> The application is NFS-RDMA. NFS keeps it's data as non-contiguous
> arrays of pages. So the motivation is that having a larger SGL allows
> you to support larger data transfers with a single operation. 

What operations are you using? RDMA writes?

If yes, it might be cleaner, and might turn out to be
more efficient, to simply post a list of multiple write WRs,
each with a small number of SGEs.

You can use selective signalling to only get a completion
on the last WR on success.


-- 
MST
_______________________________________________
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

Reply via email to