Beckmann Distribution sample, tan2theta, alphax == alphay

Percentage Accurate: 55.4% → 99.0%

Time: 3.2s

Alternatives: 8

Speedup: 2.3×

Specification

\[\left(0.0001 \leq \alpha \land \alpha \leq 1\right) \land \left(2.328306437 \cdot 10^{-10} \leq u0 \land u0 \leq 1\right)\]

Math

\[\begin{array}{l} \\ \left(\left(-\alpha\right) \cdot \alpha\right) \cdot \log \left(1 - u0\right) \end{array} \]

FPCore

(FPCore (alpha u0)
 :precision binary32
 (* (* (- alpha) alpha) (log (- 1.0 u0))))

C

float code(float alpha, float u0) {
	return (-alpha * alpha) * logf((1.0f - u0));
}

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(4) function code(alpha, u0)
use fmin_fmax_functions
    real(4), intent (in) :: alpha
    real(4), intent (in) :: u0
    code = (-alpha * alpha) * log((1.0e0 - u0))
end function

Julia

function code(alpha, u0)
	return Float32(Float32(Float32(-alpha) * alpha) * log(Float32(Float32(1.0) - u0)))
end

MATLAB

function tmp = code(alpha, u0)
	tmp = (-alpha * alpha) * log((single(1.0) - u0));
end

Initial Program: 55.4% accurate, 1.0× speedup

Math

\[\begin{array}{l} \\ \left(\left(-\alpha\right) \cdot \alpha\right) \cdot \log \left(1 - u0\right) \end{array} \]

FPCore

(FPCore (alpha u0)
 :precision binary32
 (* (* (- alpha) alpha) (log (- 1.0 u0))))

C

float code(float alpha, float u0) {
	return (-alpha * alpha) * logf((1.0f - u0));
}

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(4) function code(alpha, u0)
use fmin_fmax_functions
    real(4), intent (in) :: alpha
    real(4), intent (in) :: u0
    code = (-alpha * alpha) * log((1.0e0 - u0))
end function

Julia

function code(alpha, u0)
	return Float32(Float32(Float32(-alpha) * alpha) * log(Float32(Float32(1.0) - u0)))
end

MATLAB

function tmp = code(alpha, u0)
	tmp = (-alpha * alpha) * log((single(1.0) - u0));
end

Alternative 1: 99.0% accurate, 0.9× speedup

Math

\[\begin{array}{l} \\ \left(\left(-\alpha\right) \cdot \alpha\right) \cdot \mathsf{log1p}\left(-u0\right) \end{array} \]

FPCore

(FPCore (alpha u0) :precision binary32 (* (* (- alpha) alpha) (log1p (- u0))))

C

float code(float alpha, float u0) {
	return (-alpha * alpha) * log1pf(-u0);
}

Julia

function code(alpha, u0)
	return Float32(Float32(Float32(-alpha) * alpha) * log1p(Float32(-u0)))
end

Derivation

1. Initial program 55.4%

   \[\left(\left(-\alpha\right) \cdot \alpha\right) \cdot \log \left(1 - u0\right) \]
2. Step-by-step derivation

   1. lift--.f32 N/A

      \[\leadsto \left(\left(-\alpha\right) \cdot \alpha\right) \cdot \log \color{blue}{\left(1 - u0\right)} \]

   2. lift-log.f32 N/A

      \[\leadsto \left(\left(-\alpha\right) \cdot \alpha\right) \cdot \color{blue}{\log \left(1 - u0\right)} \]

   3. *-lft-identity N/A

      \[\leadsto \left(\left(-\alpha\right) \cdot \alpha\right) \cdot \log \left(1 - \color{blue}{1 \cdot u0}\right) \]

   4. metadata-eval N/A

      \[\leadsto \left(\left(-\alpha\right) \cdot \alpha\right) \cdot \log \left(1 - \color{blue}{\left(\mathsf{neg}\left(-1\right)\right)} \cdot u0\right) \]

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

      \[\leadsto \left(\left(-\alpha\right) \cdot \alpha\right) \cdot \log \color{blue}{\left(1 + -1 \cdot u0\right)} \]

   6. mul-1-neg N/A

      \[\leadsto \left(\left(-\alpha\right) \cdot \alpha\right) \cdot \log \left(1 + \color{blue}{\left(\mathsf{neg}\left(u0\right)\right)}\right) \]

   7. lower-log1p.f32 N/A

      \[\leadsto \left(\left(-\alpha\right) \cdot \alpha\right) \cdot \color{blue}{\mathsf{log1p}\left(\mathsf{neg}\left(u0\right)\right)} \]

   8. lower-neg.f32 99.0

      \[\leadsto \left(\left(-\alpha\right) \cdot \alpha\right) \cdot \mathsf{log1p}\left(\color{blue}{-u0}\right) \]

3. Applied rewrites 99.0%

   \[\leadsto \left(\left(-\alpha\right) \cdot \alpha\right) \cdot \color{blue}{\mathsf{log1p}\left(-u0\right)} \]
4. Add Preprocessing

Alternative 2: 96.8% accurate, 0.5× speedup

Math

\[\begin{array}{l} \\ \begin{array}{l} t_0 := \log \left(1 - u0\right)\\ \mathbf{if}\;t\_0 \leq -0.004999999888241291:\\ \;\;\;\;\left(\left(-\alpha\right) \cdot \alpha\right) \cdot t\_0\\ \mathbf{else}:\\ \;\;\;\;\mathsf{fma}\left(\alpha, \alpha, \left(\left(\alpha \cdot \alpha\right) \cdot u0\right) \cdot 0.5\right) \cdot u0\\ \end{array} \end{array} \]

FPCore

(FPCore (alpha u0)
 :precision binary32
 (let* ((t_0 (log (- 1.0 u0))))
   (if (<= t_0 -0.004999999888241291)
     (* (* (- alpha) alpha) t_0)
     (* (fma alpha alpha (* (* (* alpha alpha) u0) 0.5)) u0))))

C

float code(float alpha, float u0) {
	float t_0 = logf((1.0f - u0));
	float tmp;
	if (t_0 <= -0.004999999888241291f) {
		tmp = (-alpha * alpha) * t_0;
	} else {
		tmp = fmaf(alpha, alpha, (((alpha * alpha) * u0) * 0.5f)) * u0;
	}
	return tmp;
}

Julia

function code(alpha, u0)
	t_0 = log(Float32(Float32(1.0) - u0))
	tmp = Float32(0.0)
	if (t_0 <= Float32(-0.004999999888241291))
		tmp = Float32(Float32(Float32(-alpha) * alpha) * t_0);
	else
		tmp = Float32(fma(alpha, alpha, Float32(Float32(Float32(alpha * alpha) * u0) * Float32(0.5))) * u0);
	end
	return tmp
end

Derivation

1. Split input into 2 regimes

2. if (log.f32 (-.f32 #s(literal 1 binary32) u0)) < -0.00499999989

   1. Initial program 55.4%

      \[\left(\left(-\alpha\right) \cdot \alpha\right) \cdot \log \left(1 - u0\right) \]

   if -0.00499999989 < (log.f32 (-.f32 #s(literal 1 binary32) u0))

   1. Initial program 55.4%

      \[\left(\left(-\alpha\right) \cdot \alpha\right) \cdot \log \left(1 - u0\right) \]

   2. Taylor expanded in u0 around 0

      \[\leadsto \color{blue}{u0 \cdot \left(\frac{1}{2} \cdot \left({\alpha}^{2} \cdot u0\right) + {\alpha}^{2}\right)} \]
   3. Step-by-step derivation

      1. *-commutative N/A

         \[\leadsto \left(\frac{1}{2} \cdot \left({\alpha}^{2} \cdot u0\right) + {\alpha}^{2}\right) \cdot \color{blue}{u0} \]

      2. lower-*.f32 N/A

         \[\leadsto \left(\frac{1}{2} \cdot \left({\alpha}^{2} \cdot u0\right) + {\alpha}^{2}\right) \cdot \color{blue}{u0} \]

      3. *-commutative N/A

         \[\leadsto \left(\left({\alpha}^{2} \cdot u0\right) \cdot \frac{1}{2} + {\alpha}^{2}\right) \cdot u0 \]

      4. lower-fma.f32 N/A

         \[\leadsto \mathsf{fma}\left({\alpha}^{2} \cdot u0, \frac{1}{2}, {\alpha}^{2}\right) \cdot u0 \]

      5. lower-*.f32 N/A

         \[\leadsto \mathsf{fma}\left({\alpha}^{2} \cdot u0, \frac{1}{2}, {\alpha}^{2}\right) \cdot u0 \]

      6. unpow2 N/A

         \[\leadsto \mathsf{fma}\left(\left(\alpha \cdot \alpha\right) \cdot u0, \frac{1}{2}, {\alpha}^{2}\right) \cdot u0 \]

      7. lower-*.f32 N/A

         \[\leadsto \mathsf{fma}\left(\left(\alpha \cdot \alpha\right) \cdot u0, \frac{1}{2}, {\alpha}^{2}\right) \cdot u0 \]

      8. unpow2 N/A

         \[\leadsto \mathsf{fma}\left(\left(\alpha \cdot \alpha\right) \cdot u0, \frac{1}{2}, \alpha \cdot \alpha\right) \cdot u0 \]

      9. lower-*.f32 87.8

         \[\leadsto \mathsf{fma}\left(\left(\alpha \cdot \alpha\right) \cdot u0, 0.5, \alpha \cdot \alpha\right) \cdot u0 \]

   4. Applied rewrites 87.8%

      \[\leadsto \color{blue}{\mathsf{fma}\left(\left(\alpha \cdot \alpha\right) \cdot u0, 0.5, \alpha \cdot \alpha\right) \cdot u0} \]
   5. Step-by-step derivation

      1. lift-*.f32 N/A

         \[\leadsto \mathsf{fma}\left(\left(\alpha \cdot \alpha\right) \cdot u0, \frac{1}{2}, \alpha \cdot \alpha\right) \cdot u0 \]

      2. lift-*.f32 N/A

         \[\leadsto \mathsf{fma}\left(\left(\alpha \cdot \alpha\right) \cdot u0, \frac{1}{2}, \alpha \cdot \alpha\right) \cdot u0 \]

      3. pow2 N/A

         \[\leadsto \mathsf{fma}\left({\alpha}^{2} \cdot u0, \frac{1}{2}, \alpha \cdot \alpha\right) \cdot u0 \]

      4. lift-*.f32 N/A

         \[\leadsto \mathsf{fma}\left({\alpha}^{2} \cdot u0, \frac{1}{2}, \alpha \cdot \alpha\right) \cdot u0 \]

      5. pow2 N/A

         \[\leadsto \mathsf{fma}\left({\alpha}^{2} \cdot u0, \frac{1}{2}, {\alpha}^{2}\right) \cdot u0 \]

      6. lower-fma.f32 N/A

         \[\leadsto \left(\left({\alpha}^{2} \cdot u0\right) \cdot \frac{1}{2} + {\alpha}^{2}\right) \cdot u0 \]

      7. *-commutative N/A

         \[\leadsto \left(\frac{1}{2} \cdot \left({\alpha}^{2} \cdot u0\right) + {\alpha}^{2}\right) \cdot u0 \]

      8. +-commutative N/A

         \[\leadsto \left({\alpha}^{2} + \frac{1}{2} \cdot \left({\alpha}^{2} \cdot u0\right)\right) \cdot u0 \]

      9. pow2 N/A

         \[\leadsto \left(\alpha \cdot \alpha + \frac{1}{2} \cdot \left({\alpha}^{2} \cdot u0\right)\right) \cdot u0 \]

      10. lower-fma.f32 N/A

          \[\leadsto \mathsf{fma}\left(\alpha, \alpha, \frac{1}{2} \cdot \left({\alpha}^{2} \cdot u0\right)\right) \cdot u0 \]

      11. *-commutative N/A

          \[\leadsto \mathsf{fma}\left(\alpha, \alpha, \left({\alpha}^{2} \cdot u0\right) \cdot \frac{1}{2}\right) \cdot u0 \]

      12. lower-*.f32 N/A

          \[\leadsto \mathsf{fma}\left(\alpha, \alpha, \left({\alpha}^{2} \cdot u0\right) \cdot \frac{1}{2}\right) \cdot u0 \]

      13. pow2 N/A

          \[\leadsto \mathsf{fma}\left(\alpha, \alpha, \left(\left(\alpha \cdot \alpha\right) \cdot u0\right) \cdot \frac{1}{2}\right) \cdot u0 \]

      14. lift-*.f32 N/A

          \[\leadsto \mathsf{fma}\left(\alpha, \alpha, \left(\left(\alpha \cdot \alpha\right) \cdot u0\right) \cdot \frac{1}{2}\right) \cdot u0 \]

      15. lift-*.f32 87.9

          \[\leadsto \mathsf{fma}\left(\alpha, \alpha, \left(\left(\alpha \cdot \alpha\right) \cdot u0\right) \cdot 0.5\right) \cdot u0 \]

   6. Applied rewrites 87.9%

      \[\leadsto \mathsf{fma}\left(\alpha, \alpha, \left(\left(\alpha \cdot \alpha\right) \cdot u0\right) \cdot 0.5\right) \cdot u0 \]
3. Recombined 2 regimes into one program.
4. Add Preprocessing

Alternative 3: 96.5% accurate, 0.6× speedup

Math

\[\begin{array}{l} \\ \begin{array}{l} t_0 := \log \left(1 - u0\right)\\ \mathbf{if}\;t\_0 \leq -0.004999999888241291:\\ \;\;\;\;\left(\left(-\alpha\right) \cdot \alpha\right) \cdot t\_0\\ \mathbf{else}:\\ \;\;\;\;\left(\mathsf{fma}\left(0.5, u0, 1\right) \cdot \left(\alpha \cdot \alpha\right)\right) \cdot u0\\ \end{array} \end{array} \]

FPCore

(FPCore (alpha u0)
 :precision binary32
 (let* ((t_0 (log (- 1.0 u0))))
   (if (<= t_0 -0.004999999888241291)
     (* (* (- alpha) alpha) t_0)
     (* (* (fma 0.5 u0 1.0) (* alpha alpha)) u0))))

C

float code(float alpha, float u0) {
	float t_0 = logf((1.0f - u0));
	float tmp;
	if (t_0 <= -0.004999999888241291f) {
		tmp = (-alpha * alpha) * t_0;
	} else {
		tmp = (fmaf(0.5f, u0, 1.0f) * (alpha * alpha)) * u0;
	}
	return tmp;
}

Julia

function code(alpha, u0)
	t_0 = log(Float32(Float32(1.0) - u0))
	tmp = Float32(0.0)
	if (t_0 <= Float32(-0.004999999888241291))
		tmp = Float32(Float32(Float32(-alpha) * alpha) * t_0);
	else
		tmp = Float32(Float32(fma(Float32(0.5), u0, Float32(1.0)) * Float32(alpha * alpha)) * u0);
	end
	return tmp
end

Derivation

1. Split input into 2 regimes

2. if (log.f32 (-.f32 #s(literal 1 binary32) u0)) < -0.00499999989

   1. Initial program 55.4%

      \[\left(\left(-\alpha\right) \cdot \alpha\right) \cdot \log \left(1 - u0\right) \]

   if -0.00499999989 < (log.f32 (-.f32 #s(literal 1 binary32) u0))

   1. Initial program 55.4%

      \[\left(\left(-\alpha\right) \cdot \alpha\right) \cdot \log \left(1 - u0\right) \]

   2. Taylor expanded in u0 around 0

      \[\leadsto \color{blue}{u0 \cdot \left(u0 \cdot \left(\frac{1}{2} \cdot {\alpha}^{2} + u0 \cdot \left(\frac{1}{4} \cdot \left({\alpha}^{2} \cdot u0\right) + \frac{1}{3} \cdot {\alpha}^{2}\right)\right) + {\alpha}^{2}\right)} \]
   3. Step-by-step derivation

      1. *-commutative N/A

         \[\leadsto \left(u0 \cdot \left(\frac{1}{2} \cdot {\alpha}^{2} + u0 \cdot \left(\frac{1}{4} \cdot \left({\alpha}^{2} \cdot u0\right) + \frac{1}{3} \cdot {\alpha}^{2}\right)\right) + {\alpha}^{2}\right) \cdot \color{blue}{u0} \]

      2. lower-*.f32 N/A

         \[\leadsto \left(u0 \cdot \left(\frac{1}{2} \cdot {\alpha}^{2} + u0 \cdot \left(\frac{1}{4} \cdot \left({\alpha}^{2} \cdot u0\right) + \frac{1}{3} \cdot {\alpha}^{2}\right)\right) + {\alpha}^{2}\right) \cdot \color{blue}{u0} \]

   4. Applied rewrites 93.8%

      \[\leadsto \color{blue}{\mathsf{fma}\left(\mathsf{fma}\left(0.5, \alpha \cdot \alpha, \mathsf{fma}\left(0.25, \left(\alpha \cdot \alpha\right) \cdot u0, 0.3333333333333333 \cdot \left(\alpha \cdot \alpha\right)\right) \cdot u0\right), u0, \alpha \cdot \alpha\right) \cdot u0} \]

   5. Taylor expanded in alpha around 0

      \[\leadsto \left({\alpha}^{2} \cdot \left(1 + u0 \cdot \left(\frac{1}{2} + u0 \cdot \left(\frac{1}{3} + \frac{1}{4} \cdot u0\right)\right)\right)\right) \cdot u0 \]
   6. Step-by-step derivation

      1. *-commutative N/A

         \[\leadsto \left(\left(1 + u0 \cdot \left(\frac{1}{2} + u0 \cdot \left(\frac{1}{3} + \frac{1}{4} \cdot u0\right)\right)\right) \cdot {\alpha}^{2}\right) \cdot u0 \]

      2. lower-*.f32 N/A

         \[\leadsto \left(\left(1 + u0 \cdot \left(\frac{1}{2} + u0 \cdot \left(\frac{1}{3} + \frac{1}{4} \cdot u0\right)\right)\right) \cdot {\alpha}^{2}\right) \cdot u0 \]

      3. +-commutative N/A

         \[\leadsto \left(\left(u0 \cdot \left(\frac{1}{2} + u0 \cdot \left(\frac{1}{3} + \frac{1}{4} \cdot u0\right)\right) + 1\right) \cdot {\alpha}^{2}\right) \cdot u0 \]

      4. *-commutative N/A

         \[\leadsto \left(\left(\left(\frac{1}{2} + u0 \cdot \left(\frac{1}{3} + \frac{1}{4} \cdot u0\right)\right) \cdot u0 + 1\right) \cdot {\alpha}^{2}\right) \cdot u0 \]

      5. lower-fma.f32 N/A

         \[\leadsto \left(\mathsf{fma}\left(\frac{1}{2} + u0 \cdot \left(\frac{1}{3} + \frac{1}{4} \cdot u0\right), u0, 1\right) \cdot {\alpha}^{2}\right) \cdot u0 \]

      6. +-commutative N/A

         \[\leadsto \left(\mathsf{fma}\left(u0 \cdot \left(\frac{1}{3} + \frac{1}{4} \cdot u0\right) + \frac{1}{2}, u0, 1\right) \cdot {\alpha}^{2}\right) \cdot u0 \]

      7. *-commutative N/A

         \[\leadsto \left(\mathsf{fma}\left(\left(\frac{1}{3} + \frac{1}{4} \cdot u0\right) \cdot u0 + \frac{1}{2}, u0, 1\right) \cdot {\alpha}^{2}\right) \cdot u0 \]

      8. lower-fma.f32 N/A

         \[\leadsto \left(\mathsf{fma}\left(\mathsf{fma}\left(\frac{1}{3} + \frac{1}{4} \cdot u0, u0, \frac{1}{2}\right), u0, 1\right) \cdot {\alpha}^{2}\right) \cdot u0 \]

      9. +-commutative N/A

         \[\leadsto \left(\mathsf{fma}\left(\mathsf{fma}\left(\frac{1}{4} \cdot u0 + \frac{1}{3}, u0, \frac{1}{2}\right), u0, 1\right) \cdot {\alpha}^{2}\right) \cdot u0 \]

      10. lower-fma.f32 N/A

          \[\leadsto \left(\mathsf{fma}\left(\mathsf{fma}\left(\mathsf{fma}\left(\frac{1}{4}, u0, \frac{1}{3}\right), u0, \frac{1}{2}\right), u0, 1\right) \cdot {\alpha}^{2}\right) \cdot u0 \]

      11. pow2 N/A

          \[\leadsto \left(\mathsf{fma}\left(\mathsf{fma}\left(\mathsf{fma}\left(\frac{1}{4}, u0, \frac{1}{3}\right), u0, \frac{1}{2}\right), u0, 1\right) \cdot \left(\alpha \cdot \alpha\right)\right) \cdot u0 \]

      12. lift-*.f32 93.5

          \[\leadsto \left(\mathsf{fma}\left(\mathsf{fma}\left(\mathsf{fma}\left(0.25, u0, 0.3333333333333333\right), u0, 0.5\right), u0, 1\right) \cdot \left(\alpha \cdot \alpha\right)\right) \cdot u0 \]

   7. Applied rewrites 93.5%

      \[\leadsto \left(\mathsf{fma}\left(\mathsf{fma}\left(\mathsf{fma}\left(0.25, u0, 0.3333333333333333\right), u0, 0.5\right), u0, 1\right) \cdot \left(\alpha \cdot \alpha\right)\right) \cdot u0 \]

   8. Taylor expanded in u0 around 0

      \[\leadsto \left(\mathsf{fma}\left(\frac{1}{2}, u0, 1\right) \cdot \left(\alpha \cdot \alpha\right)\right) \cdot u0 \]
   9. Step-by-step derivation

   10. Applied rewrites 87.6%

       \[\leadsto \left(\mathsf{fma}\left(0.5, u0, 1\right) \cdot \left(\alpha \cdot \alpha\right)\right) \cdot u0 \]
3. Recombined 2 regimes into one program.
4. Add Preprocessing

Alternative 4: 96.5% accurate, 0.6× speedup

Math

\[\begin{array}{l} \\ \begin{array}{l} t_0 := \log \left(1 - u0\right)\\ \mathbf{if}\;t\_0 \leq -0.004999999888241291:\\ \;\;\;\;-\left(t\_0 \cdot \alpha\right) \cdot \alpha\\ \mathbf{else}:\\ \;\;\;\;\left(\mathsf{fma}\left(0.5, u0, 1\right) \cdot \left(\alpha \cdot \alpha\right)\right) \cdot u0\\ \end{array} \end{array} \]

FPCore

(FPCore (alpha u0)
 :precision binary32
 (let* ((t_0 (log (- 1.0 u0))))
   (if (<= t_0 -0.004999999888241291)
     (- (* (* t_0 alpha) alpha))
     (* (* (fma 0.5 u0 1.0) (* alpha alpha)) u0))))

C

float code(float alpha, float u0) {
	float t_0 = logf((1.0f - u0));
	float tmp;
	if (t_0 <= -0.004999999888241291f) {
		tmp = -((t_0 * alpha) * alpha);
	} else {
		tmp = (fmaf(0.5f, u0, 1.0f) * (alpha * alpha)) * u0;
	}
	return tmp;
}

Julia

function code(alpha, u0)
	t_0 = log(Float32(Float32(1.0) - u0))
	tmp = Float32(0.0)
	if (t_0 <= Float32(-0.004999999888241291))
		tmp = Float32(-Float32(Float32(t_0 * alpha) * alpha));
	else
		tmp = Float32(Float32(fma(Float32(0.5), u0, Float32(1.0)) * Float32(alpha * alpha)) * u0);
	end
	return tmp
end

Derivation

1. Split input into 2 regimes

2. if (log.f32 (-.f32 #s(literal 1 binary32) u0)) < -0.00499999989

   1. Initial program 55.4%

      \[\left(\left(-\alpha\right) \cdot \alpha\right) \cdot \log \left(1 - u0\right) \]
   2. Step-by-step derivation

      1. lift-*.f32 N/A

         \[\leadsto \color{blue}{\left(\left(-\alpha\right) \cdot \alpha\right) \cdot \log \left(1 - u0\right)} \]

      2. lift-*.f32 N/A

         \[\leadsto \color{blue}{\left(\left(-\alpha\right) \cdot \alpha\right)} \cdot \log \left(1 - u0\right) \]

      3. lift-neg.f32 N/A

         \[\leadsto \left(\color{blue}{\left(\mathsf{neg}\left(\alpha\right)\right)} \cdot \alpha\right) \cdot \log \left(1 - u0\right) \]

      4. distribute-lft-neg-out N/A

         \[\leadsto \color{blue}{\left(\mathsf{neg}\left(\alpha \cdot \alpha\right)\right)} \cdot \log \left(1 - u0\right) \]

      5. unpow2 N/A

         \[\leadsto \left(\mathsf{neg}\left(\color{blue}{{\alpha}^{2}}\right)\right) \cdot \log \left(1 - u0\right) \]

      6. lift--.f32 N/A

         \[\leadsto \left(\mathsf{neg}\left({\alpha}^{2}\right)\right) \cdot \log \color{blue}{\left(1 - u0\right)} \]

      7. lift-log.f32 N/A

         \[\leadsto \left(\mathsf{neg}\left({\alpha}^{2}\right)\right) \cdot \color{blue}{\log \left(1 - u0\right)} \]

      8. distribute-lft-neg-out N/A

         \[\leadsto \color{blue}{\mathsf{neg}\left({\alpha}^{2} \cdot \log \left(1 - u0\right)\right)} \]

      9. lower-neg.f32 N/A

         \[\leadsto \color{blue}{-{\alpha}^{2} \cdot \log \left(1 - u0\right)} \]

      10. *-commutative N/A

          \[\leadsto -\color{blue}{\log \left(1 - u0\right) \cdot {\alpha}^{2}} \]

      11. unpow2 N/A

          \[\leadsto -\log \left(1 - u0\right) \cdot \color{blue}{\left(\alpha \cdot \alpha\right)} \]

      12. associate-*r* N/A

          \[\leadsto -\color{blue}{\left(\log \left(1 - u0\right) \cdot \alpha\right) \cdot \alpha} \]

      13. *-commutative N/A

          \[\leadsto -\color{blue}{\left(\alpha \cdot \log \left(1 - u0\right)\right)} \cdot \alpha \]

      14. lower-*.f32 N/A

          \[\leadsto -\color{blue}{\left(\alpha \cdot \log \left(1 - u0\right)\right) \cdot \alpha} \]

      15. *-commutative N/A

          \[\leadsto -\color{blue}{\left(\log \left(1 - u0\right) \cdot \alpha\right)} \cdot \alpha \]

      16. lower-*.f32 N/A

          \[\leadsto -\color{blue}{\left(\log \left(1 - u0\right) \cdot \alpha\right)} \cdot \alpha \]

      17. lift-log.f32 N/A

          \[\leadsto -\left(\color{blue}{\log \left(1 - u0\right)} \cdot \alpha\right) \cdot \alpha \]

      18. lift--.f32 55.4

          \[\leadsto -\left(\log \color{blue}{\left(1 - u0\right)} \cdot \alpha\right) \cdot \alpha \]

   3. Applied rewrites 55.4%

      \[\leadsto \color{blue}{-\left(\log \left(1 - u0\right) \cdot \alpha\right) \cdot \alpha} \]

   if -0.00499999989 < (log.f32 (-.f32 #s(literal 1 binary32) u0))

   1. Initial program 55.4%

      \[\left(\left(-\alpha\right) \cdot \alpha\right) \cdot \log \left(1 - u0\right) \]

   2. Taylor expanded in u0 around 0

      \[\leadsto \color{blue}{u0 \cdot \left(u0 \cdot \left(\frac{1}{2} \cdot {\alpha}^{2} + u0 \cdot \left(\frac{1}{4} \cdot \left({\alpha}^{2} \cdot u0\right) + \frac{1}{3} \cdot {\alpha}^{2}\right)\right) + {\alpha}^{2}\right)} \]
   3. Step-by-step derivation

      1. *-commutative N/A

         \[\leadsto \left(u0 \cdot \left(\frac{1}{2} \cdot {\alpha}^{2} + u0 \cdot \left(\frac{1}{4} \cdot \left({\alpha}^{2} \cdot u0\right) + \frac{1}{3} \cdot {\alpha}^{2}\right)\right) + {\alpha}^{2}\right) \cdot \color{blue}{u0} \]

      2. lower-*.f32 N/A

         \[\leadsto \left(u0 \cdot \left(\frac{1}{2} \cdot {\alpha}^{2} + u0 \cdot \left(\frac{1}{4} \cdot \left({\alpha}^{2} \cdot u0\right) + \frac{1}{3} \cdot {\alpha}^{2}\right)\right) + {\alpha}^{2}\right) \cdot \color{blue}{u0} \]

   4. Applied rewrites 93.8%

      \[\leadsto \color{blue}{\mathsf{fma}\left(\mathsf{fma}\left(0.5, \alpha \cdot \alpha, \mathsf{fma}\left(0.25, \left(\alpha \cdot \alpha\right) \cdot u0, 0.3333333333333333 \cdot \left(\alpha \cdot \alpha\right)\right) \cdot u0\right), u0, \alpha \cdot \alpha\right) \cdot u0} \]

   5. Taylor expanded in alpha around 0

      \[\leadsto \left({\alpha}^{2} \cdot \left(1 + u0 \cdot \left(\frac{1}{2} + u0 \cdot \left(\frac{1}{3} + \frac{1}{4} \cdot u0\right)\right)\right)\right) \cdot u0 \]
   6. Step-by-step derivation

      1. *-commutative N/A

         \[\leadsto \left(\left(1 + u0 \cdot \left(\frac{1}{2} + u0 \cdot \left(\frac{1}{3} + \frac{1}{4} \cdot u0\right)\right)\right) \cdot {\alpha}^{2}\right) \cdot u0 \]

      2. lower-*.f32 N/A

         \[\leadsto \left(\left(1 + u0 \cdot \left(\frac{1}{2} + u0 \cdot \left(\frac{1}{3} + \frac{1}{4} \cdot u0\right)\right)\right) \cdot {\alpha}^{2}\right) \cdot u0 \]

      3. +-commutative N/A

         \[\leadsto \left(\left(u0 \cdot \left(\frac{1}{2} + u0 \cdot \left(\frac{1}{3} + \frac{1}{4} \cdot u0\right)\right) + 1\right) \cdot {\alpha}^{2}\right) \cdot u0 \]

      4. *-commutative N/A

         \[\leadsto \left(\left(\left(\frac{1}{2} + u0 \cdot \left(\frac{1}{3} + \frac{1}{4} \cdot u0\right)\right) \cdot u0 + 1\right) \cdot {\alpha}^{2}\right) \cdot u0 \]

      5. lower-fma.f32 N/A

         \[\leadsto \left(\mathsf{fma}\left(\frac{1}{2} + u0 \cdot \left(\frac{1}{3} + \frac{1}{4} \cdot u0\right), u0, 1\right) \cdot {\alpha}^{2}\right) \cdot u0 \]

      6. +-commutative N/A

         \[\leadsto \left(\mathsf{fma}\left(u0 \cdot \left(\frac{1}{3} + \frac{1}{4} \cdot u0\right) + \frac{1}{2}, u0, 1\right) \cdot {\alpha}^{2}\right) \cdot u0 \]

      7. *-commutative N/A

         \[\leadsto \left(\mathsf{fma}\left(\left(\frac{1}{3} + \frac{1}{4} \cdot u0\right) \cdot u0 + \frac{1}{2}, u0, 1\right) \cdot {\alpha}^{2}\right) \cdot u0 \]

      8. lower-fma.f32 N/A

         \[\leadsto \left(\mathsf{fma}\left(\mathsf{fma}\left(\frac{1}{3} + \frac{1}{4} \cdot u0, u0, \frac{1}{2}\right), u0, 1\right) \cdot {\alpha}^{2}\right) \cdot u0 \]

      9. +-commutative N/A

         \[\leadsto \left(\mathsf{fma}\left(\mathsf{fma}\left(\frac{1}{4} \cdot u0 + \frac{1}{3}, u0, \frac{1}{2}\right), u0, 1\right) \cdot {\alpha}^{2}\right) \cdot u0 \]

      10. lower-fma.f32 N/A

          \[\leadsto \left(\mathsf{fma}\left(\mathsf{fma}\left(\mathsf{fma}\left(\frac{1}{4}, u0, \frac{1}{3}\right), u0, \frac{1}{2}\right), u0, 1\right) \cdot {\alpha}^{2}\right) \cdot u0 \]

      11. pow2 N/A

          \[\leadsto \left(\mathsf{fma}\left(\mathsf{fma}\left(\mathsf{fma}\left(\frac{1}{4}, u0, \frac{1}{3}\right), u0, \frac{1}{2}\right), u0, 1\right) \cdot \left(\alpha \cdot \alpha\right)\right) \cdot u0 \]

      12. lift-*.f32 93.5

          \[\leadsto \left(\mathsf{fma}\left(\mathsf{fma}\left(\mathsf{fma}\left(0.25, u0, 0.3333333333333333\right), u0, 0.5\right), u0, 1\right) \cdot \left(\alpha \cdot \alpha\right)\right) \cdot u0 \]

   7. Applied rewrites 93.5%

      \[\leadsto \left(\mathsf{fma}\left(\mathsf{fma}\left(\mathsf{fma}\left(0.25, u0, 0.3333333333333333\right), u0, 0.5\right), u0, 1\right) \cdot \left(\alpha \cdot \alpha\right)\right) \cdot u0 \]

   8. Taylor expanded in u0 around 0

      \[\leadsto \left(\mathsf{fma}\left(\frac{1}{2}, u0, 1\right) \cdot \left(\alpha \cdot \alpha\right)\right) \cdot u0 \]
   9. Step-by-step derivation

   10. Applied rewrites 87.6%

       \[\leadsto \left(\mathsf{fma}\left(0.5, u0, 1\right) \cdot \left(\alpha \cdot \alpha\right)\right) \cdot u0 \]
3. Recombined 2 regimes into one program.
4. Add Preprocessing

Alternative 5: 87.6% accurate, 1.1× speedup

Math

\[\begin{array}{l} \\ \left(\mathsf{fma}\left(0.5, u0, 1\right) \cdot \left(\alpha \cdot \alpha\right)\right) \cdot u0 \end{array} \]

FPCore

(FPCore (alpha u0)
 :precision binary32
 (* (* (fma 0.5 u0 1.0) (* alpha alpha)) u0))

C

float code(float alpha, float u0) {
	return (fmaf(0.5f, u0, 1.0f) * (alpha * alpha)) * u0;
}

Julia

function code(alpha, u0)
	return Float32(Float32(fma(Float32(0.5), u0, Float32(1.0)) * Float32(alpha * alpha)) * u0)
end

Derivation

1. Initial program 55.4%

   \[\left(\left(-\alpha\right) \cdot \alpha\right) \cdot \log \left(1 - u0\right) \]

2. Taylor expanded in u0 around 0

   \[\leadsto \color{blue}{u0 \cdot \left(u0 \cdot \left(\frac{1}{2} \cdot {\alpha}^{2} + u0 \cdot \left(\frac{1}{4} \cdot \left({\alpha}^{2} \cdot u0\right) + \frac{1}{3} \cdot {\alpha}^{2}\right)\right) + {\alpha}^{2}\right)} \]
3. Step-by-step derivation

   1. *-commutative N/A

      \[\leadsto \left(u0 \cdot \left(\frac{1}{2} \cdot {\alpha}^{2} + u0 \cdot \left(\frac{1}{4} \cdot \left({\alpha}^{2} \cdot u0\right) + \frac{1}{3} \cdot {\alpha}^{2}\right)\right) + {\alpha}^{2}\right) \cdot \color{blue}{u0} \]

   2. lower-*.f32 N/A

      \[\leadsto \left(u0 \cdot \left(\frac{1}{2} \cdot {\alpha}^{2} + u0 \cdot \left(\frac{1}{4} \cdot \left({\alpha}^{2} \cdot u0\right) + \frac{1}{3} \cdot {\alpha}^{2}\right)\right) + {\alpha}^{2}\right) \cdot \color{blue}{u0} \]

4. Applied rewrites 93.8%

   \[\leadsto \color{blue}{\mathsf{fma}\left(\mathsf{fma}\left(0.5, \alpha \cdot \alpha, \mathsf{fma}\left(0.25, \left(\alpha \cdot \alpha\right) \cdot u0, 0.3333333333333333 \cdot \left(\alpha \cdot \alpha\right)\right) \cdot u0\right), u0, \alpha \cdot \alpha\right) \cdot u0} \]

5. Taylor expanded in alpha around 0

   \[\leadsto \left({\alpha}^{2} \cdot \left(1 + u0 \cdot \left(\frac{1}{2} + u0 \cdot \left(\frac{1}{3} + \frac{1}{4} \cdot u0\right)\right)\right)\right) \cdot u0 \]
6. Step-by-step derivation

   1. *-commutative N/A

      \[\leadsto \left(\left(1 + u0 \cdot \left(\frac{1}{2} + u0 \cdot \left(\frac{1}{3} + \frac{1}{4} \cdot u0\right)\right)\right) \cdot {\alpha}^{2}\right) \cdot u0 \]

   2. lower-*.f32 N/A

      \[\leadsto \left(\left(1 + u0 \cdot \left(\frac{1}{2} + u0 \cdot \left(\frac{1}{3} + \frac{1}{4} \cdot u0\right)\right)\right) \cdot {\alpha}^{2}\right) \cdot u0 \]

   3. +-commutative N/A

      \[\leadsto \left(\left(u0 \cdot \left(\frac{1}{2} + u0 \cdot \left(\frac{1}{3} + \frac{1}{4} \cdot u0\right)\right) + 1\right) \cdot {\alpha}^{2}\right) \cdot u0 \]

   4. *-commutative N/A

      \[\leadsto \left(\left(\left(\frac{1}{2} + u0 \cdot \left(\frac{1}{3} + \frac{1}{4} \cdot u0\right)\right) \cdot u0 + 1\right) \cdot {\alpha}^{2}\right) \cdot u0 \]

   5. lower-fma.f32 N/A

      \[\leadsto \left(\mathsf{fma}\left(\frac{1}{2} + u0 \cdot \left(\frac{1}{3} + \frac{1}{4} \cdot u0\right), u0, 1\right) \cdot {\alpha}^{2}\right) \cdot u0 \]

   6. +-commutative N/A

      \[\leadsto \left(\mathsf{fma}\left(u0 \cdot \left(\frac{1}{3} + \frac{1}{4} \cdot u0\right) + \frac{1}{2}, u0, 1\right) \cdot {\alpha}^{2}\right) \cdot u0 \]

   7. *-commutative N/A

      \[\leadsto \left(\mathsf{fma}\left(\left(\frac{1}{3} + \frac{1}{4} \cdot u0\right) \cdot u0 + \frac{1}{2}, u0, 1\right) \cdot {\alpha}^{2}\right) \cdot u0 \]

   8. lower-fma.f32 N/A

      \[\leadsto \left(\mathsf{fma}\left(\mathsf{fma}\left(\frac{1}{3} + \frac{1}{4} \cdot u0, u0, \frac{1}{2}\right), u0, 1\right) \cdot {\alpha}^{2}\right) \cdot u0 \]

   9. +-commutative N/A

      \[\leadsto \left(\mathsf{fma}\left(\mathsf{fma}\left(\frac{1}{4} \cdot u0 + \frac{1}{3}, u0, \frac{1}{2}\right), u0, 1\right) \cdot {\alpha}^{2}\right) \cdot u0 \]

   10. lower-fma.f32 N/A

       \[\leadsto \left(\mathsf{fma}\left(\mathsf{fma}\left(\mathsf{fma}\left(\frac{1}{4}, u0, \frac{1}{3}\right), u0, \frac{1}{2}\right), u0, 1\right) \cdot {\alpha}^{2}\right) \cdot u0 \]

   11. pow2 N/A

       \[\leadsto \left(\mathsf{fma}\left(\mathsf{fma}\left(\mathsf{fma}\left(\frac{1}{4}, u0, \frac{1}{3}\right), u0, \frac{1}{2}\right), u0, 1\right) \cdot \left(\alpha \cdot \alpha\right)\right) \cdot u0 \]

   12. lift-*.f32 93.5

       \[\leadsto \left(\mathsf{fma}\left(\mathsf{fma}\left(\mathsf{fma}\left(0.25, u0, 0.3333333333333333\right), u0, 0.5\right), u0, 1\right) \cdot \left(\alpha \cdot \alpha\right)\right) \cdot u0 \]

7. Applied rewrites 93.5%

   \[\leadsto \left(\mathsf{fma}\left(\mathsf{fma}\left(\mathsf{fma}\left(0.25, u0, 0.3333333333333333\right), u0, 0.5\right), u0, 1\right) \cdot \left(\alpha \cdot \alpha\right)\right) \cdot u0 \]

8. Taylor expanded in u0 around 0

   \[\leadsto \left(\mathsf{fma}\left(\frac{1}{2}, u0, 1\right) \cdot \left(\alpha \cdot \alpha\right)\right) \cdot u0 \]
9. Step-by-step derivation

10. Applied rewrites 87.6%

    \[\leadsto \left(\mathsf{fma}\left(0.5, u0, 1\right) \cdot \left(\alpha \cdot \alpha\right)\right) \cdot u0 \]
11. Add Preprocessing

Alternative 6: 87.6% accurate, 1.1× speedup

Math

\[\begin{array}{l} \\ \left(\mathsf{fma}\left(0.5, u0, 1\right) \cdot u0\right) \cdot \left(\alpha \cdot \alpha\right) \end{array} \]

FPCore

(FPCore (alpha u0)
 :precision binary32
 (* (* (fma 0.5 u0 1.0) u0) (* alpha alpha)))

C

float code(float alpha, float u0) {
	return (fmaf(0.5f, u0, 1.0f) * u0) * (alpha * alpha);
}

Julia

function code(alpha, u0)
	return Float32(Float32(fma(Float32(0.5), u0, Float32(1.0)) * u0) * Float32(alpha * alpha))
end

Derivation

1. Initial program 55.4%

   \[\left(\left(-\alpha\right) \cdot \alpha\right) \cdot \log \left(1 - u0\right) \]

2. Taylor expanded in u0 around 0

   \[\leadsto \color{blue}{u0 \cdot \left(\frac{1}{2} \cdot \left({\alpha}^{2} \cdot u0\right) + {\alpha}^{2}\right)} \]
3. Step-by-step derivation

   1. *-commutative N/A

      \[\leadsto \left(\frac{1}{2} \cdot \left({\alpha}^{2} \cdot u0\right) + {\alpha}^{2}\right) \cdot \color{blue}{u0} \]

   2. lower-*.f32 N/A

      \[\leadsto \left(\frac{1}{2} \cdot \left({\alpha}^{2} \cdot u0\right) + {\alpha}^{2}\right) \cdot \color{blue}{u0} \]

   3. *-commutative N/A

      \[\leadsto \left(\left({\alpha}^{2} \cdot u0\right) \cdot \frac{1}{2} + {\alpha}^{2}\right) \cdot u0 \]

   4. lower-fma.f32 N/A

      \[\leadsto \mathsf{fma}\left({\alpha}^{2} \cdot u0, \frac{1}{2}, {\alpha}^{2}\right) \cdot u0 \]

   5. lower-*.f32 N/A

      \[\leadsto \mathsf{fma}\left({\alpha}^{2} \cdot u0, \frac{1}{2}, {\alpha}^{2}\right) \cdot u0 \]

   6. unpow2 N/A

      \[\leadsto \mathsf{fma}\left(\left(\alpha \cdot \alpha\right) \cdot u0, \frac{1}{2}, {\alpha}^{2}\right) \cdot u0 \]

   7. lower-*.f32 N/A

      \[\leadsto \mathsf{fma}\left(\left(\alpha \cdot \alpha\right) \cdot u0, \frac{1}{2}, {\alpha}^{2}\right) \cdot u0 \]

   8. unpow2 N/A

      \[\leadsto \mathsf{fma}\left(\left(\alpha \cdot \alpha\right) \cdot u0, \frac{1}{2}, \alpha \cdot \alpha\right) \cdot u0 \]

   9. lower-*.f32 87.8

      \[\leadsto \mathsf{fma}\left(\left(\alpha \cdot \alpha\right) \cdot u0, 0.5, \alpha \cdot \alpha\right) \cdot u0 \]

4. Applied rewrites 87.8%

   \[\leadsto \color{blue}{\mathsf{fma}\left(\left(\alpha \cdot \alpha\right) \cdot u0, 0.5, \alpha \cdot \alpha\right) \cdot u0} \]

5. Taylor expanded in alpha around 0

   \[\leadsto {\alpha}^{2} \cdot \color{blue}{\left(u0 \cdot \left(1 + \frac{1}{2} \cdot u0\right)\right)} \]
6. Step-by-step derivation

   1. *-commutative N/A

      \[\leadsto \left(u0 \cdot \left(1 + \frac{1}{2} \cdot u0\right)\right) \cdot {\alpha}^{\color{blue}{2}} \]

   2. lower-*.f32 N/A

      \[\leadsto \left(u0 \cdot \left(1 + \frac{1}{2} \cdot u0\right)\right) \cdot {\alpha}^{\color{blue}{2}} \]

   3. *-commutative N/A

      \[\leadsto \left(\left(1 + \frac{1}{2} \cdot u0\right) \cdot u0\right) \cdot {\alpha}^{2} \]

   4. lower-*.f32 N/A

      \[\leadsto \left(\left(1 + \frac{1}{2} \cdot u0\right) \cdot u0\right) \cdot {\alpha}^{2} \]

   5. +-commutative N/A

      \[\leadsto \left(\left(\frac{1}{2} \cdot u0 + 1\right) \cdot u0\right) \cdot {\alpha}^{2} \]

   6. lower-fma.f32 N/A

      \[\leadsto \left(\mathsf{fma}\left(\frac{1}{2}, u0, 1\right) \cdot u0\right) \cdot {\alpha}^{2} \]

   7. pow2 N/A

      \[\leadsto \left(\mathsf{fma}\left(\frac{1}{2}, u0, 1\right) \cdot u0\right) \cdot \left(\alpha \cdot \alpha\right) \]

   8. lift-*.f32 87.6

      \[\leadsto \left(\mathsf{fma}\left(0.5, u0, 1\right) \cdot u0\right) \cdot \left(\alpha \cdot \alpha\right) \]

7. Applied rewrites 87.6%

   \[\leadsto \left(\mathsf{fma}\left(0.5, u0, 1\right) \cdot u0\right) \cdot \color{blue}{\left(\alpha \cdot \alpha\right)} \]
8. Add Preprocessing

Alternative 7: 75.0% accurate, 2.3× speedup

Math

\[\begin{array}{l} \\ \left(\alpha \cdot \alpha\right) \cdot u0 \end{array} \]

FPCore

(FPCore (alpha u0) :precision binary32 (* (* alpha alpha) u0))

C

float code(float alpha, float u0) {
	return (alpha * alpha) * u0;
}

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(4) function code(alpha, u0)
use fmin_fmax_functions
    real(4), intent (in) :: alpha
    real(4), intent (in) :: u0
    code = (alpha * alpha) * u0
end function

Julia

function code(alpha, u0)
	return Float32(Float32(alpha * alpha) * u0)
end

MATLAB

function tmp = code(alpha, u0)
	tmp = (alpha * alpha) * u0;
end

Derivation

1. Initial program 55.4%

   \[\left(\left(-\alpha\right) \cdot \alpha\right) \cdot \log \left(1 - u0\right) \]

2. Taylor expanded in u0 around 0

   \[\leadsto \color{blue}{{\alpha}^{2} \cdot u0} \]
3. Step-by-step derivation

   1. lower-*.f32 N/A

      \[\leadsto {\alpha}^{2} \cdot \color{blue}{u0} \]

   2. unpow2 N/A

      \[\leadsto \left(\alpha \cdot \alpha\right) \cdot u0 \]

   3. lower-*.f32 75.0

      \[\leadsto \left(\alpha \cdot \alpha\right) \cdot u0 \]

4. Applied rewrites 75.0%

   \[\leadsto \color{blue}{\left(\alpha \cdot \alpha\right) \cdot u0} \]
5. Add Preprocessing

Alternative 8: 75.0% accurate, 2.3× speedup

Math

\[\begin{array}{l} \\ \alpha \cdot \left(u0 \cdot \alpha\right) \end{array} \]

FPCore

(FPCore (alpha u0) :precision binary32 (* alpha (* u0 alpha)))

C

float code(float alpha, float u0) {
	return alpha * (u0 * alpha);
}

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(4) function code(alpha, u0)
use fmin_fmax_functions
    real(4), intent (in) :: alpha
    real(4), intent (in) :: u0
    code = alpha * (u0 * alpha)
end function

Julia

function code(alpha, u0)
	return Float32(alpha * Float32(u0 * alpha))
end

MATLAB

function tmp = code(alpha, u0)
	tmp = alpha * (u0 * alpha);
end

Derivation

1. Initial program 55.4%

   \[\left(\left(-\alpha\right) \cdot \alpha\right) \cdot \log \left(1 - u0\right) \]

2. Taylor expanded in u0 around 0

   \[\leadsto \color{blue}{{\alpha}^{2} \cdot u0} \]
3. Step-by-step derivation

   1. lower-*.f32 N/A

      \[\leadsto {\alpha}^{2} \cdot \color{blue}{u0} \]

   2. unpow2 N/A

      \[\leadsto \left(\alpha \cdot \alpha\right) \cdot u0 \]

   3. lower-*.f32 75.0

      \[\leadsto \left(\alpha \cdot \alpha\right) \cdot u0 \]

4. Applied rewrites 75.0%

   \[\leadsto \color{blue}{\left(\alpha \cdot \alpha\right) \cdot u0} \]
5. Step-by-step derivation

   1. lift-*.f32 N/A

      \[\leadsto \left(\alpha \cdot \alpha\right) \cdot \color{blue}{u0} \]

   2. lift-*.f32 N/A

      \[\leadsto \left(\alpha \cdot \alpha\right) \cdot u0 \]

   3. associate-*l* N/A

      \[\leadsto \alpha \cdot \color{blue}{\left(\alpha \cdot u0\right)} \]

   4. lower-*.f32 N/A

      \[\leadsto \alpha \cdot \color{blue}{\left(\alpha \cdot u0\right)} \]

   5. *-commutative N/A

      \[\leadsto \alpha \cdot \left(u0 \cdot \color{blue}{\alpha}\right) \]

   6. lower-*.f32 75.0

      \[\leadsto \alpha \cdot \left(u0 \cdot \color{blue}{\alpha}\right) \]

6. Applied rewrites 75.0%

   \[\leadsto \alpha \cdot \color{blue}{\left(u0 \cdot \alpha\right)} \]
7. Add Preprocessing

Reproduce

herbie shell --seed 2025140 
(FPCore (alpha u0)
  :name "Beckmann Distribution sample, tan2theta, alphax == alphay"
  :precision binary32
  :pre (and (and (<= 0.0001 alpha) (<= alpha 1.0)) (and (<= 2.328306437e-10 u0) (<= u0 1.0)))
  (* (* (- alpha) alpha) (log (- 1.0 u0))))