Fix SUNDIALS Context construction for MPICH
#2678
Conversation
![github-actions[bot]](https://avatars.githubusercontent.com/in/15368?s=60&v=4){kind=small}
dschwoerer
force-pushed
the
cast-sagitter
branch
from
c8591c3
to
aa242f8
Compare
|
clang-tidy review says "All clean, LGTM! 👍" |
|
clang-tidy review says "All clean, LGTM! 👍" |
`MPI_COMM_WORLD` can be int, thus pass the BOUT++ communicator Co-authored-by: Peter Hill <peter.hill@york.ac.uk>
Do not take the address of the pointer. Compiles with mpich and openmpi
|
clang-tidy review says "All clean, LGTM! 👍" |
This is needed so we can get a pointer to the communicator, as is needed by SUNContext
pass MPI_Comm* to SUNContext_Create as indicated in the docs https://sundials.readthedocs.io/en/latest/sundials/SUNContext_link.html
|
clang-tidy review says "All clean, LGTM! 👍" |
|
clang-tidy review says "All clean, LGTM! 👍" |
|
clang-tidy review says "All clean, LGTM! 👍" |
|
Did this change run on both MPICH and OpenMPI? If so, it can be merged right now, if not, please can someone test it? I guess this is one case where we definitely do want to test across multiple MPI implementations! |
|
Yes, I did build it against both MPIs with sundials 6.5.0-1.fc39: With these patches on top of v5.0.0: |
ZedThree
changed the title
Context construction for MPICH
Resolves #2677