
(FPCore (s u) :precision binary32 (* s (log (/ 1.0 (- 1.0 (* 4.0 u))))))
float code(float s, float u) {
return s * logf((1.0f / (1.0f - (4.0f * u))));
}
real(4) function code(s, u)
real(4), intent (in) :: s
real(4), intent (in) :: u
code = s * log((1.0e0 / (1.0e0 - (4.0e0 * u))))
end function
function code(s, u) return Float32(s * log(Float32(Float32(1.0) / Float32(Float32(1.0) - Float32(Float32(4.0) * u))))) end
function tmp = code(s, u) tmp = s * log((single(1.0) / (single(1.0) - (single(4.0) * u)))); end
\begin{array}{l}
\\
s \cdot \log \left(\frac{1}{1 - 4 \cdot u}\right)
\end{array}
Sampling outcomes in binary32 precision:
Herbie found 7 alternatives:
| Alternative | Accuracy | Speedup |
|---|
(FPCore (s u) :precision binary32 (* s (log (/ 1.0 (- 1.0 (* 4.0 u))))))
float code(float s, float u) {
return s * logf((1.0f / (1.0f - (4.0f * u))));
}
real(4) function code(s, u)
real(4), intent (in) :: s
real(4), intent (in) :: u
code = s * log((1.0e0 / (1.0e0 - (4.0e0 * u))))
end function
function code(s, u) return Float32(s * log(Float32(Float32(1.0) / Float32(Float32(1.0) - Float32(Float32(4.0) * u))))) end
function tmp = code(s, u) tmp = s * log((single(1.0) / (single(1.0) - (single(4.0) * u)))); end
\begin{array}{l}
\\
s \cdot \log \left(\frac{1}{1 - 4 \cdot u}\right)
\end{array}
(FPCore (s u)
:precision binary32
(let* ((t_0 (- 1.0 (* u 4.0))))
(if (<= t_0 0.9800000190734863)
(* (log (/ 1.0 t_0)) s)
(*
(* (* (- (* (/ (+ (/ 4.0 u) 8.0) u) u) (* -21.333333333333332 u)) u) s)
u))))
float code(float s, float u) {
float t_0 = 1.0f - (u * 4.0f);
float tmp;
if (t_0 <= 0.9800000190734863f) {
tmp = logf((1.0f / t_0)) * s;
} else {
tmp = (((((((4.0f / u) + 8.0f) / u) * u) - (-21.333333333333332f * u)) * u) * s) * u;
}
return tmp;
}
real(4) function code(s, u)
real(4), intent (in) :: s
real(4), intent (in) :: u
real(4) :: t_0
real(4) :: tmp
t_0 = 1.0e0 - (u * 4.0e0)
if (t_0 <= 0.9800000190734863e0) then
tmp = log((1.0e0 / t_0)) * s
else
tmp = (((((((4.0e0 / u) + 8.0e0) / u) * u) - ((-21.333333333333332e0) * u)) * u) * s) * u
end if
code = tmp
end function
function code(s, u) t_0 = Float32(Float32(1.0) - Float32(u * Float32(4.0))) tmp = Float32(0.0) if (t_0 <= Float32(0.9800000190734863)) tmp = Float32(log(Float32(Float32(1.0) / t_0)) * s); else tmp = Float32(Float32(Float32(Float32(Float32(Float32(Float32(Float32(Float32(4.0) / u) + Float32(8.0)) / u) * u) - Float32(Float32(-21.333333333333332) * u)) * u) * s) * u); end return tmp end
function tmp_2 = code(s, u) t_0 = single(1.0) - (u * single(4.0)); tmp = single(0.0); if (t_0 <= single(0.9800000190734863)) tmp = log((single(1.0) / t_0)) * s; else tmp = (((((((single(4.0) / u) + single(8.0)) / u) * u) - (single(-21.333333333333332) * u)) * u) * s) * u; end tmp_2 = tmp; end
\begin{array}{l}
\\
\begin{array}{l}
t_0 := 1 - u \cdot 4\\
\mathbf{if}\;t\_0 \leq 0.9800000190734863:\\
\;\;\;\;\log \left(\frac{1}{t\_0}\right) \cdot s\\
\mathbf{else}:\\
\;\;\;\;\left(\left(\left(\frac{\frac{4}{u} + 8}{u} \cdot u - -21.333333333333332 \cdot u\right) \cdot u\right) \cdot s\right) \cdot u\\
\end{array}
\end{array}
if (-.f32 #s(literal 1 binary32) (*.f32 #s(literal 4 binary32) u)) < 0.980000019Initial program 95.2%
if 0.980000019 < (-.f32 #s(literal 1 binary32) (*.f32 #s(literal 4 binary32) u)) Initial program 53.1%
Taylor expanded in u around 0
*-commutativeN/A
lower-*.f32N/A
*-commutativeN/A
distribute-rgt-inN/A
associate-*r*N/A
*-commutativeN/A
*-commutativeN/A
associate-*l*N/A
distribute-lft-outN/A
associate-*l*N/A
distribute-lft-inN/A
lower-*.f32N/A
+-commutativeN/A
*-commutativeN/A
lower-fma.f32N/A
+-commutativeN/A
lower-fma.f3281.9
Applied rewrites81.9%
Taylor expanded in u around inf
Applied rewrites98.2%
Taylor expanded in u around -inf
Applied rewrites98.4%
Applied rewrites98.5%
Final simplification97.8%
(FPCore (s u)
:precision binary32
(if (<= (- 1.0 (* u 4.0)) 0.9999899864196777)
(*
(-
(*
(/ u (fma 21.333333333333332 u -8.0))
(fma (* u u) 455.1111111111111 -64.0))
-4.0)
(* s u))
(* (* u 4.0) s)))
float code(float s, float u) {
float tmp;
if ((1.0f - (u * 4.0f)) <= 0.9999899864196777f) {
tmp = (((u / fmaf(21.333333333333332f, u, -8.0f)) * fmaf((u * u), 455.1111111111111f, -64.0f)) - -4.0f) * (s * u);
} else {
tmp = (u * 4.0f) * s;
}
return tmp;
}
function code(s, u) tmp = Float32(0.0) if (Float32(Float32(1.0) - Float32(u * Float32(4.0))) <= Float32(0.9999899864196777)) tmp = Float32(Float32(Float32(Float32(u / fma(Float32(21.333333333333332), u, Float32(-8.0))) * fma(Float32(u * u), Float32(455.1111111111111), Float32(-64.0))) - Float32(-4.0)) * Float32(s * u)); else tmp = Float32(Float32(u * Float32(4.0)) * s); end return tmp end
\begin{array}{l}
\\
\begin{array}{l}
\mathbf{if}\;1 - u \cdot 4 \leq 0.9999899864196777:\\
\;\;\;\;\left(\frac{u}{\mathsf{fma}\left(21.333333333333332, u, -8\right)} \cdot \mathsf{fma}\left(u \cdot u, 455.1111111111111, -64\right) - -4\right) \cdot \left(s \cdot u\right)\\
\mathbf{else}:\\
\;\;\;\;\left(u \cdot 4\right) \cdot s\\
\end{array}
\end{array}
if (-.f32 #s(literal 1 binary32) (*.f32 #s(literal 4 binary32) u)) < 0.999989986Initial program 79.9%
Taylor expanded in u around 0
*-commutativeN/A
lower-*.f32N/A
*-commutativeN/A
distribute-rgt-inN/A
associate-*r*N/A
*-commutativeN/A
*-commutativeN/A
associate-*l*N/A
distribute-lft-outN/A
associate-*l*N/A
distribute-lft-inN/A
lower-*.f32N/A
+-commutativeN/A
*-commutativeN/A
lower-fma.f32N/A
+-commutativeN/A
lower-fma.f3255.5
Applied rewrites55.5%
Applied rewrites55.5%
Applied rewrites55.5%
Taylor expanded in s around -inf
Applied rewrites75.4%
if 0.999989986 < (-.f32 #s(literal 1 binary32) (*.f32 #s(literal 4 binary32) u)) Initial program 38.3%
Taylor expanded in u around 0
*-commutativeN/A
lower-*.f3295.1
Applied rewrites95.1%
Final simplification84.3%
(FPCore (s u) :precision binary32 (* (* (* (- (* (/ (+ (/ 4.0 u) 8.0) u) u) (* -21.333333333333332 u)) u) s) u))
float code(float s, float u) {
return (((((((4.0f / u) + 8.0f) / u) * u) - (-21.333333333333332f * u)) * u) * s) * u;
}
real(4) function code(s, u)
real(4), intent (in) :: s
real(4), intent (in) :: u
code = (((((((4.0e0 / u) + 8.0e0) / u) * u) - ((-21.333333333333332e0) * u)) * u) * s) * u
end function
function code(s, u) return Float32(Float32(Float32(Float32(Float32(Float32(Float32(Float32(Float32(4.0) / u) + Float32(8.0)) / u) * u) - Float32(Float32(-21.333333333333332) * u)) * u) * s) * u) end
function tmp = code(s, u) tmp = (((((((single(4.0) / u) + single(8.0)) / u) * u) - (single(-21.333333333333332) * u)) * u) * s) * u; end
\begin{array}{l}
\\
\left(\left(\left(\frac{\frac{4}{u} + 8}{u} \cdot u - -21.333333333333332 \cdot u\right) \cdot u\right) \cdot s\right) \cdot u
\end{array}
Initial program 61.5%
Taylor expanded in u around 0
*-commutativeN/A
lower-*.f32N/A
*-commutativeN/A
distribute-rgt-inN/A
associate-*r*N/A
*-commutativeN/A
*-commutativeN/A
associate-*l*N/A
distribute-lft-outN/A
associate-*l*N/A
distribute-lft-inN/A
lower-*.f32N/A
+-commutativeN/A
*-commutativeN/A
lower-fma.f32N/A
+-commutativeN/A
lower-fma.f3273.0
Applied rewrites73.0%
Taylor expanded in u around inf
Applied rewrites89.9%
Taylor expanded in u around -inf
Applied rewrites90.1%
Applied rewrites90.1%
Final simplification90.1%
(FPCore (s u) :precision binary32 (* (* (* (* (- (/ (+ (/ 4.0 u) 8.0) u) -21.333333333333332) u) u) s) u))
float code(float s, float u) {
return (((((((4.0f / u) + 8.0f) / u) - -21.333333333333332f) * u) * u) * s) * u;
}
real(4) function code(s, u)
real(4), intent (in) :: s
real(4), intent (in) :: u
code = (((((((4.0e0 / u) + 8.0e0) / u) - (-21.333333333333332e0)) * u) * u) * s) * u
end function
function code(s, u) return Float32(Float32(Float32(Float32(Float32(Float32(Float32(Float32(Float32(4.0) / u) + Float32(8.0)) / u) - Float32(-21.333333333333332)) * u) * u) * s) * u) end
function tmp = code(s, u) tmp = (((((((single(4.0) / u) + single(8.0)) / u) - single(-21.333333333333332)) * u) * u) * s) * u; end
\begin{array}{l}
\\
\left(\left(\left(\left(\frac{\frac{4}{u} + 8}{u} - -21.333333333333332\right) \cdot u\right) \cdot u\right) \cdot s\right) \cdot u
\end{array}
Initial program 61.5%
Taylor expanded in u around 0
*-commutativeN/A
lower-*.f32N/A
*-commutativeN/A
distribute-rgt-inN/A
associate-*r*N/A
*-commutativeN/A
*-commutativeN/A
associate-*l*N/A
distribute-lft-outN/A
associate-*l*N/A
distribute-lft-inN/A
lower-*.f32N/A
+-commutativeN/A
*-commutativeN/A
lower-fma.f32N/A
+-commutativeN/A
lower-fma.f3273.0
Applied rewrites73.0%
Taylor expanded in u around inf
Applied rewrites89.9%
Taylor expanded in u around -inf
Applied rewrites90.1%
Final simplification90.1%
(FPCore (s u) :precision binary32 (* (* (* (- (/ (+ (/ 4.0 u) 8.0) u) -21.333333333333332) u) u) (* s u)))
float code(float s, float u) {
return ((((((4.0f / u) + 8.0f) / u) - -21.333333333333332f) * u) * u) * (s * u);
}
real(4) function code(s, u)
real(4), intent (in) :: s
real(4), intent (in) :: u
code = ((((((4.0e0 / u) + 8.0e0) / u) - (-21.333333333333332e0)) * u) * u) * (s * u)
end function
function code(s, u) return Float32(Float32(Float32(Float32(Float32(Float32(Float32(Float32(4.0) / u) + Float32(8.0)) / u) - Float32(-21.333333333333332)) * u) * u) * Float32(s * u)) end
function tmp = code(s, u) tmp = ((((((single(4.0) / u) + single(8.0)) / u) - single(-21.333333333333332)) * u) * u) * (s * u); end
\begin{array}{l}
\\
\left(\left(\left(\frac{\frac{4}{u} + 8}{u} - -21.333333333333332\right) \cdot u\right) \cdot u\right) \cdot \left(s \cdot u\right)
\end{array}
Initial program 61.5%
Taylor expanded in u around 0
*-commutativeN/A
lower-*.f32N/A
*-commutativeN/A
distribute-rgt-inN/A
associate-*r*N/A
*-commutativeN/A
*-commutativeN/A
associate-*l*N/A
distribute-lft-outN/A
associate-*l*N/A
distribute-lft-inN/A
lower-*.f32N/A
+-commutativeN/A
*-commutativeN/A
lower-fma.f32N/A
+-commutativeN/A
lower-fma.f3273.0
Applied rewrites73.0%
Taylor expanded in u around inf
Applied rewrites89.9%
Taylor expanded in u around -inf
Applied rewrites90.1%
Applied rewrites89.8%
Final simplification89.8%
(FPCore (s u) :precision binary32 (* (* u 4.0) s))
float code(float s, float u) {
return (u * 4.0f) * s;
}
real(4) function code(s, u)
real(4), intent (in) :: s
real(4), intent (in) :: u
code = (u * 4.0e0) * s
end function
function code(s, u) return Float32(Float32(u * Float32(4.0)) * s) end
function tmp = code(s, u) tmp = (u * single(4.0)) * s; end
\begin{array}{l}
\\
\left(u \cdot 4\right) \cdot s
\end{array}
Initial program 61.5%
Taylor expanded in u around 0
*-commutativeN/A
lower-*.f3273.0
Applied rewrites73.0%
Final simplification73.0%
(FPCore (s u) :precision binary32 (* 4.0 (* s u)))
float code(float s, float u) {
return 4.0f * (s * u);
}
real(4) function code(s, u)
real(4), intent (in) :: s
real(4), intent (in) :: u
code = 4.0e0 * (s * u)
end function
function code(s, u) return Float32(Float32(4.0) * Float32(s * u)) end
function tmp = code(s, u) tmp = single(4.0) * (s * u); end
\begin{array}{l}
\\
4 \cdot \left(s \cdot u\right)
\end{array}
Initial program 61.5%
Taylor expanded in u around 0
*-commutativeN/A
lower-*.f32N/A
*-commutativeN/A
distribute-rgt-inN/A
associate-*r*N/A
*-commutativeN/A
*-commutativeN/A
associate-*l*N/A
distribute-lft-outN/A
associate-*l*N/A
distribute-lft-inN/A
lower-*.f32N/A
+-commutativeN/A
*-commutativeN/A
lower-fma.f32N/A
+-commutativeN/A
lower-fma.f3273.0
Applied rewrites73.0%
Applied rewrites72.8%
Taylor expanded in u around 0
Applied rewrites72.8%
herbie shell --seed 2024255
(FPCore (s u)
:name "Disney BSSRDF, sample scattering profile, lower"
:precision binary32
:pre (and (and (<= 0.0 s) (<= s 256.0)) (and (<= 2.328306437e-10 u) (<= u 0.25)))
(* s (log (/ 1.0 (- 1.0 (* 4.0 u))))))