Add PICMI Scripts for LWFA Tests

[EZoni]

Add PICMI scripts and CI tests for LWFA in 1D, 2D, 3D and RZ geometries, mimicking the original input files in Examples/Physics_applications/laser_acceleration/.

We can probably leave the following three items to separate follow-up PRs:

• 2D w/ boosted frame
• Add electron beam to all tests
• Set same warpx.cfl in all tests

[EZoni]

@ax3l Did you want me to add these as CI tests as well? I'm not sure I understood that correctly. If not, I will have to rename the PICMI scripts avoiding the PICMI_inputs name tag, because we have a style check which requires all files in the Examples/ directory that start with inputs or PICMI_inputs to be tested automatically.

[dpgrote]

There is already the test case PICMI_inputs_laser_acceleration.py. Is there anything different in what is being tested between it and the new 3d test case? Perhaps PICMI_inputs_laser_acceleration.py can be deleted?

[EZoni]

@dpgrote Thank you, that's a good point. I think we can remove that script, because it doesn't reflect 100% the corresponding input file inputs_3d, which I understood was among the aims of having this new set of scripts. Let's see what @ax3l says. Anyways, I'll add a point to the PR description, to make sure that we don't forget.

[ax3l]

Yes, a perfect copy of the inputs_... scripts in PICMI would be fantastic.

Feel free to remove the other one to avoid confusing as Dave suggests.

yes, please add the new PICMI inputs to Warpx-tests.ini so they are covered/do not run out-of-date.

[ax3l]

warpx.cfl is set to 0.9 in the original 1D input file and to 1.0 in the original 2D, 3D and RZ input files. Do we want to set warpx.cfl = 0.9 (or some other values between 0.9 and 1.0) for all cases instead?

Let's fix that in the 1D case. 0.9 is an old value, we should not suggest that to users.
cc @prkkumar

[ax3l]

The electron beam is missing from the original 1D and 3D input files, hence it's NOT been added to the corresponding PICMI scripts either. Do we want to have an electron beam also in the 1D and 3D tests instead?

I would actually say we should add a witness beam in all those inputs. makes for a nice test and vis case as well.

This could replace #1631 then.

[ax3l]

Thanks a lot for adding these, @EZoni !! 🙏

[EZoni]

This test was associated with the old PICMI script that was removed, and I think it was not using mesh refinement (despite what the name tag MR may suggest).

[EZoni]

I suggest that we push any changes that affect the benchmarks (e.g. CFL value or adding a witness beam) in a separate PR. Then, this PR should not require any benchmark reset. Moreover, the new benchmarks for the new Python tests will be the exact copy of the existing ones, if the PICMI scripts mimic the corresponding original input files correctly.

[EZoni]

Keep in mind that this is actually a new, different test, but shows up as a benchmark change because an old test with the same name (but actually different setup) was removed (see other comment on WarpX-tests.ini).

[EZoni]

Note that the changes in

• .github/workflows/intel.yml
• .github/workflows/macos.yml
• .github/workflows/ubuntu.yml

where I simply replaced the old PICMI script PICMI_inputs_laser_acceleration.py with the new PICMI script PICMI_inputs_3d.py (also tested successfully in CI), cause some failures in the macOS and Intel builds. Have not understood why yet.

[ax3l]

I merged development now (to incorporate #2556) and I fixed the three mentioned tests (same as #2673): they lacked the first line to declare the python interpreter :)

[ax3l]

Thank you :)