> > one could do a hack with ib_modify_cq() a la IPoIB (assuming you're using > > ConnectX HCA).
> I wonder why you consider this a hack... is programming interrupt > coalescing uncommon with HBAs? with networking I understand that the > common practice is to apply both HW mitigation and "SW mitigation" > (NAPI) so with SSD IOPS rates peaking to hundreds of thousands I > assume such an approach is valid for disks as well. Just because it is ConnectX only, and SW mitigation is likely to have even more benefit. _______________________________________________ general mailing list general@lists.openfabrics.org http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
