> -----Original Message----- > From: Eelco Chaudron <[email protected]> > Sent: Wednesday, July 7, 2021 10:41 AM > To: Van Haaren, Harry <[email protected]> > Cc: Amber, Kumar <[email protected]>; [email protected]; > [email protected]; Flavio Leitner <[email protected]>; Stokes, Ian > <[email protected]> > Subject: Re: [ovs-dev] [v4 02/12] dpif-netdev: Add auto validation function > for > miniflow extract > > > > On 7 Jul 2021, at 11:33, Van Haaren, Harry wrote:
<snip> > > By removing scalar DPIF enabling of MFEX opt pointer (details below) we > remove any > > urgency on benchmark results? > > I’ll wrap up the review first, and hopefully, when you are working on > potential > changes, I can run the tests and get some results. As we're nearing the merge dates, I'd prefer to focus on getting merged. To help review & merge, v7 will contain the following patch split change: Scalar DPIF usage of the MFEX Optimized function is now in its own patch at the end of the series. This allows all other MFEX patches to be merged, without any hazard to scalar DPIF datapath performance. > I understand now what you meant with disabling it in the scalar part, so if I > still > see 1%+ deltas I’ll try it out. Eelco's testing results can inform the inclusion of Scalar DPIF usage of the MFEX function pointer. As this enabling is now in a separate patch at the end of the series, it means that the patch can be easily merged, or not merged. No rebasing or rework required. If the main MFEX code is ready for merge before the testing results are in, this allows the merge of MFEX. Scalar enabling can be merged later in the 2.16 merge window if desired, or re-visited in a future release. <snip> Regards, -Harry _______________________________________________ dev mailing list [email protected] https://mail.openvswitch.org/mailman/listinfo/ovs-dev
