math.log/1 on complex, real part

Average Error: 31.8 → 18.1

Time: 1.1s

Precision: 64

Math

\[\log \left(\sqrt{re \cdot re + im \cdot im}\right)\]

↓

\[\begin{array}{l} \mathbf{if}\;re \le -2.3383469654658497 \cdot 10^{127}:\\ \;\;\;\;\log \left(-1 \cdot re\right)\\ \mathbf{elif}\;re \le 1.99276855197564801 \cdot 10^{46}:\\ \;\;\;\;\log \left(\sqrt{re \cdot re + im \cdot im}\right)\\ \mathbf{else}:\\ \;\;\;\;\log re\\ \end{array}\]

C

double code(double re, double im) {
	return log(sqrt(((re * re) + (im * im))));
}

↓

double code(double re, double im) {
	double VAR;
	if ((re <= -2.3383469654658497e+127)) {
		VAR = log((-1.0 * re));
	} else {
		double VAR_1;
		if ((re <= 1.992768551975648e+46)) {
			VAR_1 = log(sqrt(((re * re) + (im * im))));
		} else {
			VAR_1 = log(re);
		}
		VAR = VAR_1;
	}
	return VAR;
}

Error

Bits error versus re

Bits error versus im

Derivation

1. Split input into 3 regimes

2. if re < -2.3383469654658497e+127

   1. Initial program 56.7

      \[\log \left(\sqrt{re \cdot re + im \cdot im}\right)\]

   2. Taylor expanded around -inf 7.4

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

   if -2.3383469654658497e+127 < re < 1.992768551975648e+46

   1. Initial program 22.5

      \[\log \left(\sqrt{re \cdot re + im \cdot im}\right)\]

   if 1.992768551975648e+46 < re

   1. Initial program 43.8

      \[\log \left(\sqrt{re \cdot re + im \cdot im}\right)\]

   2. Taylor expanded around inf 11.9

      \[\leadsto \log \color{blue}{re}\]
3. Recombined 3 regimes into one program.

4. Final simplification 18.1

   \[\leadsto \begin{array}{l} \mathbf{if}\;re \le -2.3383469654658497 \cdot 10^{127}:\\ \;\;\;\;\log \left(-1 \cdot re\right)\\ \mathbf{elif}\;re \le 1.99276855197564801 \cdot 10^{46}:\\ \;\;\;\;\log \left(\sqrt{re \cdot re + im \cdot im}\right)\\ \mathbf{else}:\\ \;\;\;\;\log re\\ \end{array}\]

Reproduce

herbie shell --seed 2020071 
(FPCore (re im)
  :name "math.log/1 on complex, real part"
  :precision binary64
  (log (sqrt (+ (* re re) (* im im)))))