[ofiwg] [libfabric-users] TX/RX data structures and data processing mode

[Ingerson, Alexia]

> I believe both OpenMPI and MPICH will work over the RxM and SHM providers
I can't verify that the SHM provider works on either, but I don't think it's a compatibility issue (since it works on Intel MPI) but rather a configuration/setting issue with exposing the provider. 


-----Original Message-----
From: Hefty, Sean 
Sent: Monday, March 19, 2018 10:00 AM
To: miharulidze at gmail.com
Cc: ofiwg at lists.openfabrics.org; libfabric-users at lists.openfabrics.org; Ilango, Arun <arun.ilango at intel.com>; Ingerson, Alexia <alexia.ingerson at intel.com>
Subject: RE: [libfabric-users] TX/RX data structures and data processing mode

> As far as I understand, this is sufficient, but not necessary to call 
> cq_read, for application, since we can use other functions of 
> libfabric API for data progress (fi_cntr_read) in FI_PROGRESS_MANUAL.
> Maybe I should ask this question in MPICH/OpenMPI development mailing- 
> lists, but is it possible to run these MPI implementations over 
> provider, which has only fi_cq part of API for manual data progress?

I don't know the specifics of what MPI requires.  MPICH and OpenMPI are both flexible in what features they require, however, and usually default to a fairly minimal set.

I believe both OpenMPI and MPICH will work over the RxM and SHM providers.  (The owners of those components should be able to verify).  If so, then this matrix will help identify which features are needed (though they may still implement more than what is required):
 
https://github.com/ofiwg/libfabric/wiki/Provider-Feature-Matrix-master

- Sean