physically contiguous). Removing this limitation would make the code more complex, and in general supporting huge queue depths hasn't seemed that important.
Currently lnet 1.4.11 (lustre networking layer) uses 4K work requests (page sizes). Up to 256 of these are chained together using the next pointer so that a single call to ib_post_send is made for up to a 1MB xfer. The number of work requests allocated for the QP is controlled by number of concurrent sends * 256.
At 16 concurrent sends there is no problem. At 64 there is (once we allocate recv work requests as well). It sounds like this can be alleviated by using FMR. - Galen _______________________________________________ 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
