
(FPCore (xi yi zi ux uy maxCos)
:precision binary32
(let* ((t_0 (* (* (- 1.0 ux) maxCos) ux))
(t_1 (sqrt (- 1.0 (* t_0 t_0))))
(t_2 (* (* uy 2.0) PI)))
(+ (+ (* (* (cos t_2) t_1) xi) (* (* (sin t_2) t_1) yi)) (* t_0 zi))))
float code(float xi, float yi, float zi, float ux, float uy, float maxCos) {
float t_0 = ((1.0f - ux) * maxCos) * ux;
float t_1 = sqrtf((1.0f - (t_0 * t_0)));
float t_2 = (uy * 2.0f) * ((float) M_PI);
return (((cosf(t_2) * t_1) * xi) + ((sinf(t_2) * t_1) * yi)) + (t_0 * zi);
}
function code(xi, yi, zi, ux, uy, maxCos) t_0 = Float32(Float32(Float32(Float32(1.0) - ux) * maxCos) * ux) t_1 = sqrt(Float32(Float32(1.0) - Float32(t_0 * t_0))) t_2 = Float32(Float32(uy * Float32(2.0)) * Float32(pi)) return Float32(Float32(Float32(Float32(cos(t_2) * t_1) * xi) + Float32(Float32(sin(t_2) * t_1) * yi)) + Float32(t_0 * zi)) end
function tmp = code(xi, yi, zi, ux, uy, maxCos) t_0 = ((single(1.0) - ux) * maxCos) * ux; t_1 = sqrt((single(1.0) - (t_0 * t_0))); t_2 = (uy * single(2.0)) * single(pi); tmp = (((cos(t_2) * t_1) * xi) + ((sin(t_2) * t_1) * yi)) + (t_0 * zi); end
\begin{array}{l}
\\
\begin{array}{l}
t_0 := \left(\left(1 - ux\right) \cdot maxCos\right) \cdot ux\\
t_1 := \sqrt{1 - t\_0 \cdot t\_0}\\
t_2 := \left(uy \cdot 2\right) \cdot \pi\\
\left(\left(\cos t\_2 \cdot t\_1\right) \cdot xi + \left(\sin t\_2 \cdot t\_1\right) \cdot yi\right) + t\_0 \cdot zi
\end{array}
\end{array}
Herbie found 15 alternatives:
| Alternative | Accuracy | Speedup |
|---|
(FPCore (xi yi zi ux uy maxCos)
:precision binary32
(let* ((t_0 (* (* (- 1.0 ux) maxCos) ux))
(t_1 (sqrt (- 1.0 (* t_0 t_0))))
(t_2 (* (* uy 2.0) PI)))
(+ (+ (* (* (cos t_2) t_1) xi) (* (* (sin t_2) t_1) yi)) (* t_0 zi))))
float code(float xi, float yi, float zi, float ux, float uy, float maxCos) {
float t_0 = ((1.0f - ux) * maxCos) * ux;
float t_1 = sqrtf((1.0f - (t_0 * t_0)));
float t_2 = (uy * 2.0f) * ((float) M_PI);
return (((cosf(t_2) * t_1) * xi) + ((sinf(t_2) * t_1) * yi)) + (t_0 * zi);
}
function code(xi, yi, zi, ux, uy, maxCos) t_0 = Float32(Float32(Float32(Float32(1.0) - ux) * maxCos) * ux) t_1 = sqrt(Float32(Float32(1.0) - Float32(t_0 * t_0))) t_2 = Float32(Float32(uy * Float32(2.0)) * Float32(pi)) return Float32(Float32(Float32(Float32(cos(t_2) * t_1) * xi) + Float32(Float32(sin(t_2) * t_1) * yi)) + Float32(t_0 * zi)) end
function tmp = code(xi, yi, zi, ux, uy, maxCos) t_0 = ((single(1.0) - ux) * maxCos) * ux; t_1 = sqrt((single(1.0) - (t_0 * t_0))); t_2 = (uy * single(2.0)) * single(pi); tmp = (((cos(t_2) * t_1) * xi) + ((sin(t_2) * t_1) * yi)) + (t_0 * zi); end
\begin{array}{l}
\\
\begin{array}{l}
t_0 := \left(\left(1 - ux\right) \cdot maxCos\right) \cdot ux\\
t_1 := \sqrt{1 - t\_0 \cdot t\_0}\\
t_2 := \left(uy \cdot 2\right) \cdot \pi\\
\left(\left(\cos t\_2 \cdot t\_1\right) \cdot xi + \left(\sin t\_2 \cdot t\_1\right) \cdot yi\right) + t\_0 \cdot zi
\end{array}
\end{array}
(FPCore (xi yi zi ux uy maxCos)
:precision binary32
(let* ((t_0 (* (* (- 1.0 ux) maxCos) ux))
(t_1 (* PI (* 2.0 uy)))
(t_2 (sin (acos t_0))))
(fma (* (cos t_1) t_2) xi (fma (sin t_1) (* t_2 yi) (* t_0 zi)))))
float code(float xi, float yi, float zi, float ux, float uy, float maxCos) {
float t_0 = ((1.0f - ux) * maxCos) * ux;
float t_1 = ((float) M_PI) * (2.0f * uy);
float t_2 = sinf(acosf(t_0));
return fmaf((cosf(t_1) * t_2), xi, fmaf(sinf(t_1), (t_2 * yi), (t_0 * zi)));
}
function code(xi, yi, zi, ux, uy, maxCos) t_0 = Float32(Float32(Float32(Float32(1.0) - ux) * maxCos) * ux) t_1 = Float32(Float32(pi) * Float32(Float32(2.0) * uy)) t_2 = sin(acos(t_0)) return fma(Float32(cos(t_1) * t_2), xi, fma(sin(t_1), Float32(t_2 * yi), Float32(t_0 * zi))) end
\begin{array}{l}
\\
\begin{array}{l}
t_0 := \left(\left(1 - ux\right) \cdot maxCos\right) \cdot ux\\
t_1 := \pi \cdot \left(2 \cdot uy\right)\\
t_2 := \sin \cos^{-1} t\_0\\
\mathsf{fma}\left(\cos t\_1 \cdot t\_2, xi, \mathsf{fma}\left(\sin t\_1, t\_2 \cdot yi, t\_0 \cdot zi\right)\right)
\end{array}
\end{array}
Initial program 98.9%
Applied rewrites99.0%
(FPCore (xi yi zi ux uy maxCos)
:precision binary32
(let* ((t_0 (* (- 1.0 ux) maxCos))
(t_1 (* t_0 ux))
(t_2 (sqrt (- 1.0 (* t_1 t_1))))
(t_3 (* (* uy 2.0) PI)))
(+
(+ (* (* (cos t_3) t_2) xi) (* (* (sin t_3) t_2) yi))
(* t_0 (* zi ux)))))
float code(float xi, float yi, float zi, float ux, float uy, float maxCos) {
float t_0 = (1.0f - ux) * maxCos;
float t_1 = t_0 * ux;
float t_2 = sqrtf((1.0f - (t_1 * t_1)));
float t_3 = (uy * 2.0f) * ((float) M_PI);
return (((cosf(t_3) * t_2) * xi) + ((sinf(t_3) * t_2) * yi)) + (t_0 * (zi * ux));
}
function code(xi, yi, zi, ux, uy, maxCos) t_0 = Float32(Float32(Float32(1.0) - ux) * maxCos) t_1 = Float32(t_0 * ux) t_2 = sqrt(Float32(Float32(1.0) - Float32(t_1 * t_1))) t_3 = Float32(Float32(uy * Float32(2.0)) * Float32(pi)) return Float32(Float32(Float32(Float32(cos(t_3) * t_2) * xi) + Float32(Float32(sin(t_3) * t_2) * yi)) + Float32(t_0 * Float32(zi * ux))) end
function tmp = code(xi, yi, zi, ux, uy, maxCos) t_0 = (single(1.0) - ux) * maxCos; t_1 = t_0 * ux; t_2 = sqrt((single(1.0) - (t_1 * t_1))); t_3 = (uy * single(2.0)) * single(pi); tmp = (((cos(t_3) * t_2) * xi) + ((sin(t_3) * t_2) * yi)) + (t_0 * (zi * ux)); end
\begin{array}{l}
\\
\begin{array}{l}
t_0 := \left(1 - ux\right) \cdot maxCos\\
t_1 := t\_0 \cdot ux\\
t_2 := \sqrt{1 - t\_1 \cdot t\_1}\\
t_3 := \left(uy \cdot 2\right) \cdot \pi\\
\left(\left(\cos t\_3 \cdot t\_2\right) \cdot xi + \left(\sin t\_3 \cdot t\_2\right) \cdot yi\right) + t\_0 \cdot \left(zi \cdot ux\right)
\end{array}
\end{array}
Initial program 98.9%
lift-*.f32N/A
lift-*.f32N/A
lift--.f32N/A
lift-*.f32N/A
associate-*l*N/A
lower-*.f32N/A
lift-*.f32N/A
lift--.f32N/A
*-commutativeN/A
lower-*.f3298.9
Applied rewrites98.9%
(FPCore (xi yi zi ux uy maxCos)
:precision binary32
(+
(+
(*
(*
(cos (* (* uy 2.0) PI))
(+
1.0
(* (* ux ux) (fma -0.5 (* maxCos maxCos) (* (* maxCos maxCos) ux)))))
xi)
(* (sin (* PI (* 2.0 uy))) yi))
(* (* (* (- 1.0 ux) maxCos) ux) zi)))
float code(float xi, float yi, float zi, float ux, float uy, float maxCos) {
return (((cosf(((uy * 2.0f) * ((float) M_PI))) * (1.0f + ((ux * ux) * fmaf(-0.5f, (maxCos * maxCos), ((maxCos * maxCos) * ux))))) * xi) + (sinf((((float) M_PI) * (2.0f * uy))) * yi)) + ((((1.0f - ux) * maxCos) * ux) * zi);
}
function code(xi, yi, zi, ux, uy, maxCos) return Float32(Float32(Float32(Float32(cos(Float32(Float32(uy * Float32(2.0)) * Float32(pi))) * Float32(Float32(1.0) + Float32(Float32(ux * ux) * fma(Float32(-0.5), Float32(maxCos * maxCos), Float32(Float32(maxCos * maxCos) * ux))))) * xi) + Float32(sin(Float32(Float32(pi) * Float32(Float32(2.0) * uy))) * yi)) + Float32(Float32(Float32(Float32(Float32(1.0) - ux) * maxCos) * ux) * zi)) end
\begin{array}{l}
\\
\left(\left(\cos \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \left(1 + \left(ux \cdot ux\right) \cdot \mathsf{fma}\left(-0.5, maxCos \cdot maxCos, \left(maxCos \cdot maxCos\right) \cdot ux\right)\right)\right) \cdot xi + \sin \left(\pi \cdot \left(2 \cdot uy\right)\right) \cdot yi\right) + \left(\left(\left(1 - ux\right) \cdot maxCos\right) \cdot ux\right) \cdot zi
\end{array}
Initial program 98.9%
Taylor expanded in ux around 0
associate-*r*N/A
*-commutativeN/A
lift-*.f32N/A
lift-*.f32N/A
lift-PI.f32N/A
lift-sin.f3298.8
lift-PI.f32N/A
lift-*.f32N/A
*-commutativeN/A
lower-*.f32N/A
lift-PI.f3298.8
lift-*.f32N/A
*-commutativeN/A
lower-*.f3298.8
Applied rewrites98.8%
Taylor expanded in ux around 0
lower-+.f32N/A
lower-*.f32N/A
unpow2N/A
lower-*.f32N/A
lower-fma.f32N/A
unpow2N/A
lower-*.f32N/A
lower-*.f32N/A
unpow2N/A
lower-*.f3298.8
Applied rewrites98.8%
(FPCore (xi yi zi ux uy maxCos) :precision binary32 (let* ((t_0 (* PI (* 2.0 uy)))) (fma (* maxCos ux) (* (- 1.0 ux) zi) (fma (cos t_0) xi (* (sin t_0) yi)))))
float code(float xi, float yi, float zi, float ux, float uy, float maxCos) {
float t_0 = ((float) M_PI) * (2.0f * uy);
return fmaf((maxCos * ux), ((1.0f - ux) * zi), fmaf(cosf(t_0), xi, (sinf(t_0) * yi)));
}
function code(xi, yi, zi, ux, uy, maxCos) t_0 = Float32(Float32(pi) * Float32(Float32(2.0) * uy)) return fma(Float32(maxCos * ux), Float32(Float32(Float32(1.0) - ux) * zi), fma(cos(t_0), xi, Float32(sin(t_0) * yi))) end
\begin{array}{l}
\\
\begin{array}{l}
t_0 := \pi \cdot \left(2 \cdot uy\right)\\
\mathsf{fma}\left(maxCos \cdot ux, \left(1 - ux\right) \cdot zi, \mathsf{fma}\left(\cos t\_0, xi, \sin t\_0 \cdot yi\right)\right)
\end{array}
\end{array}
Initial program 98.9%
Taylor expanded in maxCos around 0
associate-*r*N/A
lower-fma.f32N/A
lower-*.f32N/A
*-commutativeN/A
lower-*.f32N/A
lift--.f32N/A
*-commutativeN/A
Applied rewrites98.7%
(FPCore (xi yi zi ux uy maxCos) :precision binary32 (let* ((t_0 (* 2.0 (* uy PI)))) (fma maxCos (* ux (* zi (- 1.0 ux))) (fma xi (cos t_0) (* yi (sin t_0))))))
float code(float xi, float yi, float zi, float ux, float uy, float maxCos) {
float t_0 = 2.0f * (uy * ((float) M_PI));
return fmaf(maxCos, (ux * (zi * (1.0f - ux))), fmaf(xi, cosf(t_0), (yi * sinf(t_0))));
}
function code(xi, yi, zi, ux, uy, maxCos) t_0 = Float32(Float32(2.0) * Float32(uy * Float32(pi))) return fma(maxCos, Float32(ux * Float32(zi * Float32(Float32(1.0) - ux))), fma(xi, cos(t_0), Float32(yi * sin(t_0)))) end
\begin{array}{l}
\\
\begin{array}{l}
t_0 := 2 \cdot \left(uy \cdot \pi\right)\\
\mathsf{fma}\left(maxCos, ux \cdot \left(zi \cdot \left(1 - ux\right)\right), \mathsf{fma}\left(xi, \cos t\_0, yi \cdot \sin t\_0\right)\right)
\end{array}
\end{array}
Initial program 98.9%
Taylor expanded in zi around inf
*-commutativeN/A
lower-*.f32N/A
*-commutativeN/A
lower-*.f32N/A
*-commutativeN/A
lower-*.f32N/A
lift--.f3213.4
Applied rewrites13.4%
Taylor expanded in maxCos around 0
lower-fma.f32N/A
lower-*.f32N/A
lower-*.f32N/A
lift--.f32N/A
lower-fma.f32N/A
Applied rewrites98.7%
(FPCore (xi yi zi ux uy maxCos) :precision binary32 (let* ((t_0 (* 2.0 (* uy PI)))) (fma maxCos (* ux zi) (fma xi (cos t_0) (* yi (sin t_0))))))
float code(float xi, float yi, float zi, float ux, float uy, float maxCos) {
float t_0 = 2.0f * (uy * ((float) M_PI));
return fmaf(maxCos, (ux * zi), fmaf(xi, cosf(t_0), (yi * sinf(t_0))));
}
function code(xi, yi, zi, ux, uy, maxCos) t_0 = Float32(Float32(2.0) * Float32(uy * Float32(pi))) return fma(maxCos, Float32(ux * zi), fma(xi, cos(t_0), Float32(yi * sin(t_0)))) end
\begin{array}{l}
\\
\begin{array}{l}
t_0 := 2 \cdot \left(uy \cdot \pi\right)\\
\mathsf{fma}\left(maxCos, ux \cdot zi, \mathsf{fma}\left(xi, \cos t\_0, yi \cdot \sin t\_0\right)\right)
\end{array}
\end{array}
Initial program 98.9%
lift-*.f32N/A
lift-*.f32N/A
lift--.f32N/A
lift-*.f32N/A
associate-*l*N/A
lower-*.f32N/A
lift-*.f32N/A
lift--.f32N/A
*-commutativeN/A
lower-*.f3298.9
Applied rewrites98.9%
Taylor expanded in ux around 0
lower-fma.f32N/A
lower-*.f32N/A
lower-fma.f32N/A
lower-cos.f32N/A
lower-*.f32N/A
lower-*.f32N/A
lift-PI.f32N/A
lower-*.f32N/A
Applied rewrites95.9%
(FPCore (xi yi zi ux uy maxCos) :precision binary32 (let* ((t_0 (* PI (* 2.0 uy)))) (fma (* maxCos ux) zi (fma (cos t_0) xi (* (sin t_0) yi)))))
float code(float xi, float yi, float zi, float ux, float uy, float maxCos) {
float t_0 = ((float) M_PI) * (2.0f * uy);
return fmaf((maxCos * ux), zi, fmaf(cosf(t_0), xi, (sinf(t_0) * yi)));
}
function code(xi, yi, zi, ux, uy, maxCos) t_0 = Float32(Float32(pi) * Float32(Float32(2.0) * uy)) return fma(Float32(maxCos * ux), zi, fma(cos(t_0), xi, Float32(sin(t_0) * yi))) end
\begin{array}{l}
\\
\begin{array}{l}
t_0 := \pi \cdot \left(2 \cdot uy\right)\\
\mathsf{fma}\left(maxCos \cdot ux, zi, \mathsf{fma}\left(\cos t\_0, xi, \sin t\_0 \cdot yi\right)\right)
\end{array}
\end{array}
Initial program 98.9%
Taylor expanded in ux around 0
associate-*r*N/A
lower-fma.f32N/A
lower-*.f32N/A
*-commutativeN/A
lower-fma.f32N/A
Applied rewrites95.9%
(FPCore (xi yi zi ux uy maxCos)
:precision binary32
(let* ((t_0 (* PI (* 2.0 uy))))
(if (<= uy 0.0002500000118743628)
(+
(+
(*
(*
(cos (* (* uy 2.0) PI))
(+
1.0
(* (* ux ux) (fma -0.5 (* maxCos maxCos) (* (* maxCos maxCos) ux)))))
xi)
(* (* 2.0 (* uy PI)) yi))
(* (* (* (- 1.0 ux) maxCos) ux) zi))
(fma (cos t_0) xi (* (sin t_0) yi)))))
float code(float xi, float yi, float zi, float ux, float uy, float maxCos) {
float t_0 = ((float) M_PI) * (2.0f * uy);
float tmp;
if (uy <= 0.0002500000118743628f) {
tmp = (((cosf(((uy * 2.0f) * ((float) M_PI))) * (1.0f + ((ux * ux) * fmaf(-0.5f, (maxCos * maxCos), ((maxCos * maxCos) * ux))))) * xi) + ((2.0f * (uy * ((float) M_PI))) * yi)) + ((((1.0f - ux) * maxCos) * ux) * zi);
} else {
tmp = fmaf(cosf(t_0), xi, (sinf(t_0) * yi));
}
return tmp;
}
function code(xi, yi, zi, ux, uy, maxCos) t_0 = Float32(Float32(pi) * Float32(Float32(2.0) * uy)) tmp = Float32(0.0) if (uy <= Float32(0.0002500000118743628)) tmp = Float32(Float32(Float32(Float32(cos(Float32(Float32(uy * Float32(2.0)) * Float32(pi))) * Float32(Float32(1.0) + Float32(Float32(ux * ux) * fma(Float32(-0.5), Float32(maxCos * maxCos), Float32(Float32(maxCos * maxCos) * ux))))) * xi) + Float32(Float32(Float32(2.0) * Float32(uy * Float32(pi))) * yi)) + Float32(Float32(Float32(Float32(Float32(1.0) - ux) * maxCos) * ux) * zi)); else tmp = fma(cos(t_0), xi, Float32(sin(t_0) * yi)); end return tmp end
\begin{array}{l}
\\
\begin{array}{l}
t_0 := \pi \cdot \left(2 \cdot uy\right)\\
\mathbf{if}\;uy \leq 0.0002500000118743628:\\
\;\;\;\;\left(\left(\cos \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \left(1 + \left(ux \cdot ux\right) \cdot \mathsf{fma}\left(-0.5, maxCos \cdot maxCos, \left(maxCos \cdot maxCos\right) \cdot ux\right)\right)\right) \cdot xi + \left(2 \cdot \left(uy \cdot \pi\right)\right) \cdot yi\right) + \left(\left(\left(1 - ux\right) \cdot maxCos\right) \cdot ux\right) \cdot zi\\
\mathbf{else}:\\
\;\;\;\;\mathsf{fma}\left(\cos t\_0, xi, \sin t\_0 \cdot yi\right)\\
\end{array}
\end{array}
if uy < 2.50000012e-4Initial program 99.3%
Taylor expanded in ux around 0
associate-*r*N/A
*-commutativeN/A
lift-*.f32N/A
lift-*.f32N/A
lift-PI.f32N/A
lift-sin.f3299.2
lift-PI.f32N/A
lift-*.f32N/A
*-commutativeN/A
lower-*.f32N/A
lift-PI.f3299.2
lift-*.f32N/A
*-commutativeN/A
lower-*.f3299.2
Applied rewrites99.2%
Taylor expanded in ux around 0
lower-+.f32N/A
lower-*.f32N/A
unpow2N/A
lower-*.f32N/A
lower-fma.f32N/A
unpow2N/A
lower-*.f32N/A
lower-*.f32N/A
unpow2N/A
lower-*.f3299.2
Applied rewrites99.2%
Taylor expanded in uy around 0
lift-*.f32N/A
lift-PI.f32N/A
lift-*.f3299.0
Applied rewrites99.0%
if 2.50000012e-4 < uy Initial program 98.3%
Taylor expanded in ux around 0
*-commutativeN/A
lower-fma.f32N/A
Applied rewrites90.9%
(FPCore (xi yi zi ux uy maxCos)
:precision binary32
(+
(+
(*
(*
(cos (* (* uy 2.0) PI))
(+
1.0
(* (* ux ux) (fma -0.5 (* maxCos maxCos) (* (* maxCos maxCos) ux)))))
xi)
(* (* 2.0 (* uy PI)) yi))
(* (* (* (- 1.0 ux) maxCos) ux) zi)))
float code(float xi, float yi, float zi, float ux, float uy, float maxCos) {
return (((cosf(((uy * 2.0f) * ((float) M_PI))) * (1.0f + ((ux * ux) * fmaf(-0.5f, (maxCos * maxCos), ((maxCos * maxCos) * ux))))) * xi) + ((2.0f * (uy * ((float) M_PI))) * yi)) + ((((1.0f - ux) * maxCos) * ux) * zi);
}
function code(xi, yi, zi, ux, uy, maxCos) return Float32(Float32(Float32(Float32(cos(Float32(Float32(uy * Float32(2.0)) * Float32(pi))) * Float32(Float32(1.0) + Float32(Float32(ux * ux) * fma(Float32(-0.5), Float32(maxCos * maxCos), Float32(Float32(maxCos * maxCos) * ux))))) * xi) + Float32(Float32(Float32(2.0) * Float32(uy * Float32(pi))) * yi)) + Float32(Float32(Float32(Float32(Float32(1.0) - ux) * maxCos) * ux) * zi)) end
\begin{array}{l}
\\
\left(\left(\cos \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \left(1 + \left(ux \cdot ux\right) \cdot \mathsf{fma}\left(-0.5, maxCos \cdot maxCos, \left(maxCos \cdot maxCos\right) \cdot ux\right)\right)\right) \cdot xi + \left(2 \cdot \left(uy \cdot \pi\right)\right) \cdot yi\right) + \left(\left(\left(1 - ux\right) \cdot maxCos\right) \cdot ux\right) \cdot zi
\end{array}
Initial program 98.9%
Taylor expanded in ux around 0
associate-*r*N/A
*-commutativeN/A
lift-*.f32N/A
lift-*.f32N/A
lift-PI.f32N/A
lift-sin.f3298.8
lift-PI.f32N/A
lift-*.f32N/A
*-commutativeN/A
lower-*.f32N/A
lift-PI.f3298.8
lift-*.f32N/A
*-commutativeN/A
lower-*.f3298.8
Applied rewrites98.8%
Taylor expanded in ux around 0
lower-+.f32N/A
lower-*.f32N/A
unpow2N/A
lower-*.f32N/A
lower-fma.f32N/A
unpow2N/A
lower-*.f32N/A
lower-*.f32N/A
unpow2N/A
lower-*.f3298.8
Applied rewrites98.8%
Taylor expanded in uy around 0
lift-*.f32N/A
lift-PI.f32N/A
lift-*.f3289.4
Applied rewrites89.4%
(FPCore (xi yi zi ux uy maxCos)
:precision binary32
(+
(+
(*
(*
1.0
(+
1.0
(* (* ux ux) (fma -0.5 (* maxCos maxCos) (* (* maxCos maxCos) ux)))))
xi)
(* (sin (* PI (* 2.0 uy))) yi))
(* (* (* (- 1.0 ux) maxCos) ux) zi)))
float code(float xi, float yi, float zi, float ux, float uy, float maxCos) {
return (((1.0f * (1.0f + ((ux * ux) * fmaf(-0.5f, (maxCos * maxCos), ((maxCos * maxCos) * ux))))) * xi) + (sinf((((float) M_PI) * (2.0f * uy))) * yi)) + ((((1.0f - ux) * maxCos) * ux) * zi);
}
function code(xi, yi, zi, ux, uy, maxCos) return Float32(Float32(Float32(Float32(Float32(1.0) * Float32(Float32(1.0) + Float32(Float32(ux * ux) * fma(Float32(-0.5), Float32(maxCos * maxCos), Float32(Float32(maxCos * maxCos) * ux))))) * xi) + Float32(sin(Float32(Float32(pi) * Float32(Float32(2.0) * uy))) * yi)) + Float32(Float32(Float32(Float32(Float32(1.0) - ux) * maxCos) * ux) * zi)) end
\begin{array}{l}
\\
\left(\left(1 \cdot \left(1 + \left(ux \cdot ux\right) \cdot \mathsf{fma}\left(-0.5, maxCos \cdot maxCos, \left(maxCos \cdot maxCos\right) \cdot ux\right)\right)\right) \cdot xi + \sin \left(\pi \cdot \left(2 \cdot uy\right)\right) \cdot yi\right) + \left(\left(\left(1 - ux\right) \cdot maxCos\right) \cdot ux\right) \cdot zi
\end{array}
Initial program 98.9%
Taylor expanded in ux around 0
associate-*r*N/A
*-commutativeN/A
lift-*.f32N/A
lift-*.f32N/A
lift-PI.f32N/A
lift-sin.f3298.8
lift-PI.f32N/A
lift-*.f32N/A
*-commutativeN/A
lower-*.f32N/A
lift-PI.f3298.8
lift-*.f32N/A
*-commutativeN/A
lower-*.f3298.8
Applied rewrites98.8%
Taylor expanded in ux around 0
lower-+.f32N/A
lower-*.f32N/A
unpow2N/A
lower-*.f32N/A
lower-fma.f32N/A
unpow2N/A
lower-*.f32N/A
lower-*.f32N/A
unpow2N/A
lower-*.f3298.8
Applied rewrites98.8%
Taylor expanded in uy around 0
Applied rewrites88.0%
(FPCore (xi yi zi ux uy maxCos) :precision binary32 (let* ((t_0 (* (* (- 1.0 ux) maxCos) ux)) (t_1 (sqrt (- 1.0 (* t_0 t_0))))) (+ (+ (* (* 1.0 t_1) xi) (* (* (* PI (* 2.0 uy)) t_1) yi)) (* t_0 zi))))
float code(float xi, float yi, float zi, float ux, float uy, float maxCos) {
float t_0 = ((1.0f - ux) * maxCos) * ux;
float t_1 = sqrtf((1.0f - (t_0 * t_0)));
return (((1.0f * t_1) * xi) + (((((float) M_PI) * (2.0f * uy)) * t_1) * yi)) + (t_0 * zi);
}
function code(xi, yi, zi, ux, uy, maxCos) t_0 = Float32(Float32(Float32(Float32(1.0) - ux) * maxCos) * ux) t_1 = sqrt(Float32(Float32(1.0) - Float32(t_0 * t_0))) return Float32(Float32(Float32(Float32(Float32(1.0) * t_1) * xi) + Float32(Float32(Float32(Float32(pi) * Float32(Float32(2.0) * uy)) * t_1) * yi)) + Float32(t_0 * zi)) end
function tmp = code(xi, yi, zi, ux, uy, maxCos) t_0 = ((single(1.0) - ux) * maxCos) * ux; t_1 = sqrt((single(1.0) - (t_0 * t_0))); tmp = (((single(1.0) * t_1) * xi) + (((single(pi) * (single(2.0) * uy)) * t_1) * yi)) + (t_0 * zi); end
\begin{array}{l}
\\
\begin{array}{l}
t_0 := \left(\left(1 - ux\right) \cdot maxCos\right) \cdot ux\\
t_1 := \sqrt{1 - t\_0 \cdot t\_0}\\
\left(\left(1 \cdot t\_1\right) \cdot xi + \left(\left(\pi \cdot \left(2 \cdot uy\right)\right) \cdot t\_1\right) \cdot yi\right) + t\_0 \cdot zi
\end{array}
\end{array}
Initial program 98.9%
Taylor expanded in uy around 0
associate-*r*N/A
*-commutativeN/A
lift-*.f32N/A
*-commutativeN/A
lower-*.f32N/A
lift-PI.f3289.5
lift-*.f32N/A
*-commutativeN/A
lower-*.f3289.5
Applied rewrites89.5%
Taylor expanded in uy around 0
Applied rewrites80.8%
(FPCore (xi yi zi ux uy maxCos) :precision binary32 (let* ((t_0 (* maxCos (* ux (- 1.0 ux))))) (fma maxCos (* ux (* zi (- 1.0 ux))) (* xi (sqrt (- 1.0 (* t_0 t_0)))))))
float code(float xi, float yi, float zi, float ux, float uy, float maxCos) {
float t_0 = maxCos * (ux * (1.0f - ux));
return fmaf(maxCos, (ux * (zi * (1.0f - ux))), (xi * sqrtf((1.0f - (t_0 * t_0)))));
}
function code(xi, yi, zi, ux, uy, maxCos) t_0 = Float32(maxCos * Float32(ux * Float32(Float32(1.0) - ux))) return fma(maxCos, Float32(ux * Float32(zi * Float32(Float32(1.0) - ux))), Float32(xi * sqrt(Float32(Float32(1.0) - Float32(t_0 * t_0))))) end
\begin{array}{l}
\\
\begin{array}{l}
t_0 := maxCos \cdot \left(ux \cdot \left(1 - ux\right)\right)\\
\mathsf{fma}\left(maxCos, ux \cdot \left(zi \cdot \left(1 - ux\right)\right), xi \cdot \sqrt{1 - t\_0 \cdot t\_0}\right)
\end{array}
\end{array}
Initial program 98.9%
Applied rewrites99.0%
Taylor expanded in uy around 0
lower-fma.f32N/A
lower-*.f32N/A
lower-*.f32N/A
lift--.f32N/A
lower-*.f32N/A
lower-sin.f32N/A
lower-acos.f32N/A
lower-*.f32N/A
lower-*.f32N/A
lift--.f3250.7
Applied rewrites50.7%
lift-sin.f32N/A
lift-acos.f32N/A
lift-*.f32N/A
lift--.f32N/A
lift-*.f32N/A
sin-acosN/A
lower-sqrt.f32N/A
lower--.f32N/A
lower-*.f32N/A
lift-*.f32N/A
lift--.f32N/A
lift-*.f32N/A
lift-*.f32N/A
lift--.f32N/A
lift-*.f3250.7
Applied rewrites50.7%
(FPCore (xi yi zi ux uy maxCos) :precision binary32 (let* ((t_0 (* maxCos (* ux (- 1.0 ux))))) (* xi (sqrt (- 1.0 (* t_0 t_0))))))
float code(float xi, float yi, float zi, float ux, float uy, float maxCos) {
float t_0 = maxCos * (ux * (1.0f - ux));
return xi * sqrtf((1.0f - (t_0 * t_0)));
}
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(xi, yi, zi, ux, uy, maxcos)
use fmin_fmax_functions
real(4), intent (in) :: xi
real(4), intent (in) :: yi
real(4), intent (in) :: zi
real(4), intent (in) :: ux
real(4), intent (in) :: uy
real(4), intent (in) :: maxcos
real(4) :: t_0
t_0 = maxcos * (ux * (1.0e0 - ux))
code = xi * sqrt((1.0e0 - (t_0 * t_0)))
end function
function code(xi, yi, zi, ux, uy, maxCos) t_0 = Float32(maxCos * Float32(ux * Float32(Float32(1.0) - ux))) return Float32(xi * sqrt(Float32(Float32(1.0) - Float32(t_0 * t_0)))) end
function tmp = code(xi, yi, zi, ux, uy, maxCos) t_0 = maxCos * (ux * (single(1.0) - ux)); tmp = xi * sqrt((single(1.0) - (t_0 * t_0))); end
\begin{array}{l}
\\
\begin{array}{l}
t_0 := maxCos \cdot \left(ux \cdot \left(1 - ux\right)\right)\\
xi \cdot \sqrt{1 - t\_0 \cdot t\_0}
\end{array}
\end{array}
Initial program 98.9%
Applied rewrites99.0%
Taylor expanded in uy around 0
lower-fma.f32N/A
lower-*.f32N/A
lower-*.f32N/A
lift--.f32N/A
lower-*.f32N/A
lower-sin.f32N/A
lower-acos.f32N/A
lower-*.f32N/A
lower-*.f32N/A
lift--.f3250.7
Applied rewrites50.7%
Taylor expanded in xi around inf
lift-*.f32N/A
lift--.f32N/A
lift-*.f32N/A
lift-acos.f32N/A
lift-sin.f32N/A
lift-*.f3244.6
Applied rewrites44.6%
lift-sin.f32N/A
lift-acos.f32N/A
lift-*.f32N/A
lift--.f32N/A
lift-*.f32N/A
sin-acosN/A
lower-sqrt.f32N/A
lower--.f32N/A
lower-*.f32N/A
lift-*.f32N/A
lift--.f32N/A
lift-*.f32N/A
lift-*.f32N/A
lift--.f32N/A
lift-*.f3244.6
Applied rewrites44.6%
(FPCore (xi yi zi ux uy maxCos) :precision binary32 (* (* (* (- 1.0 ux) zi) ux) maxCos))
float code(float xi, float yi, float zi, float ux, float uy, float maxCos) {
return (((1.0f - ux) * zi) * ux) * 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(xi, yi, zi, ux, uy, maxcos)
use fmin_fmax_functions
real(4), intent (in) :: xi
real(4), intent (in) :: yi
real(4), intent (in) :: zi
real(4), intent (in) :: ux
real(4), intent (in) :: uy
real(4), intent (in) :: maxcos
code = (((1.0e0 - ux) * zi) * ux) * maxcos
end function
function code(xi, yi, zi, ux, uy, maxCos) return Float32(Float32(Float32(Float32(Float32(1.0) - ux) * zi) * ux) * maxCos) end
function tmp = code(xi, yi, zi, ux, uy, maxCos) tmp = (((single(1.0) - ux) * zi) * ux) * maxCos; end
\begin{array}{l}
\\
\left(\left(\left(1 - ux\right) \cdot zi\right) \cdot ux\right) \cdot maxCos
\end{array}
Initial program 98.9%
Taylor expanded in zi around inf
*-commutativeN/A
lower-*.f32N/A
*-commutativeN/A
lower-*.f32N/A
*-commutativeN/A
lower-*.f32N/A
lift--.f3213.4
Applied rewrites13.4%
(FPCore (xi yi zi ux uy maxCos) :precision binary32 (* (* zi ux) maxCos))
float code(float xi, float yi, float zi, float ux, float uy, float maxCos) {
return (zi * ux) * 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(xi, yi, zi, ux, uy, maxcos)
use fmin_fmax_functions
real(4), intent (in) :: xi
real(4), intent (in) :: yi
real(4), intent (in) :: zi
real(4), intent (in) :: ux
real(4), intent (in) :: uy
real(4), intent (in) :: maxcos
code = (zi * ux) * maxcos
end function
function code(xi, yi, zi, ux, uy, maxCos) return Float32(Float32(zi * ux) * maxCos) end
function tmp = code(xi, yi, zi, ux, uy, maxCos) tmp = (zi * ux) * maxCos; end
\begin{array}{l}
\\
\left(zi \cdot ux\right) \cdot maxCos
\end{array}
Initial program 98.9%
Taylor expanded in zi around inf
*-commutativeN/A
lower-*.f32N/A
*-commutativeN/A
lower-*.f32N/A
*-commutativeN/A
lower-*.f32N/A
lift--.f3213.4
Applied rewrites13.4%
Taylor expanded in ux around 0
Applied rewrites11.9%
herbie shell --seed 2025095
(FPCore (xi yi zi ux uy maxCos)
:name "UniformSampleCone 2"
:precision binary32
:pre (and (and (and (and (and (and (<= -10000.0 xi) (<= xi 10000.0)) (and (<= -10000.0 yi) (<= yi 10000.0))) (and (<= -10000.0 zi) (<= zi 10000.0))) (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) maxCos) ux) (* (* (- 1.0 ux) maxCos) ux))))) xi) (* (* (sin (* (* uy 2.0) PI)) (sqrt (- 1.0 (* (* (* (- 1.0 ux) maxCos) ux) (* (* (- 1.0 ux) maxCos) ux))))) yi)) (* (* (* (- 1.0 ux) maxCos) ux) zi)))