Average Error: 0.5 → 0.5
Time: 37.5s
Precision: 64
Internal Precision: 320
\[\left(a - \frac{1.0}{3.0}\right) \cdot \left(1 + \frac{1}{\sqrt{9 \cdot \left(a - \frac{1.0}{3.0}\right)}} \cdot rand\right)\]
↓
\[\left(a - \frac{1.0}{3.0}\right) \cdot 1 + \left(a - \frac{1.0}{3.0}\right) \cdot \left(\frac{1}{\sqrt{9 \cdot \left(a - \frac{1.0}{3.0}\right)}} \cdot rand\right)\]

Error

Bits error versus a

Bits error versus rand

Derivation

  1. Initial program 0.5

    \[\left(a - \left(\frac{\left(real->posit(1.0)\right)}{\left(real->posit(3.0)\right)}\right)\right) \cdot \left(\frac{\left(real->posit(1)\right)}{\left(\left(\frac{\left(real->posit(1)\right)}{\left(\sqrt{\left(\left(real->posit(9)\right) \cdot \left(a - \left(\frac{\left(real->posit(1.0)\right)}{\left(real->posit(3.0)\right)}\right)\right)\right)}\right)}\right) \cdot rand\right)}\right)\]
  2. Using strategy rm

  3. Applied distribute-lft-in0.5

    \[\leadsto \color{blue}{\frac{\left(\left(a - \left(\frac{\left(real->posit(1.0)\right)}{\left(real->posit(3.0)\right)}\right)\right) \cdot \left(real->posit(1)\right)\right)}{\left(\left(a - \left(\frac{\left(real->posit(1.0)\right)}{\left(real->posit(3.0)\right)}\right)\right) \cdot \left(\left(\frac{\left(real->posit(1)\right)}{\left(\sqrt{\left(\left(real->posit(9)\right) \cdot \left(a - \left(\frac{\left(real->posit(1.0)\right)}{\left(real->posit(3.0)\right)}\right)\right)\right)}\right)}\right) \cdot rand\right)\right)}}\]

  4. Final simplification0.5

    \[\leadsto \left(a - \frac{1.0}{3.0}\right) \cdot 1 + \left(a - \frac{1.0}{3.0}\right) \cdot \left(\frac{1}{\sqrt{9 \cdot \left(a - \frac{1.0}{3.0}\right)}} \cdot rand\right)\]

Reproduce

herbie shell --seed 2019053 +o rules:numerics
(FPCore (a rand)
  :name "Octave 3.8, oct_fill_randg"
  (*.p16 (-.p16 a (/.p16 (real->posit16 1.0) (real->posit16 3.0))) (+.p16 (real->posit16 1) (*.p16 (/.p16 (real->posit16 1) (sqrt.p16 (*.p16 (real->posit16 9) (-.p16 a (/.p16 (real->posit16 1.0) (real->posit16 3.0)))))) rand))))