
(FPCore (ux uy maxCos) :precision binary32 (let* ((t_0 (+ (- 1.0 ux) (* ux maxCos)))) (* (cos (* (* uy 2.0) (PI))) (sqrt (- 1.0 (* t_0 t_0))))))
\begin{array}{l}
\\
\begin{array}{l}
t_0 := \left(1 - ux\right) + ux \cdot maxCos\\
\cos \left(\left(uy \cdot 2\right) \cdot \mathsf{PI}\left(\right)\right) \cdot \sqrt{1 - t\_0 \cdot t\_0}
\end{array}
\end{array}
Sampling outcomes in binary32 precision:
Herbie found 14 alternatives:
| Alternative | Accuracy | Speedup |
|---|
(FPCore (ux uy maxCos) :precision binary32 (let* ((t_0 (+ (- 1.0 ux) (* ux maxCos)))) (* (cos (* (* uy 2.0) (PI))) (sqrt (- 1.0 (* t_0 t_0))))))
\begin{array}{l}
\\
\begin{array}{l}
t_0 := \left(1 - ux\right) + ux \cdot maxCos\\
\cos \left(\left(uy \cdot 2\right) \cdot \mathsf{PI}\left(\right)\right) \cdot \sqrt{1 - t\_0 \cdot t\_0}
\end{array}
\end{array}
(FPCore (ux uy maxCos) :precision binary32 (* (sqrt (* (fma maxCos -2.0 (- 2.0 (* (pow (- maxCos 1.0) 2.0) ux))) ux)) (sin (* (PI) (fma -2.0 uy 0.5)))))
\begin{array}{l}
\\
\sqrt{\mathsf{fma}\left(maxCos, -2, 2 - {\left(maxCos - 1\right)}^{2} \cdot ux\right) \cdot ux} \cdot \sin \left(\mathsf{PI}\left(\right) \cdot \mathsf{fma}\left(-2, uy, 0.5\right)\right)
\end{array}
Initial program 59.1%
Taylor expanded in ux around 0
*-commutativeN/A
lower-*.f32N/A
fp-cancel-sub-sign-invN/A
metadata-evalN/A
+-commutativeN/A
lower-fma.f32N/A
associate-*r*N/A
mul-1-negN/A
fp-cancel-sub-signN/A
lower--.f32N/A
*-commutativeN/A
lower-*.f32N/A
lower-pow.f32N/A
lower--.f3298.9
Applied rewrites98.9%
lift-cos.f32N/A
cos-neg-revN/A
sin-+PI/2-revN/A
lower-sin.f32N/A
lift-*.f32N/A
*-commutativeN/A
distribute-lft-neg-inN/A
lower-fma.f32N/A
lower-neg.f32N/A
lift-*.f32N/A
*-commutativeN/A
lower-*.f32N/A
lift-PI.f32N/A
lower-/.f3299.0
Applied rewrites99.0%
Taylor expanded in uy around 0
*-commutativeN/A
lower-fma.f32N/A
*-commutativeN/A
lower-*.f32N/A
lower-PI.f32N/A
lower-*.f32N/A
lower-PI.f3299.0
Applied rewrites99.0%
lift-*.f32N/A
*-commutativeN/A
lower-*.f3299.0
Applied rewrites99.1%
Final simplification99.1%
(FPCore (ux uy maxCos)
:precision binary32
(let* ((t_0 (+ (- 1.0 ux) (* ux maxCos))))
(if (<=
(* (cos (* (* uy 2.0) (PI))) (sqrt (- 1.0 (* t_0 t_0))))
0.017999999225139618)
(sqrt (* (- (fma -1.0 (- maxCos 1.0) 1.0) maxCos) ux))
(sqrt
(- (* (- (fma maxCos ux 1.0) ux) (- ux (fma maxCos ux 1.0))) -1.0)))))\begin{array}{l}
\\
\begin{array}{l}
t_0 := \left(1 - ux\right) + ux \cdot maxCos\\
\mathbf{if}\;\cos \left(\left(uy \cdot 2\right) \cdot \mathsf{PI}\left(\right)\right) \cdot \sqrt{1 - t\_0 \cdot t\_0} \leq 0.017999999225139618:\\
\;\;\;\;\sqrt{\left(\mathsf{fma}\left(-1, maxCos - 1, 1\right) - maxCos\right) \cdot ux}\\
\mathbf{else}:\\
\;\;\;\;\sqrt{\left(\mathsf{fma}\left(maxCos, ux, 1\right) - ux\right) \cdot \left(ux - \mathsf{fma}\left(maxCos, ux, 1\right)\right) - -1}\\
\end{array}
\end{array}
if (*.f32 (cos.f32 (*.f32 (*.f32 uy #s(literal 2 binary32)) (PI.f32))) (sqrt.f32 (-.f32 #s(literal 1 binary32) (*.f32 (+.f32 (-.f32 #s(literal 1 binary32) ux) (*.f32 ux maxCos)) (+.f32 (-.f32 #s(literal 1 binary32) ux) (*.f32 ux maxCos)))))) < 0.0179999992Initial program 38.7%
lift--.f32N/A
lift-*.f32N/A
fp-cancel-sub-sign-invN/A
+-commutativeN/A
*-commutativeN/A
lower-fma.f32N/A
lift-+.f32N/A
+-commutativeN/A
lift-*.f32N/A
*-commutativeN/A
lower-fma.f32N/A
lower-neg.f3238.8
lift-+.f32N/A
+-commutativeN/A
lift-*.f32N/A
*-commutativeN/A
lower-fma.f3238.8
Applied rewrites38.8%
Taylor expanded in uy around 0
lower-sqrt.f32N/A
+-commutativeN/A
lower-fma.f32N/A
lower--.f32N/A
+-commutativeN/A
lower-fma.f32N/A
lower--.f32N/A
+-commutativeN/A
lower-fma.f3232.8
Applied rewrites32.8%
Taylor expanded in ux around 0
Applied rewrites74.5%
if 0.0179999992 < (*.f32 (cos.f32 (*.f32 (*.f32 uy #s(literal 2 binary32)) (PI.f32))) (sqrt.f32 (-.f32 #s(literal 1 binary32) (*.f32 (+.f32 (-.f32 #s(literal 1 binary32) ux) (*.f32 ux maxCos)) (+.f32 (-.f32 #s(literal 1 binary32) ux) (*.f32 ux maxCos)))))) Initial program 89.2%
lift--.f32N/A
lift-*.f32N/A
fp-cancel-sub-sign-invN/A
+-commutativeN/A
*-commutativeN/A
lower-fma.f32N/A
lift-+.f32N/A
+-commutativeN/A
lift-*.f32N/A
*-commutativeN/A
lower-fma.f32N/A
lower-neg.f3288.7
lift-+.f32N/A
+-commutativeN/A
lift-*.f32N/A
*-commutativeN/A
lower-fma.f3288.7
Applied rewrites88.7%
Taylor expanded in uy around 0
lower-sqrt.f32N/A
+-commutativeN/A
lower-fma.f32N/A
lower--.f32N/A
+-commutativeN/A
lower-fma.f32N/A
lower--.f32N/A
+-commutativeN/A
lower-fma.f3275.7
Applied rewrites75.7%
Applied rewrites76.2%
(FPCore (ux uy maxCos)
:precision binary32
(if (<= (cos (* (* uy 2.0) (PI))) 0.9999650120735168)
(*
(fma (* (* uy uy) -2.0) (* (PI) (PI)) 1.0)
(sqrt (* (fma -2.0 maxCos 2.0) ux)))
(sqrt
(*
(-
(- (fma (* (- 1.0 maxCos) ux) (- maxCos 1.0) (- (- maxCos) -1.0)) -1.0)
maxCos)
ux))))\begin{array}{l}
\\
\begin{array}{l}
\mathbf{if}\;\cos \left(\left(uy \cdot 2\right) \cdot \mathsf{PI}\left(\right)\right) \leq 0.9999650120735168:\\
\;\;\;\;\mathsf{fma}\left(\left(uy \cdot uy\right) \cdot -2, \mathsf{PI}\left(\right) \cdot \mathsf{PI}\left(\right), 1\right) \cdot \sqrt{\mathsf{fma}\left(-2, maxCos, 2\right) \cdot ux}\\
\mathbf{else}:\\
\;\;\;\;\sqrt{\left(\left(\mathsf{fma}\left(\left(1 - maxCos\right) \cdot ux, maxCos - 1, \left(-maxCos\right) - -1\right) - -1\right) - maxCos\right) \cdot ux}\\
\end{array}
\end{array}
if (cos.f32 (*.f32 (*.f32 uy #s(literal 2 binary32)) (PI.f32))) < 0.999965012Initial program 58.8%
Taylor expanded in uy around 0
+-commutativeN/A
associate-*r*N/A
lower-fma.f32N/A
*-commutativeN/A
lower-*.f32N/A
unpow2N/A
lower-*.f32N/A
unpow2N/A
lower-*.f32N/A
lower-PI.f32N/A
lower-PI.f3241.3
Applied rewrites41.3%
Taylor expanded in ux around 0
metadata-evalN/A
fp-cancel-sign-sub-invN/A
*-commutativeN/A
lower-*.f32N/A
+-commutativeN/A
lower-fma.f3254.2
Applied rewrites54.2%
if 0.999965012 < (cos.f32 (*.f32 (*.f32 uy #s(literal 2 binary32)) (PI.f32))) Initial program 59.2%
lift--.f32N/A
lift-*.f32N/A
fp-cancel-sub-sign-invN/A
+-commutativeN/A
*-commutativeN/A
lower-fma.f32N/A
lift-+.f32N/A
+-commutativeN/A
lift-*.f32N/A
*-commutativeN/A
lower-fma.f32N/A
lower-neg.f3259.0
lift-+.f32N/A
+-commutativeN/A
lift-*.f32N/A
*-commutativeN/A
lower-fma.f3259.0
Applied rewrites59.0%
Taylor expanded in uy around 0
lower-sqrt.f32N/A
+-commutativeN/A
lower-fma.f32N/A
lower--.f32N/A
+-commutativeN/A
lower-fma.f32N/A
lower--.f32N/A
+-commutativeN/A
lower-fma.f3258.0
Applied rewrites58.0%
Taylor expanded in ux around 0
Applied rewrites96.7%
Final simplification84.4%
(FPCore (ux uy maxCos)
:precision binary32
(*
(cos (* (* uy 2.0) (PI)))
(sqrt
(*
(- (- 2.0 (* (* (- maxCos 1.0) (+ -1.0 maxCos)) ux)) (* 2.0 maxCos))
ux))))\begin{array}{l}
\\
\cos \left(\left(uy \cdot 2\right) \cdot \mathsf{PI}\left(\right)\right) \cdot \sqrt{\left(\left(2 - \left(\left(maxCos - 1\right) \cdot \left(-1 + maxCos\right)\right) \cdot ux\right) - 2 \cdot maxCos\right) \cdot ux}
\end{array}
Initial program 59.1%
lift--.f32N/A
lift-*.f32N/A
lift-+.f32N/A
distribute-rgt-inN/A
associate--r+N/A
lower--.f32N/A
lower--.f32N/A
*-commutativeN/A
lower-*.f32N/A
lift-+.f32N/A
+-commutativeN/A
lift-*.f32N/A
*-commutativeN/A
lower-fma.f32N/A
*-commutativeN/A
lift-*.f32N/A
*-commutativeN/A
Applied rewrites57.6%
Taylor expanded in ux around 0
*-commutativeN/A
lower-*.f32N/A
Applied rewrites98.9%
Final simplification98.9%
(FPCore (ux uy maxCos)
:precision binary32
(if (<= uy 7.999999797903001e-5)
(sqrt
(*
(-
(- (fma (* (- 1.0 maxCos) ux) (- maxCos 1.0) (- (- maxCos) -1.0)) -1.0)
maxCos)
ux))
(* (sin (fma (- (PI)) (* 2.0 uy) (/ (PI) 2.0))) (sqrt (* (- 2.0 ux) ux)))))\begin{array}{l}
\\
\begin{array}{l}
\mathbf{if}\;uy \leq 7.999999797903001 \cdot 10^{-5}:\\
\;\;\;\;\sqrt{\left(\left(\mathsf{fma}\left(\left(1 - maxCos\right) \cdot ux, maxCos - 1, \left(-maxCos\right) - -1\right) - -1\right) - maxCos\right) \cdot ux}\\
\mathbf{else}:\\
\;\;\;\;\sin \left(\mathsf{fma}\left(-\mathsf{PI}\left(\right), 2 \cdot uy, \frac{\mathsf{PI}\left(\right)}{2}\right)\right) \cdot \sqrt{\left(2 - ux\right) \cdot ux}\\
\end{array}
\end{array}
if uy < 7.9999998e-5Initial program 58.5%
lift--.f32N/A
lift-*.f32N/A
fp-cancel-sub-sign-invN/A
+-commutativeN/A
*-commutativeN/A
lower-fma.f32N/A
lift-+.f32N/A
+-commutativeN/A
lift-*.f32N/A
*-commutativeN/A
lower-fma.f32N/A
lower-neg.f3258.3
lift-+.f32N/A
+-commutativeN/A
lift-*.f32N/A
*-commutativeN/A
lower-fma.f3258.3
Applied rewrites58.3%
Taylor expanded in uy around 0
lower-sqrt.f32N/A
+-commutativeN/A
lower-fma.f32N/A
lower--.f32N/A
+-commutativeN/A
lower-fma.f32N/A
lower--.f32N/A
+-commutativeN/A
lower-fma.f3258.0
Applied rewrites58.0%
Taylor expanded in ux around 0
Applied rewrites99.1%
if 7.9999998e-5 < uy Initial program 59.9%
Taylor expanded in ux around 0
*-commutativeN/A
lower-*.f32N/A
fp-cancel-sub-sign-invN/A
metadata-evalN/A
+-commutativeN/A
lower-fma.f32N/A
associate-*r*N/A
mul-1-negN/A
fp-cancel-sub-signN/A
lower--.f32N/A
*-commutativeN/A
lower-*.f32N/A
lower-pow.f32N/A
lower--.f3298.3
Applied rewrites98.3%
lift-cos.f32N/A
cos-neg-revN/A
sin-+PI/2-revN/A
lower-sin.f32N/A
lift-*.f32N/A
*-commutativeN/A
distribute-lft-neg-inN/A
lower-fma.f32N/A
lower-neg.f32N/A
lift-*.f32N/A
*-commutativeN/A
lower-*.f32N/A
lift-PI.f32N/A
lower-/.f3298.6
Applied rewrites98.6%
Taylor expanded in maxCos around 0
Applied rewrites93.3%
Final simplification96.7%
(FPCore (ux uy maxCos) :precision binary32 (* (cos (* (* uy 2.0) (PI))) (sqrt (- (* (- 2.0 ux) ux) (* (* (fma -2.0 ux 2.0) ux) maxCos)))))
\begin{array}{l}
\\
\cos \left(\left(uy \cdot 2\right) \cdot \mathsf{PI}\left(\right)\right) \cdot \sqrt{\left(2 - ux\right) \cdot ux - \left(\mathsf{fma}\left(-2, ux, 2\right) \cdot ux\right) \cdot maxCos}
\end{array}
Initial program 59.1%
Taylor expanded in ux around 0
*-commutativeN/A
lower-*.f32N/A
fp-cancel-sub-sign-invN/A
metadata-evalN/A
+-commutativeN/A
lower-fma.f32N/A
associate-*r*N/A
mul-1-negN/A
fp-cancel-sub-signN/A
lower--.f32N/A
*-commutativeN/A
lower-*.f32N/A
lower-pow.f32N/A
lower--.f3298.9
Applied rewrites98.9%
Taylor expanded in maxCos around 0
Applied rewrites97.3%
Final simplification97.3%
(FPCore (ux uy maxCos) :precision binary32 (* (cos (* (* uy 2.0) (PI))) (sqrt (- (* (- (fma (- maxCos 1.0) ux 2.0) maxCos) ux) (* maxCos ux)))))
\begin{array}{l}
\\
\cos \left(\left(uy \cdot 2\right) \cdot \mathsf{PI}\left(\right)\right) \cdot \sqrt{\left(\mathsf{fma}\left(maxCos - 1, ux, 2\right) - maxCos\right) \cdot ux - maxCos \cdot ux}
\end{array}
Initial program 59.1%
lift--.f32N/A
lift-*.f32N/A
lift-+.f32N/A
distribute-rgt-inN/A
associate--r+N/A
lower--.f32N/A
lower--.f32N/A
*-commutativeN/A
lower-*.f32N/A
lift-+.f32N/A
+-commutativeN/A
lift-*.f32N/A
*-commutativeN/A
lower-fma.f32N/A
*-commutativeN/A
lift-*.f32N/A
*-commutativeN/A
Applied rewrites57.6%
Taylor expanded in ux around -inf
*-commutativeN/A
lower-*.f32N/A
Applied rewrites98.8%
Taylor expanded in ux around 0
lower-*.f3296.7
Applied rewrites96.7%
Taylor expanded in ux around 0
*-commutativeN/A
lower-*.f32N/A
lower--.f32N/A
+-commutativeN/A
*-commutativeN/A
lower-fma.f32N/A
lower--.f3296.8
Applied rewrites96.8%
(FPCore (ux uy maxCos)
:precision binary32
(if (<= uy 7.999999797903001e-5)
(sqrt
(*
(-
(- (fma (* (- 1.0 maxCos) ux) (- maxCos 1.0) (- (- maxCos) -1.0)) -1.0)
maxCos)
ux))
(* (cos (* (* uy 2.0) (PI))) (sqrt (* (- 2.0 ux) ux)))))\begin{array}{l}
\\
\begin{array}{l}
\mathbf{if}\;uy \leq 7.999999797903001 \cdot 10^{-5}:\\
\;\;\;\;\sqrt{\left(\left(\mathsf{fma}\left(\left(1 - maxCos\right) \cdot ux, maxCos - 1, \left(-maxCos\right) - -1\right) - -1\right) - maxCos\right) \cdot ux}\\
\mathbf{else}:\\
\;\;\;\;\cos \left(\left(uy \cdot 2\right) \cdot \mathsf{PI}\left(\right)\right) \cdot \sqrt{\left(2 - ux\right) \cdot ux}\\
\end{array}
\end{array}
if uy < 7.9999998e-5Initial program 58.5%
lift--.f32N/A
lift-*.f32N/A
fp-cancel-sub-sign-invN/A
+-commutativeN/A
*-commutativeN/A
lower-fma.f32N/A
lift-+.f32N/A
+-commutativeN/A
lift-*.f32N/A
*-commutativeN/A
lower-fma.f32N/A
lower-neg.f3258.3
lift-+.f32N/A
+-commutativeN/A
lift-*.f32N/A
*-commutativeN/A
lower-fma.f3258.3
Applied rewrites58.3%
Taylor expanded in uy around 0
lower-sqrt.f32N/A
+-commutativeN/A
lower-fma.f32N/A
lower--.f32N/A
+-commutativeN/A
lower-fma.f32N/A
lower--.f32N/A
+-commutativeN/A
lower-fma.f3258.0
Applied rewrites58.0%
Taylor expanded in ux around 0
Applied rewrites99.1%
if 7.9999998e-5 < uy Initial program 59.9%
Taylor expanded in ux around 0
*-commutativeN/A
lower-*.f32N/A
fp-cancel-sub-sign-invN/A
metadata-evalN/A
+-commutativeN/A
lower-fma.f32N/A
associate-*r*N/A
mul-1-negN/A
fp-cancel-sub-signN/A
lower--.f32N/A
*-commutativeN/A
lower-*.f32N/A
lower-pow.f32N/A
lower--.f3298.3
Applied rewrites98.3%
Taylor expanded in maxCos around 0
Applied rewrites93.0%
Final simplification96.6%
(FPCore (ux uy maxCos)
:precision binary32
(let* ((t_0 (+ (- 1.0 ux) (* ux maxCos))))
(if (<= (* t_0 t_0) 0.9997000098228455)
(sqrt (fma (- ux (fma maxCos ux 1.0)) (- (fma maxCos ux 1.0) ux) 1.0))
(sqrt (* (- (fma -1.0 (- maxCos 1.0) 1.0) maxCos) ux)))))
float code(float ux, float uy, float maxCos) {
float t_0 = (1.0f - ux) + (ux * maxCos);
float tmp;
if ((t_0 * t_0) <= 0.9997000098228455f) {
tmp = sqrtf(fmaf((ux - fmaf(maxCos, ux, 1.0f)), (fmaf(maxCos, ux, 1.0f) - ux), 1.0f));
} else {
tmp = sqrtf(((fmaf(-1.0f, (maxCos - 1.0f), 1.0f) - maxCos) * ux));
}
return tmp;
}
function code(ux, uy, maxCos) t_0 = Float32(Float32(Float32(1.0) - ux) + Float32(ux * maxCos)) tmp = Float32(0.0) if (Float32(t_0 * t_0) <= Float32(0.9997000098228455)) tmp = sqrt(fma(Float32(ux - fma(maxCos, ux, Float32(1.0))), Float32(fma(maxCos, ux, Float32(1.0)) - ux), Float32(1.0))); else tmp = sqrt(Float32(Float32(fma(Float32(-1.0), Float32(maxCos - Float32(1.0)), Float32(1.0)) - maxCos) * ux)); end return tmp end
\begin{array}{l}
\\
\begin{array}{l}
t_0 := \left(1 - ux\right) + ux \cdot maxCos\\
\mathbf{if}\;t\_0 \cdot t\_0 \leq 0.9997000098228455:\\
\;\;\;\;\sqrt{\mathsf{fma}\left(ux - \mathsf{fma}\left(maxCos, ux, 1\right), \mathsf{fma}\left(maxCos, ux, 1\right) - ux, 1\right)}\\
\mathbf{else}:\\
\;\;\;\;\sqrt{\left(\mathsf{fma}\left(-1, maxCos - 1, 1\right) - maxCos\right) \cdot ux}\\
\end{array}
\end{array}
if (*.f32 (+.f32 (-.f32 #s(literal 1 binary32) ux) (*.f32 ux maxCos)) (+.f32 (-.f32 #s(literal 1 binary32) ux) (*.f32 ux maxCos))) < 0.99970001Initial program 88.6%
lift--.f32N/A
lift-*.f32N/A
fp-cancel-sub-sign-invN/A
+-commutativeN/A
*-commutativeN/A
lower-fma.f32N/A
lift-+.f32N/A
+-commutativeN/A
lift-*.f32N/A
*-commutativeN/A
lower-fma.f32N/A
lower-neg.f3288.2
lift-+.f32N/A
+-commutativeN/A
lift-*.f32N/A
*-commutativeN/A
lower-fma.f3288.2
Applied rewrites88.2%
Taylor expanded in uy around 0
lower-sqrt.f32N/A
+-commutativeN/A
lower-fma.f32N/A
lower--.f32N/A
+-commutativeN/A
lower-fma.f32N/A
lower--.f32N/A
+-commutativeN/A
lower-fma.f3272.2
Applied rewrites72.2%
if 0.99970001 < (*.f32 (+.f32 (-.f32 #s(literal 1 binary32) ux) (*.f32 ux maxCos)) (+.f32 (-.f32 #s(literal 1 binary32) ux) (*.f32 ux maxCos))) Initial program 36.8%
lift--.f32N/A
lift-*.f32N/A
fp-cancel-sub-sign-invN/A
+-commutativeN/A
*-commutativeN/A
lower-fma.f32N/A
lift-+.f32N/A
+-commutativeN/A
lift-*.f32N/A
*-commutativeN/A
lower-fma.f32N/A
lower-neg.f3236.8
lift-+.f32N/A
+-commutativeN/A
lift-*.f32N/A
*-commutativeN/A
lower-fma.f3236.8
Applied rewrites36.8%
Taylor expanded in uy around 0
lower-sqrt.f32N/A
+-commutativeN/A
lower-fma.f32N/A
lower--.f32N/A
+-commutativeN/A
lower-fma.f32N/A
lower--.f32N/A
+-commutativeN/A
lower-fma.f3233.3
Applied rewrites33.3%
Taylor expanded in ux around 0
Applied rewrites77.2%
(FPCore (ux uy maxCos)
:precision binary32
(let* ((t_0 (+ (- 1.0 ux) (* ux maxCos))))
(if (<= (* t_0 t_0) 0.9997000098228455)
(sqrt (fma (- ux 1.0) (- (fma maxCos ux 1.0) ux) 1.0))
(sqrt (* (- (fma -1.0 (- maxCos 1.0) 1.0) maxCos) ux)))))
float code(float ux, float uy, float maxCos) {
float t_0 = (1.0f - ux) + (ux * maxCos);
float tmp;
if ((t_0 * t_0) <= 0.9997000098228455f) {
tmp = sqrtf(fmaf((ux - 1.0f), (fmaf(maxCos, ux, 1.0f) - ux), 1.0f));
} else {
tmp = sqrtf(((fmaf(-1.0f, (maxCos - 1.0f), 1.0f) - maxCos) * ux));
}
return tmp;
}
function code(ux, uy, maxCos) t_0 = Float32(Float32(Float32(1.0) - ux) + Float32(ux * maxCos)) tmp = Float32(0.0) if (Float32(t_0 * t_0) <= Float32(0.9997000098228455)) tmp = sqrt(fma(Float32(ux - Float32(1.0)), Float32(fma(maxCos, ux, Float32(1.0)) - ux), Float32(1.0))); else tmp = sqrt(Float32(Float32(fma(Float32(-1.0), Float32(maxCos - Float32(1.0)), Float32(1.0)) - maxCos) * ux)); end return tmp end
\begin{array}{l}
\\
\begin{array}{l}
t_0 := \left(1 - ux\right) + ux \cdot maxCos\\
\mathbf{if}\;t\_0 \cdot t\_0 \leq 0.9997000098228455:\\
\;\;\;\;\sqrt{\mathsf{fma}\left(ux - 1, \mathsf{fma}\left(maxCos, ux, 1\right) - ux, 1\right)}\\
\mathbf{else}:\\
\;\;\;\;\sqrt{\left(\mathsf{fma}\left(-1, maxCos - 1, 1\right) - maxCos\right) \cdot ux}\\
\end{array}
\end{array}
if (*.f32 (+.f32 (-.f32 #s(literal 1 binary32) ux) (*.f32 ux maxCos)) (+.f32 (-.f32 #s(literal 1 binary32) ux) (*.f32 ux maxCos))) < 0.99970001Initial program 88.6%
lift--.f32N/A
lift-*.f32N/A
fp-cancel-sub-sign-invN/A
+-commutativeN/A
*-commutativeN/A
lower-fma.f32N/A
lift-+.f32N/A
+-commutativeN/A
lift-*.f32N/A
*-commutativeN/A
lower-fma.f32N/A
lower-neg.f3288.2
lift-+.f32N/A
+-commutativeN/A
lift-*.f32N/A
*-commutativeN/A
lower-fma.f3288.2
Applied rewrites88.2%
Taylor expanded in uy around 0
lower-sqrt.f32N/A
+-commutativeN/A
lower-fma.f32N/A
lower--.f32N/A
+-commutativeN/A
lower-fma.f32N/A
lower--.f32N/A
+-commutativeN/A
lower-fma.f3272.2
Applied rewrites72.2%
Taylor expanded in maxCos around 0
Applied rewrites68.2%
if 0.99970001 < (*.f32 (+.f32 (-.f32 #s(literal 1 binary32) ux) (*.f32 ux maxCos)) (+.f32 (-.f32 #s(literal 1 binary32) ux) (*.f32 ux maxCos))) Initial program 36.8%
lift--.f32N/A
lift-*.f32N/A
fp-cancel-sub-sign-invN/A
+-commutativeN/A
*-commutativeN/A
lower-fma.f32N/A
lift-+.f32N/A
+-commutativeN/A
lift-*.f32N/A
*-commutativeN/A
lower-fma.f32N/A
lower-neg.f3236.8
lift-+.f32N/A
+-commutativeN/A
lift-*.f32N/A
*-commutativeN/A
lower-fma.f3236.8
Applied rewrites36.8%
Taylor expanded in uy around 0
lower-sqrt.f32N/A
+-commutativeN/A
lower-fma.f32N/A
lower--.f32N/A
+-commutativeN/A
lower-fma.f32N/A
lower--.f32N/A
+-commutativeN/A
lower-fma.f3233.3
Applied rewrites33.3%
Taylor expanded in ux around 0
Applied rewrites77.2%
(FPCore (ux uy maxCos)
:precision binary32
(let* ((t_0 (+ (- 1.0 ux) (* ux maxCos))))
(if (<= (* t_0 t_0) 0.9997000098228455)
(sqrt (fma (- ux 1.0) (- 1.0 ux) 1.0))
(sqrt (* (- (fma -1.0 (- maxCos 1.0) 1.0) maxCos) ux)))))
float code(float ux, float uy, float maxCos) {
float t_0 = (1.0f - ux) + (ux * maxCos);
float tmp;
if ((t_0 * t_0) <= 0.9997000098228455f) {
tmp = sqrtf(fmaf((ux - 1.0f), (1.0f - ux), 1.0f));
} else {
tmp = sqrtf(((fmaf(-1.0f, (maxCos - 1.0f), 1.0f) - maxCos) * ux));
}
return tmp;
}
function code(ux, uy, maxCos) t_0 = Float32(Float32(Float32(1.0) - ux) + Float32(ux * maxCos)) tmp = Float32(0.0) if (Float32(t_0 * t_0) <= Float32(0.9997000098228455)) tmp = sqrt(fma(Float32(ux - Float32(1.0)), Float32(Float32(1.0) - ux), Float32(1.0))); else tmp = sqrt(Float32(Float32(fma(Float32(-1.0), Float32(maxCos - Float32(1.0)), Float32(1.0)) - maxCos) * ux)); end return tmp end
\begin{array}{l}
\\
\begin{array}{l}
t_0 := \left(1 - ux\right) + ux \cdot maxCos\\
\mathbf{if}\;t\_0 \cdot t\_0 \leq 0.9997000098228455:\\
\;\;\;\;\sqrt{\mathsf{fma}\left(ux - 1, 1 - ux, 1\right)}\\
\mathbf{else}:\\
\;\;\;\;\sqrt{\left(\mathsf{fma}\left(-1, maxCos - 1, 1\right) - maxCos\right) \cdot ux}\\
\end{array}
\end{array}
if (*.f32 (+.f32 (-.f32 #s(literal 1 binary32) ux) (*.f32 ux maxCos)) (+.f32 (-.f32 #s(literal 1 binary32) ux) (*.f32 ux maxCos))) < 0.99970001Initial program 88.6%
lift--.f32N/A
lift-*.f32N/A
fp-cancel-sub-sign-invN/A
+-commutativeN/A
*-commutativeN/A
lower-fma.f32N/A
lift-+.f32N/A
+-commutativeN/A
lift-*.f32N/A
*-commutativeN/A
lower-fma.f32N/A
lower-neg.f3288.2
lift-+.f32N/A
+-commutativeN/A
lift-*.f32N/A
*-commutativeN/A
lower-fma.f3288.2
Applied rewrites88.2%
Taylor expanded in uy around 0
lower-sqrt.f32N/A
+-commutativeN/A
lower-fma.f32N/A
lower--.f32N/A
+-commutativeN/A
lower-fma.f32N/A
lower--.f32N/A
+-commutativeN/A
lower-fma.f3272.2
Applied rewrites72.2%
Taylor expanded in maxCos around 0
Applied rewrites67.9%
if 0.99970001 < (*.f32 (+.f32 (-.f32 #s(literal 1 binary32) ux) (*.f32 ux maxCos)) (+.f32 (-.f32 #s(literal 1 binary32) ux) (*.f32 ux maxCos))) Initial program 36.8%
lift--.f32N/A
lift-*.f32N/A
fp-cancel-sub-sign-invN/A
+-commutativeN/A
*-commutativeN/A
lower-fma.f32N/A
lift-+.f32N/A
+-commutativeN/A
lift-*.f32N/A
*-commutativeN/A
lower-fma.f32N/A
lower-neg.f3236.8
lift-+.f32N/A
+-commutativeN/A
lift-*.f32N/A
*-commutativeN/A
lower-fma.f3236.8
Applied rewrites36.8%
Taylor expanded in uy around 0
lower-sqrt.f32N/A
+-commutativeN/A
lower-fma.f32N/A
lower--.f32N/A
+-commutativeN/A
lower-fma.f32N/A
lower--.f32N/A
+-commutativeN/A
lower-fma.f3233.3
Applied rewrites33.3%
Taylor expanded in ux around 0
Applied rewrites77.2%
(FPCore (ux uy maxCos)
:precision binary32
(sqrt
(*
(-
(- (fma (* (- 1.0 maxCos) ux) (- maxCos 1.0) (- (- maxCos) -1.0)) -1.0)
maxCos)
ux)))
float code(float ux, float uy, float maxCos) {
return sqrtf((((fmaf(((1.0f - maxCos) * ux), (maxCos - 1.0f), (-maxCos - -1.0f)) - -1.0f) - maxCos) * ux));
}
function code(ux, uy, maxCos) return sqrt(Float32(Float32(Float32(fma(Float32(Float32(Float32(1.0) - maxCos) * ux), Float32(maxCos - Float32(1.0)), Float32(Float32(-maxCos) - Float32(-1.0))) - Float32(-1.0)) - maxCos) * ux)) end
\begin{array}{l}
\\
\sqrt{\left(\left(\mathsf{fma}\left(\left(1 - maxCos\right) \cdot ux, maxCos - 1, \left(-maxCos\right) - -1\right) - -1\right) - maxCos\right) \cdot ux}
\end{array}
Initial program 59.1%
lift--.f32N/A
lift-*.f32N/A
fp-cancel-sub-sign-invN/A
+-commutativeN/A
*-commutativeN/A
lower-fma.f32N/A
lift-+.f32N/A
+-commutativeN/A
lift-*.f32N/A
*-commutativeN/A
lower-fma.f32N/A
lower-neg.f3258.9
lift-+.f32N/A
+-commutativeN/A
lift-*.f32N/A
*-commutativeN/A
lower-fma.f3258.9
Applied rewrites58.9%
Taylor expanded in uy around 0
lower-sqrt.f32N/A
+-commutativeN/A
lower-fma.f32N/A
lower--.f32N/A
+-commutativeN/A
lower-fma.f32N/A
lower--.f32N/A
+-commutativeN/A
lower-fma.f3250.0
Applied rewrites50.0%
Taylor expanded in ux around 0
Applied rewrites80.5%
Final simplification80.5%
(FPCore (ux uy maxCos) :precision binary32 (sqrt (fma (- ux 1.0) (- 1.0 ux) 1.0)))
float code(float ux, float uy, float maxCos) {
return sqrtf(fmaf((ux - 1.0f), (1.0f - ux), 1.0f));
}
function code(ux, uy, maxCos) return sqrt(fma(Float32(ux - Float32(1.0)), Float32(Float32(1.0) - ux), Float32(1.0))) end
\begin{array}{l}
\\
\sqrt{\mathsf{fma}\left(ux - 1, 1 - ux, 1\right)}
\end{array}
Initial program 59.1%
lift--.f32N/A
lift-*.f32N/A
fp-cancel-sub-sign-invN/A
+-commutativeN/A
*-commutativeN/A
lower-fma.f32N/A
lift-+.f32N/A
+-commutativeN/A
lift-*.f32N/A
*-commutativeN/A
lower-fma.f32N/A
lower-neg.f3258.9
lift-+.f32N/A
+-commutativeN/A
lift-*.f32N/A
*-commutativeN/A
lower-fma.f3258.9
Applied rewrites58.9%
Taylor expanded in uy around 0
lower-sqrt.f32N/A
+-commutativeN/A
lower-fma.f32N/A
lower--.f32N/A
+-commutativeN/A
lower-fma.f32N/A
lower--.f32N/A
+-commutativeN/A
lower-fma.f3250.0
Applied rewrites50.0%
Taylor expanded in maxCos around 0
Applied rewrites47.8%
(FPCore (ux uy maxCos) :precision binary32 (* (* (sqrt -1.0) ux) maxCos))
float code(float ux, float uy, float maxCos) {
return (sqrtf(-1.0f) * 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(ux, uy, maxcos)
use fmin_fmax_functions
real(4), intent (in) :: ux
real(4), intent (in) :: uy
real(4), intent (in) :: maxcos
code = (sqrt((-1.0e0)) * ux) * maxcos
end function
function code(ux, uy, maxCos) return Float32(Float32(sqrt(Float32(-1.0)) * ux) * maxCos) end
function tmp = code(ux, uy, maxCos) tmp = (sqrt(single(-1.0)) * ux) * maxCos; end
\begin{array}{l}
\\
\left(\sqrt{-1} \cdot ux\right) \cdot maxCos
\end{array}
Initial program 59.1%
lift--.f32N/A
lift-*.f32N/A
fp-cancel-sub-sign-invN/A
+-commutativeN/A
*-commutativeN/A
lower-fma.f32N/A
lift-+.f32N/A
+-commutativeN/A
lift-*.f32N/A
*-commutativeN/A
lower-fma.f32N/A
lower-neg.f3258.9
lift-+.f32N/A
+-commutativeN/A
lift-*.f32N/A
*-commutativeN/A
lower-fma.f3258.9
Applied rewrites58.9%
Taylor expanded in uy around 0
lower-sqrt.f32N/A
+-commutativeN/A
lower-fma.f32N/A
lower--.f32N/A
+-commutativeN/A
lower-fma.f32N/A
lower--.f32N/A
+-commutativeN/A
lower-fma.f3250.0
Applied rewrites50.0%
Taylor expanded in maxCos around inf
Applied rewrites-0.0%
herbie shell --seed 2025010
(FPCore (ux uy maxCos)
:name "UniformSampleCone, x"
:precision binary32
:pre (and (and (and (<= 2.328306437e-10 ux) (<= ux 1.0)) (and (<= 2.328306437e-10 uy) (<= uy 1.0))) (and (<= 0.0 maxCos) (<= maxCos 1.0)))
(* (cos (* (* uy 2.0) (PI))) (sqrt (- 1.0 (* (+ (- 1.0 ux) (* ux maxCos)) (+ (- 1.0 ux) (* ux maxCos)))))))