> Quoting Roland Dreier <[EMAIL PROTECTED]>: > Subject: Re: [PATCH 0 of 3] comp_vector kernel support > > > How about just patches 1 and 2? > > They don't do anything to *kernel* ULPs by themselves, > > and give userspace ULPs opportunity to start using the feature. > > We'll learn from that, and enhance kernel ULPs by 2.6.23. > > I guess I could see doing the first patch (just support multiple > vectors in the kernel without changing any drivers). That way it > would be easy to experiment with patched drivers that enable multiple > vectors.
OK, let's do it. > I think there's still some figuring out to do about how many EQs to > enable, etc, and I think it would be better to prevent drivers from > escaping into the wild before we have a better handle on the issues. How about applying mthca patch, but changing MTHCA_COMP_VECTORS to 1? That would make it easy to do experiments at the ULP level. -- 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
