Closed TomMelt closed 4 days ago
In order to send halo regions via MPI in the dynamics, we need to know the size of the domains and the starting indices of the halo regions.
This PR adds functionality we need to parallelize nextsimdg dynamics.
The following improvements have been made.
DomainUtils
get_width
get_height
Edge
edges
NNBRS
std::cerr
exit(EXIT_FAILURE)
halo_starts
In order to send halo regions via MPI in the dynamics, we need to know the size of the domains and the starting indices of the halo regions.
This PR adds functionality we need to parallelize nextsimdg dynamics.
The following improvements have been made.
DomainUtils
to add helper methodsget_width
get_height
Edge
enums and arrayedges
for more readable loops over neighboursNNBRS
to use the newedges
enumsstd::cerr
messages are followed byexit(EXIT_FAILURE)
.TODO
halo_starts
)