On Nov 12, 2010 06:23 PM, "Hefty, Sean" <sean.he...@intel.com> wrote:
> > I would like to perform test about RDMA server/client about multiple > > reads or/and writes. > > Can server/client store few memory regions? If yes, how? > > An app can register multiple memory regions. Just call register mr for > each region that you want to register. > > > Client writes on different MRs with only one send WR. Is that > > possible? > > No - an RDMA write targets a single remote MR. > > > The goal is: build an array of addresses (MRs) and write (or read) > > with > > a First Input Fist Output (FIFO) model using RDMA. > > > > Any ideas? > > If you have an array of addresses, you can register each address and > store the mr handle with each. The address, rkey associations must be > transferred to the remote peer, so it can issue the RDMA operations. > > Unless the addresses are scattered, you could also look at registering > a single, larger memory region, with each RDMA operation targeting > only a portion of that region. --> I register a single larger memory region with rdma_reg_msgs. What are parameters in rdma_post_recv function to use these "big" memory regions few times? In other words, on that MR can I write/read different smaller messages? Andrea Gozzelino > Although, based on this brief description of having a FIFO, posted > receive buffers may work just as well for you. And a receive would > allow incoming data to be scattered to different memory buffers. > > - Sean > -- > To unsubscribe from this list: send the line "unsubscribe linux-rdma" > in > the body of a message to majord...@vger.kernel.org > More majordomo info at http://vger.kernel.org/majordomo-info.html > -- To unsubscribe from this list: send the line "unsubscribe linux-rdma" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
