[ofiwg] getting hardware details from libfabric

Hefty, Sean sean.hefty at intel.com
Tue May 29 10:16:14 PDT 2018


> This caught my eye on Friday but it was Beer-thirty.  ;-)
> 
> I'm also not sure "protocol" is sufficient.  There are many devices
> which support multiple protocols: OmniPath == verbs _and_ PSM, RoCE
> == sockets (TCP, UDP, HTTP, SSL?) _and_ roce, usNIC == Sockets and
> libfabric... ?  (what is the underlying "protocol" here?  Does it
> matter?) ???

Well, I consider verbs, psm, and sockets APIs, not protocols.  How much beer did you ingest?  :)

I guess you could argue for psm as a protocol, and roce is as well.


> I just end up asking myself "what is a protocol"?  There are so many
> definitions and depending on the level you are discussing it gets
> pretty confusing.
> 
> Should we have a list here?  Or do you get multiple fid_attr for
> each "protocol"?

To be clear, the proposed attributes are associated with a specific struct fi_info, plus other substructures (domain_attr, ep_attr, etc.).  The data is retrieved as part of calling fi_getinfo(), NOT some sort of get_device_info() call.


> What about listing APIs supported?  Verbs, Sockets, whatever GNI,
> BlueGene, usNIC are?

This is for libfabric, so I don't see a reason for it to tell the app that, hey, other APIs are supported.  As soon as the device supports TCP, it technically supports nearly all other APIs.

- Sean


More information about the ofiwg mailing list