-
Notifications
You must be signed in to change notification settings - Fork 11
Open
Description
At least on my MacBook Air M2 16GB, which is the machine on which I'm seeing best PyChOpMarg performance, the GPU doesn't appear to be engaging.
I think that means we're leaving lots of performance on the table, because the equalization optimization task, which is what consumes all the time, is highly parallelizable.
The initial results of the multiprocessing effort have shown this to be true.
However, that effort is limited, by Python's need to have completely separate memory spaces for each worker process, to about x4.
I suspect we'll get much more from a GPU.
==> Verify this and fix it if it's true.
Reactions are currently unavailable
Metadata
Metadata
Assignees
Labels
No labels