I am getting an overflow error on the KB_morphtools branch within the connected components routine while trying to run on dogwood at 6^3 subdomains, a total discretization of 1560^3, 10 nodes, and 22 tasks per node (this is unevenly loaded -> 10*22 = 220 > 216. Evenly loading will cause the same problem, but 22 subdomains per node the is max you can do per node without running into a "cpu binding" error")
The relevant files are pushed to KB_morphtools and the error should repeat if you submit the "submit.sh" file on dogwood. The error will appear ~8 minutes into the run time. Go all the way to the bottom of the posted slurm file, the errors at the top do not prevent it from running successfully (yet).
Maybe this can be resolved by changing some long type variables to long long?
I am getting an overflow error on the KB_morphtools branch within the connected components routine while trying to run on dogwood at 6^3 subdomains, a total discretization of 1560^3, 10 nodes, and 22 tasks per node (this is unevenly loaded -> 10*22 = 220 > 216. Evenly loading will cause the same problem, but 22 subdomains per node the is max you can do per node without running into a "cpu binding" error")
The relevant files are pushed to KB_morphtools and the error should repeat if you submit the "submit.sh" file on dogwood. The error will appear ~8 minutes into the run time. Go all the way to the bottom of the posted slurm file, the errors at the top do not prevent it from running successfully (yet).
Maybe this can be resolved by changing some long type variables to long long?