Overly long running physics tests

[ax3l]

Let's reduce the runtime of these CI tests to speed up CI runs:

• 80s: averaged_galilean_3d_psatd -> @oshapoval CI: Reduce Runtime of Some Galilean Tests #2625
• 80s: comoving_2d_psatd_hybrid -> @oshapoval -> add to CI: Reduce Runtime of Some Galilean Tests #2625?
• 83s: galilean_2d_psatd_hybrid -> @oshapoval -> add to CI: Reduce Runtime of Some Galilean Tests #2625?
• 80s: multi_J_2d_psatd -> @EZoni CI: Reduce Runtime of Test multi_J_2d_psatd #2638
• 45s: subcyclingMR -> @EZoni CI: Reduce Runtime of Test subcyclingMR #2632
• 336-370s: particles_in_pml_3d_MR
• 128s: pml_psatd_dive_divb_cleaning
• 168s: multi_J_rz_psatd

Goal: 2-20 sec are the usual runtimes :)

[EZoni]

I'll work on multi_J_2d_psatd because I need anyways to set up a similar test with PML boundary conditions instead of damped for #2603.

[ax3l]

Wuhu, thank you for the help @oshapoval @EZoni 🎉

[ax3l]

I saw in recent CI (#2556) that

• averaged_galilean_3d_psatd still runs around 45 seconds.
• averaged_galilean_3d_psatd_hybrid 69 seconds
• galilean_3d_psatd 71 seconds
• galilean_3d_psatd_current_correction 55 seconds

Do you think we can tune it slightly more?

I also noticed that these tests use 1 MPI rank and 1 OMP thread, we could use a 2 in either numprocs or numthreads.

[EZoni]

We can probably use more than one core provided that we reduce the spectral order (otherwise the number of guard cells will become incompatible with the size of each subdomain, since the mesh is very coarse in these 3D tests). I will take a look at this.

[ax3l]

Found another few long runners:

Cartesian 3D

particles_in_pml_3d_MR

mpiexec -n 2 ./particles_in_pml_3d_MR.ex inputs_mr_3d  diag1.file_prefix=particles_in_pml_3d_MR_plt   amrex.abort_on_unused_inputs=1 amrex.fpe_trap_invalid=1 amrex.fpe_trap_zero=1 amrex.fpe_trap_overflow=1
   Execution time: 336.776 s

pml_psatd_dive_divb_cleaning

mpiexec -n 2 ./pml_psatd_dive_divb_cleaning.ex inputs_3d  diag1.file_prefix=pml_psatd_dive_divb_cleaning_plt   amrex.abort_on_unused_inputs=1 amrex.fpe_trap_invalid=1 amrex.fpe_trap_zero=1 amrex.fpe_trap_overflow=1 warpx.do_similar_dm_pml = 0
   Execution time: 182.536 s

RZ or NOMPI

multi_J_rz_psatd

mpiexec -n 2 ./multi_J_rz_psatd.ex inputs_rz  diag1.file_prefix=multi_J_rz_psatd_plt   amrex.abort_on_unused_inputs=1 amrex.fpe_trap_invalid=1 amrex.fpe_trap_zero=1 amrex.fpe_trap_overflow=1 warpx.do_dynamic_scheduling=0 warpx.serialize_ics=1
   Execution time: 167.843 s

[ax3l]

X-ref: #2704 for the last long tests at the moment :)

Thank you @EZoni and @oshapoval 🎉