There is a use case with the rxm provider where it needs the maximum size of the CQ that the underlying message provider can create. For everyone who's forgotten, the 1.7 release will introduce a new set of hardware attributes through a fid_nic structure.
https://ofiwg.github.io/libfabric/master/man/fi_nic.3.html I'm considering adding an fi_device_limits structure to this in order to expose hardware specific limitations, with at least the following fields being defined: - ep_cnt, tx_size, tx_cnt, rx_size, rx_cnt, cq_size, cn_cnt, stx_cnt, srx_cnt, av_size Does anyone have opinions on this? _______________________________________________ ofiwg mailing list [email protected] https://lists.openfabrics.org/mailman/listinfo/ofiwg
