Data.Colour.RGBSpace.HSV:hsv from colour-2.3.3, J

Percentage Accurate: 95.8% → 99.9%

Time: 9.7s

Alternatives: 10

Speedup: 0.8×

• 21.5% of points produce a very large (infinite) output. You may want to add a precondition.

Specification

Math

\[\begin{array}{l} \\ x \cdot \left(1 - \left(1 - y\right) \cdot z\right) \end{array} \]

FPCore

(FPCore (x y z) :precision binary64 (* x (- 1.0 (* (- 1.0 y) z))))

C

double code(double x, double y, double z) {
	return x * (1.0 - ((1.0 - y) * z));
}

Fortran

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(x, y, z)
use fmin_fmax_functions
    real(8), intent (in) :: x
    real(8), intent (in) :: y
    real(8), intent (in) :: z
    code = x * (1.0d0 - ((1.0d0 - y) * z))
end function

Java

public static double code(double x, double y, double z) {
	return x * (1.0 - ((1.0 - y) * z));
}

Python

def code(x, y, z):
	return x * (1.0 - ((1.0 - y) * z))

Julia

function code(x, y, z)
	return Float64(x * Float64(1.0 - Float64(Float64(1.0 - y) * z)))
end

MATLAB

function tmp = code(x, y, z)
	tmp = x * (1.0 - ((1.0 - y) * z));
end

Wolfram

code[x_, y_, z_] := N[(x * N[(1.0 - N[(N[(1.0 - y), $MachinePrecision] * z), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]

Initial Program: 95.8% accurate, 1.0× speedup

Math

\[\begin{array}{l} \\ x \cdot \left(1 - \left(1 - y\right) \cdot z\right) \end{array} \]

FPCore

(FPCore (x y z) :precision binary64 (* x (- 1.0 (* (- 1.0 y) z))))

C

double code(double x, double y, double z) {
	return x * (1.0 - ((1.0 - y) * z));
}

Fortran

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(x, y, z)
use fmin_fmax_functions
    real(8), intent (in) :: x
    real(8), intent (in) :: y
    real(8), intent (in) :: z
    code = x * (1.0d0 - ((1.0d0 - y) * z))
end function

Java

public static double code(double x, double y, double z) {
	return x * (1.0 - ((1.0 - y) * z));
}

Python

def code(x, y, z):
	return x * (1.0 - ((1.0 - y) * z))

Julia

function code(x, y, z)
	return Float64(x * Float64(1.0 - Float64(Float64(1.0 - y) * z)))
end

MATLAB

function tmp = code(x, y, z)
	tmp = x * (1.0 - ((1.0 - y) * z));
end

Wolfram

code[x_, y_, z_] := N[(x * N[(1.0 - N[(N[(1.0 - y), $MachinePrecision] * z), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]

Alternative 1: 99.9% accurate, 0.8× speedup

Math

\[\begin{array}{l} x\_m = \left|x\right| \\ x\_s = \mathsf{copysign}\left(1, x\right) \\ x\_s \cdot \begin{array}{l} \mathbf{if}\;x\_m \leq 2 \cdot 10^{-77}:\\ \;\;\;\;\mathsf{fma}\left(\left(y - 1\right) \cdot x\_m, z, x\_m\right)\\ \mathbf{else}:\\ \;\;\;\;\mathsf{fma}\left(y - 1, z \cdot x\_m, x\_m\right)\\ \end{array} \end{array} \]

FPCore

x\_m = (fabs.f64 x)
x\_s = (copysign.f64 #s(literal 1 binary64) x)
(FPCore (x_s x_m y z)
 :precision binary64
 (*
  x_s
  (if (<= x_m 2e-77)
    (fma (* (- y 1.0) x_m) z x_m)
    (fma (- y 1.0) (* z x_m) x_m))))

C

x\_m = fabs(x);
x\_s = copysign(1.0, x);
double code(double x_s, double x_m, double y, double z) {
	double tmp;
	if (x_m <= 2e-77) {
		tmp = fma(((y - 1.0) * x_m), z, x_m);
	} else {
		tmp = fma((y - 1.0), (z * x_m), x_m);
	}
	return x_s * tmp;
}

Julia

x\_m = abs(x)
x\_s = copysign(1.0, x)
function code(x_s, x_m, y, z)
	tmp = 0.0
	if (x_m <= 2e-77)
		tmp = fma(Float64(Float64(y - 1.0) * x_m), z, x_m);
	else
		tmp = fma(Float64(y - 1.0), Float64(z * x_m), x_m);
	end
	return Float64(x_s * tmp)
end

Wolfram

x\_m = N[Abs[x], $MachinePrecision]
x\_s = N[With[{TMP1 = Abs[1.0], TMP2 = Sign[x]}, TMP1 * If[TMP2 == 0, 1, TMP2]], $MachinePrecision]
code[x$95$s_, x$95$m_, y_, z_] := N[(x$95$s * If[LessEqual[x$95$m, 2e-77], N[(N[(N[(y - 1.0), $MachinePrecision] * x$95$m), $MachinePrecision] * z + x$95$m), $MachinePrecision], N[(N[(y - 1.0), $MachinePrecision] * N[(z * x$95$m), $MachinePrecision] + x$95$m), $MachinePrecision]]), $MachinePrecision]

Derivation

1. Split input into 2 regimes

2. if x < 1.9999999999999999e-77

   1. Initial program 95.2%

      \[x \cdot \left(1 - \left(1 - y\right) \cdot z\right) \]
   2. Add Preprocessing

   3. Taylor expanded in x around 0

      \[\leadsto \color{blue}{x \cdot \left(1 - z \cdot \left(1 - y\right)\right)} \]

   5. Applied rewrites 94.1%

      \[\leadsto \color{blue}{\mathsf{fma}\left(\left(y - 1\right) \cdot x, z, x\right)} \]

   if 1.9999999999999999e-77 < x

   1. Initial program 99.9%

      \[x \cdot \left(1 - \left(1 - y\right) \cdot z\right) \]
   2. Add Preprocessing

   3. Taylor expanded in x around 0

      \[\leadsto \color{blue}{x \cdot \left(1 - z \cdot \left(1 - y\right)\right)} \]

   5. Applied rewrites 92.3%

      \[\leadsto \color{blue}{\mathsf{fma}\left(\left(y - 1\right) \cdot x, z, x\right)} \]
   6. Step-by-step derivation

      1. lift-fma.f64 N/A

         \[\leadsto \left(\left(y - 1\right) \cdot x\right) \cdot z + \color{blue}{x} \]

      2. lift-*.f64 N/A

         \[\leadsto \left(\left(y - 1\right) \cdot x\right) \cdot z + x \]

      3. associate-*l* N/A

         \[\leadsto \left(y - 1\right) \cdot \left(x \cdot z\right) + x \]

      4. lower-fma.f64 N/A

         \[\leadsto \mathsf{fma}\left(y - 1, \color{blue}{x \cdot z}, x\right) \]

      5. *-commutative N/A

         \[\leadsto \mathsf{fma}\left(y - 1, z \cdot \color{blue}{x}, x\right) \]

      6. lower-*.f64 99.9

         \[\leadsto \mathsf{fma}\left(y - 1, z \cdot \color{blue}{x}, x\right) \]

   7. Applied rewrites 99.9%

      \[\leadsto \mathsf{fma}\left(y - 1, \color{blue}{z \cdot x}, x\right) \]
3. Recombined 2 regimes into one program.
4. Add Preprocessing

Alternative 2: 99.8% accurate, 0.6× speedup

Math

\[\begin{array}{l} x\_m = \left|x\right| \\ x\_s = \mathsf{copysign}\left(1, x\right) \\ x\_s \cdot \begin{array}{l} \mathbf{if}\;z \leq -1 \cdot 10^{-17} \lor \neg \left(z \leq 1.2 \cdot 10^{-51}\right):\\ \;\;\;\;\mathsf{fma}\left(\left(y - 1\right) \cdot x\_m, z, x\_m\right)\\ \mathbf{else}:\\ \;\;\;\;x\_m \cdot \mathsf{fma}\left(y, z, 1\right)\\ \end{array} \end{array} \]

FPCore

x\_m = (fabs.f64 x)
x\_s = (copysign.f64 #s(literal 1 binary64) x)
(FPCore (x_s x_m y z)
 :precision binary64
 (*
  x_s
  (if (or (<= z -1e-17) (not (<= z 1.2e-51)))
    (fma (* (- y 1.0) x_m) z x_m)
    (* x_m (fma y z 1.0)))))

C

x\_m = fabs(x);
x\_s = copysign(1.0, x);
double code(double x_s, double x_m, double y, double z) {
	double tmp;
	if ((z <= -1e-17) || !(z <= 1.2e-51)) {
		tmp = fma(((y - 1.0) * x_m), z, x_m);
	} else {
		tmp = x_m * fma(y, z, 1.0);
	}
	return x_s * tmp;
}

Julia

x\_m = abs(x)
x\_s = copysign(1.0, x)
function code(x_s, x_m, y, z)
	tmp = 0.0
	if ((z <= -1e-17) || !(z <= 1.2e-51))
		tmp = fma(Float64(Float64(y - 1.0) * x_m), z, x_m);
	else
		tmp = Float64(x_m * fma(y, z, 1.0));
	end
	return Float64(x_s * tmp)
end

Wolfram

x\_m = N[Abs[x], $MachinePrecision]
x\_s = N[With[{TMP1 = Abs[1.0], TMP2 = Sign[x]}, TMP1 * If[TMP2 == 0, 1, TMP2]], $MachinePrecision]
code[x$95$s_, x$95$m_, y_, z_] := N[(x$95$s * If[Or[LessEqual[z, -1e-17], N[Not[LessEqual[z, 1.2e-51]], $MachinePrecision]], N[(N[(N[(y - 1.0), $MachinePrecision] * x$95$m), $MachinePrecision] * z + x$95$m), $MachinePrecision], N[(x$95$m * N[(y * z + 1.0), $MachinePrecision]), $MachinePrecision]]), $MachinePrecision]

Derivation

1. Split input into 2 regimes

2. if z < -1.00000000000000007e-17 or 1.2e-51 < z

   1. Initial program 93.9%

      \[x \cdot \left(1 - \left(1 - y\right) \cdot z\right) \]
   2. Add Preprocessing

   3. Taylor expanded in x around 0

      \[\leadsto \color{blue}{x \cdot \left(1 - z \cdot \left(1 - y\right)\right)} \]

   5. Applied rewrites 99.9%

      \[\leadsto \color{blue}{\mathsf{fma}\left(\left(y - 1\right) \cdot x, z, x\right)} \]

   if -1.00000000000000007e-17 < z < 1.2e-51

   1. Initial program 99.9%

      \[x \cdot \left(1 - \left(1 - y\right) \cdot z\right) \]
   2. Add Preprocessing

   3. Taylor expanded in y around 0

      \[\leadsto x \cdot \color{blue}{\left(\left(1 + y \cdot z\right) - z\right)} \]
   4. Step-by-step derivation

      1. associate--l+ N/A

         \[\leadsto x \cdot \left(1 + \color{blue}{\left(y \cdot z - z\right)}\right) \]

      2. +-commutative N/A

         \[\leadsto x \cdot \left(\left(y \cdot z - z\right) + \color{blue}{1}\right) \]

      3. *-lft-identity N/A

         \[\leadsto x \cdot \left(\left(y \cdot z - 1 \cdot z\right) + 1\right) \]

      4. metadata-eval N/A

         \[\leadsto x \cdot \left(\left(y \cdot z - \left(\mathsf{neg}\left(-1\right)\right) \cdot z\right) + 1\right) \]

      5. fp-cancel-sign-sub-inv N/A

         \[\leadsto x \cdot \left(\left(y \cdot z + -1 \cdot z\right) + 1\right) \]

      6. remove-double-neg N/A

         \[\leadsto x \cdot \left(\left(\left(\mathsf{neg}\left(\left(\mathsf{neg}\left(y \cdot z\right)\right)\right)\right) + -1 \cdot z\right) + 1\right) \]

      7. mul-1-neg N/A

         \[\leadsto x \cdot \left(\left(\left(\mathsf{neg}\left(-1 \cdot \left(y \cdot z\right)\right)\right) + -1 \cdot z\right) + 1\right) \]

      8. mul-1-neg N/A

         \[\leadsto x \cdot \left(\left(\left(\mathsf{neg}\left(-1 \cdot \left(y \cdot z\right)\right)\right) + \left(\mathsf{neg}\left(z\right)\right)\right) + 1\right) \]

      9. distribute-neg-in N/A

         \[\leadsto x \cdot \left(\left(\mathsf{neg}\left(\left(-1 \cdot \left(y \cdot z\right) + z\right)\right)\right) + 1\right) \]

      10. associate-*r* N/A

          \[\leadsto x \cdot \left(\left(\mathsf{neg}\left(\left(\left(-1 \cdot y\right) \cdot z + z\right)\right)\right) + 1\right) \]

      11. distribute-lft1-in N/A

          \[\leadsto x \cdot \left(\left(\mathsf{neg}\left(\left(-1 \cdot y + 1\right) \cdot z\right)\right) + 1\right) \]

      12. +-commutative N/A

          \[\leadsto x \cdot \left(\left(\mathsf{neg}\left(\left(1 + -1 \cdot y\right) \cdot z\right)\right) + 1\right) \]

      13. fp-cancel-sign-sub-inv N/A

          \[\leadsto x \cdot \left(\left(\mathsf{neg}\left(\left(1 - \left(\mathsf{neg}\left(-1\right)\right) \cdot y\right) \cdot z\right)\right) + 1\right) \]

      14. metadata-eval N/A

          \[\leadsto x \cdot \left(\left(\mathsf{neg}\left(\left(1 - 1 \cdot y\right) \cdot z\right)\right) + 1\right) \]

      15. *-lft-identity N/A

          \[\leadsto x \cdot \left(\left(\mathsf{neg}\left(\left(1 - y\right) \cdot z\right)\right) + 1\right) \]

      16. *-commutative N/A

          \[\leadsto x \cdot \left(\left(\mathsf{neg}\left(z \cdot \left(1 - y\right)\right)\right) + 1\right) \]

      17. distribute-rgt-neg-in N/A

          \[\leadsto x \cdot \left(z \cdot \left(\mathsf{neg}\left(\left(1 - y\right)\right)\right) + 1\right) \]

      18. mul-1-neg N/A

          \[\leadsto x \cdot \left(z \cdot \left(-1 \cdot \left(1 - y\right)\right) + 1\right) \]

      19. *-commutative N/A

          \[\leadsto x \cdot \left(\left(-1 \cdot \left(1 - y\right)\right) \cdot z + 1\right) \]

      20. lower-fma.f64 N/A

          \[\leadsto x \cdot \mathsf{fma}\left(-1 \cdot \left(1 - y\right), \color{blue}{z}, 1\right) \]

   5. Applied rewrites 99.9%

      \[\leadsto x \cdot \color{blue}{\mathsf{fma}\left(y - 1, z, 1\right)} \]

   6. Taylor expanded in y around inf

      \[\leadsto x \cdot \mathsf{fma}\left(y, z, 1\right) \]
   7. Step-by-step derivation

   8. Applied rewrites 99.9%

      \[\leadsto x \cdot \mathsf{fma}\left(y, z, 1\right) \]
3. Recombined 2 regimes into one program.

4. Final simplification 99.9%

   \[\leadsto \begin{array}{l} \mathbf{if}\;z \leq -1 \cdot 10^{-17} \lor \neg \left(z \leq 1.2 \cdot 10^{-51}\right):\\ \;\;\;\;\mathsf{fma}\left(\left(y - 1\right) \cdot x, z, x\right)\\ \mathbf{else}:\\ \;\;\;\;x \cdot \mathsf{fma}\left(y, z, 1\right)\\ \end{array} \]
5. Add Preprocessing

Alternative 3: 98.8% accurate, 0.7× speedup

Math

\[\begin{array}{l} x\_m = \left|x\right| \\ x\_s = \mathsf{copysign}\left(1, x\right) \\ x\_s \cdot \begin{array}{l} \mathbf{if}\;z \leq -0.146 \lor \neg \left(z \leq 0.000114\right):\\ \;\;\;\;\left(\left(y - 1\right) \cdot x\_m\right) \cdot z\\ \mathbf{else}:\\ \;\;\;\;x\_m \cdot \mathsf{fma}\left(y, z, 1\right)\\ \end{array} \end{array} \]

FPCore

x\_m = (fabs.f64 x)
x\_s = (copysign.f64 #s(literal 1 binary64) x)
(FPCore (x_s x_m y z)
 :precision binary64
 (*
  x_s
  (if (or (<= z -0.146) (not (<= z 0.000114)))
    (* (* (- y 1.0) x_m) z)
    (* x_m (fma y z 1.0)))))

C

x\_m = fabs(x);
x\_s = copysign(1.0, x);
double code(double x_s, double x_m, double y, double z) {
	double tmp;
	if ((z <= -0.146) || !(z <= 0.000114)) {
		tmp = ((y - 1.0) * x_m) * z;
	} else {
		tmp = x_m * fma(y, z, 1.0);
	}
	return x_s * tmp;
}

Julia

x\_m = abs(x)
x\_s = copysign(1.0, x)
function code(x_s, x_m, y, z)
	tmp = 0.0
	if ((z <= -0.146) || !(z <= 0.000114))
		tmp = Float64(Float64(Float64(y - 1.0) * x_m) * z);
	else
		tmp = Float64(x_m * fma(y, z, 1.0));
	end
	return Float64(x_s * tmp)
end

Wolfram

x\_m = N[Abs[x], $MachinePrecision]
x\_s = N[With[{TMP1 = Abs[1.0], TMP2 = Sign[x]}, TMP1 * If[TMP2 == 0, 1, TMP2]], $MachinePrecision]
code[x$95$s_, x$95$m_, y_, z_] := N[(x$95$s * If[Or[LessEqual[z, -0.146], N[Not[LessEqual[z, 0.000114]], $MachinePrecision]], N[(N[(N[(y - 1.0), $MachinePrecision] * x$95$m), $MachinePrecision] * z), $MachinePrecision], N[(x$95$m * N[(y * z + 1.0), $MachinePrecision]), $MachinePrecision]]), $MachinePrecision]

Derivation

1. Split input into 2 regimes

2. if z < -0.145999999999999991 or 1.1400000000000001e-4 < z

   1. Initial program 93.3%

      \[x \cdot \left(1 - \left(1 - y\right) \cdot z\right) \]
   2. Add Preprocessing

   3. Taylor expanded in z around inf

      \[\leadsto \color{blue}{x \cdot \left(z \cdot \left(y - 1\right)\right)} \]

   5. Applied rewrites 99.3%

      \[\leadsto \color{blue}{\left(\left(y - 1\right) \cdot x\right) \cdot z} \]

   if -0.145999999999999991 < z < 1.1400000000000001e-4

   1. Initial program 99.9%

      \[x \cdot \left(1 - \left(1 - y\right) \cdot z\right) \]
   2. Add Preprocessing

   3. Taylor expanded in y around 0

      \[\leadsto x \cdot \color{blue}{\left(\left(1 + y \cdot z\right) - z\right)} \]
   4. Step-by-step derivation

      1. associate--l+ N/A

         \[\leadsto x \cdot \left(1 + \color{blue}{\left(y \cdot z - z\right)}\right) \]

      2. +-commutative N/A

         \[\leadsto x \cdot \left(\left(y \cdot z - z\right) + \color{blue}{1}\right) \]

      3. *-lft-identity N/A

         \[\leadsto x \cdot \left(\left(y \cdot z - 1 \cdot z\right) + 1\right) \]

      4. metadata-eval N/A

         \[\leadsto x \cdot \left(\left(y \cdot z - \left(\mathsf{neg}\left(-1\right)\right) \cdot z\right) + 1\right) \]

      5. fp-cancel-sign-sub-inv N/A

         \[\leadsto x \cdot \left(\left(y \cdot z + -1 \cdot z\right) + 1\right) \]

      6. remove-double-neg N/A

         \[\leadsto x \cdot \left(\left(\left(\mathsf{neg}\left(\left(\mathsf{neg}\left(y \cdot z\right)\right)\right)\right) + -1 \cdot z\right) + 1\right) \]

      7. mul-1-neg N/A

         \[\leadsto x \cdot \left(\left(\left(\mathsf{neg}\left(-1 \cdot \left(y \cdot z\right)\right)\right) + -1 \cdot z\right) + 1\right) \]

      8. mul-1-neg N/A

         \[\leadsto x \cdot \left(\left(\left(\mathsf{neg}\left(-1 \cdot \left(y \cdot z\right)\right)\right) + \left(\mathsf{neg}\left(z\right)\right)\right) + 1\right) \]

      9. distribute-neg-in N/A

         \[\leadsto x \cdot \left(\left(\mathsf{neg}\left(\left(-1 \cdot \left(y \cdot z\right) + z\right)\right)\right) + 1\right) \]

      10. associate-*r* N/A

          \[\leadsto x \cdot \left(\left(\mathsf{neg}\left(\left(\left(-1 \cdot y\right) \cdot z + z\right)\right)\right) + 1\right) \]

      11. distribute-lft1-in N/A

          \[\leadsto x \cdot \left(\left(\mathsf{neg}\left(\left(-1 \cdot y + 1\right) \cdot z\right)\right) + 1\right) \]

      12. +-commutative N/A

          \[\leadsto x \cdot \left(\left(\mathsf{neg}\left(\left(1 + -1 \cdot y\right) \cdot z\right)\right) + 1\right) \]

      13. fp-cancel-sign-sub-inv N/A

          \[\leadsto x \cdot \left(\left(\mathsf{neg}\left(\left(1 - \left(\mathsf{neg}\left(-1\right)\right) \cdot y\right) \cdot z\right)\right) + 1\right) \]

      14. metadata-eval N/A

          \[\leadsto x \cdot \left(\left(\mathsf{neg}\left(\left(1 - 1 \cdot y\right) \cdot z\right)\right) + 1\right) \]

      15. *-lft-identity N/A

          \[\leadsto x \cdot \left(\left(\mathsf{neg}\left(\left(1 - y\right) \cdot z\right)\right) + 1\right) \]

      16. *-commutative N/A

          \[\leadsto x \cdot \left(\left(\mathsf{neg}\left(z \cdot \left(1 - y\right)\right)\right) + 1\right) \]

      17. distribute-rgt-neg-in N/A

          \[\leadsto x \cdot \left(z \cdot \left(\mathsf{neg}\left(\left(1 - y\right)\right)\right) + 1\right) \]

      18. mul-1-neg N/A

          \[\leadsto x \cdot \left(z \cdot \left(-1 \cdot \left(1 - y\right)\right) + 1\right) \]

      19. *-commutative N/A

          \[\leadsto x \cdot \left(\left(-1 \cdot \left(1 - y\right)\right) \cdot z + 1\right) \]

      20. lower-fma.f64 N/A

          \[\leadsto x \cdot \mathsf{fma}\left(-1 \cdot \left(1 - y\right), \color{blue}{z}, 1\right) \]

   5. Applied rewrites 99.9%

      \[\leadsto x \cdot \color{blue}{\mathsf{fma}\left(y - 1, z, 1\right)} \]

   6. Taylor expanded in y around inf

      \[\leadsto x \cdot \mathsf{fma}\left(y, z, 1\right) \]
   7. Step-by-step derivation

   8. Applied rewrites 98.6%

      \[\leadsto x \cdot \mathsf{fma}\left(y, z, 1\right) \]
3. Recombined 2 regimes into one program.

4. Final simplification 99.0%

   \[\leadsto \begin{array}{l} \mathbf{if}\;z \leq -0.146 \lor \neg \left(z \leq 0.000114\right):\\ \;\;\;\;\left(\left(y - 1\right) \cdot x\right) \cdot z\\ \mathbf{else}:\\ \;\;\;\;x \cdot \mathsf{fma}\left(y, z, 1\right)\\ \end{array} \]
5. Add Preprocessing

Alternative 4: 97.3% accurate, 0.7× speedup

Math

\[\begin{array}{l} x\_m = \left|x\right| \\ x\_s = \mathsf{copysign}\left(1, x\right) \\ x\_s \cdot \begin{array}{l} \mathbf{if}\;y \leq -1 \lor \neg \left(y \leq 1\right):\\ \;\;\;\;\mathsf{fma}\left(z \cdot x\_m, y, x\_m\right)\\ \mathbf{else}:\\ \;\;\;\;x\_m \cdot \left(1 - z\right)\\ \end{array} \end{array} \]

FPCore

x\_m = (fabs.f64 x)
x\_s = (copysign.f64 #s(literal 1 binary64) x)
(FPCore (x_s x_m y z)
 :precision binary64
 (*
  x_s
  (if (or (<= y -1.0) (not (<= y 1.0)))
    (fma (* z x_m) y x_m)
    (* x_m (- 1.0 z)))))

C

x\_m = fabs(x);
x\_s = copysign(1.0, x);
double code(double x_s, double x_m, double y, double z) {
	double tmp;
	if ((y <= -1.0) || !(y <= 1.0)) {
		tmp = fma((z * x_m), y, x_m);
	} else {
		tmp = x_m * (1.0 - z);
	}
	return x_s * tmp;
}

Julia

x\_m = abs(x)
x\_s = copysign(1.0, x)
function code(x_s, x_m, y, z)
	tmp = 0.0
	if ((y <= -1.0) || !(y <= 1.0))
		tmp = fma(Float64(z * x_m), y, x_m);
	else
		tmp = Float64(x_m * Float64(1.0 - z));
	end
	return Float64(x_s * tmp)
end

Wolfram

x\_m = N[Abs[x], $MachinePrecision]
x\_s = N[With[{TMP1 = Abs[1.0], TMP2 = Sign[x]}, TMP1 * If[TMP2 == 0, 1, TMP2]], $MachinePrecision]
code[x$95$s_, x$95$m_, y_, z_] := N[(x$95$s * If[Or[LessEqual[y, -1.0], N[Not[LessEqual[y, 1.0]], $MachinePrecision]], N[(N[(z * x$95$m), $MachinePrecision] * y + x$95$m), $MachinePrecision], N[(x$95$m * N[(1.0 - z), $MachinePrecision]), $MachinePrecision]]), $MachinePrecision]

Derivation

1. Split input into 2 regimes

2. if y < -1 or 1 < y

   1. Initial program 92.7%

      \[x \cdot \left(1 - \left(1 - y\right) \cdot z\right) \]
   2. Add Preprocessing

   3. Taylor expanded in x around 0

      \[\leadsto \color{blue}{x \cdot \left(1 - z \cdot \left(1 - y\right)\right)} \]

   5. Applied rewrites 86.2%

      \[\leadsto \color{blue}{\mathsf{fma}\left(\left(y - 1\right) \cdot x, z, x\right)} \]

   6. Taylor expanded in y around inf

      \[\leadsto \mathsf{fma}\left(y \cdot x, z, x\right) \]
   7. Step-by-step derivation

   8. Applied rewrites 85.0%

      \[\leadsto \mathsf{fma}\left(y \cdot x, z, x\right) \]
   9. Step-by-step derivation

      1. lift-fma.f64 N/A

         \[\leadsto \left(y \cdot x\right) \cdot z + \color{blue}{x} \]

      2. *-commutative N/A

         \[\leadsto z \cdot \left(y \cdot x\right) + x \]

      3. lift-*.f64 N/A

         \[\leadsto z \cdot \left(y \cdot x\right) + x \]

      4. *-commutative N/A

         \[\leadsto z \cdot \left(x \cdot y\right) + x \]

      5. associate-*r* N/A

         \[\leadsto \left(z \cdot x\right) \cdot y + x \]

      6. lift-*.f64 N/A

         \[\leadsto \left(z \cdot x\right) \cdot y + x \]

      7. lower-fma.f64 93.7

         \[\leadsto \mathsf{fma}\left(z \cdot x, \color{blue}{y}, x\right) \]

   10. Applied rewrites 93.7%

       \[\leadsto \mathsf{fma}\left(z \cdot x, \color{blue}{y}, x\right) \]

   if -1 < y < 1

   1. Initial program 100.0%

      \[x \cdot \left(1 - \left(1 - y\right) \cdot z\right) \]
   2. Add Preprocessing

   3. Taylor expanded in y around 0

      \[\leadsto x \cdot \left(1 - \color{blue}{z}\right) \]
   4. Step-by-step derivation

   5. Applied rewrites 99.4%

      \[\leadsto x \cdot \left(1 - \color{blue}{z}\right) \]
3. Recombined 2 regimes into one program.

4. Final simplification 96.8%

   \[\leadsto \begin{array}{l} \mathbf{if}\;y \leq -1 \lor \neg \left(y \leq 1\right):\\ \;\;\;\;\mathsf{fma}\left(z \cdot x, y, x\right)\\ \mathbf{else}:\\ \;\;\;\;x \cdot \left(1 - z\right)\\ \end{array} \]
5. Add Preprocessing

Alternative 5: 96.0% accurate, 0.7× speedup

Math

\[\begin{array}{l} x\_m = \left|x\right| \\ x\_s = \mathsf{copysign}\left(1, x\right) \\ x\_s \cdot \begin{array}{l} \mathbf{if}\;y \leq -1:\\ \;\;\;\;x\_m \cdot \mathsf{fma}\left(y, z, 1\right)\\ \mathbf{elif}\;y \leq 1:\\ \;\;\;\;x\_m \cdot \left(1 - z\right)\\ \mathbf{else}:\\ \;\;\;\;\mathsf{fma}\left(z \cdot x\_m, y, x\_m\right)\\ \end{array} \end{array} \]

FPCore

x\_m = (fabs.f64 x)
x\_s = (copysign.f64 #s(literal 1 binary64) x)
(FPCore (x_s x_m y z)
 :precision binary64
 (*
  x_s
  (if (<= y -1.0)
    (* x_m (fma y z 1.0))
    (if (<= y 1.0) (* x_m (- 1.0 z)) (fma (* z x_m) y x_m)))))

C

x\_m = fabs(x);
x\_s = copysign(1.0, x);
double code(double x_s, double x_m, double y, double z) {
	double tmp;
	if (y <= -1.0) {
		tmp = x_m * fma(y, z, 1.0);
	} else if (y <= 1.0) {
		tmp = x_m * (1.0 - z);
	} else {
		tmp = fma((z * x_m), y, x_m);
	}
	return x_s * tmp;
}

Julia

x\_m = abs(x)
x\_s = copysign(1.0, x)
function code(x_s, x_m, y, z)
	tmp = 0.0
	if (y <= -1.0)
		tmp = Float64(x_m * fma(y, z, 1.0));
	elseif (y <= 1.0)
		tmp = Float64(x_m * Float64(1.0 - z));
	else
		tmp = fma(Float64(z * x_m), y, x_m);
	end
	return Float64(x_s * tmp)
end

Wolfram

x\_m = N[Abs[x], $MachinePrecision]
x\_s = N[With[{TMP1 = Abs[1.0], TMP2 = Sign[x]}, TMP1 * If[TMP2 == 0, 1, TMP2]], $MachinePrecision]
code[x$95$s_, x$95$m_, y_, z_] := N[(x$95$s * If[LessEqual[y, -1.0], N[(x$95$m * N[(y * z + 1.0), $MachinePrecision]), $MachinePrecision], If[LessEqual[y, 1.0], N[(x$95$m * N[(1.0 - z), $MachinePrecision]), $MachinePrecision], N[(N[(z * x$95$m), $MachinePrecision] * y + x$95$m), $MachinePrecision]]]), $MachinePrecision]

Derivation

1. Split input into 3 regimes

2. if y < -1

   1. Initial program 93.6%

      \[x \cdot \left(1 - \left(1 - y\right) \cdot z\right) \]
   2. Add Preprocessing

   3. Taylor expanded in y around 0

      \[\leadsto x \cdot \color{blue}{\left(\left(1 + y \cdot z\right) - z\right)} \]
   4. Step-by-step derivation

      1. associate--l+ N/A

         \[\leadsto x \cdot \left(1 + \color{blue}{\left(y \cdot z - z\right)}\right) \]

      2. +-commutative N/A

         \[\leadsto x \cdot \left(\left(y \cdot z - z\right) + \color{blue}{1}\right) \]

      3. *-lft-identity N/A

         \[\leadsto x \cdot \left(\left(y \cdot z - 1 \cdot z\right) + 1\right) \]

      4. metadata-eval N/A

         \[\leadsto x \cdot \left(\left(y \cdot z - \left(\mathsf{neg}\left(-1\right)\right) \cdot z\right) + 1\right) \]

      5. fp-cancel-sign-sub-inv N/A

         \[\leadsto x \cdot \left(\left(y \cdot z + -1 \cdot z\right) + 1\right) \]

      6. remove-double-neg N/A

         \[\leadsto x \cdot \left(\left(\left(\mathsf{neg}\left(\left(\mathsf{neg}\left(y \cdot z\right)\right)\right)\right) + -1 \cdot z\right) + 1\right) \]

      7. mul-1-neg N/A

         \[\leadsto x \cdot \left(\left(\left(\mathsf{neg}\left(-1 \cdot \left(y \cdot z\right)\right)\right) + -1 \cdot z\right) + 1\right) \]

      8. mul-1-neg N/A

         \[\leadsto x \cdot \left(\left(\left(\mathsf{neg}\left(-1 \cdot \left(y \cdot z\right)\right)\right) + \left(\mathsf{neg}\left(z\right)\right)\right) + 1\right) \]

      9. distribute-neg-in N/A

         \[\leadsto x \cdot \left(\left(\mathsf{neg}\left(\left(-1 \cdot \left(y \cdot z\right) + z\right)\right)\right) + 1\right) \]

      10. associate-*r* N/A

          \[\leadsto x \cdot \left(\left(\mathsf{neg}\left(\left(\left(-1 \cdot y\right) \cdot z + z\right)\right)\right) + 1\right) \]

      11. distribute-lft1-in N/A

          \[\leadsto x \cdot \left(\left(\mathsf{neg}\left(\left(-1 \cdot y + 1\right) \cdot z\right)\right) + 1\right) \]

      12. +-commutative N/A

          \[\leadsto x \cdot \left(\left(\mathsf{neg}\left(\left(1 + -1 \cdot y\right) \cdot z\right)\right) + 1\right) \]

      13. fp-cancel-sign-sub-inv N/A

          \[\leadsto x \cdot \left(\left(\mathsf{neg}\left(\left(1 - \left(\mathsf{neg}\left(-1\right)\right) \cdot y\right) \cdot z\right)\right) + 1\right) \]

      14. metadata-eval N/A

          \[\leadsto x \cdot \left(\left(\mathsf{neg}\left(\left(1 - 1 \cdot y\right) \cdot z\right)\right) + 1\right) \]

      15. *-lft-identity N/A

          \[\leadsto x \cdot \left(\left(\mathsf{neg}\left(\left(1 - y\right) \cdot z\right)\right) + 1\right) \]

      16. *-commutative N/A

          \[\leadsto x \cdot \left(\left(\mathsf{neg}\left(z \cdot \left(1 - y\right)\right)\right) + 1\right) \]

      17. distribute-rgt-neg-in N/A

          \[\leadsto x \cdot \left(z \cdot \left(\mathsf{neg}\left(\left(1 - y\right)\right)\right) + 1\right) \]

      18. mul-1-neg N/A

          \[\leadsto x \cdot \left(z \cdot \left(-1 \cdot \left(1 - y\right)\right) + 1\right) \]

      19. *-commutative N/A

          \[\leadsto x \cdot \left(\left(-1 \cdot \left(1 - y\right)\right) \cdot z + 1\right) \]

      20. lower-fma.f64 N/A

          \[\leadsto x \cdot \mathsf{fma}\left(-1 \cdot \left(1 - y\right), \color{blue}{z}, 1\right) \]

   5. Applied rewrites 93.6%

      \[\leadsto x \cdot \color{blue}{\mathsf{fma}\left(y - 1, z, 1\right)} \]

   6. Taylor expanded in y around inf

      \[\leadsto x \cdot \mathsf{fma}\left(y, z, 1\right) \]
   7. Step-by-step derivation

   8. Applied rewrites 92.4%

      \[\leadsto x \cdot \mathsf{fma}\left(y, z, 1\right) \]

   if -1 < y < 1

   1. Initial program 100.0%

      \[x \cdot \left(1 - \left(1 - y\right) \cdot z\right) \]
   2. Add Preprocessing

   3. Taylor expanded in y around 0

      \[\leadsto x \cdot \left(1 - \color{blue}{z}\right) \]
   4. Step-by-step derivation

   5. Applied rewrites 99.4%

      \[\leadsto x \cdot \left(1 - \color{blue}{z}\right) \]

   if 1 < y

   1. Initial program 91.7%

      \[x \cdot \left(1 - \left(1 - y\right) \cdot z\right) \]
   2. Add Preprocessing

   3. Taylor expanded in x around 0

      \[\leadsto \color{blue}{x \cdot \left(1 - z \cdot \left(1 - y\right)\right)} \]

   5. Applied rewrites 83.7%

      \[\leadsto \color{blue}{\mathsf{fma}\left(\left(y - 1\right) \cdot x, z, x\right)} \]

   6. Taylor expanded in y around inf

      \[\leadsto \mathsf{fma}\left(y \cdot x, z, x\right) \]
   7. Step-by-step derivation

   8. Applied rewrites 82.5%

      \[\leadsto \mathsf{fma}\left(y \cdot x, z, x\right) \]
   9. Step-by-step derivation

      1. lift-fma.f64 N/A

         \[\leadsto \left(y \cdot x\right) \cdot z + \color{blue}{x} \]

      2. *-commutative N/A

         \[\leadsto z \cdot \left(y \cdot x\right) + x \]

      3. lift-*.f64 N/A

         \[\leadsto z \cdot \left(y \cdot x\right) + x \]

      4. *-commutative N/A

         \[\leadsto z \cdot \left(x \cdot y\right) + x \]

      5. associate-*r* N/A

         \[\leadsto \left(z \cdot x\right) \cdot y + x \]

      6. lift-*.f64 N/A

         \[\leadsto \left(z \cdot x\right) \cdot y + x \]

      7. lower-fma.f64 96.5

         \[\leadsto \mathsf{fma}\left(z \cdot x, \color{blue}{y}, x\right) \]

   10. Applied rewrites 96.5%

       \[\leadsto \mathsf{fma}\left(z \cdot x, \color{blue}{y}, x\right) \]
3. Recombined 3 regimes into one program.
4. Add Preprocessing

Alternative 6: 82.9% accurate, 0.7× speedup

Math

\[\begin{array}{l} x\_m = \left|x\right| \\ x\_s = \mathsf{copysign}\left(1, x\right) \\ x\_s \cdot \begin{array}{l} \mathbf{if}\;y \leq -3.5 \cdot 10^{+48} \lor \neg \left(y \leq 5.8 \cdot 10^{+147}\right):\\ \;\;\;\;\left(y \cdot x\_m\right) \cdot z\\ \mathbf{else}:\\ \;\;\;\;x\_m \cdot \left(1 - z\right)\\ \end{array} \end{array} \]

FPCore

x\_m = (fabs.f64 x)
x\_s = (copysign.f64 #s(literal 1 binary64) x)
(FPCore (x_s x_m y z)
 :precision binary64
 (*
  x_s
  (if (or (<= y -3.5e+48) (not (<= y 5.8e+147)))
    (* (* y x_m) z)
    (* x_m (- 1.0 z)))))

C

x\_m = fabs(x);
x\_s = copysign(1.0, x);
double code(double x_s, double x_m, double y, double z) {
	double tmp;
	if ((y <= -3.5e+48) || !(y <= 5.8e+147)) {
		tmp = (y * x_m) * z;
	} else {
		tmp = x_m * (1.0 - z);
	}
	return x_s * tmp;
}

Fortran

x\_m =     private
x\_s =     private
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(x_s, x_m, y, z)
use fmin_fmax_functions
    real(8), intent (in) :: x_s
    real(8), intent (in) :: x_m
    real(8), intent (in) :: y
    real(8), intent (in) :: z
    real(8) :: tmp
    if ((y <= (-3.5d+48)) .or. (.not. (y <= 5.8d+147))) then
        tmp = (y * x_m) * z
    else
        tmp = x_m * (1.0d0 - z)
    end if
    code = x_s * tmp
end function

Java

x\_m = Math.abs(x);
x\_s = Math.copySign(1.0, x);
public static double code(double x_s, double x_m, double y, double z) {
	double tmp;
	if ((y <= -3.5e+48) || !(y <= 5.8e+147)) {
		tmp = (y * x_m) * z;
	} else {
		tmp = x_m * (1.0 - z);
	}
	return x_s * tmp;
}

Python

x\_m = math.fabs(x)
x\_s = math.copysign(1.0, x)
def code(x_s, x_m, y, z):
	tmp = 0
	if (y <= -3.5e+48) or not (y <= 5.8e+147):
		tmp = (y * x_m) * z
	else:
		tmp = x_m * (1.0 - z)
	return x_s * tmp

Julia

x\_m = abs(x)
x\_s = copysign(1.0, x)
function code(x_s, x_m, y, z)
	tmp = 0.0
	if ((y <= -3.5e+48) || !(y <= 5.8e+147))
		tmp = Float64(Float64(y * x_m) * z);
	else
		tmp = Float64(x_m * Float64(1.0 - z));
	end
	return Float64(x_s * tmp)
end

MATLAB

x\_m = abs(x);
x\_s = sign(x) * abs(1.0);
function tmp_2 = code(x_s, x_m, y, z)
	tmp = 0.0;
	if ((y <= -3.5e+48) || ~((y <= 5.8e+147)))
		tmp = (y * x_m) * z;
	else
		tmp = x_m * (1.0 - z);
	end
	tmp_2 = x_s * tmp;
end

Wolfram

x\_m = N[Abs[x], $MachinePrecision]
x\_s = N[With[{TMP1 = Abs[1.0], TMP2 = Sign[x]}, TMP1 * If[TMP2 == 0, 1, TMP2]], $MachinePrecision]
code[x$95$s_, x$95$m_, y_, z_] := N[(x$95$s * If[Or[LessEqual[y, -3.5e+48], N[Not[LessEqual[y, 5.8e+147]], $MachinePrecision]], N[(N[(y * x$95$m), $MachinePrecision] * z), $MachinePrecision], N[(x$95$m * N[(1.0 - z), $MachinePrecision]), $MachinePrecision]]), $MachinePrecision]

Derivation

1. Split input into 2 regimes

2. if y < -3.4999999999999997e48 or 5.7999999999999997e147 < y

   1. Initial program 90.0%

      \[x \cdot \left(1 - \left(1 - y\right) \cdot z\right) \]
   2. Add Preprocessing
   3. Step-by-step derivation

      1. lift-*.f64 N/A

         \[\leadsto \color{blue}{x \cdot \left(1 - \left(1 - y\right) \cdot z\right)} \]

      2. *-commutative N/A

         \[\leadsto \color{blue}{\left(1 - \left(1 - y\right) \cdot z\right) \cdot x} \]

      3. lift--.f64 N/A

         \[\leadsto \color{blue}{\left(1 - \left(1 - y\right) \cdot z\right)} \cdot x \]

      4. flip-- N/A

         \[\leadsto \color{blue}{\frac{1 \cdot 1 - \left(\left(1 - y\right) \cdot z\right) \cdot \left(\left(1 - y\right) \cdot z\right)}{1 + \left(1 - y\right) \cdot z}} \cdot x \]

      5. associate-*l/ N/A

         \[\leadsto \color{blue}{\frac{\left(1 \cdot 1 - \left(\left(1 - y\right) \cdot z\right) \cdot \left(\left(1 - y\right) \cdot z\right)\right) \cdot x}{1 + \left(1 - y\right) \cdot z}} \]

      6. lower-/.f64 N/A

         \[\leadsto \color{blue}{\frac{\left(1 \cdot 1 - \left(\left(1 - y\right) \cdot z\right) \cdot \left(\left(1 - y\right) \cdot z\right)\right) \cdot x}{1 + \left(1 - y\right) \cdot z}} \]

   4. Applied rewrites 49.0%

      \[\leadsto \color{blue}{\frac{\left(1 - {\left(z \cdot \left(1 - y\right)\right)}^{2}\right) \cdot x}{\mathsf{fma}\left(z, 1 - y, 1\right)}} \]

   5. Taylor expanded in y around inf

      \[\leadsto \color{blue}{x \cdot \left(y \cdot z\right)} \]
   6. Step-by-step derivation

      1. associate-*r* N/A

         \[\leadsto \left(x \cdot y\right) \cdot \color{blue}{z} \]

      2. lower-*.f64 N/A

         \[\leadsto \left(x \cdot y\right) \cdot \color{blue}{z} \]

      3. *-commutative N/A

         \[\leadsto \left(y \cdot x\right) \cdot z \]

      4. lower-*.f64 71.2

         \[\leadsto \left(y \cdot x\right) \cdot z \]

   7. Applied rewrites 71.2%

      \[\leadsto \color{blue}{\left(y \cdot x\right) \cdot z} \]

   if -3.4999999999999997e48 < y < 5.7999999999999997e147

   1. Initial program 100.0%

      \[x \cdot \left(1 - \left(1 - y\right) \cdot z\right) \]
   2. Add Preprocessing

   3. Taylor expanded in y around 0

      \[\leadsto x \cdot \left(1 - \color{blue}{z}\right) \]
   4. Step-by-step derivation

   5. Applied rewrites 91.1%

      \[\leadsto x \cdot \left(1 - \color{blue}{z}\right) \]
3. Recombined 2 regimes into one program.

4. Final simplification 84.3%

   \[\leadsto \begin{array}{l} \mathbf{if}\;y \leq -3.5 \cdot 10^{+48} \lor \neg \left(y \leq 5.8 \cdot 10^{+147}\right):\\ \;\;\;\;\left(y \cdot x\right) \cdot z\\ \mathbf{else}:\\ \;\;\;\;x \cdot \left(1 - z\right)\\ \end{array} \]
5. Add Preprocessing

Alternative 7: 83.0% accurate, 0.7× speedup

Math

\[\begin{array}{l} x\_m = \left|x\right| \\ x\_s = \mathsf{copysign}\left(1, x\right) \\ x\_s \cdot \begin{array}{l} \mathbf{if}\;y \leq -3.5 \cdot 10^{+48}:\\ \;\;\;\;\left(y \cdot x\_m\right) \cdot z\\ \mathbf{elif}\;y \leq 5.8 \cdot 10^{+147}:\\ \;\;\;\;x\_m \cdot \left(1 - z\right)\\ \mathbf{else}:\\ \;\;\;\;\left(z \cdot x\_m\right) \cdot y\\ \end{array} \end{array} \]

FPCore

x\_m = (fabs.f64 x)
x\_s = (copysign.f64 #s(literal 1 binary64) x)
(FPCore (x_s x_m y z)
 :precision binary64
 (*
  x_s
  (if (<= y -3.5e+48)
    (* (* y x_m) z)
    (if (<= y 5.8e+147) (* x_m (- 1.0 z)) (* (* z x_m) y)))))

C

x\_m = fabs(x);
x\_s = copysign(1.0, x);
double code(double x_s, double x_m, double y, double z) {
	double tmp;
	if (y <= -3.5e+48) {
		tmp = (y * x_m) * z;
	} else if (y <= 5.8e+147) {
		tmp = x_m * (1.0 - z);
	} else {
		tmp = (z * x_m) * y;
	}
	return x_s * tmp;
}

Fortran

x\_m =     private
x\_s =     private
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(x_s, x_m, y, z)
use fmin_fmax_functions
    real(8), intent (in) :: x_s
    real(8), intent (in) :: x_m
    real(8), intent (in) :: y
    real(8), intent (in) :: z
    real(8) :: tmp
    if (y <= (-3.5d+48)) then
        tmp = (y * x_m) * z
    else if (y <= 5.8d+147) then
        tmp = x_m * (1.0d0 - z)
    else
        tmp = (z * x_m) * y
    end if
    code = x_s * tmp
end function

Java

x\_m = Math.abs(x);
x\_s = Math.copySign(1.0, x);
public static double code(double x_s, double x_m, double y, double z) {
	double tmp;
	if (y <= -3.5e+48) {
		tmp = (y * x_m) * z;
	} else if (y <= 5.8e+147) {
		tmp = x_m * (1.0 - z);
	} else {
		tmp = (z * x_m) * y;
	}
	return x_s * tmp;
}

Python

x\_m = math.fabs(x)
x\_s = math.copysign(1.0, x)
def code(x_s, x_m, y, z):
	tmp = 0
	if y <= -3.5e+48:
		tmp = (y * x_m) * z
	elif y <= 5.8e+147:
		tmp = x_m * (1.0 - z)
	else:
		tmp = (z * x_m) * y
	return x_s * tmp

Julia

x\_m = abs(x)
x\_s = copysign(1.0, x)
function code(x_s, x_m, y, z)
	tmp = 0.0
	if (y <= -3.5e+48)
		tmp = Float64(Float64(y * x_m) * z);
	elseif (y <= 5.8e+147)
		tmp = Float64(x_m * Float64(1.0 - z));
	else
		tmp = Float64(Float64(z * x_m) * y);
	end
	return Float64(x_s * tmp)
end

MATLAB

x\_m = abs(x);
x\_s = sign(x) * abs(1.0);
function tmp_2 = code(x_s, x_m, y, z)
	tmp = 0.0;
	if (y <= -3.5e+48)
		tmp = (y * x_m) * z;
	elseif (y <= 5.8e+147)
		tmp = x_m * (1.0 - z);
	else
		tmp = (z * x_m) * y;
	end
	tmp_2 = x_s * tmp;
end

Wolfram

x\_m = N[Abs[x], $MachinePrecision]
x\_s = N[With[{TMP1 = Abs[1.0], TMP2 = Sign[x]}, TMP1 * If[TMP2 == 0, 1, TMP2]], $MachinePrecision]
code[x$95$s_, x$95$m_, y_, z_] := N[(x$95$s * If[LessEqual[y, -3.5e+48], N[(N[(y * x$95$m), $MachinePrecision] * z), $MachinePrecision], If[LessEqual[y, 5.8e+147], N[(x$95$m * N[(1.0 - z), $MachinePrecision]), $MachinePrecision], N[(N[(z * x$95$m), $MachinePrecision] * y), $MachinePrecision]]]), $MachinePrecision]

Derivation

1. Split input into 3 regimes

2. if y < -3.4999999999999997e48

   1. Initial program 93.0%

      \[x \cdot \left(1 - \left(1 - y\right) \cdot z\right) \]
   2. Add Preprocessing
   3. Step-by-step derivation

      1. lift-*.f64 N/A

         \[\leadsto \color{blue}{x \cdot \left(1 - \left(1 - y\right) \cdot z\right)} \]

      2. *-commutative N/A

         \[\leadsto \color{blue}{\left(1 - \left(1 - y\right) \cdot z\right) \cdot x} \]

      3. lift--.f64 N/A

         \[\leadsto \color{blue}{\left(1 - \left(1 - y\right) \cdot z\right)} \cdot x \]

      4. flip-- N/A

         \[\leadsto \color{blue}{\frac{1 \cdot 1 - \left(\left(1 - y\right) \cdot z\right) \cdot \left(\left(1 - y\right) \cdot z\right)}{1 + \left(1 - y\right) \cdot z}} \cdot x \]

      5. associate-*l/ N/A

         \[\leadsto \color{blue}{\frac{\left(1 \cdot 1 - \left(\left(1 - y\right) \cdot z\right) \cdot \left(\left(1 - y\right) \cdot z\right)\right) \cdot x}{1 + \left(1 - y\right) \cdot z}} \]

      6. lower-/.f64 N/A

         \[\leadsto \color{blue}{\frac{\left(1 \cdot 1 - \left(\left(1 - y\right) \cdot z\right) \cdot \left(\left(1 - y\right) \cdot z\right)\right) \cdot x}{1 + \left(1 - y\right) \cdot z}} \]

   4. Applied rewrites 52.5%

      \[\leadsto \color{blue}{\frac{\left(1 - {\left(z \cdot \left(1 - y\right)\right)}^{2}\right) \cdot x}{\mathsf{fma}\left(z, 1 - y, 1\right)}} \]

   5. Taylor expanded in y around inf

      \[\leadsto \color{blue}{x \cdot \left(y \cdot z\right)} \]
   6. Step-by-step derivation

      1. associate-*r* N/A

         \[\leadsto \left(x \cdot y\right) \cdot \color{blue}{z} \]

      2. lower-*.f64 N/A

         \[\leadsto \left(x \cdot y\right) \cdot \color{blue}{z} \]

      3. *-commutative N/A

         \[\leadsto \left(y \cdot x\right) \cdot z \]

      4. lower-*.f64 71.4

         \[\leadsto \left(y \cdot x\right) \cdot z \]

   7. Applied rewrites 71.4%

      \[\leadsto \color{blue}{\left(y \cdot x\right) \cdot z} \]

   if -3.4999999999999997e48 < y < 5.7999999999999997e147

   1. Initial program 100.0%

      \[x \cdot \left(1 - \left(1 - y\right) \cdot z\right) \]
   2. Add Preprocessing

   3. Taylor expanded in y around 0

      \[\leadsto x \cdot \left(1 - \color{blue}{z}\right) \]
   4. Step-by-step derivation

   5. Applied rewrites 91.1%

      \[\leadsto x \cdot \left(1 - \color{blue}{z}\right) \]

   if 5.7999999999999997e147 < y

   1. Initial program 85.2%

      \[x \cdot \left(1 - \left(1 - y\right) \cdot z\right) \]
   2. Add Preprocessing
   3. Step-by-step derivation

      1. lift-*.f64 N/A

         \[\leadsto \color{blue}{x \cdot \left(1 - \left(1 - y\right) \cdot z\right)} \]

      2. *-commutative N/A

         \[\leadsto \color{blue}{\left(1 - \left(1 - y\right) \cdot z\right) \cdot x} \]

      3. lift--.f64 N/A

         \[\leadsto \color{blue}{\left(1 - \left(1 - y\right) \cdot z\right)} \cdot x \]

      4. flip-- N/A

         \[\leadsto \color{blue}{\frac{1 \cdot 1 - \left(\left(1 - y\right) \cdot z\right) \cdot \left(\left(1 - y\right) \cdot z\right)}{1 + \left(1 - y\right) \cdot z}} \cdot x \]

      5. associate-*l/ N/A

         \[\leadsto \color{blue}{\frac{\left(1 \cdot 1 - \left(\left(1 - y\right) \cdot z\right) \cdot \left(\left(1 - y\right) \cdot z\right)\right) \cdot x}{1 + \left(1 - y\right) \cdot z}} \]

      6. lower-/.f64 N/A

         \[\leadsto \color{blue}{\frac{\left(1 \cdot 1 - \left(\left(1 - y\right) \cdot z\right) \cdot \left(\left(1 - y\right) \cdot z\right)\right) \cdot x}{1 + \left(1 - y\right) \cdot z}} \]

   4. Applied rewrites 43.2%

      \[\leadsto \color{blue}{\frac{\left(1 - {\left(z \cdot \left(1 - y\right)\right)}^{2}\right) \cdot x}{\mathsf{fma}\left(z, 1 - y, 1\right)}} \]

   5. Taylor expanded in y around inf

      \[\leadsto \color{blue}{x \cdot \left(y \cdot z\right)} \]
   6. Step-by-step derivation

      1. associate-*r* N/A

         \[\leadsto \left(x \cdot y\right) \cdot \color{blue}{z} \]

      2. lower-*.f64 N/A

         \[\leadsto \left(x \cdot y\right) \cdot \color{blue}{z} \]

      3. *-commutative N/A

         \[\leadsto \left(y \cdot x\right) \cdot z \]

      4. lower-*.f64 70.8

         \[\leadsto \left(y \cdot x\right) \cdot z \]

   7. Applied rewrites 70.8%

      \[\leadsto \color{blue}{\left(y \cdot x\right) \cdot z} \]
   8. Step-by-step derivation

      1. lift-*.f64 N/A

         \[\leadsto \left(y \cdot x\right) \cdot \color{blue}{z} \]

      2. *-commutative N/A

         \[\leadsto z \cdot \color{blue}{\left(y \cdot x\right)} \]

      3. lift-*.f64 N/A

         \[\leadsto z \cdot \left(y \cdot \color{blue}{x}\right) \]

      4. *-commutative N/A

         \[\leadsto z \cdot \left(x \cdot \color{blue}{y}\right) \]

      5. associate-*r* N/A

         \[\leadsto \left(z \cdot x\right) \cdot \color{blue}{y} \]

      6. lift-*.f64 N/A

         \[\leadsto \left(z \cdot x\right) \cdot y \]

      7. lower-*.f64 85.6

         \[\leadsto \left(z \cdot x\right) \cdot \color{blue}{y} \]

   9. Applied rewrites 85.6%

      \[\leadsto \left(z \cdot x\right) \cdot \color{blue}{y} \]
3. Recombined 3 regimes into one program.
4. Add Preprocessing

Alternative 8: 65.0% accurate, 0.8× speedup

Math

\[\begin{array}{l} x\_m = \left|x\right| \\ x\_s = \mathsf{copysign}\left(1, x\right) \\ x\_s \cdot \begin{array}{l} \mathbf{if}\;z \leq -0.146 \lor \neg \left(z \leq 2300000000\right):\\ \;\;\;\;\left(-x\_m\right) \cdot z\\ \mathbf{else}:\\ \;\;\;\;x\_m\\ \end{array} \end{array} \]

FPCore

x\_m = (fabs.f64 x)
x\_s = (copysign.f64 #s(literal 1 binary64) x)
(FPCore (x_s x_m y z)
 :precision binary64
 (* x_s (if (or (<= z -0.146) (not (<= z 2300000000.0))) (* (- x_m) z) x_m)))

C

x\_m = fabs(x);
x\_s = copysign(1.0, x);
double code(double x_s, double x_m, double y, double z) {
	double tmp;
	if ((z <= -0.146) || !(z <= 2300000000.0)) {
		tmp = -x_m * z;
	} else {
		tmp = x_m;
	}
	return x_s * tmp;
}

Fortran

x\_m =     private
x\_s =     private
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(x_s, x_m, y, z)
use fmin_fmax_functions
    real(8), intent (in) :: x_s
    real(8), intent (in) :: x_m
    real(8), intent (in) :: y
    real(8), intent (in) :: z
    real(8) :: tmp
    if ((z <= (-0.146d0)) .or. (.not. (z <= 2300000000.0d0))) then
        tmp = -x_m * z
    else
        tmp = x_m
    end if
    code = x_s * tmp
end function

Java

x\_m = Math.abs(x);
x\_s = Math.copySign(1.0, x);
public static double code(double x_s, double x_m, double y, double z) {
	double tmp;
	if ((z <= -0.146) || !(z <= 2300000000.0)) {
		tmp = -x_m * z;
	} else {
		tmp = x_m;
	}
	return x_s * tmp;
}

Python

x\_m = math.fabs(x)
x\_s = math.copysign(1.0, x)
def code(x_s, x_m, y, z):
	tmp = 0
	if (z <= -0.146) or not (z <= 2300000000.0):
		tmp = -x_m * z
	else:
		tmp = x_m
	return x_s * tmp

Julia

x\_m = abs(x)
x\_s = copysign(1.0, x)
function code(x_s, x_m, y, z)
	tmp = 0.0
	if ((z <= -0.146) || !(z <= 2300000000.0))
		tmp = Float64(Float64(-x_m) * z);
	else
		tmp = x_m;
	end
	return Float64(x_s * tmp)
end

MATLAB

x\_m = abs(x);
x\_s = sign(x) * abs(1.0);
function tmp_2 = code(x_s, x_m, y, z)
	tmp = 0.0;
	if ((z <= -0.146) || ~((z <= 2300000000.0)))
		tmp = -x_m * z;
	else
		tmp = x_m;
	end
	tmp_2 = x_s * tmp;
end

Wolfram

x\_m = N[Abs[x], $MachinePrecision]
x\_s = N[With[{TMP1 = Abs[1.0], TMP2 = Sign[x]}, TMP1 * If[TMP2 == 0, 1, TMP2]], $MachinePrecision]
code[x$95$s_, x$95$m_, y_, z_] := N[(x$95$s * If[Or[LessEqual[z, -0.146], N[Not[LessEqual[z, 2300000000.0]], $MachinePrecision]], N[((-x$95$m) * z), $MachinePrecision], x$95$m]), $MachinePrecision]

Derivation

1. Split input into 2 regimes

2. if z < -0.145999999999999991 or 2.3e9 < z

   1. Initial program 93.2%

      \[x \cdot \left(1 - \left(1 - y\right) \cdot z\right) \]
   2. Add Preprocessing

   3. Taylor expanded in z around inf

      \[\leadsto \color{blue}{x \cdot \left(z \cdot \left(y - 1\right)\right)} \]

   5. Applied rewrites 99.3%

      \[\leadsto \color{blue}{\left(\left(y - 1\right) \cdot x\right) \cdot z} \]

   6. Taylor expanded in y around 0

      \[\leadsto \left(-1 \cdot x\right) \cdot z \]
   7. Step-by-step derivation

      1. mul-1-neg N/A

         \[\leadsto \left(\mathsf{neg}\left(x\right)\right) \cdot z \]

      2. lower-neg.f64 62.2

         \[\leadsto \left(-x\right) \cdot z \]

   8. Applied rewrites 62.2%

      \[\leadsto \left(-x\right) \cdot z \]

   if -0.145999999999999991 < z < 2.3e9

   1. Initial program 99.9%

      \[x \cdot \left(1 - \left(1 - y\right) \cdot z\right) \]
   2. Add Preprocessing

   3. Taylor expanded in z around 0

      \[\leadsto \color{blue}{x} \]
   4. Step-by-step derivation

   5. Applied rewrites 74.3%

      \[\leadsto \color{blue}{x} \]
3. Recombined 2 regimes into one program.

4. Final simplification 68.3%

   \[\leadsto \begin{array}{l} \mathbf{if}\;z \leq -0.146 \lor \neg \left(z \leq 2300000000\right):\\ \;\;\;\;\left(-x\right) \cdot z\\ \mathbf{else}:\\ \;\;\;\;x\\ \end{array} \]
5. Add Preprocessing

Alternative 9: 66.2% accurate, 1.9× speedup

Math

\[\begin{array}{l} x\_m = \left|x\right| \\ x\_s = \mathsf{copysign}\left(1, x\right) \\ x\_s \cdot \left(x\_m \cdot \left(1 - z\right)\right) \end{array} \]

FPCore

x\_m = (fabs.f64 x)
x\_s = (copysign.f64 #s(literal 1 binary64) x)
(FPCore (x_s x_m y z) :precision binary64 (* x_s (* x_m (- 1.0 z))))

C

x\_m = fabs(x);
x\_s = copysign(1.0, x);
double code(double x_s, double x_m, double y, double z) {
	return x_s * (x_m * (1.0 - z));
}

Fortran

x\_m =     private
x\_s =     private
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(x_s, x_m, y, z)
use fmin_fmax_functions
    real(8), intent (in) :: x_s
    real(8), intent (in) :: x_m
    real(8), intent (in) :: y
    real(8), intent (in) :: z
    code = x_s * (x_m * (1.0d0 - z))
end function

Java

x\_m = Math.abs(x);
x\_s = Math.copySign(1.0, x);
public static double code(double x_s, double x_m, double y, double z) {
	return x_s * (x_m * (1.0 - z));
}

Python

x\_m = math.fabs(x)
x\_s = math.copysign(1.0, x)
def code(x_s, x_m, y, z):
	return x_s * (x_m * (1.0 - z))

Julia

x\_m = abs(x)
x\_s = copysign(1.0, x)
function code(x_s, x_m, y, z)
	return Float64(x_s * Float64(x_m * Float64(1.0 - z)))
end

MATLAB

x\_m = abs(x);
x\_s = sign(x) * abs(1.0);
function tmp = code(x_s, x_m, y, z)
	tmp = x_s * (x_m * (1.0 - z));
end

Wolfram

x\_m = N[Abs[x], $MachinePrecision]
x\_s = N[With[{TMP1 = Abs[1.0], TMP2 = Sign[x]}, TMP1 * If[TMP2 == 0, 1, TMP2]], $MachinePrecision]
code[x$95$s_, x$95$m_, y_, z_] := N[(x$95$s * N[(x$95$m * N[(1.0 - z), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]

Derivation

1. Initial program 96.6%

   \[x \cdot \left(1 - \left(1 - y\right) \cdot z\right) \]
2. Add Preprocessing

3. Taylor expanded in y around 0

   \[\leadsto x \cdot \left(1 - \color{blue}{z}\right) \]
4. Step-by-step derivation

5. Applied rewrites 69.2%

   \[\leadsto x \cdot \left(1 - \color{blue}{z}\right) \]
6. Add Preprocessing

Alternative 10: 38.0% accurate, 17.0× speedup

Math

\[\begin{array}{l} x\_m = \left|x\right| \\ x\_s = \mathsf{copysign}\left(1, x\right) \\ x\_s \cdot x\_m \end{array} \]

FPCore

x\_m = (fabs.f64 x)
x\_s = (copysign.f64 #s(literal 1 binary64) x)
(FPCore (x_s x_m y z) :precision binary64 (* x_s x_m))

C

x\_m = fabs(x);
x\_s = copysign(1.0, x);
double code(double x_s, double x_m, double y, double z) {
	return x_s * x_m;
}

Fortran

x\_m =     private
x\_s =     private
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(x_s, x_m, y, z)
use fmin_fmax_functions
    real(8), intent (in) :: x_s
    real(8), intent (in) :: x_m
    real(8), intent (in) :: y
    real(8), intent (in) :: z
    code = x_s * x_m
end function

Java

x\_m = Math.abs(x);
x\_s = Math.copySign(1.0, x);
public static double code(double x_s, double x_m, double y, double z) {
	return x_s * x_m;
}

Python

x\_m = math.fabs(x)
x\_s = math.copysign(1.0, x)
def code(x_s, x_m, y, z):
	return x_s * x_m

Julia

x\_m = abs(x)
x\_s = copysign(1.0, x)
function code(x_s, x_m, y, z)
	return Float64(x_s * x_m)
end

MATLAB

x\_m = abs(x);
x\_s = sign(x) * abs(1.0);
function tmp = code(x_s, x_m, y, z)
	tmp = x_s * x_m;
end

Wolfram

x\_m = N[Abs[x], $MachinePrecision]
x\_s = N[With[{TMP1 = Abs[1.0], TMP2 = Sign[x]}, TMP1 * If[TMP2 == 0, 1, TMP2]], $MachinePrecision]
code[x$95$s_, x$95$m_, y_, z_] := N[(x$95$s * x$95$m), $MachinePrecision]

Derivation

1. Initial program 96.6%

   \[x \cdot \left(1 - \left(1 - y\right) \cdot z\right) \]
2. Add Preprocessing

3. Taylor expanded in z around 0

   \[\leadsto \color{blue}{x} \]
4. Step-by-step derivation

5. Applied rewrites 38.9%

   \[\leadsto \color{blue}{x} \]
6. Add Preprocessing

Developer Target 1: 99.6% accurate, 0.3× speedup

Math

\[\begin{array}{l} \\ \begin{array}{l} t_0 := x \cdot \left(1 - \left(1 - y\right) \cdot z\right)\\ t_1 := x + \left(1 - y\right) \cdot \left(\left(-z\right) \cdot x\right)\\ \mathbf{if}\;t\_0 < -1.618195973607049 \cdot 10^{+50}:\\ \;\;\;\;t\_1\\ \mathbf{elif}\;t\_0 < 3.892237649663903 \cdot 10^{+134}:\\ \;\;\;\;\left(x \cdot y\right) \cdot z - \left(x \cdot z - x\right)\\ \mathbf{else}:\\ \;\;\;\;t\_1\\ \end{array} \end{array} \]

FPCore

(FPCore (x y z)
 :precision binary64
 (let* ((t_0 (* x (- 1.0 (* (- 1.0 y) z))))
        (t_1 (+ x (* (- 1.0 y) (* (- z) x)))))
   (if (< t_0 -1.618195973607049e+50)
     t_1
     (if (< t_0 3.892237649663903e+134) (- (* (* x y) z) (- (* x z) x)) t_1))))

C

double code(double x, double y, double z) {
	double t_0 = x * (1.0 - ((1.0 - y) * z));
	double t_1 = x + ((1.0 - y) * (-z * x));
	double tmp;
	if (t_0 < -1.618195973607049e+50) {
		tmp = t_1;
	} else if (t_0 < 3.892237649663903e+134) {
		tmp = ((x * y) * z) - ((x * z) - x);
	} else {
		tmp = t_1;
	}
	return tmp;
}

Fortran

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(x, y, z)
use fmin_fmax_functions
    real(8), intent (in) :: x
    real(8), intent (in) :: y
    real(8), intent (in) :: z
    real(8) :: t_0
    real(8) :: t_1
    real(8) :: tmp
    t_0 = x * (1.0d0 - ((1.0d0 - y) * z))
    t_1 = x + ((1.0d0 - y) * (-z * x))
    if (t_0 < (-1.618195973607049d+50)) then
        tmp = t_1
    else if (t_0 < 3.892237649663903d+134) then
        tmp = ((x * y) * z) - ((x * z) - x)
    else
        tmp = t_1
    end if
    code = tmp
end function

Java

public static double code(double x, double y, double z) {
	double t_0 = x * (1.0 - ((1.0 - y) * z));
	double t_1 = x + ((1.0 - y) * (-z * x));
	double tmp;
	if (t_0 < -1.618195973607049e+50) {
		tmp = t_1;
	} else if (t_0 < 3.892237649663903e+134) {
		tmp = ((x * y) * z) - ((x * z) - x);
	} else {
		tmp = t_1;
	}
	return tmp;
}

Python

def code(x, y, z):
	t_0 = x * (1.0 - ((1.0 - y) * z))
	t_1 = x + ((1.0 - y) * (-z * x))
	tmp = 0
	if t_0 < -1.618195973607049e+50:
		tmp = t_1
	elif t_0 < 3.892237649663903e+134:
		tmp = ((x * y) * z) - ((x * z) - x)
	else:
		tmp = t_1
	return tmp

Julia

function code(x, y, z)
	t_0 = Float64(x * Float64(1.0 - Float64(Float64(1.0 - y) * z)))
	t_1 = Float64(x + Float64(Float64(1.0 - y) * Float64(Float64(-z) * x)))
	tmp = 0.0
	if (t_0 < -1.618195973607049e+50)
		tmp = t_1;
	elseif (t_0 < 3.892237649663903e+134)
		tmp = Float64(Float64(Float64(x * y) * z) - Float64(Float64(x * z) - x));
	else
		tmp = t_1;
	end
	return tmp
end

MATLAB

function tmp_2 = code(x, y, z)
	t_0 = x * (1.0 - ((1.0 - y) * z));
	t_1 = x + ((1.0 - y) * (-z * x));
	tmp = 0.0;
	if (t_0 < -1.618195973607049e+50)
		tmp = t_1;
	elseif (t_0 < 3.892237649663903e+134)
		tmp = ((x * y) * z) - ((x * z) - x);
	else
		tmp = t_1;
	end
	tmp_2 = tmp;
end

Wolfram

code[x_, y_, z_] := Block[{t$95$0 = N[(x * N[(1.0 - N[(N[(1.0 - y), $MachinePrecision] * z), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]}, Block[{t$95$1 = N[(x + N[(N[(1.0 - y), $MachinePrecision] * N[((-z) * x), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]}, If[Less[t$95$0, -1.618195973607049e+50], t$95$1, If[Less[t$95$0, 3.892237649663903e+134], N[(N[(N[(x * y), $MachinePrecision] * z), $MachinePrecision] - N[(N[(x * z), $MachinePrecision] - x), $MachinePrecision]), $MachinePrecision], t$95$1]]]]

Reproduce

herbie shell --seed 2025026 
(FPCore (x y z)
  :name "Data.Colour.RGBSpace.HSV:hsv from colour-2.3.3, J"
  :precision binary64

  :alt
  (! :herbie-platform default (if (< (* x (- 1 (* (- 1 y) z))) -161819597360704900000000000000000000000000000000000) (+ x (* (- 1 y) (* (- z) x))) (if (< (* x (- 1 (* (- 1 y) z))) 389223764966390300000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000) (- (* (* x y) z) (- (* x z) x)) (+ x (* (- 1 y) (* (- z) x))))))

  (* x (- 1.0 (* (- 1.0 y) z))))