Average Error: 1.2 → 0.7
Time: 3.1s
Precision: binary64
\[\]
↓
\[\]
↓
double code(double x, double y, double z) {
	return ((double) fabs(((double) (((double) (((double) (x + 4.0)) / y)) - ((double) (((double) (x / y)) * z))))));
}

↓
double code(double x, double y, double z) {
	double VAR;
	if ((y <= 7.06822395138139e+98)) {
		VAR = ((double) fabs(((double) (((double) (4.0 / y)) + ((double) (((double) (x / y)) * ((double) (1.0 - z))))))));
	} else {
		VAR = ((double) fabs(((double) (((double) (((double) (4.0 + x)) / y)) - ((double) (x * ((double) (z / y))))))));
	}
	return VAR;
}

Error
Bits error versus x
Bits error versus y
Bits error versus z
Try it out
Your Program's Arguments
Results
Enter valid numbers for all inputs
Derivation
  1. Split input into 2 regimes
  2. if  y  < 7.0682239513813901e98
    1. Initial program 0.8
      \[\]
    2. Taylor expanded around 0 2.7
      \[\leadsto \]
    3. Simplified0.8
      \[\leadsto \]
    if 7.0682239513813901e98 <  y 
    1. Initial program 2.8
      \[\]
    2. Simplified0.1
      \[\leadsto \]
  3. Recombined 2 regimes into one program.
  4. Final simplification0.7
    \[\leadsto \]
Reproduce
herbie shell --seed 2020180 
(FPCore (x y z)
  :name "fabs fraction 1"
  :precision binary64
  (fabs (- (/ (+ x 4.0) y) (* (/ x y) z))))