Thanks a lot Eelco. MFEX v7 will be available shorty and I have addressed all the reviews comments hopefully.
Br Amber > -----Original Message----- > From: Eelco Chaudron <echau...@redhat.com> > Sent: Thursday, July 8, 2021 6:03 PM > To: Ferriter, Cian <cian.ferri...@intel.com> > Cc: ovs-dev@openvswitch.org; f...@sysclose.org; i.maxim...@ovn.org; Van > Haaren, Harry <harry.van.haa...@intel.com>; Amber, Kumar > <kumar.am...@intel.com>; Stokes, Ian <ian.sto...@intel.com> > Subject: Re: [v6 11/11] dpif-netdev/mfex: add more AVX512 traffic profiles > > > > On 6 Jul 2021, at 15:11, Cian Ferriter wrote: > > > From: Harry van Haaren <harry.van.haa...@intel.com> > > > > This commit adds 3 new traffic profile implementations to the existing > > avx512 miniflow extract infrastructure. The profiles added are: > > - Ether()/IP()/TCP() > > - Ether()/Dot1Q()/IP()/UDP() > > - Ether()/Dot1Q()/IP()/TCP() > > > > The design of the avx512 code here is for scalability to add more > > traffic profiles, as well as enabling CPU ISA. Note that an > > implementation is primarily adding static const data, which the > > compiler then specializes away when the profile specific function is > > declared > below. > > > > As a result, the code is relatively maintainable, and scalable for new > > traffic profiles as well as new ISA, and does not lower performance > > compared with manually written code for each profile/ISA. > > > > Note that confidence in the correctness of each implementation is > > achieved through autovalidation, unit tests with known packets, and > > fuzz tested packets. > > > > Signed-off-by: Harry van Haaren <harry.van.haa...@intel.com> > > These changes look good to me! > > This concludes my review of v6. I will get my system ready to do some > performance runs, and hopefully, I can get some results tomorrow or Monday. > > Acked-by: Eelco Chaudron <echau...@redhat.com> _______________________________________________ dev mailing list d...@openvswitch.org https://mail.openvswitch.org/mailman/listinfo/ovs-dev
