Parallelize nwl #122
Parallelize nwl #122
Conversation
| diff = np.linalg.norm(pt - fw_pt) | ||
|
|
||
| return diff | ||
|
|
||
|
|
||
| def find_coords(vmec, wall_s, phi, pt): | ||
| def find_coords(data): |
There was a problem hiding this comment.
this function takes a tuple so it plays nicely with concurrent.futures.ProcessPoolExecutor
|
I have autopep8 installed, not sure if others would prefer something else. |
|
I changed to black since autopep8 wasn't being aggressive enough for the way I write code apparently 🙃. It does a better job of enforcing quotations and whitespace (and stops me from putting closing brackets on the line I'm not supposed to). It did require that I manually specify an 80 character line limit, I think it's default is 88. |
|
To test that this parallelization is working as expected I ran 10000 particles from a surface source using the main branch, and wrote the resulting matrix to file, then did the same but with this branch, using 9 threads. I then loaded the files to compare the values and found that they were identical. Not a very formal test, but I could make a PR adding testing for nwl_utils then once that's good to go this PR would be tested against that? |
Edgar-21
force-pushed
the
parallelize_nwl
branch
from
04f656f
to
37c4ce9
Compare
addresses #120