Dear all,
Following my presentation of Geodesic at the previous weekly users call, I would like to give a brief status on the one item I flagged as an open limitation: exercising the multi-process (MPI) configuration on a Carpet AMR evolution.
The single-process + AMR runs already reproduce the unigrid reference orbits with the three samplers, with no lost particles and no NaNs. Since then I have been working on the MPI path, but the work is still under test: current MPI + AMR runs still show particle loss with non-finite metric samples, and I am debugging the cause. I do not yet have a clean, reproducible result, so I have no conclusion to report this week.
In short: no new deliverable on the MPI + AMR path this week -- it is still under verification. I will report a concrete result as soon as the verification passes.
I remain available to discuss, and as before I am most comfortable continuing in writing. Any questions are welcome.
Best regards, Yuhua Li