Is there any performance or space overheads for using 'tag' in MPI?

[seongyun_k_]

Hi, 

I wonder if  there can be any performance or space overheads for using 'tag' in MPI.

It would depend on how it is implemented, but I guess if it used a sort of hash map and keep (possible) multiple buffers for the buckets. The process memory size can blow up.

Can this happen?

[Judith_W_Intel]

Please ask this here:

https://software.intel.com/en-us/forums/intel-clusters-and-hpc-technology

[James_T_Intel]

What tag are you talking about?  Is this related to the MPI tag used in point-to-point communications?  This is part of the standard and is required.