edoddridge
commented
7 years ago Fair point. The value for pi is only used in three places:
- To calculate the spectral radius of the grid for the matrix inversion algorithm. The spectral radius is used as an approximate guide to help the algorithm converge faster. It's not a high accuracy usage. https://github.com/edoddridge/MIM/blob/master/MIM.f90#L333
- To calculate the sinusoidally varying wind component - if that option is being used. https://github.com/edoddridge/MIM/blob/master/MIM.f90#L527 & https://github.com/edoddridge/MIM/blob/master/MIM.f90#L530
Given that the value isn't used directly in the computation, I'm happy just updating it.
axch
No reason to lose accuracy for a dozen source code characters.
From http://stackoverflow.com/questions/507819/pi-and-accuracy-of-a-floating-point-number, 3.1415926535897932384 should be sufficient.
I am loath to just do this myself until more of the testing is in place. Ideally, from my point of view, I would see accuracy improve (slightly) from making the change. @edoddridge, if you feel differently, don't let me stop you from just editing the file and closing this ticket.