Lisandro Dalcin

> > Limitations is not my main concern, correctness is. Also not slowing down the small-count case. > > It's one branch in BigMPI to filter the small-count case. >...

> i don't understand why an extra collective would be required to implement MPI_IN_PLACE That comment was just pure speculation from my side. What I was trying to say is...

Superseded by #628, which only handles the minimum feature set we are sure works with predefined and user-defined datatypes, that is, BigMPI's `MPI_Type_contiguous_x`. From the Python side, `MPI.Datatype.Create_contiguous(count)` will create...

@jacoterh Sorry for the long time without reply. Please note that the mpi4py-mpich on PyPI is not affiliated to the mpi4py project in any way. It was an upload made...

> Maybe somewhat related, but I noticed that [mpi4py-mpich](https://pypi.org/project/mpi4py-mpich/) wasn't updated with the 4.0.0 release. Is that still planned? @EwoutH Did you read my previous comment in this issue? https://github.com/mpi4py/mpi4py/issues/508#issuecomment-2184913330.

> I have two questions: can I dispense with `mpi4py-mpich` and simply use an appropriate version of `mpi4py` + system libraries for MPICH and OpenMPI? That is what you seem...

> but for specific project reasons I need to use 22.04. I totally understand. My comment was actually a warning/reminder for you to add the missing symlink `libmpi.so.12`, otherwise the...

@xiezhipeng-git > pip can not install mpi4py-mpich Did you read all my [previous comments](https://github.com/mpi4py/mpi4py/issues/508#issuecomment-2184913330) in this issue? > pip and conda can not install mpi4py in python 3.12 *pip*: You...

The [`mpi4py-mpich` entry on PyPI](https://pypi.org/project/mpi4py-mpich/) has been archived.

Indeed, this is an mpi4py bug trying to workaround issues for MPICH version 3.0 that was released 12 years ago.