Closed cvsik closed 3 months ago
Dear Maximilian,
thanks for the contribution. I'll take a closer look at the PR in the coming days.
Thanks again for the PR. It looks like a promising addition.
At a first glance the method seems quite slow... The alanine-dipetide already takes 42 sec on my notebook w/ 2 threads.
How did you conclude that pytest somehow disables scipy's threading? When I'm not setting OMP_NUM_THREADS=N all my cores (6 physical + 6 logical) are running at 100% utilization.
Regarding my changes in the commit:
I added geodesic to the testing
package, so the test is skipped when the package is not installed. I also fixed the can_geodesic
, which was always False
. Finally I enables --geodesic
in pysistrj
. Further options can't be passed though.
If you are fine with the change I'll merge the PR.
At a first glance the method seems quite slow...
Yes, that's true. I think their Python implementation isn't the fastest, but it's giving really good interpolation results, so I'm always more than happy to spend some more time on interpolating then 😁
The changes look good! Thanks for taking care of it! 🚀
This PR adds the geodesic interpolation for reaction pathways (https://doi.org/10.1063/1.5090303) to
interpolate
.The only strange thing I noticed is that the
pytest
session seems to disable threading forscipy
, such that the runtime is much slower onpytest
than when running thegeodesic_interpolate
command manually from the command line.... not so important though.