performance optimization of matrixDFT by caching arrays about poppy HOT 3 CLOSED

This would be more straightforward to implement now with the refactored and unified matrix DFT code. But for typical poppy and webbpsf use cases it wouldn't be a major performance gain since it's not typical to do the same transform repeatedly. Gerchberg-Saxton is a difference case of course.

So I suggest we're not going to prioritize this as part of the main development efforts - but would certainly welcome a patch from Anand or any other user if they wants to work on it.

from poppy.

From looking at (and ultimately removing) the Zernike cache, the relevant gotchas would be: returning copies of the array in the cache rather than references / marking the cached arrays as read only (to prevent accidental in-place modification), and ensuring ALL the relevant parameters are used in the cache key (to prevent the issue of changing a parameter and getting the same array back because the cache doesn't know about that parameter).

from poppy.

Stale issue with no action in the last 2 years, and no compelling need to prioritize it. I'm going to mark this wontfix & close.

from poppy.