Interrupt affinity and MSI for high performance packet forwarding

[Brett Warden]

I'm doing some evaluation of DragonFly with regard to high performance
IP forwarding on multi-core systems. My preliminary results yield
pretty poor results with 64-byte frames on GbE (comparable to NetBSD
but substantially less impressive than Linux). I was wondering if
there are any good tweaks I should use to improve performance?
Does DragonFly support MSI?

Is there any current implementation of SMP interrupt affinity (and if
so, how do I enable it)? I've only seen scattered discussion about
that on the mailing lists over a year ago.


Thanks,

--
Brett T. Warden
Intel Corporation