Average Error: 6.5 → 0.1
Time: 3.7s
Precision: 64
\[x + \frac{y \cdot y}{z}\]
↓
\[\mathsf{fma}\left(\frac{y}{z}, y, x\right)\]
x + \frac{y \cdot y}{z}
↓
\mathsf{fma}\left(\frac{y}{z}, y, x\right)
double f(double x, double y, double z) {
        double r1003121 = x;
        double r1003122 = y;
        double r1003123 = r1003122 * r1003122;
        double r1003124 = z;
        double r1003125 = r1003123 / r1003124;
        double r1003126 = r1003121 + r1003125;
        return r1003126;
}
↓
double f(double x, double y, double z) {
        double r1003127 = y;
        double r1003128 = z;
        double r1003129 = r1003127 / r1003128;
        double r1003130 = x;
        double r1003131 = fma(r1003129, r1003127, r1003130);
        return r1003131;
}

Error

Bits error versus x

Bits error versus y

Bits error versus z

Target

Original6.5
Target0.1
Herbie0.1
\[x + y \cdot \frac{y}{z}\]

Derivation

  1. Initial program 6.5

    \[x + \frac{y \cdot y}{z}\]

  2. Simplified0.1

    \[\leadsto \color{blue}{\mathsf{fma}\left(\frac{y}{z}, y, x\right)}\]

  3. Final simplification0.1

    \[\leadsto \mathsf{fma}\left(\frac{y}{z}, y, x\right)\]

Reproduce

herbie shell --seed 2019353 +o rules:numerics
(FPCore (x y z)
  :name "Crypto.Random.Test:calculate from crypto-random-0.0.9"
  :precision binary64

  :herbie-target
  (+ x (* y (/ y z)))

  (+ x (/ (* y y) z)))