
(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 17 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 (* (- ux 1.0) (* maxCos ux)))
(t_1 (sqrt (* (+ 1.0 t_0) (- 1.0 t_0))))
(t_2 (* (* uy 2.0) PI)))
(+
(+ (* (* (cos t_2) t_1) xi) (* (* (sin t_2) t_1) yi))
(* (* (* (- 1.0 ux) maxCos) ux) zi))))
float code(float xi, float yi, float zi, float ux, float uy, float maxCos) {
float t_0 = (ux - 1.0f) * (maxCos * ux);
float t_1 = sqrtf(((1.0f + t_0) * (1.0f - t_0)));
float t_2 = (uy * 2.0f) * ((float) M_PI);
return (((cosf(t_2) * t_1) * xi) + ((sinf(t_2) * t_1) * yi)) + ((((1.0f - ux) * maxCos) * ux) * zi);
}
function code(xi, yi, zi, ux, uy, maxCos) t_0 = Float32(Float32(ux - Float32(1.0)) * Float32(maxCos * ux)) t_1 = sqrt(Float32(Float32(Float32(1.0) + t_0) * Float32(Float32(1.0) - 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(Float32(Float32(Float32(Float32(1.0) - ux) * maxCos) * ux) * zi)) end
function tmp = code(xi, yi, zi, ux, uy, maxCos) t_0 = (ux - single(1.0)) * (maxCos * ux); t_1 = sqrt(((single(1.0) + t_0) * (single(1.0) - t_0))); t_2 = (uy * single(2.0)) * single(pi); tmp = (((cos(t_2) * t_1) * xi) + ((sin(t_2) * t_1) * yi)) + ((((single(1.0) - ux) * maxCos) * ux) * zi); end
\begin{array}{l}
\\
\begin{array}{l}
t_0 := \left(ux - 1\right) \cdot \left(maxCos \cdot ux\right)\\
t_1 := \sqrt{\left(1 + t\_0\right) \cdot \left(1 - t\_0\right)}\\
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) + \left(\left(\left(1 - ux\right) \cdot maxCos\right) \cdot ux\right) \cdot zi
\end{array}
\end{array}
Initial program 99.0%
lift--.f32N/A
metadata-evalN/A
lift-*.f32N/A
sqr-neg-revN/A
difference-of-squaresN/A
lower-*.f32N/A
lower-+.f32N/A
lift-*.f32N/A
lift-*.f32N/A
associate-*l*N/A
distribute-lft-neg-inN/A
lift--.f32N/A
sub-negate-revN/A
lower-*.f32N/A
lower--.f32N/A
lower-*.f32N/A
lower--.f32N/A
lift-*.f32N/A
Applied rewrites98.9%
lift--.f32N/A
metadata-evalN/A
lift-*.f32N/A
sqr-neg-revN/A
difference-of-squaresN/A
lower-*.f32N/A
lower-+.f32N/A
lift-*.f32N/A
lift-*.f32N/A
associate-*l*N/A
distribute-lft-neg-inN/A
lift--.f32N/A
sub-negate-revN/A
lower-*.f32N/A
lower--.f32N/A
lower-*.f32N/A
lower--.f32N/A
lift-*.f32N/A
Applied rewrites98.9%
(FPCore (xi yi zi ux uy maxCos)
:precision binary32
(let* ((t_0 (* PI (+ uy uy)))
(t_1
(sqrt
(fma
(* (* (- ux 1.0) (* maxCos ux)) (- 1.0 ux))
(* maxCos ux)
1.0))))
(fma
(* xi (cos t_0))
t_1
(fma (* yi t_1) (sin t_0) (* zi (* (* maxCos (- 1.0 ux)) ux))))))
float code(float xi, float yi, float zi, float ux, float uy, float maxCos) {
float t_0 = ((float) M_PI) * (uy + uy);
float t_1 = sqrtf(fmaf((((ux - 1.0f) * (maxCos * ux)) * (1.0f - ux)), (maxCos * ux), 1.0f));
return fmaf((xi * cosf(t_0)), t_1, fmaf((yi * t_1), sinf(t_0), (zi * ((maxCos * (1.0f - ux)) * ux))));
}
function code(xi, yi, zi, ux, uy, maxCos) t_0 = Float32(Float32(pi) * Float32(uy + uy)) t_1 = sqrt(fma(Float32(Float32(Float32(ux - Float32(1.0)) * Float32(maxCos * ux)) * Float32(Float32(1.0) - ux)), Float32(maxCos * ux), Float32(1.0))) return fma(Float32(xi * cos(t_0)), t_1, fma(Float32(yi * t_1), sin(t_0), Float32(zi * Float32(Float32(maxCos * Float32(Float32(1.0) - ux)) * ux)))) end
\begin{array}{l}
\\
\begin{array}{l}
t_0 := \pi \cdot \left(uy + uy\right)\\
t_1 := \sqrt{\mathsf{fma}\left(\left(\left(ux - 1\right) \cdot \left(maxCos \cdot ux\right)\right) \cdot \left(1 - ux\right), maxCos \cdot ux, 1\right)}\\
\mathsf{fma}\left(xi \cdot \cos t\_0, t\_1, \mathsf{fma}\left(yi \cdot t\_1, \sin t\_0, zi \cdot \left(\left(maxCos \cdot \left(1 - ux\right)\right) \cdot ux\right)\right)\right)
\end{array}
\end{array}
Initial program 99.0%
Applied rewrites99.0%
(FPCore (xi yi zi ux uy maxCos)
:precision binary32
(let* ((t_0 (* (* (- 1.0 ux) maxCos) ux)))
(+
(+
(* (* (cos (* (* uy 2.0) 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 * 2.0f) * ((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 * Float32(2.0)) * 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 * single(2.0)) * 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 \cdot 2\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 99.0%
Taylor expanded in ux around 0
lower-sin.f32N/A
lower-*.f32N/A
lower-*.f32N/A
lower-PI.f3298.9
Applied rewrites98.9%
(FPCore (xi yi zi ux uy maxCos)
:precision binary32
(let* ((t_0 (* (* (- 1.0 ux) maxCos) ux)))
(+
(+
(* (cos (* 2.0 (* uy PI))) xi)
(* (* (sin (* (* uy 2.0) PI)) (sqrt (- 1.0 (* t_0 t_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;
return ((cosf((2.0f * (uy * ((float) M_PI)))) * xi) + ((sinf(((uy * 2.0f) * ((float) M_PI))) * sqrtf((1.0f - (t_0 * t_0)))) * 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(cos(Float32(Float32(2.0) * Float32(uy * Float32(pi)))) * xi) + Float32(Float32(sin(Float32(Float32(uy * Float32(2.0)) * Float32(pi))) * sqrt(Float32(Float32(1.0) - Float32(t_0 * t_0)))) * 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((single(2.0) * (uy * single(pi)))) * xi) + ((sin(((uy * single(2.0)) * single(pi))) * sqrt((single(1.0) - (t_0 * t_0)))) * yi)) + (t_0 * zi); end
\begin{array}{l}
\\
\begin{array}{l}
t_0 := \left(\left(1 - ux\right) \cdot maxCos\right) \cdot ux\\
\left(\cos \left(2 \cdot \left(uy \cdot \pi\right)\right) \cdot xi + \left(\sin \left(\left(uy \cdot 2\right) \cdot \pi\right) \cdot \sqrt{1 - t\_0 \cdot t\_0}\right) \cdot yi\right) + t\_0 \cdot zi
\end{array}
\end{array}
Initial program 99.0%
Taylor expanded in ux around 0
lower-cos.f32N/A
lower-*.f32N/A
lower-*.f32N/A
lower-PI.f3298.8
Applied rewrites98.8%
(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 99.0%
Taylor expanded in maxCos around 0
lower-fma.f32N/A
lower-*.f32N/A
lower-*.f32N/A
lower--.f32N/A
lower-fma.f32N/A
Applied rewrites98.8%
(FPCore (xi yi zi ux uy maxCos)
:precision binary32
(let* ((t_0 (* maxCos (* ux (- ux 1.0))))
(t_1 (* (+ uy uy) PI))
(t_2 (sqrt (* (+ 1.0 t_0) (- 1.0 t_0)))))
(if (<= uy 0.00019999999494757503)
(+
(+ (* t_2 xi) (* (* 2.0 (* uy (* PI t_2))) yi))
(* (* (* (- 1.0 ux) maxCos) ux) zi))
(fma (cos t_1) xi (* (sin t_1) yi)))))
float code(float xi, float yi, float zi, float ux, float uy, float maxCos) {
float t_0 = maxCos * (ux * (ux - 1.0f));
float t_1 = (uy + uy) * ((float) M_PI);
float t_2 = sqrtf(((1.0f + t_0) * (1.0f - t_0)));
float tmp;
if (uy <= 0.00019999999494757503f) {
tmp = ((t_2 * xi) + ((2.0f * (uy * (((float) M_PI) * t_2))) * yi)) + ((((1.0f - ux) * maxCos) * ux) * zi);
} else {
tmp = fmaf(cosf(t_1), xi, (sinf(t_1) * yi));
}
return tmp;
}
function code(xi, yi, zi, ux, uy, maxCos) t_0 = Float32(maxCos * Float32(ux * Float32(ux - Float32(1.0)))) t_1 = Float32(Float32(uy + uy) * Float32(pi)) t_2 = sqrt(Float32(Float32(Float32(1.0) + t_0) * Float32(Float32(1.0) - t_0))) tmp = Float32(0.0) if (uy <= Float32(0.00019999999494757503)) tmp = Float32(Float32(Float32(t_2 * xi) + Float32(Float32(Float32(2.0) * Float32(uy * Float32(Float32(pi) * t_2))) * yi)) + Float32(Float32(Float32(Float32(Float32(1.0) - ux) * maxCos) * ux) * zi)); else tmp = fma(cos(t_1), xi, Float32(sin(t_1) * yi)); end return tmp end
\begin{array}{l}
\\
\begin{array}{l}
t_0 := maxCos \cdot \left(ux \cdot \left(ux - 1\right)\right)\\
t_1 := \left(uy + uy\right) \cdot \pi\\
t_2 := \sqrt{\left(1 + t\_0\right) \cdot \left(1 - t\_0\right)}\\
\mathbf{if}\;uy \leq 0.00019999999494757503:\\
\;\;\;\;\left(t\_2 \cdot xi + \left(2 \cdot \left(uy \cdot \left(\pi \cdot t\_2\right)\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\_1, xi, \sin t\_1 \cdot yi\right)\\
\end{array}
\end{array}
if uy < 1.99999995e-4Initial program 99.0%
lift--.f32N/A
metadata-evalN/A
lift-*.f32N/A
sqr-neg-revN/A
difference-of-squaresN/A
lower-*.f32N/A
lower-+.f32N/A
lift-*.f32N/A
lift-*.f32N/A
associate-*l*N/A
distribute-lft-neg-inN/A
lift--.f32N/A
sub-negate-revN/A
lower-*.f32N/A
lower--.f32N/A
lower-*.f32N/A
lower--.f32N/A
lift-*.f32N/A
Applied rewrites98.9%
lift--.f32N/A
metadata-evalN/A
lift-*.f32N/A
sqr-neg-revN/A
difference-of-squaresN/A
lower-*.f32N/A
lower-+.f32N/A
lift-*.f32N/A
lift-*.f32N/A
associate-*l*N/A
distribute-lft-neg-inN/A
lift--.f32N/A
sub-negate-revN/A
lower-*.f32N/A
lower--.f32N/A
lower-*.f32N/A
lower--.f32N/A
lift-*.f32N/A
Applied rewrites98.9%
Taylor expanded in uy around 0
lower-sqrt.f32N/A
lower-*.f32N/A
lower-+.f32N/A
lower-*.f32N/A
lower-*.f32N/A
lower--.f32N/A
lower--.f32N/A
lower-*.f32N/A
lower-*.f32N/A
lower--.f3288.5
Applied rewrites88.5%
Taylor expanded in uy around 0
lower-*.f32N/A
lower-*.f32N/A
lower-*.f32N/A
lower-PI.f32N/A
lower-sqrt.f32N/A
lower-*.f32N/A
Applied rewrites81.6%
if 1.99999995e-4 < uy Initial program 99.0%
Taylor expanded in ux around 0
lower-fma.f32N/A
lower-cos.f32N/A
lower-*.f32N/A
lower-*.f32N/A
lower-PI.f32N/A
lower-*.f32N/A
lower-sin.f32N/A
lower-*.f32N/A
lower-*.f32N/A
lower-PI.f3290.5
Applied rewrites90.5%
lift-fma.f32N/A
*-commutativeN/A
lift-*.f32N/A
lift-*.f32N/A
associate-*r*N/A
*-commutativeN/A
lift-*.f32N/A
lift-*.f32N/A
lower-fma.f3290.5
lift-*.f32N/A
*-commutativeN/A
count-2N/A
lift-+.f3290.5
lift-*.f32N/A
*-commutativeN/A
Applied rewrites90.5%
(FPCore (xi yi zi ux uy maxCos)
:precision binary32
(let* ((t_0 (* (+ uy uy) PI))
(t_1 (* maxCos (* ux (- ux 1.0))))
(t_2 (sqrt (* (+ 1.0 t_1) (- 1.0 t_1)))))
(if (<= uy 0.00019999999494757503)
(fma
2.0
(* uy (* yi (* PI t_2)))
(fma maxCos (* ux (* zi (- 1.0 ux))) (* xi t_2)))
(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 = (uy + uy) * ((float) M_PI);
float t_1 = maxCos * (ux * (ux - 1.0f));
float t_2 = sqrtf(((1.0f + t_1) * (1.0f - t_1)));
float tmp;
if (uy <= 0.00019999999494757503f) {
tmp = fmaf(2.0f, (uy * (yi * (((float) M_PI) * t_2))), fmaf(maxCos, (ux * (zi * (1.0f - ux))), (xi * t_2)));
} 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(uy + uy) * Float32(pi)) t_1 = Float32(maxCos * Float32(ux * Float32(ux - Float32(1.0)))) t_2 = sqrt(Float32(Float32(Float32(1.0) + t_1) * Float32(Float32(1.0) - t_1))) tmp = Float32(0.0) if (uy <= Float32(0.00019999999494757503)) tmp = fma(Float32(2.0), Float32(uy * Float32(yi * Float32(Float32(pi) * t_2))), fma(maxCos, Float32(ux * Float32(zi * Float32(Float32(1.0) - ux))), Float32(xi * t_2))); else tmp = fma(cos(t_0), xi, Float32(sin(t_0) * yi)); end return tmp end
\begin{array}{l}
\\
\begin{array}{l}
t_0 := \left(uy + uy\right) \cdot \pi\\
t_1 := maxCos \cdot \left(ux \cdot \left(ux - 1\right)\right)\\
t_2 := \sqrt{\left(1 + t\_1\right) \cdot \left(1 - t\_1\right)}\\
\mathbf{if}\;uy \leq 0.00019999999494757503:\\
\;\;\;\;\mathsf{fma}\left(2, uy \cdot \left(yi \cdot \left(\pi \cdot t\_2\right)\right), \mathsf{fma}\left(maxCos, ux \cdot \left(zi \cdot \left(1 - ux\right)\right), xi \cdot t\_2\right)\right)\\
\mathbf{else}:\\
\;\;\;\;\mathsf{fma}\left(\cos t\_0, xi, \sin t\_0 \cdot yi\right)\\
\end{array}
\end{array}
if uy < 1.99999995e-4Initial program 99.0%
lift--.f32N/A
metadata-evalN/A
lift-*.f32N/A
sqr-neg-revN/A
difference-of-squaresN/A
lower-*.f32N/A
lower-+.f32N/A
lift-*.f32N/A
lift-*.f32N/A
associate-*l*N/A
distribute-lft-neg-inN/A
lift--.f32N/A
sub-negate-revN/A
lower-*.f32N/A
lower--.f32N/A
lower-*.f32N/A
lower--.f32N/A
lift-*.f32N/A
Applied rewrites98.9%
lift--.f32N/A
metadata-evalN/A
lift-*.f32N/A
sqr-neg-revN/A
difference-of-squaresN/A
lower-*.f32N/A
lower-+.f32N/A
lift-*.f32N/A
lift-*.f32N/A
associate-*l*N/A
distribute-lft-neg-inN/A
lift--.f32N/A
sub-negate-revN/A
lower-*.f32N/A
lower--.f32N/A
lower-*.f32N/A
lower--.f32N/A
lift-*.f32N/A
Applied rewrites98.9%
Taylor expanded in uy around 0
lower-fma.f32N/A
Applied rewrites81.6%
if 1.99999995e-4 < uy Initial program 99.0%
Taylor expanded in ux around 0
lower-fma.f32N/A
lower-cos.f32N/A
lower-*.f32N/A
lower-*.f32N/A
lower-PI.f32N/A
lower-*.f32N/A
lower-sin.f32N/A
lower-*.f32N/A
lower-*.f32N/A
lower-PI.f3290.5
Applied rewrites90.5%
lift-fma.f32N/A
*-commutativeN/A
lift-*.f32N/A
lift-*.f32N/A
associate-*r*N/A
*-commutativeN/A
lift-*.f32N/A
lift-*.f32N/A
lower-fma.f3290.5
lift-*.f32N/A
*-commutativeN/A
count-2N/A
lift-+.f3290.5
lift-*.f32N/A
*-commutativeN/A
Applied rewrites90.5%
(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 99.0%
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
lower-PI.f32N/A
lower-*.f32N/A
Applied rewrites95.8%
(FPCore (xi yi zi ux uy maxCos)
:precision binary32
(let* ((t_0 (* maxCos (* ux (- ux 1.0))))
(t_1 (sqrt (* (+ 1.0 t_0) (- 1.0 t_0)))))
(if (<= uy 0.0011500000255182385)
(+
(fma 2.0 (* uy (* yi (* PI t_1))) (* xi t_1))
(* (* (* (- 1.0 ux) maxCos) ux) zi))
(fma xi (cos (* 2.0 (* uy PI))) (* 2.0 (* uy (* yi PI)))))))
float code(float xi, float yi, float zi, float ux, float uy, float maxCos) {
float t_0 = maxCos * (ux * (ux - 1.0f));
float t_1 = sqrtf(((1.0f + t_0) * (1.0f - t_0)));
float tmp;
if (uy <= 0.0011500000255182385f) {
tmp = fmaf(2.0f, (uy * (yi * (((float) M_PI) * t_1))), (xi * t_1)) + ((((1.0f - ux) * maxCos) * ux) * zi);
} else {
tmp = fmaf(xi, cosf((2.0f * (uy * ((float) M_PI)))), (2.0f * (uy * (yi * ((float) M_PI)))));
}
return tmp;
}
function code(xi, yi, zi, ux, uy, maxCos) t_0 = Float32(maxCos * Float32(ux * Float32(ux - Float32(1.0)))) t_1 = sqrt(Float32(Float32(Float32(1.0) + t_0) * Float32(Float32(1.0) - t_0))) tmp = Float32(0.0) if (uy <= Float32(0.0011500000255182385)) tmp = Float32(fma(Float32(2.0), Float32(uy * Float32(yi * Float32(Float32(pi) * t_1))), Float32(xi * t_1)) + Float32(Float32(Float32(Float32(Float32(1.0) - ux) * maxCos) * ux) * zi)); else tmp = fma(xi, cos(Float32(Float32(2.0) * Float32(uy * Float32(pi)))), Float32(Float32(2.0) * Float32(uy * Float32(yi * Float32(pi))))); end return tmp end
\begin{array}{l}
\\
\begin{array}{l}
t_0 := maxCos \cdot \left(ux \cdot \left(ux - 1\right)\right)\\
t_1 := \sqrt{\left(1 + t\_0\right) \cdot \left(1 - t\_0\right)}\\
\mathbf{if}\;uy \leq 0.0011500000255182385:\\
\;\;\;\;\mathsf{fma}\left(2, uy \cdot \left(yi \cdot \left(\pi \cdot t\_1\right)\right), xi \cdot t\_1\right) + \left(\left(\left(1 - ux\right) \cdot maxCos\right) \cdot ux\right) \cdot zi\\
\mathbf{else}:\\
\;\;\;\;\mathsf{fma}\left(xi, \cos \left(2 \cdot \left(uy \cdot \pi\right)\right), 2 \cdot \left(uy \cdot \left(yi \cdot \pi\right)\right)\right)\\
\end{array}
\end{array}
if uy < 0.00115000003Initial program 99.0%
lift--.f32N/A
metadata-evalN/A
lift-*.f32N/A
sqr-neg-revN/A
difference-of-squaresN/A
lower-*.f32N/A
lower-+.f32N/A
lift-*.f32N/A
lift-*.f32N/A
associate-*l*N/A
distribute-lft-neg-inN/A
lift--.f32N/A
sub-negate-revN/A
lower-*.f32N/A
lower--.f32N/A
lower-*.f32N/A
lower--.f32N/A
lift-*.f32N/A
Applied rewrites98.9%
lift--.f32N/A
metadata-evalN/A
lift-*.f32N/A
sqr-neg-revN/A
difference-of-squaresN/A
lower-*.f32N/A
lower-+.f32N/A
lift-*.f32N/A
lift-*.f32N/A
associate-*l*N/A
distribute-lft-neg-inN/A
lift--.f32N/A
sub-negate-revN/A
lower-*.f32N/A
lower--.f32N/A
lower-*.f32N/A
lower--.f32N/A
lift-*.f32N/A
Applied rewrites98.9%
Taylor expanded in uy around 0
lower-fma.f32N/A
Applied rewrites81.6%
if 0.00115000003 < uy Initial program 99.0%
Taylor expanded in ux around 0
lower-fma.f32N/A
lower-cos.f32N/A
lower-*.f32N/A
lower-*.f32N/A
lower-PI.f32N/A
lower-*.f32N/A
lower-sin.f32N/A
lower-*.f32N/A
lower-*.f32N/A
lower-PI.f3290.5
Applied rewrites90.5%
Taylor expanded in uy around 0
lower-*.f32N/A
lower-*.f32N/A
lower-*.f32N/A
lower-PI.f3282.2
Applied rewrites82.2%
(FPCore (xi yi zi ux uy maxCos)
:precision binary32
(let* ((t_0 (* maxCos (* ux (- ux 1.0))))
(t_1 (sqrt (* (+ 1.0 t_0) (- 1.0 t_0)))))
(if (<= uy 0.0011500000255182385)
(fma
2.0
(* uy (* yi (* PI t_1)))
(fma maxCos (* ux (* zi (- 1.0 ux))) (* xi t_1)))
(fma xi (cos (* 2.0 (* uy PI))) (* 2.0 (* uy (* yi PI)))))))
float code(float xi, float yi, float zi, float ux, float uy, float maxCos) {
float t_0 = maxCos * (ux * (ux - 1.0f));
float t_1 = sqrtf(((1.0f + t_0) * (1.0f - t_0)));
float tmp;
if (uy <= 0.0011500000255182385f) {
tmp = fmaf(2.0f, (uy * (yi * (((float) M_PI) * t_1))), fmaf(maxCos, (ux * (zi * (1.0f - ux))), (xi * t_1)));
} else {
tmp = fmaf(xi, cosf((2.0f * (uy * ((float) M_PI)))), (2.0f * (uy * (yi * ((float) M_PI)))));
}
return tmp;
}
function code(xi, yi, zi, ux, uy, maxCos) t_0 = Float32(maxCos * Float32(ux * Float32(ux - Float32(1.0)))) t_1 = sqrt(Float32(Float32(Float32(1.0) + t_0) * Float32(Float32(1.0) - t_0))) tmp = Float32(0.0) if (uy <= Float32(0.0011500000255182385)) tmp = fma(Float32(2.0), Float32(uy * Float32(yi * Float32(Float32(pi) * t_1))), fma(maxCos, Float32(ux * Float32(zi * Float32(Float32(1.0) - ux))), Float32(xi * t_1))); else tmp = fma(xi, cos(Float32(Float32(2.0) * Float32(uy * Float32(pi)))), Float32(Float32(2.0) * Float32(uy * Float32(yi * Float32(pi))))); end return tmp end
\begin{array}{l}
\\
\begin{array}{l}
t_0 := maxCos \cdot \left(ux \cdot \left(ux - 1\right)\right)\\
t_1 := \sqrt{\left(1 + t\_0\right) \cdot \left(1 - t\_0\right)}\\
\mathbf{if}\;uy \leq 0.0011500000255182385:\\
\;\;\;\;\mathsf{fma}\left(2, uy \cdot \left(yi \cdot \left(\pi \cdot t\_1\right)\right), \mathsf{fma}\left(maxCos, ux \cdot \left(zi \cdot \left(1 - ux\right)\right), xi \cdot t\_1\right)\right)\\
\mathbf{else}:\\
\;\;\;\;\mathsf{fma}\left(xi, \cos \left(2 \cdot \left(uy \cdot \pi\right)\right), 2 \cdot \left(uy \cdot \left(yi \cdot \pi\right)\right)\right)\\
\end{array}
\end{array}
if uy < 0.00115000003Initial program 99.0%
lift--.f32N/A
metadata-evalN/A
lift-*.f32N/A
sqr-neg-revN/A
difference-of-squaresN/A
lower-*.f32N/A
lower-+.f32N/A
lift-*.f32N/A
lift-*.f32N/A
associate-*l*N/A
distribute-lft-neg-inN/A
lift--.f32N/A
sub-negate-revN/A
lower-*.f32N/A
lower--.f32N/A
lower-*.f32N/A
lower--.f32N/A
lift-*.f32N/A
Applied rewrites98.9%
lift--.f32N/A
metadata-evalN/A
lift-*.f32N/A
sqr-neg-revN/A
difference-of-squaresN/A
lower-*.f32N/A
lower-+.f32N/A
lift-*.f32N/A
lift-*.f32N/A
associate-*l*N/A
distribute-lft-neg-inN/A
lift--.f32N/A
sub-negate-revN/A
lower-*.f32N/A
lower--.f32N/A
lower-*.f32N/A
lower--.f32N/A
lift-*.f32N/A
Applied rewrites98.9%
Taylor expanded in uy around 0
lower-fma.f32N/A
Applied rewrites81.6%
if 0.00115000003 < uy Initial program 99.0%
Taylor expanded in ux around 0
lower-fma.f32N/A
lower-cos.f32N/A
lower-*.f32N/A
lower-*.f32N/A
lower-PI.f32N/A
lower-*.f32N/A
lower-sin.f32N/A
lower-*.f32N/A
lower-*.f32N/A
lower-PI.f3290.5
Applied rewrites90.5%
Taylor expanded in uy around 0
lower-*.f32N/A
lower-*.f32N/A
lower-*.f32N/A
lower-PI.f3282.2
Applied rewrites82.2%
(FPCore (xi yi zi ux uy maxCos) :precision binary32 (fma xi (cos (* 2.0 (* uy PI))) (* 2.0 (* uy (* yi PI)))))
float code(float xi, float yi, float zi, float ux, float uy, float maxCos) {
return fmaf(xi, cosf((2.0f * (uy * ((float) M_PI)))), (2.0f * (uy * (yi * ((float) M_PI)))));
}
function code(xi, yi, zi, ux, uy, maxCos) return fma(xi, cos(Float32(Float32(2.0) * Float32(uy * Float32(pi)))), Float32(Float32(2.0) * Float32(uy * Float32(yi * Float32(pi))))) end
\begin{array}{l}
\\
\mathsf{fma}\left(xi, \cos \left(2 \cdot \left(uy \cdot \pi\right)\right), 2 \cdot \left(uy \cdot \left(yi \cdot \pi\right)\right)\right)
\end{array}
Initial program 99.0%
Taylor expanded in ux around 0
lower-fma.f32N/A
lower-cos.f32N/A
lower-*.f32N/A
lower-*.f32N/A
lower-PI.f32N/A
lower-*.f32N/A
lower-sin.f32N/A
lower-*.f32N/A
lower-*.f32N/A
lower-PI.f3290.5
Applied rewrites90.5%
Taylor expanded in uy around 0
lower-*.f32N/A
lower-*.f32N/A
lower-*.f32N/A
lower-PI.f3282.2
Applied rewrites82.2%
(FPCore (xi yi zi ux uy maxCos) :precision binary32 (if (<= uy 0.05999999865889549) (+ xi (* uy (fma -2.0 (* uy (* xi (pow PI 2.0))) (* 2.0 (* yi PI))))) (* yi (sin (* 2.0 (* uy PI))))))
float code(float xi, float yi, float zi, float ux, float uy, float maxCos) {
float tmp;
if (uy <= 0.05999999865889549f) {
tmp = xi + (uy * fmaf(-2.0f, (uy * (xi * powf(((float) M_PI), 2.0f))), (2.0f * (yi * ((float) M_PI)))));
} else {
tmp = yi * sinf((2.0f * (uy * ((float) M_PI))));
}
return tmp;
}
function code(xi, yi, zi, ux, uy, maxCos) tmp = Float32(0.0) if (uy <= Float32(0.05999999865889549)) tmp = Float32(xi + Float32(uy * fma(Float32(-2.0), Float32(uy * Float32(xi * (Float32(pi) ^ Float32(2.0)))), Float32(Float32(2.0) * Float32(yi * Float32(pi)))))); else tmp = Float32(yi * sin(Float32(Float32(2.0) * Float32(uy * Float32(pi))))); end return tmp end
\begin{array}{l}
\\
\begin{array}{l}
\mathbf{if}\;uy \leq 0.05999999865889549:\\
\;\;\;\;xi + uy \cdot \mathsf{fma}\left(-2, uy \cdot \left(xi \cdot {\pi}^{2}\right), 2 \cdot \left(yi \cdot \pi\right)\right)\\
\mathbf{else}:\\
\;\;\;\;yi \cdot \sin \left(2 \cdot \left(uy \cdot \pi\right)\right)\\
\end{array}
\end{array}
if uy < 0.0599999987Initial program 99.0%
Taylor expanded in ux around 0
lower-fma.f32N/A
lower-cos.f32N/A
lower-*.f32N/A
lower-*.f32N/A
lower-PI.f32N/A
lower-*.f32N/A
lower-sin.f32N/A
lower-*.f32N/A
lower-*.f32N/A
lower-PI.f3290.5
Applied rewrites90.5%
Taylor expanded in uy around 0
lower-+.f32N/A
lower-*.f32N/A
lower-fma.f32N/A
lower-*.f32N/A
lower-*.f32N/A
lower-pow.f32N/A
lower-PI.f32N/A
lower-*.f32N/A
lower-*.f32N/A
lower-PI.f3278.0
Applied rewrites78.0%
if 0.0599999987 < uy Initial program 99.0%
Taylor expanded in ux around 0
lower-fma.f32N/A
lower-cos.f32N/A
lower-*.f32N/A
lower-*.f32N/A
lower-PI.f32N/A
lower-*.f32N/A
lower-sin.f32N/A
lower-*.f32N/A
lower-*.f32N/A
lower-PI.f3290.5
Applied rewrites90.5%
Taylor expanded in xi around 0
lower-*.f32N/A
lower-sin.f32N/A
lower-*.f32N/A
lower-*.f32N/A
lower-PI.f3238.5
Applied rewrites38.5%
(FPCore (xi yi zi ux uy maxCos) :precision binary32 (+ xi (* uy (fma -2.0 (* uy (* xi (pow PI 2.0))) (* 2.0 (* yi PI))))))
float code(float xi, float yi, float zi, float ux, float uy, float maxCos) {
return xi + (uy * fmaf(-2.0f, (uy * (xi * powf(((float) M_PI), 2.0f))), (2.0f * (yi * ((float) M_PI)))));
}
function code(xi, yi, zi, ux, uy, maxCos) return Float32(xi + Float32(uy * fma(Float32(-2.0), Float32(uy * Float32(xi * (Float32(pi) ^ Float32(2.0)))), Float32(Float32(2.0) * Float32(yi * Float32(pi)))))) end
\begin{array}{l}
\\
xi + uy \cdot \mathsf{fma}\left(-2, uy \cdot \left(xi \cdot {\pi}^{2}\right), 2 \cdot \left(yi \cdot \pi\right)\right)
\end{array}
Initial program 99.0%
Taylor expanded in ux around 0
lower-fma.f32N/A
lower-cos.f32N/A
lower-*.f32N/A
lower-*.f32N/A
lower-PI.f32N/A
lower-*.f32N/A
lower-sin.f32N/A
lower-*.f32N/A
lower-*.f32N/A
lower-PI.f3290.5
Applied rewrites90.5%
Taylor expanded in uy around 0
lower-+.f32N/A
lower-*.f32N/A
lower-fma.f32N/A
lower-*.f32N/A
lower-*.f32N/A
lower-pow.f32N/A
lower-PI.f32N/A
lower-*.f32N/A
lower-*.f32N/A
lower-PI.f3278.0
Applied rewrites78.0%
(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 99.0%
Taylor expanded in ux around 0
lower-fma.f32N/A
lower-cos.f32N/A
lower-*.f32N/A
lower-*.f32N/A
lower-PI.f32N/A
lower-*.f32N/A
lower-sin.f32N/A
lower-*.f32N/A
lower-*.f32N/A
lower-PI.f3290.5
Applied rewrites90.5%
Taylor expanded in uy around 0
lower-+.f32N/A
lower-*.f32N/A
lower-*.f32N/A
lower-*.f32N/A
lower-PI.f3274.1
Applied rewrites74.1%
(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 99.0%
Taylor expanded in uy around 0
lower-fma.f32N/A
lower-*.f32N/A
lower-*.f32N/A
lower--.f32N/A
lower-*.f32N/A
lower-sqrt.f32N/A
lower--.f32N/A
Applied rewrites51.4%
Taylor expanded in ux around 0
lower-+.f32N/A
lower-*.f32N/A
lower-*.f3249.2
Applied rewrites49.2%
(FPCore (xi yi zi ux uy maxCos) :precision binary32 (fma (* maxCos ux) zi xi))
float code(float xi, float yi, float zi, float ux, float uy, float maxCos) {
return fmaf((maxCos * ux), zi, xi);
}
function code(xi, yi, zi, ux, uy, maxCos) return fma(Float32(maxCos * ux), zi, xi) end
\begin{array}{l}
\\
\mathsf{fma}\left(maxCos \cdot ux, zi, xi\right)
\end{array}
Initial program 99.0%
Taylor expanded in uy around 0
lower-fma.f32N/A
lower-*.f32N/A
lower-*.f32N/A
lower--.f32N/A
lower-*.f32N/A
lower-sqrt.f32N/A
lower--.f32N/A
Applied rewrites51.4%
Taylor expanded in ux around 0
lower-+.f32N/A
lower-*.f32N/A
lower-*.f3249.2
Applied rewrites49.2%
lift-+.f32N/A
+-commutativeN/A
lift-*.f32N/A
lift-*.f32N/A
associate-*r*N/A
lift-*.f32N/A
lower-fma.f3249.2
Applied rewrites49.2%
(FPCore (xi yi zi ux uy maxCos) :precision binary32 (* maxCos (* ux zi)))
float code(float xi, float yi, float zi, float ux, float uy, float maxCos) {
return 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 = maxcos * (ux * zi)
end function
function code(xi, yi, zi, ux, uy, maxCos) return Float32(maxCos * Float32(ux * zi)) end
function tmp = code(xi, yi, zi, ux, uy, maxCos) tmp = maxCos * (ux * zi); end
\begin{array}{l}
\\
maxCos \cdot \left(ux \cdot zi\right)
\end{array}
Initial program 99.0%
Taylor expanded in uy around 0
lower-fma.f32N/A
lower-*.f32N/A
lower-*.f32N/A
lower--.f32N/A
lower-*.f32N/A
lower-sqrt.f32N/A
lower--.f32N/A
Applied rewrites51.4%
Taylor expanded in ux around 0
lower-+.f32N/A
lower-*.f32N/A
lower-*.f3249.2
Applied rewrites49.2%
Taylor expanded in xi around 0
lower-*.f32N/A
lower-*.f3211.9
Applied rewrites11.9%
herbie shell --seed 2025156
(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)))