
(FPCore (a1 a2 th) :precision binary64 (let* ((t_1 (/ (cos th) (sqrt 2.0)))) (+ (* t_1 (* a1 a1)) (* t_1 (* a2 a2)))))
double code(double a1, double a2, double th) {
double t_1 = cos(th) / sqrt(2.0);
return (t_1 * (a1 * a1)) + (t_1 * (a2 * a2));
}
module fmin_fmax_functions
implicit none
private
public fmax
public fmin
interface fmax
module procedure fmax88
module procedure fmax44
module procedure fmax84
module procedure fmax48
end interface
interface fmin
module procedure fmin88
module procedure fmin44
module procedure fmin84
module procedure fmin48
end interface
contains
real(8) function fmax88(x, y) result (res)
real(8), intent (in) :: x
real(8), intent (in) :: y
res = merge(y, merge(x, max(x, y), y /= y), x /= x)
end function
real(4) function fmax44(x, y) result (res)
real(4), intent (in) :: x
real(4), intent (in) :: y
res = merge(y, merge(x, max(x, y), y /= y), x /= x)
end function
real(8) function fmax84(x, y) result(res)
real(8), intent (in) :: x
real(4), intent (in) :: y
res = merge(dble(y), merge(x, max(x, dble(y)), y /= y), x /= x)
end function
real(8) function fmax48(x, y) result(res)
real(4), intent (in) :: x
real(8), intent (in) :: y
res = merge(y, merge(dble(x), max(dble(x), y), y /= y), x /= x)
end function
real(8) function fmin88(x, y) result (res)
real(8), intent (in) :: x
real(8), intent (in) :: y
res = merge(y, merge(x, min(x, y), y /= y), x /= x)
end function
real(4) function fmin44(x, y) result (res)
real(4), intent (in) :: x
real(4), intent (in) :: y
res = merge(y, merge(x, min(x, y), y /= y), x /= x)
end function
real(8) function fmin84(x, y) result(res)
real(8), intent (in) :: x
real(4), intent (in) :: y
res = merge(dble(y), merge(x, min(x, dble(y)), y /= y), x /= x)
end function
real(8) function fmin48(x, y) result(res)
real(4), intent (in) :: x
real(8), intent (in) :: y
res = merge(y, merge(dble(x), min(dble(x), y), y /= y), x /= x)
end function
end module
real(8) function code(a1, a2, th)
use fmin_fmax_functions
real(8), intent (in) :: a1
real(8), intent (in) :: a2
real(8), intent (in) :: th
real(8) :: t_1
t_1 = cos(th) / sqrt(2.0d0)
code = (t_1 * (a1 * a1)) + (t_1 * (a2 * a2))
end function
public static double code(double a1, double a2, double th) {
double t_1 = Math.cos(th) / Math.sqrt(2.0);
return (t_1 * (a1 * a1)) + (t_1 * (a2 * a2));
}
def code(a1, a2, th): t_1 = math.cos(th) / math.sqrt(2.0) return (t_1 * (a1 * a1)) + (t_1 * (a2 * a2))
function code(a1, a2, th) t_1 = Float64(cos(th) / sqrt(2.0)) return Float64(Float64(t_1 * Float64(a1 * a1)) + Float64(t_1 * Float64(a2 * a2))) end
function tmp = code(a1, a2, th) t_1 = cos(th) / sqrt(2.0); tmp = (t_1 * (a1 * a1)) + (t_1 * (a2 * a2)); end
code[a1_, a2_, th_] := Block[{t$95$1 = N[(N[Cos[th], $MachinePrecision] / N[Sqrt[2.0], $MachinePrecision]), $MachinePrecision]}, N[(N[(t$95$1 * N[(a1 * a1), $MachinePrecision]), $MachinePrecision] + N[(t$95$1 * N[(a2 * a2), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]]
\begin{array}{l}
\\
\begin{array}{l}
t_1 := \frac{\cos th}{\sqrt{2}}\\
t\_1 \cdot \left(a1 \cdot a1\right) + t\_1 \cdot \left(a2 \cdot a2\right)
\end{array}
\end{array}
Herbie found 7 alternatives:
| Alternative | Accuracy | Speedup |
|---|
(FPCore (a1 a2 th) :precision binary64 (let* ((t_1 (/ (cos th) (sqrt 2.0)))) (+ (* t_1 (* a1 a1)) (* t_1 (* a2 a2)))))
double code(double a1, double a2, double th) {
double t_1 = cos(th) / sqrt(2.0);
return (t_1 * (a1 * a1)) + (t_1 * (a2 * a2));
}
module fmin_fmax_functions
implicit none
private
public fmax
public fmin
interface fmax
module procedure fmax88
module procedure fmax44
module procedure fmax84
module procedure fmax48
end interface
interface fmin
module procedure fmin88
module procedure fmin44
module procedure fmin84
module procedure fmin48
end interface
contains
real(8) function fmax88(x, y) result (res)
real(8), intent (in) :: x
real(8), intent (in) :: y
res = merge(y, merge(x, max(x, y), y /= y), x /= x)
end function
real(4) function fmax44(x, y) result (res)
real(4), intent (in) :: x
real(4), intent (in) :: y
res = merge(y, merge(x, max(x, y), y /= y), x /= x)
end function
real(8) function fmax84(x, y) result(res)
real(8), intent (in) :: x
real(4), intent (in) :: y
res = merge(dble(y), merge(x, max(x, dble(y)), y /= y), x /= x)
end function
real(8) function fmax48(x, y) result(res)
real(4), intent (in) :: x
real(8), intent (in) :: y
res = merge(y, merge(dble(x), max(dble(x), y), y /= y), x /= x)
end function
real(8) function fmin88(x, y) result (res)
real(8), intent (in) :: x
real(8), intent (in) :: y
res = merge(y, merge(x, min(x, y), y /= y), x /= x)
end function
real(4) function fmin44(x, y) result (res)
real(4), intent (in) :: x
real(4), intent (in) :: y
res = merge(y, merge(x, min(x, y), y /= y), x /= x)
end function
real(8) function fmin84(x, y) result(res)
real(8), intent (in) :: x
real(4), intent (in) :: y
res = merge(dble(y), merge(x, min(x, dble(y)), y /= y), x /= x)
end function
real(8) function fmin48(x, y) result(res)
real(4), intent (in) :: x
real(8), intent (in) :: y
res = merge(y, merge(dble(x), min(dble(x), y), y /= y), x /= x)
end function
end module
real(8) function code(a1, a2, th)
use fmin_fmax_functions
real(8), intent (in) :: a1
real(8), intent (in) :: a2
real(8), intent (in) :: th
real(8) :: t_1
t_1 = cos(th) / sqrt(2.0d0)
code = (t_1 * (a1 * a1)) + (t_1 * (a2 * a2))
end function
public static double code(double a1, double a2, double th) {
double t_1 = Math.cos(th) / Math.sqrt(2.0);
return (t_1 * (a1 * a1)) + (t_1 * (a2 * a2));
}
def code(a1, a2, th): t_1 = math.cos(th) / math.sqrt(2.0) return (t_1 * (a1 * a1)) + (t_1 * (a2 * a2))
function code(a1, a2, th) t_1 = Float64(cos(th) / sqrt(2.0)) return Float64(Float64(t_1 * Float64(a1 * a1)) + Float64(t_1 * Float64(a2 * a2))) end
function tmp = code(a1, a2, th) t_1 = cos(th) / sqrt(2.0); tmp = (t_1 * (a1 * a1)) + (t_1 * (a2 * a2)); end
code[a1_, a2_, th_] := Block[{t$95$1 = N[(N[Cos[th], $MachinePrecision] / N[Sqrt[2.0], $MachinePrecision]), $MachinePrecision]}, N[(N[(t$95$1 * N[(a1 * a1), $MachinePrecision]), $MachinePrecision] + N[(t$95$1 * N[(a2 * a2), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]]
\begin{array}{l}
\\
\begin{array}{l}
t_1 := \frac{\cos th}{\sqrt{2}}\\
t\_1 \cdot \left(a1 \cdot a1\right) + t\_1 \cdot \left(a2 \cdot a2\right)
\end{array}
\end{array}
a2_m = (fabs.f64 a2) NOTE: a1, a2_m, and th should be sorted in increasing order before calling this function. (FPCore (a1 a2_m th) :precision binary64 (/ (fma (* (* (cos th) a2_m) a2_m) (sqrt 2.0) (* (* (* (sqrt 2.0) a1) a1) (cos th))) 2.0))
a2_m = fabs(a2);
assert(a1 < a2_m && a2_m < th);
double code(double a1, double a2_m, double th) {
return fma(((cos(th) * a2_m) * a2_m), sqrt(2.0), (((sqrt(2.0) * a1) * a1) * cos(th))) / 2.0;
}
a2_m = abs(a2) a1, a2_m, th = sort([a1, a2_m, th]) function code(a1, a2_m, th) return Float64(fma(Float64(Float64(cos(th) * a2_m) * a2_m), sqrt(2.0), Float64(Float64(Float64(sqrt(2.0) * a1) * a1) * cos(th))) / 2.0) end
a2_m = N[Abs[a2], $MachinePrecision] NOTE: a1, a2_m, and th should be sorted in increasing order before calling this function. code[a1_, a2$95$m_, th_] := N[(N[(N[(N[(N[Cos[th], $MachinePrecision] * a2$95$m), $MachinePrecision] * a2$95$m), $MachinePrecision] * N[Sqrt[2.0], $MachinePrecision] + N[(N[(N[(N[Sqrt[2.0], $MachinePrecision] * a1), $MachinePrecision] * a1), $MachinePrecision] * N[Cos[th], $MachinePrecision]), $MachinePrecision]), $MachinePrecision] / 2.0), $MachinePrecision]
\begin{array}{l}
a2_m = \left|a2\right|
\\
[a1, a2_m, th] = \mathsf{sort}([a1, a2_m, th])\\
\\
\frac{\mathsf{fma}\left(\left(\cos th \cdot a2\_m\right) \cdot a2\_m, \sqrt{2}, \left(\left(\sqrt{2} \cdot a1\right) \cdot a1\right) \cdot \cos th\right)}{2}
\end{array}
Initial program 99.6%
lift-+.f64N/A
lift-*.f64N/A
lift-/.f64N/A
lift-cos.f64N/A
lift-sqrt.f64N/A
lift-*.f64N/A
pow2N/A
lift-*.f64N/A
lift-/.f64N/A
lift-cos.f64N/A
lift-sqrt.f64N/A
lift-*.f64N/A
pow2N/A
*-commutativeN/A
associate-/l*N/A
*-commutativeN/A
associate-/l*N/A
Applied rewrites99.6%
a2_m = (fabs.f64 a2)
NOTE: a1, a2_m, and th should be sorted in increasing order before calling this function.
(FPCore (a1 a2_m th)
:precision binary64
(let* ((t_1 (/ (cos th) (sqrt 2.0))))
(if (<= (+ (* t_1 (* a1 a1)) (* t_1 (* a2_m a2_m))) -2e-150)
(* (* 0.5 (* a2_m a2_m)) (* (sqrt 2.0) (fma -0.5 (* th th) 1.0)))
(* (* (sqrt 2.0) (fma a1 a1 (* a2_m a2_m))) 0.5))))a2_m = fabs(a2);
assert(a1 < a2_m && a2_m < th);
double code(double a1, double a2_m, double th) {
double t_1 = cos(th) / sqrt(2.0);
double tmp;
if (((t_1 * (a1 * a1)) + (t_1 * (a2_m * a2_m))) <= -2e-150) {
tmp = (0.5 * (a2_m * a2_m)) * (sqrt(2.0) * fma(-0.5, (th * th), 1.0));
} else {
tmp = (sqrt(2.0) * fma(a1, a1, (a2_m * a2_m))) * 0.5;
}
return tmp;
}
a2_m = abs(a2) a1, a2_m, th = sort([a1, a2_m, th]) function code(a1, a2_m, th) t_1 = Float64(cos(th) / sqrt(2.0)) tmp = 0.0 if (Float64(Float64(t_1 * Float64(a1 * a1)) + Float64(t_1 * Float64(a2_m * a2_m))) <= -2e-150) tmp = Float64(Float64(0.5 * Float64(a2_m * a2_m)) * Float64(sqrt(2.0) * fma(-0.5, Float64(th * th), 1.0))); else tmp = Float64(Float64(sqrt(2.0) * fma(a1, a1, Float64(a2_m * a2_m))) * 0.5); end return tmp end
a2_m = N[Abs[a2], $MachinePrecision]
NOTE: a1, a2_m, and th should be sorted in increasing order before calling this function.
code[a1_, a2$95$m_, th_] := Block[{t$95$1 = N[(N[Cos[th], $MachinePrecision] / N[Sqrt[2.0], $MachinePrecision]), $MachinePrecision]}, If[LessEqual[N[(N[(t$95$1 * N[(a1 * a1), $MachinePrecision]), $MachinePrecision] + N[(t$95$1 * N[(a2$95$m * a2$95$m), $MachinePrecision]), $MachinePrecision]), $MachinePrecision], -2e-150], N[(N[(0.5 * N[(a2$95$m * a2$95$m), $MachinePrecision]), $MachinePrecision] * N[(N[Sqrt[2.0], $MachinePrecision] * N[(-0.5 * N[(th * th), $MachinePrecision] + 1.0), $MachinePrecision]), $MachinePrecision]), $MachinePrecision], N[(N[(N[Sqrt[2.0], $MachinePrecision] * N[(a1 * a1 + N[(a2$95$m * a2$95$m), $MachinePrecision]), $MachinePrecision]), $MachinePrecision] * 0.5), $MachinePrecision]]]
\begin{array}{l}
a2_m = \left|a2\right|
\\
[a1, a2_m, th] = \mathsf{sort}([a1, a2_m, th])\\
\\
\begin{array}{l}
t_1 := \frac{\cos th}{\sqrt{2}}\\
\mathbf{if}\;t\_1 \cdot \left(a1 \cdot a1\right) + t\_1 \cdot \left(a2\_m \cdot a2\_m\right) \leq -2 \cdot 10^{-150}:\\
\;\;\;\;\left(0.5 \cdot \left(a2\_m \cdot a2\_m\right)\right) \cdot \left(\sqrt{2} \cdot \mathsf{fma}\left(-0.5, th \cdot th, 1\right)\right)\\
\mathbf{else}:\\
\;\;\;\;\left(\sqrt{2} \cdot \mathsf{fma}\left(a1, a1, a2\_m \cdot a2\_m\right)\right) \cdot 0.5\\
\end{array}
\end{array}
if (+.f64 (*.f64 (/.f64 (cos.f64 th) (sqrt.f64 #s(literal 2 binary64))) (*.f64 a1 a1)) (*.f64 (/.f64 (cos.f64 th) (sqrt.f64 #s(literal 2 binary64))) (*.f64 a2 a2))) < -2.00000000000000001e-150Initial program 99.6%
lift-+.f64N/A
lift-*.f64N/A
lift-/.f64N/A
lift-cos.f64N/A
lift-sqrt.f64N/A
lift-*.f64N/A
pow2N/A
lift-*.f64N/A
lift-/.f64N/A
lift-cos.f64N/A
lift-sqrt.f64N/A
lift-*.f64N/A
pow2N/A
*-commutativeN/A
associate-/l*N/A
*-commutativeN/A
associate-/l*N/A
Applied rewrites99.6%
Taylor expanded in a1 around 0
associate-*r*N/A
metadata-evalN/A
lower-*.f64N/A
metadata-evalN/A
lower-*.f64N/A
pow2N/A
lift-*.f64N/A
*-commutativeN/A
lower-*.f64N/A
lift-sqrt.f64N/A
lift-cos.f6476.6
Applied rewrites76.6%
Taylor expanded in th around 0
+-commutativeN/A
lower-fma.f64N/A
unpow2N/A
lower-*.f6449.1
Applied rewrites49.1%
if -2.00000000000000001e-150 < (+.f64 (*.f64 (/.f64 (cos.f64 th) (sqrt.f64 #s(literal 2 binary64))) (*.f64 a1 a1)) (*.f64 (/.f64 (cos.f64 th) (sqrt.f64 #s(literal 2 binary64))) (*.f64 a2 a2))) Initial program 99.6%
lift-+.f64N/A
lift-*.f64N/A
lift-/.f64N/A
lift-cos.f64N/A
lift-sqrt.f64N/A
lift-*.f64N/A
pow2N/A
lift-*.f64N/A
lift-/.f64N/A
lift-cos.f64N/A
lift-sqrt.f64N/A
lift-*.f64N/A
pow2N/A
*-commutativeN/A
associate-/l*N/A
*-commutativeN/A
associate-/l*N/A
Applied rewrites99.6%
Taylor expanded in th around 0
*-commutativeN/A
lower-*.f64N/A
distribute-rgt-outN/A
lower-*.f64N/A
lift-sqrt.f64N/A
pow2N/A
pow2N/A
lift-fma.f64N/A
lift-*.f6484.1
Applied rewrites84.1%
a2_m = (fabs.f64 a2) NOTE: a1, a2_m, and th should be sorted in increasing order before calling this function. (FPCore (a1 a2_m th) :precision binary64 (* (* (* (fma a1 a1 (* a2_m a2_m)) (sqrt 2.0)) (cos th)) 0.5))
a2_m = fabs(a2);
assert(a1 < a2_m && a2_m < th);
double code(double a1, double a2_m, double th) {
return ((fma(a1, a1, (a2_m * a2_m)) * sqrt(2.0)) * cos(th)) * 0.5;
}
a2_m = abs(a2) a1, a2_m, th = sort([a1, a2_m, th]) function code(a1, a2_m, th) return Float64(Float64(Float64(fma(a1, a1, Float64(a2_m * a2_m)) * sqrt(2.0)) * cos(th)) * 0.5) end
a2_m = N[Abs[a2], $MachinePrecision] NOTE: a1, a2_m, and th should be sorted in increasing order before calling this function. code[a1_, a2$95$m_, th_] := N[(N[(N[(N[(a1 * a1 + N[(a2$95$m * a2$95$m), $MachinePrecision]), $MachinePrecision] * N[Sqrt[2.0], $MachinePrecision]), $MachinePrecision] * N[Cos[th], $MachinePrecision]), $MachinePrecision] * 0.5), $MachinePrecision]
\begin{array}{l}
a2_m = \left|a2\right|
\\
[a1, a2_m, th] = \mathsf{sort}([a1, a2_m, th])\\
\\
\left(\left(\mathsf{fma}\left(a1, a1, a2\_m \cdot a2\_m\right) \cdot \sqrt{2}\right) \cdot \cos th\right) \cdot 0.5
\end{array}
Initial program 99.6%
lift-+.f64N/A
lift-*.f64N/A
lift-/.f64N/A
lift-cos.f64N/A
lift-sqrt.f64N/A
lift-*.f64N/A
pow2N/A
lift-*.f64N/A
lift-/.f64N/A
lift-cos.f64N/A
lift-sqrt.f64N/A
lift-*.f64N/A
pow2N/A
*-commutativeN/A
associate-/l*N/A
*-commutativeN/A
associate-/l*N/A
Applied rewrites99.6%
Taylor expanded in a1 around 0
distribute-lft-inN/A
*-commutativeN/A
lower-*.f64N/A
distribute-rgt-outN/A
lower-*.f64N/A
*-commutativeN/A
lower-*.f64N/A
lift-sqrt.f64N/A
lift-cos.f64N/A
pow2N/A
pow2N/A
lift-fma.f64N/A
lift-*.f6499.6
Applied rewrites99.6%
lift-*.f64N/A
lift-sqrt.f64N/A
lift-cos.f64N/A
*-commutativeN/A
lift-*.f64N/A
lift-fma.f64N/A
pow2N/A
pow2N/A
lower-*.f64N/A
associate-*r*N/A
*-commutativeN/A
lower-*.f64N/A
*-commutativeN/A
lower-*.f64N/A
pow2N/A
pow2N/A
lift-fma.f64N/A
lift-*.f64N/A
lift-sqrt.f64N/A
lift-cos.f6499.6
Applied rewrites99.6%
a2_m = (fabs.f64 a2) NOTE: a1, a2_m, and th should be sorted in increasing order before calling this function. (FPCore (a1 a2_m th) :precision binary64 (* (* (* (sqrt 2.0) (cos th)) (fma a1 a1 (* a2_m a2_m))) 0.5))
a2_m = fabs(a2);
assert(a1 < a2_m && a2_m < th);
double code(double a1, double a2_m, double th) {
return ((sqrt(2.0) * cos(th)) * fma(a1, a1, (a2_m * a2_m))) * 0.5;
}
a2_m = abs(a2) a1, a2_m, th = sort([a1, a2_m, th]) function code(a1, a2_m, th) return Float64(Float64(Float64(sqrt(2.0) * cos(th)) * fma(a1, a1, Float64(a2_m * a2_m))) * 0.5) end
a2_m = N[Abs[a2], $MachinePrecision] NOTE: a1, a2_m, and th should be sorted in increasing order before calling this function. code[a1_, a2$95$m_, th_] := N[(N[(N[(N[Sqrt[2.0], $MachinePrecision] * N[Cos[th], $MachinePrecision]), $MachinePrecision] * N[(a1 * a1 + N[(a2$95$m * a2$95$m), $MachinePrecision]), $MachinePrecision]), $MachinePrecision] * 0.5), $MachinePrecision]
\begin{array}{l}
a2_m = \left|a2\right|
\\
[a1, a2_m, th] = \mathsf{sort}([a1, a2_m, th])\\
\\
\left(\left(\sqrt{2} \cdot \cos th\right) \cdot \mathsf{fma}\left(a1, a1, a2\_m \cdot a2\_m\right)\right) \cdot 0.5
\end{array}
Initial program 99.6%
lift-+.f64N/A
lift-*.f64N/A
lift-/.f64N/A
lift-cos.f64N/A
lift-sqrt.f64N/A
lift-*.f64N/A
pow2N/A
lift-*.f64N/A
lift-/.f64N/A
lift-cos.f64N/A
lift-sqrt.f64N/A
lift-*.f64N/A
pow2N/A
*-commutativeN/A
associate-/l*N/A
*-commutativeN/A
associate-/l*N/A
Applied rewrites99.6%
Taylor expanded in a1 around 0
distribute-lft-inN/A
*-commutativeN/A
lower-*.f64N/A
distribute-rgt-outN/A
lower-*.f64N/A
*-commutativeN/A
lower-*.f64N/A
lift-sqrt.f64N/A
lift-cos.f64N/A
pow2N/A
pow2N/A
lift-fma.f64N/A
lift-*.f6499.6
Applied rewrites99.6%
a2_m = (fabs.f64 a2) NOTE: a1, a2_m, and th should be sorted in increasing order before calling this function. (FPCore (a1 a2_m th) :precision binary64 (* (* (* (* a2_m (cos th)) a2_m) (sqrt 2.0)) 0.5))
a2_m = fabs(a2);
assert(a1 < a2_m && a2_m < th);
double code(double a1, double a2_m, double th) {
return (((a2_m * cos(th)) * a2_m) * sqrt(2.0)) * 0.5;
}
a2_m = private
NOTE: a1, a2_m, and th should be sorted in increasing order before calling this function.
module fmin_fmax_functions
implicit none
private
public fmax
public fmin
interface fmax
module procedure fmax88
module procedure fmax44
module procedure fmax84
module procedure fmax48
end interface
interface fmin
module procedure fmin88
module procedure fmin44
module procedure fmin84
module procedure fmin48
end interface
contains
real(8) function fmax88(x, y) result (res)
real(8), intent (in) :: x
real(8), intent (in) :: y
res = merge(y, merge(x, max(x, y), y /= y), x /= x)
end function
real(4) function fmax44(x, y) result (res)
real(4), intent (in) :: x
real(4), intent (in) :: y
res = merge(y, merge(x, max(x, y), y /= y), x /= x)
end function
real(8) function fmax84(x, y) result(res)
real(8), intent (in) :: x
real(4), intent (in) :: y
res = merge(dble(y), merge(x, max(x, dble(y)), y /= y), x /= x)
end function
real(8) function fmax48(x, y) result(res)
real(4), intent (in) :: x
real(8), intent (in) :: y
res = merge(y, merge(dble(x), max(dble(x), y), y /= y), x /= x)
end function
real(8) function fmin88(x, y) result (res)
real(8), intent (in) :: x
real(8), intent (in) :: y
res = merge(y, merge(x, min(x, y), y /= y), x /= x)
end function
real(4) function fmin44(x, y) result (res)
real(4), intent (in) :: x
real(4), intent (in) :: y
res = merge(y, merge(x, min(x, y), y /= y), x /= x)
end function
real(8) function fmin84(x, y) result(res)
real(8), intent (in) :: x
real(4), intent (in) :: y
res = merge(dble(y), merge(x, min(x, dble(y)), y /= y), x /= x)
end function
real(8) function fmin48(x, y) result(res)
real(4), intent (in) :: x
real(8), intent (in) :: y
res = merge(y, merge(dble(x), min(dble(x), y), y /= y), x /= x)
end function
end module
real(8) function code(a1, a2_m, th)
use fmin_fmax_functions
real(8), intent (in) :: a1
real(8), intent (in) :: a2_m
real(8), intent (in) :: th
code = (((a2_m * cos(th)) * a2_m) * sqrt(2.0d0)) * 0.5d0
end function
a2_m = Math.abs(a2);
assert a1 < a2_m && a2_m < th;
public static double code(double a1, double a2_m, double th) {
return (((a2_m * Math.cos(th)) * a2_m) * Math.sqrt(2.0)) * 0.5;
}
a2_m = math.fabs(a2) [a1, a2_m, th] = sort([a1, a2_m, th]) def code(a1, a2_m, th): return (((a2_m * math.cos(th)) * a2_m) * math.sqrt(2.0)) * 0.5
a2_m = abs(a2) a1, a2_m, th = sort([a1, a2_m, th]) function code(a1, a2_m, th) return Float64(Float64(Float64(Float64(a2_m * cos(th)) * a2_m) * sqrt(2.0)) * 0.5) end
a2_m = abs(a2);
a1, a2_m, th = num2cell(sort([a1, a2_m, th])){:}
function tmp = code(a1, a2_m, th)
tmp = (((a2_m * cos(th)) * a2_m) * sqrt(2.0)) * 0.5;
end
a2_m = N[Abs[a2], $MachinePrecision] NOTE: a1, a2_m, and th should be sorted in increasing order before calling this function. code[a1_, a2$95$m_, th_] := N[(N[(N[(N[(a2$95$m * N[Cos[th], $MachinePrecision]), $MachinePrecision] * a2$95$m), $MachinePrecision] * N[Sqrt[2.0], $MachinePrecision]), $MachinePrecision] * 0.5), $MachinePrecision]
\begin{array}{l}
a2_m = \left|a2\right|
\\
[a1, a2_m, th] = \mathsf{sort}([a1, a2_m, th])\\
\\
\left(\left(\left(a2\_m \cdot \cos th\right) \cdot a2\_m\right) \cdot \sqrt{2}\right) \cdot 0.5
\end{array}
Initial program 99.6%
lift-+.f64N/A
lift-*.f64N/A
lift-/.f64N/A
lift-cos.f64N/A
lift-sqrt.f64N/A
lift-*.f64N/A
pow2N/A
lift-*.f64N/A
lift-/.f64N/A
lift-cos.f64N/A
lift-sqrt.f64N/A
lift-*.f64N/A
pow2N/A
*-commutativeN/A
associate-/l*N/A
*-commutativeN/A
associate-/l*N/A
Applied rewrites99.6%
Taylor expanded in a1 around 0
distribute-lft-inN/A
*-commutativeN/A
lower-*.f64N/A
distribute-rgt-outN/A
lower-*.f64N/A
*-commutativeN/A
lower-*.f64N/A
lift-sqrt.f64N/A
lift-cos.f64N/A
pow2N/A
pow2N/A
lift-fma.f64N/A
lift-*.f6499.6
Applied rewrites99.6%
Taylor expanded in a1 around 0
associate-*r*N/A
lower-*.f64N/A
*-commutativeN/A
pow2N/A
associate-*l*N/A
lower-*.f64N/A
*-commutativeN/A
lower-*.f64N/A
lift-cos.f64N/A
lift-sqrt.f6478.0
Applied rewrites78.0%
a2_m = (fabs.f64 a2) NOTE: a1, a2_m, and th should be sorted in increasing order before calling this function. (FPCore (a1 a2_m th) :precision binary64 (* (* (sqrt 2.0) (fma a1 a1 (* a2_m a2_m))) 0.5))
a2_m = fabs(a2);
assert(a1 < a2_m && a2_m < th);
double code(double a1, double a2_m, double th) {
return (sqrt(2.0) * fma(a1, a1, (a2_m * a2_m))) * 0.5;
}
a2_m = abs(a2) a1, a2_m, th = sort([a1, a2_m, th]) function code(a1, a2_m, th) return Float64(Float64(sqrt(2.0) * fma(a1, a1, Float64(a2_m * a2_m))) * 0.5) end
a2_m = N[Abs[a2], $MachinePrecision] NOTE: a1, a2_m, and th should be sorted in increasing order before calling this function. code[a1_, a2$95$m_, th_] := N[(N[(N[Sqrt[2.0], $MachinePrecision] * N[(a1 * a1 + N[(a2$95$m * a2$95$m), $MachinePrecision]), $MachinePrecision]), $MachinePrecision] * 0.5), $MachinePrecision]
\begin{array}{l}
a2_m = \left|a2\right|
\\
[a1, a2_m, th] = \mathsf{sort}([a1, a2_m, th])\\
\\
\left(\sqrt{2} \cdot \mathsf{fma}\left(a1, a1, a2\_m \cdot a2\_m\right)\right) \cdot 0.5
\end{array}
Initial program 99.6%
lift-+.f64N/A
lift-*.f64N/A
lift-/.f64N/A
lift-cos.f64N/A
lift-sqrt.f64N/A
lift-*.f64N/A
pow2N/A
lift-*.f64N/A
lift-/.f64N/A
lift-cos.f64N/A
lift-sqrt.f64N/A
lift-*.f64N/A
pow2N/A
*-commutativeN/A
associate-/l*N/A
*-commutativeN/A
associate-/l*N/A
Applied rewrites99.6%
Taylor expanded in th around 0
*-commutativeN/A
lower-*.f64N/A
distribute-rgt-outN/A
lower-*.f64N/A
lift-sqrt.f64N/A
pow2N/A
pow2N/A
lift-fma.f64N/A
lift-*.f6465.8
Applied rewrites65.8%
a2_m = (fabs.f64 a2) NOTE: a1, a2_m, and th should be sorted in increasing order before calling this function. (FPCore (a1 a2_m th) :precision binary64 (* (* 0.5 (* a2_m a2_m)) (sqrt 2.0)))
a2_m = fabs(a2);
assert(a1 < a2_m && a2_m < th);
double code(double a1, double a2_m, double th) {
return (0.5 * (a2_m * a2_m)) * sqrt(2.0);
}
a2_m = private
NOTE: a1, a2_m, and th should be sorted in increasing order before calling this function.
module fmin_fmax_functions
implicit none
private
public fmax
public fmin
interface fmax
module procedure fmax88
module procedure fmax44
module procedure fmax84
module procedure fmax48
end interface
interface fmin
module procedure fmin88
module procedure fmin44
module procedure fmin84
module procedure fmin48
end interface
contains
real(8) function fmax88(x, y) result (res)
real(8), intent (in) :: x
real(8), intent (in) :: y
res = merge(y, merge(x, max(x, y), y /= y), x /= x)
end function
real(4) function fmax44(x, y) result (res)
real(4), intent (in) :: x
real(4), intent (in) :: y
res = merge(y, merge(x, max(x, y), y /= y), x /= x)
end function
real(8) function fmax84(x, y) result(res)
real(8), intent (in) :: x
real(4), intent (in) :: y
res = merge(dble(y), merge(x, max(x, dble(y)), y /= y), x /= x)
end function
real(8) function fmax48(x, y) result(res)
real(4), intent (in) :: x
real(8), intent (in) :: y
res = merge(y, merge(dble(x), max(dble(x), y), y /= y), x /= x)
end function
real(8) function fmin88(x, y) result (res)
real(8), intent (in) :: x
real(8), intent (in) :: y
res = merge(y, merge(x, min(x, y), y /= y), x /= x)
end function
real(4) function fmin44(x, y) result (res)
real(4), intent (in) :: x
real(4), intent (in) :: y
res = merge(y, merge(x, min(x, y), y /= y), x /= x)
end function
real(8) function fmin84(x, y) result(res)
real(8), intent (in) :: x
real(4), intent (in) :: y
res = merge(dble(y), merge(x, min(x, dble(y)), y /= y), x /= x)
end function
real(8) function fmin48(x, y) result(res)
real(4), intent (in) :: x
real(8), intent (in) :: y
res = merge(y, merge(dble(x), min(dble(x), y), y /= y), x /= x)
end function
end module
real(8) function code(a1, a2_m, th)
use fmin_fmax_functions
real(8), intent (in) :: a1
real(8), intent (in) :: a2_m
real(8), intent (in) :: th
code = (0.5d0 * (a2_m * a2_m)) * sqrt(2.0d0)
end function
a2_m = Math.abs(a2);
assert a1 < a2_m && a2_m < th;
public static double code(double a1, double a2_m, double th) {
return (0.5 * (a2_m * a2_m)) * Math.sqrt(2.0);
}
a2_m = math.fabs(a2) [a1, a2_m, th] = sort([a1, a2_m, th]) def code(a1, a2_m, th): return (0.5 * (a2_m * a2_m)) * math.sqrt(2.0)
a2_m = abs(a2) a1, a2_m, th = sort([a1, a2_m, th]) function code(a1, a2_m, th) return Float64(Float64(0.5 * Float64(a2_m * a2_m)) * sqrt(2.0)) end
a2_m = abs(a2);
a1, a2_m, th = num2cell(sort([a1, a2_m, th])){:}
function tmp = code(a1, a2_m, th)
tmp = (0.5 * (a2_m * a2_m)) * sqrt(2.0);
end
a2_m = N[Abs[a2], $MachinePrecision] NOTE: a1, a2_m, and th should be sorted in increasing order before calling this function. code[a1_, a2$95$m_, th_] := N[(N[(0.5 * N[(a2$95$m * a2$95$m), $MachinePrecision]), $MachinePrecision] * N[Sqrt[2.0], $MachinePrecision]), $MachinePrecision]
\begin{array}{l}
a2_m = \left|a2\right|
\\
[a1, a2_m, th] = \mathsf{sort}([a1, a2_m, th])\\
\\
\left(0.5 \cdot \left(a2\_m \cdot a2\_m\right)\right) \cdot \sqrt{2}
\end{array}
Initial program 99.6%
lift-+.f64N/A
lift-*.f64N/A
lift-/.f64N/A
lift-cos.f64N/A
lift-sqrt.f64N/A
lift-*.f64N/A
pow2N/A
lift-*.f64N/A
lift-/.f64N/A
lift-cos.f64N/A
lift-sqrt.f64N/A
lift-*.f64N/A
pow2N/A
*-commutativeN/A
associate-/l*N/A
*-commutativeN/A
associate-/l*N/A
Applied rewrites99.6%
Taylor expanded in a1 around 0
associate-*r*N/A
metadata-evalN/A
lower-*.f64N/A
metadata-evalN/A
lower-*.f64N/A
pow2N/A
lift-*.f64N/A
*-commutativeN/A
lower-*.f64N/A
lift-sqrt.f64N/A
lift-cos.f6478.0
Applied rewrites78.0%
Taylor expanded in th around 0
lift-sqrt.f6452.4
Applied rewrites52.4%
herbie shell --seed 2025091
(FPCore (a1 a2 th)
:name "Migdal et al, Equation (64)"
:precision binary64
(+ (* (/ (cos th) (sqrt 2.0)) (* a1 a1)) (* (/ (cos th) (sqrt 2.0)) (* a2 a2))))