Performance


on modern, recent amd64, current Xeon, pure forwarding: cannot measure benefits
not even between software engine and offloading to hardware

analyzing profiling data: current Xeons
so much raw computing power, big and fast caches
other factors becoming the limiting ones
latency, last not least to the devices
effects are completely hidden for these microbenchmarks
different story if doing more than pure forwarding

older system (Pentium M, em(4)) seeing about 7% increase