Hui Zhou
Hui Zhou
Yes, filing as new issue will help clarifying the status.
@aditya-nishtala Could you retry the experiment using a debug-build and enable `MPIR_CVAR_DEBUG_SUMMARY=1` and then post one of the log? That may help identify whether the memory is allocated by MPICH...
Taking the difference, the memory increase are roughly linear to the number of nodes, ~55-65 MB/Node. @aditya-nishtala How many PPN (process per node)?
Thanks @nsdhaman . So that is roughly 6KB per connection.
Okay, I think the issue is we are allocating too much address table prepared for all possible connections. If we assume no application will use multiple VCI, we could configure...
> Did you do a performance test with this, or are these changes just based on the conversation earlier this week? Its on my TODOs to do some performance runs...
test:mpich/ch3/tcp
> [#7341 (comment)](https://github.com/pmodels/mpich/issues/7341#issuecomment-2726386689) > > Why 0 and not MPI_UNDEFINED? As @dalcinl mentioned, it is an MPICH backward-compatible behavior. For me, a non-existent datatype's size is `0` kinda makes sense....
> Is MPICH implementing the MPI 5.0 standard in the MPI ABI or will it make arbitrary changes? > > We discussed this semantic during a virtual meeting where you...
@dalcinl If we simply return error on unsupported builtin types, will you be able to deal with it (via error return or exception) in mpi4py?