symmetry log of sum of exp

Percentage Accurate: 54.1% → 98.3%

Time: 8.3s

Alternatives: 13

Speedup: 2.8×

• could not determine a ground truth

  1. a = -6.283272980618857e+183
  2. b = 2.2416529225755475e+140

Specification

Math

\[\begin{array}{l} \\ \log \left(e^{a} + e^{b}\right) \end{array} \]

FPCore

(FPCore (a b) :precision binary64 (log (+ (exp a) (exp b))))

C

double code(double a, double b) {
	return log((exp(a) + exp(b)));
}

Fortran

module fmin_fmax_functions
    implicit none
    private
    public fmax
    public fmin

    interface fmax
        module procedure fmax88
        module procedure fmax44
        module procedure fmax84
        module procedure fmax48
    end interface
    interface fmin
        module procedure fmin88
        module procedure fmin44
        module procedure fmin84
        module procedure fmin48
    end interface
contains
    real(8) function fmax88(x, y) result (res)
        real(8), intent (in) :: x
        real(8), intent (in) :: y
        res = merge(y, merge(x, max(x, y), y /= y), x /= x)
    end function
    real(4) function fmax44(x, y) result (res)
        real(4), intent (in) :: x
        real(4), intent (in) :: y
        res = merge(y, merge(x, max(x, y), y /= y), x /= x)
    end function
    real(8) function fmax84(x, y) result(res)
        real(8), intent (in) :: x
        real(4), intent (in) :: y
        res = merge(dble(y), merge(x, max(x, dble(y)), y /= y), x /= x)
    end function
    real(8) function fmax48(x, y) result(res)
        real(4), intent (in) :: x
        real(8), intent (in) :: y
        res = merge(y, merge(dble(x), max(dble(x), y), y /= y), x /= x)
    end function
    real(8) function fmin88(x, y) result (res)
        real(8), intent (in) :: x
        real(8), intent (in) :: y
        res = merge(y, merge(x, min(x, y), y /= y), x /= x)
    end function
    real(4) function fmin44(x, y) result (res)
        real(4), intent (in) :: x
        real(4), intent (in) :: y
        res = merge(y, merge(x, min(x, y), y /= y), x /= x)
    end function
    real(8) function fmin84(x, y) result(res)
        real(8), intent (in) :: x
        real(4), intent (in) :: y
        res = merge(dble(y), merge(x, min(x, dble(y)), y /= y), x /= x)
    end function
    real(8) function fmin48(x, y) result(res)
        real(4), intent (in) :: x
        real(8), intent (in) :: y
        res = merge(y, merge(dble(x), min(dble(x), y), y /= y), x /= x)
    end function
end module

real(8) function code(a, b)
use fmin_fmax_functions
    real(8), intent (in) :: a
    real(8), intent (in) :: b
    code = log((exp(a) + exp(b)))
end function

Java

public static double code(double a, double b) {
	return Math.log((Math.exp(a) + Math.exp(b)));
}

Python

def code(a, b):
	return math.log((math.exp(a) + math.exp(b)))

Julia

function code(a, b)
	return log(Float64(exp(a) + exp(b)))
end

MATLAB

function tmp = code(a, b)
	tmp = log((exp(a) + exp(b)));
end

Wolfram

code[a_, b_] := N[Log[N[(N[Exp[a], $MachinePrecision] + N[Exp[b], $MachinePrecision]), $MachinePrecision]], $MachinePrecision]

Initial Program: 54.1% accurate, 1.0× speedup

Math

\[\begin{array}{l} \\ \log \left(e^{a} + e^{b}\right) \end{array} \]

FPCore

(FPCore (a b) :precision binary64 (log (+ (exp a) (exp b))))

C

double code(double a, double b) {
	return log((exp(a) + exp(b)));
}

Fortran

module fmin_fmax_functions
    implicit none
    private
    public fmax
    public fmin

    interface fmax
        module procedure fmax88
        module procedure fmax44
        module procedure fmax84
        module procedure fmax48
    end interface
    interface fmin
        module procedure fmin88
        module procedure fmin44
        module procedure fmin84
        module procedure fmin48
    end interface
contains
    real(8) function fmax88(x, y) result (res)
        real(8), intent (in) :: x
        real(8), intent (in) :: y
        res = merge(y, merge(x, max(x, y), y /= y), x /= x)
    end function
    real(4) function fmax44(x, y) result (res)
        real(4), intent (in) :: x
        real(4), intent (in) :: y
        res = merge(y, merge(x, max(x, y), y /= y), x /= x)
    end function
    real(8) function fmax84(x, y) result(res)
        real(8), intent (in) :: x
        real(4), intent (in) :: y
        res = merge(dble(y), merge(x, max(x, dble(y)), y /= y), x /= x)
    end function
    real(8) function fmax48(x, y) result(res)
        real(4), intent (in) :: x
        real(8), intent (in) :: y
        res = merge(y, merge(dble(x), max(dble(x), y), y /= y), x /= x)
    end function
    real(8) function fmin88(x, y) result (res)
        real(8), intent (in) :: x
        real(8), intent (in) :: y
        res = merge(y, merge(x, min(x, y), y /= y), x /= x)
    end function
    real(4) function fmin44(x, y) result (res)
        real(4), intent (in) :: x
        real(4), intent (in) :: y
        res = merge(y, merge(x, min(x, y), y /= y), x /= x)
    end function
    real(8) function fmin84(x, y) result(res)
        real(8), intent (in) :: x
        real(4), intent (in) :: y
        res = merge(dble(y), merge(x, min(x, dble(y)), y /= y), x /= x)
    end function
    real(8) function fmin48(x, y) result(res)
        real(4), intent (in) :: x
        real(8), intent (in) :: y
        res = merge(y, merge(dble(x), min(dble(x), y), y /= y), x /= x)
    end function
end module

real(8) function code(a, b)
use fmin_fmax_functions
    real(8), intent (in) :: a
    real(8), intent (in) :: b
    code = log((exp(a) + exp(b)))
end function

Java

public static double code(double a, double b) {
	return Math.log((Math.exp(a) + Math.exp(b)));
}

Python

def code(a, b):
	return math.log((math.exp(a) + math.exp(b)))

Julia

function code(a, b)
	return log(Float64(exp(a) + exp(b)))
end

MATLAB

function tmp = code(a, b)
	tmp = log((exp(a) + exp(b)));
end

Wolfram

code[a_, b_] := N[Log[N[(N[Exp[a], $MachinePrecision] + N[Exp[b], $MachinePrecision]), $MachinePrecision]], $MachinePrecision]

Alternative 1: 98.3% accurate, 1.3× speedup

Math

\[\begin{array}{l} [a, b] = \mathsf{sort}([a, b])\\ \\ \begin{array}{l} \mathbf{if}\;a \leq -37:\\ \;\;\;\;\frac{b}{1 + e^{a}}\\ \mathbf{else}:\\ \;\;\;\;\log \left(e^{a} + \mathsf{fma}\left(\mathsf{fma}\left(\mathsf{fma}\left(0.16666666666666666, b, 0.5\right), b, 1\right), b, 1\right)\right)\\ \end{array} \end{array} \]

FPCore

NOTE: a and b should be sorted in increasing order before calling this function.
(FPCore (a b)
 :precision binary64
 (if (<= a -37.0)
   (/ b (+ 1.0 (exp a)))
   (log (+ (exp a) (fma (fma (fma 0.16666666666666666 b 0.5) b 1.0) b 1.0)))))

C

assert(a < b);
double code(double a, double b) {
	double tmp;
	if (a <= -37.0) {
		tmp = b / (1.0 + exp(a));
	} else {
		tmp = log((exp(a) + fma(fma(fma(0.16666666666666666, b, 0.5), b, 1.0), b, 1.0)));
	}
	return tmp;
}

Julia

a, b = sort([a, b])
function code(a, b)
	tmp = 0.0
	if (a <= -37.0)
		tmp = Float64(b / Float64(1.0 + exp(a)));
	else
		tmp = log(Float64(exp(a) + fma(fma(fma(0.16666666666666666, b, 0.5), b, 1.0), b, 1.0)));
	end
	return tmp
end

Wolfram

NOTE: a and b should be sorted in increasing order before calling this function.
code[a_, b_] := If[LessEqual[a, -37.0], N[(b / N[(1.0 + N[Exp[a], $MachinePrecision]), $MachinePrecision]), $MachinePrecision], N[Log[N[(N[Exp[a], $MachinePrecision] + N[(N[(N[(0.16666666666666666 * b + 0.5), $MachinePrecision] * b + 1.0), $MachinePrecision] * b + 1.0), $MachinePrecision]), $MachinePrecision]], $MachinePrecision]]

Derivation

1. Split input into 2 regimes

2. if a < -37

   1. Initial program 9.0%

      \[\log \left(e^{a} + e^{b}\right) \]
   2. Add Preprocessing

   3. Taylor expanded in b around 0

      \[\leadsto \color{blue}{\log \left(1 + e^{a}\right) + \frac{b}{1 + e^{a}}} \]
   4. Step-by-step derivation

      1. +-commutative N/A

         \[\leadsto \frac{b}{1 + e^{a}} + \color{blue}{\log \left(1 + e^{a}\right)} \]

      2. lower-+.f64 N/A

         \[\leadsto \frac{b}{1 + e^{a}} + \color{blue}{\log \left(1 + e^{a}\right)} \]

      3. lower-/.f64 N/A

         \[\leadsto \frac{b}{1 + e^{a}} + \log \color{blue}{\left(1 + e^{a}\right)} \]

      4. +-commutative N/A

         \[\leadsto \frac{b}{e^{a} + 1} + \log \left(1 + \color{blue}{e^{a}}\right) \]

      5. metadata-eval N/A

         \[\leadsto \frac{b}{e^{a} + 1 \cdot 1} + \log \left(1 + e^{a}\right) \]

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

         \[\leadsto \frac{b}{e^{a} - \left(\mathsf{neg}\left(1\right)\right) \cdot 1} + \log \left(1 + \color{blue}{e^{a}}\right) \]

      7. metadata-eval N/A

         \[\leadsto \frac{b}{e^{a} - -1 \cdot 1} + \log \left(1 + e^{a}\right) \]

      8. metadata-eval N/A

         \[\leadsto \frac{b}{e^{a} - -1} + \log \left(1 + e^{a}\right) \]

      9. lower--.f64 N/A

         \[\leadsto \frac{b}{e^{a} - -1} + \log \left(1 + \color{blue}{e^{a}}\right) \]

      10. lift-exp.f64 N/A

          \[\leadsto \frac{b}{e^{a} - -1} + \log \left(1 + e^{\color{blue}{a}}\right) \]

      11. lower-log1p.f64 N/A

          \[\leadsto \frac{b}{e^{a} - -1} + \mathsf{log1p}\left(e^{a}\right) \]

      12. lift-exp.f64 100.0

          \[\leadsto \frac{b}{e^{a} - -1} + \mathsf{log1p}\left(e^{a}\right) \]

   5. Applied rewrites 100.0%

      \[\leadsto \color{blue}{\frac{b}{e^{a} - -1} + \mathsf{log1p}\left(e^{a}\right)} \]

   6. Taylor expanded in b around inf

      \[\leadsto \frac{b}{\color{blue}{1 + e^{a}}} \]
   7. Step-by-step derivation

      1. lower-/.f64 N/A

         \[\leadsto \frac{b}{1 + \color{blue}{e^{a}}} \]

      2. lower-+.f64 N/A

         \[\leadsto \frac{b}{1 + e^{a}} \]

      3. lift-exp.f64 99.8

         \[\leadsto \frac{b}{1 + e^{a}} \]

   8. Applied rewrites 99.8%

      \[\leadsto \frac{b}{\color{blue}{1 + e^{a}}} \]

   if -37 < a

   1. Initial program 97.9%

      \[\log \left(e^{a} + e^{b}\right) \]
   2. Add Preprocessing

   3. Taylor expanded in b around 0

      \[\leadsto \log \left(e^{a} + \color{blue}{\left(1 + b \cdot \left(1 + b \cdot \left(\frac{1}{2} + \frac{1}{6} \cdot b\right)\right)\right)}\right) \]
   4. Step-by-step derivation

      1. +-commutative N/A

         \[\leadsto \log \left(e^{a} + \left(b \cdot \left(1 + b \cdot \left(\frac{1}{2} + \frac{1}{6} \cdot b\right)\right) + \color{blue}{1}\right)\right) \]

      2. *-commutative N/A

         \[\leadsto \log \left(e^{a} + \left(\left(1 + b \cdot \left(\frac{1}{2} + \frac{1}{6} \cdot b\right)\right) \cdot b + 1\right)\right) \]

      3. lower-fma.f64 N/A

         \[\leadsto \log \left(e^{a} + \mathsf{fma}\left(1 + b \cdot \left(\frac{1}{2} + \frac{1}{6} \cdot b\right), \color{blue}{b}, 1\right)\right) \]

      4. +-commutative N/A

         \[\leadsto \log \left(e^{a} + \mathsf{fma}\left(b \cdot \left(\frac{1}{2} + \frac{1}{6} \cdot b\right) + 1, b, 1\right)\right) \]

      5. *-commutative N/A

         \[\leadsto \log \left(e^{a} + \mathsf{fma}\left(\left(\frac{1}{2} + \frac{1}{6} \cdot b\right) \cdot b + 1, b, 1\right)\right) \]

      6. lower-fma.f64 N/A

         \[\leadsto \log \left(e^{a} + \mathsf{fma}\left(\mathsf{fma}\left(\frac{1}{2} + \frac{1}{6} \cdot b, b, 1\right), b, 1\right)\right) \]

      7. +-commutative N/A

         \[\leadsto \log \left(e^{a} + \mathsf{fma}\left(\mathsf{fma}\left(\frac{1}{6} \cdot b + \frac{1}{2}, b, 1\right), b, 1\right)\right) \]

      8. lower-fma.f64 96.9

         \[\leadsto \log \left(e^{a} + \mathsf{fma}\left(\mathsf{fma}\left(\mathsf{fma}\left(0.16666666666666666, b, 0.5\right), b, 1\right), b, 1\right)\right) \]

   5. Applied rewrites 96.9%

      \[\leadsto \log \left(e^{a} + \color{blue}{\mathsf{fma}\left(\mathsf{fma}\left(\mathsf{fma}\left(0.16666666666666666, b, 0.5\right), b, 1\right), b, 1\right)}\right) \]
3. Recombined 2 regimes into one program.
4. Add Preprocessing

Alternative 2: 57.0% accurate, 0.7× speedup

Math

\[\begin{array}{l} [a, b] = \mathsf{sort}([a, b])\\ \\ \begin{array}{l} \mathbf{if}\;\log \left(e^{a} + e^{b}\right) \leq 5 \cdot 10^{-10}:\\ \;\;\;\;0.5 \cdot b\\ \mathbf{else}:\\ \;\;\;\;\mathsf{fma}\left(0.5, b, \log 2\right)\\ \end{array} \end{array} \]

FPCore

NOTE: a and b should be sorted in increasing order before calling this function.
(FPCore (a b)
 :precision binary64
 (if (<= (log (+ (exp a) (exp b))) 5e-10) (* 0.5 b) (fma 0.5 b (log 2.0))))

C

assert(a < b);
double code(double a, double b) {
	double tmp;
	if (log((exp(a) + exp(b))) <= 5e-10) {
		tmp = 0.5 * b;
	} else {
		tmp = fma(0.5, b, log(2.0));
	}
	return tmp;
}

Julia

a, b = sort([a, b])
function code(a, b)
	tmp = 0.0
	if (log(Float64(exp(a) + exp(b))) <= 5e-10)
		tmp = Float64(0.5 * b);
	else
		tmp = fma(0.5, b, log(2.0));
	end
	return tmp
end

Wolfram

NOTE: a and b should be sorted in increasing order before calling this function.
code[a_, b_] := If[LessEqual[N[Log[N[(N[Exp[a], $MachinePrecision] + N[Exp[b], $MachinePrecision]), $MachinePrecision]], $MachinePrecision], 5e-10], N[(0.5 * b), $MachinePrecision], N[(0.5 * b + N[Log[2.0], $MachinePrecision]), $MachinePrecision]]

Derivation

1. Split input into 2 regimes

2. if (log.f64 (+.f64 (exp.f64 a) (exp.f64 b))) < 5.00000000000000031e-10

   1. Initial program 7.7%

      \[\log \left(e^{a} + e^{b}\right) \]
   2. Add Preprocessing

   3. Taylor expanded in b around 0

      \[\leadsto \color{blue}{\log \left(1 + e^{a}\right) + \frac{b}{1 + e^{a}}} \]
   4. Step-by-step derivation

      1. +-commutative N/A

         \[\leadsto \frac{b}{1 + e^{a}} + \color{blue}{\log \left(1 + e^{a}\right)} \]

      2. lower-+.f64 N/A

         \[\leadsto \frac{b}{1 + e^{a}} + \color{blue}{\log \left(1 + e^{a}\right)} \]

      3. lower-/.f64 N/A

         \[\leadsto \frac{b}{1 + e^{a}} + \log \color{blue}{\left(1 + e^{a}\right)} \]

      4. +-commutative N/A

         \[\leadsto \frac{b}{e^{a} + 1} + \log \left(1 + \color{blue}{e^{a}}\right) \]

      5. metadata-eval N/A

         \[\leadsto \frac{b}{e^{a} + 1 \cdot 1} + \log \left(1 + e^{a}\right) \]

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

         \[\leadsto \frac{b}{e^{a} - \left(\mathsf{neg}\left(1\right)\right) \cdot 1} + \log \left(1 + \color{blue}{e^{a}}\right) \]

      7. metadata-eval N/A

         \[\leadsto \frac{b}{e^{a} - -1 \cdot 1} + \log \left(1 + e^{a}\right) \]

      8. metadata-eval N/A

         \[\leadsto \frac{b}{e^{a} - -1} + \log \left(1 + e^{a}\right) \]

      9. lower--.f64 N/A

         \[\leadsto \frac{b}{e^{a} - -1} + \log \left(1 + \color{blue}{e^{a}}\right) \]

      10. lift-exp.f64 N/A

          \[\leadsto \frac{b}{e^{a} - -1} + \log \left(1 + e^{\color{blue}{a}}\right) \]

      11. lower-log1p.f64 N/A

          \[\leadsto \frac{b}{e^{a} - -1} + \mathsf{log1p}\left(e^{a}\right) \]

      12. lift-exp.f64 100.0

          \[\leadsto \frac{b}{e^{a} - -1} + \mathsf{log1p}\left(e^{a}\right) \]

   5. Applied rewrites 100.0%

      \[\leadsto \color{blue}{\frac{b}{e^{a} - -1} + \mathsf{log1p}\left(e^{a}\right)} \]

   6. Taylor expanded in b around inf

      \[\leadsto \frac{b}{\color{blue}{1 + e^{a}}} \]
   7. Step-by-step derivation

      1. lower-/.f64 N/A

         \[\leadsto \frac{b}{1 + \color{blue}{e^{a}}} \]

      2. lower-+.f64 N/A

         \[\leadsto \frac{b}{1 + e^{a}} \]

      3. lift-exp.f64 99.6

         \[\leadsto \frac{b}{1 + e^{a}} \]

   8. Applied rewrites 99.6%

      \[\leadsto \frac{b}{\color{blue}{1 + e^{a}}} \]

   9. Taylor expanded in a around 0

      \[\leadsto \frac{1}{2} \cdot b \]
   10. Step-by-step derivation

       1. lower-*.f64 18.7

          \[\leadsto 0.5 \cdot b \]

   11. Applied rewrites 18.7%

       \[\leadsto 0.5 \cdot b \]

   if 5.00000000000000031e-10 < (log.f64 (+.f64 (exp.f64 a) (exp.f64 b)))

   1. Initial program 96.8%

      \[\log \left(e^{a} + e^{b}\right) \]
   2. Add Preprocessing

   3. Taylor expanded in b around 0

      \[\leadsto \color{blue}{\log \left(1 + e^{a}\right) + \frac{b}{1 + e^{a}}} \]
   4. Step-by-step derivation

      1. +-commutative N/A

         \[\leadsto \frac{b}{1 + e^{a}} + \color{blue}{\log \left(1 + e^{a}\right)} \]

      2. lower-+.f64 N/A

         \[\leadsto \frac{b}{1 + e^{a}} + \color{blue}{\log \left(1 + e^{a}\right)} \]

      3. lower-/.f64 N/A

         \[\leadsto \frac{b}{1 + e^{a}} + \log \color{blue}{\left(1 + e^{a}\right)} \]

      4. +-commutative N/A

         \[\leadsto \frac{b}{e^{a} + 1} + \log \left(1 + \color{blue}{e^{a}}\right) \]

      5. metadata-eval N/A

         \[\leadsto \frac{b}{e^{a} + 1 \cdot 1} + \log \left(1 + e^{a}\right) \]

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

         \[\leadsto \frac{b}{e^{a} - \left(\mathsf{neg}\left(1\right)\right) \cdot 1} + \log \left(1 + \color{blue}{e^{a}}\right) \]

      7. metadata-eval N/A

         \[\leadsto \frac{b}{e^{a} - -1 \cdot 1} + \log \left(1 + e^{a}\right) \]

      8. metadata-eval N/A

         \[\leadsto \frac{b}{e^{a} - -1} + \log \left(1 + e^{a}\right) \]

      9. lower--.f64 N/A

         \[\leadsto \frac{b}{e^{a} - -1} + \log \left(1 + \color{blue}{e^{a}}\right) \]

      10. lift-exp.f64 N/A

          \[\leadsto \frac{b}{e^{a} - -1} + \log \left(1 + e^{\color{blue}{a}}\right) \]

      11. lower-log1p.f64 N/A

          \[\leadsto \frac{b}{e^{a} - -1} + \mathsf{log1p}\left(e^{a}\right) \]

      12. lift-exp.f64 96.8

          \[\leadsto \frac{b}{e^{a} - -1} + \mathsf{log1p}\left(e^{a}\right) \]

   5. Applied rewrites 96.8%

      \[\leadsto \color{blue}{\frac{b}{e^{a} - -1} + \mathsf{log1p}\left(e^{a}\right)} \]

   6. Taylor expanded in a around 0

      \[\leadsto \log 2 + \color{blue}{\frac{1}{2} \cdot b} \]
   7. Step-by-step derivation

      1. +-commutative N/A

         \[\leadsto \frac{1}{2} \cdot b + \log 2 \]

      2. lower-fma.f64 N/A

         \[\leadsto \mathsf{fma}\left(\frac{1}{2}, b, \log 2\right) \]

      3. lower-log.f64 92.1

         \[\leadsto \mathsf{fma}\left(0.5, b, \log 2\right) \]

   8. Applied rewrites 92.1%

      \[\leadsto \mathsf{fma}\left(0.5, \color{blue}{b}, \log 2\right) \]
3. Recombined 2 regimes into one program.
4. Add Preprocessing

Alternative 3: 98.3% accurate, 1.0× speedup

Math

\[\begin{array}{l} [a, b] = \mathsf{sort}([a, b])\\ \\ \frac{b}{e^{a} - -1} + \mathsf{log1p}\left(e^{a}\right) \end{array} \]

FPCore

NOTE: a and b should be sorted in increasing order before calling this function.
(FPCore (a b) :precision binary64 (+ (/ b (- (exp a) -1.0)) (log1p (exp a))))

C

assert(a < b);
double code(double a, double b) {
	return (b / (exp(a) - -1.0)) + log1p(exp(a));
}

Java

assert a < b;
public static double code(double a, double b) {
	return (b / (Math.exp(a) - -1.0)) + Math.log1p(Math.exp(a));
}

Python

[a, b] = sort([a, b])
def code(a, b):
	return (b / (math.exp(a) - -1.0)) + math.log1p(math.exp(a))

Julia

a, b = sort([a, b])
function code(a, b)
	return Float64(Float64(b / Float64(exp(a) - -1.0)) + log1p(exp(a)))
end

Wolfram

NOTE: a and b should be sorted in increasing order before calling this function.
code[a_, b_] := N[(N[(b / N[(N[Exp[a], $MachinePrecision] - -1.0), $MachinePrecision]), $MachinePrecision] + N[Log[1 + N[Exp[a], $MachinePrecision]], $MachinePrecision]), $MachinePrecision]

Derivation

1. Initial program 54.1%

   \[\log \left(e^{a} + e^{b}\right) \]
2. Add Preprocessing

3. Taylor expanded in b around 0

   \[\leadsto \color{blue}{\log \left(1 + e^{a}\right) + \frac{b}{1 + e^{a}}} \]
4. Step-by-step derivation

   1. +-commutative N/A

      \[\leadsto \frac{b}{1 + e^{a}} + \color{blue}{\log \left(1 + e^{a}\right)} \]

   2. lower-+.f64 N/A

      \[\leadsto \frac{b}{1 + e^{a}} + \color{blue}{\log \left(1 + e^{a}\right)} \]

   3. lower-/.f64 N/A

      \[\leadsto \frac{b}{1 + e^{a}} + \log \color{blue}{\left(1 + e^{a}\right)} \]

   4. +-commutative N/A

      \[\leadsto \frac{b}{e^{a} + 1} + \log \left(1 + \color{blue}{e^{a}}\right) \]

   5. metadata-eval N/A

      \[\leadsto \frac{b}{e^{a} + 1 \cdot 1} + \log \left(1 + e^{a}\right) \]

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

      \[\leadsto \frac{b}{e^{a} - \left(\mathsf{neg}\left(1\right)\right) \cdot 1} + \log \left(1 + \color{blue}{e^{a}}\right) \]

   7. metadata-eval N/A

      \[\leadsto \frac{b}{e^{a} - -1 \cdot 1} + \log \left(1 + e^{a}\right) \]

   8. metadata-eval N/A

      \[\leadsto \frac{b}{e^{a} - -1} + \log \left(1 + e^{a}\right) \]

   9. lower--.f64 N/A

      \[\leadsto \frac{b}{e^{a} - -1} + \log \left(1 + \color{blue}{e^{a}}\right) \]

   10. lift-exp.f64 N/A

       \[\leadsto \frac{b}{e^{a} - -1} + \log \left(1 + e^{\color{blue}{a}}\right) \]

   11. lower-log1p.f64 N/A

       \[\leadsto \frac{b}{e^{a} - -1} + \mathsf{log1p}\left(e^{a}\right) \]

   12. lift-exp.f64 98.3

       \[\leadsto \frac{b}{e^{a} - -1} + \mathsf{log1p}\left(e^{a}\right) \]

5. Applied rewrites 98.3%

   \[\leadsto \color{blue}{\frac{b}{e^{a} - -1} + \mathsf{log1p}\left(e^{a}\right)} \]
6. Add Preprocessing

Alternative 4: 98.2% accurate, 1.4× speedup

Math

\[\begin{array}{l} [a, b] = \mathsf{sort}([a, b])\\ \\ \begin{array}{l} \mathbf{if}\;a \leq -37:\\ \;\;\;\;\frac{b}{1 + e^{a}}\\ \mathbf{else}:\\ \;\;\;\;\log \left(e^{a} + \mathsf{fma}\left(\mathsf{fma}\left(0.5, b, 1\right), b, 1\right)\right)\\ \end{array} \end{array} \]

FPCore

NOTE: a and b should be sorted in increasing order before calling this function.
(FPCore (a b)
 :precision binary64
 (if (<= a -37.0)
   (/ b (+ 1.0 (exp a)))
   (log (+ (exp a) (fma (fma 0.5 b 1.0) b 1.0)))))

C

assert(a < b);
double code(double a, double b) {
	double tmp;
	if (a <= -37.0) {
		tmp = b / (1.0 + exp(a));
	} else {
		tmp = log((exp(a) + fma(fma(0.5, b, 1.0), b, 1.0)));
	}
	return tmp;
}

Julia

a, b = sort([a, b])
function code(a, b)
	tmp = 0.0
	if (a <= -37.0)
		tmp = Float64(b / Float64(1.0 + exp(a)));
	else
		tmp = log(Float64(exp(a) + fma(fma(0.5, b, 1.0), b, 1.0)));
	end
	return tmp
end

Wolfram

NOTE: a and b should be sorted in increasing order before calling this function.
code[a_, b_] := If[LessEqual[a, -37.0], N[(b / N[(1.0 + N[Exp[a], $MachinePrecision]), $MachinePrecision]), $MachinePrecision], N[Log[N[(N[Exp[a], $MachinePrecision] + N[(N[(0.5 * b + 1.0), $MachinePrecision] * b + 1.0), $MachinePrecision]), $MachinePrecision]], $MachinePrecision]]

Derivation

1. Split input into 2 regimes

2. if a < -37

   1. Initial program 9.0%

      \[\log \left(e^{a} + e^{b}\right) \]
   2. Add Preprocessing

   3. Taylor expanded in b around 0

      \[\leadsto \color{blue}{\log \left(1 + e^{a}\right) + \frac{b}{1 + e^{a}}} \]
   4. Step-by-step derivation

      1. +-commutative N/A

         \[\leadsto \frac{b}{1 + e^{a}} + \color{blue}{\log \left(1 + e^{a}\right)} \]

      2. lower-+.f64 N/A

         \[\leadsto \frac{b}{1 + e^{a}} + \color{blue}{\log \left(1 + e^{a}\right)} \]

      3. lower-/.f64 N/A

         \[\leadsto \frac{b}{1 + e^{a}} + \log \color{blue}{\left(1 + e^{a}\right)} \]

      4. +-commutative N/A

         \[\leadsto \frac{b}{e^{a} + 1} + \log \left(1 + \color{blue}{e^{a}}\right) \]

      5. metadata-eval N/A

         \[\leadsto \frac{b}{e^{a} + 1 \cdot 1} + \log \left(1 + e^{a}\right) \]

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

         \[\leadsto \frac{b}{e^{a} - \left(\mathsf{neg}\left(1\right)\right) \cdot 1} + \log \left(1 + \color{blue}{e^{a}}\right) \]

      7. metadata-eval N/A

         \[\leadsto \frac{b}{e^{a} - -1 \cdot 1} + \log \left(1 + e^{a}\right) \]

      8. metadata-eval N/A

         \[\leadsto \frac{b}{e^{a} - -1} + \log \left(1 + e^{a}\right) \]

      9. lower--.f64 N/A

         \[\leadsto \frac{b}{e^{a} - -1} + \log \left(1 + \color{blue}{e^{a}}\right) \]

      10. lift-exp.f64 N/A

          \[\leadsto \frac{b}{e^{a} - -1} + \log \left(1 + e^{\color{blue}{a}}\right) \]

      11. lower-log1p.f64 N/A

          \[\leadsto \frac{b}{e^{a} - -1} + \mathsf{log1p}\left(e^{a}\right) \]

      12. lift-exp.f64 100.0

          \[\leadsto \frac{b}{e^{a} - -1} + \mathsf{log1p}\left(e^{a}\right) \]

   5. Applied rewrites 100.0%

      \[\leadsto \color{blue}{\frac{b}{e^{a} - -1} + \mathsf{log1p}\left(e^{a}\right)} \]

   6. Taylor expanded in b around inf

      \[\leadsto \frac{b}{\color{blue}{1 + e^{a}}} \]
   7. Step-by-step derivation

      1. lower-/.f64 N/A

         \[\leadsto \frac{b}{1 + \color{blue}{e^{a}}} \]

      2. lower-+.f64 N/A

         \[\leadsto \frac{b}{1 + e^{a}} \]

      3. lift-exp.f64 99.8

         \[\leadsto \frac{b}{1 + e^{a}} \]

   8. Applied rewrites 99.8%

      \[\leadsto \frac{b}{\color{blue}{1 + e^{a}}} \]

   if -37 < a

   1. Initial program 97.9%

      \[\log \left(e^{a} + e^{b}\right) \]
   2. Add Preprocessing

   3. Taylor expanded in b around 0

      \[\leadsto \log \left(e^{a} + \color{blue}{\left(1 + b \cdot \left(1 + \frac{1}{2} \cdot b\right)\right)}\right) \]
   4. Step-by-step derivation

      1. +-commutative N/A

         \[\leadsto \log \left(e^{a} + \left(b \cdot \left(1 + \frac{1}{2} \cdot b\right) + \color{blue}{1}\right)\right) \]

      2. *-commutative N/A

         \[\leadsto \log \left(e^{a} + \left(\left(1 + \frac{1}{2} \cdot b\right) \cdot b + 1\right)\right) \]

      3. lower-fma.f64 N/A

         \[\leadsto \log \left(e^{a} + \mathsf{fma}\left(1 + \frac{1}{2} \cdot b, \color{blue}{b}, 1\right)\right) \]

      4. +-commutative N/A

         \[\leadsto \log \left(e^{a} + \mathsf{fma}\left(\frac{1}{2} \cdot b + 1, b, 1\right)\right) \]

      5. lower-fma.f64 96.7

         \[\leadsto \log \left(e^{a} + \mathsf{fma}\left(\mathsf{fma}\left(0.5, b, 1\right), b, 1\right)\right) \]

   5. Applied rewrites 96.7%

      \[\leadsto \log \left(e^{a} + \color{blue}{\mathsf{fma}\left(\mathsf{fma}\left(0.5, b, 1\right), b, 1\right)}\right) \]
3. Recombined 2 regimes into one program.
4. Add Preprocessing

Alternative 5: 98.0% accurate, 1.4× speedup

Math

\[\begin{array}{l} [a, b] = \mathsf{sort}([a, b])\\ \\ \begin{array}{l} \mathbf{if}\;a \leq -37:\\ \;\;\;\;\frac{b}{1 + e^{a}}\\ \mathbf{else}:\\ \;\;\;\;\log \left(e^{a} + \left(b - -1\right)\right)\\ \end{array} \end{array} \]

FPCore

NOTE: a and b should be sorted in increasing order before calling this function.
(FPCore (a b)
 :precision binary64
 (if (<= a -37.0) (/ b (+ 1.0 (exp a))) (log (+ (exp a) (- b -1.0)))))

C

assert(a < b);
double code(double a, double b) {
	double tmp;
	if (a <= -37.0) {
		tmp = b / (1.0 + exp(a));
	} else {
		tmp = log((exp(a) + (b - -1.0)));
	}
	return tmp;
}

Fortran

NOTE: a and b should be sorted in increasing order before calling this function.
module fmin_fmax_functions
    implicit none
    private
    public fmax
    public fmin

    interface fmax
        module procedure fmax88
        module procedure fmax44
        module procedure fmax84
        module procedure fmax48
    end interface
    interface fmin
        module procedure fmin88
        module procedure fmin44
        module procedure fmin84
        module procedure fmin48
    end interface
contains
    real(8) function fmax88(x, y) result (res)
        real(8), intent (in) :: x
        real(8), intent (in) :: y
        res = merge(y, merge(x, max(x, y), y /= y), x /= x)
    end function
    real(4) function fmax44(x, y) result (res)
        real(4), intent (in) :: x
        real(4), intent (in) :: y
        res = merge(y, merge(x, max(x, y), y /= y), x /= x)
    end function
    real(8) function fmax84(x, y) result(res)
        real(8), intent (in) :: x
        real(4), intent (in) :: y
        res = merge(dble(y), merge(x, max(x, dble(y)), y /= y), x /= x)
    end function
    real(8) function fmax48(x, y) result(res)
        real(4), intent (in) :: x
        real(8), intent (in) :: y
        res = merge(y, merge(dble(x), max(dble(x), y), y /= y), x /= x)
    end function
    real(8) function fmin88(x, y) result (res)
        real(8), intent (in) :: x
        real(8), intent (in) :: y
        res = merge(y, merge(x, min(x, y), y /= y), x /= x)
    end function
    real(4) function fmin44(x, y) result (res)
        real(4), intent (in) :: x
        real(4), intent (in) :: y
        res = merge(y, merge(x, min(x, y), y /= y), x /= x)
    end function
    real(8) function fmin84(x, y) result(res)
        real(8), intent (in) :: x
        real(4), intent (in) :: y
        res = merge(dble(y), merge(x, min(x, dble(y)), y /= y), x /= x)
    end function
    real(8) function fmin48(x, y) result(res)
        real(4), intent (in) :: x
        real(8), intent (in) :: y
        res = merge(y, merge(dble(x), min(dble(x), y), y /= y), x /= x)
    end function
end module

real(8) function code(a, b)
use fmin_fmax_functions
    real(8), intent (in) :: a
    real(8), intent (in) :: b
    real(8) :: tmp
    if (a <= (-37.0d0)) then
        tmp = b / (1.0d0 + exp(a))
    else
        tmp = log((exp(a) + (b - (-1.0d0))))
    end if
    code = tmp
end function

Java

assert a < b;
public static double code(double a, double b) {
	double tmp;
	if (a <= -37.0) {
		tmp = b / (1.0 + Math.exp(a));
	} else {
		tmp = Math.log((Math.exp(a) + (b - -1.0)));
	}
	return tmp;
}

Python

[a, b] = sort([a, b])
def code(a, b):
	tmp = 0
	if a <= -37.0:
		tmp = b / (1.0 + math.exp(a))
	else:
		tmp = math.log((math.exp(a) + (b - -1.0)))
	return tmp

Julia

a, b = sort([a, b])
function code(a, b)
	tmp = 0.0
	if (a <= -37.0)
		tmp = Float64(b / Float64(1.0 + exp(a)));
	else
		tmp = log(Float64(exp(a) + Float64(b - -1.0)));
	end
	return tmp
end

MATLAB

a, b = num2cell(sort([a, b])){:}
function tmp_2 = code(a, b)
	tmp = 0.0;
	if (a <= -37.0)
		tmp = b / (1.0 + exp(a));
	else
		tmp = log((exp(a) + (b - -1.0)));
	end
	tmp_2 = tmp;
end

Wolfram

NOTE: a and b should be sorted in increasing order before calling this function.
code[a_, b_] := If[LessEqual[a, -37.0], N[(b / N[(1.0 + N[Exp[a], $MachinePrecision]), $MachinePrecision]), $MachinePrecision], N[Log[N[(N[Exp[a], $MachinePrecision] + N[(b - -1.0), $MachinePrecision]), $MachinePrecision]], $MachinePrecision]]

Derivation

1. Split input into 2 regimes

2. if a < -37

   1. Initial program 9.0%

      \[\log \left(e^{a} + e^{b}\right) \]
   2. Add Preprocessing

   3. Taylor expanded in b around 0

      \[\leadsto \color{blue}{\log \left(1 + e^{a}\right) + \frac{b}{1 + e^{a}}} \]
   4. Step-by-step derivation

      1. +-commutative N/A

         \[\leadsto \frac{b}{1 + e^{a}} + \color{blue}{\log \left(1 + e^{a}\right)} \]

      2. lower-+.f64 N/A

         \[\leadsto \frac{b}{1 + e^{a}} + \color{blue}{\log \left(1 + e^{a}\right)} \]

      3. lower-/.f64 N/A

         \[\leadsto \frac{b}{1 + e^{a}} + \log \color{blue}{\left(1 + e^{a}\right)} \]

      4. +-commutative N/A

         \[\leadsto \frac{b}{e^{a} + 1} + \log \left(1 + \color{blue}{e^{a}}\right) \]

      5. metadata-eval N/A

         \[\leadsto \frac{b}{e^{a} + 1 \cdot 1} + \log \left(1 + e^{a}\right) \]

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

         \[\leadsto \frac{b}{e^{a} - \left(\mathsf{neg}\left(1\right)\right) \cdot 1} + \log \left(1 + \color{blue}{e^{a}}\right) \]

      7. metadata-eval N/A

         \[\leadsto \frac{b}{e^{a} - -1 \cdot 1} + \log \left(1 + e^{a}\right) \]

      8. metadata-eval N/A

         \[\leadsto \frac{b}{e^{a} - -1} + \log \left(1 + e^{a}\right) \]

      9. lower--.f64 N/A

         \[\leadsto \frac{b}{e^{a} - -1} + \log \left(1 + \color{blue}{e^{a}}\right) \]

      10. lift-exp.f64 N/A

          \[\leadsto \frac{b}{e^{a} - -1} + \log \left(1 + e^{\color{blue}{a}}\right) \]

      11. lower-log1p.f64 N/A

          \[\leadsto \frac{b}{e^{a} - -1} + \mathsf{log1p}\left(e^{a}\right) \]

      12. lift-exp.f64 100.0

          \[\leadsto \frac{b}{e^{a} - -1} + \mathsf{log1p}\left(e^{a}\right) \]

   5. Applied rewrites 100.0%

      \[\leadsto \color{blue}{\frac{b}{e^{a} - -1} + \mathsf{log1p}\left(e^{a}\right)} \]

   6. Taylor expanded in b around inf

      \[\leadsto \frac{b}{\color{blue}{1 + e^{a}}} \]
   7. Step-by-step derivation

      1. lower-/.f64 N/A

         \[\leadsto \frac{b}{1 + \color{blue}{e^{a}}} \]

      2. lower-+.f64 N/A

         \[\leadsto \frac{b}{1 + e^{a}} \]

      3. lift-exp.f64 99.8

         \[\leadsto \frac{b}{1 + e^{a}} \]

   8. Applied rewrites 99.8%

      \[\leadsto \frac{b}{\color{blue}{1 + e^{a}}} \]

   if -37 < a

   1. Initial program 97.9%

      \[\log \left(e^{a} + e^{b}\right) \]
   2. Add Preprocessing

   3. Taylor expanded in b around 0

      \[\leadsto \log \left(e^{a} + \color{blue}{\left(1 + b\right)}\right) \]
   4. Step-by-step derivation

      1. +-commutative N/A

         \[\leadsto \log \left(e^{a} + \left(b + \color{blue}{1}\right)\right) \]

      2. metadata-eval N/A

         \[\leadsto \log \left(e^{a} + \left(b + 1 \cdot \color{blue}{1}\right)\right) \]

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

         \[\leadsto \log \left(e^{a} + \left(b - \color{blue}{\left(\mathsf{neg}\left(1\right)\right) \cdot 1}\right)\right) \]

      4. metadata-eval N/A

         \[\leadsto \log \left(e^{a} + \left(b - -1 \cdot 1\right)\right) \]

      5. metadata-eval N/A

         \[\leadsto \log \left(e^{a} + \left(b - -1\right)\right) \]

      6. lower--.f64 96.3

         \[\leadsto \log \left(e^{a} + \left(b - \color{blue}{-1}\right)\right) \]

   5. Applied rewrites 96.3%

      \[\leadsto \log \left(e^{a} + \color{blue}{\left(b - -1\right)}\right) \]
3. Recombined 2 regimes into one program.
4. Add Preprocessing

Alternative 6: 97.4% accurate, 2.4× speedup

Math

\[\begin{array}{l} [a, b] = \mathsf{sort}([a, b])\\ \\ \begin{array}{l} \mathbf{if}\;a \leq -1.4:\\ \;\;\;\;\frac{b}{1 + e^{a}}\\ \mathbf{else}:\\ \;\;\;\;\mathsf{fma}\left(0.5 - 0.25 \cdot b, a, 0.5 \cdot b\right) + \log 2\\ \end{array} \end{array} \]

FPCore

NOTE: a and b should be sorted in increasing order before calling this function.
(FPCore (a b)
 :precision binary64
 (if (<= a -1.4)
   (/ b (+ 1.0 (exp a)))
   (+ (fma (- 0.5 (* 0.25 b)) a (* 0.5 b)) (log 2.0))))

C

assert(a < b);
double code(double a, double b) {
	double tmp;
	if (a <= -1.4) {
		tmp = b / (1.0 + exp(a));
	} else {
		tmp = fma((0.5 - (0.25 * b)), a, (0.5 * b)) + log(2.0);
	}
	return tmp;
}

Julia

a, b = sort([a, b])
function code(a, b)
	tmp = 0.0
	if (a <= -1.4)
		tmp = Float64(b / Float64(1.0 + exp(a)));
	else
		tmp = Float64(fma(Float64(0.5 - Float64(0.25 * b)), a, Float64(0.5 * b)) + log(2.0));
	end
	return tmp
end

Wolfram

NOTE: a and b should be sorted in increasing order before calling this function.
code[a_, b_] := If[LessEqual[a, -1.4], N[(b / N[(1.0 + N[Exp[a], $MachinePrecision]), $MachinePrecision]), $MachinePrecision], N[(N[(N[(0.5 - N[(0.25 * b), $MachinePrecision]), $MachinePrecision] * a + N[(0.5 * b), $MachinePrecision]), $MachinePrecision] + N[Log[2.0], $MachinePrecision]), $MachinePrecision]]

Derivation

1. Split input into 2 regimes

2. if a < -1.3999999999999999

   1. Initial program 9.4%

      \[\log \left(e^{a} + e^{b}\right) \]
   2. Add Preprocessing

   3. Taylor expanded in b around 0

      \[\leadsto \color{blue}{\log \left(1 + e^{a}\right) + \frac{b}{1 + e^{a}}} \]
   4. Step-by-step derivation

      1. +-commutative N/A

         \[\leadsto \frac{b}{1 + e^{a}} + \color{blue}{\log \left(1 + e^{a}\right)} \]

      2. lower-+.f64 N/A

         \[\leadsto \frac{b}{1 + e^{a}} + \color{blue}{\log \left(1 + e^{a}\right)} \]

      3. lower-/.f64 N/A

         \[\leadsto \frac{b}{1 + e^{a}} + \log \color{blue}{\left(1 + e^{a}\right)} \]

      4. +-commutative N/A

         \[\leadsto \frac{b}{e^{a} + 1} + \log \left(1 + \color{blue}{e^{a}}\right) \]

      5. metadata-eval N/A

         \[\leadsto \frac{b}{e^{a} + 1 \cdot 1} + \log \left(1 + e^{a}\right) \]

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

         \[\leadsto \frac{b}{e^{a} - \left(\mathsf{neg}\left(1\right)\right) \cdot 1} + \log \left(1 + \color{blue}{e^{a}}\right) \]

      7. metadata-eval N/A

         \[\leadsto \frac{b}{e^{a} - -1 \cdot 1} + \log \left(1 + e^{a}\right) \]

      8. metadata-eval N/A

         \[\leadsto \frac{b}{e^{a} - -1} + \log \left(1 + e^{a}\right) \]

      9. lower--.f64 N/A

         \[\leadsto \frac{b}{e^{a} - -1} + \log \left(1 + \color{blue}{e^{a}}\right) \]

      10. lift-exp.f64 N/A

          \[\leadsto \frac{b}{e^{a} - -1} + \log \left(1 + e^{\color{blue}{a}}\right) \]

      11. lower-log1p.f64 N/A

          \[\leadsto \frac{b}{e^{a} - -1} + \mathsf{log1p}\left(e^{a}\right) \]

      12. lift-exp.f64 100.0

          \[\leadsto \frac{b}{e^{a} - -1} + \mathsf{log1p}\left(e^{a}\right) \]

   5. Applied rewrites 100.0%

      \[\leadsto \color{blue}{\frac{b}{e^{a} - -1} + \mathsf{log1p}\left(e^{a}\right)} \]

   6. Taylor expanded in b around inf

      \[\leadsto \frac{b}{\color{blue}{1 + e^{a}}} \]
   7. Step-by-step derivation

      1. lower-/.f64 N/A

         \[\leadsto \frac{b}{1 + \color{blue}{e^{a}}} \]

      2. lower-+.f64 N/A

         \[\leadsto \frac{b}{1 + e^{a}} \]

      3. lift-exp.f64 99.2

         \[\leadsto \frac{b}{1 + e^{a}} \]

   8. Applied rewrites 99.2%

      \[\leadsto \frac{b}{\color{blue}{1 + e^{a}}} \]

   if -1.3999999999999999 < a

   1. Initial program 98.0%

      \[\log \left(e^{a} + e^{b}\right) \]
   2. Add Preprocessing

   3. Taylor expanded in b around 0

      \[\leadsto \color{blue}{\log \left(1 + e^{a}\right) + \frac{b}{1 + e^{a}}} \]
   4. Step-by-step derivation

      1. +-commutative N/A

         \[\leadsto \frac{b}{1 + e^{a}} + \color{blue}{\log \left(1 + e^{a}\right)} \]

      2. lower-+.f64 N/A

         \[\leadsto \frac{b}{1 + e^{a}} + \color{blue}{\log \left(1 + e^{a}\right)} \]

      3. lower-/.f64 N/A

         \[\leadsto \frac{b}{1 + e^{a}} + \log \color{blue}{\left(1 + e^{a}\right)} \]

      4. +-commutative N/A

         \[\leadsto \frac{b}{e^{a} + 1} + \log \left(1 + \color{blue}{e^{a}}\right) \]

      5. metadata-eval N/A

         \[\leadsto \frac{b}{e^{a} + 1 \cdot 1} + \log \left(1 + e^{a}\right) \]

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

         \[\leadsto \frac{b}{e^{a} - \left(\mathsf{neg}\left(1\right)\right) \cdot 1} + \log \left(1 + \color{blue}{e^{a}}\right) \]

      7. metadata-eval N/A

         \[\leadsto \frac{b}{e^{a} - -1 \cdot 1} + \log \left(1 + e^{a}\right) \]

      8. metadata-eval N/A

         \[\leadsto \frac{b}{e^{a} - -1} + \log \left(1 + e^{a}\right) \]

      9. lower--.f64 N/A

         \[\leadsto \frac{b}{e^{a} - -1} + \log \left(1 + \color{blue}{e^{a}}\right) \]

      10. lift-exp.f64 N/A

          \[\leadsto \frac{b}{e^{a} - -1} + \log \left(1 + e^{\color{blue}{a}}\right) \]

      11. lower-log1p.f64 N/A

          \[\leadsto \frac{b}{e^{a} - -1} + \mathsf{log1p}\left(e^{a}\right) \]

      12. lift-exp.f64 96.7

          \[\leadsto \frac{b}{e^{a} - -1} + \mathsf{log1p}\left(e^{a}\right) \]

   5. Applied rewrites 96.7%

      \[\leadsto \color{blue}{\frac{b}{e^{a} - -1} + \mathsf{log1p}\left(e^{a}\right)} \]

   6. Taylor expanded in a around 0

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

      1. +-commutative N/A

         \[\leadsto \left(\frac{1}{2} \cdot b + a \cdot \left(\frac{1}{2} - \frac{1}{4} \cdot b\right)\right) + \log 2 \]

      2. lower-+.f64 N/A

         \[\leadsto \left(\frac{1}{2} \cdot b + a \cdot \left(\frac{1}{2} - \frac{1}{4} \cdot b\right)\right) + \log 2 \]

      3. +-commutative N/A

         \[\leadsto \left(a \cdot \left(\frac{1}{2} - \frac{1}{4} \cdot b\right) + \frac{1}{2} \cdot b\right) + \log 2 \]

      4. *-commutative N/A

         \[\leadsto \left(\left(\frac{1}{2} - \frac{1}{4} \cdot b\right) \cdot a + \frac{1}{2} \cdot b\right) + \log 2 \]

      5. lower-fma.f64 N/A

         \[\leadsto \mathsf{fma}\left(\frac{1}{2} - \frac{1}{4} \cdot b, a, \frac{1}{2} \cdot b\right) + \log 2 \]

      6. lower--.f64 N/A

         \[\leadsto \mathsf{fma}\left(\frac{1}{2} - \frac{1}{4} \cdot b, a, \frac{1}{2} \cdot b\right) + \log 2 \]

      7. lower-*.f64 N/A

         \[\leadsto \mathsf{fma}\left(\frac{1}{2} - \frac{1}{4} \cdot b, a, \frac{1}{2} \cdot b\right) + \log 2 \]

      8. lower-*.f64 N/A

         \[\leadsto \mathsf{fma}\left(\frac{1}{2} - \frac{1}{4} \cdot b, a, \frac{1}{2} \cdot b\right) + \log 2 \]

      9. lower-log.f64 95.7

         \[\leadsto \mathsf{fma}\left(0.5 - 0.25 \cdot b, a, 0.5 \cdot b\right) + \log 2 \]

   8. Applied rewrites 95.7%

      \[\leadsto \mathsf{fma}\left(0.5 - 0.25 \cdot b, a, 0.5 \cdot b\right) + \color{blue}{\log 2} \]
3. Recombined 2 regimes into one program.
4. Add Preprocessing

Alternative 7: 97.1% accurate, 2.5× speedup

Math

\[\begin{array}{l} [a, b] = \mathsf{sort}([a, b])\\ \\ \begin{array}{l} \mathbf{if}\;a \leq -56:\\ \;\;\;\;\frac{b}{1 + e^{a}}\\ \mathbf{else}:\\ \;\;\;\;\mathsf{fma}\left(\mathsf{fma}\left(0.125, b, 0.5\right), b, \log 2\right)\\ \end{array} \end{array} \]

FPCore

NOTE: a and b should be sorted in increasing order before calling this function.
(FPCore (a b)
 :precision binary64
 (if (<= a -56.0) (/ b (+ 1.0 (exp a))) (fma (fma 0.125 b 0.5) b (log 2.0))))

C

assert(a < b);
double code(double a, double b) {
	double tmp;
	if (a <= -56.0) {
		tmp = b / (1.0 + exp(a));
	} else {
		tmp = fma(fma(0.125, b, 0.5), b, log(2.0));
	}
	return tmp;
}

Julia

a, b = sort([a, b])
function code(a, b)
	tmp = 0.0
	if (a <= -56.0)
		tmp = Float64(b / Float64(1.0 + exp(a)));
	else
		tmp = fma(fma(0.125, b, 0.5), b, log(2.0));
	end
	return tmp
end

Wolfram

NOTE: a and b should be sorted in increasing order before calling this function.
code[a_, b_] := If[LessEqual[a, -56.0], N[(b / N[(1.0 + N[Exp[a], $MachinePrecision]), $MachinePrecision]), $MachinePrecision], N[(N[(0.125 * b + 0.5), $MachinePrecision] * b + N[Log[2.0], $MachinePrecision]), $MachinePrecision]]

Derivation

1. Split input into 2 regimes

2. if a < -56

   1. Initial program 9.0%

      \[\log \left(e^{a} + e^{b}\right) \]
   2. Add Preprocessing

   3. Taylor expanded in b around 0

      \[\leadsto \color{blue}{\log \left(1 + e^{a}\right) + \frac{b}{1 + e^{a}}} \]
   4. Step-by-step derivation

      1. +-commutative N/A

         \[\leadsto \frac{b}{1 + e^{a}} + \color{blue}{\log \left(1 + e^{a}\right)} \]

      2. lower-+.f64 N/A

         \[\leadsto \frac{b}{1 + e^{a}} + \color{blue}{\log \left(1 + e^{a}\right)} \]

      3. lower-/.f64 N/A

         \[\leadsto \frac{b}{1 + e^{a}} + \log \color{blue}{\left(1 + e^{a}\right)} \]

      4. +-commutative N/A

         \[\leadsto \frac{b}{e^{a} + 1} + \log \left(1 + \color{blue}{e^{a}}\right) \]

      5. metadata-eval N/A

         \[\leadsto \frac{b}{e^{a} + 1 \cdot 1} + \log \left(1 + e^{a}\right) \]

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

         \[\leadsto \frac{b}{e^{a} - \left(\mathsf{neg}\left(1\right)\right) \cdot 1} + \log \left(1 + \color{blue}{e^{a}}\right) \]

      7. metadata-eval N/A

         \[\leadsto \frac{b}{e^{a} - -1 \cdot 1} + \log \left(1 + e^{a}\right) \]

      8. metadata-eval N/A

         \[\leadsto \frac{b}{e^{a} - -1} + \log \left(1 + e^{a}\right) \]

      9. lower--.f64 N/A

         \[\leadsto \frac{b}{e^{a} - -1} + \log \left(1 + \color{blue}{e^{a}}\right) \]

      10. lift-exp.f64 N/A

          \[\leadsto \frac{b}{e^{a} - -1} + \log \left(1 + e^{\color{blue}{a}}\right) \]

      11. lower-log1p.f64 N/A

          \[\leadsto \frac{b}{e^{a} - -1} + \mathsf{log1p}\left(e^{a}\right) \]

      12. lift-exp.f64 100.0

          \[\leadsto \frac{b}{e^{a} - -1} + \mathsf{log1p}\left(e^{a}\right) \]

   5. Applied rewrites 100.0%

      \[\leadsto \color{blue}{\frac{b}{e^{a} - -1} + \mathsf{log1p}\left(e^{a}\right)} \]

   6. Taylor expanded in b around inf

      \[\leadsto \frac{b}{\color{blue}{1 + e^{a}}} \]
   7. Step-by-step derivation

      1. lower-/.f64 N/A

         \[\leadsto \frac{b}{1 + \color{blue}{e^{a}}} \]

      2. lower-+.f64 N/A

         \[\leadsto \frac{b}{1 + e^{a}} \]

      3. lift-exp.f64 99.8

         \[\leadsto \frac{b}{1 + e^{a}} \]

   8. Applied rewrites 99.8%

      \[\leadsto \frac{b}{\color{blue}{1 + e^{a}}} \]

   if -56 < a

   1. Initial program 97.9%

      \[\log \left(e^{a} + e^{b}\right) \]
   2. Add Preprocessing

   3. Taylor expanded in a around 0

      \[\leadsto \color{blue}{\log \left(1 + e^{b}\right)} \]
   4. Step-by-step derivation

      1. lower-log1p.f64 N/A

         \[\leadsto \mathsf{log1p}\left(e^{b}\right) \]

      2. lift-exp.f64 95.5

         \[\leadsto \mathsf{log1p}\left(e^{b}\right) \]

   5. Applied rewrites 95.5%

      \[\leadsto \color{blue}{\mathsf{log1p}\left(e^{b}\right)} \]

   6. Taylor expanded in b around 0

      \[\leadsto \log 2 + \color{blue}{b \cdot \left(\frac{1}{2} + \frac{1}{8} \cdot b\right)} \]
   7. Step-by-step derivation

      1. +-commutative N/A

         \[\leadsto b \cdot \left(\frac{1}{2} + \frac{1}{8} \cdot b\right) + \log 2 \]

      2. *-commutative N/A

         \[\leadsto \left(\frac{1}{2} + \frac{1}{8} \cdot b\right) \cdot b + \log 2 \]

      3. lower-fma.f64 N/A

         \[\leadsto \mathsf{fma}\left(\frac{1}{2} + \frac{1}{8} \cdot b, b, \log 2\right) \]

      4. +-commutative N/A

         \[\leadsto \mathsf{fma}\left(\frac{1}{8} \cdot b + \frac{1}{2}, b, \log 2\right) \]

      5. lower-fma.f64 N/A

         \[\leadsto \mathsf{fma}\left(\mathsf{fma}\left(\frac{1}{8}, b, \frac{1}{2}\right), b, \log 2\right) \]

      6. lower-log.f64 94.6

         \[\leadsto \mathsf{fma}\left(\mathsf{fma}\left(0.125, b, 0.5\right), b, \log 2\right) \]

   8. Applied rewrites 94.6%

      \[\leadsto \mathsf{fma}\left(\mathsf{fma}\left(0.125, b, 0.5\right), \color{blue}{b}, \log 2\right) \]
3. Recombined 2 regimes into one program.
4. Add Preprocessing

Alternative 8: 57.2% accurate, 2.6× speedup

Math

\[\begin{array}{l} [a, b] = \mathsf{sort}([a, b])\\ \\ \begin{array}{l} \mathbf{if}\;a \leq -195:\\ \;\;\;\;0.5 \cdot b\\ \mathbf{else}:\\ \;\;\;\;\mathsf{fma}\left(\mathsf{fma}\left(0.125, b, 0.5\right), b, \log 2\right)\\ \end{array} \end{array} \]

FPCore

NOTE: a and b should be sorted in increasing order before calling this function.
(FPCore (a b)
 :precision binary64
 (if (<= a -195.0) (* 0.5 b) (fma (fma 0.125 b 0.5) b (log 2.0))))

C

assert(a < b);
double code(double a, double b) {
	double tmp;
	if (a <= -195.0) {
		tmp = 0.5 * b;
	} else {
		tmp = fma(fma(0.125, b, 0.5), b, log(2.0));
	}
	return tmp;
}

Julia

a, b = sort([a, b])
function code(a, b)
	tmp = 0.0
	if (a <= -195.0)
		tmp = Float64(0.5 * b);
	else
		tmp = fma(fma(0.125, b, 0.5), b, log(2.0));
	end
	return tmp
end

Wolfram

NOTE: a and b should be sorted in increasing order before calling this function.
code[a_, b_] := If[LessEqual[a, -195.0], N[(0.5 * b), $MachinePrecision], N[(N[(0.125 * b + 0.5), $MachinePrecision] * b + N[Log[2.0], $MachinePrecision]), $MachinePrecision]]

Derivation

1. Split input into 2 regimes

2. if a < -195

   1. Initial program 9.0%

      \[\log \left(e^{a} + e^{b}\right) \]
   2. Add Preprocessing

   3. Taylor expanded in b around 0

      \[\leadsto \color{blue}{\log \left(1 + e^{a}\right) + \frac{b}{1 + e^{a}}} \]
   4. Step-by-step derivation

      1. +-commutative N/A

         \[\leadsto \frac{b}{1 + e^{a}} + \color{blue}{\log \left(1 + e^{a}\right)} \]

      2. lower-+.f64 N/A

         \[\leadsto \frac{b}{1 + e^{a}} + \color{blue}{\log \left(1 + e^{a}\right)} \]

      3. lower-/.f64 N/A

         \[\leadsto \frac{b}{1 + e^{a}} + \log \color{blue}{\left(1 + e^{a}\right)} \]

      4. +-commutative N/A

         \[\leadsto \frac{b}{e^{a} + 1} + \log \left(1 + \color{blue}{e^{a}}\right) \]

      5. metadata-eval N/A

         \[\leadsto \frac{b}{e^{a} + 1 \cdot 1} + \log \left(1 + e^{a}\right) \]

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

         \[\leadsto \frac{b}{e^{a} - \left(\mathsf{neg}\left(1\right)\right) \cdot 1} + \log \left(1 + \color{blue}{e^{a}}\right) \]

      7. metadata-eval N/A

         \[\leadsto \frac{b}{e^{a} - -1 \cdot 1} + \log \left(1 + e^{a}\right) \]

      8. metadata-eval N/A

         \[\leadsto \frac{b}{e^{a} - -1} + \log \left(1 + e^{a}\right) \]

      9. lower--.f64 N/A

         \[\leadsto \frac{b}{e^{a} - -1} + \log \left(1 + \color{blue}{e^{a}}\right) \]

      10. lift-exp.f64 N/A

          \[\leadsto \frac{b}{e^{a} - -1} + \log \left(1 + e^{\color{blue}{a}}\right) \]

      11. lower-log1p.f64 N/A

          \[\leadsto \frac{b}{e^{a} - -1} + \mathsf{log1p}\left(e^{a}\right) \]

      12. lift-exp.f64 100.0

          \[\leadsto \frac{b}{e^{a} - -1} + \mathsf{log1p}\left(e^{a}\right) \]

   5. Applied rewrites 100.0%

      \[\leadsto \color{blue}{\frac{b}{e^{a} - -1} + \mathsf{log1p}\left(e^{a}\right)} \]

   6. Taylor expanded in b around inf

      \[\leadsto \frac{b}{\color{blue}{1 + e^{a}}} \]
   7. Step-by-step derivation

      1. lower-/.f64 N/A

         \[\leadsto \frac{b}{1 + \color{blue}{e^{a}}} \]

      2. lower-+.f64 N/A

         \[\leadsto \frac{b}{1 + e^{a}} \]

      3. lift-exp.f64 100.0

         \[\leadsto \frac{b}{1 + e^{a}} \]

   8. Applied rewrites 100.0%

      \[\leadsto \frac{b}{\color{blue}{1 + e^{a}}} \]

   9. Taylor expanded in a around 0

      \[\leadsto \frac{1}{2} \cdot b \]
   10. Step-by-step derivation

       1. lower-*.f64 18.7

          \[\leadsto 0.5 \cdot b \]

   11. Applied rewrites 18.7%

       \[\leadsto 0.5 \cdot b \]

   if -195 < a

   1. Initial program 97.7%

      \[\log \left(e^{a} + e^{b}\right) \]
   2. Add Preprocessing

   3. Taylor expanded in a around 0

      \[\leadsto \color{blue}{\log \left(1 + e^{b}\right)} \]
   4. Step-by-step derivation

      1. lower-log1p.f64 N/A

         \[\leadsto \mathsf{log1p}\left(e^{b}\right) \]

      2. lift-exp.f64 95.3

         \[\leadsto \mathsf{log1p}\left(e^{b}\right) \]

   5. Applied rewrites 95.3%

      \[\leadsto \color{blue}{\mathsf{log1p}\left(e^{b}\right)} \]

   6. Taylor expanded in b around 0

      \[\leadsto \log 2 + \color{blue}{b \cdot \left(\frac{1}{2} + \frac{1}{8} \cdot b\right)} \]
   7. Step-by-step derivation

      1. +-commutative N/A

         \[\leadsto b \cdot \left(\frac{1}{2} + \frac{1}{8} \cdot b\right) + \log 2 \]

      2. *-commutative N/A

         \[\leadsto \left(\frac{1}{2} + \frac{1}{8} \cdot b\right) \cdot b + \log 2 \]

      3. lower-fma.f64 N/A

         \[\leadsto \mathsf{fma}\left(\frac{1}{2} + \frac{1}{8} \cdot b, b, \log 2\right) \]

      4. +-commutative N/A

         \[\leadsto \mathsf{fma}\left(\frac{1}{8} \cdot b + \frac{1}{2}, b, \log 2\right) \]

      5. lower-fma.f64 N/A

         \[\leadsto \mathsf{fma}\left(\mathsf{fma}\left(\frac{1}{8}, b, \frac{1}{2}\right), b, \log 2\right) \]

      6. lower-log.f64 94.4

         \[\leadsto \mathsf{fma}\left(\mathsf{fma}\left(0.125, b, 0.5\right), b, \log 2\right) \]

   8. Applied rewrites 94.4%

      \[\leadsto \mathsf{fma}\left(\mathsf{fma}\left(0.125, b, 0.5\right), \color{blue}{b}, \log 2\right) \]
3. Recombined 2 regimes into one program.
4. Add Preprocessing

Alternative 9: 57.4% accurate, 2.6× speedup

Math

\[\begin{array}{l} [a, b] = \mathsf{sort}([a, b])\\ \\ \begin{array}{l} \mathbf{if}\;a \leq -1:\\ \;\;\;\;0.5 \cdot b\\ \mathbf{else}:\\ \;\;\;\;\log \left(\left(1 + a\right) + \left(b - -1\right)\right)\\ \end{array} \end{array} \]

FPCore

NOTE: a and b should be sorted in increasing order before calling this function.
(FPCore (a b)
 :precision binary64
 (if (<= a -1.0) (* 0.5 b) (log (+ (+ 1.0 a) (- b -1.0)))))

C

assert(a < b);
double code(double a, double b) {
	double tmp;
	if (a <= -1.0) {
		tmp = 0.5 * b;
	} else {
		tmp = log(((1.0 + a) + (b - -1.0)));
	}
	return tmp;
}

Fortran

NOTE: a and b should be sorted in increasing order before calling this function.
module fmin_fmax_functions
    implicit none
    private
    public fmax
    public fmin

    interface fmax
        module procedure fmax88
        module procedure fmax44
        module procedure fmax84
        module procedure fmax48
    end interface
    interface fmin
        module procedure fmin88
        module procedure fmin44
        module procedure fmin84
        module procedure fmin48
    end interface
contains
    real(8) function fmax88(x, y) result (res)
        real(8), intent (in) :: x
        real(8), intent (in) :: y
        res = merge(y, merge(x, max(x, y), y /= y), x /= x)
    end function
    real(4) function fmax44(x, y) result (res)
        real(4), intent (in) :: x
        real(4), intent (in) :: y
        res = merge(y, merge(x, max(x, y), y /= y), x /= x)
    end function
    real(8) function fmax84(x, y) result(res)
        real(8), intent (in) :: x
        real(4), intent (in) :: y
        res = merge(dble(y), merge(x, max(x, dble(y)), y /= y), x /= x)
    end function
    real(8) function fmax48(x, y) result(res)
        real(4), intent (in) :: x
        real(8), intent (in) :: y
        res = merge(y, merge(dble(x), max(dble(x), y), y /= y), x /= x)
    end function
    real(8) function fmin88(x, y) result (res)
        real(8), intent (in) :: x
        real(8), intent (in) :: y
        res = merge(y, merge(x, min(x, y), y /= y), x /= x)
    end function
    real(4) function fmin44(x, y) result (res)
        real(4), intent (in) :: x
        real(4), intent (in) :: y
        res = merge(y, merge(x, min(x, y), y /= y), x /= x)
    end function
    real(8) function fmin84(x, y) result(res)
        real(8), intent (in) :: x
        real(4), intent (in) :: y
        res = merge(dble(y), merge(x, min(x, dble(y)), y /= y), x /= x)
    end function
    real(8) function fmin48(x, y) result(res)
        real(4), intent (in) :: x
        real(8), intent (in) :: y
        res = merge(y, merge(dble(x), min(dble(x), y), y /= y), x /= x)
    end function
end module

real(8) function code(a, b)
use fmin_fmax_functions
    real(8), intent (in) :: a
    real(8), intent (in) :: b
    real(8) :: tmp
    if (a <= (-1.0d0)) then
        tmp = 0.5d0 * b
    else
        tmp = log(((1.0d0 + a) + (b - (-1.0d0))))
    end if
    code = tmp
end function

Java

assert a < b;
public static double code(double a, double b) {
	double tmp;
	if (a <= -1.0) {
		tmp = 0.5 * b;
	} else {
		tmp = Math.log(((1.0 + a) + (b - -1.0)));
	}
	return tmp;
}

Python

[a, b] = sort([a, b])
def code(a, b):
	tmp = 0
	if a <= -1.0:
		tmp = 0.5 * b
	else:
		tmp = math.log(((1.0 + a) + (b - -1.0)))
	return tmp

Julia

a, b = sort([a, b])
function code(a, b)
	tmp = 0.0
	if (a <= -1.0)
		tmp = Float64(0.5 * b);
	else
		tmp = log(Float64(Float64(1.0 + a) + Float64(b - -1.0)));
	end
	return tmp
end

MATLAB

a, b = num2cell(sort([a, b])){:}
function tmp_2 = code(a, b)
	tmp = 0.0;
	if (a <= -1.0)
		tmp = 0.5 * b;
	else
		tmp = log(((1.0 + a) + (b - -1.0)));
	end
	tmp_2 = tmp;
end

Wolfram

NOTE: a and b should be sorted in increasing order before calling this function.
code[a_, b_] := If[LessEqual[a, -1.0], N[(0.5 * b), $MachinePrecision], N[Log[N[(N[(1.0 + a), $MachinePrecision] + N[(b - -1.0), $MachinePrecision]), $MachinePrecision]], $MachinePrecision]]

Derivation

1. Split input into 2 regimes

2. if a < -1

   1. Initial program 9.4%

      \[\log \left(e^{a} + e^{b}\right) \]
   2. Add Preprocessing

   3. Taylor expanded in b around 0

      \[\leadsto \color{blue}{\log \left(1 + e^{a}\right) + \frac{b}{1 + e^{a}}} \]
   4. Step-by-step derivation

      1. +-commutative N/A

         \[\leadsto \frac{b}{1 + e^{a}} + \color{blue}{\log \left(1 + e^{a}\right)} \]

      2. lower-+.f64 N/A

         \[\leadsto \frac{b}{1 + e^{a}} + \color{blue}{\log \left(1 + e^{a}\right)} \]

      3. lower-/.f64 N/A

         \[\leadsto \frac{b}{1 + e^{a}} + \log \color{blue}{\left(1 + e^{a}\right)} \]

      4. +-commutative N/A

         \[\leadsto \frac{b}{e^{a} + 1} + \log \left(1 + \color{blue}{e^{a}}\right) \]

      5. metadata-eval N/A

         \[\leadsto \frac{b}{e^{a} + 1 \cdot 1} + \log \left(1 + e^{a}\right) \]

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

         \[\leadsto \frac{b}{e^{a} - \left(\mathsf{neg}\left(1\right)\right) \cdot 1} + \log \left(1 + \color{blue}{e^{a}}\right) \]

      7. metadata-eval N/A

         \[\leadsto \frac{b}{e^{a} - -1 \cdot 1} + \log \left(1 + e^{a}\right) \]

      8. metadata-eval N/A

         \[\leadsto \frac{b}{e^{a} - -1} + \log \left(1 + e^{a}\right) \]

      9. lower--.f64 N/A

         \[\leadsto \frac{b}{e^{a} - -1} + \log \left(1 + \color{blue}{e^{a}}\right) \]

      10. lift-exp.f64 N/A

          \[\leadsto \frac{b}{e^{a} - -1} + \log \left(1 + e^{\color{blue}{a}}\right) \]

      11. lower-log1p.f64 N/A

          \[\leadsto \frac{b}{e^{a} - -1} + \mathsf{log1p}\left(e^{a}\right) \]

      12. lift-exp.f64 100.0

          \[\leadsto \frac{b}{e^{a} - -1} + \mathsf{log1p}\left(e^{a}\right) \]

   5. Applied rewrites 100.0%

      \[\leadsto \color{blue}{\frac{b}{e^{a} - -1} + \mathsf{log1p}\left(e^{a}\right)} \]

   6. Taylor expanded in b around inf

      \[\leadsto \frac{b}{\color{blue}{1 + e^{a}}} \]
   7. Step-by-step derivation

      1. lower-/.f64 N/A

         \[\leadsto \frac{b}{1 + \color{blue}{e^{a}}} \]

      2. lower-+.f64 N/A

         \[\leadsto \frac{b}{1 + e^{a}} \]

      3. lift-exp.f64 99.1

         \[\leadsto \frac{b}{1 + e^{a}} \]

   8. Applied rewrites 99.1%

      \[\leadsto \frac{b}{\color{blue}{1 + e^{a}}} \]

   9. Taylor expanded in a around 0

      \[\leadsto \frac{1}{2} \cdot b \]
   10. Step-by-step derivation

       1. lower-*.f64 18.6

          \[\leadsto 0.5 \cdot b \]

   11. Applied rewrites 18.6%

       \[\leadsto 0.5 \cdot b \]

   if -1 < a

   1. Initial program 98.0%

      \[\log \left(e^{a} + e^{b}\right) \]
   2. Add Preprocessing

   3. Taylor expanded in b around 0

      \[\leadsto \log \left(e^{a} + \color{blue}{\left(1 + b\right)}\right) \]
   4. Step-by-step derivation

      1. +-commutative N/A

         \[\leadsto \log \left(e^{a} + \left(b + \color{blue}{1}\right)\right) \]

      2. metadata-eval N/A

         \[\leadsto \log \left(e^{a} + \left(b + 1 \cdot \color{blue}{1}\right)\right) \]

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

         \[\leadsto \log \left(e^{a} + \left(b - \color{blue}{\left(\mathsf{neg}\left(1\right)\right) \cdot 1}\right)\right) \]

      4. metadata-eval N/A

         \[\leadsto \log \left(e^{a} + \left(b - -1 \cdot 1\right)\right) \]

      5. metadata-eval N/A

         \[\leadsto \log \left(e^{a} + \left(b - -1\right)\right) \]

      6. lower--.f64 96.5

         \[\leadsto \log \left(e^{a} + \left(b - \color{blue}{-1}\right)\right) \]

   5. Applied rewrites 96.5%

      \[\leadsto \log \left(e^{a} + \color{blue}{\left(b - -1\right)}\right) \]

   6. Taylor expanded in a around 0

      \[\leadsto \log \left(\color{blue}{\left(1 + a\right)} + \left(b - -1\right)\right) \]
   7. Step-by-step derivation

      1. lower-+.f64 95.5

         \[\leadsto \log \left(\left(1 + \color{blue}{a}\right) + \left(b - -1\right)\right) \]

   8. Applied rewrites 95.5%

      \[\leadsto \log \left(\color{blue}{\left(1 + a\right)} + \left(b - -1\right)\right) \]
3. Recombined 2 regimes into one program.
4. Add Preprocessing

Alternative 10: 56.9% accurate, 2.8× speedup

Math

\[\begin{array}{l} [a, b] = \mathsf{sort}([a, b])\\ \\ \begin{array}{l} \mathbf{if}\;a \leq -195:\\ \;\;\;\;0.5 \cdot b\\ \mathbf{else}:\\ \;\;\;\;\mathsf{log1p}\left(1 + b\right)\\ \end{array} \end{array} \]

FPCore

NOTE: a and b should be sorted in increasing order before calling this function.
(FPCore (a b)
 :precision binary64
 (if (<= a -195.0) (* 0.5 b) (log1p (+ 1.0 b))))

C

assert(a < b);
double code(double a, double b) {
	double tmp;
	if (a <= -195.0) {
		tmp = 0.5 * b;
	} else {
		tmp = log1p((1.0 + b));
	}
	return tmp;
}

Java

assert a < b;
public static double code(double a, double b) {
	double tmp;
	if (a <= -195.0) {
		tmp = 0.5 * b;
	} else {
		tmp = Math.log1p((1.0 + b));
	}
	return tmp;
}

Python

[a, b] = sort([a, b])
def code(a, b):
	tmp = 0
	if a <= -195.0:
		tmp = 0.5 * b
	else:
		tmp = math.log1p((1.0 + b))
	return tmp

Julia

a, b = sort([a, b])
function code(a, b)
	tmp = 0.0
	if (a <= -195.0)
		tmp = Float64(0.5 * b);
	else
		tmp = log1p(Float64(1.0 + b));
	end
	return tmp
end

Wolfram

NOTE: a and b should be sorted in increasing order before calling this function.
code[a_, b_] := If[LessEqual[a, -195.0], N[(0.5 * b), $MachinePrecision], N[Log[1 + N[(1.0 + b), $MachinePrecision]], $MachinePrecision]]

Derivation

1. Split input into 2 regimes

2. if a < -195

   1. Initial program 9.0%

      \[\log \left(e^{a} + e^{b}\right) \]
   2. Add Preprocessing

   3. Taylor expanded in b around 0

      \[\leadsto \color{blue}{\log \left(1 + e^{a}\right) + \frac{b}{1 + e^{a}}} \]
   4. Step-by-step derivation

      1. +-commutative N/A

         \[\leadsto \frac{b}{1 + e^{a}} + \color{blue}{\log \left(1 + e^{a}\right)} \]

      2. lower-+.f64 N/A

         \[\leadsto \frac{b}{1 + e^{a}} + \color{blue}{\log \left(1 + e^{a}\right)} \]

      3. lower-/.f64 N/A

         \[\leadsto \frac{b}{1 + e^{a}} + \log \color{blue}{\left(1 + e^{a}\right)} \]

      4. +-commutative N/A

         \[\leadsto \frac{b}{e^{a} + 1} + \log \left(1 + \color{blue}{e^{a}}\right) \]

      5. metadata-eval N/A

         \[\leadsto \frac{b}{e^{a} + 1 \cdot 1} + \log \left(1 + e^{a}\right) \]

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

         \[\leadsto \frac{b}{e^{a} - \left(\mathsf{neg}\left(1\right)\right) \cdot 1} + \log \left(1 + \color{blue}{e^{a}}\right) \]

      7. metadata-eval N/A

         \[\leadsto \frac{b}{e^{a} - -1 \cdot 1} + \log \left(1 + e^{a}\right) \]

      8. metadata-eval N/A

         \[\leadsto \frac{b}{e^{a} - -1} + \log \left(1 + e^{a}\right) \]

      9. lower--.f64 N/A

         \[\leadsto \frac{b}{e^{a} - -1} + \log \left(1 + \color{blue}{e^{a}}\right) \]

      10. lift-exp.f64 N/A

          \[\leadsto \frac{b}{e^{a} - -1} + \log \left(1 + e^{\color{blue}{a}}\right) \]

      11. lower-log1p.f64 N/A

          \[\leadsto \frac{b}{e^{a} - -1} + \mathsf{log1p}\left(e^{a}\right) \]

      12. lift-exp.f64 100.0

          \[\leadsto \frac{b}{e^{a} - -1} + \mathsf{log1p}\left(e^{a}\right) \]

   5. Applied rewrites 100.0%

      \[\leadsto \color{blue}{\frac{b}{e^{a} - -1} + \mathsf{log1p}\left(e^{a}\right)} \]

   6. Taylor expanded in b around inf

      \[\leadsto \frac{b}{\color{blue}{1 + e^{a}}} \]
   7. Step-by-step derivation

      1. lower-/.f64 N/A

         \[\leadsto \frac{b}{1 + \color{blue}{e^{a}}} \]

      2. lower-+.f64 N/A

         \[\leadsto \frac{b}{1 + e^{a}} \]

      3. lift-exp.f64 100.0

         \[\leadsto \frac{b}{1 + e^{a}} \]

   8. Applied rewrites 100.0%

      \[\leadsto \frac{b}{\color{blue}{1 + e^{a}}} \]

   9. Taylor expanded in a around 0

      \[\leadsto \frac{1}{2} \cdot b \]
   10. Step-by-step derivation

       1. lower-*.f64 18.7

          \[\leadsto 0.5 \cdot b \]

   11. Applied rewrites 18.7%

       \[\leadsto 0.5 \cdot b \]

   if -195 < a

   1. Initial program 97.7%

      \[\log \left(e^{a} + e^{b}\right) \]
   2. Add Preprocessing

   3. Taylor expanded in a around 0

      \[\leadsto \color{blue}{\log \left(1 + e^{b}\right)} \]
   4. Step-by-step derivation

      1. lower-log1p.f64 N/A

         \[\leadsto \mathsf{log1p}\left(e^{b}\right) \]

      2. lift-exp.f64 95.3

         \[\leadsto \mathsf{log1p}\left(e^{b}\right) \]

   5. Applied rewrites 95.3%

      \[\leadsto \color{blue}{\mathsf{log1p}\left(e^{b}\right)} \]

   6. Taylor expanded in b around 0

      \[\leadsto \mathsf{log1p}\left(1 + b\right) \]
   7. Step-by-step derivation

      1. lower-+.f64 93.9

         \[\leadsto \mathsf{log1p}\left(1 + b\right) \]

   8. Applied rewrites 93.9%

      \[\leadsto \mathsf{log1p}\left(1 + b\right) \]
3. Recombined 2 regimes into one program.
4. Add Preprocessing

Alternative 11: 56.8% accurate, 2.8× speedup

Math

\[\begin{array}{l} [a, b] = \mathsf{sort}([a, b])\\ \\ \begin{array}{l} \mathbf{if}\;a \leq -1:\\ \;\;\;\;0.5 \cdot b\\ \mathbf{else}:\\ \;\;\;\;\mathsf{log1p}\left(1 + a\right)\\ \end{array} \end{array} \]

FPCore

NOTE: a and b should be sorted in increasing order before calling this function.
(FPCore (a b) :precision binary64 (if (<= a -1.0) (* 0.5 b) (log1p (+ 1.0 a))))

C

assert(a < b);
double code(double a, double b) {
	double tmp;
	if (a <= -1.0) {
		tmp = 0.5 * b;
	} else {
		tmp = log1p((1.0 + a));
	}
	return tmp;
}

Java

assert a < b;
public static double code(double a, double b) {
	double tmp;
	if (a <= -1.0) {
		tmp = 0.5 * b;
	} else {
		tmp = Math.log1p((1.0 + a));
	}
	return tmp;
}

Python

[a, b] = sort([a, b])
def code(a, b):
	tmp = 0
	if a <= -1.0:
		tmp = 0.5 * b
	else:
		tmp = math.log1p((1.0 + a))
	return tmp

Julia

a, b = sort([a, b])
function code(a, b)
	tmp = 0.0
	if (a <= -1.0)
		tmp = Float64(0.5 * b);
	else
		tmp = log1p(Float64(1.0 + a));
	end
	return tmp
end

Wolfram

NOTE: a and b should be sorted in increasing order before calling this function.
code[a_, b_] := If[LessEqual[a, -1.0], N[(0.5 * b), $MachinePrecision], N[Log[1 + N[(1.0 + a), $MachinePrecision]], $MachinePrecision]]

Derivation

1. Split input into 2 regimes

2. if a < -1

   1. Initial program 9.4%

      \[\log \left(e^{a} + e^{b}\right) \]
   2. Add Preprocessing

   3. Taylor expanded in b around 0

      \[\leadsto \color{blue}{\log \left(1 + e^{a}\right) + \frac{b}{1 + e^{a}}} \]
   4. Step-by-step derivation

      1. +-commutative N/A

         \[\leadsto \frac{b}{1 + e^{a}} + \color{blue}{\log \left(1 + e^{a}\right)} \]

      2. lower-+.f64 N/A

         \[\leadsto \frac{b}{1 + e^{a}} + \color{blue}{\log \left(1 + e^{a}\right)} \]

      3. lower-/.f64 N/A

         \[\leadsto \frac{b}{1 + e^{a}} + \log \color{blue}{\left(1 + e^{a}\right)} \]

      4. +-commutative N/A

         \[\leadsto \frac{b}{e^{a} + 1} + \log \left(1 + \color{blue}{e^{a}}\right) \]

      5. metadata-eval N/A

         \[\leadsto \frac{b}{e^{a} + 1 \cdot 1} + \log \left(1 + e^{a}\right) \]

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

         \[\leadsto \frac{b}{e^{a} - \left(\mathsf{neg}\left(1\right)\right) \cdot 1} + \log \left(1 + \color{blue}{e^{a}}\right) \]

      7. metadata-eval N/A

         \[\leadsto \frac{b}{e^{a} - -1 \cdot 1} + \log \left(1 + e^{a}\right) \]

      8. metadata-eval N/A

         \[\leadsto \frac{b}{e^{a} - -1} + \log \left(1 + e^{a}\right) \]

      9. lower--.f64 N/A

         \[\leadsto \frac{b}{e^{a} - -1} + \log \left(1 + \color{blue}{e^{a}}\right) \]

      10. lift-exp.f64 N/A

          \[\leadsto \frac{b}{e^{a} - -1} + \log \left(1 + e^{\color{blue}{a}}\right) \]

      11. lower-log1p.f64 N/A

          \[\leadsto \frac{b}{e^{a} - -1} + \mathsf{log1p}\left(e^{a}\right) \]

      12. lift-exp.f64 100.0

          \[\leadsto \frac{b}{e^{a} - -1} + \mathsf{log1p}\left(e^{a}\right) \]

   5. Applied rewrites 100.0%

      \[\leadsto \color{blue}{\frac{b}{e^{a} - -1} + \mathsf{log1p}\left(e^{a}\right)} \]

   6. Taylor expanded in b around inf

      \[\leadsto \frac{b}{\color{blue}{1 + e^{a}}} \]
   7. Step-by-step derivation

      1. lower-/.f64 N/A

         \[\leadsto \frac{b}{1 + \color{blue}{e^{a}}} \]

      2. lower-+.f64 N/A

         \[\leadsto \frac{b}{1 + e^{a}} \]

      3. lift-exp.f64 99.1

         \[\leadsto \frac{b}{1 + e^{a}} \]

   8. Applied rewrites 99.1%

      \[\leadsto \frac{b}{\color{blue}{1 + e^{a}}} \]

   9. Taylor expanded in a around 0

      \[\leadsto \frac{1}{2} \cdot b \]
   10. Step-by-step derivation

       1. lower-*.f64 18.6

          \[\leadsto 0.5 \cdot b \]

   11. Applied rewrites 18.6%

       \[\leadsto 0.5 \cdot b \]

   if -1 < a

   1. Initial program 98.0%

      \[\log \left(e^{a} + e^{b}\right) \]
   2. Add Preprocessing

   3. Taylor expanded in b around 0

      \[\leadsto \color{blue}{\log \left(1 + e^{a}\right)} \]
   4. Step-by-step derivation

      1. lower-log1p.f64 N/A

         \[\leadsto \mathsf{log1p}\left(e^{a}\right) \]

      2. lift-exp.f64 95.4

         \[\leadsto \mathsf{log1p}\left(e^{a}\right) \]

   5. Applied rewrites 95.4%

      \[\leadsto \color{blue}{\mathsf{log1p}\left(e^{a}\right)} \]

   6. Taylor expanded in a around 0

      \[\leadsto \mathsf{log1p}\left(1 + a\right) \]
   7. Step-by-step derivation

      1. lower-+.f64 94.5

         \[\leadsto \mathsf{log1p}\left(1 + a\right) \]

   8. Applied rewrites 94.5%

      \[\leadsto \mathsf{log1p}\left(1 + a\right) \]
3. Recombined 2 regimes into one program.
4. Add Preprocessing

Alternative 12: 56.4% accurate, 2.8× speedup

Math

\[\begin{array}{l} [a, b] = \mathsf{sort}([a, b])\\ \\ \begin{array}{l} \mathbf{if}\;a \leq -195:\\ \;\;\;\;0.5 \cdot b\\ \mathbf{else}:\\ \;\;\;\;\mathsf{log1p}\left(1\right)\\ \end{array} \end{array} \]

FPCore

NOTE: a and b should be sorted in increasing order before calling this function.
(FPCore (a b) :precision binary64 (if (<= a -195.0) (* 0.5 b) (log1p 1.0)))

C

assert(a < b);
double code(double a, double b) {
	double tmp;
	if (a <= -195.0) {
		tmp = 0.5 * b;
	} else {
		tmp = log1p(1.0);
	}
	return tmp;
}

Java

assert a < b;
public static double code(double a, double b) {
	double tmp;
	if (a <= -195.0) {
		tmp = 0.5 * b;
	} else {
		tmp = Math.log1p(1.0);
	}
	return tmp;
}

Python

[a, b] = sort([a, b])
def code(a, b):
	tmp = 0
	if a <= -195.0:
		tmp = 0.5 * b
	else:
		tmp = math.log1p(1.0)
	return tmp

Julia

a, b = sort([a, b])
function code(a, b)
	tmp = 0.0
	if (a <= -195.0)
		tmp = Float64(0.5 * b);
	else
		tmp = log1p(1.0);
	end
	return tmp
end

Wolfram

NOTE: a and b should be sorted in increasing order before calling this function.
code[a_, b_] := If[LessEqual[a, -195.0], N[(0.5 * b), $MachinePrecision], N[Log[1 + 1.0], $MachinePrecision]]

Derivation

1. Split input into 2 regimes

2. if a < -195

   1. Initial program 9.0%

      \[\log \left(e^{a} + e^{b}\right) \]
   2. Add Preprocessing

   3. Taylor expanded in b around 0

      \[\leadsto \color{blue}{\log \left(1 + e^{a}\right) + \frac{b}{1 + e^{a}}} \]
   4. Step-by-step derivation

      1. +-commutative N/A

         \[\leadsto \frac{b}{1 + e^{a}} + \color{blue}{\log \left(1 + e^{a}\right)} \]

      2. lower-+.f64 N/A

         \[\leadsto \frac{b}{1 + e^{a}} + \color{blue}{\log \left(1 + e^{a}\right)} \]

      3. lower-/.f64 N/A

         \[\leadsto \frac{b}{1 + e^{a}} + \log \color{blue}{\left(1 + e^{a}\right)} \]

      4. +-commutative N/A

         \[\leadsto \frac{b}{e^{a} + 1} + \log \left(1 + \color{blue}{e^{a}}\right) \]

      5. metadata-eval N/A

         \[\leadsto \frac{b}{e^{a} + 1 \cdot 1} + \log \left(1 + e^{a}\right) \]

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

         \[\leadsto \frac{b}{e^{a} - \left(\mathsf{neg}\left(1\right)\right) \cdot 1} + \log \left(1 + \color{blue}{e^{a}}\right) \]

      7. metadata-eval N/A

         \[\leadsto \frac{b}{e^{a} - -1 \cdot 1} + \log \left(1 + e^{a}\right) \]

      8. metadata-eval N/A

         \[\leadsto \frac{b}{e^{a} - -1} + \log \left(1 + e^{a}\right) \]

      9. lower--.f64 N/A

         \[\leadsto \frac{b}{e^{a} - -1} + \log \left(1 + \color{blue}{e^{a}}\right) \]

      10. lift-exp.f64 N/A

          \[\leadsto \frac{b}{e^{a} - -1} + \log \left(1 + e^{\color{blue}{a}}\right) \]

      11. lower-log1p.f64 N/A

          \[\leadsto \frac{b}{e^{a} - -1} + \mathsf{log1p}\left(e^{a}\right) \]

      12. lift-exp.f64 100.0

          \[\leadsto \frac{b}{e^{a} - -1} + \mathsf{log1p}\left(e^{a}\right) \]

   5. Applied rewrites 100.0%

      \[\leadsto \color{blue}{\frac{b}{e^{a} - -1} + \mathsf{log1p}\left(e^{a}\right)} \]

   6. Taylor expanded in b around inf

      \[\leadsto \frac{b}{\color{blue}{1 + e^{a}}} \]
   7. Step-by-step derivation

      1. lower-/.f64 N/A

         \[\leadsto \frac{b}{1 + \color{blue}{e^{a}}} \]

      2. lower-+.f64 N/A

         \[\leadsto \frac{b}{1 + e^{a}} \]

      3. lift-exp.f64 100.0

         \[\leadsto \frac{b}{1 + e^{a}} \]

   8. Applied rewrites 100.0%

      \[\leadsto \frac{b}{\color{blue}{1 + e^{a}}} \]

   9. Taylor expanded in a around 0

      \[\leadsto \frac{1}{2} \cdot b \]
   10. Step-by-step derivation

       1. lower-*.f64 18.7

          \[\leadsto 0.5 \cdot b \]

   11. Applied rewrites 18.7%

       \[\leadsto 0.5 \cdot b \]

   if -195 < a

   1. Initial program 97.7%

      \[\log \left(e^{a} + e^{b}\right) \]
   2. Add Preprocessing

   3. Taylor expanded in a around 0

      \[\leadsto \color{blue}{\log \left(1 + e^{b}\right)} \]
   4. Step-by-step derivation

      1. lower-log1p.f64 N/A

         \[\leadsto \mathsf{log1p}\left(e^{b}\right) \]

      2. lift-exp.f64 95.3

         \[\leadsto \mathsf{log1p}\left(e^{b}\right) \]

   5. Applied rewrites 95.3%

      \[\leadsto \color{blue}{\mathsf{log1p}\left(e^{b}\right)} \]

   6. Taylor expanded in b around 0

      \[\leadsto \mathsf{log1p}\left(1\right) \]
   7. Step-by-step derivation

   8. Applied rewrites 92.8%

      \[\leadsto \mathsf{log1p}\left(1\right) \]
3. Recombined 2 regimes into one program.
4. Add Preprocessing

Alternative 13: 11.9% accurate, 50.7× speedup

Math

\[\begin{array}{l} [a, b] = \mathsf{sort}([a, b])\\ \\ 0.5 \cdot b \end{array} \]

FPCore

NOTE: a and b should be sorted in increasing order before calling this function.
(FPCore (a b) :precision binary64 (* 0.5 b))

C

assert(a < b);
double code(double a, double b) {
	return 0.5 * b;
}

Fortran

NOTE: a and b should be sorted in increasing order before calling this function.
module fmin_fmax_functions
    implicit none
    private
    public fmax
    public fmin

    interface fmax
        module procedure fmax88
        module procedure fmax44
        module procedure fmax84
        module procedure fmax48
    end interface
    interface fmin
        module procedure fmin88
        module procedure fmin44
        module procedure fmin84
        module procedure fmin48
    end interface
contains
    real(8) function fmax88(x, y) result (res)
        real(8), intent (in) :: x
        real(8), intent (in) :: y
        res = merge(y, merge(x, max(x, y), y /= y), x /= x)
    end function
    real(4) function fmax44(x, y) result (res)
        real(4), intent (in) :: x
        real(4), intent (in) :: y
        res = merge(y, merge(x, max(x, y), y /= y), x /= x)
    end function
    real(8) function fmax84(x, y) result(res)
        real(8), intent (in) :: x
        real(4), intent (in) :: y
        res = merge(dble(y), merge(x, max(x, dble(y)), y /= y), x /= x)
    end function
    real(8) function fmax48(x, y) result(res)
        real(4), intent (in) :: x
        real(8), intent (in) :: y
        res = merge(y, merge(dble(x), max(dble(x), y), y /= y), x /= x)
    end function
    real(8) function fmin88(x, y) result (res)
        real(8), intent (in) :: x
        real(8), intent (in) :: y
        res = merge(y, merge(x, min(x, y), y /= y), x /= x)
    end function
    real(4) function fmin44(x, y) result (res)
        real(4), intent (in) :: x
        real(4), intent (in) :: y
        res = merge(y, merge(x, min(x, y), y /= y), x /= x)
    end function
    real(8) function fmin84(x, y) result(res)
        real(8), intent (in) :: x
        real(4), intent (in) :: y
        res = merge(dble(y), merge(x, min(x, dble(y)), y /= y), x /= x)
    end function
    real(8) function fmin48(x, y) result(res)
        real(4), intent (in) :: x
        real(8), intent (in) :: y
        res = merge(y, merge(dble(x), min(dble(x), y), y /= y), x /= x)
    end function
end module

real(8) function code(a, b)
use fmin_fmax_functions
    real(8), intent (in) :: a
    real(8), intent (in) :: b
    code = 0.5d0 * b
end function

Java

assert a < b;
public static double code(double a, double b) {
	return 0.5 * b;
}

Python

[a, b] = sort([a, b])
def code(a, b):
	return 0.5 * b

Julia

a, b = sort([a, b])
function code(a, b)
	return Float64(0.5 * b)
end

MATLAB

a, b = num2cell(sort([a, b])){:}
function tmp = code(a, b)
	tmp = 0.5 * b;
end

Wolfram

NOTE: a and b should be sorted in increasing order before calling this function.
code[a_, b_] := N[(0.5 * b), $MachinePrecision]

Derivation

1. Initial program 54.1%

   \[\log \left(e^{a} + e^{b}\right) \]
2. Add Preprocessing

3. Taylor expanded in b around 0

   \[\leadsto \color{blue}{\log \left(1 + e^{a}\right) + \frac{b}{1 + e^{a}}} \]
4. Step-by-step derivation

   1. +-commutative N/A

      \[\leadsto \frac{b}{1 + e^{a}} + \color{blue}{\log \left(1 + e^{a}\right)} \]

   2. lower-+.f64 N/A

      \[\leadsto \frac{b}{1 + e^{a}} + \color{blue}{\log \left(1 + e^{a}\right)} \]

   3. lower-/.f64 N/A

      \[\leadsto \frac{b}{1 + e^{a}} + \log \color{blue}{\left(1 + e^{a}\right)} \]

   4. +-commutative N/A

      \[\leadsto \frac{b}{e^{a} + 1} + \log \left(1 + \color{blue}{e^{a}}\right) \]

   5. metadata-eval N/A

      \[\leadsto \frac{b}{e^{a} + 1 \cdot 1} + \log \left(1 + e^{a}\right) \]

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

      \[\leadsto \frac{b}{e^{a} - \left(\mathsf{neg}\left(1\right)\right) \cdot 1} + \log \left(1 + \color{blue}{e^{a}}\right) \]

   7. metadata-eval N/A

      \[\leadsto \frac{b}{e^{a} - -1 \cdot 1} + \log \left(1 + e^{a}\right) \]

   8. metadata-eval N/A

      \[\leadsto \frac{b}{e^{a} - -1} + \log \left(1 + e^{a}\right) \]

   9. lower--.f64 N/A

      \[\leadsto \frac{b}{e^{a} - -1} + \log \left(1 + \color{blue}{e^{a}}\right) \]

   10. lift-exp.f64 N/A

       \[\leadsto \frac{b}{e^{a} - -1} + \log \left(1 + e^{\color{blue}{a}}\right) \]

   11. lower-log1p.f64 N/A

       \[\leadsto \frac{b}{e^{a} - -1} + \mathsf{log1p}\left(e^{a}\right) \]

   12. lift-exp.f64 98.3

       \[\leadsto \frac{b}{e^{a} - -1} + \mathsf{log1p}\left(e^{a}\right) \]

5. Applied rewrites 98.3%

   \[\leadsto \color{blue}{\frac{b}{e^{a} - -1} + \mathsf{log1p}\left(e^{a}\right)} \]

6. Taylor expanded in b around inf

   \[\leadsto \frac{b}{\color{blue}{1 + e^{a}}} \]
7. Step-by-step derivation

   1. lower-/.f64 N/A

      \[\leadsto \frac{b}{1 + \color{blue}{e^{a}}} \]

   2. lower-+.f64 N/A

      \[\leadsto \frac{b}{1 + e^{a}} \]

   3. lift-exp.f64 51.9

      \[\leadsto \frac{b}{1 + e^{a}} \]

8. Applied rewrites 51.9%

   \[\leadsto \frac{b}{\color{blue}{1 + e^{a}}} \]

9. Taylor expanded in a around 0

   \[\leadsto \frac{1}{2} \cdot b \]
10. Step-by-step derivation

    1. lower-*.f64 11.9

       \[\leadsto 0.5 \cdot b \]

11. Applied rewrites 11.9%

    \[\leadsto 0.5 \cdot b \]
12. Add Preprocessing

Reproduce

herbie shell --seed 2025088 
(FPCore (a b)
  :name "symmetry log of sum of exp"
  :precision binary64
  (log (+ (exp a) (exp b))))