UniformSampleCone, x

Percentage Accurate: 57.5% → 99.0%
Time: 6.3s
Alternatives: 14
Speedup: 11.2×

Specification

?
\[\left(\left(2.328306437 \cdot 10^{-10} \leq ux \land ux \leq 1\right) \land \left(2.328306437 \cdot 10^{-10} \leq uy \land uy \leq 1\right)\right) \land \left(0 \leq maxCos \land maxCos \leq 1\right)\]
\[\begin{array}{l} \\ \begin{array}{l} t_0 := \left(1 - ux\right) + ux \cdot maxCos\\ \cos \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{1 - t\_0 \cdot t\_0} \end{array} \end{array} \]
(FPCore (ux uy maxCos)
 :precision binary32
 (let* ((t_0 (+ (- 1.0 ux) (* ux maxCos))))
   (* (cos (* (* 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 cosf(((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(cos(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 = cos(((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\\
\cos \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 14 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: 57.5% accurate, 1.0× speedup?

\[\begin{array}{l} \\ \begin{array}{l} t_0 := \left(1 - ux\right) + ux \cdot maxCos\\ \cos \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{1 - t\_0 \cdot t\_0} \end{array} \end{array} \]
(FPCore (ux uy maxCos)
 :precision binary32
 (let* ((t_0 (+ (- 1.0 ux) (* ux maxCos))))
   (* (cos (* (* 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 cosf(((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(cos(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 = cos(((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\\
\cos \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{1 - t\_0 \cdot t\_0}
\end{array}
\end{array}

Alternative 1: 99.0% accurate, 0.9× speedup?

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

\\
\sin \left(\left(-\left(\pi + \pi\right) \cdot uy\right) + \frac{\pi}{2}\right) \cdot \sqrt{\left(2 - \mathsf{fma}\left(ux \cdot \left(maxCos - 1\right), maxCos - 1, maxCos + maxCos\right)\right) \cdot ux}
\end{array}
Derivation
  1. Initial program 57.5%

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

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

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

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

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

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

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

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

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

      \[\leadsto \cos \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(\mathsf{fma}\left(-ux, {\left(maxCos - 1\right)}^{2}, 2\right) - 2 \cdot maxCos\right) \cdot ux} \]
    9. unpow2N/A

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

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

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

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

      \[\leadsto \cos \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(\mathsf{fma}\left(-ux, \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - \left(maxCos + maxCos\right)\right) \cdot ux} \]
    14. lower-+.f3298.9

      \[\leadsto \cos \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(\mathsf{fma}\left(-ux, \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - \left(maxCos + maxCos\right)\right) \cdot ux} \]
  4. Applied rewrites98.9%

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

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

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

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

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

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

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

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

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

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

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

      \[\leadsto \sin \left(\mathsf{fma}\left(uy \cdot 2, \pi, \frac{\color{blue}{\pi}}{2}\right)\right) \cdot \sqrt{\left(\mathsf{fma}\left(-ux, \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - \left(maxCos + maxCos\right)\right) \cdot ux} \]
  6. Applied rewrites98.8%

    \[\leadsto \color{blue}{\sin \left(\mathsf{fma}\left(uy \cdot 2, \pi, \frac{\pi}{2}\right)\right)} \cdot \sqrt{\left(\mathsf{fma}\left(-ux, \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - \left(maxCos + maxCos\right)\right) \cdot ux} \]
  7. Applied rewrites99.0%

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

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

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

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

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

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

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

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

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

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

      \[\leadsto \sin \left(\left(-\left(\pi + \pi\right) \cdot uy\right) + \color{blue}{\frac{\mathsf{PI}\left(\right)}{2}}\right) \cdot \sqrt{\left(2 - \mathsf{fma}\left(ux \cdot \left(maxCos - 1\right), maxCos - 1, maxCos + maxCos\right)\right) \cdot ux} \]
    11. lift-PI.f3299.0

      \[\leadsto \sin \left(\left(-\left(\pi + \pi\right) \cdot uy\right) + \frac{\color{blue}{\pi}}{2}\right) \cdot \sqrt{\left(2 - \mathsf{fma}\left(ux \cdot \left(maxCos - 1\right), maxCos - 1, maxCos + maxCos\right)\right) \cdot ux} \]
  9. Applied rewrites99.0%

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

Alternative 2: 99.0% accurate, 1.0× speedup?

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

\\
\cos \left(uy \cdot \left(\pi + \pi\right)\right) \cdot \sqrt{\left(2 - \mathsf{fma}\left(ux \cdot \left(maxCos - 1\right), maxCos - 1, maxCos + maxCos\right)\right) \cdot ux}
\end{array}
Derivation
  1. Initial program 57.5%

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

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

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

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

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

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

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

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

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

      \[\leadsto \cos \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(\mathsf{fma}\left(-ux, {\left(maxCos - 1\right)}^{2}, 2\right) - 2 \cdot maxCos\right) \cdot ux} \]
    9. unpow2N/A

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

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

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

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

      \[\leadsto \cos \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(\mathsf{fma}\left(-ux, \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - \left(maxCos + maxCos\right)\right) \cdot ux} \]
    14. lower-+.f3298.9

      \[\leadsto \cos \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(\mathsf{fma}\left(-ux, \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - \left(maxCos + maxCos\right)\right) \cdot ux} \]
  4. Applied rewrites98.9%

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

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

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

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

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

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

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

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

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

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

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

      \[\leadsto \sin \left(\mathsf{fma}\left(uy \cdot 2, \pi, \frac{\color{blue}{\pi}}{2}\right)\right) \cdot \sqrt{\left(\mathsf{fma}\left(-ux, \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - \left(maxCos + maxCos\right)\right) \cdot ux} \]
  6. Applied rewrites98.8%

    \[\leadsto \color{blue}{\sin \left(\mathsf{fma}\left(uy \cdot 2, \pi, \frac{\pi}{2}\right)\right)} \cdot \sqrt{\left(\mathsf{fma}\left(-ux, \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - \left(maxCos + maxCos\right)\right) \cdot ux} \]
  7. Applied rewrites99.0%

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

Alternative 3: 98.9% accurate, 1.1× speedup?

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

\\
\cos \left(uy \cdot \left(\pi + \pi\right)\right) \cdot \sqrt{\left(2 - \mathsf{fma}\left(\mathsf{fma}\left(ux, -2 + maxCos, 2\right), maxCos, ux\right)\right) \cdot ux}
\end{array}
Derivation
  1. Initial program 57.5%

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

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

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

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

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

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

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

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

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

      \[\leadsto \cos \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(\mathsf{fma}\left(-ux, {\left(maxCos - 1\right)}^{2}, 2\right) - 2 \cdot maxCos\right) \cdot ux} \]
    9. unpow2N/A

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

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

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

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

      \[\leadsto \cos \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(\mathsf{fma}\left(-ux, \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - \left(maxCos + maxCos\right)\right) \cdot ux} \]
    14. lower-+.f3298.9

      \[\leadsto \cos \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(\mathsf{fma}\left(-ux, \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - \left(maxCos + maxCos\right)\right) \cdot ux} \]
  4. Applied rewrites98.9%

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

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

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

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

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

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

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

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

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

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

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

      \[\leadsto \sin \left(\mathsf{fma}\left(uy \cdot 2, \pi, \frac{\color{blue}{\pi}}{2}\right)\right) \cdot \sqrt{\left(\mathsf{fma}\left(-ux, \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - \left(maxCos + maxCos\right)\right) \cdot ux} \]
  6. Applied rewrites98.8%

    \[\leadsto \color{blue}{\sin \left(\mathsf{fma}\left(uy \cdot 2, \pi, \frac{\pi}{2}\right)\right)} \cdot \sqrt{\left(\mathsf{fma}\left(-ux, \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - \left(maxCos + maxCos\right)\right) \cdot ux} \]
  7. Applied rewrites99.0%

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

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

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

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

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

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

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

      \[\leadsto \cos \left(uy \cdot \left(\pi + \pi\right)\right) \cdot \sqrt{\left(2 - \mathsf{fma}\left(\mathsf{fma}\left(ux, -2 + maxCos, 2\right), maxCos, ux\right)\right) \cdot ux} \]
    7. lower-+.f3298.9

      \[\leadsto \cos \left(uy \cdot \left(\pi + \pi\right)\right) \cdot \sqrt{\left(2 - \mathsf{fma}\left(\mathsf{fma}\left(ux, -2 + maxCos, 2\right), maxCos, ux\right)\right) \cdot ux} \]
  10. Applied rewrites98.9%

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

Alternative 4: 98.2% accurate, 1.1× speedup?

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

\\
\cos \left(uy \cdot \left(\pi + \pi\right)\right) \cdot \sqrt{\left(2 - \mathsf{fma}\left(\mathsf{fma}\left(-2, ux, 2\right), maxCos, ux\right)\right) \cdot ux}
\end{array}
Derivation
  1. Initial program 57.5%

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

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

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

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

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

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

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

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

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

      \[\leadsto \cos \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(\mathsf{fma}\left(-ux, {\left(maxCos - 1\right)}^{2}, 2\right) - 2 \cdot maxCos\right) \cdot ux} \]
    9. unpow2N/A

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

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

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

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

      \[\leadsto \cos \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(\mathsf{fma}\left(-ux, \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - \left(maxCos + maxCos\right)\right) \cdot ux} \]
    14. lower-+.f3298.9

      \[\leadsto \cos \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(\mathsf{fma}\left(-ux, \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - \left(maxCos + maxCos\right)\right) \cdot ux} \]
  4. Applied rewrites98.9%

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

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

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

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

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

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

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

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

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

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

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

      \[\leadsto \sin \left(\mathsf{fma}\left(uy \cdot 2, \pi, \frac{\color{blue}{\pi}}{2}\right)\right) \cdot \sqrt{\left(\mathsf{fma}\left(-ux, \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - \left(maxCos + maxCos\right)\right) \cdot ux} \]
  6. Applied rewrites98.8%

    \[\leadsto \color{blue}{\sin \left(\mathsf{fma}\left(uy \cdot 2, \pi, \frac{\pi}{2}\right)\right)} \cdot \sqrt{\left(\mathsf{fma}\left(-ux, \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - \left(maxCos + maxCos\right)\right) \cdot ux} \]
  7. Applied rewrites99.0%

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

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

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

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

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

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

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

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

Alternative 5: 97.4% accurate, 1.2× speedup?

\[\begin{array}{l} \\ \begin{array}{l} \mathbf{if}\;maxCos \leq 6.000000212225132 \cdot 10^{-7}:\\ \;\;\;\;\cos \left(uy \cdot \left(\pi + \pi\right)\right) \cdot \sqrt{\left(2 - ux\right) \cdot ux}\\ \mathbf{else}:\\ \;\;\;\;\left(1 - 2 \cdot \left(\left(uy \cdot uy\right) \cdot \left(\pi \cdot \pi\right)\right)\right) \cdot \sqrt{\left(\mathsf{fma}\left(-ux, \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - \left(maxCos + maxCos\right)\right) \cdot ux}\\ \end{array} \end{array} \]
(FPCore (ux uy maxCos)
 :precision binary32
 (if (<= maxCos 6.000000212225132e-7)
   (* (cos (* uy (+ PI PI))) (sqrt (* (- 2.0 ux) ux)))
   (*
    (- 1.0 (* 2.0 (* (* uy uy) (* PI PI))))
    (sqrt
     (*
      (- (fma (- ux) (* (- maxCos 1.0) (- maxCos 1.0)) 2.0) (+ maxCos maxCos))
      ux)))))
float code(float ux, float uy, float maxCos) {
	float tmp;
	if (maxCos <= 6.000000212225132e-7f) {
		tmp = cosf((uy * (((float) M_PI) + ((float) M_PI)))) * sqrtf(((2.0f - ux) * ux));
	} else {
		tmp = (1.0f - (2.0f * ((uy * uy) * (((float) M_PI) * ((float) M_PI))))) * sqrtf(((fmaf(-ux, ((maxCos - 1.0f) * (maxCos - 1.0f)), 2.0f) - (maxCos + maxCos)) * ux));
	}
	return tmp;
}
function code(ux, uy, maxCos)
	tmp = Float32(0.0)
	if (maxCos <= Float32(6.000000212225132e-7))
		tmp = Float32(cos(Float32(uy * Float32(Float32(pi) + Float32(pi)))) * sqrt(Float32(Float32(Float32(2.0) - ux) * ux)));
	else
		tmp = Float32(Float32(Float32(1.0) - Float32(Float32(2.0) * Float32(Float32(uy * uy) * Float32(Float32(pi) * Float32(pi))))) * sqrt(Float32(Float32(fma(Float32(-ux), Float32(Float32(maxCos - Float32(1.0)) * Float32(maxCos - Float32(1.0))), Float32(2.0)) - Float32(maxCos + maxCos)) * ux)));
	end
	return tmp
end
\begin{array}{l}

\\
\begin{array}{l}
\mathbf{if}\;maxCos \leq 6.000000212225132 \cdot 10^{-7}:\\
\;\;\;\;\cos \left(uy \cdot \left(\pi + \pi\right)\right) \cdot \sqrt{\left(2 - ux\right) \cdot ux}\\

\mathbf{else}:\\
\;\;\;\;\left(1 - 2 \cdot \left(\left(uy \cdot uy\right) \cdot \left(\pi \cdot \pi\right)\right)\right) \cdot \sqrt{\left(\mathsf{fma}\left(-ux, \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - \left(maxCos + maxCos\right)\right) \cdot ux}\\


\end{array}
\end{array}
Derivation
  1. Split input into 2 regimes
  2. if maxCos < 6.0000002e-7

    1. Initial program 57.5%

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

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

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

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

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

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

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

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

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

        \[\leadsto \cos \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(\mathsf{fma}\left(-ux, {\left(maxCos - 1\right)}^{2}, 2\right) - 2 \cdot maxCos\right) \cdot ux} \]
      9. unpow2N/A

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

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

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

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

        \[\leadsto \cos \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(\mathsf{fma}\left(-ux, \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - \left(maxCos + maxCos\right)\right) \cdot ux} \]
      14. lower-+.f3299.0

        \[\leadsto \cos \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(\mathsf{fma}\left(-ux, \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - \left(maxCos + maxCos\right)\right) \cdot ux} \]
    4. Applied rewrites99.0%

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

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

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

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

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

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

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

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

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

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

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

        \[\leadsto \sin \left(\mathsf{fma}\left(uy \cdot 2, \pi, \frac{\color{blue}{\pi}}{2}\right)\right) \cdot \sqrt{\left(\mathsf{fma}\left(-ux, \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - \left(maxCos + maxCos\right)\right) \cdot ux} \]
    6. Applied rewrites98.8%

      \[\leadsto \color{blue}{\sin \left(\mathsf{fma}\left(uy \cdot 2, \pi, \frac{\pi}{2}\right)\right)} \cdot \sqrt{\left(\mathsf{fma}\left(-ux, \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - \left(maxCos + maxCos\right)\right) \cdot ux} \]
    7. Applied rewrites99.0%

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

      \[\leadsto \cos \left(uy \cdot \left(\pi + \pi\right)\right) \cdot \sqrt{\left(2 - ux\right) \cdot ux} \]
    9. Step-by-step derivation
      1. Applied rewrites98.9%

        \[\leadsto \cos \left(uy \cdot \left(\pi + \pi\right)\right) \cdot \sqrt{\left(2 - ux\right) \cdot ux} \]

      if 6.0000002e-7 < maxCos

      1. Initial program 57.5%

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

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

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

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

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

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

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

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

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

          \[\leadsto \cos \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(\mathsf{fma}\left(-ux, {\left(maxCos - 1\right)}^{2}, 2\right) - 2 \cdot maxCos\right) \cdot ux} \]
        9. unpow2N/A

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

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

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

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

          \[\leadsto \cos \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(\mathsf{fma}\left(-ux, \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - \left(maxCos + maxCos\right)\right) \cdot ux} \]
        14. lower-+.f3298.8

          \[\leadsto \cos \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(\mathsf{fma}\left(-ux, \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - \left(maxCos + maxCos\right)\right) \cdot ux} \]
      4. Applied rewrites98.8%

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

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

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

          \[\leadsto \left(1 - \color{blue}{\left(\mathsf{neg}\left(-2\right)\right) \cdot \left({uy}^{2} \cdot {\mathsf{PI}\left(\right)}^{2}\right)}\right) \cdot \sqrt{\left(\mathsf{fma}\left(-ux, \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - \left(maxCos + maxCos\right)\right) \cdot ux} \]
        3. metadata-evalN/A

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

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

          \[\leadsto \left(1 - 2 \cdot \left({uy}^{2} \cdot \color{blue}{{\mathsf{PI}\left(\right)}^{2}}\right)\right) \cdot \sqrt{\left(\mathsf{fma}\left(-ux, \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - \left(maxCos + maxCos\right)\right) \cdot ux} \]
        6. unpow2N/A

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

          \[\leadsto \left(1 - 2 \cdot \left(\left(uy \cdot uy\right) \cdot {\color{blue}{\mathsf{PI}\left(\right)}}^{2}\right)\right) \cdot \sqrt{\left(\mathsf{fma}\left(-ux, \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - \left(maxCos + maxCos\right)\right) \cdot ux} \]
        8. unpow2N/A

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

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

          \[\leadsto \left(1 - 2 \cdot \left(\left(uy \cdot uy\right) \cdot \left(\pi \cdot \mathsf{PI}\left(\right)\right)\right)\right) \cdot \sqrt{\left(\mathsf{fma}\left(-ux, \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - \left(maxCos + maxCos\right)\right) \cdot ux} \]
        11. lift-PI.f3287.7

          \[\leadsto \left(1 - 2 \cdot \left(\left(uy \cdot uy\right) \cdot \left(\pi \cdot \pi\right)\right)\right) \cdot \sqrt{\left(\mathsf{fma}\left(-ux, \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - \left(maxCos + maxCos\right)\right) \cdot ux} \]
      7. Applied rewrites87.7%

        \[\leadsto \color{blue}{\left(1 - 2 \cdot \left(\left(uy \cdot uy\right) \cdot \left(\pi \cdot \pi\right)\right)\right)} \cdot \sqrt{\left(\mathsf{fma}\left(-ux, \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - \left(maxCos + maxCos\right)\right) \cdot ux} \]
    10. Recombined 2 regimes into one program.
    11. Add Preprocessing

    Alternative 6: 97.1% accurate, 1.1× speedup?

    \[\begin{array}{l} \\ \cos \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(\left(2 - 1 \cdot ux\right) - \left(maxCos + maxCos\right)\right) \cdot ux} \end{array} \]
    (FPCore (ux uy maxCos)
     :precision binary32
     (*
      (cos (* (* uy 2.0) PI))
      (sqrt (* (- (- 2.0 (* 1.0 ux)) (+ maxCos maxCos)) ux))))
    float code(float ux, float uy, float maxCos) {
    	return cosf(((uy * 2.0f) * ((float) M_PI))) * sqrtf((((2.0f - (1.0f * ux)) - (maxCos + maxCos)) * ux));
    }
    
    function code(ux, uy, maxCos)
    	return Float32(cos(Float32(Float32(uy * Float32(2.0)) * Float32(pi))) * sqrt(Float32(Float32(Float32(Float32(2.0) - Float32(Float32(1.0) * ux)) - Float32(maxCos + maxCos)) * ux)))
    end
    
    function tmp = code(ux, uy, maxCos)
    	tmp = cos(((uy * single(2.0)) * single(pi))) * sqrt((((single(2.0) - (single(1.0) * ux)) - (maxCos + maxCos)) * ux));
    end
    
    \begin{array}{l}
    
    \\
    \cos \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(\left(2 - 1 \cdot ux\right) - \left(maxCos + maxCos\right)\right) \cdot ux}
    \end{array}
    
    Derivation
    1. Initial program 57.5%

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

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

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

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

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

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

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

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

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

        \[\leadsto \cos \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(\mathsf{fma}\left(-ux, {\left(maxCos - 1\right)}^{2}, 2\right) - 2 \cdot maxCos\right) \cdot ux} \]
      9. unpow2N/A

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

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

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

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

        \[\leadsto \cos \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(\mathsf{fma}\left(-ux, \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - \left(maxCos + maxCos\right)\right) \cdot ux} \]
      14. lower-+.f3298.9

        \[\leadsto \cos \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(\mathsf{fma}\left(-ux, \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - \left(maxCos + maxCos\right)\right) \cdot ux} \]
    4. Applied rewrites98.9%

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

      \[\leadsto \cos \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(\left(2 + -1 \cdot ux\right) - \left(maxCos + maxCos\right)\right) \cdot ux} \]
    6. Step-by-step derivation
      1. fp-cancel-sign-sub-invN/A

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

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

        \[\leadsto \cos \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(\left(2 - 1 \cdot ux\right) - \left(maxCos + maxCos\right)\right) \cdot ux} \]
      4. lower-*.f3297.4

        \[\leadsto \cos \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(\left(2 - 1 \cdot ux\right) - \left(maxCos + maxCos\right)\right) \cdot ux} \]
    7. Applied rewrites97.4%

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

    Alternative 7: 88.2% accurate, 1.6× speedup?

    \[\begin{array}{l} \\ \left(1 - 2 \cdot \left(\left(uy \cdot uy\right) \cdot \left(\pi \cdot \pi\right)\right)\right) \cdot \sqrt{\left(\mathsf{fma}\left(-ux, \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - \left(maxCos + maxCos\right)\right) \cdot ux} \end{array} \]
    (FPCore (ux uy maxCos)
     :precision binary32
     (*
      (- 1.0 (* 2.0 (* (* uy uy) (* PI PI))))
      (sqrt
       (*
        (- (fma (- ux) (* (- maxCos 1.0) (- maxCos 1.0)) 2.0) (+ maxCos maxCos))
        ux))))
    float code(float ux, float uy, float maxCos) {
    	return (1.0f - (2.0f * ((uy * uy) * (((float) M_PI) * ((float) M_PI))))) * sqrtf(((fmaf(-ux, ((maxCos - 1.0f) * (maxCos - 1.0f)), 2.0f) - (maxCos + maxCos)) * ux));
    }
    
    function code(ux, uy, maxCos)
    	return Float32(Float32(Float32(1.0) - Float32(Float32(2.0) * Float32(Float32(uy * uy) * Float32(Float32(pi) * Float32(pi))))) * sqrt(Float32(Float32(fma(Float32(-ux), Float32(Float32(maxCos - Float32(1.0)) * Float32(maxCos - Float32(1.0))), Float32(2.0)) - Float32(maxCos + maxCos)) * ux)))
    end
    
    \begin{array}{l}
    
    \\
    \left(1 - 2 \cdot \left(\left(uy \cdot uy\right) \cdot \left(\pi \cdot \pi\right)\right)\right) \cdot \sqrt{\left(\mathsf{fma}\left(-ux, \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - \left(maxCos + maxCos\right)\right) \cdot ux}
    \end{array}
    
    Derivation
    1. Initial program 57.5%

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

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

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

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

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

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

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

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

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

        \[\leadsto \cos \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(\mathsf{fma}\left(-ux, {\left(maxCos - 1\right)}^{2}, 2\right) - 2 \cdot maxCos\right) \cdot ux} \]
      9. unpow2N/A

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

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

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

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

        \[\leadsto \cos \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(\mathsf{fma}\left(-ux, \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - \left(maxCos + maxCos\right)\right) \cdot ux} \]
      14. lower-+.f3298.9

        \[\leadsto \cos \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{\left(\mathsf{fma}\left(-ux, \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - \left(maxCos + maxCos\right)\right) \cdot ux} \]
    4. Applied rewrites98.9%

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

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

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

        \[\leadsto \left(1 - \color{blue}{\left(\mathsf{neg}\left(-2\right)\right) \cdot \left({uy}^{2} \cdot {\mathsf{PI}\left(\right)}^{2}\right)}\right) \cdot \sqrt{\left(\mathsf{fma}\left(-ux, \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - \left(maxCos + maxCos\right)\right) \cdot ux} \]
      3. metadata-evalN/A

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

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

        \[\leadsto \left(1 - 2 \cdot \left({uy}^{2} \cdot \color{blue}{{\mathsf{PI}\left(\right)}^{2}}\right)\right) \cdot \sqrt{\left(\mathsf{fma}\left(-ux, \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - \left(maxCos + maxCos\right)\right) \cdot ux} \]
      6. unpow2N/A

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

        \[\leadsto \left(1 - 2 \cdot \left(\left(uy \cdot uy\right) \cdot {\color{blue}{\mathsf{PI}\left(\right)}}^{2}\right)\right) \cdot \sqrt{\left(\mathsf{fma}\left(-ux, \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - \left(maxCos + maxCos\right)\right) \cdot ux} \]
      8. unpow2N/A

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

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

        \[\leadsto \left(1 - 2 \cdot \left(\left(uy \cdot uy\right) \cdot \left(\pi \cdot \mathsf{PI}\left(\right)\right)\right)\right) \cdot \sqrt{\left(\mathsf{fma}\left(-ux, \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - \left(maxCos + maxCos\right)\right) \cdot ux} \]
      11. lift-PI.f3288.2

        \[\leadsto \left(1 - 2 \cdot \left(\left(uy \cdot uy\right) \cdot \left(\pi \cdot \pi\right)\right)\right) \cdot \sqrt{\left(\mathsf{fma}\left(-ux, \left(maxCos - 1\right) \cdot \left(maxCos - 1\right), 2\right) - \left(maxCos + maxCos\right)\right) \cdot ux} \]
    7. Applied rewrites88.2%

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

    Alternative 8: 79.9% accurate, 2.8× speedup?

    \[\begin{array}{l} \\ \sqrt{ux \cdot \left(2 - \mathsf{fma}\left(ux \cdot \left(maxCos - 1\right), maxCos - 1, maxCos + maxCos\right)\right)} \end{array} \]
    (FPCore (ux uy maxCos)
     :precision binary32
     (sqrt
      (* ux (- 2.0 (fma (* ux (- maxCos 1.0)) (- maxCos 1.0) (+ maxCos maxCos))))))
    float code(float ux, float uy, float maxCos) {
    	return sqrtf((ux * (2.0f - fmaf((ux * (maxCos - 1.0f)), (maxCos - 1.0f), (maxCos + maxCos)))));
    }
    
    function code(ux, uy, maxCos)
    	return sqrt(Float32(ux * Float32(Float32(2.0) - fma(Float32(ux * Float32(maxCos - Float32(1.0))), Float32(maxCos - Float32(1.0)), Float32(maxCos + maxCos)))))
    end
    
    \begin{array}{l}
    
    \\
    \sqrt{ux \cdot \left(2 - \mathsf{fma}\left(ux \cdot \left(maxCos - 1\right), maxCos - 1, maxCos + maxCos\right)\right)}
    \end{array}
    
    Derivation
    1. Initial program 57.5%

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

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

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

        \[\leadsto \sqrt{1 - {\left(\left(1 + maxCos \cdot ux\right) - ux\right)}^{2}} \]
      3. unpow2N/A

        \[\leadsto \sqrt{1 - \left(\left(1 + maxCos \cdot ux\right) - ux\right) \cdot \left(\left(1 + maxCos \cdot ux\right) - ux\right)} \]
      4. lower-*.f32N/A

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

        \[\leadsto \sqrt{1 - \left(\left(1 + maxCos \cdot ux\right) - ux\right) \cdot \left(\left(1 + maxCos \cdot ux\right) - ux\right)} \]
      6. +-commutativeN/A

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

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

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

        \[\leadsto \sqrt{1 - \left(\mathsf{fma}\left(maxCos, ux, 1\right) - ux\right) \cdot \left(\left(maxCos \cdot ux + 1\right) - ux\right)} \]
      10. lower-fma.f3249.2

        \[\leadsto \sqrt{1 - \left(\mathsf{fma}\left(maxCos, ux, 1\right) - ux\right) \cdot \left(\mathsf{fma}\left(maxCos, ux, 1\right) - ux\right)} \]
    4. Applied rewrites49.2%

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

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

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

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

        \[\leadsto \sqrt{ux \cdot \left(\left(2 - \left(\mathsf{neg}\left(-1\right)\right) \cdot \left(ux \cdot {\left(maxCos - 1\right)}^{2}\right)\right) - 2 \cdot maxCos\right)} \]
      4. lower--.f32N/A

        \[\leadsto \sqrt{ux \cdot \left(\left(2 - \left(\mathsf{neg}\left(-1\right)\right) \cdot \left(ux \cdot {\left(maxCos - 1\right)}^{2}\right)\right) - 2 \cdot maxCos\right)} \]
      5. metadata-evalN/A

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

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

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

        \[\leadsto \sqrt{ux \cdot \left(\left(2 - 1 \cdot \left(ux \cdot \left(\left(maxCos - 1\right) \cdot \left(maxCos - 1\right)\right)\right)\right) - 2 \cdot maxCos\right)} \]
      9. lift--.f32N/A

        \[\leadsto \sqrt{ux \cdot \left(\left(2 - 1 \cdot \left(ux \cdot \left(\left(maxCos - 1\right) \cdot \left(maxCos - 1\right)\right)\right)\right) - 2 \cdot maxCos\right)} \]
      10. lift--.f32N/A

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

        \[\leadsto \sqrt{ux \cdot \left(\left(2 - 1 \cdot \left(ux \cdot \left(\left(maxCos - 1\right) \cdot \left(maxCos - 1\right)\right)\right)\right) - 2 \cdot maxCos\right)} \]
      12. lower-*.f3279.8

        \[\leadsto \sqrt{ux \cdot \left(\left(2 - 1 \cdot \left(ux \cdot \left(\left(maxCos - 1\right) \cdot \left(maxCos - 1\right)\right)\right)\right) - 2 \cdot maxCos\right)} \]
    7. Applied rewrites79.8%

      \[\leadsto \sqrt{ux \cdot \left(\left(2 - 1 \cdot \left(ux \cdot \left(\left(maxCos - 1\right) \cdot \left(maxCos - 1\right)\right)\right)\right) - 2 \cdot maxCos\right)} \]
    8. Step-by-step derivation
      1. lift-*.f32N/A

        \[\leadsto \sqrt{ux \cdot \left(\left(2 - 1 \cdot \left(ux \cdot \left(\left(maxCos - 1\right) \cdot \left(maxCos - 1\right)\right)\right)\right) - 2 \cdot maxCos\right)} \]
      2. lift--.f32N/A

        \[\leadsto \sqrt{ux \cdot \left(\left(2 - 1 \cdot \left(ux \cdot \left(\left(maxCos - 1\right) \cdot \left(maxCos - 1\right)\right)\right)\right) - 2 \cdot maxCos\right)} \]
      3. lift--.f32N/A

        \[\leadsto \sqrt{ux \cdot \left(\left(2 - 1 \cdot \left(ux \cdot \left(\left(maxCos - 1\right) \cdot \left(maxCos - 1\right)\right)\right)\right) - 2 \cdot maxCos\right)} \]
      4. lift-*.f32N/A

        \[\leadsto \sqrt{ux \cdot \left(\left(2 - 1 \cdot \left(ux \cdot \left(\left(maxCos - 1\right) \cdot \left(maxCos - 1\right)\right)\right)\right) - 2 \cdot maxCos\right)} \]
      5. lift-*.f32N/A

        \[\leadsto \sqrt{ux \cdot \left(\left(2 - 1 \cdot \left(ux \cdot \left(\left(maxCos - 1\right) \cdot \left(maxCos - 1\right)\right)\right)\right) - 2 \cdot maxCos\right)} \]
      6. lift-*.f32N/A

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

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

        \[\leadsto \sqrt{ux \cdot \left(\left(2 - 1 \cdot \left(ux \cdot \left(\left(maxCos - 1\right) \cdot \left(maxCos - 1\right)\right)\right)\right) - 2 \cdot maxCos\right)} \]
      9. associate--l-N/A

        \[\leadsto \sqrt{ux \cdot \left(2 - \left(1 \cdot \left(ux \cdot \left(\left(maxCos - 1\right) \cdot \left(maxCos - 1\right)\right)\right) + 2 \cdot maxCos\right)\right)} \]
      10. *-lft-identityN/A

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

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

        \[\leadsto \sqrt{ux \cdot \left(2 - \left(2 \cdot maxCos + ux \cdot {\left(maxCos - 1\right)}^{2}\right)\right)} \]
    9. Applied rewrites79.9%

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

    Alternative 9: 79.3% accurate, 3.4× speedup?

    \[\begin{array}{l} \\ \sqrt{ux \cdot \left(\mathsf{fma}\left(\left(ux + ux\right) - 2, maxCos, -ux\right) + 2\right)} \end{array} \]
    (FPCore (ux uy maxCos)
     :precision binary32
     (sqrt (* ux (+ (fma (- (+ ux ux) 2.0) maxCos (- ux)) 2.0))))
    float code(float ux, float uy, float maxCos) {
    	return sqrtf((ux * (fmaf(((ux + ux) - 2.0f), maxCos, -ux) + 2.0f)));
    }
    
    function code(ux, uy, maxCos)
    	return sqrt(Float32(ux * Float32(fma(Float32(Float32(ux + ux) - Float32(2.0)), maxCos, Float32(-ux)) + Float32(2.0))))
    end
    
    \begin{array}{l}
    
    \\
    \sqrt{ux \cdot \left(\mathsf{fma}\left(\left(ux + ux\right) - 2, maxCos, -ux\right) + 2\right)}
    \end{array}
    
    Derivation
    1. Initial program 57.5%

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

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

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

        \[\leadsto \sqrt{1 - {\left(\left(1 + maxCos \cdot ux\right) - ux\right)}^{2}} \]
      3. unpow2N/A

        \[\leadsto \sqrt{1 - \left(\left(1 + maxCos \cdot ux\right) - ux\right) \cdot \left(\left(1 + maxCos \cdot ux\right) - ux\right)} \]
      4. lower-*.f32N/A

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

        \[\leadsto \sqrt{1 - \left(\left(1 + maxCos \cdot ux\right) - ux\right) \cdot \left(\left(1 + maxCos \cdot ux\right) - ux\right)} \]
      6. +-commutativeN/A

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

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

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

        \[\leadsto \sqrt{1 - \left(\mathsf{fma}\left(maxCos, ux, 1\right) - ux\right) \cdot \left(\left(maxCos \cdot ux + 1\right) - ux\right)} \]
      10. lower-fma.f3249.2

        \[\leadsto \sqrt{1 - \left(\mathsf{fma}\left(maxCos, ux, 1\right) - ux\right) \cdot \left(\mathsf{fma}\left(maxCos, ux, 1\right) - ux\right)} \]
    4. Applied rewrites49.2%

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

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

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

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

        \[\leadsto \sqrt{ux \cdot \left(\left(2 - \left(\mathsf{neg}\left(-1\right)\right) \cdot \left(ux \cdot {\left(maxCos - 1\right)}^{2}\right)\right) - 2 \cdot maxCos\right)} \]
      4. lower--.f32N/A

        \[\leadsto \sqrt{ux \cdot \left(\left(2 - \left(\mathsf{neg}\left(-1\right)\right) \cdot \left(ux \cdot {\left(maxCos - 1\right)}^{2}\right)\right) - 2 \cdot maxCos\right)} \]
      5. metadata-evalN/A

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

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

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

        \[\leadsto \sqrt{ux \cdot \left(\left(2 - 1 \cdot \left(ux \cdot \left(\left(maxCos - 1\right) \cdot \left(maxCos - 1\right)\right)\right)\right) - 2 \cdot maxCos\right)} \]
      9. lift--.f32N/A

        \[\leadsto \sqrt{ux \cdot \left(\left(2 - 1 \cdot \left(ux \cdot \left(\left(maxCos - 1\right) \cdot \left(maxCos - 1\right)\right)\right)\right) - 2 \cdot maxCos\right)} \]
      10. lift--.f32N/A

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

        \[\leadsto \sqrt{ux \cdot \left(\left(2 - 1 \cdot \left(ux \cdot \left(\left(maxCos - 1\right) \cdot \left(maxCos - 1\right)\right)\right)\right) - 2 \cdot maxCos\right)} \]
      12. lower-*.f3279.8

        \[\leadsto \sqrt{ux \cdot \left(\left(2 - 1 \cdot \left(ux \cdot \left(\left(maxCos - 1\right) \cdot \left(maxCos - 1\right)\right)\right)\right) - 2 \cdot maxCos\right)} \]
    7. Applied rewrites79.8%

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

      \[\leadsto \sqrt{ux \cdot \left(2 - ux\right)} \]
    9. Step-by-step derivation
      1. metadata-evalN/A

        \[\leadsto \sqrt{ux \cdot \left(2 - ux\right)} \]
      2. pow2N/A

        \[\leadsto \sqrt{ux \cdot \left(2 - ux\right)} \]
      3. fp-cancel-sign-sub-invN/A

        \[\leadsto \sqrt{ux \cdot \left(2 - ux\right)} \]
      4. fp-cancel-sign-sub-invN/A

        \[\leadsto \sqrt{ux \cdot \left(2 - ux\right)} \]
      5. metadata-evalN/A

        \[\leadsto \sqrt{ux \cdot \left(2 - ux\right)} \]
      6. pow2N/A

        \[\leadsto \sqrt{ux \cdot \left(2 - ux\right)} \]
    10. Applied rewrites75.8%

      \[\leadsto \sqrt{ux \cdot \left(2 - ux\right)} \]
    11. Taylor expanded in maxCos around 0

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

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

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

        \[\leadsto \sqrt{ux \cdot \left(\left(maxCos \cdot \left(2 \cdot ux - 2\right) + -1 \cdot ux\right) + 2\right)} \]
      4. *-commutativeN/A

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

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

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

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

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

        \[\leadsto \sqrt{ux \cdot \left(\mathsf{fma}\left(\left(ux + ux\right) - 2, maxCos, \mathsf{neg}\left(ux\right)\right) + 2\right)} \]
      10. lower-neg.f3279.3

        \[\leadsto \sqrt{ux \cdot \left(\mathsf{fma}\left(\left(ux + ux\right) - 2, maxCos, -ux\right) + 2\right)} \]
    13. Applied rewrites79.3%

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

    Alternative 10: 79.3% accurate, 3.4× speedup?

    \[\begin{array}{l} \\ \sqrt{ux \cdot \left(\mathsf{fma}\left(-maxCos, \mathsf{fma}\left(-2, ux, 2\right), 2\right) - ux\right)} \end{array} \]
    (FPCore (ux uy maxCos)
     :precision binary32
     (sqrt (* ux (- (fma (- maxCos) (fma -2.0 ux 2.0) 2.0) ux))))
    float code(float ux, float uy, float maxCos) {
    	return sqrtf((ux * (fmaf(-maxCos, fmaf(-2.0f, ux, 2.0f), 2.0f) - ux)));
    }
    
    function code(ux, uy, maxCos)
    	return sqrt(Float32(ux * Float32(fma(Float32(-maxCos), fma(Float32(-2.0), ux, Float32(2.0)), Float32(2.0)) - ux)))
    end
    
    \begin{array}{l}
    
    \\
    \sqrt{ux \cdot \left(\mathsf{fma}\left(-maxCos, \mathsf{fma}\left(-2, ux, 2\right), 2\right) - ux\right)}
    \end{array}
    
    Derivation
    1. Initial program 57.5%

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

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

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

        \[\leadsto \sqrt{1 - {\left(\left(1 + maxCos \cdot ux\right) - ux\right)}^{2}} \]
      3. unpow2N/A

        \[\leadsto \sqrt{1 - \left(\left(1 + maxCos \cdot ux\right) - ux\right) \cdot \left(\left(1 + maxCos \cdot ux\right) - ux\right)} \]
      4. lower-*.f32N/A

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

        \[\leadsto \sqrt{1 - \left(\left(1 + maxCos \cdot ux\right) - ux\right) \cdot \left(\left(1 + maxCos \cdot ux\right) - ux\right)} \]
      6. +-commutativeN/A

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

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

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

        \[\leadsto \sqrt{1 - \left(\mathsf{fma}\left(maxCos, ux, 1\right) - ux\right) \cdot \left(\left(maxCos \cdot ux + 1\right) - ux\right)} \]
      10. lower-fma.f3249.2

        \[\leadsto \sqrt{1 - \left(\mathsf{fma}\left(maxCos, ux, 1\right) - ux\right) \cdot \left(\mathsf{fma}\left(maxCos, ux, 1\right) - ux\right)} \]
    4. Applied rewrites49.2%

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

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

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

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

        \[\leadsto \sqrt{ux \cdot \left(\left(2 - \left(\mathsf{neg}\left(-1\right)\right) \cdot \left(ux \cdot {\left(maxCos - 1\right)}^{2}\right)\right) - 2 \cdot maxCos\right)} \]
      4. lower--.f32N/A

        \[\leadsto \sqrt{ux \cdot \left(\left(2 - \left(\mathsf{neg}\left(-1\right)\right) \cdot \left(ux \cdot {\left(maxCos - 1\right)}^{2}\right)\right) - 2 \cdot maxCos\right)} \]
      5. metadata-evalN/A

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

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

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

        \[\leadsto \sqrt{ux \cdot \left(\left(2 - 1 \cdot \left(ux \cdot \left(\left(maxCos - 1\right) \cdot \left(maxCos - 1\right)\right)\right)\right) - 2 \cdot maxCos\right)} \]
      9. lift--.f32N/A

        \[\leadsto \sqrt{ux \cdot \left(\left(2 - 1 \cdot \left(ux \cdot \left(\left(maxCos - 1\right) \cdot \left(maxCos - 1\right)\right)\right)\right) - 2 \cdot maxCos\right)} \]
      10. lift--.f32N/A

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

        \[\leadsto \sqrt{ux \cdot \left(\left(2 - 1 \cdot \left(ux \cdot \left(\left(maxCos - 1\right) \cdot \left(maxCos - 1\right)\right)\right)\right) - 2 \cdot maxCos\right)} \]
      12. lower-*.f3279.8

        \[\leadsto \sqrt{ux \cdot \left(\left(2 - 1 \cdot \left(ux \cdot \left(\left(maxCos - 1\right) \cdot \left(maxCos - 1\right)\right)\right)\right) - 2 \cdot maxCos\right)} \]
    7. Applied rewrites79.8%

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

      \[\leadsto \sqrt{ux \cdot \left(\left(2 + -1 \cdot \left(maxCos \cdot \left(2 + -2 \cdot ux\right)\right)\right) - ux\right)} \]
    9. Step-by-step derivation
      1. metadata-evalN/A

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

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

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

        \[\leadsto \sqrt{ux \cdot \left(\left(2 + -1 \cdot \left(maxCos \cdot \left(2 + -2 \cdot ux\right)\right)\right) - ux\right)} \]
      5. metadata-evalN/A

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

        \[\leadsto \sqrt{ux \cdot \left(\left(2 + -1 \cdot \left(maxCos \cdot \left(2 + -2 \cdot ux\right)\right)\right) - ux\right)} \]
    10. Applied rewrites79.3%

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

    Alternative 11: 78.8% accurate, 4.8× speedup?

    \[\begin{array}{l} \\ \sqrt{ux \cdot \left(\left(2 - ux\right) - 2 \cdot maxCos\right)} \end{array} \]
    (FPCore (ux uy maxCos)
     :precision binary32
     (sqrt (* ux (- (- 2.0 ux) (* 2.0 maxCos)))))
    float code(float ux, float uy, float maxCos) {
    	return sqrtf((ux * ((2.0f - ux) - (2.0f * maxCos))));
    }
    
    module fmin_fmax_functions
        implicit none
        private
        public fmax
        public fmin
    
        interface fmax
            module procedure fmax88
            module procedure fmax44
            module procedure fmax84
            module procedure fmax48
        end interface
        interface fmin
            module procedure fmin88
            module procedure fmin44
            module procedure fmin84
            module procedure fmin48
        end interface
    contains
        real(8) function fmax88(x, y) result (res)
            real(8), intent (in) :: x
            real(8), intent (in) :: y
            res = merge(y, merge(x, max(x, y), y /= y), x /= x)
        end function
        real(4) function fmax44(x, y) result (res)
            real(4), intent (in) :: x
            real(4), intent (in) :: y
            res = merge(y, merge(x, max(x, y), y /= y), x /= x)
        end function
        real(8) function fmax84(x, y) result(res)
            real(8), intent (in) :: x
            real(4), intent (in) :: y
            res = merge(dble(y), merge(x, max(x, dble(y)), y /= y), x /= x)
        end function
        real(8) function fmax48(x, y) result(res)
            real(4), intent (in) :: x
            real(8), intent (in) :: y
            res = merge(y, merge(dble(x), max(dble(x), y), y /= y), x /= x)
        end function
        real(8) function fmin88(x, y) result (res)
            real(8), intent (in) :: x
            real(8), intent (in) :: y
            res = merge(y, merge(x, min(x, y), y /= y), x /= x)
        end function
        real(4) function fmin44(x, y) result (res)
            real(4), intent (in) :: x
            real(4), intent (in) :: y
            res = merge(y, merge(x, min(x, y), y /= y), x /= x)
        end function
        real(8) function fmin84(x, y) result(res)
            real(8), intent (in) :: x
            real(4), intent (in) :: y
            res = merge(dble(y), merge(x, min(x, dble(y)), y /= y), x /= x)
        end function
        real(8) function fmin48(x, y) result(res)
            real(4), intent (in) :: x
            real(8), intent (in) :: y
            res = merge(y, merge(dble(x), min(dble(x), y), y /= y), x /= x)
        end function
    end module
    
    real(4) function code(ux, uy, maxcos)
    use fmin_fmax_functions
        real(4), intent (in) :: ux
        real(4), intent (in) :: uy
        real(4), intent (in) :: maxcos
        code = sqrt((ux * ((2.0e0 - ux) - (2.0e0 * maxcos))))
    end function
    
    function code(ux, uy, maxCos)
    	return sqrt(Float32(ux * Float32(Float32(Float32(2.0) - ux) - Float32(Float32(2.0) * maxCos))))
    end
    
    function tmp = code(ux, uy, maxCos)
    	tmp = sqrt((ux * ((single(2.0) - ux) - (single(2.0) * maxCos))));
    end
    
    \begin{array}{l}
    
    \\
    \sqrt{ux \cdot \left(\left(2 - ux\right) - 2 \cdot maxCos\right)}
    \end{array}
    
    Derivation
    1. Initial program 57.5%

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

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

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

        \[\leadsto \sqrt{1 - {\left(\left(1 + maxCos \cdot ux\right) - ux\right)}^{2}} \]
      3. unpow2N/A

        \[\leadsto \sqrt{1 - \left(\left(1 + maxCos \cdot ux\right) - ux\right) \cdot \left(\left(1 + maxCos \cdot ux\right) - ux\right)} \]
      4. lower-*.f32N/A

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

        \[\leadsto \sqrt{1 - \left(\left(1 + maxCos \cdot ux\right) - ux\right) \cdot \left(\left(1 + maxCos \cdot ux\right) - ux\right)} \]
      6. +-commutativeN/A

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

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

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

        \[\leadsto \sqrt{1 - \left(\mathsf{fma}\left(maxCos, ux, 1\right) - ux\right) \cdot \left(\left(maxCos \cdot ux + 1\right) - ux\right)} \]
      10. lower-fma.f3249.2

        \[\leadsto \sqrt{1 - \left(\mathsf{fma}\left(maxCos, ux, 1\right) - ux\right) \cdot \left(\mathsf{fma}\left(maxCos, ux, 1\right) - ux\right)} \]
    4. Applied rewrites49.2%

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

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

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

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

        \[\leadsto \sqrt{ux \cdot \left(\left(2 - \left(\mathsf{neg}\left(-1\right)\right) \cdot \left(ux \cdot {\left(maxCos - 1\right)}^{2}\right)\right) - 2 \cdot maxCos\right)} \]
      4. lower--.f32N/A

        \[\leadsto \sqrt{ux \cdot \left(\left(2 - \left(\mathsf{neg}\left(-1\right)\right) \cdot \left(ux \cdot {\left(maxCos - 1\right)}^{2}\right)\right) - 2 \cdot maxCos\right)} \]
      5. metadata-evalN/A

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

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

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

        \[\leadsto \sqrt{ux \cdot \left(\left(2 - 1 \cdot \left(ux \cdot \left(\left(maxCos - 1\right) \cdot \left(maxCos - 1\right)\right)\right)\right) - 2 \cdot maxCos\right)} \]
      9. lift--.f32N/A

        \[\leadsto \sqrt{ux \cdot \left(\left(2 - 1 \cdot \left(ux \cdot \left(\left(maxCos - 1\right) \cdot \left(maxCos - 1\right)\right)\right)\right) - 2 \cdot maxCos\right)} \]
      10. lift--.f32N/A

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

        \[\leadsto \sqrt{ux \cdot \left(\left(2 - 1 \cdot \left(ux \cdot \left(\left(maxCos - 1\right) \cdot \left(maxCos - 1\right)\right)\right)\right) - 2 \cdot maxCos\right)} \]
      12. lower-*.f3279.8

        \[\leadsto \sqrt{ux \cdot \left(\left(2 - 1 \cdot \left(ux \cdot \left(\left(maxCos - 1\right) \cdot \left(maxCos - 1\right)\right)\right)\right) - 2 \cdot maxCos\right)} \]
    7. Applied rewrites79.8%

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

      \[\leadsto \sqrt{ux \cdot \left(\left(2 - ux\right) - 2 \cdot maxCos\right)} \]
    9. Step-by-step derivation
      1. Applied rewrites78.8%

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

      Alternative 12: 75.8% accurate, 7.9× speedup?

      \[\begin{array}{l} \\ \sqrt{ux \cdot \left(2 - ux\right)} \end{array} \]
      (FPCore (ux uy maxCos) :precision binary32 (sqrt (* ux (- 2.0 ux))))
      float code(float ux, float uy, float maxCos) {
      	return sqrtf((ux * (2.0f - ux)));
      }
      
      module fmin_fmax_functions
          implicit none
          private
          public fmax
          public fmin
      
          interface fmax
              module procedure fmax88
              module procedure fmax44
              module procedure fmax84
              module procedure fmax48
          end interface
          interface fmin
              module procedure fmin88
              module procedure fmin44
              module procedure fmin84
              module procedure fmin48
          end interface
      contains
          real(8) function fmax88(x, y) result (res)
              real(8), intent (in) :: x
              real(8), intent (in) :: y
              res = merge(y, merge(x, max(x, y), y /= y), x /= x)
          end function
          real(4) function fmax44(x, y) result (res)
              real(4), intent (in) :: x
              real(4), intent (in) :: y
              res = merge(y, merge(x, max(x, y), y /= y), x /= x)
          end function
          real(8) function fmax84(x, y) result(res)
              real(8), intent (in) :: x
              real(4), intent (in) :: y
              res = merge(dble(y), merge(x, max(x, dble(y)), y /= y), x /= x)
          end function
          real(8) function fmax48(x, y) result(res)
              real(4), intent (in) :: x
              real(8), intent (in) :: y
              res = merge(y, merge(dble(x), max(dble(x), y), y /= y), x /= x)
          end function
          real(8) function fmin88(x, y) result (res)
              real(8), intent (in) :: x
              real(8), intent (in) :: y
              res = merge(y, merge(x, min(x, y), y /= y), x /= x)
          end function
          real(4) function fmin44(x, y) result (res)
              real(4), intent (in) :: x
              real(4), intent (in) :: y
              res = merge(y, merge(x, min(x, y), y /= y), x /= x)
          end function
          real(8) function fmin84(x, y) result(res)
              real(8), intent (in) :: x
              real(4), intent (in) :: y
              res = merge(dble(y), merge(x, min(x, dble(y)), y /= y), x /= x)
          end function
          real(8) function fmin48(x, y) result(res)
              real(4), intent (in) :: x
              real(8), intent (in) :: y
              res = merge(y, merge(dble(x), min(dble(x), y), y /= y), x /= x)
          end function
      end module
      
      real(4) function code(ux, uy, maxcos)
      use fmin_fmax_functions
          real(4), intent (in) :: ux
          real(4), intent (in) :: uy
          real(4), intent (in) :: maxcos
          code = sqrt((ux * (2.0e0 - ux)))
      end function
      
      function code(ux, uy, maxCos)
      	return sqrt(Float32(ux * Float32(Float32(2.0) - ux)))
      end
      
      function tmp = code(ux, uy, maxCos)
      	tmp = sqrt((ux * (single(2.0) - ux)));
      end
      
      \begin{array}{l}
      
      \\
      \sqrt{ux \cdot \left(2 - ux\right)}
      \end{array}
      
      Derivation
      1. Initial program 57.5%

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

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

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

          \[\leadsto \sqrt{1 - {\left(\left(1 + maxCos \cdot ux\right) - ux\right)}^{2}} \]
        3. unpow2N/A

          \[\leadsto \sqrt{1 - \left(\left(1 + maxCos \cdot ux\right) - ux\right) \cdot \left(\left(1 + maxCos \cdot ux\right) - ux\right)} \]
        4. lower-*.f32N/A

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

          \[\leadsto \sqrt{1 - \left(\left(1 + maxCos \cdot ux\right) - ux\right) \cdot \left(\left(1 + maxCos \cdot ux\right) - ux\right)} \]
        6. +-commutativeN/A

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

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

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

          \[\leadsto \sqrt{1 - \left(\mathsf{fma}\left(maxCos, ux, 1\right) - ux\right) \cdot \left(\left(maxCos \cdot ux + 1\right) - ux\right)} \]
        10. lower-fma.f3249.2

          \[\leadsto \sqrt{1 - \left(\mathsf{fma}\left(maxCos, ux, 1\right) - ux\right) \cdot \left(\mathsf{fma}\left(maxCos, ux, 1\right) - ux\right)} \]
      4. Applied rewrites49.2%

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

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

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

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

          \[\leadsto \sqrt{ux \cdot \left(\left(2 - \left(\mathsf{neg}\left(-1\right)\right) \cdot \left(ux \cdot {\left(maxCos - 1\right)}^{2}\right)\right) - 2 \cdot maxCos\right)} \]
        4. lower--.f32N/A

          \[\leadsto \sqrt{ux \cdot \left(\left(2 - \left(\mathsf{neg}\left(-1\right)\right) \cdot \left(ux \cdot {\left(maxCos - 1\right)}^{2}\right)\right) - 2 \cdot maxCos\right)} \]
        5. metadata-evalN/A

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

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

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

          \[\leadsto \sqrt{ux \cdot \left(\left(2 - 1 \cdot \left(ux \cdot \left(\left(maxCos - 1\right) \cdot \left(maxCos - 1\right)\right)\right)\right) - 2 \cdot maxCos\right)} \]
        9. lift--.f32N/A

          \[\leadsto \sqrt{ux \cdot \left(\left(2 - 1 \cdot \left(ux \cdot \left(\left(maxCos - 1\right) \cdot \left(maxCos - 1\right)\right)\right)\right) - 2 \cdot maxCos\right)} \]
        10. lift--.f32N/A

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

          \[\leadsto \sqrt{ux \cdot \left(\left(2 - 1 \cdot \left(ux \cdot \left(\left(maxCos - 1\right) \cdot \left(maxCos - 1\right)\right)\right)\right) - 2 \cdot maxCos\right)} \]
        12. lower-*.f3279.8

          \[\leadsto \sqrt{ux \cdot \left(\left(2 - 1 \cdot \left(ux \cdot \left(\left(maxCos - 1\right) \cdot \left(maxCos - 1\right)\right)\right)\right) - 2 \cdot maxCos\right)} \]
      7. Applied rewrites79.8%

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

        \[\leadsto \sqrt{ux \cdot \left(2 - ux\right)} \]
      9. Step-by-step derivation
        1. metadata-evalN/A

          \[\leadsto \sqrt{ux \cdot \left(2 - ux\right)} \]
        2. pow2N/A

          \[\leadsto \sqrt{ux \cdot \left(2 - ux\right)} \]
        3. fp-cancel-sign-sub-invN/A

          \[\leadsto \sqrt{ux \cdot \left(2 - ux\right)} \]
        4. fp-cancel-sign-sub-invN/A

          \[\leadsto \sqrt{ux \cdot \left(2 - ux\right)} \]
        5. metadata-evalN/A

          \[\leadsto \sqrt{ux \cdot \left(2 - ux\right)} \]
        6. pow2N/A

          \[\leadsto \sqrt{ux \cdot \left(2 - ux\right)} \]
      10. Applied rewrites75.8%

        \[\leadsto \sqrt{ux \cdot \left(2 - ux\right)} \]
      11. Add Preprocessing

      Alternative 13: 62.2% accurate, 11.2× speedup?

      \[\begin{array}{l} \\ \sqrt{ux \cdot 2} \end{array} \]
      (FPCore (ux uy maxCos) :precision binary32 (sqrt (* ux 2.0)))
      float code(float ux, float uy, float maxCos) {
      	return sqrtf((ux * 2.0f));
      }
      
      module fmin_fmax_functions
          implicit none
          private
          public fmax
          public fmin
      
          interface fmax
              module procedure fmax88
              module procedure fmax44
              module procedure fmax84
              module procedure fmax48
          end interface
          interface fmin
              module procedure fmin88
              module procedure fmin44
              module procedure fmin84
              module procedure fmin48
          end interface
      contains
          real(8) function fmax88(x, y) result (res)
              real(8), intent (in) :: x
              real(8), intent (in) :: y
              res = merge(y, merge(x, max(x, y), y /= y), x /= x)
          end function
          real(4) function fmax44(x, y) result (res)
              real(4), intent (in) :: x
              real(4), intent (in) :: y
              res = merge(y, merge(x, max(x, y), y /= y), x /= x)
          end function
          real(8) function fmax84(x, y) result(res)
              real(8), intent (in) :: x
              real(4), intent (in) :: y
              res = merge(dble(y), merge(x, max(x, dble(y)), y /= y), x /= x)
          end function
          real(8) function fmax48(x, y) result(res)
              real(4), intent (in) :: x
              real(8), intent (in) :: y
              res = merge(y, merge(dble(x), max(dble(x), y), y /= y), x /= x)
          end function
          real(8) function fmin88(x, y) result (res)
              real(8), intent (in) :: x
              real(8), intent (in) :: y
              res = merge(y, merge(x, min(x, y), y /= y), x /= x)
          end function
          real(4) function fmin44(x, y) result (res)
              real(4), intent (in) :: x
              real(4), intent (in) :: y
              res = merge(y, merge(x, min(x, y), y /= y), x /= x)
          end function
          real(8) function fmin84(x, y) result(res)
              real(8), intent (in) :: x
              real(4), intent (in) :: y
              res = merge(dble(y), merge(x, min(x, dble(y)), y /= y), x /= x)
          end function
          real(8) function fmin48(x, y) result(res)
              real(4), intent (in) :: x
              real(8), intent (in) :: y
              res = merge(y, merge(dble(x), min(dble(x), y), y /= y), x /= x)
          end function
      end module
      
      real(4) function code(ux, uy, maxcos)
      use fmin_fmax_functions
          real(4), intent (in) :: ux
          real(4), intent (in) :: uy
          real(4), intent (in) :: maxcos
          code = sqrt((ux * 2.0e0))
      end function
      
      function code(ux, uy, maxCos)
      	return sqrt(Float32(ux * Float32(2.0)))
      end
      
      function tmp = code(ux, uy, maxCos)
      	tmp = sqrt((ux * single(2.0)));
      end
      
      \begin{array}{l}
      
      \\
      \sqrt{ux \cdot 2}
      \end{array}
      
      Derivation
      1. Initial program 57.5%

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

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

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

          \[\leadsto \sqrt{1 - {\left(\left(1 + maxCos \cdot ux\right) - ux\right)}^{2}} \]
        3. unpow2N/A

          \[\leadsto \sqrt{1 - \left(\left(1 + maxCos \cdot ux\right) - ux\right) \cdot \left(\left(1 + maxCos \cdot ux\right) - ux\right)} \]
        4. lower-*.f32N/A

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

          \[\leadsto \sqrt{1 - \left(\left(1 + maxCos \cdot ux\right) - ux\right) \cdot \left(\left(1 + maxCos \cdot ux\right) - ux\right)} \]
        6. +-commutativeN/A

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

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

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

          \[\leadsto \sqrt{1 - \left(\mathsf{fma}\left(maxCos, ux, 1\right) - ux\right) \cdot \left(\left(maxCos \cdot ux + 1\right) - ux\right)} \]
        10. lower-fma.f3249.2

          \[\leadsto \sqrt{1 - \left(\mathsf{fma}\left(maxCos, ux, 1\right) - ux\right) \cdot \left(\mathsf{fma}\left(maxCos, ux, 1\right) - ux\right)} \]
      4. Applied rewrites49.2%

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

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

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

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

          \[\leadsto \sqrt{ux \cdot \left(\left(2 - \left(\mathsf{neg}\left(-1\right)\right) \cdot \left(ux \cdot {\left(maxCos - 1\right)}^{2}\right)\right) - 2 \cdot maxCos\right)} \]
        4. lower--.f32N/A

          \[\leadsto \sqrt{ux \cdot \left(\left(2 - \left(\mathsf{neg}\left(-1\right)\right) \cdot \left(ux \cdot {\left(maxCos - 1\right)}^{2}\right)\right) - 2 \cdot maxCos\right)} \]
        5. metadata-evalN/A

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

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

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

          \[\leadsto \sqrt{ux \cdot \left(\left(2 - 1 \cdot \left(ux \cdot \left(\left(maxCos - 1\right) \cdot \left(maxCos - 1\right)\right)\right)\right) - 2 \cdot maxCos\right)} \]
        9. lift--.f32N/A

          \[\leadsto \sqrt{ux \cdot \left(\left(2 - 1 \cdot \left(ux \cdot \left(\left(maxCos - 1\right) \cdot \left(maxCos - 1\right)\right)\right)\right) - 2 \cdot maxCos\right)} \]
        10. lift--.f32N/A

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

          \[\leadsto \sqrt{ux \cdot \left(\left(2 - 1 \cdot \left(ux \cdot \left(\left(maxCos - 1\right) \cdot \left(maxCos - 1\right)\right)\right)\right) - 2 \cdot maxCos\right)} \]
        12. lower-*.f3279.8

          \[\leadsto \sqrt{ux \cdot \left(\left(2 - 1 \cdot \left(ux \cdot \left(\left(maxCos - 1\right) \cdot \left(maxCos - 1\right)\right)\right)\right) - 2 \cdot maxCos\right)} \]
      7. Applied rewrites79.8%

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

        \[\leadsto \sqrt{ux \cdot \left(2 - ux\right)} \]
      9. Step-by-step derivation
        1. metadata-evalN/A

          \[\leadsto \sqrt{ux \cdot \left(2 - ux\right)} \]
        2. pow2N/A

          \[\leadsto \sqrt{ux \cdot \left(2 - ux\right)} \]
        3. fp-cancel-sign-sub-invN/A

          \[\leadsto \sqrt{ux \cdot \left(2 - ux\right)} \]
        4. fp-cancel-sign-sub-invN/A

          \[\leadsto \sqrt{ux \cdot \left(2 - ux\right)} \]
        5. metadata-evalN/A

          \[\leadsto \sqrt{ux \cdot \left(2 - ux\right)} \]
        6. pow2N/A

          \[\leadsto \sqrt{ux \cdot \left(2 - ux\right)} \]
      10. Applied rewrites75.8%

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

        \[\leadsto \sqrt{ux \cdot 2} \]
      12. Step-by-step derivation
        1. Applied rewrites62.2%

          \[\leadsto \sqrt{ux \cdot 2} \]
        2. Add Preprocessing

        Alternative 14: 6.6% accurate, 12.2× speedup?

        \[\begin{array}{l} \\ \sqrt{1 - 1} \end{array} \]
        (FPCore (ux uy maxCos) :precision binary32 (sqrt (- 1.0 1.0)))
        float code(float ux, float uy, float maxCos) {
        	return sqrtf((1.0f - 1.0f));
        }
        
        module fmin_fmax_functions
            implicit none
            private
            public fmax
            public fmin
        
            interface fmax
                module procedure fmax88
                module procedure fmax44
                module procedure fmax84
                module procedure fmax48
            end interface
            interface fmin
                module procedure fmin88
                module procedure fmin44
                module procedure fmin84
                module procedure fmin48
            end interface
        contains
            real(8) function fmax88(x, y) result (res)
                real(8), intent (in) :: x
                real(8), intent (in) :: y
                res = merge(y, merge(x, max(x, y), y /= y), x /= x)
            end function
            real(4) function fmax44(x, y) result (res)
                real(4), intent (in) :: x
                real(4), intent (in) :: y
                res = merge(y, merge(x, max(x, y), y /= y), x /= x)
            end function
            real(8) function fmax84(x, y) result(res)
                real(8), intent (in) :: x
                real(4), intent (in) :: y
                res = merge(dble(y), merge(x, max(x, dble(y)), y /= y), x /= x)
            end function
            real(8) function fmax48(x, y) result(res)
                real(4), intent (in) :: x
                real(8), intent (in) :: y
                res = merge(y, merge(dble(x), max(dble(x), y), y /= y), x /= x)
            end function
            real(8) function fmin88(x, y) result (res)
                real(8), intent (in) :: x
                real(8), intent (in) :: y
                res = merge(y, merge(x, min(x, y), y /= y), x /= x)
            end function
            real(4) function fmin44(x, y) result (res)
                real(4), intent (in) :: x
                real(4), intent (in) :: y
                res = merge(y, merge(x, min(x, y), y /= y), x /= x)
            end function
            real(8) function fmin84(x, y) result(res)
                real(8), intent (in) :: x
                real(4), intent (in) :: y
                res = merge(dble(y), merge(x, min(x, dble(y)), y /= y), x /= x)
            end function
            real(8) function fmin48(x, y) result(res)
                real(4), intent (in) :: x
                real(8), intent (in) :: y
                res = merge(y, merge(dble(x), min(dble(x), y), y /= y), x /= x)
            end function
        end module
        
        real(4) function code(ux, uy, maxcos)
        use fmin_fmax_functions
            real(4), intent (in) :: ux
            real(4), intent (in) :: uy
            real(4), intent (in) :: maxcos
            code = sqrt((1.0e0 - 1.0e0))
        end function
        
        function code(ux, uy, maxCos)
        	return sqrt(Float32(Float32(1.0) - Float32(1.0)))
        end
        
        function tmp = code(ux, uy, maxCos)
        	tmp = sqrt((single(1.0) - single(1.0)));
        end
        
        \begin{array}{l}
        
        \\
        \sqrt{1 - 1}
        \end{array}
        
        Derivation
        1. Initial program 57.5%

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

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

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

            \[\leadsto \sqrt{1 - {\left(\left(1 + maxCos \cdot ux\right) - ux\right)}^{2}} \]
          3. unpow2N/A

            \[\leadsto \sqrt{1 - \left(\left(1 + maxCos \cdot ux\right) - ux\right) \cdot \left(\left(1 + maxCos \cdot ux\right) - ux\right)} \]
          4. lower-*.f32N/A

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

            \[\leadsto \sqrt{1 - \left(\left(1 + maxCos \cdot ux\right) - ux\right) \cdot \left(\left(1 + maxCos \cdot ux\right) - ux\right)} \]
          6. +-commutativeN/A

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

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

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

            \[\leadsto \sqrt{1 - \left(\mathsf{fma}\left(maxCos, ux, 1\right) - ux\right) \cdot \left(\left(maxCos \cdot ux + 1\right) - ux\right)} \]
          10. lower-fma.f3249.2

            \[\leadsto \sqrt{1 - \left(\mathsf{fma}\left(maxCos, ux, 1\right) - ux\right) \cdot \left(\mathsf{fma}\left(maxCos, ux, 1\right) - ux\right)} \]
        4. Applied rewrites49.2%

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

          \[\leadsto \sqrt{1 - 1} \]
        6. Step-by-step derivation
          1. Applied rewrites6.6%

            \[\leadsto \sqrt{1 - 1} \]
          2. Add Preprocessing

          Reproduce

          ?
          herbie shell --seed 2025112 
          (FPCore (ux uy maxCos)
            :name "UniformSampleCone, x"
            :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)))
            (* (cos (* (* uy 2.0) PI)) (sqrt (- 1.0 (* (+ (- 1.0 ux) (* ux maxCos)) (+ (- 1.0 ux) (* ux maxCos)))))))