
(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}
Sampling outcomes in binary32 precision:
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 (sqrt (- 1.0 (* t_0 t_0)))))
(+
(+
(* (* (cos (* (* uy 2.0) PI)) t_1) xi)
(* (* (* 2.0 (* (sin (* PI uy)) (cos (* PI 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 (((cosf(((uy * 2.0f) * ((float) M_PI))) * t_1) * xi) + (((2.0f * (sinf((((float) M_PI) * uy)) * cosf((((float) M_PI) * 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(cos(Float32(Float32(uy * Float32(2.0)) * Float32(pi))) * t_1) * xi) + Float32(Float32(Float32(Float32(2.0) * Float32(sin(Float32(Float32(pi) * uy)) * cos(Float32(Float32(pi) * 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 = (((cos(((uy * single(2.0)) * single(pi))) * t_1) * xi) + (((single(2.0) * (sin((single(pi) * uy)) * cos((single(pi) * 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(\cos \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot t\_1\right) \cdot xi + \left(\left(2 \cdot \left(\sin \left(\pi \cdot uy\right) \cdot \cos \left(\pi \cdot uy\right)\right)\right) \cdot t\_1\right) \cdot yi\right) + t\_0 \cdot zi
\end{array}
\end{array}
Initial program 98.7%
lift-sin.f32N/A
lift-PI.f32N/A
lift-*.f32N/A
lift-*.f32N/A
*-commutativeN/A
associate-*r*N/A
sin-2N/A
lower-*.f32N/A
lower-*.f32N/A
lower-sin.f32N/A
*-commutativeN/A
lower-*.f32N/A
lift-PI.f32N/A
lower-cos.f32N/A
*-commutativeN/A
lower-*.f32N/A
lift-PI.f3298.8
Applied rewrites98.8%
(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)))))
(+
(+
(* (* (cos (* (+ uy uy) PI)) t_1) xi)
(* (* (sin (* (* uy 2.0) PI)) 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 (((cosf(((uy + uy) * ((float) M_PI))) * t_1) * xi) + ((sinf(((uy * 2.0f) * ((float) M_PI))) * 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(cos(Float32(Float32(uy + uy) * Float32(pi))) * t_1) * xi) + Float32(Float32(sin(Float32(Float32(uy * Float32(2.0)) * Float32(pi))) * 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 = (((cos(((uy + uy) * single(pi))) * t_1) * xi) + ((sin(((uy * single(2.0)) * single(pi))) * 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(\cos \left(\left(uy + uy\right) \cdot \pi\right) \cdot t\_1\right) \cdot xi + \left(\sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot t\_1\right) \cdot yi\right) + t\_0 \cdot zi
\end{array}
\end{array}
Initial program 98.7%
lift-*.f32N/A
*-commutativeN/A
count-2-revN/A
lower-+.f3298.7
Applied rewrites98.7%
(FPCore (xi yi zi ux uy maxCos)
:precision binary32
(let* ((t_0 (* (* (- 1.0 ux) maxCos) ux)) (t_1 (* (* uy 2.0) PI)))
(+
(+
(* (* (cos t_1) (sqrt (- 1.0 (* t_0 t_0)))) xi)
(*
(*
(sin t_1)
(fma
(fma -0.5 (* maxCos maxCos) (* (* maxCos maxCos) ux))
(* ux ux)
1.0))
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 = (uy * 2.0f) * ((float) M_PI);
return (((cosf(t_1) * sqrtf((1.0f - (t_0 * t_0)))) * xi) + ((sinf(t_1) * fmaf(fmaf(-0.5f, (maxCos * maxCos), ((maxCos * maxCos) * ux)), (ux * ux), 1.0f)) * 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(uy * Float32(2.0)) * Float32(pi)) return Float32(Float32(Float32(Float32(cos(t_1) * sqrt(Float32(Float32(1.0) - Float32(t_0 * t_0)))) * xi) + Float32(Float32(sin(t_1) * fma(fma(Float32(-0.5), Float32(maxCos * maxCos), Float32(Float32(maxCos * maxCos) * ux)), Float32(ux * ux), Float32(1.0))) * 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 := \left(uy \cdot 2\right) \cdot \pi\\
\left(\left(\cos t\_1 \cdot \sqrt{1 - t\_0 \cdot t\_0}\right) \cdot xi + \left(\sin t\_1 \cdot \mathsf{fma}\left(\mathsf{fma}\left(-0.5, maxCos \cdot maxCos, \left(maxCos \cdot maxCos\right) \cdot ux\right), ux \cdot ux, 1\right)\right) \cdot yi\right) + t\_0 \cdot zi
\end{array}
\end{array}
Initial program 98.7%
Taylor expanded in ux around 0
+-commutativeN/A
*-commutativeN/A
lower-fma.f32N/A
lower-fma.f32N/A
unpow2N/A
lower-*.f32N/A
lower-*.f32N/A
unpow2N/A
lower-*.f32N/A
unpow2N/A
lower-*.f3298.7
Applied rewrites98.7%
(FPCore (xi yi zi ux uy maxCos)
:precision binary32
(let* ((t_0 (* (* (- 1.0 ux) maxCos) ux)))
(+
(+
(* (* (cos (* (+ uy uy) PI)) (sqrt (- 1.0 (* t_0 t_0)))) xi)
(* (sin (* 2.0 (* uy PI))) 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;
return (((cosf(((uy + uy) * ((float) M_PI))) * sqrtf((1.0f - (t_0 * t_0)))) * xi) + (sinf((2.0f * (uy * ((float) M_PI)))) * yi)) + (t_0 * zi);
}
function code(xi, yi, zi, ux, uy, maxCos) t_0 = Float32(Float32(Float32(Float32(1.0) - ux) * maxCos) * ux) return Float32(Float32(Float32(Float32(cos(Float32(Float32(uy + uy) * Float32(pi))) * sqrt(Float32(Float32(1.0) - Float32(t_0 * t_0)))) * xi) + Float32(sin(Float32(Float32(2.0) * Float32(uy * Float32(pi)))) * yi)) + Float32(t_0 * zi)) end
function tmp = code(xi, yi, zi, ux, uy, maxCos) t_0 = ((single(1.0) - ux) * maxCos) * ux; tmp = (((cos(((uy + uy) * single(pi))) * sqrt((single(1.0) - (t_0 * t_0)))) * xi) + (sin((single(2.0) * (uy * single(pi)))) * yi)) + (t_0 * zi); end
\begin{array}{l}
\\
\begin{array}{l}
t_0 := \left(\left(1 - ux\right) \cdot maxCos\right) \cdot ux\\
\left(\left(\cos \left(\left(uy + uy\right) \cdot \pi\right) \cdot \sqrt{1 - t\_0 \cdot t\_0}\right) \cdot xi + \sin \left(2 \cdot \left(uy \cdot \pi\right)\right) \cdot yi\right) + t\_0 \cdot zi
\end{array}
\end{array}
Initial program 98.7%
lift-*.f32N/A
*-commutativeN/A
count-2-revN/A
lower-+.f3298.7
Applied rewrites98.7%
Taylor expanded in ux around 0
lower-sin.f32N/A
lift-*.f32N/A
lift-PI.f32N/A
lift-*.f3298.6
Applied rewrites98.6%
(FPCore (xi yi zi ux uy maxCos)
:precision binary32
(let* ((t_0 (* PI (* 2.0 uy))) (t_1 (* (* (- 1.0 ux) maxCos) ux)))
(if (<= t_1 1.0000000116860974e-7)
(fma (cos t_0) xi (* (sin t_0) yi))
(+
(* (sqrt (- 1.0 (* (pow (* (- 1.0 ux) ux) 2.0) (* maxCos maxCos)))) xi)
(* t_1 zi)))))
float code(float xi, float yi, float zi, float ux, float uy, float maxCos) {
float t_0 = ((float) M_PI) * (2.0f * uy);
float t_1 = ((1.0f - ux) * maxCos) * ux;
float tmp;
if (t_1 <= 1.0000000116860974e-7f) {
tmp = fmaf(cosf(t_0), xi, (sinf(t_0) * yi));
} else {
tmp = (sqrtf((1.0f - (powf(((1.0f - ux) * ux), 2.0f) * (maxCos * maxCos)))) * xi) + (t_1 * zi);
}
return tmp;
}
function code(xi, yi, zi, ux, uy, maxCos) t_0 = Float32(Float32(pi) * Float32(Float32(2.0) * uy)) t_1 = Float32(Float32(Float32(Float32(1.0) - ux) * maxCos) * ux) tmp = Float32(0.0) if (t_1 <= Float32(1.0000000116860974e-7)) tmp = fma(cos(t_0), xi, Float32(sin(t_0) * yi)); else tmp = Float32(Float32(sqrt(Float32(Float32(1.0) - Float32((Float32(Float32(Float32(1.0) - ux) * ux) ^ Float32(2.0)) * Float32(maxCos * maxCos)))) * xi) + Float32(t_1 * zi)); end return tmp end
\begin{array}{l}
\\
\begin{array}{l}
t_0 := \pi \cdot \left(2 \cdot uy\right)\\
t_1 := \left(\left(1 - ux\right) \cdot maxCos\right) \cdot ux\\
\mathbf{if}\;t\_1 \leq 1.0000000116860974 \cdot 10^{-7}:\\
\;\;\;\;\mathsf{fma}\left(\cos t\_0, xi, \sin t\_0 \cdot yi\right)\\
\mathbf{else}:\\
\;\;\;\;\sqrt{1 - {\left(\left(1 - ux\right) \cdot ux\right)}^{2} \cdot \left(maxCos \cdot maxCos\right)} \cdot xi + t\_1 \cdot zi\\
\end{array}
\end{array}
if (*.f32 (*.f32 (-.f32 #s(literal 1 binary32) ux) maxCos) ux) < 1.00000001e-7Initial program 98.7%
Taylor expanded in ux around 0
*-commutativeN/A
lower-fma.f32N/A
Applied rewrites93.5%
if 1.00000001e-7 < (*.f32 (*.f32 (-.f32 #s(literal 1 binary32) ux) maxCos) ux) Initial program 99.4%
Taylor expanded in uy around 0
*-commutativeN/A
lower-*.f32N/A
Applied rewrites86.0%
(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 98.7%
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 rewrites98.5%
(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 98.7%
Taylor expanded in ux around 0
associate-*r*N/A
lower-fma.f32N/A
lower-*.f32N/A
*-commutativeN/A
lower-fma.f32N/A
Applied rewrites94.9%
(FPCore (xi yi zi ux uy maxCos)
:precision binary32
(let* ((t_0 (* (* (- 1.0 ux) maxCos) ux)))
(if (<= t_0 1.0000000116860974e-7)
(fma 2.0 (* yi (* uy PI)) (* xi (cos (* 2.0 (* uy PI)))))
(+
(* (sqrt (- 1.0 (* (pow (* (- 1.0 ux) ux) 2.0) (* maxCos maxCos)))) xi)
(* 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 tmp;
if (t_0 <= 1.0000000116860974e-7f) {
tmp = fmaf(2.0f, (yi * (uy * ((float) M_PI))), (xi * cosf((2.0f * (uy * ((float) M_PI))))));
} else {
tmp = (sqrtf((1.0f - (powf(((1.0f - ux) * ux), 2.0f) * (maxCos * maxCos)))) * xi) + (t_0 * zi);
}
return tmp;
}
function code(xi, yi, zi, ux, uy, maxCos) t_0 = Float32(Float32(Float32(Float32(1.0) - ux) * maxCos) * ux) tmp = Float32(0.0) if (t_0 <= Float32(1.0000000116860974e-7)) tmp = fma(Float32(2.0), Float32(yi * Float32(uy * Float32(pi))), Float32(xi * cos(Float32(Float32(2.0) * Float32(uy * Float32(pi)))))); else tmp = Float32(Float32(sqrt(Float32(Float32(1.0) - Float32((Float32(Float32(Float32(1.0) - ux) * ux) ^ Float32(2.0)) * Float32(maxCos * maxCos)))) * xi) + Float32(t_0 * zi)); end return tmp end
\begin{array}{l}
\\
\begin{array}{l}
t_0 := \left(\left(1 - ux\right) \cdot maxCos\right) \cdot ux\\
\mathbf{if}\;t\_0 \leq 1.0000000116860974 \cdot 10^{-7}:\\
\;\;\;\;\mathsf{fma}\left(2, yi \cdot \left(uy \cdot \pi\right), xi \cdot \cos \left(2 \cdot \left(uy \cdot \pi\right)\right)\right)\\
\mathbf{else}:\\
\;\;\;\;\sqrt{1 - {\left(\left(1 - ux\right) \cdot ux\right)}^{2} \cdot \left(maxCos \cdot maxCos\right)} \cdot xi + t\_0 \cdot zi\\
\end{array}
\end{array}
if (*.f32 (*.f32 (-.f32 #s(literal 1 binary32) ux) maxCos) ux) < 1.00000001e-7Initial program 98.7%
lift-sin.f32N/A
lift-PI.f32N/A
lift-*.f32N/A
lift-*.f32N/A
*-commutativeN/A
associate-*r*N/A
sin-2N/A
lower-*.f32N/A
lower-*.f32N/A
lower-sin.f32N/A
*-commutativeN/A
lower-*.f32N/A
lift-PI.f32N/A
lower-cos.f32N/A
*-commutativeN/A
lower-*.f32N/A
lift-PI.f3298.7
Applied rewrites98.7%
Taylor expanded in ux around 0
lower-fma.f32N/A
Applied rewrites93.6%
Taylor expanded in uy around 0
lift-*.f32N/A
lift-PI.f3283.9
Applied rewrites83.9%
if 1.00000001e-7 < (*.f32 (*.f32 (-.f32 #s(literal 1 binary32) ux) maxCos) ux) Initial program 99.4%
Taylor expanded in uy around 0
*-commutativeN/A
lower-*.f32N/A
Applied rewrites86.0%
(FPCore (xi yi zi ux uy maxCos) :precision binary32 (if (<= (* (* (- 1.0 ux) maxCos) ux) 1.0000000116860974e-7) (fma 2.0 (* yi (* uy PI)) (* xi (cos (* 2.0 (* uy PI))))) (fma maxCos (* ux (* zi (- 1.0 ux))) xi)))
float code(float xi, float yi, float zi, float ux, float uy, float maxCos) {
float tmp;
if ((((1.0f - ux) * maxCos) * ux) <= 1.0000000116860974e-7f) {
tmp = fmaf(2.0f, (yi * (uy * ((float) M_PI))), (xi * cosf((2.0f * (uy * ((float) M_PI))))));
} else {
tmp = fmaf(maxCos, (ux * (zi * (1.0f - ux))), xi);
}
return tmp;
}
function code(xi, yi, zi, ux, uy, maxCos) tmp = Float32(0.0) if (Float32(Float32(Float32(Float32(1.0) - ux) * maxCos) * ux) <= Float32(1.0000000116860974e-7)) tmp = fma(Float32(2.0), Float32(yi * Float32(uy * Float32(pi))), Float32(xi * cos(Float32(Float32(2.0) * Float32(uy * Float32(pi)))))); else tmp = fma(maxCos, Float32(ux * Float32(zi * Float32(Float32(1.0) - ux))), xi); end return tmp end
\begin{array}{l}
\\
\begin{array}{l}
\mathbf{if}\;\left(\left(1 - ux\right) \cdot maxCos\right) \cdot ux \leq 1.0000000116860974 \cdot 10^{-7}:\\
\;\;\;\;\mathsf{fma}\left(2, yi \cdot \left(uy \cdot \pi\right), xi \cdot \cos \left(2 \cdot \left(uy \cdot \pi\right)\right)\right)\\
\mathbf{else}:\\
\;\;\;\;\mathsf{fma}\left(maxCos, ux \cdot \left(zi \cdot \left(1 - ux\right)\right), xi\right)\\
\end{array}
\end{array}
if (*.f32 (*.f32 (-.f32 #s(literal 1 binary32) ux) maxCos) ux) < 1.00000001e-7Initial program 98.7%
lift-sin.f32N/A
lift-PI.f32N/A
lift-*.f32N/A
lift-*.f32N/A
*-commutativeN/A
associate-*r*N/A
sin-2N/A
lower-*.f32N/A
lower-*.f32N/A
lower-sin.f32N/A
*-commutativeN/A
lower-*.f32N/A
lift-PI.f32N/A
lower-cos.f32N/A
*-commutativeN/A
lower-*.f32N/A
lift-PI.f3298.7
Applied rewrites98.7%
Taylor expanded in ux around 0
lower-fma.f32N/A
Applied rewrites93.6%
Taylor expanded in uy around 0
lift-*.f32N/A
lift-PI.f3283.9
Applied rewrites83.9%
if 1.00000001e-7 < (*.f32 (*.f32 (-.f32 #s(literal 1 binary32) ux) maxCos) ux) Initial program 99.4%
lift-sin.f32N/A
lift-PI.f32N/A
lift-*.f32N/A
lift-*.f32N/A
*-commutativeN/A
associate-*r*N/A
sin-2N/A
lower-*.f32N/A
lower-*.f32N/A
lower-sin.f32N/A
*-commutativeN/A
lower-*.f32N/A
lift-PI.f32N/A
lower-cos.f32N/A
*-commutativeN/A
lower-*.f32N/A
lift-PI.f3299.4
Applied rewrites99.4%
Taylor expanded in uy around 0
lower-fma.f32N/A
lower-*.f32N/A
lower-*.f32N/A
lift--.f32N/A
lower-*.f32N/A
lower-sqrt.f32N/A
lower--.f32N/A
Applied rewrites85.6%
Taylor expanded in ux around 0
Applied rewrites85.7%
(FPCore (xi yi zi ux uy maxCos) :precision binary32 (if (<= uy 0.052000001072883606) (+ xi (* uy (fma -2.0 (* uy (* xi (* PI PI))) (* 2.0 (* yi PI))))) (* xi (cos (* 2.0 (* uy PI))))))
float code(float xi, float yi, float zi, float ux, float uy, float maxCos) {
float tmp;
if (uy <= 0.052000001072883606f) {
tmp = xi + (uy * fmaf(-2.0f, (uy * (xi * (((float) M_PI) * ((float) M_PI)))), (2.0f * (yi * ((float) M_PI)))));
} else {
tmp = xi * cosf((2.0f * (uy * ((float) M_PI))));
}
return tmp;
}
function code(xi, yi, zi, ux, uy, maxCos) tmp = Float32(0.0) if (uy <= Float32(0.052000001072883606)) tmp = Float32(xi + Float32(uy * fma(Float32(-2.0), Float32(uy * Float32(xi * Float32(Float32(pi) * Float32(pi)))), Float32(Float32(2.0) * Float32(yi * Float32(pi)))))); else tmp = Float32(xi * cos(Float32(Float32(2.0) * Float32(uy * Float32(pi))))); end return tmp end
\begin{array}{l}
\\
\begin{array}{l}
\mathbf{if}\;uy \leq 0.052000001072883606:\\
\;\;\;\;xi + uy \cdot \mathsf{fma}\left(-2, uy \cdot \left(xi \cdot \left(\pi \cdot \pi\right)\right), 2 \cdot \left(yi \cdot \pi\right)\right)\\
\mathbf{else}:\\
\;\;\;\;xi \cdot \cos \left(2 \cdot \left(uy \cdot \pi\right)\right)\\
\end{array}
\end{array}
if uy < 0.0520000011Initial program 99.1%
lift-sin.f32N/A
lift-PI.f32N/A
lift-*.f32N/A
lift-*.f32N/A
*-commutativeN/A
associate-*r*N/A
sin-2N/A
lower-*.f32N/A
lower-*.f32N/A
lower-sin.f32N/A
*-commutativeN/A
lower-*.f32N/A
lift-PI.f32N/A
lower-cos.f32N/A
*-commutativeN/A
lower-*.f32N/A
lift-PI.f3299.2
Applied rewrites99.2%
Taylor expanded in ux around 0
lower-fma.f32N/A
Applied rewrites88.7%
Taylor expanded in uy around 0
lower-+.f32N/A
lower-*.f32N/A
lower-fma.f32N/A
lower-*.f32N/A
pow2N/A
lift-*.f32N/A
lift-PI.f32N/A
lift-PI.f32N/A
lift-*.f32N/A
lower-*.f32N/A
lower-*.f32N/A
lift-PI.f3284.0
Applied rewrites84.0%
if 0.0520000011 < uy Initial program 96.7%
Taylor expanded in xi around inf
lower-*.f32N/A
Applied rewrites54.2%
Taylor expanded in ux around 0
lower-*.f32N/A
lower-cos.f32N/A
lower-*.f32N/A
lower-*.f32N/A
lift-PI.f3254.2
Applied rewrites54.2%
(FPCore (xi yi zi ux uy maxCos) :precision binary32 (if (<= (* (* (- 1.0 ux) maxCos) ux) 1.0000000116860974e-7) (+ xi (* uy (fma -2.0 (* uy (* xi (* PI PI))) (* 2.0 (* yi PI))))) (fma maxCos (* ux (* zi (- 1.0 ux))) xi)))
float code(float xi, float yi, float zi, float ux, float uy, float maxCos) {
float tmp;
if ((((1.0f - ux) * maxCos) * ux) <= 1.0000000116860974e-7f) {
tmp = xi + (uy * fmaf(-2.0f, (uy * (xi * (((float) M_PI) * ((float) M_PI)))), (2.0f * (yi * ((float) M_PI)))));
} else {
tmp = fmaf(maxCos, (ux * (zi * (1.0f - ux))), xi);
}
return tmp;
}
function code(xi, yi, zi, ux, uy, maxCos) tmp = Float32(0.0) if (Float32(Float32(Float32(Float32(1.0) - ux) * maxCos) * ux) <= Float32(1.0000000116860974e-7)) tmp = Float32(xi + Float32(uy * fma(Float32(-2.0), Float32(uy * Float32(xi * Float32(Float32(pi) * Float32(pi)))), Float32(Float32(2.0) * Float32(yi * Float32(pi)))))); else tmp = fma(maxCos, Float32(ux * Float32(zi * Float32(Float32(1.0) - ux))), xi); end return tmp end
\begin{array}{l}
\\
\begin{array}{l}
\mathbf{if}\;\left(\left(1 - ux\right) \cdot maxCos\right) \cdot ux \leq 1.0000000116860974 \cdot 10^{-7}:\\
\;\;\;\;xi + uy \cdot \mathsf{fma}\left(-2, uy \cdot \left(xi \cdot \left(\pi \cdot \pi\right)\right), 2 \cdot \left(yi \cdot \pi\right)\right)\\
\mathbf{else}:\\
\;\;\;\;\mathsf{fma}\left(maxCos, ux \cdot \left(zi \cdot \left(1 - ux\right)\right), xi\right)\\
\end{array}
\end{array}
if (*.f32 (*.f32 (-.f32 #s(literal 1 binary32) ux) maxCos) ux) < 1.00000001e-7Initial program 98.7%
lift-sin.f32N/A
lift-PI.f32N/A
lift-*.f32N/A
lift-*.f32N/A
*-commutativeN/A
associate-*r*N/A
sin-2N/A
lower-*.f32N/A
lower-*.f32N/A
lower-sin.f32N/A
*-commutativeN/A
lower-*.f32N/A
lift-PI.f32N/A
lower-cos.f32N/A
*-commutativeN/A
lower-*.f32N/A
lift-PI.f3298.7
Applied rewrites98.7%
Taylor expanded in ux around 0
lower-fma.f32N/A
Applied rewrites93.6%
Taylor expanded in uy around 0
lower-+.f32N/A
lower-*.f32N/A
lower-fma.f32N/A
lower-*.f32N/A
pow2N/A
lift-*.f32N/A
lift-PI.f32N/A
lift-PI.f32N/A
lift-*.f32N/A
lower-*.f32N/A
lower-*.f32N/A
lift-PI.f3277.9
Applied rewrites77.9%
if 1.00000001e-7 < (*.f32 (*.f32 (-.f32 #s(literal 1 binary32) ux) maxCos) ux) Initial program 99.4%
lift-sin.f32N/A
lift-PI.f32N/A
lift-*.f32N/A
lift-*.f32N/A
*-commutativeN/A
associate-*r*N/A
sin-2N/A
lower-*.f32N/A
lower-*.f32N/A
lower-sin.f32N/A
*-commutativeN/A
lower-*.f32N/A
lift-PI.f32N/A
lower-cos.f32N/A
*-commutativeN/A
lower-*.f32N/A
lift-PI.f3299.4
Applied rewrites99.4%
Taylor expanded in uy around 0
lower-fma.f32N/A
lower-*.f32N/A
lower-*.f32N/A
lift--.f32N/A
lower-*.f32N/A
lower-sqrt.f32N/A
lower--.f32N/A
Applied rewrites85.6%
Taylor expanded in ux around 0
Applied rewrites85.7%
(FPCore (xi yi zi ux uy maxCos) :precision binary32 (if (<= (* (* (- 1.0 ux) maxCos) ux) 1.0000000116860974e-7) (+ xi (* 2.0 (* uy (* yi PI)))) (fma maxCos (* ux (* zi (- 1.0 ux))) xi)))
float code(float xi, float yi, float zi, float ux, float uy, float maxCos) {
float tmp;
if ((((1.0f - ux) * maxCos) * ux) <= 1.0000000116860974e-7f) {
tmp = xi + (2.0f * (uy * (yi * ((float) M_PI))));
} else {
tmp = fmaf(maxCos, (ux * (zi * (1.0f - ux))), xi);
}
return tmp;
}
function code(xi, yi, zi, ux, uy, maxCos) tmp = Float32(0.0) if (Float32(Float32(Float32(Float32(1.0) - ux) * maxCos) * ux) <= Float32(1.0000000116860974e-7)) tmp = Float32(xi + Float32(Float32(2.0) * Float32(uy * Float32(yi * Float32(pi))))); else tmp = fma(maxCos, Float32(ux * Float32(zi * Float32(Float32(1.0) - ux))), xi); end return tmp end
\begin{array}{l}
\\
\begin{array}{l}
\mathbf{if}\;\left(\left(1 - ux\right) \cdot maxCos\right) \cdot ux \leq 1.0000000116860974 \cdot 10^{-7}:\\
\;\;\;\;xi + 2 \cdot \left(uy \cdot \left(yi \cdot \pi\right)\right)\\
\mathbf{else}:\\
\;\;\;\;\mathsf{fma}\left(maxCos, ux \cdot \left(zi \cdot \left(1 - ux\right)\right), xi\right)\\
\end{array}
\end{array}
if (*.f32 (*.f32 (-.f32 #s(literal 1 binary32) ux) maxCos) ux) < 1.00000001e-7Initial program 98.7%
lift-sin.f32N/A
lift-PI.f32N/A
lift-*.f32N/A
lift-*.f32N/A
*-commutativeN/A
associate-*r*N/A
sin-2N/A
lower-*.f32N/A
lower-*.f32N/A
lower-sin.f32N/A
*-commutativeN/A
lower-*.f32N/A
lift-PI.f32N/A
lower-cos.f32N/A
*-commutativeN/A
lower-*.f32N/A
lift-PI.f3298.7
Applied rewrites98.7%
Taylor expanded in ux around 0
lower-fma.f32N/A
Applied rewrites93.6%
Taylor expanded in uy around 0
lower-+.f32N/A
lower-*.f32N/A
lower-*.f32N/A
lower-*.f32N/A
lift-PI.f3273.9
Applied rewrites73.9%
if 1.00000001e-7 < (*.f32 (*.f32 (-.f32 #s(literal 1 binary32) ux) maxCos) ux) Initial program 99.4%
lift-sin.f32N/A
lift-PI.f32N/A
lift-*.f32N/A
lift-*.f32N/A
*-commutativeN/A
associate-*r*N/A
sin-2N/A
lower-*.f32N/A
lower-*.f32N/A
lower-sin.f32N/A
*-commutativeN/A
lower-*.f32N/A
lift-PI.f32N/A
lower-cos.f32N/A
*-commutativeN/A
lower-*.f32N/A
lift-PI.f3299.4
Applied rewrites99.4%
Taylor expanded in uy around 0
lower-fma.f32N/A
lower-*.f32N/A
lower-*.f32N/A
lift--.f32N/A
lower-*.f32N/A
lower-sqrt.f32N/A
lower--.f32N/A
Applied rewrites85.6%
Taylor expanded in ux around 0
Applied rewrites85.7%
(FPCore (xi yi zi ux uy maxCos) :precision binary32 (+ xi (* 2.0 (* uy (* yi PI)))))
float code(float xi, float yi, float zi, float ux, float uy, float maxCos) {
return xi + (2.0f * (uy * (yi * ((float) M_PI))));
}
function code(xi, yi, zi, ux, uy, maxCos) return Float32(xi + Float32(Float32(2.0) * Float32(uy * Float32(yi * Float32(pi))))) end
function tmp = code(xi, yi, zi, ux, uy, maxCos) tmp = xi + (single(2.0) * (uy * (yi * single(pi)))); end
\begin{array}{l}
\\
xi + 2 \cdot \left(uy \cdot \left(yi \cdot \pi\right)\right)
\end{array}
Initial program 98.7%
lift-sin.f32N/A
lift-PI.f32N/A
lift-*.f32N/A
lift-*.f32N/A
*-commutativeN/A
associate-*r*N/A
sin-2N/A
lower-*.f32N/A
lower-*.f32N/A
lower-sin.f32N/A
*-commutativeN/A
lower-*.f32N/A
lift-PI.f32N/A
lower-cos.f32N/A
*-commutativeN/A
lower-*.f32N/A
lift-PI.f3298.8
Applied rewrites98.8%
Taylor expanded in ux around 0
lower-fma.f32N/A
Applied rewrites89.8%
Taylor expanded in uy around 0
lower-+.f32N/A
lower-*.f32N/A
lower-*.f32N/A
lower-*.f32N/A
lift-PI.f3271.5
Applied rewrites71.5%
(FPCore (xi yi zi ux uy maxCos) :precision binary32 (+ xi (* maxCos (* ux zi))))
float code(float xi, float yi, float zi, float ux, float uy, float maxCos) {
return xi + (maxCos * (ux * zi));
}
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 = xi + (maxcos * (ux * zi))
end function
function code(xi, yi, zi, ux, uy, maxCos) return Float32(xi + Float32(maxCos * Float32(ux * zi))) end
function tmp = code(xi, yi, zi, ux, uy, maxCos) tmp = xi + (maxCos * (ux * zi)); end
\begin{array}{l}
\\
xi + maxCos \cdot \left(ux \cdot zi\right)
\end{array}
Initial program 98.7%
lift-sin.f32N/A
lift-PI.f32N/A
lift-*.f32N/A
lift-*.f32N/A
*-commutativeN/A
associate-*r*N/A
sin-2N/A
lower-*.f32N/A
lower-*.f32N/A
lower-sin.f32N/A
*-commutativeN/A
lower-*.f32N/A
lift-PI.f32N/A
lower-cos.f32N/A
*-commutativeN/A
lower-*.f32N/A
lift-PI.f3298.8
Applied rewrites98.8%
Taylor expanded in uy around 0
lower-fma.f32N/A
lower-*.f32N/A
lower-*.f32N/A
lift--.f32N/A
lower-*.f32N/A
lower-sqrt.f32N/A
lower--.f32N/A
Applied rewrites47.9%
Taylor expanded in ux around 0
lower-+.f32N/A
lower-*.f32N/A
lower-*.f3245.5
Applied rewrites45.5%
(FPCore (xi yi zi ux uy maxCos) :precision binary32 xi)
float code(float xi, float yi, float zi, float ux, float uy, float maxCos) {
return xi;
}
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 = xi
end function
function code(xi, yi, zi, ux, uy, maxCos) return xi end
function tmp = code(xi, yi, zi, ux, uy, maxCos) tmp = xi; end
\begin{array}{l}
\\
xi
\end{array}
Initial program 98.7%
lift-sin.f32N/A
lift-PI.f32N/A
lift-*.f32N/A
lift-*.f32N/A
*-commutativeN/A
associate-*r*N/A
sin-2N/A
lower-*.f32N/A
lower-*.f32N/A
lower-sin.f32N/A
*-commutativeN/A
lower-*.f32N/A
lift-PI.f32N/A
lower-cos.f32N/A
*-commutativeN/A
lower-*.f32N/A
lift-PI.f3298.8
Applied rewrites98.8%
Taylor expanded in ux around 0
lower-fma.f32N/A
Applied rewrites89.8%
Taylor expanded in uy around 0
Applied rewrites41.3%
herbie shell --seed 2025064
(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)))