Models fails due to ``failed to setup interpolation grid``

[prashjet]

Some models fail with the error message

RuntimeError: cmd_orb_start exit code 0. ERROR. Message: Note: The following floating-point exceptions are signalling: IEEE_UNDERFLOW_FLAG
STOP failed to setup interpolation grid

One very quick test suggests this happens for very flat models, i.e. with p and/or q close to 0. I was testing the suggestion from @gsantucci90 to start as flat as possible (See #174). For NGC6278 I tried to run a model with p,q,u = (0.02, 0.01, 0.03). This combination of values does have a valid de-projection, i.e. it does not trigger either of the following warnings:

STOP "p^2 or q^2 is below 0."
STOP "No triaxial deprojection possible"

however it does trigger the IEEE_UNDERFLOW_FLAG/failed to setup interpolation grid message. The above values were the flattest possible (p,q,u) with a valid de-projection. When I replace these with the mean valid values instead of the flattest i.e. p,q,u = (0.58, 0.21, 0.75), then there is no problem.

So this means (at least for this test):

• we cannot easily start at the flattest valid values of (p,q,u)
• it is not clear what the flattest valid values are which will not trigger the IEEE_UNDERFLOW_FLAG/failed to setup interpolation grid error
• starting at the mean valid values of (p,q,u) seems to work

[sthater]

Could this maybe be something @ejlilley could investigate? maybe we just think deprojection is possible but it is in fact not? Not sure if there are limitations...