Most functions take sin_theta and cos_theta as arguments (e.g.
|
virtual double rhs_f(const double& r, const double& theta, const double& sin_theta, |
|
const double& cos_theta) const = 0; |
) . If I remember right, this was added by Phillip as the analytical expressions that we used for the test cases in the paper all contained a large number of sin/cos expressions so pre-calculating improved the timings. However in the real-world non-analytical case there are few to no cos/sin expressions. These arguments therefore feel nonsensical.
