Castro
Castro copied to clipboard
revisit ppm_temp_fix
At it's peak, we had 3 options for including temperature in the reconstruction of interface states:
-
ppm_temp_fix = 1
would reconstuct T and then use the EOS to get p, and the rest of the algorithm would proceed as normal -
ppm_temp_fix = 2
modifies the Riemann solver to enforce thermodynamic consistency on the interfaces by recomputing p via the EOS -
ppm_temp_fix = 3
used a tau, u, T eigensystem for the characteristic tracing. This was removed in #764
It turns out that ppm_temp_fix = 1
only works with ppm_predict_gammae = 1
due to an oversight. We have also not been using it. In PR #780 , we will remove the gammae reconstruction, since that has not been shown to help much.
We will also be temporarily removing ppm_temp_fix = 1
in that same PR. This removal means that there will be no more EOS calls in any of the reconstruction, so we can go ahead with the port to C++ (since not having the C++ EOS is the current blocker).
Once the C++ EOS PRs are merged, we should go back and explore these temperature fixes again, in a more controlled study, to understand where they might be beneficial.