some numeric thresholds are calculated differently from galsim #189
Description
@beckermr pointed out that some numeric thresholds related to drawing images like maxk are not computed in the same way between jax galsim and galsim. As I understand it, this doesn't lead to significantly biased results and relates to the fact that some of the way these algorithms were implemented in galsim are not very jax friendly. The problem is that the implementation in jax galsim for these algorithms leads to different outputs than galsim everything else equal which might surprise the user.
We discussed in the meeting two options:
(1) Document these caveats so the users know about them
(2) Try to find new implementations in jax that match the galsim output
I think we were leaning towards (1).
@beckermr do you have any particular example of this? Does the above accurately describe the issue you had in mind?