b parameter of renormalized beta distribution

Percentage Accurate: 99.9% → 99.9%

Time: 7.9s

Alternatives: 16

Speedup: 1.0×

• 42.1% of points produce a very large (infinite) output. You may want to add a precondition.

Specification

\[\left(0 < m \land 0 < v\right) \land v < 0.25\]

Math

\[\begin{array}{l} \\ \left(\frac{m \cdot \left(1 - m\right)}{v} - 1\right) \cdot \left(1 - m\right) \end{array} \]

FPCore

(FPCore (m v) :precision binary64 (* (- (/ (* m (- 1.0 m)) v) 1.0) (- 1.0 m)))

C

double code(double m, double v) {
	return (((m * (1.0 - m)) / v) - 1.0) * (1.0 - m);
}

Fortran

real(8) function code(m, v)
    real(8), intent (in) :: m
    real(8), intent (in) :: v
    code = (((m * (1.0d0 - m)) / v) - 1.0d0) * (1.0d0 - m)
end function

Java

public static double code(double m, double v) {
	return (((m * (1.0 - m)) / v) - 1.0) * (1.0 - m);
}

Python

def code(m, v):
	return (((m * (1.0 - m)) / v) - 1.0) * (1.0 - m)

Julia

function code(m, v)
	return Float64(Float64(Float64(Float64(m * Float64(1.0 - m)) / v) - 1.0) * Float64(1.0 - m))
end

MATLAB

function tmp = code(m, v)
	tmp = (((m * (1.0 - m)) / v) - 1.0) * (1.0 - m);
end

Wolfram

code[m_, v_] := N[(N[(N[(N[(m * N[(1.0 - m), $MachinePrecision]), $MachinePrecision] / v), $MachinePrecision] - 1.0), $MachinePrecision] * N[(1.0 - m), $MachinePrecision]), $MachinePrecision]

Initial Program: 99.9% accurate, 1.0× speedup

Math

\[\begin{array}{l} \\ \left(\frac{m \cdot \left(1 - m\right)}{v} - 1\right) \cdot \left(1 - m\right) \end{array} \]

FPCore

(FPCore (m v) :precision binary64 (* (- (/ (* m (- 1.0 m)) v) 1.0) (- 1.0 m)))

C

double code(double m, double v) {
	return (((m * (1.0 - m)) / v) - 1.0) * (1.0 - m);
}

Fortran

real(8) function code(m, v)
    real(8), intent (in) :: m
    real(8), intent (in) :: v
    code = (((m * (1.0d0 - m)) / v) - 1.0d0) * (1.0d0 - m)
end function

Java

public static double code(double m, double v) {
	return (((m * (1.0 - m)) / v) - 1.0) * (1.0 - m);
}

Python

def code(m, v):
	return (((m * (1.0 - m)) / v) - 1.0) * (1.0 - m)

Julia

function code(m, v)
	return Float64(Float64(Float64(Float64(m * Float64(1.0 - m)) / v) - 1.0) * Float64(1.0 - m))
end

MATLAB

function tmp = code(m, v)
	tmp = (((m * (1.0 - m)) / v) - 1.0) * (1.0 - m);
end

Wolfram

code[m_, v_] := N[(N[(N[(N[(m * N[(1.0 - m), $MachinePrecision]), $MachinePrecision] / v), $MachinePrecision] - 1.0), $MachinePrecision] * N[(1.0 - m), $MachinePrecision]), $MachinePrecision]

Alternative 1: 99.9% accurate, 0.9× speedup

Math

\[\begin{array}{l} \\ \begin{array}{l} \mathbf{if}\;m \leq 5.8 \cdot 10^{-10}:\\ \;\;\;\;-1 + \mathsf{fma}\left(\frac{m}{v}, \mathsf{fma}\left(m, -2, 1\right), m\right)\\ \mathbf{else}:\\ \;\;\;\;\frac{\mathsf{fma}\left(m + -2, m \cdot m, m\right)}{v}\\ \end{array} \end{array} \]

FPCore

(FPCore (m v)
 :precision binary64
 (if (<= m 5.8e-10)
   (+ -1.0 (fma (/ m v) (fma m -2.0 1.0) m))
   (/ (fma (+ m -2.0) (* m m) m) v)))

C

double code(double m, double v) {
	double tmp;
	if (m <= 5.8e-10) {
		tmp = -1.0 + fma((m / v), fma(m, -2.0, 1.0), m);
	} else {
		tmp = fma((m + -2.0), (m * m), m) / v;
	}
	return tmp;
}

Julia

function code(m, v)
	tmp = 0.0
	if (m <= 5.8e-10)
		tmp = Float64(-1.0 + fma(Float64(m / v), fma(m, -2.0, 1.0), m));
	else
		tmp = Float64(fma(Float64(m + -2.0), Float64(m * m), m) / v);
	end
	return tmp
end

Wolfram

code[m_, v_] := If[LessEqual[m, 5.8e-10], N[(-1.0 + N[(N[(m / v), $MachinePrecision] * N[(m * -2.0 + 1.0), $MachinePrecision] + m), $MachinePrecision]), $MachinePrecision], N[(N[(N[(m + -2.0), $MachinePrecision] * N[(m * m), $MachinePrecision] + m), $MachinePrecision] / v), $MachinePrecision]]

Derivation

1. Split input into 2 regimes

2. if m < 5.79999999999999962e-10

   1. Initial program 100.0%

      \[\left(\frac{m \cdot \left(1 - m\right)}{v} - 1\right) \cdot \left(1 - m\right) \]
   2. Add Preprocessing

   3. Taylor expanded in m around 0

      \[\leadsto \color{blue}{m \cdot \left(1 + \left(-2 \cdot \frac{m}{v} + \frac{1}{v}\right)\right) - 1} \]
   4. Step-by-step derivation

      1. sub-neg N/A

         \[\leadsto \color{blue}{m \cdot \left(1 + \left(-2 \cdot \frac{m}{v} + \frac{1}{v}\right)\right) + \left(\mathsf{neg}\left(1\right)\right)} \]

      2. metadata-eval N/A

         \[\leadsto m \cdot \left(1 + \left(-2 \cdot \frac{m}{v} + \frac{1}{v}\right)\right) + \color{blue}{-1} \]

      3. +-commutative N/A

         \[\leadsto \color{blue}{-1 + m \cdot \left(1 + \left(-2 \cdot \frac{m}{v} + \frac{1}{v}\right)\right)} \]

      4. lower-+.f64 N/A

         \[\leadsto \color{blue}{-1 + m \cdot \left(1 + \left(-2 \cdot \frac{m}{v} + \frac{1}{v}\right)\right)} \]

      5. +-commutative N/A

         \[\leadsto -1 + m \cdot \color{blue}{\left(\left(-2 \cdot \frac{m}{v} + \frac{1}{v}\right) + 1\right)} \]

      6. distribute-lft-in N/A

         \[\leadsto -1 + \color{blue}{\left(m \cdot \left(-2 \cdot \frac{m}{v} + \frac{1}{v}\right) + m \cdot 1\right)} \]

      7. distribute-lft-in N/A

         \[\leadsto -1 + \left(\color{blue}{\left(m \cdot \left(-2 \cdot \frac{m}{v}\right) + m \cdot \frac{1}{v}\right)} + m \cdot 1\right) \]

      8. associate-*r* N/A

         \[\leadsto -1 + \left(\left(\color{blue}{\left(m \cdot -2\right) \cdot \frac{m}{v}} + m \cdot \frac{1}{v}\right) + m \cdot 1\right) \]

      9. *-commutative N/A

         \[\leadsto -1 + \left(\left(\color{blue}{\left(-2 \cdot m\right)} \cdot \frac{m}{v} + m \cdot \frac{1}{v}\right) + m \cdot 1\right) \]

      10. *-commutative N/A

          \[\leadsto -1 + \left(\left(\left(-2 \cdot m\right) \cdot \frac{m}{v} + \color{blue}{\frac{1}{v} \cdot m}\right) + m \cdot 1\right) \]

      11. associate-*l/ N/A

          \[\leadsto -1 + \left(\left(\left(-2 \cdot m\right) \cdot \frac{m}{v} + \color{blue}{\frac{1 \cdot m}{v}}\right) + m \cdot 1\right) \]

      12. associate-/l* N/A

          \[\leadsto -1 + \left(\left(\left(-2 \cdot m\right) \cdot \frac{m}{v} + \color{blue}{1 \cdot \frac{m}{v}}\right) + m \cdot 1\right) \]

      13. distribute-rgt-out N/A

          \[\leadsto -1 + \left(\color{blue}{\frac{m}{v} \cdot \left(-2 \cdot m + 1\right)} + m \cdot 1\right) \]

      14. *-rgt-identity N/A

          \[\leadsto -1 + \left(\frac{m}{v} \cdot \left(-2 \cdot m + 1\right) + \color{blue}{m}\right) \]

      15. lower-fma.f64 N/A

          \[\leadsto -1 + \color{blue}{\mathsf{fma}\left(\frac{m}{v}, -2 \cdot m + 1, m\right)} \]

      16. lower-/.f64 N/A

          \[\leadsto -1 + \mathsf{fma}\left(\color{blue}{\frac{m}{v}}, -2 \cdot m + 1, m\right) \]

      17. *-commutative N/A

          \[\leadsto -1 + \mathsf{fma}\left(\frac{m}{v}, \color{blue}{m \cdot -2} + 1, m\right) \]

      18. lower-fma.f64 100.0

          \[\leadsto -1 + \mathsf{fma}\left(\frac{m}{v}, \color{blue}{\mathsf{fma}\left(m, -2, 1\right)}, m\right) \]

   5. Applied rewrites 100.0%

      \[\leadsto \color{blue}{-1 + \mathsf{fma}\left(\frac{m}{v}, \mathsf{fma}\left(m, -2, 1\right), m\right)} \]

   if 5.79999999999999962e-10 < m

   1. Initial program 99.9%

      \[\left(\frac{m \cdot \left(1 - m\right)}{v} - 1\right) \cdot \left(1 - m\right) \]
   2. Add Preprocessing

   3. Taylor expanded in m around 0

      \[\leadsto \color{blue}{m \cdot \left(1 + \left(m \cdot \left(\frac{m}{v} - 2 \cdot \frac{1}{v}\right) + \frac{1}{v}\right)\right) - 1} \]
   4. Step-by-step derivation

      1. associate-+r+ N/A

         \[\leadsto m \cdot \color{blue}{\left(\left(1 + m \cdot \left(\frac{m}{v} - 2 \cdot \frac{1}{v}\right)\right) + \frac{1}{v}\right)} - 1 \]

      2. distribute-lft-in N/A

         \[\leadsto \color{blue}{\left(m \cdot \left(1 + m \cdot \left(\frac{m}{v} - 2 \cdot \frac{1}{v}\right)\right) + m \cdot \frac{1}{v}\right)} - 1 \]

      3. associate-*r/ N/A

         \[\leadsto \left(m \cdot \left(1 + m \cdot \left(\frac{m}{v} - 2 \cdot \frac{1}{v}\right)\right) + \color{blue}{\frac{m \cdot 1}{v}}\right) - 1 \]

      4. *-rgt-identity N/A

         \[\leadsto \left(m \cdot \left(1 + m \cdot \left(\frac{m}{v} - 2 \cdot \frac{1}{v}\right)\right) + \frac{\color{blue}{m}}{v}\right) - 1 \]

      5. associate--l+ N/A

         \[\leadsto \color{blue}{m \cdot \left(1 + m \cdot \left(\frac{m}{v} - 2 \cdot \frac{1}{v}\right)\right) + \left(\frac{m}{v} - 1\right)} \]

      6. *-lft-identity N/A

         \[\leadsto m \cdot \left(1 + m \cdot \left(\frac{m}{v} - 2 \cdot \frac{1}{v}\right)\right) + \left(\frac{\color{blue}{1 \cdot m}}{v} - 1\right) \]

      7. associate-*l/ N/A

         \[\leadsto m \cdot \left(1 + m \cdot \left(\frac{m}{v} - 2 \cdot \frac{1}{v}\right)\right) + \left(\color{blue}{\frac{1}{v} \cdot m} - 1\right) \]

      8. lower-fma.f64 N/A

         \[\leadsto \color{blue}{\mathsf{fma}\left(m, 1 + m \cdot \left(\frac{m}{v} - 2 \cdot \frac{1}{v}\right), \frac{1}{v} \cdot m - 1\right)} \]

   5. Applied rewrites 99.9%

      \[\leadsto \color{blue}{\mathsf{fma}\left(m, \mathsf{fma}\left(\frac{m}{v}, -2 + m, 1\right), -1 + \frac{m}{v}\right)} \]

   6. Taylor expanded in v around 0

      \[\leadsto \frac{m + {m}^{2} \cdot \left(m - 2\right)}{\color{blue}{v}} \]
   7. Step-by-step derivation

   8. Applied rewrites 100.0%

      \[\leadsto \frac{\mathsf{fma}\left(m, m \cdot \left(m + -2\right), m\right)}{\color{blue}{v}} \]
   9. Step-by-step derivation

   10. Applied rewrites 100.0%

       \[\leadsto \color{blue}{\frac{\mathsf{fma}\left(m + -2, m \cdot m, m\right)}{v}} \]
3. Recombined 2 regimes into one program.
4. Add Preprocessing

Alternative 2: 99.8% accurate, 0.5× speedup

Math

\[\begin{array}{l} \\ \begin{array}{l} \mathbf{if}\;\left(1 - m\right) \cdot \left(-1 + \frac{m \cdot \left(1 - m\right)}{v}\right) \leq 4000000000000:\\ \;\;\;\;-1 + \frac{m}{v}\\ \mathbf{else}:\\ \;\;\;\;\frac{\mathsf{fma}\left(m, m \cdot \left(m + -2\right), m\right)}{v}\\ \end{array} \end{array} \]

FPCore

(FPCore (m v)
 :precision binary64
 (if (<= (* (- 1.0 m) (+ -1.0 (/ (* m (- 1.0 m)) v))) 4000000000000.0)
   (+ -1.0 (/ m v))
   (/ (fma m (* m (+ m -2.0)) m) v)))

C

double code(double m, double v) {
	double tmp;
	if (((1.0 - m) * (-1.0 + ((m * (1.0 - m)) / v))) <= 4000000000000.0) {
		tmp = -1.0 + (m / v);
	} else {
		tmp = fma(m, (m * (m + -2.0)), m) / v;
	}
	return tmp;
}

Julia

function code(m, v)
	tmp = 0.0
	if (Float64(Float64(1.0 - m) * Float64(-1.0 + Float64(Float64(m * Float64(1.0 - m)) / v))) <= 4000000000000.0)
		tmp = Float64(-1.0 + Float64(m / v));
	else
		tmp = Float64(fma(m, Float64(m * Float64(m + -2.0)), m) / v);
	end
	return tmp
end

Wolfram

code[m_, v_] := If[LessEqual[N[(N[(1.0 - m), $MachinePrecision] * N[(-1.0 + N[(N[(m * N[(1.0 - m), $MachinePrecision]), $MachinePrecision] / v), $MachinePrecision]), $MachinePrecision]), $MachinePrecision], 4000000000000.0], N[(-1.0 + N[(m / v), $MachinePrecision]), $MachinePrecision], N[(N[(m * N[(m * N[(m + -2.0), $MachinePrecision]), $MachinePrecision] + m), $MachinePrecision] / v), $MachinePrecision]]

Derivation

1. Split input into 2 regimes

2. if (*.f64 (-.f64 (/.f64 (*.f64 m (-.f64 #s(literal 1 binary64) m)) v) #s(literal 1 binary64)) (-.f64 #s(literal 1 binary64) m)) < 4e12

   1. Initial program 100.0%

      \[\left(\frac{m \cdot \left(1 - m\right)}{v} - 1\right) \cdot \left(1 - m\right) \]
   2. Add Preprocessing

   3. Taylor expanded in m around 0

      \[\leadsto \color{blue}{m \cdot \left(1 + \frac{1}{v}\right) - 1} \]
   4. Step-by-step derivation

      1. sub-neg N/A

         \[\leadsto \color{blue}{m \cdot \left(1 + \frac{1}{v}\right) + \left(\mathsf{neg}\left(1\right)\right)} \]

      2. metadata-eval N/A

         \[\leadsto m \cdot \left(1 + \frac{1}{v}\right) + \color{blue}{-1} \]

      3. +-commutative N/A

         \[\leadsto \color{blue}{-1 + m \cdot \left(1 + \frac{1}{v}\right)} \]

      4. lower-+.f64 N/A

         \[\leadsto \color{blue}{-1 + m \cdot \left(1 + \frac{1}{v}\right)} \]

      5. distribute-rgt-in N/A

         \[\leadsto -1 + \color{blue}{\left(1 \cdot m + \frac{1}{v} \cdot m\right)} \]

      6. associate-*l/ N/A

         \[\leadsto -1 + \left(1 \cdot m + \color{blue}{\frac{1 \cdot m}{v}}\right) \]

      7. *-lft-identity N/A

         \[\leadsto -1 + \left(1 \cdot m + \frac{\color{blue}{m}}{v}\right) \]

      8. *-lft-identity N/A

         \[\leadsto -1 + \left(\color{blue}{m} + \frac{m}{v}\right) \]

      9. lower-+.f64 N/A

         \[\leadsto -1 + \color{blue}{\left(m + \frac{m}{v}\right)} \]

      10. lower-/.f64 100.0

          \[\leadsto -1 + \left(m + \color{blue}{\frac{m}{v}}\right) \]

   5. Applied rewrites 100.0%

      \[\leadsto \color{blue}{-1 + \left(m + \frac{m}{v}\right)} \]

   6. Taylor expanded in v around 0

      \[\leadsto -1 + \frac{m}{\color{blue}{v}} \]
   7. Step-by-step derivation

   8. Applied rewrites 100.0%

      \[\leadsto -1 + \frac{m}{\color{blue}{v}} \]

   if 4e12 < (*.f64 (-.f64 (/.f64 (*.f64 m (-.f64 #s(literal 1 binary64) m)) v) #s(literal 1 binary64)) (-.f64 #s(literal 1 binary64) m))

   1. Initial program 99.9%

      \[\left(\frac{m \cdot \left(1 - m\right)}{v} - 1\right) \cdot \left(1 - m\right) \]
   2. Add Preprocessing

   3. Taylor expanded in m around 0

      \[\leadsto \color{blue}{m \cdot \left(1 + \left(m \cdot \left(\frac{m}{v} - 2 \cdot \frac{1}{v}\right) + \frac{1}{v}\right)\right) - 1} \]
   4. Step-by-step derivation

      1. associate-+r+ N/A

         \[\leadsto m \cdot \color{blue}{\left(\left(1 + m \cdot \left(\frac{m}{v} - 2 \cdot \frac{1}{v}\right)\right) + \frac{1}{v}\right)} - 1 \]

      2. distribute-lft-in N/A

         \[\leadsto \color{blue}{\left(m \cdot \left(1 + m \cdot \left(\frac{m}{v} - 2 \cdot \frac{1}{v}\right)\right) + m \cdot \frac{1}{v}\right)} - 1 \]

      3. associate-*r/ N/A

         \[\leadsto \left(m \cdot \left(1 + m \cdot \left(\frac{m}{v} - 2 \cdot \frac{1}{v}\right)\right) + \color{blue}{\frac{m \cdot 1}{v}}\right) - 1 \]

      4. *-rgt-identity N/A

         \[\leadsto \left(m \cdot \left(1 + m \cdot \left(\frac{m}{v} - 2 \cdot \frac{1}{v}\right)\right) + \frac{\color{blue}{m}}{v}\right) - 1 \]

      5. associate--l+ N/A

         \[\leadsto \color{blue}{m \cdot \left(1 + m \cdot \left(\frac{m}{v} - 2 \cdot \frac{1}{v}\right)\right) + \left(\frac{m}{v} - 1\right)} \]

      6. *-lft-identity N/A

         \[\leadsto m \cdot \left(1 + m \cdot \left(\frac{m}{v} - 2 \cdot \frac{1}{v}\right)\right) + \left(\frac{\color{blue}{1 \cdot m}}{v} - 1\right) \]

      7. associate-*l/ N/A

         \[\leadsto m \cdot \left(1 + m \cdot \left(\frac{m}{v} - 2 \cdot \frac{1}{v}\right)\right) + \left(\color{blue}{\frac{1}{v} \cdot m} - 1\right) \]

      8. lower-fma.f64 N/A

         \[\leadsto \color{blue}{\mathsf{fma}\left(m, 1 + m \cdot \left(\frac{m}{v} - 2 \cdot \frac{1}{v}\right), \frac{1}{v} \cdot m - 1\right)} \]

   5. Applied rewrites 99.9%

      \[\leadsto \color{blue}{\mathsf{fma}\left(m, \mathsf{fma}\left(\frac{m}{v}, -2 + m, 1\right), -1 + \frac{m}{v}\right)} \]

   6. Taylor expanded in v around 0

      \[\leadsto \frac{m + {m}^{2} \cdot \left(m - 2\right)}{\color{blue}{v}} \]
   7. Step-by-step derivation

   8. Applied rewrites 99.9%

      \[\leadsto \frac{\mathsf{fma}\left(m, m \cdot \left(m + -2\right), m\right)}{\color{blue}{v}} \]
3. Recombined 2 regimes into one program.

4. Final simplification 100.0%

   \[\leadsto \begin{array}{l} \mathbf{if}\;\left(1 - m\right) \cdot \left(-1 + \frac{m \cdot \left(1 - m\right)}{v}\right) \leq 4000000000000:\\ \;\;\;\;-1 + \frac{m}{v}\\ \mathbf{else}:\\ \;\;\;\;\frac{\mathsf{fma}\left(m, m \cdot \left(m + -2\right), m\right)}{v}\\ \end{array} \]
5. Add Preprocessing

Alternative 3: 73.5% accurate, 0.6× speedup

Math

\[\begin{array}{l} \\ \begin{array}{l} \mathbf{if}\;\left(1 - m\right) \cdot \left(-1 + \frac{m \cdot \left(1 - m\right)}{v}\right) \leq -0.5:\\ \;\;\;\;-1\\ \mathbf{else}:\\ \;\;\;\;m + \frac{m}{v}\\ \end{array} \end{array} \]

FPCore

(FPCore (m v)
 :precision binary64
 (if (<= (* (- 1.0 m) (+ -1.0 (/ (* m (- 1.0 m)) v))) -0.5)
   -1.0
   (+ m (/ m v))))

C

double code(double m, double v) {
	double tmp;
	if (((1.0 - m) * (-1.0 + ((m * (1.0 - m)) / v))) <= -0.5) {
		tmp = -1.0;
	} else {
		tmp = m + (m / v);
	}
	return tmp;
}

Fortran

real(8) function code(m, v)
    real(8), intent (in) :: m
    real(8), intent (in) :: v
    real(8) :: tmp
    if (((1.0d0 - m) * ((-1.0d0) + ((m * (1.0d0 - m)) / v))) <= (-0.5d0)) then
        tmp = -1.0d0
    else
        tmp = m + (m / v)
    end if
    code = tmp
end function

Java

public static double code(double m, double v) {
	double tmp;
	if (((1.0 - m) * (-1.0 + ((m * (1.0 - m)) / v))) <= -0.5) {
		tmp = -1.0;
	} else {
		tmp = m + (m / v);
	}
	return tmp;
}

Python

def code(m, v):
	tmp = 0
	if ((1.0 - m) * (-1.0 + ((m * (1.0 - m)) / v))) <= -0.5:
		tmp = -1.0
	else:
		tmp = m + (m / v)
	return tmp

Julia

function code(m, v)
	tmp = 0.0
	if (Float64(Float64(1.0 - m) * Float64(-1.0 + Float64(Float64(m * Float64(1.0 - m)) / v))) <= -0.5)
		tmp = -1.0;
	else
		tmp = Float64(m + Float64(m / v));
	end
	return tmp
end

MATLAB

function tmp_2 = code(m, v)
	tmp = 0.0;
	if (((1.0 - m) * (-1.0 + ((m * (1.0 - m)) / v))) <= -0.5)
		tmp = -1.0;
	else
		tmp = m + (m / v);
	end
	tmp_2 = tmp;
end

Wolfram

code[m_, v_] := If[LessEqual[N[(N[(1.0 - m), $MachinePrecision] * N[(-1.0 + N[(N[(m * N[(1.0 - m), $MachinePrecision]), $MachinePrecision] / v), $MachinePrecision]), $MachinePrecision]), $MachinePrecision], -0.5], -1.0, N[(m + N[(m / v), $MachinePrecision]), $MachinePrecision]]

Derivation

1. Split input into 2 regimes

2. if (*.f64 (-.f64 (/.f64 (*.f64 m (-.f64 #s(literal 1 binary64) m)) v) #s(literal 1 binary64)) (-.f64 #s(literal 1 binary64) m)) < -0.5

   1. Initial program 100.0%

      \[\left(\frac{m \cdot \left(1 - m\right)}{v} - 1\right) \cdot \left(1 - m\right) \]
   2. Add Preprocessing

   3. Taylor expanded in m around 0

      \[\leadsto \color{blue}{-1} \]
   4. Step-by-step derivation

   5. Applied rewrites 98.1%

      \[\leadsto \color{blue}{-1} \]

   if -0.5 < (*.f64 (-.f64 (/.f64 (*.f64 m (-.f64 #s(literal 1 binary64) m)) v) #s(literal 1 binary64)) (-.f64 #s(literal 1 binary64) m))

   1. Initial program 99.9%

      \[\left(\frac{m \cdot \left(1 - m\right)}{v} - 1\right) \cdot \left(1 - m\right) \]
   2. Add Preprocessing

   3. Taylor expanded in m around 0

      \[\leadsto \color{blue}{m \cdot \left(1 + \frac{1}{v}\right) - 1} \]
   4. Step-by-step derivation

      1. sub-neg N/A

         \[\leadsto \color{blue}{m \cdot \left(1 + \frac{1}{v}\right) + \left(\mathsf{neg}\left(1\right)\right)} \]

      2. metadata-eval N/A

         \[\leadsto m \cdot \left(1 + \frac{1}{v}\right) + \color{blue}{-1} \]

      3. +-commutative N/A

         \[\leadsto \color{blue}{-1 + m \cdot \left(1 + \frac{1}{v}\right)} \]

      4. lower-+.f64 N/A

         \[\leadsto \color{blue}{-1 + m \cdot \left(1 + \frac{1}{v}\right)} \]

      5. distribute-rgt-in N/A

         \[\leadsto -1 + \color{blue}{\left(1 \cdot m + \frac{1}{v} \cdot m\right)} \]

      6. associate-*l/ N/A

         \[\leadsto -1 + \left(1 \cdot m + \color{blue}{\frac{1 \cdot m}{v}}\right) \]

      7. *-lft-identity N/A

         \[\leadsto -1 + \left(1 \cdot m + \frac{\color{blue}{m}}{v}\right) \]

      8. *-lft-identity N/A

         \[\leadsto -1 + \left(\color{blue}{m} + \frac{m}{v}\right) \]

      9. lower-+.f64 N/A

         \[\leadsto -1 + \color{blue}{\left(m + \frac{m}{v}\right)} \]

      10. lower-/.f64 68.7

          \[\leadsto -1 + \left(m + \color{blue}{\frac{m}{v}}\right) \]

   5. Applied rewrites 68.7%

      \[\leadsto \color{blue}{-1 + \left(m + \frac{m}{v}\right)} \]

   6. Taylor expanded in m around inf

      \[\leadsto m \cdot \color{blue}{\left(1 + \frac{1}{v}\right)} \]
   7. Step-by-step derivation

   8. Applied rewrites 67.0%

      \[\leadsto m + \color{blue}{\frac{m}{v}} \]
3. Recombined 2 regimes into one program.

4. Final simplification 74.4%

   \[\leadsto \begin{array}{l} \mathbf{if}\;\left(1 - m\right) \cdot \left(-1 + \frac{m \cdot \left(1 - m\right)}{v}\right) \leq -0.5:\\ \;\;\;\;-1\\ \mathbf{else}:\\ \;\;\;\;m + \frac{m}{v}\\ \end{array} \]
5. Add Preprocessing

Alternative 4: 73.5% accurate, 0.6× speedup

Math

\[\begin{array}{l} \\ \begin{array}{l} \mathbf{if}\;\left(1 - m\right) \cdot \left(-1 + \frac{m \cdot \left(1 - m\right)}{v}\right) \leq -0.5:\\ \;\;\;\;-1\\ \mathbf{else}:\\ \;\;\;\;\frac{m}{v}\\ \end{array} \end{array} \]

FPCore

(FPCore (m v)
 :precision binary64
 (if (<= (* (- 1.0 m) (+ -1.0 (/ (* m (- 1.0 m)) v))) -0.5) -1.0 (/ m v)))

C

double code(double m, double v) {
	double tmp;
	if (((1.0 - m) * (-1.0 + ((m * (1.0 - m)) / v))) <= -0.5) {
		tmp = -1.0;
	} else {
		tmp = m / v;
	}
	return tmp;
}

Fortran

real(8) function code(m, v)
    real(8), intent (in) :: m
    real(8), intent (in) :: v
    real(8) :: tmp
    if (((1.0d0 - m) * ((-1.0d0) + ((m * (1.0d0 - m)) / v))) <= (-0.5d0)) then
        tmp = -1.0d0
    else
        tmp = m / v
    end if
    code = tmp
end function

Java

public static double code(double m, double v) {
	double tmp;
	if (((1.0 - m) * (-1.0 + ((m * (1.0 - m)) / v))) <= -0.5) {
		tmp = -1.0;
	} else {
		tmp = m / v;
	}
	return tmp;
}

Python

def code(m, v):
	tmp = 0
	if ((1.0 - m) * (-1.0 + ((m * (1.0 - m)) / v))) <= -0.5:
		tmp = -1.0
	else:
		tmp = m / v
	return tmp

Julia

function code(m, v)
	tmp = 0.0
	if (Float64(Float64(1.0 - m) * Float64(-1.0 + Float64(Float64(m * Float64(1.0 - m)) / v))) <= -0.5)
		tmp = -1.0;
	else
		tmp = Float64(m / v);
	end
	return tmp
end

MATLAB

function tmp_2 = code(m, v)
	tmp = 0.0;
	if (((1.0 - m) * (-1.0 + ((m * (1.0 - m)) / v))) <= -0.5)
		tmp = -1.0;
	else
		tmp = m / v;
	end
	tmp_2 = tmp;
end

Wolfram

code[m_, v_] := If[LessEqual[N[(N[(1.0 - m), $MachinePrecision] * N[(-1.0 + N[(N[(m * N[(1.0 - m), $MachinePrecision]), $MachinePrecision] / v), $MachinePrecision]), $MachinePrecision]), $MachinePrecision], -0.5], -1.0, N[(m / v), $MachinePrecision]]

Derivation

1. Split input into 2 regimes

2. if (*.f64 (-.f64 (/.f64 (*.f64 m (-.f64 #s(literal 1 binary64) m)) v) #s(literal 1 binary64)) (-.f64 #s(literal 1 binary64) m)) < -0.5

   1. Initial program 100.0%

      \[\left(\frac{m \cdot \left(1 - m\right)}{v} - 1\right) \cdot \left(1 - m\right) \]
   2. Add Preprocessing

   3. Taylor expanded in m around 0

      \[\leadsto \color{blue}{-1} \]
   4. Step-by-step derivation

   5. Applied rewrites 98.1%

      \[\leadsto \color{blue}{-1} \]

   if -0.5 < (*.f64 (-.f64 (/.f64 (*.f64 m (-.f64 #s(literal 1 binary64) m)) v) #s(literal 1 binary64)) (-.f64 #s(literal 1 binary64) m))

   1. Initial program 99.9%

      \[\left(\frac{m \cdot \left(1 - m\right)}{v} - 1\right) \cdot \left(1 - m\right) \]
   2. Add Preprocessing

   3. Taylor expanded in m around 0

      \[\leadsto \color{blue}{m \cdot \left(1 + \frac{1}{v}\right) - 1} \]
   4. Step-by-step derivation

      1. sub-neg N/A

         \[\leadsto \color{blue}{m \cdot \left(1 + \frac{1}{v}\right) + \left(\mathsf{neg}\left(1\right)\right)} \]

      2. metadata-eval N/A

         \[\leadsto m \cdot \left(1 + \frac{1}{v}\right) + \color{blue}{-1} \]

      3. +-commutative N/A

         \[\leadsto \color{blue}{-1 + m \cdot \left(1 + \frac{1}{v}\right)} \]

      4. lower-+.f64 N/A

         \[\leadsto \color{blue}{-1 + m \cdot \left(1 + \frac{1}{v}\right)} \]

      5. distribute-rgt-in N/A

         \[\leadsto -1 + \color{blue}{\left(1 \cdot m + \frac{1}{v} \cdot m\right)} \]

      6. associate-*l/ N/A

         \[\leadsto -1 + \left(1 \cdot m + \color{blue}{\frac{1 \cdot m}{v}}\right) \]

      7. *-lft-identity N/A

         \[\leadsto -1 + \left(1 \cdot m + \frac{\color{blue}{m}}{v}\right) \]

      8. *-lft-identity N/A

         \[\leadsto -1 + \left(\color{blue}{m} + \frac{m}{v}\right) \]

      9. lower-+.f64 N/A

         \[\leadsto -1 + \color{blue}{\left(m + \frac{m}{v}\right)} \]

      10. lower-/.f64 68.7

          \[\leadsto -1 + \left(m + \color{blue}{\frac{m}{v}}\right) \]

   5. Applied rewrites 68.7%

      \[\leadsto \color{blue}{-1 + \left(m + \frac{m}{v}\right)} \]

   6. Taylor expanded in v around 0

      \[\leadsto \frac{m}{\color{blue}{v}} \]
   7. Step-by-step derivation

   8. Applied rewrites 67.0%

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

4. Final simplification 74.4%

   \[\leadsto \begin{array}{l} \mathbf{if}\;\left(1 - m\right) \cdot \left(-1 + \frac{m \cdot \left(1 - m\right)}{v}\right) \leq -0.5:\\ \;\;\;\;-1\\ \mathbf{else}:\\ \;\;\;\;\frac{m}{v}\\ \end{array} \]
5. Add Preprocessing

Alternative 5: 99.9% accurate, 0.8× speedup

Math

\[\begin{array}{l} \\ \mathsf{fma}\left(m, \mathsf{fma}\left(\frac{m}{v}, m + -2, 1\right), \frac{m}{v} + -1\right) \end{array} \]

FPCore

(FPCore (m v)
 :precision binary64
 (fma m (fma (/ m v) (+ m -2.0) 1.0) (+ (/ m v) -1.0)))

C

double code(double m, double v) {
	return fma(m, fma((m / v), (m + -2.0), 1.0), ((m / v) + -1.0));
}

Julia

function code(m, v)
	return fma(m, fma(Float64(m / v), Float64(m + -2.0), 1.0), Float64(Float64(m / v) + -1.0))
end

Wolfram

code[m_, v_] := N[(m * N[(N[(m / v), $MachinePrecision] * N[(m + -2.0), $MachinePrecision] + 1.0), $MachinePrecision] + N[(N[(m / v), $MachinePrecision] + -1.0), $MachinePrecision]), $MachinePrecision]

Derivation

1. Initial program 100.0%

   \[\left(\frac{m \cdot \left(1 - m\right)}{v} - 1\right) \cdot \left(1 - m\right) \]
2. Add Preprocessing

3. Taylor expanded in m around 0

   \[\leadsto \color{blue}{m \cdot \left(1 + \left(m \cdot \left(\frac{m}{v} - 2 \cdot \frac{1}{v}\right) + \frac{1}{v}\right)\right) - 1} \]
4. Step-by-step derivation

   1. associate-+r+ N/A

      \[\leadsto m \cdot \color{blue}{\left(\left(1 + m \cdot \left(\frac{m}{v} - 2 \cdot \frac{1}{v}\right)\right) + \frac{1}{v}\right)} - 1 \]

   2. distribute-lft-in N/A

      \[\leadsto \color{blue}{\left(m \cdot \left(1 + m \cdot \left(\frac{m}{v} - 2 \cdot \frac{1}{v}\right)\right) + m \cdot \frac{1}{v}\right)} - 1 \]

   3. associate-*r/ N/A

      \[\leadsto \left(m \cdot \left(1 + m \cdot \left(\frac{m}{v} - 2 \cdot \frac{1}{v}\right)\right) + \color{blue}{\frac{m \cdot 1}{v}}\right) - 1 \]

   4. *-rgt-identity N/A

      \[\leadsto \left(m \cdot \left(1 + m \cdot \left(\frac{m}{v} - 2 \cdot \frac{1}{v}\right)\right) + \frac{\color{blue}{m}}{v}\right) - 1 \]

   5. associate--l+ N/A

      \[\leadsto \color{blue}{m \cdot \left(1 + m \cdot \left(\frac{m}{v} - 2 \cdot \frac{1}{v}\right)\right) + \left(\frac{m}{v} - 1\right)} \]

   6. *-lft-identity N/A

      \[\leadsto m \cdot \left(1 + m \cdot \left(\frac{m}{v} - 2 \cdot \frac{1}{v}\right)\right) + \left(\frac{\color{blue}{1 \cdot m}}{v} - 1\right) \]

   7. associate-*l/ N/A

      \[\leadsto m \cdot \left(1 + m \cdot \left(\frac{m}{v} - 2 \cdot \frac{1}{v}\right)\right) + \left(\color{blue}{\frac{1}{v} \cdot m} - 1\right) \]

   8. lower-fma.f64 N/A

      \[\leadsto \color{blue}{\mathsf{fma}\left(m, 1 + m \cdot \left(\frac{m}{v} - 2 \cdot \frac{1}{v}\right), \frac{1}{v} \cdot m - 1\right)} \]

5. Applied rewrites 100.0%

   \[\leadsto \color{blue}{\mathsf{fma}\left(m, \mathsf{fma}\left(\frac{m}{v}, -2 + m, 1\right), -1 + \frac{m}{v}\right)} \]

6. Final simplification 100.0%

   \[\leadsto \mathsf{fma}\left(m, \mathsf{fma}\left(\frac{m}{v}, m + -2, 1\right), \frac{m}{v} + -1\right) \]
7. Add Preprocessing

Alternative 6: 99.8% accurate, 1.0× speedup

Math

\[\begin{array}{l} \\ \begin{array}{l} \mathbf{if}\;m \leq 1.6 \cdot 10^{-15}:\\ \;\;\;\;-1 + \frac{m}{v}\\ \mathbf{else}:\\ \;\;\;\;\frac{\mathsf{fma}\left(m + -2, m \cdot m, m\right)}{v}\\ \end{array} \end{array} \]

FPCore

(FPCore (m v)
 :precision binary64
 (if (<= m 1.6e-15) (+ -1.0 (/ m v)) (/ (fma (+ m -2.0) (* m m) m) v)))

C

double code(double m, double v) {
	double tmp;
	if (m <= 1.6e-15) {
		tmp = -1.0 + (m / v);
	} else {
		tmp = fma((m + -2.0), (m * m), m) / v;
	}
	return tmp;
}

Julia

function code(m, v)
	tmp = 0.0
	if (m <= 1.6e-15)
		tmp = Float64(-1.0 + Float64(m / v));
	else
		tmp = Float64(fma(Float64(m + -2.0), Float64(m * m), m) / v);
	end
	return tmp
end

Wolfram

code[m_, v_] := If[LessEqual[m, 1.6e-15], N[(-1.0 + N[(m / v), $MachinePrecision]), $MachinePrecision], N[(N[(N[(m + -2.0), $MachinePrecision] * N[(m * m), $MachinePrecision] + m), $MachinePrecision] / v), $MachinePrecision]]

Derivation

1. Split input into 2 regimes

2. if m < 1.6e-15

   1. Initial program 100.0%

      \[\left(\frac{m \cdot \left(1 - m\right)}{v} - 1\right) \cdot \left(1 - m\right) \]
   2. Add Preprocessing

   3. Taylor expanded in m around 0

      \[\leadsto \color{blue}{m \cdot \left(1 + \frac{1}{v}\right) - 1} \]
   4. Step-by-step derivation

      1. sub-neg N/A

         \[\leadsto \color{blue}{m \cdot \left(1 + \frac{1}{v}\right) + \left(\mathsf{neg}\left(1\right)\right)} \]

      2. metadata-eval N/A

         \[\leadsto m \cdot \left(1 + \frac{1}{v}\right) + \color{blue}{-1} \]

      3. +-commutative N/A

         \[\leadsto \color{blue}{-1 + m \cdot \left(1 + \frac{1}{v}\right)} \]

      4. lower-+.f64 N/A

         \[\leadsto \color{blue}{-1 + m \cdot \left(1 + \frac{1}{v}\right)} \]

      5. distribute-rgt-in N/A

         \[\leadsto -1 + \color{blue}{\left(1 \cdot m + \frac{1}{v} \cdot m\right)} \]

      6. associate-*l/ N/A

         \[\leadsto -1 + \left(1 \cdot m + \color{blue}{\frac{1 \cdot m}{v}}\right) \]

      7. *-lft-identity N/A

         \[\leadsto -1 + \left(1 \cdot m + \frac{\color{blue}{m}}{v}\right) \]

      8. *-lft-identity N/A

         \[\leadsto -1 + \left(\color{blue}{m} + \frac{m}{v}\right) \]

      9. lower-+.f64 N/A

         \[\leadsto -1 + \color{blue}{\left(m + \frac{m}{v}\right)} \]

      10. lower-/.f64 100.0

          \[\leadsto -1 + \left(m + \color{blue}{\frac{m}{v}}\right) \]

   5. Applied rewrites 100.0%

      \[\leadsto \color{blue}{-1 + \left(m + \frac{m}{v}\right)} \]

   6. Taylor expanded in v around 0

      \[\leadsto -1 + \frac{m}{\color{blue}{v}} \]
   7. Step-by-step derivation

   8. Applied rewrites 100.0%

      \[\leadsto -1 + \frac{m}{\color{blue}{v}} \]

   if 1.6e-15 < m

   1. Initial program 99.9%

      \[\left(\frac{m \cdot \left(1 - m\right)}{v} - 1\right) \cdot \left(1 - m\right) \]
   2. Add Preprocessing

   3. Taylor expanded in m around 0

      \[\leadsto \color{blue}{m \cdot \left(1 + \left(m \cdot \left(\frac{m}{v} - 2 \cdot \frac{1}{v}\right) + \frac{1}{v}\right)\right) - 1} \]
   4. Step-by-step derivation

      1. associate-+r+ N/A

         \[\leadsto m \cdot \color{blue}{\left(\left(1 + m \cdot \left(\frac{m}{v} - 2 \cdot \frac{1}{v}\right)\right) + \frac{1}{v}\right)} - 1 \]

      2. distribute-lft-in N/A

         \[\leadsto \color{blue}{\left(m \cdot \left(1 + m \cdot \left(\frac{m}{v} - 2 \cdot \frac{1}{v}\right)\right) + m \cdot \frac{1}{v}\right)} - 1 \]

      3. associate-*r/ N/A

         \[\leadsto \left(m \cdot \left(1 + m \cdot \left(\frac{m}{v} - 2 \cdot \frac{1}{v}\right)\right) + \color{blue}{\frac{m \cdot 1}{v}}\right) - 1 \]

      4. *-rgt-identity N/A

         \[\leadsto \left(m \cdot \left(1 + m \cdot \left(\frac{m}{v} - 2 \cdot \frac{1}{v}\right)\right) + \frac{\color{blue}{m}}{v}\right) - 1 \]

      5. associate--l+ N/A

         \[\leadsto \color{blue}{m \cdot \left(1 + m \cdot \left(\frac{m}{v} - 2 \cdot \frac{1}{v}\right)\right) + \left(\frac{m}{v} - 1\right)} \]

      6. *-lft-identity N/A

         \[\leadsto m \cdot \left(1 + m \cdot \left(\frac{m}{v} - 2 \cdot \frac{1}{v}\right)\right) + \left(\frac{\color{blue}{1 \cdot m}}{v} - 1\right) \]

      7. associate-*l/ N/A

         \[\leadsto m \cdot \left(1 + m \cdot \left(\frac{m}{v} - 2 \cdot \frac{1}{v}\right)\right) + \left(\color{blue}{\frac{1}{v} \cdot m} - 1\right) \]

      8. lower-fma.f64 N/A

         \[\leadsto \color{blue}{\mathsf{fma}\left(m, 1 + m \cdot \left(\frac{m}{v} - 2 \cdot \frac{1}{v}\right), \frac{1}{v} \cdot m - 1\right)} \]

   5. Applied rewrites 99.9%

      \[\leadsto \color{blue}{\mathsf{fma}\left(m, \mathsf{fma}\left(\frac{m}{v}, -2 + m, 1\right), -1 + \frac{m}{v}\right)} \]

   6. Taylor expanded in v around 0

      \[\leadsto \frac{m + {m}^{2} \cdot \left(m - 2\right)}{\color{blue}{v}} \]
   7. Step-by-step derivation

   8. Applied rewrites 99.9%

      \[\leadsto \frac{\mathsf{fma}\left(m, m \cdot \left(m + -2\right), m\right)}{\color{blue}{v}} \]
   9. Step-by-step derivation

   10. Applied rewrites 99.9%

       \[\leadsto \color{blue}{\frac{\mathsf{fma}\left(m + -2, m \cdot m, m\right)}{v}} \]
3. Recombined 2 regimes into one program.
4. Add Preprocessing

Alternative 7: 98.5% accurate, 1.0× speedup

Math

\[\begin{array}{l} \\ \begin{array}{l} \mathbf{if}\;m \leq 1.6:\\ \;\;\;\;\left(1 - m\right) \cdot \left(-1 + \frac{m}{v}\right)\\ \mathbf{else}:\\ \;\;\;\;m \cdot \frac{m \cdot \left(m + -2\right)}{v}\\ \end{array} \end{array} \]

FPCore

(FPCore (m v)
 :precision binary64
 (if (<= m 1.6) (* (- 1.0 m) (+ -1.0 (/ m v))) (* m (/ (* m (+ m -2.0)) v))))

C

double code(double m, double v) {
	double tmp;
	if (m <= 1.6) {
		tmp = (1.0 - m) * (-1.0 + (m / v));
	} else {
		tmp = m * ((m * (m + -2.0)) / v);
	}
	return tmp;
}

Fortran

real(8) function code(m, v)
    real(8), intent (in) :: m
    real(8), intent (in) :: v
    real(8) :: tmp
    if (m <= 1.6d0) then
        tmp = (1.0d0 - m) * ((-1.0d0) + (m / v))
    else
        tmp = m * ((m * (m + (-2.0d0))) / v)
    end if
    code = tmp
end function

Java

public static double code(double m, double v) {
	double tmp;
	if (m <= 1.6) {
		tmp = (1.0 - m) * (-1.0 + (m / v));
	} else {
		tmp = m * ((m * (m + -2.0)) / v);
	}
	return tmp;
}

Python

def code(m, v):
	tmp = 0
	if m <= 1.6:
		tmp = (1.0 - m) * (-1.0 + (m / v))
	else:
		tmp = m * ((m * (m + -2.0)) / v)
	return tmp

Julia

function code(m, v)
	tmp = 0.0
	if (m <= 1.6)
		tmp = Float64(Float64(1.0 - m) * Float64(-1.0 + Float64(m / v)));
	else
		tmp = Float64(m * Float64(Float64(m * Float64(m + -2.0)) / v));
	end
	return tmp
end

MATLAB

function tmp_2 = code(m, v)
	tmp = 0.0;
	if (m <= 1.6)
		tmp = (1.0 - m) * (-1.0 + (m / v));
	else
		tmp = m * ((m * (m + -2.0)) / v);
	end
	tmp_2 = tmp;
end

Wolfram

code[m_, v_] := If[LessEqual[m, 1.6], N[(N[(1.0 - m), $MachinePrecision] * N[(-1.0 + N[(m / v), $MachinePrecision]), $MachinePrecision]), $MachinePrecision], N[(m * N[(N[(m * N[(m + -2.0), $MachinePrecision]), $MachinePrecision] / v), $MachinePrecision]), $MachinePrecision]]

Derivation

1. Split input into 2 regimes

2. if m < 1.6000000000000001

   1. Initial program 100.0%

      \[\left(\frac{m \cdot \left(1 - m\right)}{v} - 1\right) \cdot \left(1 - m\right) \]
   2. Add Preprocessing

   3. Taylor expanded in m around 0

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

      1. lower-/.f64 97.1

         \[\leadsto \left(\color{blue}{\frac{m}{v}} - 1\right) \cdot \left(1 - m\right) \]

   5. Applied rewrites 97.1%

      \[\leadsto \left(\color{blue}{\frac{m}{v}} - 1\right) \cdot \left(1 - m\right) \]

   if 1.6000000000000001 < m

   1. Initial program 100.0%

      \[\left(\frac{m \cdot \left(1 - m\right)}{v} - 1\right) \cdot \left(1 - m\right) \]
   2. Add Preprocessing

   3. Taylor expanded in m around 0

      \[\leadsto \color{blue}{m \cdot \left(1 + \left(m \cdot \left(\frac{m}{v} - 2 \cdot \frac{1}{v}\right) + \frac{1}{v}\right)\right) - 1} \]
   4. Step-by-step derivation

      1. associate-+r+ N/A

         \[\leadsto m \cdot \color{blue}{\left(\left(1 + m \cdot \left(\frac{m}{v} - 2 \cdot \frac{1}{v}\right)\right) + \frac{1}{v}\right)} - 1 \]

      2. distribute-lft-in N/A

         \[\leadsto \color{blue}{\left(m \cdot \left(1 + m \cdot \left(\frac{m}{v} - 2 \cdot \frac{1}{v}\right)\right) + m \cdot \frac{1}{v}\right)} - 1 \]

      3. associate-*r/ N/A

         \[\leadsto \left(m \cdot \left(1 + m \cdot \left(\frac{m}{v} - 2 \cdot \frac{1}{v}\right)\right) + \color{blue}{\frac{m \cdot 1}{v}}\right) - 1 \]

      4. *-rgt-identity N/A

         \[\leadsto \left(m \cdot \left(1 + m \cdot \left(\frac{m}{v} - 2 \cdot \frac{1}{v}\right)\right) + \frac{\color{blue}{m}}{v}\right) - 1 \]

      5. associate--l+ N/A

         \[\leadsto \color{blue}{m \cdot \left(1 + m \cdot \left(\frac{m}{v} - 2 \cdot \frac{1}{v}\right)\right) + \left(\frac{m}{v} - 1\right)} \]

      6. *-lft-identity N/A

         \[\leadsto m \cdot \left(1 + m \cdot \left(\frac{m}{v} - 2 \cdot \frac{1}{v}\right)\right) + \left(\frac{\color{blue}{1 \cdot m}}{v} - 1\right) \]

      7. associate-*l/ N/A

         \[\leadsto m \cdot \left(1 + m \cdot \left(\frac{m}{v} - 2 \cdot \frac{1}{v}\right)\right) + \left(\color{blue}{\frac{1}{v} \cdot m} - 1\right) \]

      8. lower-fma.f64 N/A

         \[\leadsto \color{blue}{\mathsf{fma}\left(m, 1 + m \cdot \left(\frac{m}{v} - 2 \cdot \frac{1}{v}\right), \frac{1}{v} \cdot m - 1\right)} \]

   5. Applied rewrites 99.9%

      \[\leadsto \color{blue}{\mathsf{fma}\left(m, \mathsf{fma}\left(\frac{m}{v}, -2 + m, 1\right), -1 + \frac{m}{v}\right)} \]

   6. Taylor expanded in m around inf

      \[\leadsto \color{blue}{{m}^{3} \cdot \left(\frac{1}{v} - 2 \cdot \frac{1}{m \cdot v}\right)} \]

   8. Applied rewrites 98.2%

      \[\leadsto \color{blue}{m \cdot \frac{m \cdot \left(m + -2\right)}{v}} \]
3. Recombined 2 regimes into one program.

4. Final simplification 97.7%

   \[\leadsto \begin{array}{l} \mathbf{if}\;m \leq 1.6:\\ \;\;\;\;\left(1 - m\right) \cdot \left(-1 + \frac{m}{v}\right)\\ \mathbf{else}:\\ \;\;\;\;m \cdot \frac{m \cdot \left(m + -2\right)}{v}\\ \end{array} \]
5. Add Preprocessing

Alternative 8: 98.5% accurate, 1.0× speedup

Math

\[\begin{array}{l} \\ \begin{array}{l} \mathbf{if}\;m \leq 1.6:\\ \;\;\;\;\left(1 - m\right) \cdot \left(-1 + \frac{m}{v}\right)\\ \mathbf{else}:\\ \;\;\;\;m \cdot \left(m \cdot \frac{m + -2}{v}\right)\\ \end{array} \end{array} \]

FPCore

(FPCore (m v)
 :precision binary64
 (if (<= m 1.6) (* (- 1.0 m) (+ -1.0 (/ m v))) (* m (* m (/ (+ m -2.0) v)))))

C

double code(double m, double v) {
	double tmp;
	if (m <= 1.6) {
		tmp = (1.0 - m) * (-1.0 + (m / v));
	} else {
		tmp = m * (m * ((m + -2.0) / v));
	}
	return tmp;
}

Fortran

real(8) function code(m, v)
    real(8), intent (in) :: m
    real(8), intent (in) :: v
    real(8) :: tmp
    if (m <= 1.6d0) then
        tmp = (1.0d0 - m) * ((-1.0d0) + (m / v))
    else
        tmp = m * (m * ((m + (-2.0d0)) / v))
    end if
    code = tmp
end function

Java

public static double code(double m, double v) {
	double tmp;
	if (m <= 1.6) {
		tmp = (1.0 - m) * (-1.0 + (m / v));
	} else {
		tmp = m * (m * ((m + -2.0) / v));
	}
	return tmp;
}

Python

def code(m, v):
	tmp = 0
	if m <= 1.6:
		tmp = (1.0 - m) * (-1.0 + (m / v))
	else:
		tmp = m * (m * ((m + -2.0) / v))
	return tmp

Julia

function code(m, v)
	tmp = 0.0
	if (m <= 1.6)
		tmp = Float64(Float64(1.0 - m) * Float64(-1.0 + Float64(m / v)));
	else
		tmp = Float64(m * Float64(m * Float64(Float64(m + -2.0) / v)));
	end
	return tmp
end

MATLAB

function tmp_2 = code(m, v)
	tmp = 0.0;
	if (m <= 1.6)
		tmp = (1.0 - m) * (-1.0 + (m / v));
	else
		tmp = m * (m * ((m + -2.0) / v));
	end
	tmp_2 = tmp;
end

Wolfram

code[m_, v_] := If[LessEqual[m, 1.6], N[(N[(1.0 - m), $MachinePrecision] * N[(-1.0 + N[(m / v), $MachinePrecision]), $MachinePrecision]), $MachinePrecision], N[(m * N[(m * N[(N[(m + -2.0), $MachinePrecision] / v), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]]

Derivation

1. Split input into 2 regimes

2. if m < 1.6000000000000001

   1. Initial program 100.0%

      \[\left(\frac{m \cdot \left(1 - m\right)}{v} - 1\right) \cdot \left(1 - m\right) \]
   2. Add Preprocessing

   3. Taylor expanded in m around 0

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

      1. lower-/.f64 97.1

         \[\leadsto \left(\color{blue}{\frac{m}{v}} - 1\right) \cdot \left(1 - m\right) \]

   5. Applied rewrites 97.1%

      \[\leadsto \left(\color{blue}{\frac{m}{v}} - 1\right) \cdot \left(1 - m\right) \]

   if 1.6000000000000001 < m

   1. Initial program 100.0%

      \[\left(\frac{m \cdot \left(1 - m\right)}{v} - 1\right) \cdot \left(1 - m\right) \]
   2. Add Preprocessing

   3. Taylor expanded in m around inf

      \[\leadsto \color{blue}{{m}^{3} \cdot \left(\frac{1}{v} - 2 \cdot \frac{1}{m \cdot v}\right)} \]

   5. Applied rewrites 98.2%

      \[\leadsto \color{blue}{m \cdot \left(\frac{m}{v} \cdot \left(-2 + m\right)\right)} \]
   6. Step-by-step derivation

   7. Applied rewrites 98.2%

      \[\leadsto m \cdot \left(m \cdot \color{blue}{\frac{m + -2}{v}}\right) \]
3. Recombined 2 regimes into one program.

4. Final simplification 97.7%

   \[\leadsto \begin{array}{l} \mathbf{if}\;m \leq 1.6:\\ \;\;\;\;\left(1 - m\right) \cdot \left(-1 + \frac{m}{v}\right)\\ \mathbf{else}:\\ \;\;\;\;m \cdot \left(m \cdot \frac{m + -2}{v}\right)\\ \end{array} \]
5. Add Preprocessing

Alternative 9: 99.9% accurate, 1.0× speedup

Math

\[\begin{array}{l} \\ \left(1 - m\right) \cdot \left(-1 + \frac{m \cdot \left(1 - m\right)}{v}\right) \end{array} \]

FPCore

(FPCore (m v) :precision binary64 (* (- 1.0 m) (+ -1.0 (/ (* m (- 1.0 m)) v))))

C

double code(double m, double v) {
	return (1.0 - m) * (-1.0 + ((m * (1.0 - m)) / v));
}

Fortran

real(8) function code(m, v)
    real(8), intent (in) :: m
    real(8), intent (in) :: v
    code = (1.0d0 - m) * ((-1.0d0) + ((m * (1.0d0 - m)) / v))
end function

Java

public static double code(double m, double v) {
	return (1.0 - m) * (-1.0 + ((m * (1.0 - m)) / v));
}

Python

def code(m, v):
	return (1.0 - m) * (-1.0 + ((m * (1.0 - m)) / v))

Julia

function code(m, v)
	return Float64(Float64(1.0 - m) * Float64(-1.0 + Float64(Float64(m * Float64(1.0 - m)) / v)))
end

MATLAB

function tmp = code(m, v)
	tmp = (1.0 - m) * (-1.0 + ((m * (1.0 - m)) / v));
end

Wolfram

code[m_, v_] := N[(N[(1.0 - m), $MachinePrecision] * N[(-1.0 + N[(N[(m * N[(1.0 - m), $MachinePrecision]), $MachinePrecision] / v), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]

Derivation

1. Initial program 100.0%

   \[\left(\frac{m \cdot \left(1 - m\right)}{v} - 1\right) \cdot \left(1 - m\right) \]
2. Add Preprocessing

3. Final simplification 100.0%

   \[\leadsto \left(1 - m\right) \cdot \left(-1 + \frac{m \cdot \left(1 - m\right)}{v}\right) \]
4. Add Preprocessing

Alternative 10: 98.0% accurate, 1.1× speedup

Math

\[\begin{array}{l} \\ \begin{array}{l} \mathbf{if}\;m \leq 0.42:\\ \;\;\;\;\left(1 - m\right) \cdot \left(-1 + \frac{m}{v}\right)\\ \mathbf{else}:\\ \;\;\;\;m \cdot \frac{m \cdot m}{v}\\ \end{array} \end{array} \]

FPCore

(FPCore (m v)
 :precision binary64
 (if (<= m 0.42) (* (- 1.0 m) (+ -1.0 (/ m v))) (* m (/ (* m m) v))))

C

double code(double m, double v) {
	double tmp;
	if (m <= 0.42) {
		tmp = (1.0 - m) * (-1.0 + (m / v));
	} else {
		tmp = m * ((m * m) / v);
	}
	return tmp;
}

Fortran

real(8) function code(m, v)
    real(8), intent (in) :: m
    real(8), intent (in) :: v
    real(8) :: tmp
    if (m <= 0.42d0) then
        tmp = (1.0d0 - m) * ((-1.0d0) + (m / v))
    else
        tmp = m * ((m * m) / v)
    end if
    code = tmp
end function

Java

public static double code(double m, double v) {
	double tmp;
	if (m <= 0.42) {
		tmp = (1.0 - m) * (-1.0 + (m / v));
	} else {
		tmp = m * ((m * m) / v);
	}
	return tmp;
}

Python

def code(m, v):
	tmp = 0
	if m <= 0.42:
		tmp = (1.0 - m) * (-1.0 + (m / v))
	else:
		tmp = m * ((m * m) / v)
	return tmp

Julia

function code(m, v)
	tmp = 0.0
	if (m <= 0.42)
		tmp = Float64(Float64(1.0 - m) * Float64(-1.0 + Float64(m / v)));
	else
		tmp = Float64(m * Float64(Float64(m * m) / v));
	end
	return tmp
end

MATLAB

function tmp_2 = code(m, v)
	tmp = 0.0;
	if (m <= 0.42)
		tmp = (1.0 - m) * (-1.0 + (m / v));
	else
		tmp = m * ((m * m) / v);
	end
	tmp_2 = tmp;
end

Wolfram

code[m_, v_] := If[LessEqual[m, 0.42], N[(N[(1.0 - m), $MachinePrecision] * N[(-1.0 + N[(m / v), $MachinePrecision]), $MachinePrecision]), $MachinePrecision], N[(m * N[(N[(m * m), $MachinePrecision] / v), $MachinePrecision]), $MachinePrecision]]

Derivation

1. Split input into 2 regimes

2. if m < 0.419999999999999984

   1. Initial program 100.0%

      \[\left(\frac{m \cdot \left(1 - m\right)}{v} - 1\right) \cdot \left(1 - m\right) \]
   2. Add Preprocessing

   3. Taylor expanded in m around 0

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

      1. lower-/.f64 97.1

         \[\leadsto \left(\color{blue}{\frac{m}{v}} - 1\right) \cdot \left(1 - m\right) \]

   5. Applied rewrites 97.1%

      \[\leadsto \left(\color{blue}{\frac{m}{v}} - 1\right) \cdot \left(1 - m\right) \]

   if 0.419999999999999984 < m

   1. Initial program 100.0%

      \[\left(\frac{m \cdot \left(1 - m\right)}{v} - 1\right) \cdot \left(1 - m\right) \]
   2. Add Preprocessing

   3. Taylor expanded in m around inf

      \[\leadsto \color{blue}{{m}^{3} \cdot \left(\frac{1}{v} - 2 \cdot \frac{1}{m \cdot v}\right)} \]

   5. Applied rewrites 98.2%

      \[\leadsto \color{blue}{m \cdot \left(\frac{m}{v} \cdot \left(-2 + m\right)\right)} \]

   6. Taylor expanded in m around inf

      \[\leadsto m \cdot \frac{{m}^{2}}{\color{blue}{v}} \]
   7. Step-by-step derivation

   8. Applied rewrites 96.3%

      \[\leadsto m \cdot \frac{m \cdot m}{\color{blue}{v}} \]
3. Recombined 2 regimes into one program.

4. Final simplification 96.7%

   \[\leadsto \begin{array}{l} \mathbf{if}\;m \leq 0.42:\\ \;\;\;\;\left(1 - m\right) \cdot \left(-1 + \frac{m}{v}\right)\\ \mathbf{else}:\\ \;\;\;\;m \cdot \frac{m \cdot m}{v}\\ \end{array} \]
5. Add Preprocessing

Alternative 11: 97.9% accurate, 1.1× speedup

Math

\[\begin{array}{l} \\ \begin{array}{l} \mathbf{if}\;m \leq 2.6:\\ \;\;\;\;-1 + \frac{m}{v}\\ \mathbf{else}:\\ \;\;\;\;m \cdot \frac{m \cdot m}{v}\\ \end{array} \end{array} \]

FPCore

(FPCore (m v)
 :precision binary64
 (if (<= m 2.6) (+ -1.0 (/ m v)) (* m (/ (* m m) v))))

C

double code(double m, double v) {
	double tmp;
	if (m <= 2.6) {
		tmp = -1.0 + (m / v);
	} else {
		tmp = m * ((m * m) / v);
	}
	return tmp;
}

Fortran

real(8) function code(m, v)
    real(8), intent (in) :: m
    real(8), intent (in) :: v
    real(8) :: tmp
    if (m <= 2.6d0) then
        tmp = (-1.0d0) + (m / v)
    else
        tmp = m * ((m * m) / v)
    end if
    code = tmp
end function

Java

public static double code(double m, double v) {
	double tmp;
	if (m <= 2.6) {
		tmp = -1.0 + (m / v);
	} else {
		tmp = m * ((m * m) / v);
	}
	return tmp;
}

Python

def code(m, v):
	tmp = 0
	if m <= 2.6:
		tmp = -1.0 + (m / v)
	else:
		tmp = m * ((m * m) / v)
	return tmp

Julia

function code(m, v)
	tmp = 0.0
	if (m <= 2.6)
		tmp = Float64(-1.0 + Float64(m / v));
	else
		tmp = Float64(m * Float64(Float64(m * m) / v));
	end
	return tmp
end

MATLAB

function tmp_2 = code(m, v)
	tmp = 0.0;
	if (m <= 2.6)
		tmp = -1.0 + (m / v);
	else
		tmp = m * ((m * m) / v);
	end
	tmp_2 = tmp;
end

Wolfram

code[m_, v_] := If[LessEqual[m, 2.6], N[(-1.0 + N[(m / v), $MachinePrecision]), $MachinePrecision], N[(m * N[(N[(m * m), $MachinePrecision] / v), $MachinePrecision]), $MachinePrecision]]

Derivation

1. Split input into 2 regimes

2. if m < 2.60000000000000009

   1. Initial program 100.0%

      \[\left(\frac{m \cdot \left(1 - m\right)}{v} - 1\right) \cdot \left(1 - m\right) \]
   2. Add Preprocessing

   3. Taylor expanded in m around 0

      \[\leadsto \color{blue}{m \cdot \left(1 + \frac{1}{v}\right) - 1} \]
   4. Step-by-step derivation

      1. sub-neg N/A

         \[\leadsto \color{blue}{m \cdot \left(1 + \frac{1}{v}\right) + \left(\mathsf{neg}\left(1\right)\right)} \]

      2. metadata-eval N/A

         \[\leadsto m \cdot \left(1 + \frac{1}{v}\right) + \color{blue}{-1} \]

      3. +-commutative N/A

         \[\leadsto \color{blue}{-1 + m \cdot \left(1 + \frac{1}{v}\right)} \]

      4. lower-+.f64 N/A

         \[\leadsto \color{blue}{-1 + m \cdot \left(1 + \frac{1}{v}\right)} \]

      5. distribute-rgt-in N/A

         \[\leadsto -1 + \color{blue}{\left(1 \cdot m + \frac{1}{v} \cdot m\right)} \]

      6. associate-*l/ N/A

         \[\leadsto -1 + \left(1 \cdot m + \color{blue}{\frac{1 \cdot m}{v}}\right) \]

      7. *-lft-identity N/A

         \[\leadsto -1 + \left(1 \cdot m + \frac{\color{blue}{m}}{v}\right) \]

      8. *-lft-identity N/A

         \[\leadsto -1 + \left(\color{blue}{m} + \frac{m}{v}\right) \]

      9. lower-+.f64 N/A

         \[\leadsto -1 + \color{blue}{\left(m + \frac{m}{v}\right)} \]

      10. lower-/.f64 97.0

          \[\leadsto -1 + \left(m + \color{blue}{\frac{m}{v}}\right) \]

   5. Applied rewrites 97.0%

      \[\leadsto \color{blue}{-1 + \left(m + \frac{m}{v}\right)} \]

   6. Taylor expanded in v around 0

      \[\leadsto -1 + \frac{m}{\color{blue}{v}} \]
   7. Step-by-step derivation

   8. Applied rewrites 97.0%

      \[\leadsto -1 + \frac{m}{\color{blue}{v}} \]

   if 2.60000000000000009 < m

   1. Initial program 100.0%

      \[\left(\frac{m \cdot \left(1 - m\right)}{v} - 1\right) \cdot \left(1 - m\right) \]
   2. Add Preprocessing

   3. Taylor expanded in m around inf

      \[\leadsto \color{blue}{{m}^{3} \cdot \left(\frac{1}{v} - 2 \cdot \frac{1}{m \cdot v}\right)} \]

   5. Applied rewrites 98.2%

      \[\leadsto \color{blue}{m \cdot \left(\frac{m}{v} \cdot \left(-2 + m\right)\right)} \]

   6. Taylor expanded in m around inf

      \[\leadsto m \cdot \frac{{m}^{2}}{\color{blue}{v}} \]
   7. Step-by-step derivation

   8. Applied rewrites 96.3%

      \[\leadsto m \cdot \frac{m \cdot m}{\color{blue}{v}} \]
3. Recombined 2 regimes into one program.
4. Add Preprocessing

Alternative 12: 81.3% accurate, 1.3× speedup

Math

\[\begin{array}{l} \\ \begin{array}{l} \mathbf{if}\;m \leq 1.32 \cdot 10^{+154}:\\ \;\;\;\;-1 + \left(m + \frac{m}{v}\right)\\ \mathbf{else}:\\ \;\;\;\;\frac{\mathsf{fma}\left(m, m, -1\right)}{1}\\ \end{array} \end{array} \]

FPCore

(FPCore (m v)
 :precision binary64
 (if (<= m 1.32e+154) (+ -1.0 (+ m (/ m v))) (/ (fma m m -1.0) 1.0)))

C

double code(double m, double v) {
	double tmp;
	if (m <= 1.32e+154) {
		tmp = -1.0 + (m + (m / v));
	} else {
		tmp = fma(m, m, -1.0) / 1.0;
	}
	return tmp;
}

Julia

function code(m, v)
	tmp = 0.0
	if (m <= 1.32e+154)
		tmp = Float64(-1.0 + Float64(m + Float64(m / v)));
	else
		tmp = Float64(fma(m, m, -1.0) / 1.0);
	end
	return tmp
end

Wolfram

code[m_, v_] := If[LessEqual[m, 1.32e+154], N[(-1.0 + N[(m + N[(m / v), $MachinePrecision]), $MachinePrecision]), $MachinePrecision], N[(N[(m * m + -1.0), $MachinePrecision] / 1.0), $MachinePrecision]]

Derivation

1. Split input into 2 regimes

2. if m < 1.31999999999999998e154

   1. Initial program 99.9%

      \[\left(\frac{m \cdot \left(1 - m\right)}{v} - 1\right) \cdot \left(1 - m\right) \]
   2. Add Preprocessing

   3. Taylor expanded in m around 0

      \[\leadsto \color{blue}{m \cdot \left(1 + \frac{1}{v}\right) - 1} \]
   4. Step-by-step derivation

      1. sub-neg N/A

         \[\leadsto \color{blue}{m \cdot \left(1 + \frac{1}{v}\right) + \left(\mathsf{neg}\left(1\right)\right)} \]

      2. metadata-eval N/A

         \[\leadsto m \cdot \left(1 + \frac{1}{v}\right) + \color{blue}{-1} \]

      3. +-commutative N/A

         \[\leadsto \color{blue}{-1 + m \cdot \left(1 + \frac{1}{v}\right)} \]

      4. lower-+.f64 N/A

         \[\leadsto \color{blue}{-1 + m \cdot \left(1 + \frac{1}{v}\right)} \]

      5. distribute-rgt-in N/A

         \[\leadsto -1 + \color{blue}{\left(1 \cdot m + \frac{1}{v} \cdot m\right)} \]

      6. associate-*l/ N/A

         \[\leadsto -1 + \left(1 \cdot m + \color{blue}{\frac{1 \cdot m}{v}}\right) \]

      7. *-lft-identity N/A

         \[\leadsto -1 + \left(1 \cdot m + \frac{\color{blue}{m}}{v}\right) \]

      8. *-lft-identity N/A

         \[\leadsto -1 + \left(\color{blue}{m} + \frac{m}{v}\right) \]

      9. lower-+.f64 N/A

         \[\leadsto -1 + \color{blue}{\left(m + \frac{m}{v}\right)} \]

      10. lower-/.f64 76.3

          \[\leadsto -1 + \left(m + \color{blue}{\frac{m}{v}}\right) \]

   5. Applied rewrites 76.3%

      \[\leadsto \color{blue}{-1 + \left(m + \frac{m}{v}\right)} \]

   if 1.31999999999999998e154 < m

   1. Initial program 100.0%

      \[\left(\frac{m \cdot \left(1 - m\right)}{v} - 1\right) \cdot \left(1 - m\right) \]
   2. Add Preprocessing

   3. Taylor expanded in v around inf

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

      1. mul-1-neg N/A

         \[\leadsto \color{blue}{\mathsf{neg}\left(\left(1 - m\right)\right)} \]

      2. neg-sub0 N/A

         \[\leadsto \color{blue}{0 - \left(1 - m\right)} \]

      3. associate--r- N/A

         \[\leadsto \color{blue}{\left(0 - 1\right) + m} \]

      4. metadata-eval N/A

         \[\leadsto \color{blue}{-1} + m \]

      5. lower-+.f64 6.7

         \[\leadsto \color{blue}{-1 + m} \]

   5. Applied rewrites 6.7%

      \[\leadsto \color{blue}{-1 + m} \]
   6. Step-by-step derivation

   7. Applied rewrites 100.0%

      \[\leadsto \frac{\mathsf{fma}\left(m, m, -1\right)}{\color{blue}{m + 1}} \]

   8. Taylor expanded in m around 0

      \[\leadsto \frac{\mathsf{fma}\left(m, m, -1\right)}{1} \]
   9. Step-by-step derivation

   10. Applied rewrites 100.0%

       \[\leadsto \frac{\mathsf{fma}\left(m, m, -1\right)}{1} \]
3. Recombined 2 regimes into one program.
4. Add Preprocessing

Alternative 13: 75.5% accurate, 1.7× speedup

Math

\[\begin{array}{l} \\ -1 + \left(m + \frac{m}{v}\right) \end{array} \]

FPCore

(FPCore (m v) :precision binary64 (+ -1.0 (+ m (/ m v))))

C

double code(double m, double v) {
	return -1.0 + (m + (m / v));
}

Fortran

real(8) function code(m, v)
    real(8), intent (in) :: m
    real(8), intent (in) :: v
    code = (-1.0d0) + (m + (m / v))
end function

Java

public static double code(double m, double v) {
	return -1.0 + (m + (m / v));
}

Python

def code(m, v):
	return -1.0 + (m + (m / v))

Julia

function code(m, v)
	return Float64(-1.0 + Float64(m + Float64(m / v)))
end

MATLAB

function tmp = code(m, v)
	tmp = -1.0 + (m + (m / v));
end

Wolfram

code[m_, v_] := N[(-1.0 + N[(m + N[(m / v), $MachinePrecision]), $MachinePrecision]), $MachinePrecision]

Derivation

1. Initial program 100.0%

   \[\left(\frac{m \cdot \left(1 - m\right)}{v} - 1\right) \cdot \left(1 - m\right) \]
2. Add Preprocessing

3. Taylor expanded in m around 0

   \[\leadsto \color{blue}{m \cdot \left(1 + \frac{1}{v}\right) - 1} \]
4. Step-by-step derivation

   1. sub-neg N/A

      \[\leadsto \color{blue}{m \cdot \left(1 + \frac{1}{v}\right) + \left(\mathsf{neg}\left(1\right)\right)} \]

   2. metadata-eval N/A

      \[\leadsto m \cdot \left(1 + \frac{1}{v}\right) + \color{blue}{-1} \]

   3. +-commutative N/A

      \[\leadsto \color{blue}{-1 + m \cdot \left(1 + \frac{1}{v}\right)} \]

   4. lower-+.f64 N/A

      \[\leadsto \color{blue}{-1 + m \cdot \left(1 + \frac{1}{v}\right)} \]

   5. distribute-rgt-in N/A

      \[\leadsto -1 + \color{blue}{\left(1 \cdot m + \frac{1}{v} \cdot m\right)} \]

   6. associate-*l/ N/A

      \[\leadsto -1 + \left(1 \cdot m + \color{blue}{\frac{1 \cdot m}{v}}\right) \]

   7. *-lft-identity N/A

      \[\leadsto -1 + \left(1 \cdot m + \frac{\color{blue}{m}}{v}\right) \]

   8. *-lft-identity N/A

      \[\leadsto -1 + \left(\color{blue}{m} + \frac{m}{v}\right) \]

   9. lower-+.f64 N/A

      \[\leadsto -1 + \color{blue}{\left(m + \frac{m}{v}\right)} \]

   10. lower-/.f64 76.2

       \[\leadsto -1 + \left(m + \color{blue}{\frac{m}{v}}\right) \]

5. Applied rewrites 76.2%

   \[\leadsto \color{blue}{-1 + \left(m + \frac{m}{v}\right)} \]
6. Add Preprocessing

Alternative 14: 75.5% accurate, 2.1× speedup

Math

\[\begin{array}{l} \\ -1 + \frac{m}{v} \end{array} \]

FPCore

(FPCore (m v) :precision binary64 (+ -1.0 (/ m v)))

C

double code(double m, double v) {
	return -1.0 + (m / v);
}

Fortran

real(8) function code(m, v)
    real(8), intent (in) :: m
    real(8), intent (in) :: v
    code = (-1.0d0) + (m / v)
end function

Java

public static double code(double m, double v) {
	return -1.0 + (m / v);
}

Python

def code(m, v):
	return -1.0 + (m / v)

Julia

function code(m, v)
	return Float64(-1.0 + Float64(m / v))
end

MATLAB

function tmp = code(m, v)
	tmp = -1.0 + (m / v);
end

Wolfram

code[m_, v_] := N[(-1.0 + N[(m / v), $MachinePrecision]), $MachinePrecision]

Derivation

1. Initial program 100.0%

   \[\left(\frac{m \cdot \left(1 - m\right)}{v} - 1\right) \cdot \left(1 - m\right) \]
2. Add Preprocessing

3. Taylor expanded in m around 0

   \[\leadsto \color{blue}{m \cdot \left(1 + \frac{1}{v}\right) - 1} \]
4. Step-by-step derivation

   1. sub-neg N/A

      \[\leadsto \color{blue}{m \cdot \left(1 + \frac{1}{v}\right) + \left(\mathsf{neg}\left(1\right)\right)} \]

   2. metadata-eval N/A

      \[\leadsto m \cdot \left(1 + \frac{1}{v}\right) + \color{blue}{-1} \]

   3. +-commutative N/A

      \[\leadsto \color{blue}{-1 + m \cdot \left(1 + \frac{1}{v}\right)} \]

   4. lower-+.f64 N/A

      \[\leadsto \color{blue}{-1 + m \cdot \left(1 + \frac{1}{v}\right)} \]

   5. distribute-rgt-in N/A

      \[\leadsto -1 + \color{blue}{\left(1 \cdot m + \frac{1}{v} \cdot m\right)} \]

   6. associate-*l/ N/A

      \[\leadsto -1 + \left(1 \cdot m + \color{blue}{\frac{1 \cdot m}{v}}\right) \]

   7. *-lft-identity N/A

      \[\leadsto -1 + \left(1 \cdot m + \frac{\color{blue}{m}}{v}\right) \]

   8. *-lft-identity N/A

      \[\leadsto -1 + \left(\color{blue}{m} + \frac{m}{v}\right) \]

   9. lower-+.f64 N/A

      \[\leadsto -1 + \color{blue}{\left(m + \frac{m}{v}\right)} \]

   10. lower-/.f64 76.2

       \[\leadsto -1 + \left(m + \color{blue}{\frac{m}{v}}\right) \]

5. Applied rewrites 76.2%

   \[\leadsto \color{blue}{-1 + \left(m + \frac{m}{v}\right)} \]

6. Taylor expanded in v around 0

   \[\leadsto -1 + \frac{m}{\color{blue}{v}} \]
7. Step-by-step derivation

8. Applied rewrites 76.2%

   \[\leadsto -1 + \frac{m}{\color{blue}{v}} \]
9. Add Preprocessing

Alternative 15: 27.1% accurate, 7.8× speedup

Math

\[\begin{array}{l} \\ m + -1 \end{array} \]

FPCore

(FPCore (m v) :precision binary64 (+ m -1.0))

C

double code(double m, double v) {
	return m + -1.0;
}

Fortran

real(8) function code(m, v)
    real(8), intent (in) :: m
    real(8), intent (in) :: v
    code = m + (-1.0d0)
end function

Java

public static double code(double m, double v) {
	return m + -1.0;
}

Python

def code(m, v):
	return m + -1.0

Julia

function code(m, v)
	return Float64(m + -1.0)
end

MATLAB

function tmp = code(m, v)
	tmp = m + -1.0;
end

Wolfram

code[m_, v_] := N[(m + -1.0), $MachinePrecision]

Derivation

1. Initial program 100.0%

   \[\left(\frac{m \cdot \left(1 - m\right)}{v} - 1\right) \cdot \left(1 - m\right) \]
2. Add Preprocessing

3. Taylor expanded in v around inf

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

   1. mul-1-neg N/A

      \[\leadsto \color{blue}{\mathsf{neg}\left(\left(1 - m\right)\right)} \]

   2. neg-sub0 N/A

      \[\leadsto \color{blue}{0 - \left(1 - m\right)} \]

   3. associate--r- N/A

      \[\leadsto \color{blue}{\left(0 - 1\right) + m} \]

   4. metadata-eval N/A

      \[\leadsto \color{blue}{-1} + m \]

   5. lower-+.f64 26.3

      \[\leadsto \color{blue}{-1 + m} \]

5. Applied rewrites 26.3%

   \[\leadsto \color{blue}{-1 + m} \]

6. Final simplification 26.3%

   \[\leadsto m + -1 \]
7. Add Preprocessing

Alternative 16: 24.6% accurate, 31.0× speedup

Math

\[\begin{array}{l} \\ -1 \end{array} \]

FPCore

(FPCore (m v) :precision binary64 -1.0)

C

double code(double m, double v) {
	return -1.0;
}

Fortran

real(8) function code(m, v)
    real(8), intent (in) :: m
    real(8), intent (in) :: v
    code = -1.0d0
end function

Java

public static double code(double m, double v) {
	return -1.0;
}

Python

def code(m, v):
	return -1.0

Julia

function code(m, v)
	return -1.0
end

MATLAB

function tmp = code(m, v)
	tmp = -1.0;
end

Wolfram

code[m_, v_] := -1.0

Derivation

1. Initial program 100.0%

   \[\left(\frac{m \cdot \left(1 - m\right)}{v} - 1\right) \cdot \left(1 - m\right) \]
2. Add Preprocessing

3. Taylor expanded in m around 0

   \[\leadsto \color{blue}{-1} \]
4. Step-by-step derivation

5. Applied rewrites 24.1%

   \[\leadsto \color{blue}{-1} \]
6. Add Preprocessing

Reproduce

herbie shell --seed 2024226 
(FPCore (m v)
  :name "b parameter of renormalized beta distribution"
  :precision binary64
  :pre (and (and (< 0.0 m) (< 0.0 v)) (< v 0.25))
  (* (- (/ (* m (- 1.0 m)) v) 1.0) (- 1.0 m)))