Reference data

The tests compare Runes2P with the output of DREAM and of GSL stored in six reference sets, so they do not run DREAM themselves. The DREAM cases are DREAM's public examples, run with their own settings and a few stated overrides, and synthetic cases written for one comparison each. The generator scripts, the build of DREAM and the regeneration are in reference/ of the repository.

The sets

Set	Source	Files
`dream-v26.5`	DREAM release v26.5, commit `ecdd5e146537c77602c9d7cc76b36100200e4b9a`	16
`dream-v26.5-disruption`	DREAM release v26.5, commit `ecdd5e146537c77602c9d7cc76b36100200e4b9a`: the disruption chain from its initialization	2
`dream-v26.5-initial-state`	DREAM release v26.5, commit `ecdd5e146537c77602c9d7cc76b36100200e4b9a`: the starts of a run and the collisional heating of the cold electrons	4
`dream-v26.5-runaway-grid`	DREAM release v26.5, commit `ecdd5e146537c77602c9d7cc76b36100200e4b9a`: the disruption with a runaway grid	1
`dream-boltzmann-eqterm-93d27e17`	DREAM branch `boltzmann-eqterm`, commit `93d27e1763827e19dafb1fe790e99c2688ad6d62`	6
`gsl-2.8`	GSL 2.8 of conda-forge	1

DREAM is at https://github.com/chalmersplasmatheory/DREAM. The branch boltzmann-eqterm holds the Boltzmann knock-on operator, which is in no release, so the cases of that operator need the branch. The release v26.5 is not an ancestor of the branch. The set gsl-2.8 holds the values of GSL's two-dimensional interpolation as DREAM uses it for the ADAS coefficients, on a synthetic table (reference/gsl/bicubic_reference.c).

The DREAM file and line citations in the code and on these pages refer to DREAM commit 93d27e1 (branch boltzmann-eqterm) unless they name another version.

Each set is published as a tarball with its files at the top level:

The files and their tests

File	Set	Case script	Test
`dream_avalanche_kinetic.h5`	`dream-boltzmann-eqterm-93d27e17`	`avalanche_kinetic.py`	`test_avalanche_dream.jl`
`dream_cold_energy.h5`	`dream-v26.5`	`cold_energy_fluid.py`	`test_cold_energy_dream.jl`
`dream_collisions.h5`	`dream-v26.5`	`collision_frequencies.py`	`test_collisions_dream.jl`
`dream_cooling_positivity.h5`	`dream-v26.5`	`cooling_positivity.py`	`test_cooling_dream.jl`
`dream_coronal_start.h5`	`dream-v26.5-initial-state`	`initial_state.py case=coronal`	`test_initial_state_dream.jl`
`dream_disruption_chain_small.h5`	`dream-v26.5-disruption`	`disruption_kinetic.py`	`test_disruption_initialization_dream.jl` (runs 0 to 2, the initialization), `test_disruption_chain_dream.jl` (runs 0 to 4, end to end)
`dream_disruption_runaway_small.h5`	`dream-v26.5-runaway-grid`	`disruption_kinetic.py --runaway-grid --window-start 170`	`test_disruption_runaway_dream.jl`
`dream_disruption_small.h5`	`dream-v26.5`	`disruption_kinetic.py`	`test_disruption_dream.jl`, `test_disruption_runaway.jl` (the quench with a runaway grid, from `f_re` zero)
`dream_disruption_superthermal_channels_small.h5`	`dream-v26.5-initial-state`	`disruption_kinetic.py --mode superthermal --test-data-channels`	`test_collisional_heating_dream.jl`
`dream_disruption_superthermal_small.h5`	`dream-v26.5-disruption`	`disruption_kinetic.py --mode superthermal`	`test_disruption_superthermal_dream.jl` (runs 0 to 4, end to end, and the phases from DREAM's state)
`dream_dreicer_kinetic.h5`	`dream-v26.5`	`dreicer_kinetic.py`	`test_runaway_physics.jl`
`dream_example_avalanche.h5`	`dream-boltzmann-eqterm-93d27e17`	`example_avalanche.py`	`test_example_avalanche.jl`
`dream_example_runaway.h5`	`dream-v26.5`	`example_runaway.py`	`test_example_runaway.jl`
`dream_flux_limiters.h5`	`dream-v26.5`	`flux_limiters.py`	`test_flux_limiters_dream.jl`
`dream_ion_energy.h5`	`dream-v26.5`	`ion_energy_fluid.py`	`test_ion_energy_dream.jl`
`dream_ion_rates.h5`	`dream-v26.5`	`ion_rates_fluid.py`	`test_ion_rates_dream.jl`
`dream_kinetic_ionization.h5`	`dream-v26.5`	`kinetic_ionization.py`	`test_kinetic_ionization_dream.jl`
`dream_kinetic_operator.h5`	`dream-v26.5`	`kinetic_operator.py`	`test_kinetic_operator_dream.jl`
`dream_knockon_operator.h5`	`dream-boltzmann-eqterm-93d27e17`	`knockon_operator.py`	`test_knockon_operator.jl`
`dream_pmax_boundary.h5`	`dream-v26.5`	`pmax_boundary.py`	`test_pmax_boundary_dream.jl`
`dream_radial_transport.h5`	`dream-v26.5`	`radial_transport.py`	`test_radial_transport_dream.jl`
`dream_rp_avalanche.h5`	`dream-v26.5`	`rp_avalanche.py`	`test_avalanche_rp_dream.jl`
`dream_runaway_heating.h5`	`dream-v26.5-initial-state`	`initial_state.py case=heating`	`test_collisional_heating_dream.jl`, `test_runaway_heating_dream.jl`
`dream_runaway_rates.h5`	`dream-v26.5`	`runaway_rates.py`	`test_runaway_rates.jl`, `test_runaway_physics.jl`
`dream_runaway_shapes.h5`	`dream-v26.5-initial-state`	`initial_state.py case=shapes`	`test_initial_state_dream.jl`
`dream_self_consistent_field.h5`	`dream-v26.5`	`self_consistent_field.py`	`test_self_consistent_dream.jl`
`dream_time_stepping.h5`	`dream-boltzmann-eqterm-93d27e17`	`time_stepping.py`	`test_time_stepping_dream.jl`
`dream_time_stepping_adaptive.h5`	`dream-boltzmann-eqterm-93d27e17`	`time_stepping_adaptive.py`	`test_time_stepping_dream.jl`
`dream_two_grids.h5`	`dream-boltzmann-eqterm-93d27e17`	`two_grids.py`	`test_two_grids_dream.jl`
`gsl_bicubic_reference.txt`	`gsl-2.8`	`reference/gsl/bicubic_reference.c`	`test_adas.jl`

The case scripts are in reference/dream/cases/, and each states its case, its overrides and the test that reads its output. The tests are in test/.

How the tests use them

test/Artifacts.toml declares each set as a lazy artifact, with the URL of its tarball and its hashes. reference_file(name) of test/helpers.jl gives the path of a data file in the artifact of its set, which is downloaded at its first use. The table REFERENCE_FILES of test/helpers.jl names the set of every file. For the output of a DREAM example it also names the last DREAM commit that changed the example.

Every DREAM file records dream_commit, dream_branch and petsc_version as attributes. test/test_kinetic_operator_dream.jl checks the commit and the branch of every DREAM file against its set. With the DREAM checkout of a set's build at reference/dream/<build>/DREAM it also checks that the checkout is at that commit, and that each example is unchanged since its recorded commit. The build is named after its set; dream-v26.5-disruption, dream-v26.5-initial-state and dream-v26.5-runaway-grid use that of dream-v26.5. Without the checkout these checks are skipped.

Test nodes without network need the sets installed beforehand, on a node that has network:

julia --project=test -e 'using Pkg; Pkg.Artifacts.ensure_all_artifacts_installed("test/Artifacts.toml"; include_lazy=true)'

Regeneration

reference/dream/README.md describes the build and the regeneration. The toolchain is the pixi environment of reference/dream/pixi.toml, with PETSc built from source for both DREAM builds. build_all.sh checks out the commit of each DREAM build and builds it, and run_all.sh runs the cases of the named sets into out/<set>/. run_all.sh checks that the DREAM checkout of each set is at its commit before it runs a case. The HDF5 files hold a created time stamp, so a rerun with the same build gives the same data but not the same bytes.