Jeff Hammond

I am less interested in configuring it and more interested in being able to figure out what the limit is, so I don't write bad tests. I know I can't...

@hzhou the specific request I would make on this issue is for MPICH to provide a way to query the maximum number of live requests permitted at a time so...

Can't you do that by writing PMPI interposition of functions that produce and consume requests? https://github.com/LLNL/wrap will autogenerate them? That's going to be a lot easier than modifying MPICH.

> > I didn't include any RMA or reductions, but I'd be happy to do it if there's interest. > > RMA accumulate routines cannot be implemented in BigMPI, as...

> > However, if BigMPI is ever added, it has to be as comprehensible as possible, and even more importantly, it has to be correct. > > The BigMPI paper...

> 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. > Perhaps we should...

> In private communication, @jeffhammond warned me that BigMPI never cared about user-defined datatypes. Looking at the BigMPI datatype is constructed, I'm not sure things would work for arbitrary user-defined...

i don't understand why an extra collective would be required to implement `MPI_IN_PLACE`. as i said, i don't even know that it's broken. i just didn't test it. at this...

Installing BLAS with Apt works around this, but I would really like to be able to point CMake at a specific BLAS library and have it use that.

It's probably my bug. I'll work on it tomorrow or Friday. Please just use MPICH 4.2+ for now. I haven't worked too much on implementation portability because the features were...