Marc T. Henry de Frahan

Ran the new reg test with MPI and everything on and I don't see any diffs in the particle files. I would like to double check netcdf. But here's the...

This is the diff with the big case I am using to benchmark (40M probes): ``` absolute_error relative_error uid 0.000000e+00 0.000000e+00 set_id 0.000000e+00 0.000000e+00 probe_id 0.000000e+00 0.000000e+00 xco 1.104635e-08 1.092440e-16...

Ok here's the updated plots with the improvements I made:  And now we are back to something reasonable: ``` Time spent in InitData(): 16.04659739 Time spent in Evolve(): 4.62255488...

Checklist: - [x] check netcdf - [ ] rerun all the tests - [x] check abl_sampling test and large test files from a GPU run - [x] check free surface...

Because I can't leave well enough alone... That last simplification means that now we are even faster than the main branch in the init (2X speedup). Turns out TANSTAAFL is...

> not sure if you've incorporated my additions to probe sampler, but hopefully those won't be too bad to address. Yup I merged those in when you made the changes....

That last one made the init 10% faster than the one before. ``` Time spent in InitData(): 4.615414819 Time spent in Evolve(): 5.122051878 ``` Not enough to claim that the...

Free surface sampler looks fine. [plot_sampling_native.pdf](https://github.com/user-attachments/files/17344725/plot_sampling_native.pdf)

abl_sampling_netcdf with 10 ranks on Kestrel: This PR: ``` Time spent in InitData(): 0.38666686 Time spent in Evolve(): 1.077089947 ``` `main` branch: ``` Time spent in InitData(): 0.483688814 Time spent...

got nccmp added to kestrel: ``` ❯ nccmp -d post_processing_dev/volume_sampling00000.nc post_processing_main/volume_sampling00000.nc ❯ nccmp -d post_processing_dev/volume_sampling200000.nc post_processing_main/volume_sampling200000.nc ``` good to go for netcdf