Closed carterbox closed 3 years ago
This is actually, not ready. When setting num_gpu to >=2 it fails because I forgot to divide theta
amongst the processes.
OK. Now, I think this PR is ready because it is actually being tested in a multi-gpu environment.
Purpose
Add multi-device parallelism to the laminography solver.
Approach
Ported the same map reduce approach used in the ptycho cgrad solver.
Pre-Merge Checklists
Submitter
yapf
to format python code.Reviewer