Ben Wibking
Ben Wibking
Since the electron abundance normalization depends on the charge of all of the species, it should be moved into the network-specific files in Microphysics, or somehow re-written so it does...
This is nontrivial to do in a way that minimizes the inconsistency at the coarse-fine level boundaries. In order of increasing consistency (and complexity), there are three possible approaches one...
At outflow-outflow boundary corners, the ghost cells must take into account additional constraints to be consistent. See https://doi.org/10.2514/1.J055885. The fix is straightforward, but will require users to call a different...
This might improve stability in difficult situations, since the waves that are traveling away from the interface are deleted, and thus don't enter into the Riemann solver. We should not...
From the expressions given in [1][2][3], we can compute a discrete entropy inequality in each cell, and then fallback to first order when it is violated [3]. More work is...
Castro and other AMReX codes output a `jobinfo` file inside the plt00000/ directory that includes the entire ParmParse database, the git hash, the AMReX version, the compiler used, the cluster...
We should add a runtime parameter than shrinks the initial timestep by a given factor in order to avoid start-up error/stability issues. In Castro, this is done with this parameter:...
We have regression tests for problems that have exact (or semi-analytic) solutions, but we don't test whether we're getting better or worse answers for complicated problems that lack exact solutions....
A small fraction (about 3%) of plotfiles seem to be corrupted when running with `amrex.async_out=1` on Gadi in production AMR simulations (with the shock-cloud problem). This *only* affects runs with...
Managed memory has a very negative performance impact when using GPU-aware MPI. The last uses of managed memory in the code are in the Grackle data tables: https://github.com/search?q=repo%3Aquokka-astro%2Fquokka%20The_Managed_Arena&type=code