Port wgtfac_c

[jonasjucker]

Port stencil from src/atm_dyn_iconam/mo_vertical_grid.f90.

       DO jk = 2, nlev
          p_nh(jg)%metrics%wgtfac_c(1:nlen,jk,jb) =  &
           (p_nh(jg)%metrics%z_ifc(1:nlen,jk-1,jb) - &
            p_nh(jg)%metrics%z_ifc(1:nlen,jk,jb)) /  &
           (p_nh(jg)%metrics%z_ifc(1:nlen,jk-1,jb) - &
            p_nh(jg)%metrics%z_ifc(1:nlen,jk+1,jb))
        ENDDO

It does not include the edge-cases k=1 and k=nlevp1
What to do about that?

[jonasjucker]

launch jenkins spack

[jonasjucker]

cscs-ci run default

[halungge]

Port stencil from src/atm_dyn_iconam/mo_vertical_grid.f90.

       DO jk = 2, nlev
          p_nh(jg)%metrics%wgtfac_c(1:nlen,jk,jb) =  &
           (p_nh(jg)%metrics%z_ifc(1:nlen,jk-1,jb) - &
            p_nh(jg)%metrics%z_ifc(1:nlen,jk,jb)) /  &
           (p_nh(jg)%metrics%z_ifc(1:nlen,jk-1,jb) - &
            p_nh(jg)%metrics%z_ifc(1:nlen,jk+1,jb))
        ENDDO

It does not include the edge-cases k=1 and k=nlevp1 What to do about that?

Port stencil from src/atm_dyn_iconam/mo_vertical_grid.f90.

       DO jk = 2, nlev
          p_nh(jg)%metrics%wgtfac_c(1:nlen,jk,jb) =  &
           (p_nh(jg)%metrics%z_ifc(1:nlen,jk-1,jb) - &
            p_nh(jg)%metrics%z_ifc(1:nlen,jk,jb)) /  &
           (p_nh(jg)%metrics%z_ifc(1:nlen,jk-1,jb) - &
            p_nh(jg)%metrics%z_ifc(1:nlen,jk+1,jb))
        ENDDO

It does not include the edge-cases k=1 and k=nlevp1 What to do about that?

I assume the originally initalize with 0 everywhere and then overwrite the "inner" levels. That is what usually happens in ICON. Eventually we will then do the same thing. Allocate with the right size and 0 and overwrite the inner levels by setting the vertical_start, vertical_end correctly. Exactly as you do in the verification test.

[halungge]

Looks good, only some small clean ups in the test.

[halungge]

If you have several test files in this folder you might want to put the fixture imports into a conftest.py pytest checks that one automatically for fixtures and makes them available to the tests. conftest.py would essentially just look like this:

from icon4py.model.common.test_utils.datatest_fixtures import (  # noqa: F401
    damping_height,
    data_provider,
    datapath,
    decomposition_info,
    download_ser_data,
    experiment,
    grid_savepoint,
    icon_grid,
    interpolation_savepoint,
    processor_props,
    ranked_data_path,
)

[jonasjucker]

Thanks for this info.

I would keep that for the next PR that add another stencil to metrics.
For now I think it is fine as is

[jonasjucker]

The thing with the edge-case is that the field is filled for these indices as well, just below the loop:

          p_nh(jg)%metrics%wgtfac_c(1:nlen,1,jb) = &
           (p_nh(jg)%metrics%z_ifc(1:nlen,2,jb) -  &
            p_nh(jg)%metrics%z_ifc(1:nlen,1,jb)) / &
           (p_nh(jg)%metrics%z_ifc(1:nlen,3,jb) -  &
            p_nh(jg)%metrics%z_ifc(1:nlen,1,jb))
          p_nh(jg)%metrics%wgtfac_c(1:nlen,nlevp1,jb) = &
           (p_nh(jg)%metrics%z_ifc(1:nlen,nlev,jb) -    &
            p_nh(jg)%metrics%z_ifc(1:nlen,nlevp1,jb)) / &
           (p_nh(jg)%metrics%z_ifc(1:nlen,nlev-1,jb) -  &
            p_nh(jg)%metrics%z_ifc(1:nlen,nlevp1,jb))

So the current implementation of this stencil does not return the correct field.

Is this correct that way?
How do you plan to solve the edge-cases then?

[github-actions]

Mandatory Tests

Please make sure you run these tests via comment before you merge!

• cscs-ci run default
• launch jenkins spack

Optional Tests

To run benchmarks you can use:

• cscs-ci run benchmark

In case your change might affect downstream icon-exclaim, please consider running

• launch jenkins icon

For more detailed information please look at CI in the EXCLAIM universe.

[jonasjucker]

launch jenkins spack

[jonasjucker]

cscs-ci run default