Closed stvdwtt closed 2 years ago
@Rombur -- I just realized I didn't merge the current master into this so the thermal conductivity input files are wrong. I'll fix that early this afternoon.
Do you know what we should do about the cuda failures in set_with_experimental_data
? For now should I only have that test run on the host?
There is a bug with rank
, you need to get it before using it (dealii::Utilities::MPI::this_mpi_process(comm)
will give to you). It also look like clang-sanitizer is not happy.
Clang-tidy and CUDA fails on the same error inside deal.II https://cloud.cees.ornl.gov/jenkins-ci/blue/organizations/jenkins/adamantine/detail/adamantine/136/tests/
An error occurred in line <2241> of file </scratch/source/dealii/source/dofs/dof_tools.cc> in function
void dealii::DoFTools::internal::{anonymous}::map_dofs_to_support_points(const dealii::hp::MappingCollection<dim, spacedim>&, const dealii::DoFHandler<dim, spacedim>&, std::map<long unsigned int, dealii::Point<spacedim> >&, const dealii::ComponentMask&) [with int dim = 3; int spacedim = 3]
The violated condition was:
fe_collection[fe_index].has_support_points()
Additional information:
You are trying to access the support points of a finite element that
either has no support points at all, or for which the corresponding
tables have not been implemented.
The reason it passes on the CPU is because we use the Release
version and so we skip, deal.II's assert. You should be able to reproduce the error using the Debug
version.
@Rombur, I just want to confirm -- since the deal.II PR removed the assert (instead of moving it inside the masked off loop), the code is in this PR is good as it is and doesn't need the mask?
That's right. The mask is actually for the vector component not the fe index, so there is no point in using it.
This PR adds data assimilation capabilities to the adamantine app.
Main changes:
Other changes:
run_ensemble
to make it easier to read through (at least for me)Notes: