Average Error: 0.5 → 0.5
Time: 8.7s
Precision: binary64
\[\frac{\cos th}{\sqrt{2}} \cdot \left(a1 \cdot a1\right) + \frac{\cos th}{\sqrt{2}} \cdot \left(a2 \cdot a2\right)\]
↓
\[\left(\left(a1 \cdot a1 + a2 \cdot a2\right) \cdot \cos th\right) \cdot \frac{\frac{1}{\sqrt{\sqrt{2}}}}{\sqrt{\sqrt{2}}}\]
\frac{\cos th}{\sqrt{2}} \cdot \left(a1 \cdot a1\right) + \frac{\cos th}{\sqrt{2}} \cdot \left(a2 \cdot a2\right)
↓
\left(\left(a1 \cdot a1 + a2 \cdot a2\right) \cdot \cos th\right) \cdot \frac{\frac{1}{\sqrt{\sqrt{2}}}}{\sqrt{\sqrt{2}}}
(FPCore (a1 a2 th)
 :precision binary64
 (+
  (* (/ (cos th) (sqrt 2.0)) (* a1 a1))
  (* (/ (cos th) (sqrt 2.0)) (* a2 a2))))
↓
(FPCore (a1 a2 th)
 :precision binary64
 (*
  (* (+ (* a1 a1) (* a2 a2)) (cos th))
  (/ (/ 1.0 (sqrt (sqrt 2.0))) (sqrt (sqrt 2.0)))))
double code(double a1, double a2, double th) {
	return ((double) (((double) ((((double) cos(th)) / ((double) sqrt(2.0))) * ((double) (a1 * a1)))) + ((double) ((((double) cos(th)) / ((double) sqrt(2.0))) * ((double) (a2 * a2))))));
}
↓
double code(double a1, double a2, double th) {
	return ((double) (((double) (((double) (((double) (a1 * a1)) + ((double) (a2 * a2)))) * ((double) cos(th)))) * ((1.0 / ((double) sqrt(((double) sqrt(2.0))))) / ((double) sqrt(((double) sqrt(2.0)))))));
}

Error

Bits error versus a1

Bits error versus a2

Bits error versus th

Try it out

Your Program's Arguments

Results

Enter valid numbers for all inputs

Derivation

  1. Initial program Error: 0.5 bits

    \[\frac{\cos th}{\sqrt{2}} \cdot \left(a1 \cdot a1\right) + \frac{\cos th}{\sqrt{2}} \cdot \left(a2 \cdot a2\right)\]

  2. SimplifiedError: 0.5 bits

    \[\leadsto \color{blue}{\cos th \cdot \frac{a1 \cdot a1 + a2 \cdot a2}{\sqrt{2}}}\]
  3. Using strategy rm

  4. Applied add-sqr-sqrtError: 0.5 bits

    \[\leadsto \cos th \cdot \frac{a1 \cdot a1 + a2 \cdot a2}{\sqrt{\color{blue}{\sqrt{2} \cdot \sqrt{2}}}}\]

  5. Applied sqrt-prodError: 0.6 bits

    \[\leadsto \cos th \cdot \frac{a1 \cdot a1 + a2 \cdot a2}{\color{blue}{\sqrt{\sqrt{2}} \cdot \sqrt{\sqrt{2}}}}\]

  6. Applied associate-/r*Error: 0.5 bits

    \[\leadsto \cos th \cdot \color{blue}{\frac{\frac{a1 \cdot a1 + a2 \cdot a2}{\sqrt{\sqrt{2}}}}{\sqrt{\sqrt{2}}}}\]
  7. Using strategy rm

  8. Applied *-un-lft-identityError: 0.5 bits

    \[\leadsto \cos th \cdot \frac{\frac{a1 \cdot a1 + a2 \cdot a2}{\sqrt{\sqrt{2}}}}{\sqrt{\sqrt{\color{blue}{1 \cdot 2}}}}\]

  9. Applied sqrt-prodError: 0.5 bits

    \[\leadsto \cos th \cdot \frac{\frac{a1 \cdot a1 + a2 \cdot a2}{\sqrt{\sqrt{2}}}}{\sqrt{\color{blue}{\sqrt{1} \cdot \sqrt{2}}}}\]

  10. Applied sqrt-prodError: 0.5 bits

    \[\leadsto \cos th \cdot \frac{\frac{a1 \cdot a1 + a2 \cdot a2}{\sqrt{\sqrt{2}}}}{\color{blue}{\sqrt{\sqrt{1}} \cdot \sqrt{\sqrt{2}}}}\]

  11. Applied div-invError: 0.7 bits

    \[\leadsto \cos th \cdot \frac{\color{blue}{\left(a1 \cdot a1 + a2 \cdot a2\right) \cdot \frac{1}{\sqrt{\sqrt{2}}}}}{\sqrt{\sqrt{1}} \cdot \sqrt{\sqrt{2}}}\]

  12. Applied times-fracError: 0.5 bits

    \[\leadsto \cos th \cdot \color{blue}{\left(\frac{a1 \cdot a1 + a2 \cdot a2}{\sqrt{\sqrt{1}}} \cdot \frac{\frac{1}{\sqrt{\sqrt{2}}}}{\sqrt{\sqrt{2}}}\right)}\]

  13. Applied associate-*r*Error: 0.5 bits

    \[\leadsto \color{blue}{\left(\cos th \cdot \frac{a1 \cdot a1 + a2 \cdot a2}{\sqrt{\sqrt{1}}}\right) \cdot \frac{\frac{1}{\sqrt{\sqrt{2}}}}{\sqrt{\sqrt{2}}}}\]

  14. SimplifiedError: 0.5 bits

    \[\leadsto \color{blue}{\left(\left(a1 \cdot a1 + a2 \cdot a2\right) \cdot \cos th\right)} \cdot \frac{\frac{1}{\sqrt{\sqrt{2}}}}{\sqrt{\sqrt{2}}}\]

  15. Final simplificationError: 0.5 bits

    \[\leadsto \left(\left(a1 \cdot a1 + a2 \cdot a2\right) \cdot \cos th\right) \cdot \frac{\frac{1}{\sqrt{\sqrt{2}}}}{\sqrt{\sqrt{2}}}\]

Reproduce

herbie shell --seed 2020203 
(FPCore (a1 a2 th)
  :name "Migdal et al, Equation (64)"
  :precision binary64
  (+ (* (/ (cos th) (sqrt 2.0)) (* a1 a1)) (* (/ (cos th) (sqrt 2.0)) (* a2 a2))))