On 25/01/2016 18:20, "Andy Zhou" <az...@ovn.org> wrote:
>On Sun, Jan 24, 2016 at 8:32 AM, Ben Pfaff <b...@ovn.org> wrote: > >> Before this commit, emc_processing() copied a netdev_flow_key if there >>was >> no exact-match cache (EMC) hit. This commit eliminates the copy by >> constructing the netdev_flow_key in the place it would be copied. >> >> Found by inspection. >> >> Shahbaz (CCed) reports that this reduces the cost of an EMC miss by 72 >> cycles in his test case in which the EMC is disabled. Presumably this >> is similarly valuable in cases where the EMC merely has few hits. >> >> CC: Muhammad Shahbaz <mshah...@cs.princeton.edu> >> Signed-off-by: Ben Pfaff <b...@ovn.org> >> > >Logic looks good to me. > >Acked-by: Andy Zhou <az...@ovn.org> > >I think Daniele is testing on performance impact on this patch. so it is >may be worthwhile >to wait for his input before pushing the change. If I disable the EMC I observe an increase in throughput when this patch is applied (from ~4.8 Mpps to ~5.4 Mpps). If the EMC is enabled, instead, I see a drop in throughput (from ~12.4 Mpps to ~12.0 Mpps) with this patch. At a first glance it looks like this patch is preventing my compiler (GCC 4.8) to perform some optimization. It is probably a good idea to apply this anyway given the numbers, but I'd like to do some more testing to figure out if we can avoid the drop in the EMC-only case. Thanks, Daniele _______________________________________________ dev mailing list dev@openvswitch.org http://openvswitch.org/mailman/listinfo/dev
