
(FPCore (sinTheta_O h eta)
:precision binary32
(asin
(/
h
(sqrt
(-
(* eta eta)
(/
(* sinTheta_O sinTheta_O)
(sqrt (- 1.0 (* sinTheta_O sinTheta_O)))))))))
float code(float sinTheta_O, float h, float eta) {
return asinf((h / sqrtf(((eta * eta) - ((sinTheta_O * sinTheta_O) / sqrtf((1.0f - (sinTheta_O * sinTheta_O))))))));
}
real(4) function code(sintheta_o, h, eta)
real(4), intent (in) :: sintheta_o
real(4), intent (in) :: h
real(4), intent (in) :: eta
code = asin((h / sqrt(((eta * eta) - ((sintheta_o * sintheta_o) / sqrt((1.0e0 - (sintheta_o * sintheta_o))))))))
end function
function code(sinTheta_O, h, eta) return asin(Float32(h / sqrt(Float32(Float32(eta * eta) - Float32(Float32(sinTheta_O * sinTheta_O) / sqrt(Float32(Float32(1.0) - Float32(sinTheta_O * sinTheta_O)))))))) end
function tmp = code(sinTheta_O, h, eta) tmp = asin((h / sqrt(((eta * eta) - ((sinTheta_O * sinTheta_O) / sqrt((single(1.0) - (sinTheta_O * sinTheta_O)))))))); end
\begin{array}{l}
\\
\sin^{-1} \left(\frac{h}{\sqrt{eta \cdot eta - \frac{sinTheta\_O \cdot sinTheta\_O}{\sqrt{1 - sinTheta\_O \cdot sinTheta\_O}}}}\right)
\end{array}
Sampling outcomes in binary32 precision:
Herbie found 5 alternatives:
| Alternative | Accuracy | Speedup |
|---|
(FPCore (sinTheta_O h eta)
:precision binary32
(asin
(/
h
(sqrt
(-
(* eta eta)
(/
(* sinTheta_O sinTheta_O)
(sqrt (- 1.0 (* sinTheta_O sinTheta_O)))))))))
float code(float sinTheta_O, float h, float eta) {
return asinf((h / sqrtf(((eta * eta) - ((sinTheta_O * sinTheta_O) / sqrtf((1.0f - (sinTheta_O * sinTheta_O))))))));
}
real(4) function code(sintheta_o, h, eta)
real(4), intent (in) :: sintheta_o
real(4), intent (in) :: h
real(4), intent (in) :: eta
code = asin((h / sqrt(((eta * eta) - ((sintheta_o * sintheta_o) / sqrt((1.0e0 - (sintheta_o * sintheta_o))))))))
end function
function code(sinTheta_O, h, eta) return asin(Float32(h / sqrt(Float32(Float32(eta * eta) - Float32(Float32(sinTheta_O * sinTheta_O) / sqrt(Float32(Float32(1.0) - Float32(sinTheta_O * sinTheta_O)))))))) end
function tmp = code(sinTheta_O, h, eta) tmp = asin((h / sqrt(((eta * eta) - ((sinTheta_O * sinTheta_O) / sqrt((single(1.0) - (sinTheta_O * sinTheta_O)))))))); end
\begin{array}{l}
\\
\sin^{-1} \left(\frac{h}{\sqrt{eta \cdot eta - \frac{sinTheta\_O \cdot sinTheta\_O}{\sqrt{1 - sinTheta\_O \cdot sinTheta\_O}}}}\right)
\end{array}
sinTheta_O_m = (fabs.f32 sinTheta_O)
(FPCore (sinTheta_O_m h eta)
:precision binary32
(let* ((t_0 (pow (- 1.0 (pow sinTheta_O_m 2.0)) -0.25)))
(asin
(/
(* h (pow (- eta (* sinTheta_O_m t_0)) -0.5))
(sqrt (fma sinTheta_O_m t_0 eta))))))sinTheta_O_m = fabs(sinTheta_O);
float code(float sinTheta_O_m, float h, float eta) {
float t_0 = powf((1.0f - powf(sinTheta_O_m, 2.0f)), -0.25f);
return asinf(((h * powf((eta - (sinTheta_O_m * t_0)), -0.5f)) / sqrtf(fmaf(sinTheta_O_m, t_0, eta))));
}
sinTheta_O_m = abs(sinTheta_O) function code(sinTheta_O_m, h, eta) t_0 = Float32(Float32(1.0) - (sinTheta_O_m ^ Float32(2.0))) ^ Float32(-0.25) return asin(Float32(Float32(h * (Float32(eta - Float32(sinTheta_O_m * t_0)) ^ Float32(-0.5))) / sqrt(fma(sinTheta_O_m, t_0, eta)))) end
\begin{array}{l}
sinTheta_O_m = \left|sinTheta\_O\right|
\\
\begin{array}{l}
t_0 := {\left(1 - {sinTheta\_O\_m}^{2}\right)}^{-0.25}\\
\sin^{-1} \left(\frac{h \cdot {\left(eta - sinTheta\_O\_m \cdot t\_0\right)}^{-0.5}}{\sqrt{\mathsf{fma}\left(sinTheta\_O\_m, t\_0, eta\right)}}\right)
\end{array}
\end{array}
Initial program 94.4%
add-sqr-sqrt94.4%
difference-of-squares94.4%
sqrt-div94.4%
sqrt-prod51.4%
add-sqr-sqrt90.5%
pow1/290.5%
sqrt-pow190.5%
pow290.5%
metadata-eval90.5%
sqrt-div90.5%
sqrt-prod51.5%
add-sqr-sqrt94.4%
pow1/294.4%
sqrt-pow194.4%
Applied egg-rr94.4%
*-un-lft-identity94.4%
sqrt-prod98.6%
times-frac98.1%
+-commutative98.1%
div-inv98.1%
fma-def98.1%
pow-flip98.1%
metadata-eval98.1%
div-inv98.1%
pow-flip98.1%
metadata-eval98.1%
Applied egg-rr98.1%
associate-*l/98.2%
*-lft-identity98.2%
Simplified98.2%
div-inv98.2%
pow1/298.2%
pow-flip98.7%
metadata-eval98.7%
Applied egg-rr98.7%
Final simplification98.7%
sinTheta_O_m = (fabs.f32 sinTheta_O)
(FPCore (sinTheta_O_m h eta)
:precision binary32
(let* ((t_0 (pow (- 1.0 (pow sinTheta_O_m 2.0)) -0.25)))
(asin
(*
h
(/
(pow (- eta (* sinTheta_O_m t_0)) -0.5)
(sqrt (fma sinTheta_O_m t_0 eta)))))))sinTheta_O_m = fabs(sinTheta_O);
float code(float sinTheta_O_m, float h, float eta) {
float t_0 = powf((1.0f - powf(sinTheta_O_m, 2.0f)), -0.25f);
return asinf((h * (powf((eta - (sinTheta_O_m * t_0)), -0.5f) / sqrtf(fmaf(sinTheta_O_m, t_0, eta)))));
}
sinTheta_O_m = abs(sinTheta_O) function code(sinTheta_O_m, h, eta) t_0 = Float32(Float32(1.0) - (sinTheta_O_m ^ Float32(2.0))) ^ Float32(-0.25) return asin(Float32(h * Float32((Float32(eta - Float32(sinTheta_O_m * t_0)) ^ Float32(-0.5)) / sqrt(fma(sinTheta_O_m, t_0, eta))))) end
\begin{array}{l}
sinTheta_O_m = \left|sinTheta\_O\right|
\\
\begin{array}{l}
t_0 := {\left(1 - {sinTheta\_O\_m}^{2}\right)}^{-0.25}\\
\sin^{-1} \left(h \cdot \frac{{\left(eta - sinTheta\_O\_m \cdot t\_0\right)}^{-0.5}}{\sqrt{\mathsf{fma}\left(sinTheta\_O\_m, t\_0, eta\right)}}\right)
\end{array}
\end{array}
Initial program 94.4%
add-sqr-sqrt94.4%
difference-of-squares94.4%
sqrt-div94.4%
sqrt-prod51.4%
add-sqr-sqrt90.5%
pow1/290.5%
sqrt-pow190.5%
pow290.5%
metadata-eval90.5%
sqrt-div90.5%
sqrt-prod51.5%
add-sqr-sqrt94.4%
pow1/294.4%
sqrt-pow194.4%
Applied egg-rr94.4%
*-un-lft-identity94.4%
sqrt-prod98.6%
times-frac98.1%
+-commutative98.1%
div-inv98.1%
fma-def98.1%
pow-flip98.1%
metadata-eval98.1%
div-inv98.1%
pow-flip98.1%
metadata-eval98.1%
Applied egg-rr98.1%
associate-*l/98.2%
*-lft-identity98.2%
Simplified98.2%
div-inv98.2%
*-un-lft-identity98.2%
times-frac98.3%
add-sqr-sqrt50.9%
associate-*l/50.9%
/-rgt-identity50.9%
add-sqr-sqrt98.3%
pow1/298.3%
pow-flip98.7%
metadata-eval98.7%
Applied egg-rr98.7%
Final simplification98.7%
sinTheta_O_m = (fabs.f32 sinTheta_O)
(FPCore (sinTheta_O_m h eta)
:precision binary32
(let* ((t_0 (pow (- 1.0 (pow sinTheta_O_m 2.0)) -0.25)))
(asin
(/
(/ h (sqrt (- eta (* sinTheta_O_m t_0))))
(sqrt (fma sinTheta_O_m t_0 eta))))))sinTheta_O_m = fabs(sinTheta_O);
float code(float sinTheta_O_m, float h, float eta) {
float t_0 = powf((1.0f - powf(sinTheta_O_m, 2.0f)), -0.25f);
return asinf(((h / sqrtf((eta - (sinTheta_O_m * t_0)))) / sqrtf(fmaf(sinTheta_O_m, t_0, eta))));
}
sinTheta_O_m = abs(sinTheta_O) function code(sinTheta_O_m, h, eta) t_0 = Float32(Float32(1.0) - (sinTheta_O_m ^ Float32(2.0))) ^ Float32(-0.25) return asin(Float32(Float32(h / sqrt(Float32(eta - Float32(sinTheta_O_m * t_0)))) / sqrt(fma(sinTheta_O_m, t_0, eta)))) end
\begin{array}{l}
sinTheta_O_m = \left|sinTheta\_O\right|
\\
\begin{array}{l}
t_0 := {\left(1 - {sinTheta\_O\_m}^{2}\right)}^{-0.25}\\
\sin^{-1} \left(\frac{\frac{h}{\sqrt{eta - sinTheta\_O\_m \cdot t\_0}}}{\sqrt{\mathsf{fma}\left(sinTheta\_O\_m, t\_0, eta\right)}}\right)
\end{array}
\end{array}
Initial program 94.4%
add-sqr-sqrt94.4%
difference-of-squares94.4%
sqrt-div94.4%
sqrt-prod51.4%
add-sqr-sqrt90.5%
pow1/290.5%
sqrt-pow190.5%
pow290.5%
metadata-eval90.5%
sqrt-div90.5%
sqrt-prod51.5%
add-sqr-sqrt94.4%
pow1/294.4%
sqrt-pow194.4%
Applied egg-rr94.4%
*-un-lft-identity94.4%
sqrt-prod98.6%
times-frac98.1%
+-commutative98.1%
div-inv98.1%
fma-def98.1%
pow-flip98.1%
metadata-eval98.1%
div-inv98.1%
pow-flip98.1%
metadata-eval98.1%
Applied egg-rr98.1%
associate-*l/98.2%
*-lft-identity98.2%
Simplified98.2%
Final simplification98.2%
sinTheta_O_m = (fabs.f32 sinTheta_O) (FPCore (sinTheta_O_m h eta) :precision binary32 (asin (/ h (+ eta (* -0.5 (* sinTheta_O_m (/ sinTheta_O_m eta)))))))
sinTheta_O_m = fabs(sinTheta_O);
float code(float sinTheta_O_m, float h, float eta) {
return asinf((h / (eta + (-0.5f * (sinTheta_O_m * (sinTheta_O_m / eta))))));
}
sinTheta_O_m = abs(sinTheta_O)
real(4) function code(sintheta_o_m, h, eta)
real(4), intent (in) :: sintheta_o_m
real(4), intent (in) :: h
real(4), intent (in) :: eta
code = asin((h / (eta + ((-0.5e0) * (sintheta_o_m * (sintheta_o_m / eta))))))
end function
sinTheta_O_m = abs(sinTheta_O) function code(sinTheta_O_m, h, eta) return asin(Float32(h / Float32(eta + Float32(Float32(-0.5) * Float32(sinTheta_O_m * Float32(sinTheta_O_m / eta)))))) end
sinTheta_O_m = abs(sinTheta_O); function tmp = code(sinTheta_O_m, h, eta) tmp = asin((h / (eta + (single(-0.5) * (sinTheta_O_m * (sinTheta_O_m / eta)))))); end
\begin{array}{l}
sinTheta_O_m = \left|sinTheta\_O\right|
\\
\sin^{-1} \left(\frac{h}{eta + -0.5 \cdot \left(sinTheta\_O\_m \cdot \frac{sinTheta\_O\_m}{eta}\right)}\right)
\end{array}
Initial program 94.4%
Taylor expanded in sinTheta_O around 0 97.2%
unpow297.2%
*-un-lft-identity97.2%
times-frac97.6%
Applied egg-rr97.6%
Final simplification97.6%
sinTheta_O_m = (fabs.f32 sinTheta_O) (FPCore (sinTheta_O_m h eta) :precision binary32 (asin (/ h eta)))
sinTheta_O_m = fabs(sinTheta_O);
float code(float sinTheta_O_m, float h, float eta) {
return asinf((h / eta));
}
sinTheta_O_m = abs(sinTheta_O)
real(4) function code(sintheta_o_m, h, eta)
real(4), intent (in) :: sintheta_o_m
real(4), intent (in) :: h
real(4), intent (in) :: eta
code = asin((h / eta))
end function
sinTheta_O_m = abs(sinTheta_O) function code(sinTheta_O_m, h, eta) return asin(Float32(h / eta)) end
sinTheta_O_m = abs(sinTheta_O); function tmp = code(sinTheta_O_m, h, eta) tmp = asin((h / eta)); end
\begin{array}{l}
sinTheta_O_m = \left|sinTheta\_O\right|
\\
\sin^{-1} \left(\frac{h}{eta}\right)
\end{array}
Initial program 94.4%
Taylor expanded in eta around inf 95.4%
Final simplification95.4%
herbie shell --seed 2024029
(FPCore (sinTheta_O h eta)
:name "HairBSDF, gamma for a refracted ray"
:precision binary32
:pre (and (and (and (<= -1.0 sinTheta_O) (<= sinTheta_O 1.0)) (and (<= -1.0 h) (<= h 1.0))) (and (<= 0.0 eta) (<= eta 10.0)))
(asin (/ h (sqrt (- (* eta eta) (/ (* sinTheta_O sinTheta_O) (sqrt (- 1.0 (* sinTheta_O sinTheta_O)))))))))