AFD-Illinois / iharm3d

Hybrid MPI/OpenMP 3D HARM with vectorization
Other
12 stars 9 forks source link

Default to a single MPI process for all problems #39

Closed bprather closed 3 years ago

bprather commented 3 years ago

This has now proven an issue for both reviewers of the iharm3D paper, and understandably so as it's maddeningly counter-intuitive. Better to default to the configurations people will run first, not what's useful for stress-testing.

The tests already set the number of MPI processes for themselves in most cases, but it will take a bit to verify that this doesn't break the automated testing in the 'dev' branch.