Computing the PSF in both Python and IDL takes a really long time. In Python, this process is sped up significantly with a GPU, but is still significantly slower than the actual deconvolution.
I would suggest precomputing these once and then storing them remotely somewhere and downloading them as needed (similar to how we handle the data needed for the wavelength response calculation. We could still maintain the code to compute the PSF functions if people really wanted to do that computation, but the default behavior would just be to use the remote files.
In GitLab by @wtbarnes on Sep 3, 2023, 13:48
Computing the PSF in both Python and IDL takes a really long time. In Python, this process is sped up significantly with a GPU, but is still significantly slower than the actual deconvolution.
I would suggest precomputing these once and then storing them remotely somewhere and downloading them as needed (similar to how we handle the data needed for the wavelength response calculation. We could still maintain the code to compute the PSF functions if people really wanted to do that computation, but the default behavior would just be to use the remote files.