Closed stvdwtt closed 1 month ago
@stvdwtt the build_time_log.txt
and the .csv
files are missing
Sorry about that, here they are: observations.zip
For me the code crashes inside ArborX after using all the memory available. I think that the localization cutoff distance
is too large. There are 58140 supports points but the for first 1000 points, ArborX finds 22091450 points inside the cutoff radius. If the cutoff distance is physical, we need to rethink that part of the code.
Closing since this isn't an ArborX issue, it's a user-issue.
For the particular input files below, rank 0 of adamantine hangs in
DataAssimilator::update_covariance_sparsity_pattern
.It happens during this ArborX call: https://github.com/adamantine-sim/adamantine/blob/36131aaaa13433d2d5b6cb4b998e6c5d8f8862f8/source/DataAssimilator.cc#L541
The backtrace is rather long:
Inputs: failing_example.zip
For the three ensemble members in his failing test case, I've seen it hang for 1, 3, and 9 MPI ranks. Fortunately, that means it can be debugged with serial tests.
I wonder if this and #314 have the same root cause.
@Rombur, can you take a look?