Skip to content

Conversation

@dalcinl
Copy link
Contributor

@dalcinl dalcinl commented Feb 8, 2025

No description provided.

@dalcinl dalcinl requested review from hzhou and jeffhammond February 8, 2025 10:39
@dalcinl
Copy link
Contributor Author

dalcinl commented Feb 10, 2025

@hzhou We need these new routines in MPICH.

@dalcinl
Copy link
Contributor Author

dalcinl commented Feb 10, 2025

@jeffhammond In the new MPI std document we have "It is erroneous to call this function with an invalid handle argument". However, these routines do not follow the usual "return error code" pattern, and therefore is not clear at all how error handling should be implemented and eventually handled by users. IMHO, this should be amended following the model of c2f/f2c routines, that is, a) null handles can be properly mapped back and forth, but invalid MPI handles are mapped to invalid C int handles and vice versa. If an MPI implementation can afford the overhead of distinguishing valid vs invalid handles via some internal mechanism, then the 0 for the C int value and NULL (e.g. for MPI_Comm in the ABI) would be great choices for returning invalid handles (at least within the new ABI).

hzhou added a commit to hzhou/mpich that referenced this pull request Feb 11, 2025
Sync up mpi_abi.h from upstream -
mpi-forum/mpi-abi-stubs#39
hzhou added a commit to hzhou/mpich that referenced this pull request Feb 11, 2025
Sync up mpi_abi.h from upstream -
mpi-forum/mpi-abi-stubs#39
@jeffhammond
Copy link
Member

We say it is erroneous to place the burden of checking entirely on the user, to allow trivial implementations.

@dalcinl
Copy link
Contributor Author

dalcinl commented Feb 13, 2025

We say it is erroneous to place the burden of checking entirely on the user, to allow trivial implementations.

OK, I clearly misunderstood the actual meaning.

hzhou added a commit to hzhou/mpich that referenced this pull request Feb 24, 2025
Sync up mpi_abi.h from upstream -
mpi-forum/mpi-abi-stubs#39
hzhou added a commit to hzhou/mpich that referenced this pull request Mar 3, 2025
Sync up mpi_abi.h from upstream -
mpi-forum/mpi-abi-stubs#39
hzhou added a commit to pmodels/mpich that referenced this pull request Mar 4, 2025
Sync up mpi_abi.h from upstream -
mpi-forum/mpi-abi-stubs#39
@dalcinl
Copy link
Contributor Author

dalcinl commented Mar 6, 2025

@jeffhammond I think this one is also ready for merge.

@jeffhammond
Copy link
Member

Does this remove all f2c/c2f?

@dalcinl
Copy link
Contributor Author

dalcinl commented Mar 6, 2025

Does this remove all f2c/c2f?

No, not at all! Should it??

@dalcinl dalcinl self-assigned this Mar 6, 2025
@dalcinl
Copy link
Contributor Author

dalcinl commented Mar 6, 2025

@jeffhammond I brought your commit from #48 to update this PR. Now the mpi4py build test pass, although with a branch other than master.

If everything is good with this PR from your side, please approve.
Leave the merge to me such that I can address the testing part.

@dalcinl dalcinl marked this pull request as draft March 6, 2025 13:18
@dalcinl dalcinl marked this pull request as ready for review March 6, 2025 15:17
@dalcinl dalcinl merged commit fca839f into main Mar 6, 2025
13 checks passed
@dalcinl dalcinl deleted the int-handle branch March 6, 2025 15:19
hzhou added a commit to pmodels/mpich that referenced this pull request Aug 25, 2025
Sync up mpi_abi.h from upstream -
mpi-forum/mpi-abi-stubs#39
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants