RRTMGP test

[nataraj2]

Hi,

Ran this RRTMGP test to generate the baseline, and plotted and printed out the flux components (2d arrays of dimensions 128 x 43). The images (as 2d arrays) for the fluxes are below. The order is sw_flux_dir, sw_flux_up, lw_flux_dn, lw_flux_up, sw_flux_dn and a zoomed in version of sw_flux_dn. All fluxes look alright, except the sw_flux_dn which looks overall correct, but at 2 points the values are NaNs. The last image (zoomed in view) shows the 2 points with NaNs. Just wanted to know if this is what you observe as well?

Thanks.

[mahf708]

@nataraj2 this is just a dummy-data generation. Have you tried running the other two tests in the same folder? That would be a far better check. I can also run this generation and attach the file here in a bit

[nataraj2]

Thanks @mahf708
If you could give a data file with what you get doing this baseline test case, that would be great.
Meanwhile, I will try to run the other tests as well.

[mahf708]

This link will download the generated file. Can you take a look?

I looked inside and I couldn't find any NaNs. I created the test on perlmutter gpu ...

[nataraj2]

Thanks a lot @mahf708

I hadn't realized that the code writes out a netcdf file for the baseline as output.

I was writing my own data file and accessing the fluxes directly as sw_flux_up(i,j), for eg. It gave no errors. And the data looked overall correct except for the glitches. I now realize that it should be done as

auto sw_flux_up_data = sw_flux_up.data();
for (int i=0; i<sw_flux_up.extent(0); i++) {
    for (int j=0; j<sw_flux_up.extent(1); j++) {
        printf("value = %f at %i,%i\n", sw_flux_up_data[j*nx + i], i, j);
    }
  }

That fixed my issue of writing the data. Thanks a lot for the help. Really appreciate it.

[mahf708]

Ok, sounds good! Happy to help :)