2016-06-15 6:37 GMT-07:00 Ryan Moats <rmo...@us.ibm.com>: > "dev" <dev-boun...@openvswitch.org> wrote on 06/14/2016 08:58:33 PM: > > > From: Daniele Di Proietto <diproiet...@ovn.org> > > To: Nirapada Ghosh/San Jose/IBM@IBMUS > > Cc: "dev@openvswitch.org" <dev@openvswitch.org> > > Date: 06/14/2016 08:59 PM > > Subject: Re: [ovs-dev] [PATCH V5] Function tracer to trace all function > calls > > Sent by: "dev" <dev-boun...@openvswitch.org> > > > > Hi Nirapada, > > > > When optimizing for the DPDK datapath we have a very similar problem and, > > usually, running a simple profiler like perf ( > > https://perf.wiki.kernel.org/index.php/Main_Page) is enough to highlight > > the bottlenecks in terms of CPU usage. > > > > Have you tried perf? Does this infrastructure provide advantages over > > sample profiling? I'm not saying that one is better than the other, I'm > > just curious! > > > > Thanks, > > > > Daniele > > Daniele- > > Nirapada's patch set is driven by the requirement to extract automated > profiling under scale while using the same bytes that are to be > deployed for operation. > > This pretty much precludes the use of any external profiler. > > Ryan > Ok, I understand, no objections here.
Thanks, Daniele _______________________________________________ dev mailing list dev@openvswitch.org http://openvswitch.org/mailman/listinfo/dev
