From mboxrd@z Thu Jan 1 00:00:00 1970 From: Roland Dreier Subject: Re: InfiniBand/RDMA merge plans for 2.6.35 Date: Thu, 10 Jun 2010 13:28:26 -0700 Message-ID: References: <4C0D778C.8070108@oracle.com> <4C114223.1010309@oracle.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Return-path: In-Reply-To: <4C114223.1010309-QHcLZuEGTsvQT0dZR+AlfA@public.gmane.org> (Andy Grover's message of "Thu, 10 Jun 2010 12:50:59 -0700") Sender: linux-rdma-owner-u79uwXL29TY76Z2rM5mHXA@public.gmane.org To: Andy Grover Cc: linux-rdma-u79uwXL29TY76Z2rM5mHXA@public.gmane.org, Yevgeny Petrilin List-Id: linux-rdma@vger.kernel.org > > However, I haven't replied to these patches in detail but in general I > > don't like this approach of "pick a random vector" since it is > > non-deterministic and not likely to end up with an optimal result. > What is the optimal way to do this, if it isn't to spread CQs evenly > across all available vectors? (least attached vector != random.) Since there is no way to know whether a given vector has a bunch of CQs that generate very few events or maybe a single CQ that generates a heavy load of events, the number of attached CQs is really pretty useless as a basis to decide. I think it's much better to try and attach your CQ to a vector that is directed at the CPU where you want to process the work. - R. -- Roland Dreier || For corporate legal information go to: http://www.cisco.com/web/about/doing_business/legal/cri/index.html -- To unsubscribe from this list: send the line "unsubscribe linux-rdma" in the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org More majordomo info at http://vger.kernel.org/majordomo-info.html