Closed sbailey closed 4 years ago
On the whole this looks good (although I didn't go through in great detail)-- thanks @sbailey !
Do I understand correctly that this means the bug/bugs now occur somewhere after projection_matrix?
Yep, any remaining discrepancies occur post-projection_matrix
, and ex2d_patch
is up next. I'll put together some example comparisons but I think I will punt that one to you for debugging. Thanks.
This PR:
get_spots
andprojection_matrix
Note: unlike what I suggested at our weekly meeting, the spots differences with specter was not due to an equivalently correct but different of definition for the corners; it was a bug in gpu_specter that I now fixed.
Tests require specter master, since this work also uncovered a bug there that was fixed in desihub/specter#79.
CPU code in gpu_specter now agrees with specter up through projection_matrix.
Assigning this to @dmargala but also pinging @lastephey and @ziyaointl .