> Quoting Roland Dreier <[EMAIL PROTECTED]>: > Subject: Re: [Bug 508] IPoIB CM multicast is hogging interrupts > > > This seems to hurt top bandwidth a bit in my testing, so this needs some > more > > work. Meanwhile, Scott, could you please check whether the following > > patch helps in your test-case? > > The common CQ helped bandwidth, right? What's the advantage of this approach?
Unfortunately, further testing demonstrated a problem with the common CQ approach: when an error is discovered on some TX QP, I can't just stop using that TX and immediately re-enable the interface as I currently do: I have to wait until I get all completions with error or until the QP is reset to get rid of outstanding completions, and this might (and does) take a long time since this all involves slow path operations. -- MST _______________________________________________ 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
