Difference between project Riemann1 and Shocktest?

[henry2004y]

The two projects Riemann1 and Shocktest has very similar setups. The required input parameters for the two projects are identical. The differences seem to be

1. Shocktest is a subclass of TriAxisSearch (which is a subclass of Project), whereas Riemann1 is a direct subclass of Project.
2. The default cfg files have a different set of input parameters.
3. Riemann1 has more additional files.

---

I then check what makes the class TriAxisSearch different from a base Project class. It seems that for TriAxisSearch an optimized findBlocksToInitialize() is defined to start the loop from the initial bulk velocity, which makes it faster.

The findBlocksToInitialize() method for base TriAxisSearch has the following comment

vlasiator/projects/projectTriAxisSearch.cpp

Lines 32 to 34 in af6b9b6

	* WARNING This assumes that the velocity space is isotropic (same resolution in vx, vy, vz).
	*/
	std::vector<vmesh::GlobalID> TriAxisSearch::findBlocksToInitialize(SpatialCell* cell,const uint popID) const {

but inside the function the vspace resolutions dvx, dvy, dvz seems to be completely decoupled such that different resolutions in the vx, vy, vz dimension are allowed?

vlasiator/projects/projectTriAxisSearch.cpp

Lines 47 to 49 in af6b9b6

	creal dvxCell = cell->get_velocity_grid_cell_size(popID,refLevel)[0];
	creal dvyCell = cell->get_velocity_grid_cell_size(popID,refLevel)[1];
	creal dvzCell = cell->get_velocity_grid_cell_size(popID,refLevel)[2];

---

Should we keep both projects? Which one should we use to perform a 1D shock tube test?

[ykempf]

I suppose the one using the TriAxisSearch will be more recent, as that's the optimised block initialisation version. Now you will have the most recent shock code in the IPShock project, that should easily be "simplified" to do a shock tube if the perturbations are set to zero and the shock width is made small.

[markusbattarbee]

Indeed recent shock tests have been made with the IPshock class, but that one makes some assumptions about the frame of reference. So I think there is merit in perhaps keeping a simpler shocktest version alive as well.
If indeed those are the only differences then I'd say keep the shocktest and ditch the Riemann, but even ditching that one is optional and won't really change things.