Roland Dreier wrote:
Have you tested this to know whether it matters? Increasing the limit
uses more memory per QP...
It give some benefit but not as substantial as in ConnectX, so I guess
we do not need this after all.
Does the rdma read latency test in OFED queue up enough work requests
to measure this?
yes
Tziporet
_______________________________________________
general mailing list
[email protected]
http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general
To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
