
(FPCore (cosTheta_i u1 u2) :precision binary32 (* (sqrt (- (log (- 1.0 u1)))) (sin (* (* 2.0 PI) u2))))
float code(float cosTheta_i, float u1, float u2) {
return sqrtf(-logf((1.0f - u1))) * sinf(((2.0f * ((float) M_PI)) * u2));
}
function code(cosTheta_i, u1, u2) return Float32(sqrt(Float32(-log(Float32(Float32(1.0) - u1)))) * sin(Float32(Float32(Float32(2.0) * Float32(pi)) * u2))) end
function tmp = code(cosTheta_i, u1, u2) tmp = sqrt(-log((single(1.0) - u1))) * sin(((single(2.0) * single(pi)) * u2)); end
\begin{array}{l}
\\
\sqrt{-\log \left(1 - u1\right)} \cdot \sin \left(\left(2 \cdot \pi\right) \cdot u2\right)
\end{array}
Sampling outcomes in binary32 precision:
Herbie found 11 alternatives:
| Alternative | Accuracy | Speedup |
|---|
(FPCore (cosTheta_i u1 u2) :precision binary32 (* (sqrt (- (log (- 1.0 u1)))) (sin (* (* 2.0 PI) u2))))
float code(float cosTheta_i, float u1, float u2) {
return sqrtf(-logf((1.0f - u1))) * sinf(((2.0f * ((float) M_PI)) * u2));
}
function code(cosTheta_i, u1, u2) return Float32(sqrt(Float32(-log(Float32(Float32(1.0) - u1)))) * sin(Float32(Float32(Float32(2.0) * Float32(pi)) * u2))) end
function tmp = code(cosTheta_i, u1, u2) tmp = sqrt(-log((single(1.0) - u1))) * sin(((single(2.0) * single(pi)) * u2)); end
\begin{array}{l}
\\
\sqrt{-\log \left(1 - u1\right)} \cdot \sin \left(\left(2 \cdot \pi\right) \cdot u2\right)
\end{array}
(FPCore (cosTheta_i u1 u2) :precision binary32 (* (sqrt (- (log1p (- u1)))) (sin (* PI (+ u2 u2)))))
float code(float cosTheta_i, float u1, float u2) {
return sqrtf(-log1pf(-u1)) * sinf((((float) M_PI) * (u2 + u2)));
}
function code(cosTheta_i, u1, u2) return Float32(sqrt(Float32(-log1p(Float32(-u1)))) * sin(Float32(Float32(pi) * Float32(u2 + u2)))) end
\begin{array}{l}
\\
\sqrt{-\mathsf{log1p}\left(-u1\right)} \cdot \sin \left(\pi \cdot \left(u2 + u2\right)\right)
\end{array}
Initial program 58.0%
Applied rewrites98.2%
lift--.f32N/A
lift-log1p.f32N/A
lift-log1p.f32N/A
diff-logN/A
clear-numN/A
lift-neg.f32N/A
unsub-negN/A
metadata-evalN/A
lift-*.f32N/A
flip--N/A
neg-logN/A
lower-neg.f32N/A
sub-negN/A
lower-log1p.f32N/A
lower-neg.f3298.3
Applied rewrites98.3%
(FPCore (cosTheta_i u1 u2)
:precision binary32
(if (<= (* u2 (* PI 2.0)) 0.05999999865889549)
(*
(sqrt (- (log1p (- u1))))
(* u2 (fma u2 (* u2 (* PI (* (* PI PI) -1.3333333333333333))) (* PI 2.0))))
(*
(sin (* PI (+ u2 u2)))
(sqrt (fma (* u1 u1) (fma u1 (fma u1 0.25 0.3333333333333333) 0.5) u1)))))
float code(float cosTheta_i, float u1, float u2) {
float tmp;
if ((u2 * (((float) M_PI) * 2.0f)) <= 0.05999999865889549f) {
tmp = sqrtf(-log1pf(-u1)) * (u2 * fmaf(u2, (u2 * (((float) M_PI) * ((((float) M_PI) * ((float) M_PI)) * -1.3333333333333333f))), (((float) M_PI) * 2.0f)));
} else {
tmp = sinf((((float) M_PI) * (u2 + u2))) * sqrtf(fmaf((u1 * u1), fmaf(u1, fmaf(u1, 0.25f, 0.3333333333333333f), 0.5f), u1));
}
return tmp;
}
function code(cosTheta_i, u1, u2) tmp = Float32(0.0) if (Float32(u2 * Float32(Float32(pi) * Float32(2.0))) <= Float32(0.05999999865889549)) tmp = Float32(sqrt(Float32(-log1p(Float32(-u1)))) * Float32(u2 * fma(u2, Float32(u2 * Float32(Float32(pi) * Float32(Float32(Float32(pi) * Float32(pi)) * Float32(-1.3333333333333333)))), Float32(Float32(pi) * Float32(2.0))))); else tmp = Float32(sin(Float32(Float32(pi) * Float32(u2 + u2))) * sqrt(fma(Float32(u1 * u1), fma(u1, fma(u1, Float32(0.25), Float32(0.3333333333333333)), Float32(0.5)), u1))); end return tmp end
\begin{array}{l}
\\
\begin{array}{l}
\mathbf{if}\;u2 \cdot \left(\pi \cdot 2\right) \leq 0.05999999865889549:\\
\;\;\;\;\sqrt{-\mathsf{log1p}\left(-u1\right)} \cdot \left(u2 \cdot \mathsf{fma}\left(u2, u2 \cdot \left(\pi \cdot \left(\left(\pi \cdot \pi\right) \cdot -1.3333333333333333\right)\right), \pi \cdot 2\right)\right)\\
\mathbf{else}:\\
\;\;\;\;\sin \left(\pi \cdot \left(u2 + u2\right)\right) \cdot \sqrt{\mathsf{fma}\left(u1 \cdot u1, \mathsf{fma}\left(u1, \mathsf{fma}\left(u1, 0.25, 0.3333333333333333\right), 0.5\right), u1\right)}\\
\end{array}
\end{array}
if (*.f32 (*.f32 #s(literal 2 binary32) (PI.f32)) u2) < 0.0599999987Initial program 57.8%
lift-log.f32N/A
lift--.f32N/A
sub-negN/A
lower-log1p.f32N/A
lower-neg.f3298.5
Applied rewrites98.5%
Taylor expanded in u2 around 0
+-commutativeN/A
*-commutativeN/A
associate-*r*N/A
lower-*.f32N/A
associate-*r*N/A
*-commutativeN/A
+-commutativeN/A
*-commutativeN/A
associate-*r*N/A
*-commutativeN/A
unpow2N/A
associate-*l*N/A
lower-fma.f32N/A
Applied rewrites98.5%
if 0.0599999987 < (*.f32 (*.f32 #s(literal 2 binary32) (PI.f32)) u2) Initial program 59.2%
Taylor expanded in u1 around 0
lower-*.f32N/A
sub-negN/A
metadata-evalN/A
lower-fma.f32N/A
sub-negN/A
metadata-evalN/A
lower-fma.f32N/A
sub-negN/A
*-commutativeN/A
metadata-evalN/A
lower-fma.f3291.3
Applied rewrites91.3%
lift-sin.f32N/A
lift-*.f32N/A
lift-*.f32N/A
associate-*l*N/A
sin-2N/A
lower-*.f32N/A
lower-*.f32N/A
lower-sin.f32N/A
lower-*.f32N/A
lower-cos.f32N/A
lower-*.f3291.0
Applied rewrites91.0%
lift-*.f32N/A
lift-*.f32N/A
lift-sin.f32N/A
lift-cos.f32N/A
2-sinN/A
lift-*.f32N/A
associate-*l*N/A
lift-PI.f32N/A
lift-PI.f32N/A
lift-*.f32N/A
lift-*.f32N/A
lift-sin.f3291.3
lift-*.f32N/A
lift-*.f32N/A
associate-*l*N/A
lift-*.f32N/A
count-2N/A
Applied rewrites91.3%
Taylor expanded in u1 around 0
+-commutativeN/A
distribute-lft-inN/A
associate-*r*N/A
unpow2N/A
*-rgt-identityN/A
lower-fma.f32N/A
unpow2N/A
lower-*.f32N/A
+-commutativeN/A
lower-fma.f32N/A
+-commutativeN/A
*-commutativeN/A
lower-fma.f3291.6
Applied rewrites91.6%
Final simplification97.4%
(FPCore (cosTheta_i u1 u2)
:precision binary32
(if (<= (- 1.0 u1) 0.9300000071525574)
(*
(sqrt (- (log (- 1.0 u1))))
(* u2 (* PI (fma (* -1.3333333333333333 (* u2 u2)) (* PI PI) 2.0))))
(*
(sin (* PI (+ u2 u2)))
(sqrt (fma (* u1 u1) (fma u1 (fma u1 0.25 0.3333333333333333) 0.5) u1)))))
float code(float cosTheta_i, float u1, float u2) {
float tmp;
if ((1.0f - u1) <= 0.9300000071525574f) {
tmp = sqrtf(-logf((1.0f - u1))) * (u2 * (((float) M_PI) * fmaf((-1.3333333333333333f * (u2 * u2)), (((float) M_PI) * ((float) M_PI)), 2.0f)));
} else {
tmp = sinf((((float) M_PI) * (u2 + u2))) * sqrtf(fmaf((u1 * u1), fmaf(u1, fmaf(u1, 0.25f, 0.3333333333333333f), 0.5f), u1));
}
return tmp;
}
function code(cosTheta_i, u1, u2) tmp = Float32(0.0) if (Float32(Float32(1.0) - u1) <= Float32(0.9300000071525574)) tmp = Float32(sqrt(Float32(-log(Float32(Float32(1.0) - u1)))) * Float32(u2 * Float32(Float32(pi) * fma(Float32(Float32(-1.3333333333333333) * Float32(u2 * u2)), Float32(Float32(pi) * Float32(pi)), Float32(2.0))))); else tmp = Float32(sin(Float32(Float32(pi) * Float32(u2 + u2))) * sqrt(fma(Float32(u1 * u1), fma(u1, fma(u1, Float32(0.25), Float32(0.3333333333333333)), Float32(0.5)), u1))); end return tmp end
\begin{array}{l}
\\
\begin{array}{l}
\mathbf{if}\;1 - u1 \leq 0.9300000071525574:\\
\;\;\;\;\sqrt{-\log \left(1 - u1\right)} \cdot \left(u2 \cdot \left(\pi \cdot \mathsf{fma}\left(-1.3333333333333333 \cdot \left(u2 \cdot u2\right), \pi \cdot \pi, 2\right)\right)\right)\\
\mathbf{else}:\\
\;\;\;\;\sin \left(\pi \cdot \left(u2 + u2\right)\right) \cdot \sqrt{\mathsf{fma}\left(u1 \cdot u1, \mathsf{fma}\left(u1, \mathsf{fma}\left(u1, 0.25, 0.3333333333333333\right), 0.5\right), u1\right)}\\
\end{array}
\end{array}
if (-.f32 #s(literal 1 binary32) u1) < 0.930000007Initial program 98.2%
Taylor expanded in u2 around 0
*-commutativeN/A
associate-*r*N/A
*-commutativeN/A
+-commutativeN/A
lower-*.f32N/A
+-commutativeN/A
*-commutativeN/A
associate-*r*N/A
*-commutativeN/A
associate-*r*N/A
unpow3N/A
associate-*r*N/A
distribute-rgt-outN/A
Applied rewrites92.6%
if 0.930000007 < (-.f32 #s(literal 1 binary32) u1) Initial program 51.9%
Taylor expanded in u1 around 0
lower-*.f32N/A
sub-negN/A
metadata-evalN/A
lower-fma.f32N/A
sub-negN/A
metadata-evalN/A
lower-fma.f32N/A
sub-negN/A
*-commutativeN/A
metadata-evalN/A
lower-fma.f3297.9
Applied rewrites97.9%
lift-sin.f32N/A
lift-*.f32N/A
lift-*.f32N/A
associate-*l*N/A
sin-2N/A
lower-*.f32N/A
lower-*.f32N/A
lower-sin.f32N/A
lower-*.f32N/A
lower-cos.f32N/A
lower-*.f3297.8
Applied rewrites97.8%
lift-*.f32N/A
lift-*.f32N/A
lift-sin.f32N/A
lift-cos.f32N/A
2-sinN/A
lift-*.f32N/A
associate-*l*N/A
lift-PI.f32N/A
lift-PI.f32N/A
lift-*.f32N/A
lift-*.f32N/A
lift-sin.f3297.9
lift-*.f32N/A
lift-*.f32N/A
associate-*l*N/A
lift-*.f32N/A
count-2N/A
Applied rewrites97.9%
Taylor expanded in u1 around 0
+-commutativeN/A
distribute-lft-inN/A
associate-*r*N/A
unpow2N/A
*-rgt-identityN/A
lower-fma.f32N/A
unpow2N/A
lower-*.f32N/A
+-commutativeN/A
lower-fma.f32N/A
+-commutativeN/A
*-commutativeN/A
lower-fma.f3298.0
Applied rewrites98.0%
Final simplification97.3%
(FPCore (cosTheta_i u1 u2)
:precision binary32
(if (<= (* u2 (* PI 2.0)) 0.001820000004954636)
(* (sqrt (- (log1p (- u1)))) (* 2.0 (* PI u2)))
(*
(sin (* PI (+ u2 u2)))
(sqrt (fma (* u1 u1) (fma u1 0.3333333333333333 0.5) u1)))))
float code(float cosTheta_i, float u1, float u2) {
float tmp;
if ((u2 * (((float) M_PI) * 2.0f)) <= 0.001820000004954636f) {
tmp = sqrtf(-log1pf(-u1)) * (2.0f * (((float) M_PI) * u2));
} else {
tmp = sinf((((float) M_PI) * (u2 + u2))) * sqrtf(fmaf((u1 * u1), fmaf(u1, 0.3333333333333333f, 0.5f), u1));
}
return tmp;
}
function code(cosTheta_i, u1, u2) tmp = Float32(0.0) if (Float32(u2 * Float32(Float32(pi) * Float32(2.0))) <= Float32(0.001820000004954636)) tmp = Float32(sqrt(Float32(-log1p(Float32(-u1)))) * Float32(Float32(2.0) * Float32(Float32(pi) * u2))); else tmp = Float32(sin(Float32(Float32(pi) * Float32(u2 + u2))) * sqrt(fma(Float32(u1 * u1), fma(u1, Float32(0.3333333333333333), Float32(0.5)), u1))); end return tmp end
\begin{array}{l}
\\
\begin{array}{l}
\mathbf{if}\;u2 \cdot \left(\pi \cdot 2\right) \leq 0.001820000004954636:\\
\;\;\;\;\sqrt{-\mathsf{log1p}\left(-u1\right)} \cdot \left(2 \cdot \left(\pi \cdot u2\right)\right)\\
\mathbf{else}:\\
\;\;\;\;\sin \left(\pi \cdot \left(u2 + u2\right)\right) \cdot \sqrt{\mathsf{fma}\left(u1 \cdot u1, \mathsf{fma}\left(u1, 0.3333333333333333, 0.5\right), u1\right)}\\
\end{array}
\end{array}
if (*.f32 (*.f32 #s(literal 2 binary32) (PI.f32)) u2) < 0.00182Initial program 58.8%
lift-log.f32N/A
lift--.f32N/A
sub-negN/A
lower-log1p.f32N/A
lower-neg.f3298.5
Applied rewrites98.5%
Taylor expanded in u2 around 0
lower-*.f32N/A
lower-*.f32N/A
lower-PI.f3298.0
Applied rewrites98.0%
if 0.00182 < (*.f32 (*.f32 #s(literal 2 binary32) (PI.f32)) u2) Initial program 56.4%
Taylor expanded in u1 around 0
lower-*.f32N/A
sub-negN/A
metadata-evalN/A
lower-fma.f32N/A
sub-negN/A
metadata-evalN/A
lower-fma.f32N/A
sub-negN/A
*-commutativeN/A
metadata-evalN/A
lower-fma.f3293.5
Applied rewrites93.5%
lift-sin.f32N/A
lift-*.f32N/A
lift-*.f32N/A
associate-*l*N/A
sin-2N/A
lower-*.f32N/A
lower-*.f32N/A
lower-sin.f32N/A
lower-*.f32N/A
lower-cos.f32N/A
lower-*.f3293.4
Applied rewrites93.4%
lift-*.f32N/A
lift-*.f32N/A
lift-sin.f32N/A
lift-cos.f32N/A
2-sinN/A
lift-*.f32N/A
associate-*l*N/A
lift-PI.f32N/A
lift-PI.f32N/A
lift-*.f32N/A
lift-*.f32N/A
lift-sin.f3293.5
lift-*.f32N/A
lift-*.f32N/A
associate-*l*N/A
lift-*.f32N/A
count-2N/A
Applied rewrites93.5%
Taylor expanded in u1 around 0
+-commutativeN/A
distribute-lft-inN/A
associate-*r*N/A
unpow2N/A
*-rgt-identityN/A
lower-fma.f32N/A
unpow2N/A
lower-*.f32N/A
+-commutativeN/A
*-commutativeN/A
lower-fma.f3291.7
Applied rewrites91.7%
Final simplification96.0%
(FPCore (cosTheta_i u1 u2)
:precision binary32
(if (<= (- 1.0 u1) 0.9200000166893005)
(* (sqrt (- (log (- 1.0 u1)))) (* 2.0 (* PI u2)))
(*
(sin (* PI (+ u2 u2)))
(sqrt (fma (* u1 u1) (fma u1 (fma u1 0.25 0.3333333333333333) 0.5) u1)))))
float code(float cosTheta_i, float u1, float u2) {
float tmp;
if ((1.0f - u1) <= 0.9200000166893005f) {
tmp = sqrtf(-logf((1.0f - u1))) * (2.0f * (((float) M_PI) * u2));
} else {
tmp = sinf((((float) M_PI) * (u2 + u2))) * sqrtf(fmaf((u1 * u1), fmaf(u1, fmaf(u1, 0.25f, 0.3333333333333333f), 0.5f), u1));
}
return tmp;
}
function code(cosTheta_i, u1, u2) tmp = Float32(0.0) if (Float32(Float32(1.0) - u1) <= Float32(0.9200000166893005)) tmp = Float32(sqrt(Float32(-log(Float32(Float32(1.0) - u1)))) * Float32(Float32(2.0) * Float32(Float32(pi) * u2))); else tmp = Float32(sin(Float32(Float32(pi) * Float32(u2 + u2))) * sqrt(fma(Float32(u1 * u1), fma(u1, fma(u1, Float32(0.25), Float32(0.3333333333333333)), Float32(0.5)), u1))); end return tmp end
\begin{array}{l}
\\
\begin{array}{l}
\mathbf{if}\;1 - u1 \leq 0.9200000166893005:\\
\;\;\;\;\sqrt{-\log \left(1 - u1\right)} \cdot \left(2 \cdot \left(\pi \cdot u2\right)\right)\\
\mathbf{else}:\\
\;\;\;\;\sin \left(\pi \cdot \left(u2 + u2\right)\right) \cdot \sqrt{\mathsf{fma}\left(u1 \cdot u1, \mathsf{fma}\left(u1, \mathsf{fma}\left(u1, 0.25, 0.3333333333333333\right), 0.5\right), u1\right)}\\
\end{array}
\end{array}
if (-.f32 #s(literal 1 binary32) u1) < 0.920000017Initial program 98.3%
Taylor expanded in u2 around 0
lower-*.f32N/A
lower-*.f32N/A
lower-PI.f3288.9
Applied rewrites88.9%
if 0.920000017 < (-.f32 #s(literal 1 binary32) u1) Initial program 52.5%
Taylor expanded in u1 around 0
lower-*.f32N/A
sub-negN/A
metadata-evalN/A
lower-fma.f32N/A
sub-negN/A
metadata-evalN/A
lower-fma.f32N/A
sub-negN/A
*-commutativeN/A
metadata-evalN/A
lower-fma.f3297.7
Applied rewrites97.7%
lift-sin.f32N/A
lift-*.f32N/A
lift-*.f32N/A
associate-*l*N/A
sin-2N/A
lower-*.f32N/A
lower-*.f32N/A
lower-sin.f32N/A
lower-*.f32N/A
lower-cos.f32N/A
lower-*.f3297.6
Applied rewrites97.6%
lift-*.f32N/A
lift-*.f32N/A
lift-sin.f32N/A
lift-cos.f32N/A
2-sinN/A
lift-*.f32N/A
associate-*l*N/A
lift-PI.f32N/A
lift-PI.f32N/A
lift-*.f32N/A
lift-*.f32N/A
lift-sin.f3297.7
lift-*.f32N/A
lift-*.f32N/A
associate-*l*N/A
lift-*.f32N/A
count-2N/A
Applied rewrites97.7%
Taylor expanded in u1 around 0
+-commutativeN/A
distribute-lft-inN/A
associate-*r*N/A
unpow2N/A
*-rgt-identityN/A
lower-fma.f32N/A
unpow2N/A
lower-*.f32N/A
+-commutativeN/A
lower-fma.f32N/A
+-commutativeN/A
*-commutativeN/A
lower-fma.f3297.8
Applied rewrites97.8%
Final simplification96.7%
(FPCore (cosTheta_i u1 u2) :precision binary32 (if (<= (* u2 (* PI 2.0)) 0.001820000004954636) (* (sqrt (- (log1p (- u1)))) (* 2.0 (* PI u2))) (* (sin (* PI (+ u2 u2))) (sqrt (fma u1 (* u1 0.5) u1)))))
float code(float cosTheta_i, float u1, float u2) {
float tmp;
if ((u2 * (((float) M_PI) * 2.0f)) <= 0.001820000004954636f) {
tmp = sqrtf(-log1pf(-u1)) * (2.0f * (((float) M_PI) * u2));
} else {
tmp = sinf((((float) M_PI) * (u2 + u2))) * sqrtf(fmaf(u1, (u1 * 0.5f), u1));
}
return tmp;
}
function code(cosTheta_i, u1, u2) tmp = Float32(0.0) if (Float32(u2 * Float32(Float32(pi) * Float32(2.0))) <= Float32(0.001820000004954636)) tmp = Float32(sqrt(Float32(-log1p(Float32(-u1)))) * Float32(Float32(2.0) * Float32(Float32(pi) * u2))); else tmp = Float32(sin(Float32(Float32(pi) * Float32(u2 + u2))) * sqrt(fma(u1, Float32(u1 * Float32(0.5)), u1))); end return tmp end
\begin{array}{l}
\\
\begin{array}{l}
\mathbf{if}\;u2 \cdot \left(\pi \cdot 2\right) \leq 0.001820000004954636:\\
\;\;\;\;\sqrt{-\mathsf{log1p}\left(-u1\right)} \cdot \left(2 \cdot \left(\pi \cdot u2\right)\right)\\
\mathbf{else}:\\
\;\;\;\;\sin \left(\pi \cdot \left(u2 + u2\right)\right) \cdot \sqrt{\mathsf{fma}\left(u1, u1 \cdot 0.5, u1\right)}\\
\end{array}
\end{array}
if (*.f32 (*.f32 #s(literal 2 binary32) (PI.f32)) u2) < 0.00182Initial program 58.8%
lift-log.f32N/A
lift--.f32N/A
sub-negN/A
lower-log1p.f32N/A
lower-neg.f3298.5
Applied rewrites98.5%
Taylor expanded in u2 around 0
lower-*.f32N/A
lower-*.f32N/A
lower-PI.f3298.0
Applied rewrites98.0%
if 0.00182 < (*.f32 (*.f32 #s(literal 2 binary32) (PI.f32)) u2) Initial program 56.4%
Taylor expanded in u1 around 0
lower-*.f32N/A
sub-negN/A
metadata-evalN/A
lower-fma.f32N/A
sub-negN/A
metadata-evalN/A
lower-fma.f32N/A
sub-negN/A
*-commutativeN/A
metadata-evalN/A
lower-fma.f3293.5
Applied rewrites93.5%
lift-sin.f32N/A
lift-*.f32N/A
lift-*.f32N/A
associate-*l*N/A
sin-2N/A
lower-*.f32N/A
lower-*.f32N/A
lower-sin.f32N/A
lower-*.f32N/A
lower-cos.f32N/A
lower-*.f3293.4
Applied rewrites93.4%
lift-*.f32N/A
lift-*.f32N/A
lift-sin.f32N/A
lift-cos.f32N/A
2-sinN/A
lift-*.f32N/A
associate-*l*N/A
lift-PI.f32N/A
lift-PI.f32N/A
lift-*.f32N/A
lift-*.f32N/A
lift-sin.f3293.5
lift-*.f32N/A
lift-*.f32N/A
associate-*l*N/A
lift-*.f32N/A
count-2N/A
Applied rewrites93.5%
Taylor expanded in u1 around 0
+-commutativeN/A
distribute-lft-inN/A
*-rgt-identityN/A
lower-fma.f32N/A
*-commutativeN/A
lower-*.f3288.1
Applied rewrites88.1%
Final simplification94.8%
(FPCore (cosTheta_i u1 u2)
:precision binary32
(let* ((t_0 (* u2 (* PI 2.0))))
(if (<= t_0 0.006000000052154064)
(* (sqrt (- (log1p (- u1)))) (* 2.0 (* PI u2)))
(* (sqrt u1) (sin t_0)))))
float code(float cosTheta_i, float u1, float u2) {
float t_0 = u2 * (((float) M_PI) * 2.0f);
float tmp;
if (t_0 <= 0.006000000052154064f) {
tmp = sqrtf(-log1pf(-u1)) * (2.0f * (((float) M_PI) * u2));
} else {
tmp = sqrtf(u1) * sinf(t_0);
}
return tmp;
}
function code(cosTheta_i, u1, u2) t_0 = Float32(u2 * Float32(Float32(pi) * Float32(2.0))) tmp = Float32(0.0) if (t_0 <= Float32(0.006000000052154064)) tmp = Float32(sqrt(Float32(-log1p(Float32(-u1)))) * Float32(Float32(2.0) * Float32(Float32(pi) * u2))); else tmp = Float32(sqrt(u1) * sin(t_0)); end return tmp end
\begin{array}{l}
\\
\begin{array}{l}
t_0 := u2 \cdot \left(\pi \cdot 2\right)\\
\mathbf{if}\;t\_0 \leq 0.006000000052154064:\\
\;\;\;\;\sqrt{-\mathsf{log1p}\left(-u1\right)} \cdot \left(2 \cdot \left(\pi \cdot u2\right)\right)\\
\mathbf{else}:\\
\;\;\;\;\sqrt{u1} \cdot \sin t\_0\\
\end{array}
\end{array}
if (*.f32 (*.f32 #s(literal 2 binary32) (PI.f32)) u2) < 0.00600000005Initial program 59.1%
lift-log.f32N/A
lift--.f32N/A
sub-negN/A
lower-log1p.f32N/A
lower-neg.f3298.5
Applied rewrites98.5%
Taylor expanded in u2 around 0
lower-*.f32N/A
lower-*.f32N/A
lower-PI.f3296.7
Applied rewrites96.7%
if 0.00600000005 < (*.f32 (*.f32 #s(literal 2 binary32) (PI.f32)) u2) Initial program 54.7%
Applied rewrites50.9%
Taylor expanded in u1 around 0
lower-sqrt.f3277.3
Applied rewrites77.3%
Final simplification91.8%
(FPCore (cosTheta_i u1 u2)
:precision binary32
(let* ((t_0 (* u2 (* PI 2.0))))
(if (<= t_0 0.18000000715255737)
(*
(*
u2
(fma u2 (* u2 (* PI (* (* PI PI) -1.3333333333333333))) (* PI 2.0)))
(sqrt
(-
(*
u1
(fma u1 (fma u1 (fma u1 -0.25 -0.3333333333333333) -0.5) -1.0)))))
(* (sqrt u1) (sin t_0)))))
float code(float cosTheta_i, float u1, float u2) {
float t_0 = u2 * (((float) M_PI) * 2.0f);
float tmp;
if (t_0 <= 0.18000000715255737f) {
tmp = (u2 * fmaf(u2, (u2 * (((float) M_PI) * ((((float) M_PI) * ((float) M_PI)) * -1.3333333333333333f))), (((float) M_PI) * 2.0f))) * sqrtf(-(u1 * fmaf(u1, fmaf(u1, fmaf(u1, -0.25f, -0.3333333333333333f), -0.5f), -1.0f)));
} else {
tmp = sqrtf(u1) * sinf(t_0);
}
return tmp;
}
function code(cosTheta_i, u1, u2) t_0 = Float32(u2 * Float32(Float32(pi) * Float32(2.0))) tmp = Float32(0.0) if (t_0 <= Float32(0.18000000715255737)) tmp = Float32(Float32(u2 * fma(u2, Float32(u2 * Float32(Float32(pi) * Float32(Float32(Float32(pi) * Float32(pi)) * Float32(-1.3333333333333333)))), Float32(Float32(pi) * Float32(2.0)))) * sqrt(Float32(-Float32(u1 * fma(u1, fma(u1, fma(u1, Float32(-0.25), Float32(-0.3333333333333333)), Float32(-0.5)), Float32(-1.0)))))); else tmp = Float32(sqrt(u1) * sin(t_0)); end return tmp end
\begin{array}{l}
\\
\begin{array}{l}
t_0 := u2 \cdot \left(\pi \cdot 2\right)\\
\mathbf{if}\;t\_0 \leq 0.18000000715255737:\\
\;\;\;\;\left(u2 \cdot \mathsf{fma}\left(u2, u2 \cdot \left(\pi \cdot \left(\left(\pi \cdot \pi\right) \cdot -1.3333333333333333\right)\right), \pi \cdot 2\right)\right) \cdot \sqrt{-u1 \cdot \mathsf{fma}\left(u1, \mathsf{fma}\left(u1, \mathsf{fma}\left(u1, -0.25, -0.3333333333333333\right), -0.5\right), -1\right)}\\
\mathbf{else}:\\
\;\;\;\;\sqrt{u1} \cdot \sin t\_0\\
\end{array}
\end{array}
if (*.f32 (*.f32 #s(literal 2 binary32) (PI.f32)) u2) < 0.180000007Initial program 58.2%
Taylor expanded in u1 around 0
lower-*.f32N/A
sub-negN/A
metadata-evalN/A
lower-fma.f32N/A
sub-negN/A
metadata-evalN/A
lower-fma.f32N/A
sub-negN/A
*-commutativeN/A
metadata-evalN/A
lower-fma.f3293.1
Applied rewrites93.1%
Taylor expanded in u2 around 0
+-commutativeN/A
*-commutativeN/A
associate-*r*N/A
lower-*.f32N/A
associate-*r*N/A
*-commutativeN/A
+-commutativeN/A
*-commutativeN/A
associate-*r*N/A
*-commutativeN/A
unpow2N/A
associate-*l*N/A
lower-fma.f32N/A
Applied rewrites92.8%
if 0.180000007 < (*.f32 (*.f32 #s(literal 2 binary32) (PI.f32)) u2) Initial program 56.6%
Applied rewrites51.6%
Taylor expanded in u1 around 0
lower-sqrt.f3275.0
Applied rewrites75.0%
Final simplification90.4%
(FPCore (cosTheta_i u1 u2) :precision binary32 (* (* u2 (fma u2 (* u2 (* PI (* (* PI PI) -1.3333333333333333))) (* PI 2.0))) (sqrt (- (* u1 (fma u1 (fma u1 (fma u1 -0.25 -0.3333333333333333) -0.5) -1.0))))))
float code(float cosTheta_i, float u1, float u2) {
return (u2 * fmaf(u2, (u2 * (((float) M_PI) * ((((float) M_PI) * ((float) M_PI)) * -1.3333333333333333f))), (((float) M_PI) * 2.0f))) * sqrtf(-(u1 * fmaf(u1, fmaf(u1, fmaf(u1, -0.25f, -0.3333333333333333f), -0.5f), -1.0f)));
}
function code(cosTheta_i, u1, u2) return Float32(Float32(u2 * fma(u2, Float32(u2 * Float32(Float32(pi) * Float32(Float32(Float32(pi) * Float32(pi)) * Float32(-1.3333333333333333)))), Float32(Float32(pi) * Float32(2.0)))) * sqrt(Float32(-Float32(u1 * fma(u1, fma(u1, fma(u1, Float32(-0.25), Float32(-0.3333333333333333)), Float32(-0.5)), Float32(-1.0)))))) end
\begin{array}{l}
\\
\left(u2 \cdot \mathsf{fma}\left(u2, u2 \cdot \left(\pi \cdot \left(\left(\pi \cdot \pi\right) \cdot -1.3333333333333333\right)\right), \pi \cdot 2\right)\right) \cdot \sqrt{-u1 \cdot \mathsf{fma}\left(u1, \mathsf{fma}\left(u1, \mathsf{fma}\left(u1, -0.25, -0.3333333333333333\right), -0.5\right), -1\right)}
\end{array}
Initial program 58.0%
Taylor expanded in u1 around 0
lower-*.f32N/A
sub-negN/A
metadata-evalN/A
lower-fma.f32N/A
sub-negN/A
metadata-evalN/A
lower-fma.f32N/A
sub-negN/A
*-commutativeN/A
metadata-evalN/A
lower-fma.f3292.8
Applied rewrites92.8%
Taylor expanded in u2 around 0
+-commutativeN/A
*-commutativeN/A
associate-*r*N/A
lower-*.f32N/A
associate-*r*N/A
*-commutativeN/A
+-commutativeN/A
*-commutativeN/A
associate-*r*N/A
*-commutativeN/A
unpow2N/A
associate-*l*N/A
lower-fma.f32N/A
Applied rewrites86.0%
Final simplification86.0%
(FPCore (cosTheta_i u1 u2) :precision binary32 (* (* 2.0 (* PI u2)) (sqrt (- (* u1 (fma u1 (fma u1 (fma u1 -0.25 -0.3333333333333333) -0.5) -1.0))))))
float code(float cosTheta_i, float u1, float u2) {
return (2.0f * (((float) M_PI) * u2)) * sqrtf(-(u1 * fmaf(u1, fmaf(u1, fmaf(u1, -0.25f, -0.3333333333333333f), -0.5f), -1.0f)));
}
function code(cosTheta_i, u1, u2) return Float32(Float32(Float32(2.0) * Float32(Float32(pi) * u2)) * sqrt(Float32(-Float32(u1 * fma(u1, fma(u1, fma(u1, Float32(-0.25), Float32(-0.3333333333333333)), Float32(-0.5)), Float32(-1.0)))))) end
\begin{array}{l}
\\
\left(2 \cdot \left(\pi \cdot u2\right)\right) \cdot \sqrt{-u1 \cdot \mathsf{fma}\left(u1, \mathsf{fma}\left(u1, \mathsf{fma}\left(u1, -0.25, -0.3333333333333333\right), -0.5\right), -1\right)}
\end{array}
Initial program 58.0%
Taylor expanded in u1 around 0
lower-*.f32N/A
sub-negN/A
metadata-evalN/A
lower-fma.f32N/A
sub-negN/A
metadata-evalN/A
lower-fma.f32N/A
sub-negN/A
*-commutativeN/A
metadata-evalN/A
lower-fma.f3292.8
Applied rewrites92.8%
Taylor expanded in u2 around 0
lower-*.f32N/A
lower-*.f32N/A
lower-PI.f3280.0
Applied rewrites80.0%
Final simplification80.0%
(FPCore (cosTheta_i u1 u2) :precision binary32 (* (* 2.0 (* PI u2)) (sqrt u1)))
float code(float cosTheta_i, float u1, float u2) {
return (2.0f * (((float) M_PI) * u2)) * sqrtf(u1);
}
function code(cosTheta_i, u1, u2) return Float32(Float32(Float32(2.0) * Float32(Float32(pi) * u2)) * sqrt(u1)) end
function tmp = code(cosTheta_i, u1, u2) tmp = (single(2.0) * (single(pi) * u2)) * sqrt(u1); end
\begin{array}{l}
\\
\left(2 \cdot \left(\pi \cdot u2\right)\right) \cdot \sqrt{u1}
\end{array}
Initial program 58.0%
Applied rewrites84.1%
Taylor expanded in u2 around 0
associate-*r*N/A
*-commutativeN/A
associate-*r*N/A
*-commutativeN/A
lower-*.f32N/A
*-commutativeN/A
associate-*r*N/A
*-commutativeN/A
lower-*.f32N/A
lower-*.f32N/A
lower-PI.f32N/A
lower-sqrt.f32N/A
lower--.f32N/A
lower-log1p.f32N/A
+-commutativeN/A
unpow2N/A
lower-fma.f32N/A
lower-log1p.f32N/A
Applied rewrites73.3%
Taylor expanded in u1 around 0
Applied rewrites66.8%
Final simplification66.8%
herbie shell --seed 2024233
(FPCore (cosTheta_i u1 u2)
:name "Beckmann Sample, near normal, slope_y"
: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 (- (log (- 1.0 u1)))) (sin (* (* 2.0 PI) u2))))