)), ux, Float32(1.0))
        	t_1 = Float32(Float32(Float32(1.0) - ux) + Float32(ux * maxCos))
        	tmp = Float32(0.0)
        	if (Float32(t_1 * t_1) <= Float32(0.9998000264167786))
        		tmp = Float32(Float32(sqrt(Float32(Float32(1.0) - Float32(t_0 * t_0))) * Float32(pi)) * Float32(uy + uy));
        	else
        		tmp = Float32(sqrt(fma(Float32(maxCos * ux), Float32(-2.0), Float32(ux + ux))) * Float32(Float32(uy + uy) * Float32(pi)));
        	end
        	return tmp
        end
        
        \begin{array}{l}
        
        \\
        \begin{array}{l}
        t_0 := \mathsf{fma}\left(maxCos - 1, ux, 1\right)\\
        t_1 := \left(1 - ux\right) + ux \cdot maxCos\\
        \mathbf{if}\;t\_1 \cdot t\_1 \leq 0.9998000264167786:\\
        \;\;\;\;\left(\sqrt{1 - t\_0 \cdot t\_0} \cdot \pi\right) \cdot \left(uy + uy\right)\\
        
        \mathbf{else}:\\
        \;\;\;\;\sqrt{\mathsf{fma}\left(maxCos \cdot ux, -2, ux + ux\right)} \cdot \left(\left(uy + uy\right) \cdot \pi\right)\\
        
        
        \end{array}
        \end{array}
        
        Derivation
        1. Split input into 2 regimes
        2. if (*.f32 (+.f32 (-.f32 #s(literal 1 binary32) ux) (*.f32 ux maxCos)) (+.f32 (-.f32 #s(literal 1 binary32) ux) (*.f32 ux maxCos))) < 0.999800026

          1. Initial program 88.0%

            \[\sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{1 - \left(\left(1 - ux\right) + ux \cdot maxCos\right) \cdot \left(\left(1 - ux\right) + ux \cdot maxCos\right)} \]
          2. Taylor expanded in uy around 0

            \[\leadsto \color{blue}{2 \cdot \left(\left(uy \cdot \mathsf{PI}\left(\right)\right) \cdot \sqrt{1 - {\left(\left(1 + maxCos \cdot ux\right) - ux\right)}^{2}}\right)} \]
          3. Applied rewrites75.0%

            \[\leadsto \color{blue}{\left(\sqrt{1 - \mathsf{fma}\left(maxCos, ux, 1 - ux\right) \cdot \mathsf{fma}\left(maxCos, ux, 1 - ux\right)} \cdot \pi\right) \cdot \left(uy + uy\right)} \]
          4. Taylor expanded in ux around 0

            \[\leadsto \left(\sqrt{1 - \left(1 + ux \cdot \left(maxCos - 1\right)\right) \cdot \mathsf{fma}\left(maxCos, ux, 1 - ux\right)} \cdot \pi\right) \cdot \left(uy + uy\right) \]
          5. Step-by-step derivation
            1. *-commutativeN/A

              \[\leadsto \left(\sqrt{1 - \left(1 + \left(maxCos - 1\right) \cdot ux\right) \cdot \mathsf{fma}\left(maxCos, ux, 1 - ux\right)} \cdot \pi\right) \cdot \left(uy + uy\right) \]
            2. +-commutativeN/A

              \[\leadsto \left(\sqrt{1 - \left(\left(maxCos - 1\right) \cdot ux + 1\right) \cdot \mathsf{fma}\left(maxCos, ux, 1 - ux\right)} \cdot \pi\right) \cdot \left(uy + uy\right) \]
            3. lower-fma.f32N/A

              \[\leadsto \left(\sqrt{1 - \mathsf{fma}\left(maxCos - 1, ux, 1\right) \cdot \mathsf{fma}\left(maxCos, ux, 1 - ux\right)} \cdot \pi\right) \cdot \left(uy + uy\right) \]
            4. lift--.f3275.1

              \[\leadsto \left(\sqrt{1 - \mathsf{fma}\left(maxCos - 1, ux, 1\right) \cdot \mathsf{fma}\left(maxCos, ux, 1 - ux\right)} \cdot \pi\right) \cdot \left(uy + uy\right) \]
          6. Applied rewrites75.1%

            \[\leadsto \left(\sqrt{1 - \mathsf{fma}\left(maxCos - 1, ux, 1\right) \cdot \mathsf{fma}\left(maxCos, ux, 1 - ux\right)} \cdot \pi\right) \cdot \left(uy + uy\right) \]
          7. Taylor expanded in ux around 0

            \[\leadsto \left(\sqrt{1 - \mathsf{fma}\left(maxCos - 1, ux, 1\right) \cdot \left(1 + ux \cdot \left(maxCos - 1\right)\right)} \cdot \pi\right) \cdot \left(uy + uy\right) \]
          8. Step-by-step derivation
            1. *-commutativeN/A

              \[\leadsto \left(\sqrt{1 - \mathsf{fma}\left(maxCos - 1, ux, 1\right) \cdot \left(1 + \left(maxCos - 1\right) \cdot ux\right)} \cdot \pi\right) \cdot \left(uy + uy\right) \]
            2. +-commutativeN/A

              \[\leadsto \left(\sqrt{1 - \mathsf{fma}\left(maxCos - 1, ux, 1\right) \cdot \left(\left(maxCos - 1\right) \cdot ux + 1\right)} \cdot \pi\right) \cdot \left(uy + uy\right) \]
            3. lower-fma.f32N/A

              \[\leadsto \left(\sqrt{1 - \mathsf{fma}\left(maxCos - 1, ux, 1\right) \cdot \mathsf{fma}\left(maxCos - 1, ux, 1\right)} \cdot \pi\right) \cdot \left(uy + uy\right) \]
            4. lift--.f3275.1

              \[\leadsto \left(\sqrt{1 - \mathsf{fma}\left(maxCos - 1, ux, 1\right) \cdot \mathsf{fma}\left(maxCos - 1, ux, 1\right)} \cdot \pi\right) \cdot \left(uy + uy\right) \]
          9. Applied rewrites75.1%

            \[\leadsto \left(\sqrt{1 - \mathsf{fma}\left(maxCos - 1, ux, 1\right) \cdot \mathsf{fma}\left(maxCos - 1, ux, 1\right)} \cdot \pi\right) \cdot \left(uy + uy\right) \]

          if 0.999800026 < (*.f32 (+.f32 (-.f32 #s(literal 1 binary32) ux) (*.f32 ux maxCos)) (+.f32 (-.f32 #s(literal 1 binary32) ux) (*.f32 ux maxCos)))

          1. Initial program 35.4%

            \[\sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{1 - \left(\left(1 - ux\right) + ux \cdot maxCos\right) \cdot \left(\left(1 - ux\right) + ux \cdot maxCos\right)} \]
          2. Taylor expanded in uy around 0

            \[\leadsto \color{blue}{2 \cdot \left(\left(uy \cdot \mathsf{PI}\left(\right)\right) \cdot \sqrt{1 - {\left(\left(1 + maxCos \cdot ux\right) - ux\right)}^{2}}\right)} \]
          3. Applied rewrites33.0%

            \[\leadsto \color{blue}{\left(\sqrt{1 - \mathsf{fma}\left(maxCos, ux, 1 - ux\right) \cdot \mathsf{fma}\left(maxCos, ux, 1 - ux\right)} \cdot \pi\right) \cdot \left(uy + uy\right)} \]
          4. Taylor expanded in ux around 0

            \[\leadsto 2 \cdot \color{blue}{\left(\sqrt{ux \cdot \left(2 - 2 \cdot maxCos\right)} \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right)} \]
          5. Step-by-step derivation
            1. count-2-revN/A

              \[\leadsto \sqrt{ux \cdot \left(2 - 2 \cdot maxCos\right)} \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right) + \sqrt{ux \cdot \left(2 - 2 \cdot maxCos\right)} \cdot \color{blue}{\left(uy \cdot \mathsf{PI}\left(\right)\right)} \]
            2. distribute-lft-outN/A

              \[\leadsto \sqrt{ux \cdot \left(2 - 2 \cdot maxCos\right)} \cdot \left(uy \cdot \mathsf{PI}\left(\right) + \color{blue}{uy \cdot \mathsf{PI}\left(\right)}\right) \]
            3. count-2-revN/A

              \[\leadsto \sqrt{ux \cdot \left(2 - 2 \cdot maxCos\right)} \cdot \left(2 \cdot \left(uy \cdot \color{blue}{\mathsf{PI}\left(\right)}\right)\right) \]
            4. lower-*.f32N/A

              \[\leadsto \sqrt{ux \cdot \left(2 - 2 \cdot maxCos\right)} \cdot \left(2 \cdot \color{blue}{\left(uy \cdot \mathsf{PI}\left(\right)\right)}\right) \]
            5. lower-sqrt.f32N/A

              \[\leadsto \sqrt{ux \cdot \left(2 - 2 \cdot maxCos\right)} \cdot \left(2 \cdot \left(\color{blue}{uy} \cdot \mathsf{PI}\left(\right)\right)\right) \]
            6. *-commutativeN/A

              \[\leadsto \sqrt{\left(2 - 2 \cdot maxCos\right) \cdot ux} \cdot \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \]
            7. lower-*.f32N/A

              \[\leadsto \sqrt{\left(2 - 2 \cdot maxCos\right) \cdot ux} \cdot \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \]
            8. count-2-revN/A

              \[\leadsto \sqrt{\left(2 - \left(maxCos + maxCos\right)\right) \cdot ux} \cdot \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \]
            9. associate--r+N/A

              \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \]
            10. lower--.f32N/A

              \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \]
            11. lower--.f32N/A

              \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \]
            12. associate-*r*N/A

              \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(\left(2 \cdot uy\right) \cdot \mathsf{PI}\left(\right)\right) \]
            13. count-2-revN/A

              \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \mathsf{PI}\left(\right)\right) \]
            14. lift-+.f32N/A

              \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \mathsf{PI}\left(\right)\right) \]
            15. lower-*.f32N/A

              \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \mathsf{PI}\left(\right)\right) \]
            16. lift-PI.f3277.8

              \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
          6. Applied rewrites77.8%

            \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \color{blue}{\left(\left(uy + uy\right) \cdot \pi\right)} \]
          7. Step-by-step derivation
            1. lift-*.f32N/A

              \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
            2. lift--.f32N/A

              \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
            3. lift--.f32N/A

              \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
            4. associate--l-N/A

              \[\leadsto \sqrt{\left(2 - \left(maxCos + maxCos\right)\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
            5. count-2-revN/A

              \[\leadsto \sqrt{\left(2 - 2 \cdot maxCos\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
            6. fp-cancel-sub-sign-invN/A

              \[\leadsto \sqrt{\left(2 + \left(\mathsf{neg}\left(2\right)\right) \cdot maxCos\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
            7. metadata-evalN/A

              \[\leadsto \sqrt{\left(2 + -2 \cdot maxCos\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
            8. *-commutativeN/A

              \[\leadsto \sqrt{ux \cdot \left(2 + -2 \cdot maxCos\right)} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
            9. distribute-rgt-inN/A

              \[\leadsto \sqrt{2 \cdot ux + \left(-2 \cdot maxCos\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
            10. associate-*r*N/A

              \[\leadsto \sqrt{2 \cdot ux + -2 \cdot \left(maxCos \cdot ux\right)} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
            11. +-commutativeN/A

              \[\leadsto \sqrt{-2 \cdot \left(maxCos \cdot ux\right) + 2 \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
            12. *-commutativeN/A

              \[\leadsto \sqrt{\left(maxCos \cdot ux\right) \cdot -2 + 2 \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
            13. lower-fma.f32N/A

              \[\leadsto \sqrt{\mathsf{fma}\left(maxCos \cdot ux, -2, 2 \cdot ux\right)} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
            14. lower-*.f32N/A

              \[\leadsto \sqrt{\mathsf{fma}\left(maxCos \cdot ux, -2, 2 \cdot ux\right)} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
            15. count-2-revN/A

              \[\leadsto \sqrt{\mathsf{fma}\left(maxCos \cdot ux, -2, ux + ux\right)} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
            16. lower-+.f3277.8

              \[\leadsto \sqrt{\mathsf{fma}\left(maxCos \cdot ux, -2, ux + ux\right)} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
          8. Applied rewrites77.8%

            \[\leadsto \sqrt{\mathsf{fma}\left(maxCos \cdot ux, -2, ux + ux\right)} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
        3. Recombined 2 regimes into one program.
        4. Add Preprocessing

        Alternative 13: 75.4% accurate, 1.5× speedup?

        \[\begin{array}{l} \\ \begin{array}{l} t_0 := \left(1 - ux\right) + ux \cdot maxCos\\ \mathbf{if}\;t\_0 \cdot t\_0 \leq 0.9995800256729126:\\ \;\;\;\;\left(\sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \cdot \pi\right) \cdot \left(uy + uy\right)\\ \mathbf{else}:\\ \;\;\;\;\sqrt{\mathsf{fma}\left(maxCos \cdot ux, -2, ux + ux\right)} \cdot \left(\left(uy + uy\right) \cdot \pi\right)\\ \end{array} \end{array} \]
        (FPCore (ux uy maxCos)
         :precision binary32
         (let* ((t_0 (+ (- 1.0 ux) (* ux maxCos))))
           (if (<= (* t_0 t_0) 0.9995800256729126)
             (* (* (sqrt (- 1.0 (* (- 1.0 ux) (- 1.0 ux)))) PI) (+ uy uy))
             (* (sqrt (fma (* maxCos ux) -2.0 (+ ux ux))) (* (+ uy uy) PI)))))
        float code(float ux, float uy, float maxCos) {
        	float t_0 = (1.0f - ux) + (ux * maxCos);
        	float tmp;
        	if ((t_0 * t_0) <= 0.9995800256729126f) {
        		tmp = (sqrtf((1.0f - ((1.0f - ux) * (1.0f - ux)))) * ((float) M_PI)) * (uy + uy);
        	} else {
        		tmp = sqrtf(fmaf((maxCos * ux), -2.0f, (ux + ux))) * ((uy + uy) * ((float) M_PI));
        	}
        	return tmp;
        }
        
        function code(ux, uy, maxCos)
        	t_0 = Float32(Float32(Float32(1.0) - ux) + Float32(ux * maxCos))
        	tmp = Float32(0.0)
        	if (Float32(t_0 * t_0) <= Float32(0.9995800256729126))
        		tmp = Float32(Float32(sqrt(Float32(Float32(1.0) - Float32(Float32(Float32(1.0) - ux) * Float32(Float32(1.0) - ux)))) * Float32(pi)) * Float32(uy + uy));
        	else
        		tmp = Float32(sqrt(fma(Float32(maxCos * ux), Float32(-2.0), Float32(ux + ux))) * Float32(Float32(uy + uy) * Float32(pi)));
        	end
        	return tmp
        end
        
        \begin{array}{l}
        
        \\
        \begin{array}{l}
        t_0 := \left(1 - ux\right) + ux \cdot maxCos\\
        \mathbf{if}\;t\_0 \cdot t\_0 \leq 0.9995800256729126:\\
        \;\;\;\;\left(\sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \cdot \pi\right) \cdot \left(uy + uy\right)\\
        
        \mathbf{else}:\\
        \;\;\;\;\sqrt{\mathsf{fma}\left(maxCos \cdot ux, -2, ux + ux\right)} \cdot \left(\left(uy + uy\right) \cdot \pi\right)\\
        
        
        \end{array}
        \end{array}
        
        Derivation
        1. Split input into 2 regimes
        2. if (*.f32 (+.f32 (-.f32 #s(literal 1 binary32) ux) (*.f32 ux maxCos)) (+.f32 (-.f32 #s(literal 1 binary32) ux) (*.f32 ux maxCos))) < 0.999580026

          1. Initial program 89.6%

            \[\sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{1 - \left(\left(1 - ux\right) + ux \cdot maxCos\right) \cdot \left(\left(1 - ux\right) + ux \cdot maxCos\right)} \]
          2. Taylor expanded in uy around 0

            \[\leadsto \color{blue}{2 \cdot \left(\left(uy \cdot \mathsf{PI}\left(\right)\right) \cdot \sqrt{1 - {\left(\left(1 + maxCos \cdot ux\right) - ux\right)}^{2}}\right)} \]
          3. Applied rewrites76.0%

            \[\leadsto \color{blue}{\left(\sqrt{1 - \mathsf{fma}\left(maxCos, ux, 1 - ux\right) \cdot \mathsf{fma}\left(maxCos, ux, 1 - ux\right)} \cdot \pi\right) \cdot \left(uy + uy\right)} \]
          4. Taylor expanded in maxCos around 0

            \[\leadsto \left(\sqrt{1 - \left(1 - ux\right) \cdot \mathsf{fma}\left(maxCos, ux, 1 - ux\right)} \cdot \pi\right) \cdot \left(uy + uy\right) \]
          5. Step-by-step derivation
            1. lift--.f3273.0

              \[\leadsto \left(\sqrt{1 - \left(1 - ux\right) \cdot \mathsf{fma}\left(maxCos, ux, 1 - ux\right)} \cdot \pi\right) \cdot \left(uy + uy\right) \]
          6. Applied rewrites73.0%

            \[\leadsto \left(\sqrt{1 - \left(1 - ux\right) \cdot \mathsf{fma}\left(maxCos, ux, 1 - ux\right)} \cdot \pi\right) \cdot \left(uy + uy\right) \]
          7. Taylor expanded in maxCos around 0

            \[\leadsto \left(\sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \cdot \pi\right) \cdot \left(uy + uy\right) \]
          8. Step-by-step derivation
            1. lift--.f3272.7

              \[\leadsto \left(\sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \cdot \pi\right) \cdot \left(uy + uy\right) \]
          9. Applied rewrites72.7%

            \[\leadsto \left(\sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \cdot \pi\right) \cdot \left(uy + uy\right) \]

          if 0.999580026 < (*.f32 (+.f32 (-.f32 #s(literal 1 binary32) ux) (*.f32 ux maxCos)) (+.f32 (-.f32 #s(literal 1 binary32) ux) (*.f32 ux maxCos)))

          1. Initial program 37.6%

            \[\sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{1 - \left(\left(1 - ux\right) + ux \cdot maxCos\right) \cdot \left(\left(1 - ux\right) + ux \cdot maxCos\right)} \]
          2. Taylor expanded in uy around 0

            \[\leadsto \color{blue}{2 \cdot \left(\left(uy \cdot \mathsf{PI}\left(\right)\right) \cdot \sqrt{1 - {\left(\left(1 + maxCos \cdot ux\right) - ux\right)}^{2}}\right)} \]
          3. Applied rewrites35.0%

            \[\leadsto \color{blue}{\left(\sqrt{1 - \mathsf{fma}\left(maxCos, ux, 1 - ux\right) \cdot \mathsf{fma}\left(maxCos, ux, 1 - ux\right)} \cdot \pi\right) \cdot \left(uy + uy\right)} \]
          4. Taylor expanded in ux around 0

            \[\leadsto 2 \cdot \color{blue}{\left(\sqrt{ux \cdot \left(2 - 2 \cdot maxCos\right)} \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right)} \]
          5. Step-by-step derivation
            1. count-2-revN/A

              \[\leadsto \sqrt{ux \cdot \left(2 - 2 \cdot maxCos\right)} \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right) + \sqrt{ux \cdot \left(2 - 2 \cdot maxCos\right)} \cdot \color{blue}{\left(uy \cdot \mathsf{PI}\left(\right)\right)} \]
            2. distribute-lft-outN/A

              \[\leadsto \sqrt{ux \cdot \left(2 - 2 \cdot maxCos\right)} \cdot \left(uy \cdot \mathsf{PI}\left(\right) + \color{blue}{uy \cdot \mathsf{PI}\left(\right)}\right) \]
            3. count-2-revN/A

              \[\leadsto \sqrt{ux \cdot \left(2 - 2 \cdot maxCos\right)} \cdot \left(2 \cdot \left(uy \cdot \color{blue}{\mathsf{PI}\left(\right)}\right)\right) \]
            4. lower-*.f32N/A

              \[\leadsto \sqrt{ux \cdot \left(2 - 2 \cdot maxCos\right)} \cdot \left(2 \cdot \color{blue}{\left(uy \cdot \mathsf{PI}\left(\right)\right)}\right) \]
            5. lower-sqrt.f32N/A

              \[\leadsto \sqrt{ux \cdot \left(2 - 2 \cdot maxCos\right)} \cdot \left(2 \cdot \left(\color{blue}{uy} \cdot \mathsf{PI}\left(\right)\right)\right) \]
            6. *-commutativeN/A

              \[\leadsto \sqrt{\left(2 - 2 \cdot maxCos\right) \cdot ux} \cdot \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \]
            7. lower-*.f32N/A

              \[\leadsto \sqrt{\left(2 - 2 \cdot maxCos\right) \cdot ux} \cdot \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \]
            8. count-2-revN/A

              \[\leadsto \sqrt{\left(2 - \left(maxCos + maxCos\right)\right) \cdot ux} \cdot \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \]
            9. associate--r+N/A

              \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \]
            10. lower--.f32N/A

              \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \]
            11. lower--.f32N/A

              \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \]
            12. associate-*r*N/A

              \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(\left(2 \cdot uy\right) \cdot \mathsf{PI}\left(\right)\right) \]
            13. count-2-revN/A

              \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \mathsf{PI}\left(\right)\right) \]
            14. lift-+.f32N/A

              \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \mathsf{PI}\left(\right)\right) \]
            15. lower-*.f32N/A

              \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \mathsf{PI}\left(\right)\right) \]
            16. lift-PI.f3277.1

              \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
          6. Applied rewrites77.1%

            \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \color{blue}{\left(\left(uy + uy\right) \cdot \pi\right)} \]
          7. Step-by-step derivation
            1. lift-*.f32N/A

              \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
            2. lift--.f32N/A

              \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
            3. lift--.f32N/A

              \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
            4. associate--l-N/A

              \[\leadsto \sqrt{\left(2 - \left(maxCos + maxCos\right)\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
            5. count-2-revN/A

              \[\leadsto \sqrt{\left(2 - 2 \cdot maxCos\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
            6. fp-cancel-sub-sign-invN/A

              \[\leadsto \sqrt{\left(2 + \left(\mathsf{neg}\left(2\right)\right) \cdot maxCos\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
            7. metadata-evalN/A

              \[\leadsto \sqrt{\left(2 + -2 \cdot maxCos\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
            8. *-commutativeN/A

              \[\leadsto \sqrt{ux \cdot \left(2 + -2 \cdot maxCos\right)} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
            9. distribute-rgt-inN/A

              \[\leadsto \sqrt{2 \cdot ux + \left(-2 \cdot maxCos\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
            10. associate-*r*N/A

              \[\leadsto \sqrt{2 \cdot ux + -2 \cdot \left(maxCos \cdot ux\right)} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
            11. +-commutativeN/A

              \[\leadsto \sqrt{-2 \cdot \left(maxCos \cdot ux\right) + 2 \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
            12. *-commutativeN/A

              \[\leadsto \sqrt{\left(maxCos \cdot ux\right) \cdot -2 + 2 \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
            13. lower-fma.f32N/A

              \[\leadsto \sqrt{\mathsf{fma}\left(maxCos \cdot ux, -2, 2 \cdot ux\right)} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
            14. lower-*.f32N/A

              \[\leadsto \sqrt{\mathsf{fma}\left(maxCos \cdot ux, -2, 2 \cdot ux\right)} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
            15. count-2-revN/A

              \[\leadsto \sqrt{\mathsf{fma}\left(maxCos \cdot ux, -2, ux + ux\right)} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
            16. lower-+.f3277.1

              \[\leadsto \sqrt{\mathsf{fma}\left(maxCos \cdot ux, -2, ux + ux\right)} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
          8. Applied rewrites77.1%

            \[\leadsto \sqrt{\mathsf{fma}\left(maxCos \cdot ux, -2, ux + ux\right)} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
        3. Recombined 2 regimes into one program.
        4. Add Preprocessing

        Alternative 14: 75.4% accurate, 1.5× speedup?

        \[\begin{array}{l} \\ \begin{array}{l} t_0 := \left(1 - ux\right) + ux \cdot maxCos\\ \mathbf{if}\;t\_0 \cdot t\_0 \leq 0.9995800256729126:\\ \;\;\;\;\left(\sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \cdot \pi\right) \cdot \left(uy + uy\right)\\ \mathbf{else}:\\ \;\;\;\;\sqrt{\mathsf{fma}\left(maxCos, -2, 2\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right)\\ \end{array} \end{array} \]
        (FPCore (ux uy maxCos)
         :precision binary32
         (let* ((t_0 (+ (- 1.0 ux) (* ux maxCos))))
           (if (<= (* t_0 t_0) 0.9995800256729126)
             (* (* (sqrt (- 1.0 (* (- 1.0 ux) (- 1.0 ux)))) PI) (+ uy uy))
             (* (sqrt (* (fma maxCos -2.0 2.0) ux)) (* (+ uy uy) PI)))))
        float code(float ux, float uy, float maxCos) {
        	float t_0 = (1.0f - ux) + (ux * maxCos);
        	float tmp;
        	if ((t_0 * t_0) <= 0.9995800256729126f) {
        		tmp = (sqrtf((1.0f - ((1.0f - ux) * (1.0f - ux)))) * ((float) M_PI)) * (uy + uy);
        	} else {
        		tmp = sqrtf((fmaf(maxCos, -2.0f, 2.0f) * ux)) * ((uy + uy) * ((float) M_PI));
        	}
        	return tmp;
        }
        
        function code(ux, uy, maxCos)
        	t_0 = Float32(Float32(Float32(1.0) - ux) + Float32(ux * maxCos))
        	tmp = Float32(0.0)
        	if (Float32(t_0 * t_0) <= Float32(0.9995800256729126))
        		tmp = Float32(Float32(sqrt(Float32(Float32(1.0) - Float32(Float32(Float32(1.0) - ux) * Float32(Float32(1.0) - ux)))) * Float32(pi)) * Float32(uy + uy));
        	else
        		tmp = Float32(sqrt(Float32(fma(maxCos, Float32(-2.0), Float32(2.0)) * ux)) * Float32(Float32(uy + uy) * Float32(pi)));
        	end
        	return tmp
        end
        
        \begin{array}{l}
        
        \\
        \begin{array}{l}
        t_0 := \left(1 - ux\right) + ux \cdot maxCos\\
        \mathbf{if}\;t\_0 \cdot t\_0 \leq 0.9995800256729126:\\
        \;\;\;\;\left(\sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \cdot \pi\right) \cdot \left(uy + uy\right)\\
        
        \mathbf{else}:\\
        \;\;\;\;\sqrt{\mathsf{fma}\left(maxCos, -2, 2\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right)\\
        
        
        \end{array}
        \end{array}
        
        Derivation
        1. Split input into 2 regimes
        2. if (*.f32 (+.f32 (-.f32 #s(literal 1 binary32) ux) (*.f32 ux maxCos)) (+.f32 (-.f32 #s(literal 1 binary32) ux) (*.f32 ux maxCos))) < 0.999580026

          1. Initial program 89.6%

            \[\sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{1 - \left(\left(1 - ux\right) + ux \cdot maxCos\right) \cdot \left(\left(1 - ux\right) + ux \cdot maxCos\right)} \]
          2. Taylor expanded in uy around 0

            \[\leadsto \color{blue}{2 \cdot \left(\left(uy \cdot \mathsf{PI}\left(\right)\right) \cdot \sqrt{1 - {\left(\left(1 + maxCos \cdot ux\right) - ux\right)}^{2}}\right)} \]
          3. Applied rewrites76.0%

            \[\leadsto \color{blue}{\left(\sqrt{1 - \mathsf{fma}\left(maxCos, ux, 1 - ux\right) \cdot \mathsf{fma}\left(maxCos, ux, 1 - ux\right)} \cdot \pi\right) \cdot \left(uy + uy\right)} \]
          4. Taylor expanded in maxCos around 0

            \[\leadsto \left(\sqrt{1 - \left(1 - ux\right) \cdot \mathsf{fma}\left(maxCos, ux, 1 - ux\right)} \cdot \pi\right) \cdot \left(uy + uy\right) \]
          5. Step-by-step derivation
            1. lift--.f3273.0

              \[\leadsto \left(\sqrt{1 - \left(1 - ux\right) \cdot \mathsf{fma}\left(maxCos, ux, 1 - ux\right)} \cdot \pi\right) \cdot \left(uy + uy\right) \]
          6. Applied rewrites73.0%

            \[\leadsto \left(\sqrt{1 - \left(1 - ux\right) \cdot \mathsf{fma}\left(maxCos, ux, 1 - ux\right)} \cdot \pi\right) \cdot \left(uy + uy\right) \]
          7. Taylor expanded in maxCos around 0

            \[\leadsto \left(\sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \cdot \pi\right) \cdot \left(uy + uy\right) \]
          8. Step-by-step derivation
            1. lift--.f3272.7

              \[\leadsto \left(\sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \cdot \pi\right) \cdot \left(uy + uy\right) \]
          9. Applied rewrites72.7%

            \[\leadsto \left(\sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \cdot \pi\right) \cdot \left(uy + uy\right) \]

          if 0.999580026 < (*.f32 (+.f32 (-.f32 #s(literal 1 binary32) ux) (*.f32 ux maxCos)) (+.f32 (-.f32 #s(literal 1 binary32) ux) (*.f32 ux maxCos)))

          1. Initial program 37.6%

            \[\sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{1 - \left(\left(1 - ux\right) + ux \cdot maxCos\right) \cdot \left(\left(1 - ux\right) + ux \cdot maxCos\right)} \]
          2. Taylor expanded in uy around 0

            \[\leadsto \color{blue}{2 \cdot \left(\left(uy \cdot \mathsf{PI}\left(\right)\right) \cdot \sqrt{1 - {\left(\left(1 + maxCos \cdot ux\right) - ux\right)}^{2}}\right)} \]
          3. Applied rewrites35.0%

            \[\leadsto \color{blue}{\left(\sqrt{1 - \mathsf{fma}\left(maxCos, ux, 1 - ux\right) \cdot \mathsf{fma}\left(maxCos, ux, 1 - ux\right)} \cdot \pi\right) \cdot \left(uy + uy\right)} \]
          4. Taylor expanded in ux around 0

            \[\leadsto 2 \cdot \color{blue}{\left(\sqrt{ux \cdot \left(2 - 2 \cdot maxCos\right)} \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right)} \]
          5. Step-by-step derivation
            1. count-2-revN/A

              \[\leadsto \sqrt{ux \cdot \left(2 - 2 \cdot maxCos\right)} \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right) + \sqrt{ux \cdot \left(2 - 2 \cdot maxCos\right)} \cdot \color{blue}{\left(uy \cdot \mathsf{PI}\left(\right)\right)} \]
            2. distribute-lft-outN/A

              \[\leadsto \sqrt{ux \cdot \left(2 - 2 \cdot maxCos\right)} \cdot \left(uy \cdot \mathsf{PI}\left(\right) + \color{blue}{uy \cdot \mathsf{PI}\left(\right)}\right) \]
            3. count-2-revN/A

              \[\leadsto \sqrt{ux \cdot \left(2 - 2 \cdot maxCos\right)} \cdot \left(2 \cdot \left(uy \cdot \color{blue}{\mathsf{PI}\left(\right)}\right)\right) \]
            4. lower-*.f32N/A

              \[\leadsto \sqrt{ux \cdot \left(2 - 2 \cdot maxCos\right)} \cdot \left(2 \cdot \color{blue}{\left(uy \cdot \mathsf{PI}\left(\right)\right)}\right) \]
            5. lower-sqrt.f32N/A

              \[\leadsto \sqrt{ux \cdot \left(2 - 2 \cdot maxCos\right)} \cdot \left(2 \cdot \left(\color{blue}{uy} \cdot \mathsf{PI}\left(\right)\right)\right) \]
            6. *-commutativeN/A

              \[\leadsto \sqrt{\left(2 - 2 \cdot maxCos\right) \cdot ux} \cdot \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \]
            7. lower-*.f32N/A

              \[\leadsto \sqrt{\left(2 - 2 \cdot maxCos\right) \cdot ux} \cdot \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \]
            8. count-2-revN/A

              \[\leadsto \sqrt{\left(2 - \left(maxCos + maxCos\right)\right) \cdot ux} \cdot \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \]
            9. associate--r+N/A

              \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \]
            10. lower--.f32N/A

              \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \]
            11. lower--.f32N/A

              \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \]
            12. associate-*r*N/A

              \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(\left(2 \cdot uy\right) \cdot \mathsf{PI}\left(\right)\right) \]
            13. count-2-revN/A

              \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \mathsf{PI}\left(\right)\right) \]
            14. lift-+.f32N/A

              \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \mathsf{PI}\left(\right)\right) \]
            15. lower-*.f32N/A

              \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \mathsf{PI}\left(\right)\right) \]
            16. lift-PI.f3277.1

              \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
          6. Applied rewrites77.1%

            \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \color{blue}{\left(\left(uy + uy\right) \cdot \pi\right)} \]
          7. Step-by-step derivation
            1. lift--.f32N/A

              \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
            2. lift--.f32N/A

              \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
            3. associate--l-N/A

              \[\leadsto \sqrt{\left(2 - \left(maxCos + maxCos\right)\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
            4. count-2-revN/A

              \[\leadsto \sqrt{\left(2 - 2 \cdot maxCos\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
            5. fp-cancel-sub-sign-invN/A

              \[\leadsto \sqrt{\left(2 + \left(\mathsf{neg}\left(2\right)\right) \cdot maxCos\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
            6. metadata-evalN/A

              \[\leadsto \sqrt{\left(2 + -2 \cdot maxCos\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
            7. +-commutativeN/A

              \[\leadsto \sqrt{\left(-2 \cdot maxCos + 2\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
            8. *-commutativeN/A

              \[\leadsto \sqrt{\left(maxCos \cdot -2 + 2\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
            9. lower-fma.f3277.1

              \[\leadsto \sqrt{\mathsf{fma}\left(maxCos, -2, 2\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
          8. Applied rewrites77.1%

            \[\leadsto \sqrt{\mathsf{fma}\left(maxCos, -2, 2\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
        3. Recombined 2 regimes into one program.
        4. Add Preprocessing

        Alternative 15: 75.4% accurate, 1.5× speedup?

        \[\begin{array}{l} \\ \begin{array}{l} t_0 := \left(1 - ux\right) + ux \cdot maxCos\\ t_1 := \left(uy + uy\right) \cdot \pi\\ \mathbf{if}\;t\_0 \cdot t\_0 \leq 0.9995800256729126:\\ \;\;\;\;t\_1 \cdot \sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)}\\ \mathbf{else}:\\ \;\;\;\;\sqrt{\mathsf{fma}\left(maxCos, -2, 2\right) \cdot ux} \cdot t\_1\\ \end{array} \end{array} \]
        (FPCore (ux uy maxCos)
         :precision binary32
         (let* ((t_0 (+ (- 1.0 ux) (* ux maxCos))) (t_1 (* (+ uy uy) PI)))
           (if (<= (* t_0 t_0) 0.9995800256729126)
             (* t_1 (sqrt (- 1.0 (* (- 1.0 ux) (- 1.0 ux)))))
             (* (sqrt (* (fma maxCos -2.0 2.0) ux)) t_1))))
        float code(float ux, float uy, float maxCos) {
        	float t_0 = (1.0f - ux) + (ux * maxCos);
        	float t_1 = (uy + uy) * ((float) M_PI);
        	float tmp;
        	if ((t_0 * t_0) <= 0.9995800256729126f) {
        		tmp = t_1 * sqrtf((1.0f - ((1.0f - ux) * (1.0f - ux))));
        	} else {
        		tmp = sqrtf((fmaf(maxCos, -2.0f, 2.0f) * ux)) * t_1;
        	}
        	return tmp;
        }
        
        function code(ux, uy, maxCos)
        	t_0 = Float32(Float32(Float32(1.0) - ux) + Float32(ux * maxCos))
        	t_1 = Float32(Float32(uy + uy) * Float32(pi))
        	tmp = Float32(0.0)
        	if (Float32(t_0 * t_0) <= Float32(0.9995800256729126))
        		tmp = Float32(t_1 * sqrt(Float32(Float32(1.0) - Float32(Float32(Float32(1.0) - ux) * Float32(Float32(1.0) - ux)))));
        	else
        		tmp = Float32(sqrt(Float32(fma(maxCos, Float32(-2.0), Float32(2.0)) * ux)) * t_1);
        	end
        	return tmp
        end
        
        \begin{array}{l}
        
        \\
        \begin{array}{l}
        t_0 := \left(1 - ux\right) + ux \cdot maxCos\\
        t_1 := \left(uy + uy\right) \cdot \pi\\
        \mathbf{if}\;t\_0 \cdot t\_0 \leq 0.9995800256729126:\\
        \;\;\;\;t\_1 \cdot \sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)}\\
        
        \mathbf{else}:\\
        \;\;\;\;\sqrt{\mathsf{fma}\left(maxCos, -2, 2\right) \cdot ux} \cdot t\_1\\
        
        
        \end{array}
        \end{array}
        
        Derivation
        1. Split input into 2 regimes
        2. if (*.f32 (+.f32 (-.f32 #s(literal 1 binary32) ux) (*.f32 ux maxCos)) (+.f32 (-.f32 #s(literal 1 binary32) ux) (*.f32 ux maxCos))) < 0.999580026

          1. Initial program 89.6%

            \[\sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{1 - \left(\left(1 - ux\right) + ux \cdot maxCos\right) \cdot \left(\left(1 - ux\right) + ux \cdot maxCos\right)} \]
          2. Taylor expanded in uy around 0

            \[\leadsto \color{blue}{2 \cdot \left(\left(uy \cdot \mathsf{PI}\left(\right)\right) \cdot \sqrt{1 - {\left(\left(1 + maxCos \cdot ux\right) - ux\right)}^{2}}\right)} \]
          3. Applied rewrites76.0%

            \[\leadsto \color{blue}{\left(\sqrt{1 - \mathsf{fma}\left(maxCos, ux, 1 - ux\right) \cdot \mathsf{fma}\left(maxCos, ux, 1 - ux\right)} \cdot \pi\right) \cdot \left(uy + uy\right)} \]
          4. Taylor expanded in maxCos around 0

            \[\leadsto 2 \cdot \color{blue}{\left(\left(uy \cdot \mathsf{PI}\left(\right)\right) \cdot \sqrt{1 - {\left(1 - ux\right)}^{2}}\right)} \]
          5. Step-by-step derivation
            1. associate-*r*N/A

              \[\leadsto \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \cdot \sqrt{1 - {\left(1 - ux\right)}^{2}} \]
            2. lower-*.f32N/A

              \[\leadsto \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \cdot \sqrt{1 - {\left(1 - ux\right)}^{2}} \]
            3. associate-*r*N/A

              \[\leadsto \left(\left(2 \cdot uy\right) \cdot \mathsf{PI}\left(\right)\right) \cdot \sqrt{1 - {\left(1 - ux\right)}^{2}} \]
            4. count-2-revN/A

              \[\leadsto \left(\left(uy + uy\right) \cdot \mathsf{PI}\left(\right)\right) \cdot \sqrt{1 - {\left(1 - ux\right)}^{2}} \]
            5. lift-+.f32N/A

              \[\leadsto \left(\left(uy + uy\right) \cdot \mathsf{PI}\left(\right)\right) \cdot \sqrt{1 - {\left(1 - ux\right)}^{2}} \]
            6. lower-*.f32N/A

              \[\leadsto \left(\left(uy + uy\right) \cdot \mathsf{PI}\left(\right)\right) \cdot \sqrt{1 - {\left(1 - ux\right)}^{2}} \]
            7. lift-PI.f32N/A

              \[\leadsto \left(\left(uy + uy\right) \cdot \pi\right) \cdot \sqrt{1 - {\left(1 - ux\right)}^{2}} \]
            8. lower-sqrt.f32N/A

              \[\leadsto \left(\left(uy + uy\right) \cdot \pi\right) \cdot \sqrt{1 - {\left(1 - ux\right)}^{2}} \]
            9. pow2N/A

              \[\leadsto \left(\left(uy + uy\right) \cdot \pi\right) \cdot \sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \]
            10. lift--.f32N/A

              \[\leadsto \left(\left(uy + uy\right) \cdot \pi\right) \cdot \sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \]
            11. lift--.f32N/A

              \[\leadsto \left(\left(uy + uy\right) \cdot \pi\right) \cdot \sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \]
            12. lift-*.f32N/A

              \[\leadsto \left(\left(uy + uy\right) \cdot \pi\right) \cdot \sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \]
            13. lift--.f3272.7

              \[\leadsto \left(\left(uy + uy\right) \cdot \pi\right) \cdot \sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)} \]
          6. Applied rewrites72.7%

            \[\leadsto \left(\left(uy + uy\right) \cdot \pi\right) \cdot \color{blue}{\sqrt{1 - \left(1 - ux\right) \cdot \left(1 - ux\right)}} \]

          if 0.999580026 < (*.f32 (+.f32 (-.f32 #s(literal 1 binary32) ux) (*.f32 ux maxCos)) (+.f32 (-.f32 #s(literal 1 binary32) ux) (*.f32 ux maxCos)))

          1. Initial program 37.6%

            \[\sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{1 - \left(\left(1 - ux\right) + ux \cdot maxCos\right) \cdot \left(\left(1 - ux\right) + ux \cdot maxCos\right)} \]
          2. Taylor expanded in uy around 0

            \[\leadsto \color{blue}{2 \cdot \left(\left(uy \cdot \mathsf{PI}\left(\right)\right) \cdot \sqrt{1 - {\left(\left(1 + maxCos \cdot ux\right) - ux\right)}^{2}}\right)} \]
          3. Applied rewrites35.0%

            \[\leadsto \color{blue}{\left(\sqrt{1 - \mathsf{fma}\left(maxCos, ux, 1 - ux\right) \cdot \mathsf{fma}\left(maxCos, ux, 1 - ux\right)} \cdot \pi\right) \cdot \left(uy + uy\right)} \]
          4. Taylor expanded in ux around 0

            \[\leadsto 2 \cdot \color{blue}{\left(\sqrt{ux \cdot \left(2 - 2 \cdot maxCos\right)} \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right)} \]
          5. Step-by-step derivation
            1. count-2-revN/A

              \[\leadsto \sqrt{ux \cdot \left(2 - 2 \cdot maxCos\right)} \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right) + \sqrt{ux \cdot \left(2 - 2 \cdot maxCos\right)} \cdot \color{blue}{\left(uy \cdot \mathsf{PI}\left(\right)\right)} \]
            2. distribute-lft-outN/A

              \[\leadsto \sqrt{ux \cdot \left(2 - 2 \cdot maxCos\right)} \cdot \left(uy \cdot \mathsf{PI}\left(\right) + \color{blue}{uy \cdot \mathsf{PI}\left(\right)}\right) \]
            3. count-2-revN/A

              \[\leadsto \sqrt{ux \cdot \left(2 - 2 \cdot maxCos\right)} \cdot \left(2 \cdot \left(uy \cdot \color{blue}{\mathsf{PI}\left(\right)}\right)\right) \]
            4. lower-*.f32N/A

              \[\leadsto \sqrt{ux \cdot \left(2 - 2 \cdot maxCos\right)} \cdot \left(2 \cdot \color{blue}{\left(uy \cdot \mathsf{PI}\left(\right)\right)}\right) \]
            5. lower-sqrt.f32N/A

              \[\leadsto \sqrt{ux \cdot \left(2 - 2 \cdot maxCos\right)} \cdot \left(2 \cdot \left(\color{blue}{uy} \cdot \mathsf{PI}\left(\right)\right)\right) \]
            6. *-commutativeN/A

              \[\leadsto \sqrt{\left(2 - 2 \cdot maxCos\right) \cdot ux} \cdot \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \]
            7. lower-*.f32N/A

              \[\leadsto \sqrt{\left(2 - 2 \cdot maxCos\right) \cdot ux} \cdot \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \]
            8. count-2-revN/A

              \[\leadsto \sqrt{\left(2 - \left(maxCos + maxCos\right)\right) \cdot ux} \cdot \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \]
            9. associate--r+N/A

              \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \]
            10. lower--.f32N/A

              \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \]
            11. lower--.f32N/A

              \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \]
            12. associate-*r*N/A

              \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(\left(2 \cdot uy\right) \cdot \mathsf{PI}\left(\right)\right) \]
            13. count-2-revN/A

              \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \mathsf{PI}\left(\right)\right) \]
            14. lift-+.f32N/A

              \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \mathsf{PI}\left(\right)\right) \]
            15. lower-*.f32N/A

              \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \mathsf{PI}\left(\right)\right) \]
            16. lift-PI.f3277.1

              \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
          6. Applied rewrites77.1%

            \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \color{blue}{\left(\left(uy + uy\right) \cdot \pi\right)} \]
          7. Step-by-step derivation
            1. lift--.f32N/A

              \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
            2. lift--.f32N/A

              \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
            3. associate--l-N/A

              \[\leadsto \sqrt{\left(2 - \left(maxCos + maxCos\right)\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
            4. count-2-revN/A

              \[\leadsto \sqrt{\left(2 - 2 \cdot maxCos\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
            5. fp-cancel-sub-sign-invN/A

              \[\leadsto \sqrt{\left(2 + \left(\mathsf{neg}\left(2\right)\right) \cdot maxCos\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
            6. metadata-evalN/A

              \[\leadsto \sqrt{\left(2 + -2 \cdot maxCos\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
            7. +-commutativeN/A

              \[\leadsto \sqrt{\left(-2 \cdot maxCos + 2\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
            8. *-commutativeN/A

              \[\leadsto \sqrt{\left(maxCos \cdot -2 + 2\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
            9. lower-fma.f3277.1

              \[\leadsto \sqrt{\mathsf{fma}\left(maxCos, -2, 2\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
          8. Applied rewrites77.1%

            \[\leadsto \sqrt{\mathsf{fma}\left(maxCos, -2, 2\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
        3. Recombined 2 regimes into one program.
        4. Add Preprocessing

        Alternative 16: 65.9% accurate, 3.4× speedup?

        \[\begin{array}{l} \\ \sqrt{\mathsf{fma}\left(maxCos, -2, 2\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \end{array} \]
        (FPCore (ux uy maxCos)
         :precision binary32
         (* (sqrt (* (fma maxCos -2.0 2.0) ux)) (* (+ uy uy) PI)))
        float code(float ux, float uy, float maxCos) {
        	return sqrtf((fmaf(maxCos, -2.0f, 2.0f) * ux)) * ((uy + uy) * ((float) M_PI));
        }
        
        function code(ux, uy, maxCos)
        	return Float32(sqrt(Float32(fma(maxCos, Float32(-2.0), Float32(2.0)) * ux)) * Float32(Float32(uy + uy) * Float32(pi)))
        end
        
        \begin{array}{l}
        
        \\
        \sqrt{\mathsf{fma}\left(maxCos, -2, 2\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right)
        \end{array}
        
        Derivation
        1. Initial program 57.6%

          \[\sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{1 - \left(\left(1 - ux\right) + ux \cdot maxCos\right) \cdot \left(\left(1 - ux\right) + ux \cdot maxCos\right)} \]
        2. Taylor expanded in uy around 0

          \[\leadsto \color{blue}{2 \cdot \left(\left(uy \cdot \mathsf{PI}\left(\right)\right) \cdot \sqrt{1 - {\left(\left(1 + maxCos \cdot ux\right) - ux\right)}^{2}}\right)} \]
        3. Applied rewrites50.7%

          \[\leadsto \color{blue}{\left(\sqrt{1 - \mathsf{fma}\left(maxCos, ux, 1 - ux\right) \cdot \mathsf{fma}\left(maxCos, ux, 1 - ux\right)} \cdot \pi\right) \cdot \left(uy + uy\right)} \]
        4. Taylor expanded in ux around 0

          \[\leadsto 2 \cdot \color{blue}{\left(\sqrt{ux \cdot \left(2 - 2 \cdot maxCos\right)} \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right)} \]
        5. Step-by-step derivation
          1. count-2-revN/A

            \[\leadsto \sqrt{ux \cdot \left(2 - 2 \cdot maxCos\right)} \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right) + \sqrt{ux \cdot \left(2 - 2 \cdot maxCos\right)} \cdot \color{blue}{\left(uy \cdot \mathsf{PI}\left(\right)\right)} \]
          2. distribute-lft-outN/A

            \[\leadsto \sqrt{ux \cdot \left(2 - 2 \cdot maxCos\right)} \cdot \left(uy \cdot \mathsf{PI}\left(\right) + \color{blue}{uy \cdot \mathsf{PI}\left(\right)}\right) \]
          3. count-2-revN/A

            \[\leadsto \sqrt{ux \cdot \left(2 - 2 \cdot maxCos\right)} \cdot \left(2 \cdot \left(uy \cdot \color{blue}{\mathsf{PI}\left(\right)}\right)\right) \]
          4. lower-*.f32N/A

            \[\leadsto \sqrt{ux \cdot \left(2 - 2 \cdot maxCos\right)} \cdot \left(2 \cdot \color{blue}{\left(uy \cdot \mathsf{PI}\left(\right)\right)}\right) \]
          5. lower-sqrt.f32N/A

            \[\leadsto \sqrt{ux \cdot \left(2 - 2 \cdot maxCos\right)} \cdot \left(2 \cdot \left(\color{blue}{uy} \cdot \mathsf{PI}\left(\right)\right)\right) \]
          6. *-commutativeN/A

            \[\leadsto \sqrt{\left(2 - 2 \cdot maxCos\right) \cdot ux} \cdot \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \]
          7. lower-*.f32N/A

            \[\leadsto \sqrt{\left(2 - 2 \cdot maxCos\right) \cdot ux} \cdot \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \]
          8. count-2-revN/A

            \[\leadsto \sqrt{\left(2 - \left(maxCos + maxCos\right)\right) \cdot ux} \cdot \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \]
          9. associate--r+N/A

            \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \]
          10. lower--.f32N/A

            \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \]
          11. lower--.f32N/A

            \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \]
          12. associate-*r*N/A

            \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(\left(2 \cdot uy\right) \cdot \mathsf{PI}\left(\right)\right) \]
          13. count-2-revN/A

            \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \mathsf{PI}\left(\right)\right) \]
          14. lift-+.f32N/A

            \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \mathsf{PI}\left(\right)\right) \]
          15. lower-*.f32N/A

            \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \mathsf{PI}\left(\right)\right) \]
          16. lift-PI.f3265.9

            \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
        6. Applied rewrites65.9%

          \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \color{blue}{\left(\left(uy + uy\right) \cdot \pi\right)} \]
        7. Step-by-step derivation
          1. lift--.f32N/A

            \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
          2. lift--.f32N/A

            \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
          3. associate--l-N/A

            \[\leadsto \sqrt{\left(2 - \left(maxCos + maxCos\right)\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
          4. count-2-revN/A

            \[\leadsto \sqrt{\left(2 - 2 \cdot maxCos\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
          5. fp-cancel-sub-sign-invN/A

            \[\leadsto \sqrt{\left(2 + \left(\mathsf{neg}\left(2\right)\right) \cdot maxCos\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
          6. metadata-evalN/A

            \[\leadsto \sqrt{\left(2 + -2 \cdot maxCos\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
          7. +-commutativeN/A

            \[\leadsto \sqrt{\left(-2 \cdot maxCos + 2\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
          8. *-commutativeN/A

            \[\leadsto \sqrt{\left(maxCos \cdot -2 + 2\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
          9. lower-fma.f3265.9

            \[\leadsto \sqrt{\mathsf{fma}\left(maxCos, -2, 2\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
        8. Applied rewrites65.9%

          \[\leadsto \sqrt{\mathsf{fma}\left(maxCos, -2, 2\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
        9. Add Preprocessing

        Alternative 17: 65.9% accurate, 3.4× speedup?

        \[\begin{array}{l} \\ \left(\sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(uy + uy\right)\right) \cdot \pi \end{array} \]
        (FPCore (ux uy maxCos)
         :precision binary32
         (* (* (sqrt (* (- (- 2.0 maxCos) maxCos) ux)) (+ uy uy)) PI))
        float code(float ux, float uy, float maxCos) {
        	return (sqrtf((((2.0f - maxCos) - maxCos) * ux)) * (uy + uy)) * ((float) M_PI);
        }
        
        function code(ux, uy, maxCos)
        	return Float32(Float32(sqrt(Float32(Float32(Float32(Float32(2.0) - maxCos) - maxCos) * ux)) * Float32(uy + uy)) * Float32(pi))
        end
        
        function tmp = code(ux, uy, maxCos)
        	tmp = (sqrt((((single(2.0) - maxCos) - maxCos) * ux)) * (uy + uy)) * single(pi);
        end
        
        \begin{array}{l}
        
        \\
        \left(\sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(uy + uy\right)\right) \cdot \pi
        \end{array}
        
        Derivation
        1. Initial program 57.6%

          \[\sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{1 - \left(\left(1 - ux\right) + ux \cdot maxCos\right) \cdot \left(\left(1 - ux\right) + ux \cdot maxCos\right)} \]
        2. Taylor expanded in uy around 0

          \[\leadsto \color{blue}{2 \cdot \left(\left(uy \cdot \mathsf{PI}\left(\right)\right) \cdot \sqrt{1 - {\left(\left(1 + maxCos \cdot ux\right) - ux\right)}^{2}}\right)} \]
        3. Applied rewrites50.7%

          \[\leadsto \color{blue}{\left(\sqrt{1 - \mathsf{fma}\left(maxCos, ux, 1 - ux\right) \cdot \mathsf{fma}\left(maxCos, ux, 1 - ux\right)} \cdot \pi\right) \cdot \left(uy + uy\right)} \]
        4. Taylor expanded in ux around 0

          \[\leadsto 2 \cdot \color{blue}{\left(\sqrt{ux \cdot \left(2 - 2 \cdot maxCos\right)} \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right)} \]
        5. Step-by-step derivation
          1. count-2-revN/A

            \[\leadsto \sqrt{ux \cdot \left(2 - 2 \cdot maxCos\right)} \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right) + \sqrt{ux \cdot \left(2 - 2 \cdot maxCos\right)} \cdot \color{blue}{\left(uy \cdot \mathsf{PI}\left(\right)\right)} \]
          2. distribute-lft-outN/A

            \[\leadsto \sqrt{ux \cdot \left(2 - 2 \cdot maxCos\right)} \cdot \left(uy \cdot \mathsf{PI}\left(\right) + \color{blue}{uy \cdot \mathsf{PI}\left(\right)}\right) \]
          3. count-2-revN/A

            \[\leadsto \sqrt{ux \cdot \left(2 - 2 \cdot maxCos\right)} \cdot \left(2 \cdot \left(uy \cdot \color{blue}{\mathsf{PI}\left(\right)}\right)\right) \]
          4. lower-*.f32N/A

            \[\leadsto \sqrt{ux \cdot \left(2 - 2 \cdot maxCos\right)} \cdot \left(2 \cdot \color{blue}{\left(uy \cdot \mathsf{PI}\left(\right)\right)}\right) \]
          5. lower-sqrt.f32N/A

            \[\leadsto \sqrt{ux \cdot \left(2 - 2 \cdot maxCos\right)} \cdot \left(2 \cdot \left(\color{blue}{uy} \cdot \mathsf{PI}\left(\right)\right)\right) \]
          6. *-commutativeN/A

            \[\leadsto \sqrt{\left(2 - 2 \cdot maxCos\right) \cdot ux} \cdot \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \]
          7. lower-*.f32N/A

            \[\leadsto \sqrt{\left(2 - 2 \cdot maxCos\right) \cdot ux} \cdot \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \]
          8. count-2-revN/A

            \[\leadsto \sqrt{\left(2 - \left(maxCos + maxCos\right)\right) \cdot ux} \cdot \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \]
          9. associate--r+N/A

            \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \]
          10. lower--.f32N/A

            \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \]
          11. lower--.f32N/A

            \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \]
          12. associate-*r*N/A

            \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(\left(2 \cdot uy\right) \cdot \mathsf{PI}\left(\right)\right) \]
          13. count-2-revN/A

            \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \mathsf{PI}\left(\right)\right) \]
          14. lift-+.f32N/A

            \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \mathsf{PI}\left(\right)\right) \]
          15. lower-*.f32N/A

            \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \mathsf{PI}\left(\right)\right) \]
          16. lift-PI.f3265.9

            \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
        6. Applied rewrites65.9%

          \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \color{blue}{\left(\left(uy + uy\right) \cdot \pi\right)} \]
        7. Applied rewrites65.9%

          \[\leadsto \color{blue}{\left(\sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(uy + uy\right)\right) \cdot \pi} \]
        8. Add Preprocessing

        Alternative 18: 63.6% accurate, 3.9× speedup?

        \[\begin{array}{l} \\ \sqrt{\left(2 - maxCos\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \end{array} \]
        (FPCore (ux uy maxCos)
         :precision binary32
         (* (sqrt (* (- 2.0 maxCos) ux)) (* (+ uy uy) PI)))
        float code(float ux, float uy, float maxCos) {
        	return sqrtf(((2.0f - maxCos) * ux)) * ((uy + uy) * ((float) M_PI));
        }
        
        function code(ux, uy, maxCos)
        	return Float32(sqrt(Float32(Float32(Float32(2.0) - maxCos) * ux)) * Float32(Float32(uy + uy) * Float32(pi)))
        end
        
        function tmp = code(ux, uy, maxCos)
        	tmp = sqrt(((single(2.0) - maxCos) * ux)) * ((uy + uy) * single(pi));
        end
        
        \begin{array}{l}
        
        \\
        \sqrt{\left(2 - maxCos\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right)
        \end{array}
        
        Derivation
        1. Initial program 57.6%

          \[\sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{1 - \left(\left(1 - ux\right) + ux \cdot maxCos\right) \cdot \left(\left(1 - ux\right) + ux \cdot maxCos\right)} \]
        2. Taylor expanded in uy around 0

          \[\leadsto \color{blue}{2 \cdot \left(\left(uy \cdot \mathsf{PI}\left(\right)\right) \cdot \sqrt{1 - {\left(\left(1 + maxCos \cdot ux\right) - ux\right)}^{2}}\right)} \]
        3. Applied rewrites50.7%

          \[\leadsto \color{blue}{\left(\sqrt{1 - \mathsf{fma}\left(maxCos, ux, 1 - ux\right) \cdot \mathsf{fma}\left(maxCos, ux, 1 - ux\right)} \cdot \pi\right) \cdot \left(uy + uy\right)} \]
        4. Taylor expanded in ux around 0

          \[\leadsto 2 \cdot \color{blue}{\left(\sqrt{ux \cdot \left(2 - 2 \cdot maxCos\right)} \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right)} \]
        5. Step-by-step derivation
          1. count-2-revN/A

            \[\leadsto \sqrt{ux \cdot \left(2 - 2 \cdot maxCos\right)} \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right) + \sqrt{ux \cdot \left(2 - 2 \cdot maxCos\right)} \cdot \color{blue}{\left(uy \cdot \mathsf{PI}\left(\right)\right)} \]
          2. distribute-lft-outN/A

            \[\leadsto \sqrt{ux \cdot \left(2 - 2 \cdot maxCos\right)} \cdot \left(uy \cdot \mathsf{PI}\left(\right) + \color{blue}{uy \cdot \mathsf{PI}\left(\right)}\right) \]
          3. count-2-revN/A

            \[\leadsto \sqrt{ux \cdot \left(2 - 2 \cdot maxCos\right)} \cdot \left(2 \cdot \left(uy \cdot \color{blue}{\mathsf{PI}\left(\right)}\right)\right) \]
          4. lower-*.f32N/A

            \[\leadsto \sqrt{ux \cdot \left(2 - 2 \cdot maxCos\right)} \cdot \left(2 \cdot \color{blue}{\left(uy \cdot \mathsf{PI}\left(\right)\right)}\right) \]
          5. lower-sqrt.f32N/A

            \[\leadsto \sqrt{ux \cdot \left(2 - 2 \cdot maxCos\right)} \cdot \left(2 \cdot \left(\color{blue}{uy} \cdot \mathsf{PI}\left(\right)\right)\right) \]
          6. *-commutativeN/A

            \[\leadsto \sqrt{\left(2 - 2 \cdot maxCos\right) \cdot ux} \cdot \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \]
          7. lower-*.f32N/A

            \[\leadsto \sqrt{\left(2 - 2 \cdot maxCos\right) \cdot ux} \cdot \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \]
          8. count-2-revN/A

            \[\leadsto \sqrt{\left(2 - \left(maxCos + maxCos\right)\right) \cdot ux} \cdot \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \]
          9. associate--r+N/A

            \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \]
          10. lower--.f32N/A

            \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \]
          11. lower--.f32N/A

            \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \]
          12. associate-*r*N/A

            \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(\left(2 \cdot uy\right) \cdot \mathsf{PI}\left(\right)\right) \]
          13. count-2-revN/A

            \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \mathsf{PI}\left(\right)\right) \]
          14. lift-+.f32N/A

            \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \mathsf{PI}\left(\right)\right) \]
          15. lower-*.f32N/A

            \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \mathsf{PI}\left(\right)\right) \]
          16. lift-PI.f3265.9

            \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
        6. Applied rewrites65.9%

          \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \color{blue}{\left(\left(uy + uy\right) \cdot \pi\right)} \]
        7. Taylor expanded in maxCos around 0

          \[\leadsto \sqrt{\left(2 - maxCos\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
        8. Step-by-step derivation
          1. Applied rewrites63.6%

            \[\leadsto \sqrt{\left(2 - maxCos\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
          2. Add Preprocessing

          Alternative 19: 63.4% accurate, 3.9× speedup?

          \[\begin{array}{l} \\ \left(\left(\sqrt{ux} \cdot 2\right) \cdot uy\right) \cdot \left(\sqrt{2} \cdot \pi\right) \end{array} \]
          (FPCore (ux uy maxCos)
           :precision binary32
           (* (* (* (sqrt ux) 2.0) uy) (* (sqrt 2.0) PI)))
          float code(float ux, float uy, float maxCos) {
          	return ((sqrtf(ux) * 2.0f) * uy) * (sqrtf(2.0f) * ((float) M_PI));
          }
          
          function code(ux, uy, maxCos)
          	return Float32(Float32(Float32(sqrt(ux) * Float32(2.0)) * uy) * Float32(sqrt(Float32(2.0)) * Float32(pi)))
          end
          
          function tmp = code(ux, uy, maxCos)
          	tmp = ((sqrt(ux) * single(2.0)) * uy) * (sqrt(single(2.0)) * single(pi));
          end
          
          \begin{array}{l}
          
          \\
          \left(\left(\sqrt{ux} \cdot 2\right) \cdot uy\right) \cdot \left(\sqrt{2} \cdot \pi\right)
          \end{array}
          
          Derivation
          1. Initial program 57.6%

            \[\sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{1 - \left(\left(1 - ux\right) + ux \cdot maxCos\right) \cdot \left(\left(1 - ux\right) + ux \cdot maxCos\right)} \]
          2. Taylor expanded in uy around 0

            \[\leadsto \color{blue}{2 \cdot \left(\left(uy \cdot \mathsf{PI}\left(\right)\right) \cdot \sqrt{1 - {\left(\left(1 + maxCos \cdot ux\right) - ux\right)}^{2}}\right)} \]
          3. Applied rewrites50.7%

            \[\leadsto \color{blue}{\left(\sqrt{1 - \mathsf{fma}\left(maxCos, ux, 1 - ux\right) \cdot \mathsf{fma}\left(maxCos, ux, 1 - ux\right)} \cdot \pi\right) \cdot \left(uy + uy\right)} \]
          4. Taylor expanded in ux around 0

            \[\leadsto -1 \cdot \left(\sqrt{\frac{{ux}^{3}}{2 - 2 \cdot maxCos}} \cdot \left(uy \cdot \left(\mathsf{PI}\left(\right) \cdot {\left(maxCos - 1\right)}^{2}\right)\right)\right) + \color{blue}{2 \cdot \left(\sqrt{ux \cdot \left(2 - 2 \cdot maxCos\right)} \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right)} \]
          5. Applied rewrites75.7%

            \[\leadsto \mathsf{fma}\left(-\left(\left(\left(maxCos - 1\right) \cdot \left(maxCos - 1\right)\right) \cdot \pi\right) \cdot uy, \color{blue}{\sqrt{\frac{\left(ux \cdot ux\right) \cdot ux}{\left(2 - maxCos\right) - maxCos}}}, \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right)\right) \]
          6. Taylor expanded in maxCos around 0

            \[\leadsto -1 \cdot \left(\sqrt{{ux}^{3}} \cdot \left(uy \cdot \left(\mathsf{PI}\left(\right) \cdot \sqrt{\frac{1}{2}}\right)\right)\right) + 2 \cdot \color{blue}{\left(\sqrt{ux} \cdot \left(uy \cdot \left(\mathsf{PI}\left(\right) \cdot \sqrt{2}\right)\right)\right)} \]
          7. Step-by-step derivation
            1. +-commutativeN/A

              \[\leadsto 2 \cdot \left(\sqrt{ux} \cdot \left(uy \cdot \left(\mathsf{PI}\left(\right) \cdot \sqrt{2}\right)\right)\right) + -1 \cdot \left(\sqrt{{ux}^{3}} \cdot \color{blue}{\left(uy \cdot \left(\mathsf{PI}\left(\right) \cdot \sqrt{\frac{1}{2}}\right)\right)}\right) \]
            2. associate-*r*N/A

              \[\leadsto \left(2 \cdot \sqrt{ux}\right) \cdot \left(uy \cdot \left(\mathsf{PI}\left(\right) \cdot \sqrt{2}\right)\right) + -1 \cdot \left(\sqrt{{ux}^{3}} \cdot \left(\color{blue}{uy} \cdot \left(\mathsf{PI}\left(\right) \cdot \sqrt{\frac{1}{2}}\right)\right)\right) \]
            3. lower-fma.f32N/A

              \[\leadsto \mathsf{fma}\left(2 \cdot \sqrt{ux}, uy \cdot \left(\mathsf{PI}\left(\right) \cdot \color{blue}{\sqrt{2}}\right), -1 \cdot \left(\sqrt{{ux}^{3}} \cdot \left(uy \cdot \left(\mathsf{PI}\left(\right) \cdot \sqrt{\frac{1}{2}}\right)\right)\right)\right) \]
            4. lower-*.f32N/A

              \[\leadsto \mathsf{fma}\left(2 \cdot \sqrt{ux}, uy \cdot \left(\mathsf{PI}\left(\right) \cdot \sqrt{\color{blue}{2}}\right), -1 \cdot \left(\sqrt{{ux}^{3}} \cdot \left(uy \cdot \left(\mathsf{PI}\left(\right) \cdot \sqrt{\frac{1}{2}}\right)\right)\right)\right) \]
            5. lower-sqrt.f32N/A

              \[\leadsto \mathsf{fma}\left(2 \cdot \sqrt{ux}, uy \cdot \left(\mathsf{PI}\left(\right) \cdot \sqrt{2}\right), -1 \cdot \left(\sqrt{{ux}^{3}} \cdot \left(uy \cdot \left(\mathsf{PI}\left(\right) \cdot \sqrt{\frac{1}{2}}\right)\right)\right)\right) \]
            6. lower-*.f32N/A

              \[\leadsto \mathsf{fma}\left(2 \cdot \sqrt{ux}, uy \cdot \left(\mathsf{PI}\left(\right) \cdot \sqrt{2}\right), -1 \cdot \left(\sqrt{{ux}^{3}} \cdot \left(uy \cdot \left(\mathsf{PI}\left(\right) \cdot \sqrt{\frac{1}{2}}\right)\right)\right)\right) \]
            7. lower-*.f32N/A

              \[\leadsto \mathsf{fma}\left(2 \cdot \sqrt{ux}, uy \cdot \left(\mathsf{PI}\left(\right) \cdot \sqrt{2}\right), -1 \cdot \left(\sqrt{{ux}^{3}} \cdot \left(uy \cdot \left(\mathsf{PI}\left(\right) \cdot \sqrt{\frac{1}{2}}\right)\right)\right)\right) \]
            8. lift-PI.f32N/A

              \[\leadsto \mathsf{fma}\left(2 \cdot \sqrt{ux}, uy \cdot \left(\pi \cdot \sqrt{2}\right), -1 \cdot \left(\sqrt{{ux}^{3}} \cdot \left(uy \cdot \left(\mathsf{PI}\left(\right) \cdot \sqrt{\frac{1}{2}}\right)\right)\right)\right) \]
            9. lower-sqrt.f32N/A

              \[\leadsto \mathsf{fma}\left(2 \cdot \sqrt{ux}, uy \cdot \left(\pi \cdot \sqrt{2}\right), -1 \cdot \left(\sqrt{{ux}^{3}} \cdot \left(uy \cdot \left(\mathsf{PI}\left(\right) \cdot \sqrt{\frac{1}{2}}\right)\right)\right)\right) \]
          8. Applied rewrites72.2%

            \[\leadsto \mathsf{fma}\left(2 \cdot \sqrt{ux}, uy \cdot \color{blue}{\left(\pi \cdot \sqrt{2}\right)}, -\sqrt{\left(ux \cdot ux\right) \cdot ux} \cdot \left(uy \cdot \left(\pi \cdot \sqrt{0.5}\right)\right)\right) \]
          9. Taylor expanded in ux around 0

            \[\leadsto 2 \cdot \left(\sqrt{ux} \cdot \left(uy \cdot \color{blue}{\left(\mathsf{PI}\left(\right) \cdot \sqrt{2}\right)}\right)\right) \]
          10. Step-by-step derivation
            1. associate-*l*N/A

              \[\leadsto \left(2 \cdot \sqrt{ux}\right) \cdot \left(uy \cdot \left(\mathsf{PI}\left(\right) \cdot \sqrt{2}\right)\right) \]
            2. associate-*r*N/A

              \[\leadsto \left(\left(2 \cdot \sqrt{ux}\right) \cdot uy\right) \cdot \left(\mathsf{PI}\left(\right) \cdot \sqrt{2}\right) \]
            3. lower-*.f32N/A

              \[\leadsto \left(\left(2 \cdot \sqrt{ux}\right) \cdot uy\right) \cdot \left(\mathsf{PI}\left(\right) \cdot \sqrt{2}\right) \]
            4. lower-*.f32N/A

              \[\leadsto \left(\left(2 \cdot \sqrt{ux}\right) \cdot uy\right) \cdot \left(\mathsf{PI}\left(\right) \cdot \sqrt{2}\right) \]
            5. *-commutativeN/A

              \[\leadsto \left(\left(\sqrt{ux} \cdot 2\right) \cdot uy\right) \cdot \left(\mathsf{PI}\left(\right) \cdot \sqrt{2}\right) \]
            6. lower-*.f32N/A

              \[\leadsto \left(\left(\sqrt{ux} \cdot 2\right) \cdot uy\right) \cdot \left(\mathsf{PI}\left(\right) \cdot \sqrt{2}\right) \]
            7. lift-sqrt.f32N/A

              \[\leadsto \left(\left(\sqrt{ux} \cdot 2\right) \cdot uy\right) \cdot \left(\mathsf{PI}\left(\right) \cdot \sqrt{2}\right) \]
            8. *-commutativeN/A

              \[\leadsto \left(\left(\sqrt{ux} \cdot 2\right) \cdot uy\right) \cdot \left(\sqrt{2} \cdot \mathsf{PI}\left(\right)\right) \]
            9. lower-*.f32N/A

              \[\leadsto \left(\left(\sqrt{ux} \cdot 2\right) \cdot uy\right) \cdot \left(\sqrt{2} \cdot \mathsf{PI}\left(\right)\right) \]
            10. lift-sqrt.f32N/A

              \[\leadsto \left(\left(\sqrt{ux} \cdot 2\right) \cdot uy\right) \cdot \left(\sqrt{2} \cdot \mathsf{PI}\left(\right)\right) \]
            11. lift-PI.f3263.4

              \[\leadsto \left(\left(\sqrt{ux} \cdot 2\right) \cdot uy\right) \cdot \left(\sqrt{2} \cdot \pi\right) \]
          11. Applied rewrites63.4%

            \[\leadsto \left(\left(\sqrt{ux} \cdot 2\right) \cdot uy\right) \cdot \left(\sqrt{2} \cdot \pi\right) \]
          12. Add Preprocessing

          Alternative 20: 63.4% accurate, 4.7× speedup?

          \[\begin{array}{l} \\ \sqrt{ux + ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \end{array} \]
          (FPCore (ux uy maxCos)
           :precision binary32
           (* (sqrt (+ ux ux)) (* (+ uy uy) PI)))
          float code(float ux, float uy, float maxCos) {
          	return sqrtf((ux + ux)) * ((uy + uy) * ((float) M_PI));
          }
          
          function code(ux, uy, maxCos)
          	return Float32(sqrt(Float32(ux + ux)) * Float32(Float32(uy + uy) * Float32(pi)))
          end
          
          function tmp = code(ux, uy, maxCos)
          	tmp = sqrt((ux + ux)) * ((uy + uy) * single(pi));
          end
          
          \begin{array}{l}
          
          \\
          \sqrt{ux + ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right)
          \end{array}
          
          Derivation
          1. Initial program 57.6%

            \[\sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{1 - \left(\left(1 - ux\right) + ux \cdot maxCos\right) \cdot \left(\left(1 - ux\right) + ux \cdot maxCos\right)} \]
          2. Taylor expanded in uy around 0

            \[\leadsto \color{blue}{2 \cdot \left(\left(uy \cdot \mathsf{PI}\left(\right)\right) \cdot \sqrt{1 - {\left(\left(1 + maxCos \cdot ux\right) - ux\right)}^{2}}\right)} \]
          3. Applied rewrites50.7%

            \[\leadsto \color{blue}{\left(\sqrt{1 - \mathsf{fma}\left(maxCos, ux, 1 - ux\right) \cdot \mathsf{fma}\left(maxCos, ux, 1 - ux\right)} \cdot \pi\right) \cdot \left(uy + uy\right)} \]
          4. Taylor expanded in ux around 0

            \[\leadsto 2 \cdot \color{blue}{\left(\sqrt{ux \cdot \left(2 - 2 \cdot maxCos\right)} \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right)} \]
          5. Step-by-step derivation
            1. count-2-revN/A

              \[\leadsto \sqrt{ux \cdot \left(2 - 2 \cdot maxCos\right)} \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right) + \sqrt{ux \cdot \left(2 - 2 \cdot maxCos\right)} \cdot \color{blue}{\left(uy \cdot \mathsf{PI}\left(\right)\right)} \]
            2. distribute-lft-outN/A

              \[\leadsto \sqrt{ux \cdot \left(2 - 2 \cdot maxCos\right)} \cdot \left(uy \cdot \mathsf{PI}\left(\right) + \color{blue}{uy \cdot \mathsf{PI}\left(\right)}\right) \]
            3. count-2-revN/A

              \[\leadsto \sqrt{ux \cdot \left(2 - 2 \cdot maxCos\right)} \cdot \left(2 \cdot \left(uy \cdot \color{blue}{\mathsf{PI}\left(\right)}\right)\right) \]
            4. lower-*.f32N/A

              \[\leadsto \sqrt{ux \cdot \left(2 - 2 \cdot maxCos\right)} \cdot \left(2 \cdot \color{blue}{\left(uy \cdot \mathsf{PI}\left(\right)\right)}\right) \]
            5. lower-sqrt.f32N/A

              \[\leadsto \sqrt{ux \cdot \left(2 - 2 \cdot maxCos\right)} \cdot \left(2 \cdot \left(\color{blue}{uy} \cdot \mathsf{PI}\left(\right)\right)\right) \]
            6. *-commutativeN/A

              \[\leadsto \sqrt{\left(2 - 2 \cdot maxCos\right) \cdot ux} \cdot \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \]
            7. lower-*.f32N/A

              \[\leadsto \sqrt{\left(2 - 2 \cdot maxCos\right) \cdot ux} \cdot \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \]
            8. count-2-revN/A

              \[\leadsto \sqrt{\left(2 - \left(maxCos + maxCos\right)\right) \cdot ux} \cdot \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \]
            9. associate--r+N/A

              \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \]
            10. lower--.f32N/A

              \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \]
            11. lower--.f32N/A

              \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(2 \cdot \left(uy \cdot \mathsf{PI}\left(\right)\right)\right) \]
            12. associate-*r*N/A

              \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(\left(2 \cdot uy\right) \cdot \mathsf{PI}\left(\right)\right) \]
            13. count-2-revN/A

              \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \mathsf{PI}\left(\right)\right) \]
            14. lift-+.f32N/A

              \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \mathsf{PI}\left(\right)\right) \]
            15. lower-*.f32N/A

              \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \mathsf{PI}\left(\right)\right) \]
            16. lift-PI.f3265.9

              \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
          6. Applied rewrites65.9%

            \[\leadsto \sqrt{\left(\left(2 - maxCos\right) - maxCos\right) \cdot ux} \cdot \color{blue}{\left(\left(uy + uy\right) \cdot \pi\right)} \]
          7. Taylor expanded in maxCos around 0

            \[\leadsto \sqrt{2 \cdot ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
          8. Step-by-step derivation
            1. count-2-revN/A

              \[\leadsto \sqrt{ux + ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
            2. lower-+.f3263.4

              \[\leadsto \sqrt{ux + ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
          9. Applied rewrites63.4%

            \[\leadsto \sqrt{ux + ux} \cdot \left(\left(uy + uy\right) \cdot \pi\right) \]
          10. Add Preprocessing

          Alternative 21: 7.1% accurate, 4.7× speedup?

          \[\begin{array}{l} \\ \left(\sqrt{1 - 1} \cdot \pi\right) \cdot \left(uy + uy\right) \end{array} \]
          (FPCore (ux uy maxCos)
           :precision binary32
           (* (* (sqrt (- 1.0 1.0)) PI) (+ uy uy)))
          float code(float ux, float uy, float maxCos) {
          	return (sqrtf((1.0f - 1.0f)) * ((float) M_PI)) * (uy + uy);
          }
          
          function code(ux, uy, maxCos)
          	return Float32(Float32(sqrt(Float32(Float32(1.0) - Float32(1.0))) * Float32(pi)) * Float32(uy + uy))
          end
          
          function tmp = code(ux, uy, maxCos)
          	tmp = (sqrt((single(1.0) - single(1.0))) * single(pi)) * (uy + uy);
          end
          
          \begin{array}{l}
          
          \\
          \left(\sqrt{1 - 1} \cdot \pi\right) \cdot \left(uy + uy\right)
          \end{array}
          
          Derivation
          1. Initial program 57.6%

            \[\sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{1 - \left(\left(1 - ux\right) + ux \cdot maxCos\right) \cdot \left(\left(1 - ux\right) + ux \cdot maxCos\right)} \]
          2. Taylor expanded in uy around 0

            \[\leadsto \color{blue}{2 \cdot \left(\left(uy \cdot \mathsf{PI}\left(\right)\right) \cdot \sqrt{1 - {\left(\left(1 + maxCos \cdot ux\right) - ux\right)}^{2}}\right)} \]
          3. Applied rewrites50.7%

            \[\leadsto \color{blue}{\left(\sqrt{1 - \mathsf{fma}\left(maxCos, ux, 1 - ux\right) \cdot \mathsf{fma}\left(maxCos, ux, 1 - ux\right)} \cdot \pi\right) \cdot \left(uy + uy\right)} \]
          4. Taylor expanded in ux around 0

            \[\leadsto \left(\sqrt{1 - 1} \cdot \pi\right) \cdot \left(uy + uy\right) \]
          5. Step-by-step derivation
            1. Applied rewrites7.1%

              \[\leadsto \left(\sqrt{1 - 1} \cdot \pi\right) \cdot \left(uy + uy\right) \]
            2. Add Preprocessing

            Reproduce

            ?
            herbie shell --seed 2025130 
            (FPCore (ux uy maxCos)
              :name "UniformSampleCone, y"
              :precision binary32
              :pre (and (and (and (<= 2.328306437e-10 ux) (<= ux 1.0)) (and (<= 2.328306437e-10 uy) (<= uy 1.0))) (and (<= 0.0 maxCos) (<= maxCos 1.0)))
              (* (sin (* (* uy 2.0) PI)) (sqrt (- 1.0 (* (+ (- 1.0 ux) (* ux maxCos)) (+ (- 1.0 ux) (* ux maxCos)))))))