
(FPCore (cosTheta alpha)
:precision binary32
(let* ((t_0 (- (* alpha alpha) 1.0)))
(/
t_0
(* (* PI (log (* alpha alpha))) (+ 1.0 (* (* t_0 cosTheta) cosTheta))))))
float code(float cosTheta, float alpha) {
float t_0 = (alpha * alpha) - 1.0f;
return t_0 / ((((float) M_PI) * logf((alpha * alpha))) * (1.0f + ((t_0 * cosTheta) * cosTheta)));
}
function code(cosTheta, alpha) t_0 = Float32(Float32(alpha * alpha) - Float32(1.0)) return Float32(t_0 / Float32(Float32(Float32(pi) * log(Float32(alpha * alpha))) * Float32(Float32(1.0) + Float32(Float32(t_0 * cosTheta) * cosTheta)))) end
function tmp = code(cosTheta, alpha) t_0 = (alpha * alpha) - single(1.0); tmp = t_0 / ((single(pi) * log((alpha * alpha))) * (single(1.0) + ((t_0 * cosTheta) * cosTheta))); end
\begin{array}{l}
\\
\begin{array}{l}
t_0 := \alpha \cdot \alpha - 1\\
\frac{t\_0}{\left(\pi \cdot \log \left(\alpha \cdot \alpha\right)\right) \cdot \left(1 + \left(t\_0 \cdot cosTheta\right) \cdot cosTheta\right)}
\end{array}
\end{array}
Sampling outcomes in binary32 precision:
Herbie found 10 alternatives:
| Alternative | Accuracy | Speedup |
|---|
(FPCore (cosTheta alpha)
:precision binary32
(let* ((t_0 (- (* alpha alpha) 1.0)))
(/
t_0
(* (* PI (log (* alpha alpha))) (+ 1.0 (* (* t_0 cosTheta) cosTheta))))))
float code(float cosTheta, float alpha) {
float t_0 = (alpha * alpha) - 1.0f;
return t_0 / ((((float) M_PI) * logf((alpha * alpha))) * (1.0f + ((t_0 * cosTheta) * cosTheta)));
}
function code(cosTheta, alpha) t_0 = Float32(Float32(alpha * alpha) - Float32(1.0)) return Float32(t_0 / Float32(Float32(Float32(pi) * log(Float32(alpha * alpha))) * Float32(Float32(1.0) + Float32(Float32(t_0 * cosTheta) * cosTheta)))) end
function tmp = code(cosTheta, alpha) t_0 = (alpha * alpha) - single(1.0); tmp = t_0 / ((single(pi) * log((alpha * alpha))) * (single(1.0) + ((t_0 * cosTheta) * cosTheta))); end
\begin{array}{l}
\\
\begin{array}{l}
t_0 := \alpha \cdot \alpha - 1\\
\frac{t\_0}{\left(\pi \cdot \log \left(\alpha \cdot \alpha\right)\right) \cdot \left(1 + \left(t\_0 \cdot cosTheta\right) \cdot cosTheta\right)}
\end{array}
\end{array}
(FPCore (cosTheta alpha)
:precision binary32
(let* ((t_0 (+ (* alpha alpha) -1.0)))
(/
t_0
(* (* PI (log (* alpha alpha))) (+ 1.0 (* t_0 (* cosTheta cosTheta)))))))
float code(float cosTheta, float alpha) {
float t_0 = (alpha * alpha) + -1.0f;
return t_0 / ((((float) M_PI) * logf((alpha * alpha))) * (1.0f + (t_0 * (cosTheta * cosTheta))));
}
function code(cosTheta, alpha) t_0 = Float32(Float32(alpha * alpha) + Float32(-1.0)) return Float32(t_0 / Float32(Float32(Float32(pi) * log(Float32(alpha * alpha))) * Float32(Float32(1.0) + Float32(t_0 * Float32(cosTheta * cosTheta))))) end
function tmp = code(cosTheta, alpha) t_0 = (alpha * alpha) + single(-1.0); tmp = t_0 / ((single(pi) * log((alpha * alpha))) * (single(1.0) + (t_0 * (cosTheta * cosTheta)))); end
\begin{array}{l}
\\
\begin{array}{l}
t_0 := \alpha \cdot \alpha + -1\\
\frac{t\_0}{\left(\pi \cdot \log \left(\alpha \cdot \alpha\right)\right) \cdot \left(1 + t\_0 \cdot \left(cosTheta \cdot cosTheta\right)\right)}
\end{array}
\end{array}
Initial program 98.5%
associate-*l*N/A
*-lowering-*.f32N/A
sub-negN/A
+-lowering-+.f32N/A
*-lowering-*.f32N/A
metadata-evalN/A
*-lowering-*.f3298.5%
Applied egg-rr98.5%
Final simplification98.5%
(FPCore (cosTheta alpha)
:precision binary32
(let* ((t_0 (+ (* alpha alpha) -1.0)))
(/
t_0
(* (* PI (log (* alpha alpha))) (+ 1.0 (* cosTheta (* t_0 cosTheta)))))))
float code(float cosTheta, float alpha) {
float t_0 = (alpha * alpha) + -1.0f;
return t_0 / ((((float) M_PI) * logf((alpha * alpha))) * (1.0f + (cosTheta * (t_0 * cosTheta))));
}
function code(cosTheta, alpha) t_0 = Float32(Float32(alpha * alpha) + Float32(-1.0)) return Float32(t_0 / Float32(Float32(Float32(pi) * log(Float32(alpha * alpha))) * Float32(Float32(1.0) + Float32(cosTheta * Float32(t_0 * cosTheta))))) end
function tmp = code(cosTheta, alpha) t_0 = (alpha * alpha) + single(-1.0); tmp = t_0 / ((single(pi) * log((alpha * alpha))) * (single(1.0) + (cosTheta * (t_0 * cosTheta)))); end
\begin{array}{l}
\\
\begin{array}{l}
t_0 := \alpha \cdot \alpha + -1\\
\frac{t\_0}{\left(\pi \cdot \log \left(\alpha \cdot \alpha\right)\right) \cdot \left(1 + cosTheta \cdot \left(t\_0 \cdot cosTheta\right)\right)}
\end{array}
\end{array}
Initial program 98.5%
Final simplification98.5%
(FPCore (cosTheta alpha) :precision binary32 (/ (+ (* alpha alpha) -1.0) (* (* PI (log (* alpha alpha))) (- 1.0 (* cosTheta cosTheta)))))
float code(float cosTheta, float alpha) {
return ((alpha * alpha) + -1.0f) / ((((float) M_PI) * logf((alpha * alpha))) * (1.0f - (cosTheta * cosTheta)));
}
function code(cosTheta, alpha) return Float32(Float32(Float32(alpha * alpha) + Float32(-1.0)) / Float32(Float32(Float32(pi) * log(Float32(alpha * alpha))) * Float32(Float32(1.0) - Float32(cosTheta * cosTheta)))) end
function tmp = code(cosTheta, alpha) tmp = ((alpha * alpha) + single(-1.0)) / ((single(pi) * log((alpha * alpha))) * (single(1.0) - (cosTheta * cosTheta))); end
\begin{array}{l}
\\
\frac{\alpha \cdot \alpha + -1}{\left(\pi \cdot \log \left(\alpha \cdot \alpha\right)\right) \cdot \left(1 - cosTheta \cdot cosTheta\right)}
\end{array}
Initial program 98.5%
Taylor expanded in alpha around 0
mul-1-negN/A
neg-lowering-neg.f3297.9%
Simplified97.9%
Taylor expanded in cosTheta around 0
associate-*r*N/A
distribute-lft1-inN/A
+-commutativeN/A
*-lowering-*.f32N/A
mul-1-negN/A
unsub-negN/A
--lowering--.f32N/A
unpow2N/A
*-lowering-*.f32N/A
*-lowering-*.f32N/A
PI-lowering-PI.f32N/A
log-lowering-log.f32N/A
unpow2N/A
*-lowering-*.f3297.9%
Simplified97.9%
Final simplification97.9%
(FPCore (cosTheta alpha) :precision binary32 (/ (+ (* alpha alpha) -1.0) (* (* (- 1.0 (* cosTheta cosTheta)) (log alpha)) (* PI 2.0))))
float code(float cosTheta, float alpha) {
return ((alpha * alpha) + -1.0f) / (((1.0f - (cosTheta * cosTheta)) * logf(alpha)) * (((float) M_PI) * 2.0f));
}
function code(cosTheta, alpha) return Float32(Float32(Float32(alpha * alpha) + Float32(-1.0)) / Float32(Float32(Float32(Float32(1.0) - Float32(cosTheta * cosTheta)) * log(alpha)) * Float32(Float32(pi) * Float32(2.0)))) end
function tmp = code(cosTheta, alpha) tmp = ((alpha * alpha) + single(-1.0)) / (((single(1.0) - (cosTheta * cosTheta)) * log(alpha)) * (single(pi) * single(2.0))); end
\begin{array}{l}
\\
\frac{\alpha \cdot \alpha + -1}{\left(\left(1 - cosTheta \cdot cosTheta\right) \cdot \log \alpha\right) \cdot \left(\pi \cdot 2\right)}
\end{array}
Initial program 98.5%
Taylor expanded in alpha around 0
associate-*r*N/A
*-commutativeN/A
*-lowering-*.f32N/A
*-lowering-*.f32N/A
log-lowering-log.f32N/A
mul-1-negN/A
unsub-negN/A
--lowering--.f32N/A
unpow2N/A
*-lowering-*.f32N/A
*-commutativeN/A
*-lowering-*.f32N/A
PI-lowering-PI.f3297.8%
Simplified97.8%
Final simplification97.8%
(FPCore (cosTheta alpha) :precision binary32 (* (/ (+ (* alpha alpha) -1.0) (* PI (log (* alpha alpha)))) (+ 1.0 (* cosTheta cosTheta))))
float code(float cosTheta, float alpha) {
return (((alpha * alpha) + -1.0f) / (((float) M_PI) * logf((alpha * alpha)))) * (1.0f + (cosTheta * cosTheta));
}
function code(cosTheta, alpha) return Float32(Float32(Float32(Float32(alpha * alpha) + Float32(-1.0)) / Float32(Float32(pi) * log(Float32(alpha * alpha)))) * Float32(Float32(1.0) + Float32(cosTheta * cosTheta))) end
function tmp = code(cosTheta, alpha) tmp = (((alpha * alpha) + single(-1.0)) / (single(pi) * log((alpha * alpha)))) * (single(1.0) + (cosTheta * cosTheta)); end
\begin{array}{l}
\\
\frac{\alpha \cdot \alpha + -1}{\pi \cdot \log \left(\alpha \cdot \alpha\right)} \cdot \left(1 + cosTheta \cdot cosTheta\right)
\end{array}
Initial program 98.5%
Taylor expanded in alpha around 0
mul-1-negN/A
neg-lowering-neg.f3297.9%
Simplified97.9%
Taylor expanded in cosTheta around 0
associate--l+N/A
associate-/l*N/A
div-subN/A
*-lft-identityN/A
distribute-rgt-outN/A
*-lowering-*.f32N/A
Simplified97.1%
Final simplification97.1%
(FPCore (cosTheta alpha) :precision binary32 (/ (+ (* alpha alpha) -1.0) (* PI (log (* alpha alpha)))))
float code(float cosTheta, float alpha) {
return ((alpha * alpha) + -1.0f) / (((float) M_PI) * logf((alpha * alpha)));
}
function code(cosTheta, alpha) return Float32(Float32(Float32(alpha * alpha) + Float32(-1.0)) / Float32(Float32(pi) * log(Float32(alpha * alpha)))) end
function tmp = code(cosTheta, alpha) tmp = ((alpha * alpha) + single(-1.0)) / (single(pi) * log((alpha * alpha))); end
\begin{array}{l}
\\
\frac{\alpha \cdot \alpha + -1}{\pi \cdot \log \left(\alpha \cdot \alpha\right)}
\end{array}
Initial program 98.5%
Taylor expanded in cosTheta around 0
*-lowering-*.f32N/A
PI-lowering-PI.f32N/A
log-lowering-log.f32N/A
unpow2N/A
*-lowering-*.f3295.6%
Simplified95.6%
Final simplification95.6%
(FPCore (cosTheta alpha) :precision binary32 (/ (+ (* alpha alpha) -1.0) (* 2.0 (* PI (log alpha)))))
float code(float cosTheta, float alpha) {
return ((alpha * alpha) + -1.0f) / (2.0f * (((float) M_PI) * logf(alpha)));
}
function code(cosTheta, alpha) return Float32(Float32(Float32(alpha * alpha) + Float32(-1.0)) / Float32(Float32(2.0) * Float32(Float32(pi) * log(alpha)))) end
function tmp = code(cosTheta, alpha) tmp = ((alpha * alpha) + single(-1.0)) / (single(2.0) * (single(pi) * log(alpha))); end
\begin{array}{l}
\\
\frac{\alpha \cdot \alpha + -1}{2 \cdot \left(\pi \cdot \log \alpha\right)}
\end{array}
Initial program 98.5%
Taylor expanded in alpha around 0
distribute-lft-outN/A
*-lowering-*.f32N/A
associate-*r*N/A
associate-*r*N/A
*-commutativeN/A
distribute-lft-outN/A
*-lowering-*.f32N/A
*-lowering-*.f32N/A
PI-lowering-PI.f32N/A
log-lowering-log.f32N/A
+-lowering-+.f32N/A
Simplified98.4%
Taylor expanded in cosTheta around 0
*-lowering-*.f32N/A
*-lowering-*.f32N/A
PI-lowering-PI.f32N/A
log-lowering-log.f3295.5%
Simplified95.5%
Final simplification95.5%
(FPCore (cosTheta alpha) :precision binary32 (/ (/ -0.5 PI) (log alpha)))
float code(float cosTheta, float alpha) {
return (-0.5f / ((float) M_PI)) / logf(alpha);
}
function code(cosTheta, alpha) return Float32(Float32(Float32(-0.5) / Float32(pi)) / log(alpha)) end
function tmp = code(cosTheta, alpha) tmp = (single(-0.5) / single(pi)) / log(alpha); end
\begin{array}{l}
\\
\frac{\frac{-0.5}{\pi}}{\log \alpha}
\end{array}
Initial program 98.5%
Taylor expanded in alpha around inf
*-lowering-*.f32N/A
unpow2N/A
*-lowering-*.f32N/A
sub-negN/A
+-lowering-+.f32N/A
distribute-neg-fracN/A
metadata-evalN/A
/-lowering-/.f32N/A
unpow2N/A
*-lowering-*.f3298.4%
Simplified98.4%
Taylor expanded in cosTheta around 0
associate-/l*N/A
*-lowering-*.f32N/A
unpow2N/A
*-lowering-*.f32N/A
associate-/r*N/A
/-lowering-/.f32N/A
/-lowering-/.f32N/A
sub-negN/A
+-lowering-+.f32N/A
distribute-neg-fracN/A
metadata-evalN/A
/-lowering-/.f32N/A
unpow2N/A
*-lowering-*.f32N/A
PI-lowering-PI.f32N/A
log-lowering-log.f32N/A
unpow2N/A
*-lowering-*.f3295.5%
Simplified95.5%
Taylor expanded in alpha around inf
sub-negN/A
+-lowering-+.f32N/A
/-lowering-/.f32N/A
PI-lowering-PI.f32N/A
distribute-neg-fracN/A
metadata-evalN/A
/-lowering-/.f32N/A
*-commutativeN/A
*-lowering-*.f32N/A
PI-lowering-PI.f32N/A
unpow2N/A
*-lowering-*.f3295.3%
Simplified95.3%
Taylor expanded in alpha around 0
associate-/r*N/A
/-lowering-/.f32N/A
/-lowering-/.f32N/A
PI-lowering-PI.f32N/A
log-lowering-log.f3264.3%
Simplified64.3%
(FPCore (cosTheta alpha) :precision binary32 (/ -0.5 (* PI (log alpha))))
float code(float cosTheta, float alpha) {
return -0.5f / (((float) M_PI) * logf(alpha));
}
function code(cosTheta, alpha) return Float32(Float32(-0.5) / Float32(Float32(pi) * log(alpha))) end
function tmp = code(cosTheta, alpha) tmp = single(-0.5) / (single(pi) * log(alpha)); end
\begin{array}{l}
\\
\frac{-0.5}{\pi \cdot \log \alpha}
\end{array}
Initial program 98.5%
Taylor expanded in alpha around inf
*-lowering-*.f32N/A
unpow2N/A
*-lowering-*.f32N/A
sub-negN/A
+-lowering-+.f32N/A
distribute-neg-fracN/A
metadata-evalN/A
/-lowering-/.f32N/A
unpow2N/A
*-lowering-*.f3298.4%
Simplified98.4%
Taylor expanded in cosTheta around 0
associate-/l*N/A
*-lowering-*.f32N/A
unpow2N/A
*-lowering-*.f32N/A
associate-/r*N/A
/-lowering-/.f32N/A
/-lowering-/.f32N/A
sub-negN/A
+-lowering-+.f32N/A
distribute-neg-fracN/A
metadata-evalN/A
/-lowering-/.f32N/A
unpow2N/A
*-lowering-*.f32N/A
PI-lowering-PI.f32N/A
log-lowering-log.f32N/A
unpow2N/A
*-lowering-*.f3295.5%
Simplified95.5%
Taylor expanded in alpha around 0
/-lowering-/.f32N/A
*-lowering-*.f32N/A
PI-lowering-PI.f32N/A
log-lowering-log.f3264.2%
Simplified64.2%
(FPCore (cosTheta alpha) :precision binary32 (/ (/ 1.0 (* PI (/ 0.0 0.0))) (+ -1.0 (* cosTheta cosTheta))))
float code(float cosTheta, float alpha) {
return (1.0f / (((float) M_PI) * (0.0f / 0.0f))) / (-1.0f + (cosTheta * cosTheta));
}
function code(cosTheta, alpha) return Float32(Float32(Float32(1.0) / Float32(Float32(pi) * Float32(Float32(0.0) / Float32(0.0)))) / Float32(Float32(-1.0) + Float32(cosTheta * cosTheta))) end
function tmp = code(cosTheta, alpha) tmp = (single(1.0) / (single(pi) * (single(0.0) / single(0.0)))) / (single(-1.0) + (cosTheta * cosTheta)); end
\begin{array}{l}
\\
\frac{\frac{1}{\pi \cdot \frac{0}{0}}}{-1 + cosTheta \cdot cosTheta}
\end{array}
Initial program 98.5%
associate-/r*N/A
frac-2negN/A
/-lowering-/.f32N/A
Applied egg-rr-0.0%
Taylor expanded in alpha around 0
Simplified-0.0%
Taylor expanded in alpha around 0
Simplified-0.0%
Taylor expanded in cosTheta around 0
unpow2N/A
*-lowering-*.f32-0.0%
Simplified-0.0%
herbie shell --seed 2024191
(FPCore (cosTheta alpha)
:name "GTR1 distribution"
:precision binary32
:pre (and (and (<= 0.0 cosTheta) (<= cosTheta 1.0)) (and (<= 0.0001 alpha) (<= alpha 1.0)))
(/ (- (* alpha alpha) 1.0) (* (* PI (log (* alpha alpha))) (+ 1.0 (* (* (- (* alpha alpha) 1.0) cosTheta) cosTheta)))))