
(FPCore (cosTheta_i u1 u2) :precision binary32 (* (sqrt (/ u1 (- 1.0 u1))) (cos (* 6.28318530718 u2))))
float code(float cosTheta_i, float u1, float u2) {
return sqrtf((u1 / (1.0f - u1))) * cosf((6.28318530718f * u2));
}
real(4) function code(costheta_i, u1, u2)
real(4), intent (in) :: costheta_i
real(4), intent (in) :: u1
real(4), intent (in) :: u2
code = sqrt((u1 / (1.0e0 - u1))) * cos((6.28318530718e0 * u2))
end function
function code(cosTheta_i, u1, u2) return Float32(sqrt(Float32(u1 / Float32(Float32(1.0) - u1))) * cos(Float32(Float32(6.28318530718) * u2))) end
function tmp = code(cosTheta_i, u1, u2) tmp = sqrt((u1 / (single(1.0) - u1))) * cos((single(6.28318530718) * u2)); end
\begin{array}{l}
\\
\sqrt{\frac{u1}{1 - u1}} \cdot \cos \left(6.28318530718 \cdot u2\right)
\end{array}
Sampling outcomes in binary32 precision:
Herbie found 8 alternatives:
| Alternative | Accuracy | Speedup |
|---|
(FPCore (cosTheta_i u1 u2) :precision binary32 (* (sqrt (/ u1 (- 1.0 u1))) (cos (* 6.28318530718 u2))))
float code(float cosTheta_i, float u1, float u2) {
return sqrtf((u1 / (1.0f - u1))) * cosf((6.28318530718f * u2));
}
real(4) function code(costheta_i, u1, u2)
real(4), intent (in) :: costheta_i
real(4), intent (in) :: u1
real(4), intent (in) :: u2
code = sqrt((u1 / (1.0e0 - u1))) * cos((6.28318530718e0 * u2))
end function
function code(cosTheta_i, u1, u2) return Float32(sqrt(Float32(u1 / Float32(Float32(1.0) - u1))) * cos(Float32(Float32(6.28318530718) * u2))) end
function tmp = code(cosTheta_i, u1, u2) tmp = sqrt((u1 / (single(1.0) - u1))) * cos((single(6.28318530718) * u2)); end
\begin{array}{l}
\\
\sqrt{\frac{u1}{1 - u1}} \cdot \cos \left(6.28318530718 \cdot u2\right)
\end{array}
(FPCore (cosTheta_i u1 u2) :precision binary32 (* (pow (- (+ (/ (/ 1.0 u1) u1) 1.0) (/ 2.0 u1)) -0.25) (cos (* 6.28318530718 u2))))
float code(float cosTheta_i, float u1, float u2) {
return powf(((((1.0f / u1) / u1) + 1.0f) - (2.0f / u1)), -0.25f) * cosf((6.28318530718f * u2));
}
real(4) function code(costheta_i, u1, u2)
real(4), intent (in) :: costheta_i
real(4), intent (in) :: u1
real(4), intent (in) :: u2
code = (((((1.0e0 / u1) / u1) + 1.0e0) - (2.0e0 / u1)) ** (-0.25e0)) * cos((6.28318530718e0 * u2))
end function
function code(cosTheta_i, u1, u2) return Float32((Float32(Float32(Float32(Float32(Float32(1.0) / u1) / u1) + Float32(1.0)) - Float32(Float32(2.0) / u1)) ^ Float32(-0.25)) * cos(Float32(Float32(6.28318530718) * u2))) end
function tmp = code(cosTheta_i, u1, u2) tmp = (((((single(1.0) / u1) / u1) + single(1.0)) - (single(2.0) / u1)) ^ single(-0.25)) * cos((single(6.28318530718) * u2)); end
\begin{array}{l}
\\
{\left(\left(\frac{\frac{1}{u1}}{u1} + 1\right) - \frac{2}{u1}\right)}^{-0.25} \cdot \cos \left(6.28318530718 \cdot u2\right)
\end{array}
Initial program 98.8%
Applied rewrites98.8%
Taylor expanded in u1 around inf
lower--.f32N/A
+-commutativeN/A
lower-+.f32N/A
unpow2N/A
associate-/r*N/A
lower-/.f32N/A
lower-/.f32N/A
associate-*r/N/A
metadata-evalN/A
lower-/.f3298.9
Applied rewrites98.9%
(FPCore (cosTheta_i u1 u2)
:precision binary32
(let* ((t_0 (cos (* 6.28318530718 u2))))
(if (<= t_0 0.9955000281333923)
(* (sqrt u1) t_0)
(* (sqrt (/ u1 (- 1.0 u1))) (+ (* -19.739208802181317 (* u2 u2)) 1.0)))))
float code(float cosTheta_i, float u1, float u2) {
float t_0 = cosf((6.28318530718f * u2));
float tmp;
if (t_0 <= 0.9955000281333923f) {
tmp = sqrtf(u1) * t_0;
} else {
tmp = sqrtf((u1 / (1.0f - u1))) * ((-19.739208802181317f * (u2 * u2)) + 1.0f);
}
return tmp;
}
real(4) function code(costheta_i, u1, u2)
real(4), intent (in) :: costheta_i
real(4), intent (in) :: u1
real(4), intent (in) :: u2
real(4) :: t_0
real(4) :: tmp
t_0 = cos((6.28318530718e0 * u2))
if (t_0 <= 0.9955000281333923e0) then
tmp = sqrt(u1) * t_0
else
tmp = sqrt((u1 / (1.0e0 - u1))) * (((-19.739208802181317e0) * (u2 * u2)) + 1.0e0)
end if
code = tmp
end function
function code(cosTheta_i, u1, u2) t_0 = cos(Float32(Float32(6.28318530718) * u2)) tmp = Float32(0.0) if (t_0 <= Float32(0.9955000281333923)) tmp = Float32(sqrt(u1) * t_0); else tmp = Float32(sqrt(Float32(u1 / Float32(Float32(1.0) - u1))) * Float32(Float32(Float32(-19.739208802181317) * Float32(u2 * u2)) + Float32(1.0))); end return tmp end
function tmp_2 = code(cosTheta_i, u1, u2) t_0 = cos((single(6.28318530718) * u2)); tmp = single(0.0); if (t_0 <= single(0.9955000281333923)) tmp = sqrt(u1) * t_0; else tmp = sqrt((u1 / (single(1.0) - u1))) * ((single(-19.739208802181317) * (u2 * u2)) + single(1.0)); end tmp_2 = tmp; end
\begin{array}{l}
\\
\begin{array}{l}
t_0 := \cos \left(6.28318530718 \cdot u2\right)\\
\mathbf{if}\;t\_0 \leq 0.9955000281333923:\\
\;\;\;\;\sqrt{u1} \cdot t\_0\\
\mathbf{else}:\\
\;\;\;\;\sqrt{\frac{u1}{1 - u1}} \cdot \left(-19.739208802181317 \cdot \left(u2 \cdot u2\right) + 1\right)\\
\end{array}
\end{array}
if (cos.f32 (*.f32 #s(literal 314159265359/50000000000 binary32) u2)) < 0.995500028Initial program 96.5%
Taylor expanded in u1 around 0
lower-sqrt.f3274.6
Applied rewrites74.6%
if 0.995500028 < (cos.f32 (*.f32 #s(literal 314159265359/50000000000 binary32) u2)) Initial program 99.4%
Taylor expanded in u2 around 0
+-commutativeN/A
*-commutativeN/A
lower-fma.f32N/A
unpow2N/A
lower-*.f3292.0
Applied rewrites91.8%
Applied rewrites98.9%
(FPCore (cosTheta_i u1 u2) :precision binary32 (* (sqrt (/ u1 (- 1.0 u1))) (cos (* 6.28318530718 u2))))
float code(float cosTheta_i, float u1, float u2) {
return sqrtf((u1 / (1.0f - u1))) * cosf((6.28318530718f * u2));
}
real(4) function code(costheta_i, u1, u2)
real(4), intent (in) :: costheta_i
real(4), intent (in) :: u1
real(4), intent (in) :: u2
code = sqrt((u1 / (1.0e0 - u1))) * cos((6.28318530718e0 * u2))
end function
function code(cosTheta_i, u1, u2) return Float32(sqrt(Float32(u1 / Float32(Float32(1.0) - u1))) * cos(Float32(Float32(6.28318530718) * u2))) end
function tmp = code(cosTheta_i, u1, u2) tmp = sqrt((u1 / (single(1.0) - u1))) * cos((single(6.28318530718) * u2)); end
\begin{array}{l}
\\
\sqrt{\frac{u1}{1 - u1}} \cdot \cos \left(6.28318530718 \cdot u2\right)
\end{array}
Initial program 98.8%
(FPCore (cosTheta_i u1 u2) :precision binary32 (* (sqrt (/ u1 (- 1.0 u1))) (+ (* -19.739208802181317 (* u2 u2)) 1.0)))
float code(float cosTheta_i, float u1, float u2) {
return sqrtf((u1 / (1.0f - u1))) * ((-19.739208802181317f * (u2 * u2)) + 1.0f);
}
real(4) function code(costheta_i, u1, u2)
real(4), intent (in) :: costheta_i
real(4), intent (in) :: u1
real(4), intent (in) :: u2
code = sqrt((u1 / (1.0e0 - u1))) * (((-19.739208802181317e0) * (u2 * u2)) + 1.0e0)
end function
function code(cosTheta_i, u1, u2) return Float32(sqrt(Float32(u1 / Float32(Float32(1.0) - u1))) * Float32(Float32(Float32(-19.739208802181317) * Float32(u2 * u2)) + Float32(1.0))) end
function tmp = code(cosTheta_i, u1, u2) tmp = sqrt((u1 / (single(1.0) - u1))) * ((single(-19.739208802181317) * (u2 * u2)) + single(1.0)); end
\begin{array}{l}
\\
\sqrt{\frac{u1}{1 - u1}} \cdot \left(-19.739208802181317 \cdot \left(u2 \cdot u2\right) + 1\right)
\end{array}
Initial program 98.8%
Taylor expanded in u2 around 0
+-commutativeN/A
*-commutativeN/A
lower-fma.f32N/A
unpow2N/A
lower-*.f3280.5
Applied rewrites80.3%
Applied rewrites88.8%
(FPCore (cosTheta_i u1 u2) :precision binary32 (sqrt (/ u1 (- 1.0 u1))))
float code(float cosTheta_i, float u1, float u2) {
return sqrtf((u1 / (1.0f - u1)));
}
real(4) function code(costheta_i, u1, u2)
real(4), intent (in) :: costheta_i
real(4), intent (in) :: u1
real(4), intent (in) :: u2
code = sqrt((u1 / (1.0e0 - u1)))
end function
function code(cosTheta_i, u1, u2) return sqrt(Float32(u1 / Float32(Float32(1.0) - u1))) end
function tmp = code(cosTheta_i, u1, u2) tmp = sqrt((u1 / (single(1.0) - u1))); end
\begin{array}{l}
\\
\sqrt{\frac{u1}{1 - u1}}
\end{array}
Initial program 98.8%
Taylor expanded in u2 around 0
*-rgt-identityN/A
sub-negN/A
rgt-mult-inverseN/A
mul-1-negN/A
distribute-neg-frac2N/A
mul-1-negN/A
*-rgt-identityN/A
distribute-lft-inN/A
+-commutativeN/A
sub-negN/A
associate-*r*N/A
lower-sqrt.f32N/A
*-rgt-identityN/A
lower-/.f32N/A
associate-*r*N/A
sub-negN/A
+-commutativeN/A
distribute-lft-inN/A
Applied rewrites80.5%
(FPCore (cosTheta_i u1 u2) :precision binary32 (sqrt (+ (* u1 u1) u1)))
float code(float cosTheta_i, float u1, float u2) {
return sqrtf(((u1 * u1) + u1));
}
real(4) function code(costheta_i, u1, u2)
real(4), intent (in) :: costheta_i
real(4), intent (in) :: u1
real(4), intent (in) :: u2
code = sqrt(((u1 * u1) + u1))
end function
function code(cosTheta_i, u1, u2) return sqrt(Float32(Float32(u1 * u1) + u1)) end
function tmp = code(cosTheta_i, u1, u2) tmp = sqrt(((u1 * u1) + u1)); end
\begin{array}{l}
\\
\sqrt{u1 \cdot u1 + u1}
\end{array}
Initial program 98.8%
Taylor expanded in u2 around 0
*-rgt-identityN/A
sub-negN/A
rgt-mult-inverseN/A
mul-1-negN/A
distribute-neg-frac2N/A
mul-1-negN/A
*-rgt-identityN/A
distribute-lft-inN/A
+-commutativeN/A
sub-negN/A
associate-*r*N/A
lower-sqrt.f32N/A
*-rgt-identityN/A
lower-/.f32N/A
associate-*r*N/A
sub-negN/A
+-commutativeN/A
distribute-lft-inN/A
Applied rewrites80.5%
Taylor expanded in u1 around 0
Applied rewrites62.7%
Applied rewrites70.7%
(FPCore (cosTheta_i u1 u2) :precision binary32 (sqrt (* (- u1 -1.0) u1)))
float code(float cosTheta_i, float u1, float u2) {
return sqrtf(((u1 - -1.0f) * u1));
}
real(4) function code(costheta_i, u1, u2)
real(4), intent (in) :: costheta_i
real(4), intent (in) :: u1
real(4), intent (in) :: u2
code = sqrt(((u1 - (-1.0e0)) * u1))
end function
function code(cosTheta_i, u1, u2) return sqrt(Float32(Float32(u1 - Float32(-1.0)) * u1)) end
function tmp = code(cosTheta_i, u1, u2) tmp = sqrt(((u1 - single(-1.0)) * u1)); end
\begin{array}{l}
\\
\sqrt{\left(u1 - -1\right) \cdot u1}
\end{array}
Initial program 98.8%
Taylor expanded in u2 around 0
*-rgt-identityN/A
sub-negN/A
rgt-mult-inverseN/A
mul-1-negN/A
distribute-neg-frac2N/A
mul-1-negN/A
*-rgt-identityN/A
distribute-lft-inN/A
+-commutativeN/A
sub-negN/A
associate-*r*N/A
lower-sqrt.f32N/A
*-rgt-identityN/A
lower-/.f32N/A
associate-*r*N/A
sub-negN/A
+-commutativeN/A
distribute-lft-inN/A
Applied rewrites80.5%
Taylor expanded in u1 around 0
Applied rewrites62.7%
Applied rewrites70.6%
(FPCore (cosTheta_i u1 u2) :precision binary32 (sqrt u1))
float code(float cosTheta_i, float u1, float u2) {
return sqrtf(u1);
}
real(4) function code(costheta_i, u1, u2)
real(4), intent (in) :: costheta_i
real(4), intent (in) :: u1
real(4), intent (in) :: u2
code = sqrt(u1)
end function
function code(cosTheta_i, u1, u2) return sqrt(u1) end
function tmp = code(cosTheta_i, u1, u2) tmp = sqrt(u1); end
\begin{array}{l}
\\
\sqrt{u1}
\end{array}
Initial program 98.8%
Taylor expanded in u2 around 0
*-rgt-identityN/A
sub-negN/A
rgt-mult-inverseN/A
mul-1-negN/A
distribute-neg-frac2N/A
mul-1-negN/A
*-rgt-identityN/A
distribute-lft-inN/A
+-commutativeN/A
sub-negN/A
associate-*r*N/A
lower-sqrt.f32N/A
*-rgt-identityN/A
lower-/.f32N/A
associate-*r*N/A
sub-negN/A
+-commutativeN/A
distribute-lft-inN/A
Applied rewrites80.5%
Taylor expanded in u1 around 0
Applied rewrites62.7%
herbie shell --seed 2024313
(FPCore (cosTheta_i u1 u2)
:name "Trowbridge-Reitz Sample, near normal, slope_x"
:precision binary32
:pre (and (and (and (> cosTheta_i 0.9999) (<= cosTheta_i 1.0)) (and (<= 2.328306437e-10 u1) (<= u1 1.0))) (and (<= 2.328306437e-10 u2) (<= u2 1.0)))
(* (sqrt (/ u1 (- 1.0 u1))) (cos (* 6.28318530718 u2))))