> I am thinking about detecting that we are getting a completion each > couple of microseconds, and just polling a bit more (even if CQ is empty) > simply to avoid paying 10s micro for the next interrupt. > This is just a fuzzy idea, but maybe this can work?
It's a good area to look at (assuming it's worth spending yet more time on IPoIB performance), but I think it's quite difficult to come up with a good solution. For example search for Jamal's paper "When NAPI Comes to Town" to see some related work that was a failure in the end. _______________________________________________ openib-general mailing list openib-general@openib.org http://openib.org/mailman/listinfo/openib-general To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general