Data.Colour.CIE:cieLABView from colour-2.3.3, B

Percentage Accurate: 100.0% → 100.0%
Time: 2.9s
Alternatives: 4
Speedup: 1.0×

Specification

?
\[\begin{array}{l} \\ 500 \cdot \left(x - y\right) \end{array} \]
(FPCore (x y) :precision binary64 (* 500.0 (- x y)))
double code(double x, double y) {
	return 500.0 * (x - y);
}

real(8) function code(x, y)
    real(8), intent (in) :: x
    real(8), intent (in) :: y
    code = 500.0d0 * (x - y)
end function

public static double code(double x, double y) {
	return 500.0 * (x - y);
}

def code(x, y):
	return 500.0 * (x - y)

function code(x, y)
	return Float64(500.0 * Float64(x - y))
end

function tmp = code(x, y)
	tmp = 500.0 * (x - y);
end

code[x_, y_] := N[(500.0 * N[(x - y), $MachinePrecision]), $MachinePrecision]

\begin{array}{l}

\\
500 \cdot \left(x - y\right)
\end{array}

Sampling outcomes in binary64 precision:

Local Percentage Accuracy vs ?

The average percentage accuracy by input value. Horizontal axis shows value of an input variable; the variable is choosen in the title. Vertical axis is accuracy; higher is better. Red represent the original program, while blue represents Herbie's suggestion. These can be toggled with buttons below the plot. The line is an average while dots represent individual samples.

Accuracy vs Speed?

Herbie found 4 alternatives:

AlternativeAccuracySpeedup
The accuracy (vertical axis) and speed (horizontal axis) of each alternatives. Up and to the right is better. The red square shows the initial program, and each blue circle shows an alternative.The line shows the best available speed-accuracy tradeoffs.

Initial Program: 100.0% accurate, 1.0× speedup?

\[\begin{array}{l} \\ 500 \cdot \left(x - y\right) \end{array} \]
(FPCore (x y) :precision binary64 (* 500.0 (- x y)))
double code(double x, double y) {
	return 500.0 * (x - y);
}

real(8) function code(x, y)
    real(8), intent (in) :: x
    real(8), intent (in) :: y
    code = 500.0d0 * (x - y)
end function

public static double code(double x, double y) {
	return 500.0 * (x - y);
}

def code(x, y):
	return 500.0 * (x - y)

function code(x, y)
	return Float64(500.0 * Float64(x - y))
end

function tmp = code(x, y)
	tmp = 500.0 * (x - y);
end

code[x_, y_] := N[(500.0 * N[(x - y), $MachinePrecision]), $MachinePrecision]

\begin{array}{l}

\\
500 \cdot \left(x - y\right)
\end{array}

Alternative 1: 100.0% accurate, 0.0× speedup?

\[\begin{array}{l} \\ \mathsf{fma}\left(x, 500, 500 \cdot \left(-y\right)\right) \end{array} \]
(FPCore (x y) :precision binary64 (fma x 500.0 (* 500.0 (- y))))
double code(double x, double y) {
	return fma(x, 500.0, (500.0 * -y));
}

function code(x, y)
	return fma(x, 500.0, Float64(500.0 * Float64(-y)))
end

code[x_, y_] := N[(x * 500.0 + N[(500.0 * (-y)), $MachinePrecision]), $MachinePrecision]

\begin{array}{l}

\\
\mathsf{fma}\left(x, 500, 500 \cdot \left(-y\right)\right)
\end{array}
Derivation

  1. Initial program 100.0%

    \[500 \cdot \left(x - y\right) \]
  2. Add Preprocessing
  3. Step-by-step derivation

    1. sub-neg100.0%

      \[\leadsto 500 \cdot \color{blue}{\left(x + \left(-y\right)\right)} \]

    2. distribute-rgt-in100.0%

      \[\leadsto \color{blue}{x \cdot 500 + \left(-y\right) \cdot 500} \]

    3. fma-define100.0%

      \[\leadsto \color{blue}{\mathsf{fma}\left(x, 500, \left(-y\right) \cdot 500\right)} \]

  4. Applied egg-rr100.0%

    \[\leadsto \color{blue}{\mathsf{fma}\left(x, 500, \left(-y\right) \cdot 500\right)} \]

  5. Final simplification100.0%

    \[\leadsto \mathsf{fma}\left(x, 500, 500 \cdot \left(-y\right)\right) \]
  6. Add Preprocessing

Alternative 2: 71.8% accurate, 0.4× speedup?

\[\begin{array}{l} \\ \begin{array}{l} \mathbf{if}\;y \leq -3300000 \lor \neg \left(y \leq 3 \cdot 10^{-143}\right):\\ \;\;\;\;y \cdot -500\\ \mathbf{else}:\\ \;\;\;\;x \cdot 500\\ \end{array} \end{array} \]
(FPCore (x y)
 :precision binary64
 (if (or (<= y -3300000.0) (not (<= y 3e-143))) (* y -500.0) (* x 500.0)))
double code(double x, double y) {
	double tmp;
	if ((y <= -3300000.0) || !(y <= 3e-143)) {
		tmp = y * -500.0;
	} else {
		tmp = x * 500.0;
	}
	return tmp;
}

real(8) function code(x, y)
    real(8), intent (in) :: x
    real(8), intent (in) :: y
    real(8) :: tmp
    if ((y <= (-3300000.0d0)) .or. (.not. (y <= 3d-143))) then
        tmp = y * (-500.0d0)
    else
        tmp = x * 500.0d0
    end if
    code = tmp
end function

public static double code(double x, double y) {
	double tmp;
	if ((y <= -3300000.0) || !(y <= 3e-143)) {
		tmp = y * -500.0;
	} else {
		tmp = x * 500.0;
	}
	return tmp;
}

def code(x, y):
	tmp = 0
	if (y <= -3300000.0) or not (y <= 3e-143):
		tmp = y * -500.0
	else:
		tmp = x * 500.0
	return tmp

function code(x, y)
	tmp = 0.0
	if ((y <= -3300000.0) || !(y <= 3e-143))
		tmp = Float64(y * -500.0);
	else
		tmp = Float64(x * 500.0);
	end
	return tmp
end

function tmp_2 = code(x, y)
	tmp = 0.0;
	if ((y <= -3300000.0) || ~((y <= 3e-143)))
		tmp = y * -500.0;
	else
		tmp = x * 500.0;
	end
	tmp_2 = tmp;
end

code[x_, y_] := If[Or[LessEqual[y, -3300000.0], N[Not[LessEqual[y, 3e-143]], $MachinePrecision]], N[(y * -500.0), $MachinePrecision], N[(x * 500.0), $MachinePrecision]]

\begin{array}{l}

\\
\begin{array}{l}
\mathbf{if}\;y \leq -3300000 \lor \neg \left(y \leq 3 \cdot 10^{-143}\right):\\
\;\;\;\;y \cdot -500\\

\mathbf{else}:\\
\;\;\;\;x \cdot 500\\


\end{array}
\end{array}
Derivation
  1. Split input into 2 regimes

  2. if y < -3.3e6 or 2.99999999999999985e-143 < y

    1. Initial program 100.0%

      \[500 \cdot \left(x - y\right) \]
    2. Add Preprocessing

    3. Taylor expanded in x around 0 74.5%

      \[\leadsto \color{blue}{-500 \cdot y} \]

    if -3.3e6 < y < 2.99999999999999985e-143

    1. Initial program 100.0%

      \[500 \cdot \left(x - y\right) \]
    2. Add Preprocessing

    3. Taylor expanded in x around inf 87.7%

      \[\leadsto \color{blue}{500 \cdot x} \]
  3. Recombined 2 regimes into one program.

  4. Final simplification79.6%

    \[\leadsto \begin{array}{l} \mathbf{if}\;y \leq -3300000 \lor \neg \left(y \leq 3 \cdot 10^{-143}\right):\\ \;\;\;\;y \cdot -500\\ \mathbf{else}:\\ \;\;\;\;x \cdot 500\\ \end{array} \]
  5. Add Preprocessing

Alternative 3: 100.0% accurate, 1.0× speedup?

\[\begin{array}{l} \\ 500 \cdot \left(x - y\right) \end{array} \]
(FPCore (x y) :precision binary64 (* 500.0 (- x y)))
double code(double x, double y) {
	return 500.0 * (x - y);
}

real(8) function code(x, y)
    real(8), intent (in) :: x
    real(8), intent (in) :: y
    code = 500.0d0 * (x - y)
end function

public static double code(double x, double y) {
	return 500.0 * (x - y);
}

def code(x, y):
	return 500.0 * (x - y)

function code(x, y)
	return Float64(500.0 * Float64(x - y))
end

function tmp = code(x, y)
	tmp = 500.0 * (x - y);
end

code[x_, y_] := N[(500.0 * N[(x - y), $MachinePrecision]), $MachinePrecision]

\begin{array}{l}

\\
500 \cdot \left(x - y\right)
\end{array}
Derivation

  1. Initial program 100.0%

    \[500 \cdot \left(x - y\right) \]
  2. Add Preprocessing
  3. Add Preprocessing

Alternative 4: 50.5% accurate, 1.7× speedup?

\[\begin{array}{l} \\ y \cdot -500 \end{array} \]
(FPCore (x y) :precision binary64 (* y -500.0))
double code(double x, double y) {
	return y * -500.0;
}

real(8) function code(x, y)
    real(8), intent (in) :: x
    real(8), intent (in) :: y
    code = y * (-500.0d0)
end function

public static double code(double x, double y) {
	return y * -500.0;
}

def code(x, y):
	return y * -500.0

function code(x, y)
	return Float64(y * -500.0)
end

function tmp = code(x, y)
	tmp = y * -500.0;
end

code[x_, y_] := N[(y * -500.0), $MachinePrecision]

\begin{array}{l}

\\
y \cdot -500
\end{array}
Derivation

  1. Initial program 100.0%

    \[500 \cdot \left(x - y\right) \]
  2. Add Preprocessing

  3. Taylor expanded in x around 0 51.0%

    \[\leadsto \color{blue}{-500 \cdot y} \]

  4. Final simplification51.0%

    \[\leadsto y \cdot -500 \]
  5. Add Preprocessing

Reproduce

?
herbie shell --seed 2024116 
(FPCore (x y)
  :name "Data.Colour.CIE:cieLABView from colour-2.3.3, B"
  :precision binary64
  (* 500.0 (- x y)))