Average Error: 0.2 → 0.2
Time: 4.0s
Precision: binary64
\[\cos a1 \cdot \sin a2 - \sin a1 \cdot \cos a2\]
↓
\[\cos a1 \cdot \sin a2 - \sin a1 \cdot \cos a2\]
\cos a1 \cdot \sin a2 - \sin a1 \cdot \cos a2
↓
\cos a1 \cdot \sin a2 - \sin a1 \cdot \cos a2
double code(double a1, double a2) {
	return ((double) (((double) (((double) cos(a1)) * ((double) sin(a2)))) - ((double) (((double) sin(a1)) * ((double) cos(a2))))));
}
↓
double code(double a1, double a2) {
	return ((double) (((double) (((double) cos(a1)) * ((double) sin(a2)))) - ((double) (((double) sin(a1)) * ((double) cos(a2))))));
}

Error

Bits error versus a1

Bits error versus a2

Try it out

Your Program's Arguments

Results

Enter valid numbers for all inputs

Derivation

  1. Initial program 0.2

    \[\cos a1 \cdot \sin a2 - \sin a1 \cdot \cos a2\]

  2. Final simplification0.2

    \[\leadsto \cos a1 \cdot \sin a2 - \sin a1 \cdot \cos a2\]

Reproduce

herbie shell --seed 2020153 
(FPCore (a1 a2)
  :name "(- (* (cos a1) (sin a2)) (* (sin a1) (cos a2)))"
  :precision binary64
  (- (* (cos a1) (sin a2)) (* (sin a1) (cos a2))))