When executing the test, the "optimized" version (using
static variables in the function
for values that are expensive to calculate, e.g. involving
std::cbrt) is surprisingly slower
than the version without static variables (with compiler optimizations).
Without compiler optimization, the opposite is true (which would be expected).
Making the variables
static constexpr (and adapting the relevant methods involved), makes the "optimized" version
as fast as the other one.
How should we deal with this? Might be related to #441 (closed)
Closes #389 (closed)