One thing to keep in mind is that the IBTA workgroup responsible for the transport wanted to eliminate immediate data support entirely but it was retained solely to enable VIA application migration (even though the application base was quite small).  If that requirement could have been eliminated, then it would have been gone in a heart beat.  Given a RDMA-WRITE followed by a SEND provides the same application semantics based on the use models, iWARP chose not to support immediate data. 

 

Mike,

 

I was not part of the original IBTA discussions and I won’t argue whether this facility should or shouldn’t have been include.  Nevertheless, it is part of the specification, there are HCA vendors that implement it, and we have applications that make use of it.  I would, however, disagree with your assertion that write followed by a send is semantically equivalent to write immediate.  Ordering may be semantically the same, but the service is not.  Receive work completions are explicitly indicated as being associated with immediate data and therefore an associated write completion.  A write followed by a send does not provide the same indication semantic.

 

Roy

_______________________________________________
openib-general mailing list
[email protected]
http://openib.org/mailman/listinfo/openib-general

To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general

Reply via email to